Electroplating bath controls for copper interconnects
11/01/1998
Electroplating bath control for copper interconnects
Thomas Taylor, Thomas Ritzdorf, Fred Lindberg, Brad Carpenter, Semitool Inc., Kalispell, Montana
Mark LeFebvre, Shipley Co., Marlborough, Massachusetts
Electroplating bath composition plays a significant role in deposited copper film properties. Methods were developed to replenish plating bath additives depleted while depositing copper on semiconductor substrates under conditions simulating high-volume wafer fabrication. Advantages and drawbacks of the analytical methods were noted, and a candidate topology for a closed-loop electrolyte monitoring and control system was proposed.
Damascene process flows - involving the deposition of metal into high-aspect-ratio, barrier-lined features etched into an underlying dielectric film - are the de facto standard for the formation of copper interconnects. Damascene Cu will likely be introduced using electroplating or electrochemical deposition (ECD) processes [1, 2].
Though application of electroplating to fill submicron interconnect features is relatively recent, it has long been employed in multilevel printed wiring board (PWB) manufacturing. Electrolytes employed for PWB copper metallization include those based on pyrophosphate, fluoroborate, and sulfate salts. Influenced by considerations of cost, safety and convenience of use, acid copper sulfate baths have been increasingly dominant in PWB fabrication, and are among those first examined for IC interconnect deposition. For any of these electrolytes, however, the ability to monitor and control bath composition is a key factor in ensuring uniform and reproducible deposit properties.
In semiconductor applications, the electronic and morphological properties of copper interconnect films are of principal importance in determining final device performance and reliability. Important mechanical properties include modulus, ductility, hardness, and surface texture - all controlled or strongly influenced by the composition of the electrolyte, including (for acid copper) the concentrations of the copper cations, sulfate anions, chloride ions, free acid, and metallic impurities.
Of particular importance is measurement and control of proprietary organic compounds that modify the deposit properties through adsorption onto and desorption from the cathode surface during plating, affecting the diffusion rate of copper cations to nucleation and growth sites. These compounds are typically delivered as multicomponent packages from plating chemistry vendors.
One of the most important functions of the additive packages is to influence the throwing power of the electrolyte - the relative insensitivity of plating rate to variations in cathodic current density across the wafer or to ion concentration gradients in complex topographical features. The throwing power of the electrolyte has a major effect on the within-wafer uniformity of plated film thickness, and on the void-free filling of ultrafine trenches and vias. Organic additives have also been shown to have dramatic effects on mechanical film properties [3-8]. Detection and quantification of these important bath constituents is complicated by the fact that they are effective at very low concentrations in the electrolyte (often at several ppm or less).
Today, the majority of bath replenishment systems rely on open-loop control. Often this takes the form of a semi-automated replenishment system that makes periodic additions of concentrates to the electrolyte based on consumption models determined from preliminary characterization. However, open-loop replenishment may fail to accommodate variations in consumption associated with plating parameter fluctuations or changing chemical equilibria as the electrolyte solution ages.
Electroanalytical methods
Faradaic electroanalysis studies the electrochemical activity of the bath sample under applied electrical stimulus; the measured responses are thus fundamentally related to the electrolyte properties, which influence the quality of the metal deposition. Electroanalysis further offers the opportunity to study the mechanisms and kinetics of the plating process, and the influences the various bath components exert on plating rate suppression and acceleration.
The most widely applied electroanalytical techniques for plating bath analysis are cyclic voltammetric stripping (CVS) and cyclic pulsed voltammetric stripping (CPVS). Both techniques control the voltage between working and auxiliary electrodes with a potentiostat such that the working electrode cycles between cathodic and anodic potentials while in contact with the metal salt electrolyte. A metal film is alternately reduced on the working electrode surface and then stripped by anodic dissolution. During stripping, the current is integrated over time to quantify the electric charge required for complete film dissolution. The charge is directly related to the molar quantity of metal stripped (thus to the amount initially deposited) by Faraday`s laws. The stripping charge is monitored, instead of the deposition charge, because it is less sensitive to changing electrode surface states and impurity currents. Monitoring a process that proceeds to a well-defined endpoint also has inherent advantages.
The current/voltage/time relationship during analysis is extremely sensitive to variations in electrolyte composition and, not incidentally, measurement conditions such as temperature. If sufficient care is taken in methods development and measurement technique, then stripping voltammetry can be employed to generate calibration curves that provide reasonably accurate quantification of analyte composition.
Voltammetric stripping techniques are most often employed for evaluating levels of organic additives such as "suppressing" and "brightening" agents. Suppressors (also termed "carriers" or "levellers," depending on specific functionality) are macromolecule deposition inhibitors that tend to adsorb over the wafer surface and reduce local deposition rates. Brighteners are organic molecules that tend to improve the specularity of the deposit by reducing both surface roughness and grain-size variation. Brighteners interact with the suppressor molecules, competing for surface adsorption sites and locally accelerating deposition rates.
For ULSI interconnect applications, composition and concentrations of brighteners and suppressors are selected such that brightener surface concentration dominates that of suppressors on the interior surfaces of trenches and vias. Local deposition rates are thus surpressed at the top of topographical features relative to the insides, leading to the desired "bottom-up" deposition and void-free metal filling.
Brighteners are particularly prone to breakdown under the heavy electrolyte use of production plating conditions, and variations in concentrations can have dramatic effects in deposit film properties. A number of calibration methods have been proposed for correlating stripping charges and bath species` concentrations; several have been put into practice with good results [9, 10].
The main limitation of voltammetric analyses is their sensitivity to "matrix effects." Many bath components and their breakdown products display convoluted electrochemical interactions, so that stripping charge response is ambiguous if several components simultaneously change concentration. Conversely, compounds and ions that are not electrochemically active over the ranges of potentials explored, and which therefore do not directly affect the deposition or stripping rates under measurement conditions, cannot be detected. For instance, bath impurities that may be occluded in the deposit, or components that affect film properties without appreciably altering plating rate, are likely to escape notice.
These attributes make the voltammetric techniques less than ideal if the user`s goal is certain speciation of all bath components. Nonetheless, under conditions where the principal requirement is to detect and correct changes in one or two electrochemically active components that have a rapid rate of consumption as compared to others, the voltammetric methods are promising candidates for incorporation in closed-loop control systems.
 |
Figure 1. High-performance liquid chromatography (HPLC) schematic.
Chromatographic methods
Chromatographic separation of multicomponent liquids has been a standard quantitative analysis technique. Liquid chromatography can pinpoint almost all important electrolyte species, including inorganic ions; transition and heavy metals; chelating agents and complexes; organic additives and their breakdown products; and others specific to certain ECD metallization processes [11].
The chromatographic methods of greatest interest to electroplaters are high-performance liquid chromatography (HPLC) and ion chromatography (IC). Both methods require an eluent technique with provisions for sample introduction, a separation module, and a detection module (Fig. 1). It is a carrier solvent that transports small sample volumes through the separation and detection modules; the eluent is pumped under relatively high pressure and the analyte sample is introduced into the flowing stream with an injection valve. The eluent stream carries the sample through a separation column packed with resin beads. The constituent species of the sample display differing affinities for adsorption onto the bead materials; those adsorbed weakly are eluted through the column most rapidly, while those with increasing affinity display decreasing mobility and are separated into discrete bands before entering the detector module.
By judicious selection of eluent compositions, column resin coatings, and detector types, liquid chromatography can speciate and quantify most of the electrolyte components of interest without ambiguity due to matrix effects. It can also reveal the presence and build-up of impurities, even if their existence is initially unsuspected and they exert no direct influence on electrochemical activity of the bath.
Shortcomings of the liquid chromatography methods include waste generation, particularly the significant volumes of eluent necessary to perform frequent and successive analyses (~100 ml/measurement in continuous operation). As opposed to voltammetry, the waste streams differ in makeup from the plating electrolyte, and cannot simply be routed to a common electrolyte collector or discharged into the active plating reservoir. Also, the time required to perform the analyses can be long, particularly if dilute eluents are used to improve the separation phase resolution of similarly mobile species.
Organic additives in the electrolyte, which frequently exhibit rates of change exceeding those of the inorganic components, are usually present in extremely low volumetric concentrations. Quantizing these materials with sub-ppm accuracy can be extremely challenging with UV absorption. Since these materi als often induce a large electrochemical response, a promising technique to improve signal strength involves an electrochemical cell as the post-column detector. Amperometric detectors are available for this purpose from instrument suppliers, and methods development for high-resolution separation and measurement has been reported [12].
 |
Figure 2. CVS voltammograms of Enthone Cubath M electrolyte with different additives.
Experiment
We focused on developing reliable methods of analysis for acid copper electrolytes for plating thin (1.0-1.5 ?m) copper films on 200-mm wafers patterned with interconnect structures emulating damascene features. Commercially supplied acid copper sulfate plating electrolytes were examined, representing two types of commonly available additive packages. From Enthone-OMI, the CUBATH M bath chemistry reportedly uses a brightening agent based on polyether sulfide (PES). Shipley Co. supplied samples of its ELECTROPOSIT 1100 acid copper bath, which from examination of voltammetric behavior apparently employs an additive package based on sulfonium alkane sulfonates (SAS).
In work on CVS methods development [13, 14], significant differences were deduced in the deposition kinetics between electrolytes with these general additive types. These electrolytes were each originally developed for electrodeposition of PWB copper. Other copper electrolyte types and/or additive packages may ultimately be better suited to submicron interconnect fabrication.
Electrochemical deposition (ECD) of copper on 200-mm silicon wafers was performed using Semitool`s LT-210 ECD systems, or in some cases a manually operated version of this ECD reactor of the same configuration. CVS analyses were performed on an EG&G Potentiostat/Galvanostat Model 263A, or on a Qualiplate 4000 bath analyzer from ECI. Typical operating parameters were 100 mV/sec sweep rate and 2500 rpm on the working electrode. The CPVS analyses were performed using Shipley Co.`s ELECTROPOSIT Bath Analyzer. All ion chromatography and HPLC analyses were performed on a Dionex DX 500 system.
Cyclic voltammetric stripping
A typical CVS technique involves titration to measure changes in the electrochemical activity of the sample with cumulative volume of successive additive-containing aliquots. Figure 2 shows a series of voltammograms taken from an acid copper plating solution with the Enthone-OMI PES-type additive system. The curves in this figure represent the electrolyte solution without any additives, with the carrier additive (M D), with the brightener/leveler (M LO 70/30 Special), and with both components in recommended use concentrations. Suppression and enhancement of the plating rate can be seen in the cathodic portion of the curves, and by the area under the stripping peaks. The repeatability of CVS is approximately ?5% of the normalized peak area.
For convenient measurement of multicomponent mixtures, it is desirable to analyze directly each of the additives with no matrix effects from the other components. Unfortunately, the CVS data indicate that Enthone-OMI`s CUBATH M exhibits strong interactions between the organic additives over most of the region of interest (Fig. 3).
With no independent effect of one additive in the desired concentration range, an alternative technique must be employed. A method of titrating small amounts of the bath into a reference electrolyte solution containing no additives was used for analysis [15]. Volumes of samples (or additives) necessary to suppress stripping peak area by a preselected amount (70%) can be correlated to produce a calibration curve.
We used titration to make multiple measurements of baths with various additive concentrations to quantify the effects of additive concentration on the reported brightener and suppressor values reported by ECI`s Qualiplate analysis system. The brightener value displays direct linear correlation to the concentration of Enthone-OMI`s M D additive, while the suppressor value correlates somewhat to the volume of M LO additive. The M LO concentration can also be correlated to a linear regression of both the brightener and suppressor values.
As the CVS techniques were used to control bath concentration over an extended period of time, we observed that the reported suppressor value would continue to climb, although no additions of the M LO additive were made to the bath (Fig. 4). This suggests that the suppressor analysis is affected by one or more of the breakdown products from the M D additive.
 |
Figure 3. Normalized stripping peak area vs. titration volume shows organic additive matrix effects.
 |
Figure 4. Organic analysis vs. bath life, showing that CVS registers a false increase in suppressor value when no M LO is added.
 |
Figure 5. CPVS analysis of CUBATH M, showing good correlation to M LO concentration.
Cyclic pulsed voltammetric stripping
CPVS can also measure organic additives in acid copper sulfate plating baths. This technique has been optimized by Shipley Co. for use with its ELECTROPOSIT series copper baths and the brightening agents contained in them, and used to determine the brightener levels in these baths for PWB production purposes [12]. In Shipley`s implementation, the stripping charge is correlated to brightener concentration through calibration, and is thereafter reported as Total Brightener Analysis (TBA) units in evaluation of subsequent samples.
CPVS was also applied to the PES-type Enthone-OMI CUBATH M, without modification or optimization of the measurement parameters, for comparison. The technique is relatively sensitive to the M LO additive concentration (Fig. 5). As long as the carrier concentration is close to its recommended level, it is a simple matter to estimate the level of the brightener/leveler additive in the mixture. Alternatively, if the M D additive level is determined from a prior analysis (such as HPLC) then the M LO additive level can be calculated from the MD additive level and the reported TBA value.
 |
Figure 6. a) HPLC chromatogram of fresh acid copper plating bath with organic additives; b) the same bath, showing organic reaction byproducts after use
High-performance liquid chromatography
Significant HPLC parameters include eluent composition and concentrations, eluent flow rate, and detector parameters such as wavelength for a UV/VIS spectrophotometric detector. Various detectors exist to maximize the signal for the species of interest while separating it from other components with similar elution times.
Data were accumulated after optimizing the eluent concentration and detector settings in order to resolve separate peaks and maximize signal-to-noise with UV/VIS detection. HPLC spectra clearly show that a fresh CUBATH M electrolyte bath evolves new species over time (Fig. 6 a, b).
HPLC measurements monitored CUBATH M additive concentration during a marathon run of an LT-210 ECD system. These data indicate that as the peak labeled "MD" is held constant through bath replenishment techniques, another unidentified peak increases markedly with time, and a third peak, which appears only after running the bath, maintains an equilibrium value (Fig. 7). We hypothesize that these unidentified peaks correspond to two or more organic byproducts whose concentrations change as the bath ages. This is consistent with CVS analyses of the same electrolyte under extended-run conditions, in which increasing rate suppression was presumed to result from the build-up of brightener breakdown products in the PES-based additive package.
 |
Figure 7. HPLC analysis of marathon run.
Conclusion
Several analytical methods exist that may be appropriate for inclusion in a closed-loop, feedback-controlled replenishment system to stabilize the compositions of acid copper electrolytes employed for advanced semiconductor interconnect metallization. Electroanalytical techniques such as CVS and CPVS can provide rapid measurements with low hardware investment, minimum reagent cost, and little added waste disposal burden. However, matrix effects make it difficult precisely to correlate observed changes of electrochemical activity to variations in individual bath components over extended times. Liquid chromatography can speciate a wide variety of electrolyte components with less ambiguity, though it results in greater overhead in methods development, measurement time, waste generation, and system/operating cost.
At this time, it appears that the most workable approach is one in which both voltammetry and chromatography are integrated into a hierarchical method that exploits their relative strengths while minimizing their shortcomings. Such a multilevel approach might include CVS or CPVS systems integrated into individual ECD systems, and dedicated to tracking concentration changes of organic additives. An on-board CPU could perform trend analysis and make calls to a replenishment unit. Frequent analysis minimizes the chance that large changes in any component (organic or ionic) could result in electrolyte changes seriously masked by matrix effects.
Since voltammetry cannot resolve all significant bath changes, a chromatographic system could be configured to accept sample streams from a number of plating systems. HPLC or IC analysis would take place on a less frequent basis, gated either by a periodic sampling schedule or episodically as dictated by individual bath component excursions. The chromatography results could modify the transfer functions linking a tool`s on-board voltammetry-based analyses to organic additive replenishment, initiate less-frequent dosing of inorganic species, or indicate the need for electrolyte treatment (e.g., activated carbon filtration for removal of organic contamination) or replacement. n
Acknowledgments
The authors wish to thank the engineers and technicians of Semitool`s Advanced Technology Group laboratory; the chemists and technologists at Shipley Co. involved in electrolyte development and characterization; the members of SEMATECH`s advanced interconnect team focused on copper metallization and integration; Traci Deglow and Jennie Hollingsworth of Semitool for manuscript production; and the employees of Dionex Corp., ECI, EG&G, and Enthone-OMI for their material and intellectual contributions to advancing the state-of-the-art in copper electroplating for microelectronics applications.
Registered trademarks include LT-210 , Semitool; ELECTROPOSIT 1100 and ELECTROPOSIT Bath Analyzer, Shipley Co; CUBATH M, M D, and M LO 70/30 Special, Enthone-OMI; Potentiostat/Galvanostat Model 263A, EG&G; Qualiplate 4000, ECI; and DX 500, Dionex.
References
1. D. Edelstein et al., "Full Copper Wiring in a Sub-0.25 ?m CMOS ULSI Technology," Proc. IEEE IEDM, pp. 773-776, 1997.
2. E.M. Zielinski et al., "Damascene Integration of Cu and Ultra-Low-k Xerogel for High-Performance Interconnects," Proc. IEEE IEDM, pp. 936-938, 1997.
3. R. Haak, C. Ogden, D. Tench, "Cyclic Voltammetric Stripping Analysis of Acid Copper Sulfate Plating Baths, Part 1: Polyether-Sulfide-Based Additives," Plat. & Surf. Fin., April 1981.
4. E.K. Yung, L.T. Romankiwand, R.C. Alkire, "Plating of Copper into Through-Holes and Vias," J. Electrochem. Soc., pp. 206-215, 136, 1, 1989.
5. T. Pearson, J.K. Dennis, "Effect of Pulsed Reverse Current on the Structure and Hardness of Copper Deposits Obtained from Acidic Electrolytes Containing Organic Additives," Surface and Coatings Tech., 42, pp. 69-79, 1990.
6. V.A. Lamb, et al., "Physical and Mechanical Properties of Electrodeposited Copper: I. Literature Survey," Plating, pp. 1289-1311, Dec. 1965.
7. H.J. Wiesner, W.P. Frey, "Some Mechanical Properties of Copper Electrodeposited from Pyrophosphate and Sulfate Solutions," Plat. & Surf. Fin., pp. 51-56, Feb. 1979.
8. D. Anderson, R. Haak, C. Ogden, D. Tench, J. White, "Tensile Properties of Acid Copper Electrodeposits," J. Appl. Electrochem., pp. 631-637, 13, 1985.
9. W. Freitag, et al., "Determination of the Individual Additive Components in Acid Copper Plating Baths," Plat. & Surf. Fin., Vol. 70, (10), 55, 1983.
10. P. Bratin, "New Developmenets in Use of CVS for Analysis of Plating Solutions," Proceedings of AES Analytical Methods Symposium, Chicago, IL, 1985.
11. K. Haak, "Ion Chromatography in the Electroplating Industry," Plat. & Surf. Fin., Sept., 1983.
12. Dahmen, p. 346.
13. R. Haak, et al., "Cyclic Voltammetric Stripping Analysis of Acid Copper Sulfate Baths, Part 1: Polyether-Sulfide-Based Additives," Plat. & Surf. Fin., April 1981.
14. R. Haak, C. Ogden, D. Tench, "Cyclic Voltammetric Stripping Analysis of Acid Copper Sulfate Baths, Part 2: Sulfoniumalkanesulfonate-Based Additives," Plat. & Surf. Fin., March 1982.
15. Qualiplate QP-4000 User`s Manual, Vol. 1.4, Jan. 26, 1993.
THOMAS TAYLOR received his BS in chemical engineering from the University of Utah. He is general manager of the Electrochemical Deposition Division of Semitool Inc., 655 W. Reserve Drive, Kalispell, MT 59901; ph 406/752-2107, fax 406/752-5522.
TOM RITZDORF received his BS in chemical engineering from Montana State University and his MS in chemical engineering from the University of Minnesota. At Semitool, he has been intensely involved in the development of electrochemical deposition processes.
FRED LINDBERG received his BS degrees in both chemistry and geology from Metropolitan State College, Denver, CO. He is chief analyst in charge of quality assurance of ECD plating baths at the Semitool lab.
BRAD CARPENTER received his BS in chemical engineering from the University of Minnesota. At Semitool, he supports the process development of copper metalization with SEMATECH`s Interconnect program.
MARK LEFEBVRE received his BS in chemical engineering from the University of Lowell. He is PWB R&D license and acquistion manager for Shipley Co.
Adapted, with permission, from a paper presented at the 193rd meeting of the Electrochemical Society, San Diego, CA, May 3-8, 1998; proceedings to be published in Electrochemical Processing in ULSI Fabrication.