Factors influencing damascene feature fill using copper PVD and electroplating
07/01/2000
Jon Reid, Steve Mayer, Eliot Broadbent, Novellus Systems, Portland, Oregon
Erich Klawuhn, Kaihan Ashtiani, Novellus Systems, San Jose, California
overview
This article presents several determinant factors for successful integration of PVD copper seed and electroplating. A general description of copper-plating chemistries is given. Based on a large volume of experimental work, a summary of filling trends and observations is provided, along with two proposed mechanisms for bottom-up fill performance.
 Novellus' SABRE-xT copper electrofill system is shown moving a 300mm wafer into the plating cell. |
Copper physical vapor deposition (PVD) seeding and subsequent fill by electroplating have become the processes of choice to produce void-free fill of high-aspect-ratio (AR) damascene features. To achieve this filling performance, as well as to extend it to future device geometries, the incoming feature profile, seed layer attributes, and key electroplating process and chemistry parameters must be optimized to encourage acceleration of deposition near the base of damascene features ("bottom-up" fill). The success of copper plating to fill high AR features is built upon the achievement of successful nucleation followed by rapidly accelerated Cu deposition at the feature base.
The initial process chemistries developed for printed circuit board (PCB) applications in conjunction with PVD copper seed provided sufficient capability to fill sub-0.25mm generation integrated circuit (IC) features. PCB chemistries were developed to achieve high copper ductility, a uniform thickness distribution across large geometric areas, and a similar plating rate between the board surface and within high AR openings (through-holes). The latter requirement differs distinctly for the IC application in which a sharply accelerated plating rate is desired within submicron IC features to achieve void-free fill.
Presently, many IC manufacturers are implementing copper at device dimensions where use of PCB process chemistries (and early PVD seed technologies) does not produce void-free filling. The extendibility requirement for filling smaller geometries has driven extensive development of the copper-plating and seed-layer deposition processes. As a result, improved seed-layer integrity, modifications of plating current waveforms, and new plating process chemistries [1] have led to significant filling improvements.
PVD seed characteristics
All electroplating processes require an electrically conductive surface film (seed) to promote nucleation and growth [2]. Our understanding and experience indicate that there are several key requirements for PVD seed to realize void-free integration with electroplating [3, 4]. These attributes are described briefly below and are summarized in Fig. 1 by example illustrations and fill results.
Coverage. Via and trench features must have sufficient, continuous seed coverage over their entire structure. Deposition from a highly ionized (directional) PVD source is typically required to achieve optimum coverage on the lower sidewalls and base regions [5]. In some cases, seed re-sputtering from the feature base can redistribute material onto the lower sidewalls (where typically the lowest PVD step coverage occurs) and enhance fill results.
 Figure 1. Example set of illustrations and cross-sectional images showing key seed attributes and their impact on electroplating fill results. |
Morphology. The seed layer must possess a smooth and continuous morphology. Copper agglomeration at the sidewall can cause center and/or bottom voiding during fill. Active wafer cooling during seed deposition (such as that provided by a cold electrostatic chuck operated at a very low temperature) is required to avoid agglomeration of Cu material on the underlying barrier. With a continuous seed layer, plating readily initiates on all surfaces, thereby allowing subsequent bottom-up fill, as shown in the sequence in Fig. 2a. In contrast, plating onto agglomerated or discontinuous seed material can result in voids near the base, which develop when nucleation fails on the lower sidewalls or base of the vias. For the seed shown in Fig. 2b, both a failure to initiate plating and evidence of seed dissolution near the via base have produced a large bottom void following plating.
Oxidation. Seed oxidation and corrosion must be minimized both prior to plating and during bath immersion. Oxidation or corrosion can cause seed discontinuity and fill voiding. The seed surface must be largely free of oxides for efficient charge transfer to begin during plating. PVD copper seed films with thin (10-20Å), acid-soluble oxide layers easily meet this criterion for damascene plating. Where thicker (50-100Å) cuprous oxides have been produced, however, a greater degree of bottom voiding is observed (Fig. 3).
 The SABRE xT plating cell. |
Feature overhang. Barrier and seed deposition must not have significant overhang. Excessive overhang can cause premature feature close-off during electroplated fill. Deposition from a highly ionized (directional) PVD source provides a greater degree of uniform coverage while minimizing any resulting overhang. Minimal re-sputtering is desired to maintain minimal overhang. (This is in contrast to the beneficial effect of re-sputtering noted for the preceding coverage requirement.)
Asymmetry. The seed layer must possess minimal coverage asymmetry between radially inward- and radially outward-facing sidewalls. Too thin a seed on one side can produce nucleation difficulties and result in void formation. Control of directional deposition must be maintained across the wafer plane from center to edge.
Seed layer immersion
Cuprous oxide is difficult to reduce electrochemically to copper [6]. However, the Pourbaix diagram shows that both Cu2O and CuO are unstable in a strongly acidic medium (pH <3) [7] because of their spontaneous decomposition with hydrogen ion, according to the reactions below:
Cu2O + 2H+ -> Cu+2 + Cu + H2O
CuO + 2H+ -> Cu+2 + H2O
 |
The seed layer also can be dissolved when it is exposed to the plating bath. One process by which this can occur is corrosion of the copper by dissolved oxygen. Additives that oxidize copper can have a similar effect.
RIGHT - Figure 2. Cross-sectional via images showing (left to right): a) smooth seed coverage only, metal profile following partial fill, and complete fill by electroplating, and b) agglomerated seed coverage, metal profile following some electroplating, and resulting void at completion of electroplating.
The growth of an electrodeposited metal film requires a nucleation center that has reached a critical dimension such that the free energy change of deposition is negative. Unfortunately, the reverse of this process can occur if a seed layer has growth centers that are smaller than these critical dimensions. This process is termed "Ostwald corrosion" [6]. This phenomenon occurs because of the difference in potential between metal deposition on a flat local surface compared to one with substantial curvature (such as that with a rough or agglomerated surface). The micro-rough metal surface within a feature can produce a local micro-corrosion cell.
Rapid uniform wetting of the wafer surface and the inside of high AR features by the plating solution is required for good filling performance. For a solution with good wetting properties on the copper surface (a low contact angle), it can be easily shown that capillary action should force solution several meters into submicron features. The contact angle of the plating solution on the wafer surface is a function of both the plating bath chemistry and the oxidation properties of the copper surface. Contact angle measurements (Fig. 4 on p. 92) show a high degree of variability between different seeds and across (center-to-edge) a single-seeded wafer. In general, oxidation and poor filling performance are associated with high contact angles, but the nature of the relationship is still under study.
Electroplating process
Figure 5 on p. 92 shows an example of a two-step DC plating process with three important features identified. The first feature is termed "zero current induction time," or "induction time" for short. Induction time is the time that the wafer is exposed to the plating solution prior to the application of cathodic current (or potential). Usually the wafer is being wetted during this period. It represents the time period over which the part is exposed to the plating electrolyte at an uncontrolled (floating) potential. If the part is not polarized during this time, copper corrosion, Ostwald ripening, and cuprous oxide dissolution can occur. If the wafer is introduced into the bath with an immediate small current flowing (or a small cathodic potential applied), these corrosion reactions can be reduced or reversed. This process is commonly referred to as cathodic protection.
 Figure 3. Cross-sectional via images showing that the extent of bottom void formation correlates well to the degree of seed layer oxidation. |
The second feature in Fig. 5 is termed "initiation current." During this step, a relatively small current is applied to the wafer. Because the current is small, the surface resistance is large, the rate of deposition is slow, and deposition is conformal. During this time, the seed layer within the feature will thicken. If discontinuous, it can also become continuous during this step. (In some cases, a thicker continuous seed might be required to support the nucleation and high current density required in the subsequent fill step.)
The last feature in Fig. 5 is the fill step, which typically lasts 20-40 sec, depending on bath conditions and feature size. Smaller features tend to fill earlier because they are less accessible to the electrolyte and develop conditions necessary for bottom-up fill more rapidly.
A final required process step (not shown) is the large feature-filling step. This step is required to fill features with ARs much less than unity (such as test or bond pads), and typically requires the deposition of copper to a thickness somewhat greater than that of the damascene dielectric film stack.
Copper-plating chemistries
Copper electroplating baths are normally formulated using a highly stable base electrolyte solution containing copper sulfate and sulfuric acid [8]. The basic kinetics and solution properties of these solutions have been studied for more Than 50 years and are well understood.
In IC damascene applications, the sole important criterion for copper sulfate concentration is to avoid the depletion of cupric ion within high AR features during filling processes. Typical cupric ion concentrations in use today are in the range of 17.5-60g/L. Sulfuric acid is usually added to the plating electrolyte (45-325g/L) to increase solution conductivity and improve wetting or oxide dissolution on seed surfaces. In general, more conductive solutions result in a system where plating thickness distribution is less dependent on plating cell geometry, while low acid electrolytes result in a system with less dependence on seed layer resistivity.
 Figure 4. Contact angle measurements obtained from several different seeds using a small droplet of plating solution placed onto the surface. |
Organic additives added to copper electroplating baths fall into three basic categories. Accelerators are molecules that contain pendant sulfur atoms (such as organic mercapto species) that locally accelerate current at a given voltage where they adsorb. Accelerators are usually present in the plating bath in the concentration range of 1-25ppm. Suppressors are polymers such as polyethylene glycol, which tend to form a current-suppressing film on the entire wafer surface, especially in the presence of chloride ion (which can be considered a co-suppressor). Suppressors are usually present in the plating bath at high concentrations (200-2000ppm) so that their concentration at the interface is not strongly dependent on their rate of mass transfer or diffusion to the surface.
Levelers are a second class of current-suppressing molecules, which are usually added to the plating bath at a low concentration. Hence, unlike suppressors, the concentration of levelers at the interface is mass-transfer-dependent. In this way, isolated locations such as the inside of a via (where mass transfer is limited) are less suppressed, while protruding surfaces or corners (to which mass transfer by diffusion or migration is more efficient) are more suppressed.
Filling trends and observations
Any proposed mechanisms for bottom-up fill should explain several measured trends in filling performance as a function of process chemistry conditions. (These trends relate strongly to the elimination of center voids in fill experiments and somewhat less to the ability to eliminate bottom voids.) The following outline of points can be made in summarizing our experience and observations in working with Cu plating bath additive systems [9]:
1. Strongly accelerated fill has not been achieved using suppressing polymers alone, or suppressors with chloride ion. (Growth in several tested chemistry systems appears to be largely conformal.)
2. Suppression of polymer concentration can be increased to at least 4x beyond an optimal (plateau) level without impacting fill. These concentrations can extend upward of 1000mg/L, well beyond a level at which concentration gradients within features are likely to develop.
3. Numerous polymers have been found to yield bottom-up fill when used as accelerator, although very strongly adsorbing polymers generally result in conformal fill.
 Figure 5. Example of a two-step DC plating process with three key features identified. |
4. Bottom-up filling can be achieved in the absence of levelers. Time evolution profiles obtained using an additive system containing only a polymer suppressor, chloride, and an accelerator show that an accelerated bottom-up growth component is operative.
5. The addition of levelers to bottom-up filling chemistries often results in top center voids, indicating a disruption of bottom-up growth. At adequately low concentrations, little impact to filling is observed.
6. As chloride ion concentration is increased from zero, the filling acceleration in vias increases to a maximum value but decreases as chloride concentration is further increased. (Conformal fill has been noted at very high chloride levels using some additive systems.)
7. As accelerator concentration is increased, bottom-up fill increases from near zero to a maximum rate and then diminishes to conformal behavior.
8. For a given additive system, too low a copper concentration diminishes bottom-up fill capability in high AR features.
9. In the absence of a leveler-like species, accelerated copper growth continues over a damascene feature set following bottom-up fill. The metal thickness profile over a set of trenches that exhibit bottom-up fill can be twice that of the adjacent field following formation of a 1mm plated deposit. The nearly 2x thickness increase of copper over dense features reflects a continuation of the accelerated copper growth beyond the time of filling completion. The addition of a leveler component serves to suppress current on the rapidly growing surface after it protrudes above the field, thereby leading to a relatively uniform deposit thickness.
10. In additive systems that fail to exhibit strong bottom-up fill using a polymer suppressor and an accelerator, only certain levelers can enhance filling. In this type of system, filling is often more complete in smaller rather than larger features, probably reflecting suppression of growth in large features that exhibit mass transfer characteristics similar to the field.
Mechanisms of superfilling
To date, the bottom-up fill process has been described in literature as being driven primarily by the establishment of a diffusion gradient of suppressing polymers [10]. The effect of leveling additives, which can reduce the copper growth rate at the entrance to features, also has been widely discussed [11]. Based on the observations of the preceding section, a primary mechanism (dependent on several factors), as well as a secondary mechanism (involving the leveler component), are proposed for the establishment and propagation of bottom-up fill.
Primary fill mechanism
When a wafer is first immersed in a plating solution, a concentration gradient of suppressing and accelerating species can exist on the wafer surface between the via base and the field. This will happen when the quantity of additive species required to form an adsorbed layer on the surface within the via exceeds the amount of additive contained in the solution volume within the via. This effect could account for less suppression at the via base by slow diffusing polymer species when relatively low suppressing polymer concentrations (<100mg/liter) are present in solution.
A simple calculation, however, indicates that unless a species is consumed rapidly, this concentration gradient will last one second or less for suppressing polymer species. These polymer species are present at up to several hundred mg/liter in solution and should not exist at all when polymer concentrations exceed 1000mg/liter (where bottom-up fill is still observed). It is well known that most polymer species used as suppressors are not readily consumed or decomposed at the cathode. Also, as noted in the filling observations, increasing the polymer level beyond that required for optimal fill does not degrade fill, as would be expected for a fill mechanism driven by a polymer diffusion gradient. Beyond this, it is observed that bottom-up fill will take place on a well-seeded surface if either:
a) a wafer is immersed in a plating solution with the current on, or b) the wafer is placed in the bath and allowed to equilibrate with the additives in solution for at least 10 sec.
 Figure 6. Illustration of additive adsorption behavior during plating fill. A mechanism for the establishment and propagation of bottom-up fill is suggested. |
For these reasons, it appears that diffusion-limited adsorption of the polymer is not critical to bottom-up fill in at least some commercially available additive systems.
When current flow begins, we will assume that all additive species have reached an equilibrium level on all surfaces of the wafer. (Refer to the illustrated sequence provided in Figure 6 for this discussion.) This should be expected to lead to initial currents that are approximately equivalent on all surfaces. This effect agrees with the observed relatively small amount of bottom-up (i.e., conformal) growth seen in the first 5-10 sec of a filling process.
After some time (which depends on current density, chemical concentrations, and feature size), two phenomena might begin to contribute to fill. First, the accumulation of accelerating mercapto species (or their more accelerating breakdown products) on the surfaces within the features takes place. Accumulation results as surface area within the via decreases and adsorbed mercapto species (which are neither incorporated in the deposit nor desorbed into solution) are thereby concentrated. Current increases in the areas of geometric concentration (bottom corners) as chloride and suppressing polymer is displaced.
Too much accelerator in a bath disrupts fill because the accumulation of accelerating species also begins to take place on the wafer surface, and differentiation from the via base is lost. The accumulation of catalytic species on the growing surfaces within features is strongly supported by the continued rapid Cu growth above features in the absence of leveler. Discontinuing current flow to allow polymer re-equilibration does not interrupt such behavior. It is, however, disrupted by reversal of interfacial potential to a value causing oxidation or desorption of the adsorbed catalytic material, or by addition of a leveling additive which suppresses current at protruding geometries.
To explain the observed effect of chloride on fill, a second phenomenon that contributes to fill initiation can involve the co-deposition of chloride in the Cu film. This results in a depleted chloride concentration within features, or competitive displacement of Cl on the copper surface by the accumulating mercapto species. Low chloride concentration leads to weak or negligible polymer adsorption and higher current for a given potential, as has been well documented in literature [12]. This effect results in poor suppression of the current at the wafer surface and, therefore, a lack of possible relative acceleration in the feature.
As Cl concentration is increased, good surface suppression is obtained. The levels of Cl within the via might not be adequate to maintain polymer adsorption in the presence of the accumulating accelerator mercapto species, however, and good bottom-up fill is obtained. Further increase in Cl can disrupt fill because Cl is able to compete successfully with the mercapto species for adsorption sites and subsequently attract suppressing polymer adsorption within the via.
Finally, after bottom-up fill begins, the high current that is initially established will tend to continue as adsorbed accelerator species accumulate on the growing front.
Secondary fill mechanism
The secondary fill mechanism involves the leveler component and is simply understood by the enhanced suppression of current flow at surfaces to which the mass transfer of leveler is most rapid. This results in a growth rate that increases in low mass-transfer areas such as narrow or deep features. In general, it appears that this mechanism yields only moderately increased growth rates within features unless combined with the effects of accelerator accumulation. The details of this type of mechanism have been described elsewhere [13].
Conclusion
Improved integration of copper PVD and electroplating processes has provided significant advances in void-free fill of high AR features. A discussion of factors influencing feature fill has been presented, including seed layer attributes, electroplating process considerations, and bath chemistry formulation. Founded on a large volume of experimental work, a mechanism for bottom-up fill performance was presented based primarily on the accumulation of accelerating species within damascene features. Such understanding opens the way for continued extendibility of fill by copper seed and electroplating methods.
Acknowledgments
Our thanks to the many engineers from the Novellus Portland and Novellus San Jose facilities who contributed to the technical work presented in this article.
References
- J. Reid et al., "Optimization of Damascene Feature Fill for Copper Electroplating Process," Proc. of the Int'l. Interconnect Technology Conf., p. 284, IEEE Cat. No. 99EX247, 1999.
- E.K. Broadbent et al., "Experimental and Analytical Study of Seed Layer Resistance for Copper Damascene Electroplating," J. Vac. Sci. & Technol. B17, p. 2584, Nov./Dec. 1999.
- J. Reid and S. Mayer, "Factors Influencing Fill of IC Features Using Electroplated Copper," Proc. of the Advanced Metallization Conference, 1999.
- S. Mayer et al., "Integration of Copper PVD and Electroplating Processes for Fill of Damascene Features," Ext. Abstracts of the 196th Mtg. of the Electrochemical Society, Vol. 99-2, Abstract No. 732, 1999.
- E. Klawuhn et al., "Ionized Physical Vapor Deposition (PVD) using Hollow-Cathode Magnetron (HCM) Source for Advanced Metallization," 46th Int'l. AVS Symposium, Abstract No. 1466, 1999.
- S.T. Mayer, "In Situ Studies of Electrochemical Cu, Ag, and Zn Film Formation," UC Berkeley Ph.D. thesis, LBL Report #28085, pp. 1-37, 76-78, Dec. 1989.
- N. De Zoubov, C. Vanleugenhaghe, and M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, Marcel Pourbaix, Ed., Pergamon Press, New York, 1966.
- F.A. Lowenheim, Modern Electroplating, Chapter 7, Wiley & Sons, New York, 1963.
- J. Reid and S. Mayer, "Mechanisms of Filling Sub-Micron Features Using Electroplated Copper," Proc. of SEMI Korea Technical Symposium 2000, Semicon Korea, February 15-17, 2000.
- H. Deligianni et al., "A Model of Superfilling in Damascene Electroplating," Ext. Abstracts of the 195th Mtg. of the Electrochem. Soc., Vol. 99-1, Abstract No. 267, 1999.
- J. Kelly, C. Tian, and A. West, "Leveling and Microstructural Effects of Additives for Copper Electrodeposition," J. Electrochem. Soc., Vol. 146, p. 2540, 1999.
- M. Yokio, T. Hayashi, and S. Konishi, "Adsorption Behavior of Polyethylene Glycol on the Copper Surface in an Acid Copper Sulfate Bath," Denki Kagaku, Vol. 52, p. 218, 1994.
- A. West, "Theory of Filling of High-Aspect Ratio Trenches and Vias in Presence of Additives," J. Electrochem. Soc., Vol. 147, p. 227, 2000.
Jon Reid received his PhD in chemistry from the University of North Carolina at Chapel Hill, after which he was involved in electroplating process development at IBM. Since 1996, he has been at Novellus Systems as a technologist for the Sabre copper program. Novellus Systems Inc., 26277 SW 95th Ave., Suite 402, Wilsonville, OR 97070; ph 503/685-8353, fax 503/685-8399, e-mail [email protected].
Steve Mayer received his PhD in chemical engineering from UC Berkeley in 1989. He spent four years at Lawrence Livermore National Laboratories developing damascene copper electrodepositions and electropolishing processes, as well as performing energy-storage research. In 1993, he founded PolyStor Corp. He is currently a process technologist at Novellus Systems.
Eliot Broadbent is chief scientist for Novellus Systems, where he has worked for 11 years on the development of thin film process systems. He previously worked for Philips, specializing in the areas of CVD metals, silicide formation, diffusion barriers, and aluminum interconnection. He has written extensively and holds 22 patents.
Erich Klawuhn is process development manager for Novellus Systems. He has four years of experience in development of the copper barrier seed process used for copper interconnect technology.
Kaihan Ashtiani received his PhD in electrical and computer engineering with emphasis on plasma physics from the University of Wisconsin-Madison. He worked in the PVD division of Materials Research Corp. before joining Novellus Systems in 1997. He is an engineering program manager involved in development of the hollow cathode magnetron ionized PVD source.