Optimized bath control for void-free copper deposition
11/01/2001
Zhi-Wen Sun, Girish Dixit, Applied Materials Inc.
overview
A method for optimizing the control of organic additives in a plating bath for copper interconnect applications is achieved by careful choice of the CVS cycling parameters and the supporting electrolyte makeup.
Copper electro-deposition is the standard process used for sub-micron trench/via filling in advanced Cu interconnect applications [1, 2]. Organic additives with concentrations at ppm (parts per million) levels in the plating bath control the Cu deposition mechanism and the deposit quality. Additive concentration analysis, with excellent accuracy, precision and high resolution, as well as control, are therefore critical to ensure sub-100nm trench/via filling with high-quality, void-free Cu deposits free of killer defects. Electro-analytical techniques like cyclic voltammetric stripping (CVS) are especially suited for this purpose [3, 4].
Used in the printed circuit board industry for many years, CVS has gained widespread acceptance as the standard on-line organic additive analysis method for the semiconductor processing industry. Because this industry was unfamiliar with such electro-analytical techniques, CVS analysis development was done solely by the instrument vendors. Since a complete understanding and characterization of chemical management methodologies by equipment makers is necessary for production-worthy process solutions, this approach limited the industry's capability to optimize the CVS method for controlling plating applications.
To fully utilize and optimize this methodology for critical Cu interconnect fill application, Applied Materials' in-house CVS analysis team developed custom applications for on-line monitoring and control of advanced Cu plating bath systems. These contain three organic additives: the accelerator (A), suppressor (S), and leveler (L). This effort has resulted in the creation of a fully automated chemical management system for production electroplating, with very tight process control suitable for device generations at 100nm and below.
Role of organic additives
Successful plating of high-aspect ratio features requires reproducible control of the copper film growth and morphology of the plated film. The growth rate must be balanced to minimize seam/void formation, typically using a bottom-up fill process known as "superfill." The correct balance of additives to the basic electrolyte provides this bottom-up capability.
Conventional copper-plating baths contain copper sulfate (CuSO4), sulfuric acid (H2SO4), and chloride ions plus several essential organic additives that control the copper electroplating rate and morphology [5, 6]. Common organic additives include:
- accelerator (sometimes called brightener or anti-suppressor) that catalyzes and speeds-up the conformal overfilling of vias and trenches. It contains sulfur-containing molecules, typically sulfonic acid groups or disulfides such as SPS (Bis- (sodium sulfopropyl)-disulfide) with the chemical formula of NaSO3(CH2)3S-S(CH2)3SO3Na.
- suppressor, a surfactant or wetting agent (sometimes called the carrier), that suppresses Cu growth at the top edges of vias and trenches. Surfactants can be long chain polymers such as polyethylene glycol (PEG) or co-polymers of polyoxyethylene and polyoxypropylene having mean molecular weights >1000.
- leveler (also called a grain refiner or over-plate inhibitor) that controls the grain size of the plated copper and inhibits over-plate of the copper above the top of the trench (overplating). Levelers are usually high-molecular weight polymers with amine (-NH3) or amide (-NH2) functional groups [5].
During plating, these organic additives are consumed so concentrations decrease to negatively impact filling properties. Additive breakdown is caused by anodic oxidation, cathodic reduction, and subsequent complex formation. Breakdown occurs even in the absence of plating by oxidation by the dissolved air and hydrolysis and complex formation in the plating solution (tool idle depletion). Highest organic breakdown rates are associated with high acid chemistries and high plating currents.
Cyclic voltammetric stripping analysis
CVS is the standard technique for on-line analysis of such organic additives. Output from this analysis drives automated replenishment algorithms for additive concentration control within predefined limits in copper plating baths. This kind of bath control is absolutely necessary to maintain the precise concentrations of the three organic components required for void-free gap fill. Figure 1 shows the need for plating bath control in electro-plating: Fig. 1a shows void-free fill when additive concentrations are controlled within specified range. The SEM in Fig. 1b shows voiding that can arise with additives outside the control range.
The CVS technique is based on the concept that the plating rate varies with additive concentration. In CVS, a miniature test cell mimics the wafer environment and monitors changes in the
Cu-plating rate on a test electrode. The amount of Cu plated is dependent on the additive concentration: in general, suppressors and levelers decrease the amount of plated Cu, while accelerators increase the plated Cu.
The CVS test cell characterized in Fig. 2 consists of a platinum (Pt) rotating disc electrode (the working electrode), a copper rod as the counter (or auxiliary) electrode, and an Ag/AgCl reference electrode (Fig 2a). All three electrodes are immersed in a solution consisting of a supporting electrolyte (organic free virgin make-up solution [VMS] plus pre-determined organic content) and a sample of plating bath. A cyclical scanning voltage is applied to the Pt electrode; the current flow to the electrode is measured and displayed as a function of the applied voltage (Fig 2b).
 Figure 2. Current flow to electrode is measured/displayed as a function of applied voltage. |
In the figure, the negative potential limit (NPL) is a cathodic potential where Cu deposition occurs. The positive potential limit (PPL) is an anodic potential where not only the copper deposited on the Pt surface during cathodic scan is stripped, but also the Pt surface is cleaned of adsorbed organic contaminants. Therefore, during an entire scan, a small amount of copper is alternately deposited and stripped from the Pt disk electrode surface. The amount of Cu stripped during the anodic scan is the same as that deposited during the cathodic scan, and is indicated by the hatched peak area (Ar). The CVS instrument integrates the current to determine Ar in milli-coulombs (mC). The response curve for a specific additive in an optimized supporting electrolyte is used as the calibration curve for that additive. The bath solution can be analyzed by adding a small amount to be tested to the supporting electrolyte and comparing the results to established calibration curves. As will be discussed later, the Ar values are controlled not only by the individual additive concentrations, but also by the interaction of the various additives (the matrix effect) as well as the specific process conditions of the CVS analysis itself.
The challenge in developing a production-worthy, reliable CVS analytical technique is to fully understand the inter-relationships of the additives and process conditions and to accurately measure concentrations and calculate needed dosing volumes. The important CVS cycling parameters are the rotation speed of the Pt electrode, the potential scan rate, and the maximum NPL. Judicious choice of these cycling parameters, as well as the chemistry of the supporting electrolyte, maximizes analysis sensitivity and minimizes the matrix (interference) effect during additive analysis.
Optimizing additive response curves
To ensure the most robust measurement methodology, the approach was validated using CVS analysis on the Electra Cu Integrated ECP system with multiple chemistries from several suppliers. Standard chemical solutions each containing three additives were mixed with a VMS composed of CuSO4, H2SO4 and CuCl2. The additive standards were prepared immediately prior to analysis in order to ensure accuracy of the calibration curves. Additive response curves show the relation between the plating rate and a particular additive concentration for selected CVS cycling parameters.
Figures 3a and 3b show representative response curves: normalized peak area (the ratio of Cu plated with additive to Cu plated without additive, Ar/Ar0 where Ar0 is the value with no additive,) as a function of each individual additive concentration and CVS cycling parameters. The figure legends list the three CVS cycling parameters in the order of: electrode rotation speed (rpm), scan rate (mV/s), and negative potential limit (mV); e.g. 2500-100-300 indicates 2500rpm, with 100mV/s and -300mV cycling conditions. Figure 3a shows the response to the suppressor concentration; Fig. 3b shows the response to the leveler concentration.
Note that both the suppressor and the leveler inhibit the Cu deposition rate since the normalized peak area values are <1.0. The degree of suppression increases with increasing electrode rotation speed. At higher electrode rotation speeds, the boundary layer is thinner and suppressive additives can more readily diffuse to the electrode, adsorb on the surface and thus inhibit Cu growth. On the other hand, the degree of suppression decreases with increasing potential scan rate: there is less time for additives (suppressor and leveler) to reach the electrode surface during the scanning cycle. For both additives, increasing the negative potential limit increases the additive concentration required for saturation suppression (defined as no Cu plated regardless of additive concentration). These trends are true for all types of levelers and suppressors.
Figures 3a and 3b also illustrate the two major distinctions between the suppressor and the leveler. First, the saturation suppression occurs at significantly lower concentrations (an order-of-magnitude) for the suppressor than for the leveler. An amount of suppressor <0.3 ml/liter is needed to completely inhibit Cu deposition. This is due to the fact that the adsorption of a monolayer of the suppressor polymer on the surface is sufficient to completely inhibit the Cu deposition [7]. Since the saturation suppression concentration of the suppressor is so small relative to the other additives, the suppressor analysis can be done with little interference from the other additives. In contrast, the leveler concentrations impact the plating rate over a much broader concentration range. Second, the leveler is much more sensitive to CVS parameters. The leveler is a large, bulky molecule that diffuses very slowly; low rotation speeds and high voltage scan rates render the leveler completely inactive [8]. Thus, in a typical plating bath, CVS conditions can be chosen so that leveler interference in other additives is minimized.
The response curve for the accelerator concentration is less complicated and is not shown here; the Cu deposition initially increases rapidly up to 2ml/liter; more accelerator has less impact until a saturation point is reached. The disulfide accelerator acts as a Cu deposition catalyst by forming surface complexes such as, Cu+S(CH2)3SO3, between Cu ions and the disulfide reduction product formed at Cu plating potentials [9].
Minimizing the matrix effect
Figure 4 shows several response curves: a standard suppressor calibration curve and the response curves of 1) a leveler and suppressor solution, 2) an accelerator and suppressor solution, and 3) a three component leveler, suppressor, and accelerator solution (typical plating bath). Note that all curves strongly resemble a suppressor-only response curve. Since all solutions contain the same suppressor concentration, this clearly illustrates the dominant effect of the suppressor. By comparing the plating bath response curves to the suppressor calibration curve, we can calculate the amount of suppressor in the plating bath.
 Figure 5. a) Accelerator effect in the presence of excessive suppressor; b) excessive leveler concentration acts as a suppressor. |
The accelerator and leveler analyses must, of necessity, be made under the strong influence of the suppressor. The suppressor concentration must be high enough such that the response of the additive is independent of suppressor concentration. Figure 5a shows the accelerator effect in the presence of excessive suppressor. It is evident that the accelerator acts as a very effective anti-suppressor: the Cu deposition rate, while almost totally inhibited by the suppressor, is significantly reactivated by the addition of the accelerator. The anti-suppression effect originates from the adsorption of the additives on the electrode surface and the subsequent suppressor desorption and reduction under the influence of accelerators [10]. Note that the leveler in the presence of low accelerator concentrations has little impact on the plating rate. At high leveler concentrations, the leveler acts as a suppressor to reduce the plating rate in spite of the accelerator.
The data shown in Fig. 5a suggest that it is possible to take advantage of the high anti-suppression characteristic at low accelerator concentrations to measure the accelerator concentration by diluting the solution under test with the supporting electrolyte comprising the VMS and excessive amount of suppressor. In fact, this has been the standard approach used for accelerator additive analysis.
Figure 6 also suggests that the leveler analysis could be done in a supporting electrolyte containing suppressor and accelerator, if the leveler can compete with accelerator for surface adsorption or compete with the accelerator to reduce the effective anti-suppression effect. The response curve in Fig. 5b shows that the leveler does indeed behave as a suppressor under this condition; this calibration curve is then used to analyze the leveler concentration [11, 12].
 Figure 6. Response curves for several aged solutions, each measured using a different set of CVS conditions. |
All data presented so far have been collected using fresh solutions or standards. During the actual plating process, however, the organic additives are consumed and form various byproducts. The decrease in additive concentrations and increase in byproduct concentrations negatively impact filling properties. Also, these byproducts are known to impact CVS analysis, in that the CVS process window must be chosen to be large enough to accurately measure additive concentrations in the presence of these inherently occurring byproducts.
Figures 6a and 6b show response curves for several aged solutions, each measured using a different set of CVS conditions. On the left, the nonoptimized CVS analysis cannot distinguish between the two aged baths. On the right, the optimized CVS analysis maintains sensitivity even on an aged bath.
 Figure 7. Schematic of ECP system with dedicated chemical cabinet having a separate titrator cell. |
The bath control approach used in manufacturing utilizes auto-dosing algorithms to supplement additives as they are consumed. Dosing amounts are determined by analyzing the existing concentration by CVS relative to predetermined levels. In addition, a bleed-and-feed algorithm is used to reduce the amount of byproduct build up. Optimized CVS conditions accurately measure concentrations to maintain a stable bath and minimize the amount of bleed-and-feed required. This is done by maintaining a high accuracy of additive measurement in spite of increased additive byproduct concentrations as the bath ages.
 Figure 8. Additive concentration stability using automated three-component CVS control. |
Plug-and-play chemical analysis for production
For production plating, the Electra Cu iECP system utilizes a closed-loop chemical cabinet that measures (by CVS) and adjusts the concentration of all three organic components. The inorganic concentrations are measured using standard volumetric titration methods. However, the titration is performed in a separate, dedicated titrator cell (Fig. 7). Conventional metrology cabinets combine both the CVS and titration analyses in a single test cell. Separating the inorganic and organic analyses eliminates cross-contamination to increase analysis accuracy, and, since the analyses can be done in parallel, the analysis time is significantly reduced. The duration of the analysis is important to provide real-time process control for the tightest possible process window.
Figure 8 shows additive concentration stability using the automated three-component CVS control for the concentrations of leveler, suppressor, and accelerator as a function of bath age or amp-hours of use.
Conclusion
Optimized control of organic additives (accelerator, suppressor, and leveler) in a plating bath for copper interconnect applications is achieved by careful choice of the CVS cycling parameters and the supporting electrolyte makeup. The optimized analytical methodology relies on a detailed understanding of the relation between copper deposition rate, the additive transport in electrolyte solution, the adsorption/desorption of the additives on the electrode surface, and the interaction with additive byproducts. An optimized analysis and control technique maximizes plating performance (fill, defects, process repeatability) while minimizing the amount of bleed-and-feed needed. Furthermore, a detailed understanding of the additive responses under various conditions enhances our understanding of the Cu plating deposition mechanisms and reaction kinetics.
Acknowledgments
The following members of the ECP Division of Applied Materials contributed to this work: Chunman Yu, Nicolay Kovarsky, Srinivas Gandikota, Brian Metzger, and David W. Nguyen. Technical discussions with ECI personnel Alex Kogan, Gene Chalyt, Mike Pavlov, and their contributions to the on-line CVS analyzer development are greatly appreciated.
References
- M.E. Gross, C. Lingk, W.L. Brown, R. Drese, "Implications of Damascene Topography for Electroplated Copper Interconnects," Solid State Technology, 42(8), pp. 47-52, August 1999.
- S. Parikh, et al., "Defect and Electromigration Characterization of a Two Level Copper Interconnect," Proceedings of the 2001 International Interconnect Technology Conf., pp. 183-185, IEEE, 2001.
- Application Note, "Fundamentals of Cyclic Voltammetric Stripping," ECI Technology, East Rutherford, NJ.
- D.M. Tench, C.A. Ogden, US Patent 4,132,605, "Method for Evaluating the Quality of Electroplating Baths," 1979.
- J. Reid, "An HPLC Study of Acid Copper Brightener Properties," PC Fab 10, p. 65, Nov. 1987.
- A.C. West, "Theory of Filling of High-Aspect Ratio Trenches and Vias in Presence of Additives," J. Electrochemical Soc. 147, p. 227, 2000.
- J.J. Kelly and A.C. West, "Copper Deposition in the Presence of Polyethylene Glycol," J. Electrochemical Soc. 145, pp. 3472-3476, 1998.
- L. Mirkova, St. Rashkov, Chr. Nanev, "The Leveling Mechanism During Bright Copper Plating," Surface Technology 15, pp. 181-190, 1982.
- E.E. Farndon, F.C. Walsh, S.A. Campbell, "Effects of Thiourea, Benzotriazole, and 4,5-dithiaoctane-1,8-disulfonic Acid on the Kinetics of Copper Deposition from Dilute Acid Sulfate Solutions," J. Appl. Electrochem. 25, pp. 574-583, 1995.
- N. Kovarsky, Z. Sun, "Mechanisms of Cu Interconnect Plating: Effect of Accelerator and Suppressor Additives on Trench and Via Filling," to be submitted to J. Electrochemical Soc.
- P. Bratin, G. Chalyt, M. Pavlov, R. Gluzman "Control of Tin and Tin / Lead Plating Bath by Cyclic Voltammetry," Proceedings of IPC 96 Printed Circuit Expo, p. S16-1-1-14, 1996.
- P. Bratin, G. Chalyt, M. Pavlov, "Automated Online Control of Plating Bath Additives Increases Wafer Yield," Semiconductor Fabtech, 14th Edition, pp. 205-207, 2001.
Zhi-Wen Sun received his PhD in Materials Science from University of Bordeaux, France. He is a senior member of the technical staff in the ECP Technology department of the CPI Business Group at Applied Materials, 3303 Scott Blvd., M/S 10851, Santa Clara CA 95054; e-mail: [email protected].
Girish A. Dixit received his PhD in Materials Science and Engineering from State University of New York at Stony Brook. He is senior director of technology in the ECP Technology department of the CPI Business Group at Applied Materials, 3303 Scott Blvd., M/S 1158, Santa Clara CA 95054; e-mail: [email protected].