Chemistry is Key to ECMP Efficiency

Electrochemical mechanical polishing (ECMP) technology uses electrical charge to facilitate planarization. Applied potential oxidizes metallic copper to its ions (Cu+ and Cu++). These copper ions then react with chemistry in the electrolyte to form a passivation layer. The inhibitor is a critical component in the electrolyte used for ECMP because it affects the passivation film, planarization, removal rate and surface finish.

Passivation effects on a copper surface have been studied for many years.^1-4 During conventional chemical mechanical planarization (CMP), the passivation film was found to be closely related to removal rate, surface finish and planarization efficiency.^5-8 The passivation layer involves a copper surface that is at least partially oxidized to copper oxide by an oxidizer and then an inhibitor. Passivation film formation is critical to the CMP process, and depends largely on choosing the right inhibitor compound and concentration. Since ECMP technology uses electrical charge, the oxidizer is no longer needed. Therefore, the passivation film is formed by the reaction between the inhibitor and cupric (Cu++) ions. This makes its bonding to the copper surface weak and easy to remove with a minimum of downforce and/or abrasion. Thus, one of the main factors affecting the ECMP passivation film is the concentration and efficiency of the inhibitors.

System evaluation

We used a modified three-electrode system consisting of a working, counter (platinum wire mesh) and a reference electrolyte (saturated silver chloride/silver [AgCl/Ag]) in benchtop experiments (Fig. 1) to simulate the inhibitor’s passivation capability in trench and field areas on patterned wafers (Fig. 2 ). We performed the design of experiments (DoE) with and without rotation of the working electrode (copper rod). The counter electrode material was platinum mesh, and the pad material was polyurethane IC1000.

1. Test cell to understand inhibitor behaviors and simulate the differences between trench and field areas during ECMP.

In this design, the pad could be removed and put back to simulate non-contact and contact conditions during patterned wafer polishing using ECMP. The working electrode was connected to a rotary motor whose velocity was controlled by electrical potential. The copper rods used in the electrochemical experiments were polished mechanically in the electrolyte for 120 sec before each experiment to ensure a fresh copper surface.

2. Before ECMP, the patterned wafer has overburden in dense arrays and recess in large trenches.

A computer-controlled potentiostat/galvanostat was used in the electrochemical experiments. All experiments were performed in a 300 mL glass cell. Potentiodynamic polarization curves were generated by scanning the potential from -0.5 to 2.7 V at a scan rate of 10 mV/sec, with and without rotation. Each experiment was run with fresh electrolyte.

To simulate the inhibitor’s passivation capability in trench and field areas on a patterned wafer, two extreme conditions were used. The trench was assumed to be 100% not contacted by the pad, while the field area was 100% in contact with the pad during wafer polishing. A no-rotation I-V curve simulated the trench area of the pattered wafer because there was no abrasion between the copper rod and pad. An I-V curve with rotation simulated the field and protruded area of the patterned wafer during polishing, because the rotation of the copper rod generated abrasion between the rod and pad.

The passivation capability of the inhibitor is reflected by the current intensity difference during cyclic voltametric experiments. In the case of the I-V curve with rotation, the current density reflected the removal rate in the field area where the passivation film was removed by pad abrasion. In the case of the I-V curve with no rotation, the current density reflected the removal rate in the trench. The copper rod could be rotated at different rpm to simulate the speed effect of the platen on the copper wafer.

We conducted a DoE with different downforce and bias voltage and a proprietary polishing pad configured with a counter electrode underneath it. This counter electrode was divided into three concentric areas that formed the electrical zone control of the polishing process. Each zone operated independently with regard to electric potential and time. Experiments were performed to tune the control of each zone and achieve a flat profile.

Inhibitor comparisons

The inhibitor choice affects factors such as planarization, removal rate and surface finish. We tested many inhibitor reagents using dynamic anodic polarization. Figure 3 summarizes the anodic polarization study for current vs. voltage against a reference electrode (Ag/AgCl). The results showed a wide range of passivation among different inhibitors with the same molar concentration. Inhibitor A attained the highest passivation strength.

3. The passivation layer is weak, therefore copper dissolution occurs at low potential (~0.5 V). As effective inhibitor concentration increases, the passivation layer forms on the copper surface, preventing copper dissolution.

We also considered inhibitor concentration. Without inhibitors, the current gradually increases with the potential. This indicates that the passivation layer is weak and that copper dissolution occurs immediately with a potential of 0.5 V and above. However, as the effective inhibitor concentration increases, a passivation layer is formed on the copper surface to prevent copper dissolution. The passivation layer became stronger with increased concentration.

Figure 4 shows anodic polarization curves for copper in the electrolyte. The top I-V curve was generated with abrasion of the copper rod and bottom curve without abrasion of the copper rod. The large current intensity difference between the two curves illustrates the removal rate difference. Removal rate was high when abrasion was used because it simulated the field area and constant passivation layer removal. Based on the same principle, the removal rate in the trench area was near zero over a wide potential range. Therefore, we concluded that the planarization efficiency was high in the electrolyte.

4. Simulation of the field area (abrasion of copper rod) compared with the trench area (no abrasion) shows a substantial removal difference, indicating high removal efficiency in the electrolyte.

Finally, we examined removal rate as a function of potential. We measured the electrical potential static polarizations of the copper rod in the electrolyte with and without rotation in different time periods and with varying potential (Fig. 5). In region 1, each potential polarization reached its maximum current in a short time and stayed at that level over time, for a wide potential range (1.0-2.5 V). We stopped the rotation of the copper rod in region 2, causing the abrasion between the rod and pad to stop. A fast current drop occurred in the first 3 sec, indicating the rapid formation of a thin passivation film when the abrasion stopped. The removal current then decreased gradually. In region 3, the rotation of the copper rod was restored, and the current went back to the same level as in region 1. There is a linear relationship between the removal current and voltage against a reference electrode (Fig. 6 ), indicating that the copper polishing removal rate is proportional to the electrical potential.

5. Over a wide potential range (1.0-2.5 V) with both active (650 rpm) and static copper rod rotation conditions, the removal potential rapidly reaches a constant maximum current, drops during static conditions when a thin passivation film forms, and returns to the previous level when polishing is resumed.

The passivation layer formed on a copper surface is the result of a reaction between the inhibitor with copper ions or copper oxide. In conventional CMP, it is believed that the passivation layer is formed through the reaction of the inhibitor with cuprous oxide (Cu₂O) and ions.^9-11 However, in ECMP, it is proposed that the passivation film is formed because of the reaction between the inhibitor and cupric ions (Cu++), and the bonding to the copper surface, which is much weaker than in conventional CMP. Therefore, the passivation layer in ECMP can be more easily removed with a low downforce.

6. Polishing removal rate of copper is proportional to applied potential.

Planarization efficiency

ECMP has the potential to deliver higher planarization efficiency than conventional CMP because of the direct correlation between removal rate and electrical potential combined with low necessary downforce. Our results demonstrated that the removal rate in ECMP was only proportional to the electrical potential and not to downforce.

During ECMP, the current will follow the least resistive route. Therefore, even a thin passivation layer on trench areas can prevent copper dissolution from underneath while the field area is being polished. This leads to a great removal difference between the field and trench area. Since removal efficiency is defined as the ratio between the step-height change and removal, the high downforce of conventional CMP tends to remove the trench bottom material to some degree, lowering the planarization efficiency, compared with that in ECMP.