1. Introduction

Chemical mechanical polishing (CMP) is an effective important material removal and surface planarization technology, which has been widely adopted to achieve global planarization of the copper interconnections during the dual-Damascus process^[1-4]. As is known to all, a complete CMP process mainly includes three steps. A higher removal rate of Cu is required for the first step to efficiently eliminate the step height of the Cu films. Then a lower removal rate of Cu is necessary to precisely control the end-point over the barrier metals during the second step. In addition, in order to ensure that the Cu film on the barrier layer was removed completely, a period of time is also needed for the over-polishing step. The last step is the barrier layer and dielectric layer polishing^[5].

However, there are two major issues. The first is the dishing of Cu lines and erosion of dielectric films during the over-polishing step. The dishing value is generally around 500 Å after the over-polishing. To reduce the dishing value, an appropriate removal rate selectivity of TaN:Cu and TEOS:Cu should be obtained, respectively^[6]. If the polishing rate of the barrier and TEOS is faster than the polishing rate of Cu, a Cu line will protrude. On the contrary, if the polishing rate of Cu is too fast, the dishing value will become bigger after the barrier CMP. So only by the choice of an appropriate removal rate selectivity of different materials, will the dishing value decrease in the barrier CMP process and a good planarization can be achieved. The other is the high surface roughness. In order to control the surface roughness, the concentration of abrasive and polishing pressure must be reduced at the same time.

How to obtain an appropriate selectivity and reduce the surface roughness value in barrier polishing has become a focus of experts at home and abroad. Chiu^[7] studied the effects of slurry pH and alkaline ions on the polishing behaviors of Cu, Ta and TEOS. He found that the ideal removal selectivity of Ta/Cu and TEOS/Cu can be achieved by adjusting the pH value of the slurry. Li^[8] studied the effect of guanidine hydrochloride on removal rate selectivity and wafer topography modification in barrier CMP. He also received a good removal rate selectivity of Ta/Cu. In addition, a number of scientists studied the relevant reaction mechanism by the electrochemical analysis method, such as Chen, Janjama, and Rock^[9-13]. In this paper, the influence of three kinds of guanidine salt on the removal rate selectivity of different materials was studied. It effectively improved the removal rate selectivity and reduced the dishing and surface roughness value under the premise of low abrasive concentration and low polishing pressure when adding a small amount of guanidine hydrochloride. Guanidine hydrochloride has a better performance than the other two kinds of guanidine salt.

2. Experiment

2.1 Polishing experiments

Polishing experiments were performed on the E460 polishing system and the polishing pad was the IC1000, provided by Rohm Hass. The polishing pad was conditioned by using a grid abrade stainless steel conditioner with diamond pieces embedded into the surface. The correction time is 1 min before each polishing experiment. For the removal rate measurements, a 3 inch Cu piece, a 3 inch Ta piece and a 3 inch TEOS piece were used (the purity of the three kinds of materials was 99.99%). A 12 inch M1 pattern wafer was used to test the planarization ability of the different slurries. Table 1 shows the whole CMP process parameters, and Table 2 shows us the composition ratio of the three kinds of slurries.

2.2 pH and particle size test

The pH values of the three kinds of slurries were tested by the PHB-4 pH meter produced by INESA (precision is 0.01). Before each measurement, the pH meter was calibrated again with the special calibration fluid. The particle sizes of the slurries were tested by NiComp380 DLS. Before the test, the three kinds of slurries were continuously stirred for 30 s, then they were diluted 10 times with deionized water. In this paper, the mean diameter of the slurries was achieved by an intensity analysis technique, which is more accurate for measuring the small particle size.

2.3 Removal rate dishing and roughness test

The removal rate was calculated by Eq. (1) as follows:

$\begin{equation} {\rm RR}=\frac{\Delta m}{\rho \pi R^2T}, \end{equation}$

(1)

where RR is the removal rate, $\Delta m$ is the mass lost before and after polishing, R is the radius of the 3 inch piece (= 3.81 cm), $\rho$ is the density of materials ($\rho_{\rm Cu}$ = 8.9 g/cm$^{3}$, $\rho _{\rm TEOS}$ = 2.3 g/cm$^{3}$, $\rho _{\rm Ta}$ = 16.6 g/cm$^{3})$, and T is the polishing time. The mass losses of materials before and after polishing were measured by analytical balance (Mettle Toledo AB204-N). The cross-section view of the dishing pad (50 $\times$ 70 $\mu \text{m}^{2})$ and the erosion pad (50 $\times$ 70 $\mu \text{m}^{2})$ of the pattern wafer can be seen in Fig. 1. After the barrier polishing, the Xp-300 profiler was used to measure the dishing value and erosion value. The surface roughness of the pattern wafer was tested by an atomic force microscope (5600LS, produced by Agilent). The scanning area was (10 $\times$ 10 $\mu \text{m}^{2})$.

2.4 Electrochemical test

All the electrochemical experiments were done by the CHI660C electrochemical workstation (produced by Shanghai Hua Chen Instruments). The corrosion current of Cu and Ta were analyzed in the three kinds of slurries, respectively. In order to complete the experiment, we made a Cu electrode (1 cm² and a Ta electrode (1 cm². Before each experiment, the Cu electrode and Ta electrode were ground for 1 min on sandpaper. Then the surface of the Cu electrode and Ta electrode were cleaned by anhydrous ethanol to remove the organic matter. In order to prevent an abrasive adhesive on the surface of the electrode, the electrochemical testing was performed without adding the SiO₂ abrasive.

3. Results and discussion

3.1 pH, particle size and removal rate

The pH value and particle size value of the three kinds of slurries are shown in Fig. 2. It is obvious that the pH of the slurry with guanidine nitrate is lower than other slurries, but there is not a big difference regarding the particle size. As shown in Fig. 3, we can find that the addition of guanidine hydrochloride is effective in improving the removal rate of TEOS and Ta, and the Cu removal rate will decrease slightly compared with the two other kinds of slurries at the same time. Therefore, the addition of guanidine hydrochloride can improve the removal rate selectivity (TEOS:Cu and Ta:Cu) more effectively than the two other kinds of guanidine salt. The removal rates of Cu, Ta and TEOS are 342, 365 and 872 Å/min, respectively. Due to the fact that the Ta removal rate and Cu removal rate is close to 1 : 1, the dishing value will not increase during Ta polishing. In addition, the TEOS removal rate and Cu removal rate is close to 2 : 1, the dishing value will be decreased during the TEOS polishing. Therefore, a small amount of guanidine salt can effectively modify the dishing value after the elaborate polishing (called p2) process.

3.2 Dishing and erosion

As we can see in Figs. 4 and 5, guanidine hydrochloride is more effective in controlling the dishing and erosion values during the barrier CMP process than guanidine carbonate and guanidine nitrate. The dishing value and erosion value are 355 Å and 27 Å after the barrier CMP process, respectively; thus it meets the requirements of the 45 nm technology node. The decrease of the dishing value and erosion value is due to the fact that the addition of guanidine hydrochloride is effective in improving the removal rate of Ta and TEOS, and the Cu removal rate is reduced at the same time. Therefore, the dishing value and erosion value have been effectively improved and controlled in the polishing process of TEOS. The other important reason is that the working pressure is very low, just 1 psi during the pattern wafer polishing. In addition, a low concentration of abrasive is conducive to reducing the dishing and erosion values. So, a low working pressure and a low abrasive concentration are the future development trends for barrier polishing.

3.3 Surface roughness

The root-mean square (RMS) value was used to evaluate the surface quality. Figure 6 shows the RMS values and surface topography photos of the pattern wafer after the barrier CMP. We can find that the surface topography is better after the barrier polishing with the slurry containing guanidine hydrochloride and the RMS value of the pattern wafer is only 1.12 nm. The surface roughness values of the pattern wafer polished by the slurry with guanidine carbonate and guanidine nitrate are 1.85 nm and 2.03 nm, respectively. So the addition of guanidine hydrochloride has a better performance on the roughness than the other two kinds of guanidine salt. This is because guanidine hydrochloride has lower corrosiveness on the Cu and Ta materials. In addition, the low erosion value is good for reducing the surface roughness.

3.4 Electrochemical analysis

From the Tafel plot, we can give an overview of the surface reaction of Cu and Ta. The corrosion current of Cu and Ta are presented in Figs. 7 and 8, respectively. We can conclude that for the slurry with guanidine hydrochloride, the corrosion currents of Cu and Ta are relatively large compared to the slurry containing guanidine carbonate and guanidine hydrochloride. So, it is clear that a certain thickness of the passivation layer formed on the surface of Cu and Ta, respectively. Because the passivation layer of Cu and Ta is thicker when the slurry containing guanidine hydrochloride is used, so the corrosion current is smaller than the other two slurries, and it is easy to get a better removal rate selectivity and planarization efficiency. The hole electrochemical reaction is very complex. The main chemical reactions occuring at the anode are as follows. Firstly, Cu and Ta are oxidized at the anode.

$\begin{equation} 2{\rm Cu}+{\rm H}_2 {\rm O}\to {\rm Cu}_2 {\rm O}+2{\rm H}^++2{\rm e}^-, \end{equation}$

(2)

$\begin{equation} {\rm Cu}_2{\rm O}+{\rm H}_2{\rm O}\to 2{\rm CuO}+2{\rm H}^++2{\rm e}^-, \end{equation}$

(3)

$\begin{equation} {\rm Cu}_2{\rm O}+3{\rm H}_2{\rm O} \to 2{\rm Cu}\left( {\rm OH} \right)_2 +2{\rm H}^++2{\rm e}^-, \end{equation}$

(4)

$\begin{equation} 2{\rm Ta}+5{\rm H}_2{\rm O} \to {\rm Ta}_2{\rm O}_5 +10{\rm H}^++10{\rm e}^-. \end{equation}$

(5)

Then an oxygen reduction reaction occurs at the cathode.

$\begin{equation} {\rm O}_2 +2{\rm H}_2{\rm O}+4{\rm e}^-\to 4{\rm OH}^-, \end{equation}$

(6)

$\begin{equation} {\rm H}_2{\rm O}_2 +2{\rm e}^-\to 2{\rm OH}^-. \end{equation}$

(7)

Finally, the guanidine salt and chelating agent are effective in complexing Cu $^{+}$, Cu$^{2 +}$ and Ta₂O₅; however, the chemical reaction mechanism of metal ions and the guanidine salt is very complex, which is out of the scope of this paper, so it needs more investigative research.

4. Conclusion

In this paper, the influence of three kinds of guanidine salt on the performance of the alkaline barrier slurry has been compared on the basis of a low concentration of abrasive. The results reveal that the removal rate selectivity is effectively improved and the dishing value and surface roughness value are reduced by adding a small amount of guanidine salt. Also, it was found that the performance of guanidine hydrochloride is better than guanidine nitrate or guanidine carbonate. The biggest significance for this study is that the removal rate selectivity, planarization efficiency and surface roughness are improved under the condition of low abrasive concentration by adding a small amount of guanidine salt. The alkaline slurry can very well meet the requirements of the 45 nm technology node for barrier polishing.