1. Introduction

CMP is known as the only way to achieve global planarization with the perfect wafer surface obtained. With the integrated development of the wafer fabrication process, the number of wiring layers has reached more than ten. For the dual-damascene structure, the interlayer dielectrics with low-$k$ materials is introduced to isolate wiring layers and decrease the signal delay. However, some problems are generated in the integration process of Cu and the low-$k$ dielectric^{[1, 2, 3, 4]}. First of all, the diffusion of Cu ions to the dielectric layers is a vital problem. To solve such a problem, a kind of material such as tantalum (Ta)/tantalum nitride (TaN) deposited as the diffusion barrier layer was chosen to prevent the diffusion of Cu ions in the copper lines to the dielectric layers.

The barrier layer polishing is a critical step during the Cu CMP process. In this step, different materials were polished together, such as Cu, Ta and oxide (TEOS). Mostly, before the barrier polishing, the wafer surface suffered from a rugged surface, where the recessed region is called dishing (as shown in Figure 1). The goal of barrier polishing is to correct the dishing and to achieve global planarization.

Extensive work had been done in studying barrier layer polishing. In past research, we proposed a barrier polishing slurry containing a FA/O I chelating agent. It was free of any corrosion inhibitors and hydrogen peroxide. The slurry had an acceptable dishing correction and Cu/Ta /TEOS polishing speed selectivity. However, the copper polishing rate seemed to be not high enough. As a result, post the barrier CMP, the roughness of the wafer was unacceptable. The excessive surface micro roughness (RMS) of the wafer could cause the resistance, the noise and the current leakage to be increased with the IC chip, generating the extreme deterioration with the EM (electro migration)^[6]. Meanwhile, due to such changes, it initiated the micro structure of the interconnect lines to be destroyed, leading the circuits to be short or open with both of the device reliability and the yield impacted. Therefore, it has great significance to enhance the IC performance and the reliability by the means of decreasing the RMS, the damaged layer, and the surface defects with the perfect surface of the wafer obtained^[7]. What is more, the pH of the slurry containing the FA/O I chelating agent was about more than 10, so it is very easy to break the dielectric layer while barrier polishing. In this paper, we studied a new weakly chelating agent which had a lower pH and stronger chelating ability than the FA/O I chelating agents researched before. The performance of the slurry containing the new weakly chelating agent was evaluated in terms of the Cu/Ta/TEOS polishing speed selectivity, dishing correction, surface roughness and capacitance of the wiring line. The chelating reaction was discussed in this paper also.

2. Experiment

The experiment was performed on the E460E polisher produced by the France Alpsitec company. The pad was a Rohm Hass IC1000 hard pad. To investigate the materials removal rate (MRR) selectivity of the Cu/Ta/TEOS, it was adopted by the materials with 3 inch copper blanket wafers, Ta wafers and TEOS wafers during the CMP. The polishing recipe is shown in Table 1. It used the Mettle ToledoAB204-N analytical balance ($\pm$0.1 mg accuracy) to weight the above wafers before CMP and post. Then the MRR of Cu/Ta/TEOS calculated with Equation (1) below:

${\rm RR}=\frac{\Delta m}{\rho \pi R^2t}.$

(1)

With the above formula, the RR was meant of the MRR. The $\Delta m$ was the quality difference of the wafers between before and post CMP. The $\rho$ was the density of the material. The $R$ was the radius, and $t$ was the polishing time.

12 inch MIT854 pattern wafers were used to investigate the modification ability of the barrier slurry and the roughness of the wafers. A cross-section view of the wafers is shown in Figure 2. It was adopted by the xp-300 film step profiler (US Ambios company) to measure the removal thickness of the copper lines with the modification ability inspected. It was also used by the AFM (US Agilent 5600LS) to detect the surface micro roughness of the copper lines with the results contrasted. The polishing recipe is shown in Table 2.

The experiment used the slurry that mainly contained silica abrasives (average size 20-30 nm), the chelating agent (product self-developed with the Institute of Microelectronics, Hebei university) and the surfactant included. The slurry that contained the FA/O I chelating agent, researched before, was noted as slurry A. The novel slurry that contained the new weakly alkali chelating agent was noted as slurry B. As the agent of the H$_{2}$O$_{2}$ was decomposed easily, to ensure the stability of the barrier slurry, it was not added with the H$_{2}$O$_{2}$ during the preparation process of the slurry.

All the experiments were performed in a super clean laboratory. The room temperature was set at 21-23 C. All the wafers post CMP were rinsed with DIW and dried by pure N$_{2}$ before measuring in the experiments.

3. Result discussion

Figure 3 shows a comparison between the new weakly alkaline chelating agent (Slurry B) and FA/O I chelating agent (Slurry A). The experiment was performed under the same condition, and the slurries used in experiment contained the same amount of the two chelating agents. It was indicated that it obtained a copper removal rate of 265.64 /min with Slurry B and the roughness of the wafer was 1.01 nm (10 $\times$ 10 $\mu$m$^2$), shown in Figure 4(b). Meanwhile, Slurry A had a copper removal rate of 122.4 /min and the roughness of the wafer was 1.51~nm (10 $\times$ 10 $\mu$m$^2$), shown in Figure 4(a). The removal rates of Ta and TEOS with Slurry A and Slurry B showed no obvious difference.

The difference between the two polishing rates obtained by the slurries with two kinds of chelating agent implied that the new weakly alkaline chelating agent had a stronger chelating ability than the FA/O I chelating agent. The FA/O I chelating agent had a 13-chelating-ring structure, and it belonged to an amine alkali^[8]. It contained a hydroxyl radical, which would generate the ionized OH$^{-}$ as it dissolved into the aqueous solution. However, the new weakly alkaline chelating agent was a kind of amine salt without a hydroxyl radical. So the new chelating agent in an aqueous solution had a lower pH value of 8.89.

In the surface chemical reaction process of CMP, Cu was oxidized into CuO and Cu$_{2}$O by the small amount of oxygen dissolved from air. CuO would react with H$_{2}$O generating Cu(OH)$_{2}$, which had a weak ionization equilibrium, shown in Equation (5). Equations (4) and (5) were reversible.

The pH value of Slurry A was much higher than Slurry B. That is to say, there were more OH$^{-}$ released in Slurry A than in Slurry B. According to the common ion effect, when there were more OH$^{-}$, Equation (5) reacted toward the left and thus the degree of ionization of Cu(OH)$_{2}$ in Slurry A turned weaker than in Slurry B. That is to say, there were more Cu$^{2+}$ ions in Slurry B than Slurry A. Then the chelating agent would react with more Cu$^{2+}$ forming a kind of soluble product. In this way, Cu could be removed more quickly by this chemical accelerating process in Slurry B. The reaction processes were as follows.

$4{\rm Cu}+{\rm O}_{2} \to 2{\rm Cu}_{2}{\rm O},$

(2)

$2{\rm Cu}+{\rm O}_{2} \to 2{\rm CuO},$

(3)

${\rm Cu}_{2}{\rm O}+{\rm H}_{2}{\rm O} \leftrightarrow 2{\rm Cu(OH)} \leftrightarrow 2{\rm Cu}^{+}+2{\rm OH}^{-},$

(4)

${\rm CuO}+{\rm H}_{2}{\rm O} \leftrightarrow {\rm Cu(OH)}_{2} \leftrightarrow {\rm Cu}^{2+}+2{\rm OH}^{-},$

(5)

${\rm Cu}^{+}+{\rm RNH}_{2} \to [{\rm Cu(RNH}_{2})_{2}]^{+},$

(6)

${\rm Cu}^{2+}+{\rm RNH}_{2} \to [{\rm Cu(RNH}_{2})_{4}]^{2+}.$

(7)

Figure 4 shows in the same experiment condition, the roughness of the wafer with Slurry B was lower than Slurry A. Due to the stronger chelating ability of Slurry B, the mechanical action and chemical action matched better. So the grinding effect with the colloidal abrasives did not cause the excessive RMS of the wafer.

Due to the small amount of oxygen dissolved in the slurry, tantalum was oxidized to Ta$_{2}$O$_{5}$. In an alkaline environment, the oxide then turned into Ta$^{3+}$, this anion reacted with the chelating agent to form a soluble complex to remove tantalum barrier materials. Benefiting from the advanced fabrication process, the barrier layer was only 8 nm. A high tantalum removal rate was not necessary during barrier CMP. It was required that the tantalum had a 30-40 nm/min polishing speed in processing. The data obtained from the experiment showed that the tantalum had a minimum polishing speed at 30-35 nm/min.

Figure 5 shows the trend of the polishing speed when the concentration of the new weakly alkaline chelating agent changed. We can see that the Cu polishing speed was the highest at the new chelating agent concentration of 2.5 mL/L, which was 31.082 nm/min. Initially, without the new chelating agent changing in the polishing slurry, the Cu can be removed mainly by the grinding effect, so the Cu polishing rates measured were found to be negligible (9.792 nm/min). However, as the new chelating agent concentration is increased from 0 to 2.5~mL/L, the Cu polishing rate increases significantly from 9.792 to 31.082 nm/min due to the stronger chelating ability of the new weakly chelating agent. As the removal rate of copper is the result of the chelating agent and oxidant cooperative effect, the oxygen dissolved in the slurry is limited, the CuO and Cu$_{2}$O generated by oxidation is limited, so the increase of the chelating agent is invalid for increasing the Cu polishing rate^[9]. When the new weakly chelating agent concentration is greater than 2.5 mL/L, the effects of the new chelating agent on the polishing rate are reduced, so the removal rate of copper changes to flatten out.

When the new weakly chelating agent concentration is 2.5~mL/L, the Cu polishing speed was the highest, and the polishing rate selectivity (ratio of Cu polishing rate and the sum of Ta and TEOS polishing rate) was about 1 : 3, meeting the process demand of the barrier CMP.

Thus we applied the slurry containing 2.5 mL/L new weakly alkaline chelating agent, 500 mL/L silica abrasives (average size 20-30 nm) and 15 mL/L of the surfactant in a 12~inch pattern wafer polishing to test the dishing correcting performance.

Figure 6 shows the dishing data after polishing for 30/40/50/60/70 s, the dishing value declined to 41 nm from 117~nm, correcting 76 nm. It indicated that the slurry containing the new chelating agent had an effective correction of dishing.

In the roughness comparison experiment, we found that the surface of the wafer suffered from a high roughness of 5.78 nm before barrier CMP, shown in Figure 7, while after using the new weakly alkaline chelating agent, the surface had a lower roughness value of 0.693 nm, shown as Figure 8. It was~preliminarily~concluded that the low copper polishing rate would cause bad surface roughness. The detailed principle of the relevance between the polishing rate and the roughness will be discussed in further study.

The FA/O I chelating agent was a kind of organic amine alkali, it had a pH of 10.1. Different from the FA/O I chelating agent, the new weakly alkaline chelating was a kind of amine salt, which had a lower pH value (8.89) than FA/O I. The new chelating agent also played the role of a pH buffer agent. The barrier slurry containing the new chelating agent had a pH value of 9.36, which was hardly changed over time. The low pH of the slurry makes it possible to protect the low-$k$ dielectric under the barrier layer from structurally breaking. Figure 9 shows the capacitance of the interconnect wiring as a function of the polishing time. In the first 40 s, the capacitance raised to 2.4 pF, then after 70 s of polishing, the capacitance almost remained unchanged. The dielectric was loose and porous, so the $k$ value of the dielectric was easily increased when the dielectric was exposed to the wet environment^[11]. The increase of the capacitance was caused by the slurry infiltration to low-$k$ dielectrics.

4. Conclusion

In this paper, we proposed a new kind of chelating agent. This chelating agent was a kind of amine salt. It had a low pH value of 8.89 and a stronger chelating ability. The polishing rate of copper using the new chelating agent was higher than the FA/O I chelating agent. The results from 12 inch MIT 854 wafer polishing revealed that the slurry containing the new chelating agent had an effective dishing correction. By comparing the surface roughness after using the slurry containing the FA/O I chelating agent and the new chelating agent, we found that the new chelating agent had the advantage in optimizing surface roughness. We also found that the roughness was in relation with the materials polishing rate. The capacitance test results suggested that the slurry containing the new chelating agent hardly had a negative effect on the $k$ value profiting from the low pH value of the slurry.