Planarization by Electrochemical and Mechanical Actions

As noted above, while electropolishing is a cost-effective method of Cu overburden removal, it does not have significant planarization capability, especially for large features with low aspect ratio. A combination of electrochemical and mechanical means of Cu removal has been studied as a novel planarization technique. This has involved mainly two types of planarization techniques as shown in Figure 12.4 [36,37]. In one approach, the planarization is achieved during electroplating using electrochemical mechanical deposition (ECMD) which is followed by electropolishing to remove the overburden, while in the other approach electroplating is followed by ECMR Both approaches emphasize electropolishing as the key metal removal method, thus positioning electropolishing at the center stage in the planarization of interconnect structures.

FIGURE 12.4 Two types of planarization techniques: one involving (a) ECMD followed by electropolishing and the other (b) involving electroplating followed by ECMP [36].

Electrochemical Mechanical Deposition

ECMD process described by Basol et al. [37-40] and Jeong et al. [41] involves simultaneous electrochemical metal deposition and removal. Mechanical action during electrodeposition is accomplished by the sweeping action of a pad on the wafer surface. The plating bath does not contain any abrasive slurry. Therefore, the mechanical action is purely due to the sweeping action of the pad. Sweeping of the pad removes Cu from the top surface, thus allowing Cu growth rate to be higher in the features. Chemical modifications of the plating bath in combination with ECMD may make the planarization action more effective without the need to apply pressure to the pad.

Basol et al. [39] investigated the planarization efficiency of the ECMD technique as a function of various process parameters. Planarization w'as found to be a strong function of the nature of the organic additives in the copper plating bath. While no planarization was observed in an additive-free bath and a bath containing only accelerators, a small degree of planarization was found in a bath containing only suppressors. The best copper layer planarization results and efficiencies over 70% were observed in plating electrolytes containing both accelerators and suppressors. This is due to a differential in the relative surface coverage of additives on the top surface vs. the cavities. Sweeping by the planarization pad reduces additive coverage at the top surface but leaves the cavities with preferential coverage at the bottom. During the plating period, more of the current flows into the cavities where the accelerator-to-suppressor surface coverage ratio is higher compared to the top surface. The addition of levelers into the plating bath was found to be detrimental for planarization efficiency. Planarization efficiency was found to increase with increased accelerator concentration within the process window used in this study. Higher planarization was observed in the high-acid electrolyte compared to the low'-acid electrolyte.

Figure 12.5 shows an example of planar deposition by ECMD followed by electropolishing to remove copper dowm to 100 nm thickness [40]. In principle, ECMD has unique capabilities as compared w'ith the standard electrodeposition process. Metal layers deposited by ECMD are planar and the thickness of the overburden is smaller than the films deposited by standard approaches. ECMD achieves

FIGURE 12.5 Top: Schematic diagram of the ECMD tool. Bottom: Cross section of parts planar deposited by ECMD followed by electropolishing, (a) After planar electrodeposition and (b) after copper electropolishing down to 100 nm thickness [40].

this result by enhancing material deposition rate into the cavities while retarding or minimizing deposition on the substrate top surface. The thin and planar copper deposits such as those shown in Figure 12.5 are very attractive for etching, electropolishing, and CMP. For etching and electropolishing, a planar layer offers the possibility of removing the overburden in a planar manner without causing excessive dishing into the large features. For CMP, ECMD offers significant cost advantage due to thinner copper to be removed.

Electrochemical Mechanical Planarization

In electrochemical mechanical planarization (ECMP), metal removal and planarization are accomplished by a combination of the virtues of electropolishing and CMP. Compared to conventional CMP, ECMP allows ten times lower down-force for planarization. A pad is used in conjunction with an electrolyte for electropolishing. Figure 12.6 shows the schematic of a typical ECMP equipment. It consists of a cathode plate between the platen and the pad, while the wafer is attached to a rotating mount which is connected as an anode. There is provision for the supply of the electrolyte above the pad. The holes in the pad are filled with the electrolyte and form electrical contact between the wafer anode and the cathode. The planarization efficiency of ECMP depends on the electrolyte composition. The electrolyte used in ECMP is generally H,P0₄ based, containing additives that can form a complex on the Cu surface. Electrolytes based on H₂S0₄ or HNO, have also been reported. The passive film formed must provide a certain degree of resistance to copper dissolution in the recessed areas but at the same time, they must be soft enough to be removed by the mechanical abrasion of the polishing pad at low down pressures (<0.5 psi).

Copper film on the Si wafer is polarized anodically while the polishing pad mechanically wears the metal surface. At the surface where the pad is in contact with the wafer, more Cu removal takes place due to both electrochemical and mechanical action. On the other hand, only electrochemical action removes material from

FIGURE 12.6 Schematic diagram of an ECMP tool.

locations of the wafer where contact with the pad is absent. The contact of the polishing pad with the copper surface is maintained at a low-down pressure so that the metal film does not delaminate from the low-k/ULK film. As a result, defects such as dishing, erosion, and scratch that are common in the conventional CMP process are minimized. The applied charge controls the material removal rate (MRR). Electrochemical dissolution leads to the formation of metal ions and passivation film on the metal surface. The low-lying areas of the film are protected by the passivation film while the protruding features are polished by the polishing pad. However, the problem of dishing, particularly in wide features, continues to remain the key concern in the ECMP thus requiring optimization of the process parameters. In the following, some of the published literature related to the influence of electrolyte composition and mechanical factors on the ECMP process is briefly described.

Mechanical Factors

The mechanical factors such as the pad material, pad design, and the abrasive in the electrolyte play a significant role in influencing the ECMP performance. A polymeric pad must have mechanical integrity and chemical resistance to survive the rigors of polishing. Mechanically, a polishing pad should have acceptable levels of hardness, and modulus, and good abrasion resistance to endure the Cu ECMP process. Chemically, a polishing pad should be able to survive the electrolyte chemistries, which include either highly alkaline or highly acidic electrolytes. Jeong et al. [42] measured compressibility, elastic recovery, permanent deformation, viscoelastic property, and other time related physical properties of two types of pads: Polyurethane pad (IC 1400 к-groove), polymer impregnated felts pad (Suba 600). The mechanical properties of the polishing pad were measured under three different conditions; dry pad, pad soaked in the deionized water, and pad soaked in the electrolyte containing a mixture of H,P0₄ 6 wt.%, H,0₂ 0.5 wt.%, BTA 0.5 wt.%, glycine 0.5 wt.%, and citric ammonium 5 wt.% for 16 hours. Based on their results, they recommended the use of the polyurethane pad, which has stable viscoelastic behavior and high chemical attack resistance in the electrolyte. Because the hardness of the polyurethane pad was higher than that of the polymer impregnated felts pad, better global uniformity was achieved, and the metal removal rate was also high during the ECMP process. Jeong et al. [42] also investigated the effect of abrasive on the uniformity of the wafer scale. They used Colloidal silica consisting of stable dispersion of amorphous silica particles. To achieve stable dispersion, not affected by gravity, the silica particle size of the order of 20 nm were used (Ludox TM colloidal silica with mean particle diameter: 22 mm, silica concentration: 50 wt.%). Different concentrations of the colloidal silica abrasive ranging from 0% to 50% of the colloidal silica were used in the electrolyte cited above. Their results showed that the surface roughness, which is around 50 nm without the abrasive, goes through a minimum (16 nm) at 10% abrasive. Within the wafer nonuniformity, which is around 10%-12% without the abrasive, improves to 2% at an abrasive concentration of 10% or higher.

ECMP pads require punching holes to ensure electrolyte contact between the cathode and the substrate surface during polishing. The differences in the arrangement of holes on the polishing pad may cause uneven distribution of electrochemical action on the substrate, which may result in different MRRs. Liu et al. [43] investigated the nonuniformity of material removal in ECMP by using track point density distribution. The coefficient of variation of track point density and density distribution of two polishing pads at different speed ratios were simulated to represent the uniformity of electrochemical action across the substrate. The two types of pad designs used by Liu et al. [43] consisted of concentric-type holes and phyllotactic-type holes. The simulation results were verified by experiments. The simulation and experimental results showed that in pads with concentric-type arrangement, the track points produced by the holes have a ring-shaped ripple distribution on the substrate, and the track point density at the peak ring is large. For pads with the phyllotactic arrangement, the track point density distribution is generally uniform. With increasing radius, the track point density decreased slightly, indicating that the electrochemical effect will decrease gradually as the radius increases. Pads with phyllotactic holes showed improvement in substrate flatness. Both the simulation results and the experimental results showed that the material removal nonuniformity of the phyllotactic arrangement polishing pad is better than the concentric arrangement pad, indicating that the uniformity of the ECMP material removal can be represented by the density distribution of the track point on the substrate.

Kondo et al. [44] developed a carbon polishing pad for the ECMP process and conducted experiments on an orbital 300 mm CMP machine. Figure 12.7 shows a schematic diagram of the carbon pad. The pad consisted of a surface carbon layer acting as an anode, an intermediate insulating layer, and an underlying cathode sheet. More than 100 electro-cells were fabricated within this tri-layered structure, which was about 5-mm-thick. The intermediate insulating layer acted as a cushion layer to improve within-wafer nonuniformity. Soft carbon material was chosen so that the copper surface would not be damaged. The power supply was connected with the cathode at the edge of the pad. The carbon pad was stuck to the CMP platen with an adhesive sheet and could be easily replaced. This enabled easy changing a CMP

FIGURE 12.7 Schematic diagram of a carbon polishing pad for the ECMP process [44].

system into an ECMP system by replacing the conventional polyurethane CMP pad with a carbon one. Kondo et al. claimed that the use of carbon pad in the ECMP process resolved several issues such as scratching, copper residues, and cathode regeneration. They also claimed to have successfully fabricated 45-nm-node porous low-k/ Cu interconnects using an ECMP process followed by removal of the TaN barrier layer by CMP.

Electrolyte Composition

During ECMP, an anodic potential supplies the driving force for the dissolution Cu and transport of copper ions. Planarization in ECMP possibly occurs due to accelerated electrochemical Cu removal from surface elevations, where the passive film is abraded, coupled with little or no removal at surface recesses, where the passive film remains intact. Hence, the nature of the passive film formed during ECMP largely determines the planarization efficiency and the Cu surface quality. The material removal differential in ECMP can be made more effective by selecting conditions whereby surface passivation plays a major role in the electrochemical reaction. The nature of the passive film formed during ECMP is determined by the inhibiting or passivating agents in the electrolyte which in turn determine the copper removal rate and planarization efficiency. Benzotriazole (BTA), which is the most commonly used inhibiting species used in CMP, is ineffective in acid electrolytes and at high anode potentials encountered in ECMP. Several investigations involving the influence of different inhibitors and chelating agents have been reported in the literature [35,45-56]. Among several inhibitors studied, 5 phenyl- 1-H-tetrazole (PTA) was found to be the most effective for copper CMP at lower pH [51]. The importance of the use of a chelating agent in the electrolyte is the fact that the passive film thickness in such a system can be modulated by the applied anode potential. Hydroxyethylidenediphosphoric acid (HEDP) and oxalic acid have been found to be effective chelating agents for copper ECMP [49,50]. Another interesting manifestation of the dependence of metal removal on applied current is the possibility of varying applied current density at different radial zone on the wafer, thereby permitting desired differential removal rates at different locations [56].

All of these factors have permitted the planarization of copper during ECMP with significantly minimized to no particulates.

In general, copper is etched in acidic solutions because Cu is readily ionized to Cu²⁺ ions. In many cases the etch rate may be too high to control under very low pH conditions. In order to precisely control the polishing rate and reduce the dishing effect in the ECMP process, Oh et al. [54] used alkali conditions by forming copper oxide or copper ions which are readily ionized to HCuOy or CuO,^2- in an alkali electrolyte. The use of KOH electrolyte produces Cu hydroxide on the Cu surface which on the application of an anodic potential to the Cu wafer ionizes the Cu ions from Cu hydroxide to Cu²⁺ or Cu0₂^2-, and mechanical force is also applied to the Cu surface employing a soft polymer polishing pad. The mechanical effect induced by the polishing pad increases the uniformity and the polishing rate of the Cu layer. The chemical state of the Cu surface was studied by X-ray photoelectron spectroscopy analysis and found that the oxidation rate was increased by adding H₂0₂, which increases the polishing rate by increasing the surface oxidation of Cu to Cu(OH)₂. Furthermore, to prevent the dishing effect, Oh et al. [54] added a BTA inhibitor to the KOH electrolyte. It is generally known that BTA is easily adsorbed on the Cu surface and prevents the excessive etching of Cu during the ECMP process. The ECMP process was performed for the 3 pm trenched wafer with 20 wt.% KOH electrolyte, containing 0.005 M BTA at a current density of 9 mA/cm². Figure 12.8 shows that the Cu film was successfully polished and planarized, with the Cu remaining inside the trenches. These experimental results support that the BTA acts as an inhibitor and suppresses the dishing phenomenon. However, the process is very slow, and it takes 14 minutes to complete the planarization process. The addition of H₂0₂ to the electrolyte containing BTA, increased the etching rate, by augmenting the oxidation rate of the Cu surface to Cu(OH)₂. The etch rate increased as the concentration of H₂0, was increased from 1 to 7 wt.%. However, the etch rate was drastically reduced in the electrolyte containing 10 wt.% of H₂0₂. This indicated that H₂0₂ promotes the

FIGURE 12.8 Cross-sectional SEM images after the ECMP process in 20 wt.% KOH solution containing 0.005 M BTA [54].

generation of Cu(OH)₂, but an excess of H₂0₂ induces the generation of CuO rather than Cu(OH)₂ and the formation of CuO reduces the ionization of Cu to Cu²⁺.

Shattuck et al. [57,58] explored the use of phosphate salt as an electrolyte component to conduct the ECMP process at low pH. An electrolyte containing 1.0 M potassium phosphate salt concentration with a pH value of 2 and a BTA concentration of 0.001 M was tested for its planarization capability on patterned Cu structures using a custom-built ECMP tool. Planarization was achieved on patterned Cu structures with an abrasive-free electrolyte. Results indicated that using a low pH phosphate-based electrolyte and BTA, as an inhibitor, ECMP can provide the necessary metal removal rate to polish Cu while providing adequate passivation in the recessed regions to allow for planarization to occur.

Tripathi et al. [53] described electrochemical studies and polishing results during ECMP of blanket and patterned Cu wafers using electrolytes containing HEDP, PTA, and oxalic acid. Voltammetry measurements were made with and without abrasion. The nature of the passive film formed at different potentials was investigated by EIS and electrochemical quartz crystal microbalance (EQCM). Their results indicated that the Cu removal rate and the planarization efficiency during Cu ECMP can be approximated using electrochemical measurements of the Cu removal rate, with and without surface abrasion. These results predicted a 500 mV potential window within which the Cu removal rate is greater than 600 nm/min and the planarization efficiency is greater than 0.90. However, high planarization efficiencies are only obtained when silica abrasives are included within the ECMP electrolyte. In-situ EIS results indicate that the interfacial impedance is increased by the presence of silica, suggesting that silica is incorporated into the PTA-based passive film and is thus needed for effective planarization. EQCM experiments indicate that PTA may provide better Cu surface passivation at a high anodic potential than BTA, which is widely used during Cu CMP.

Choi et al. [55] investigated the influence of copper ion concentration on the kinetics of formation of a protective layer on copper in an acidic solution containing BTA and glycine. They observed that even modest additions of copper ions to the electrolyte impacted the kinetics of the formation of the protective layer on the copper surface. This is due to Cu²⁺ ions complexing with BTA and eventually nucleating and precipitating Cu(II)BTA₂. The higher the concentration of Cu²⁺ in the electrolyte, the more easily Cu(II)BTA₂ nucleates. Nucleation and growth of Cu(II)BTA₂ causes the desorption of BTA species from the adsorbed protective layer. This nucleation of Cu(II)BTA₂ followed by the desorption of BTA could be particularly problematic at recessed regions on the wafer surface. At recessed regions there would be less agitation, thus BTA has enough time to be adsorbed onto the copper surface and form protective layers or multilayers. If the electrolyte contains a sufficiently high concentration of Cu²⁺ ions and BTA, however, the adlayer formed on the surface may be destroyed. With the elimination of passivity, the recessed regions of the wafer would not be protected, which could lead to roughened topography, or at least longer times needed to achieve planarization. Even for low concentrations of copper ions in the bulk, high concentrations of copper ions may develop if mass transport of dissolved Cu²⁺ into the bulk is hindered. This suggests that the influence of copper ions on the kinetics of the formation of the protective layer must be considered when formulating new electrolytes or slurries for ECMP or CMP of copper. Also, the concentration of copper ions in the electrolyte must be controlled below a certain level to ensure the integrity of the protective material during the process. This might be achieved with lower residence times for slurries, and passage of the withdrawn slurry through a bed of a cation exchange resin that could replace Cu²⁺ by H⁺ ions.