1. Introduction

Ruthenium (Ru) metal has been investigated as a copper diffusion barrier for copper dual-damascene interconnects for many years^{[1, 2, 3, 4, 5, 6]}. The traditional barrier structure of tantalum (Ta)/tantalum nitride (TaN) bilayer will cause serious copper deposition problems such as pinched-off trench opening and inadequate gap-fill. In order to solve the above problems, Ru provides a thin seed layer from which Cu may be directly grown by electrochemical deposition, and the resulting metal lines tend to show reduced interconnect electrical resistance and resistance capacitance signal delay compared to lines grown from PVD Cu seeds^[7].

Hence, Ru is being considered for replacing Ta/TaN as barrier materials for Cu interconnects wiring structures in advanced devices. Ru liners have good barrier properties such as a high melting point (2310 ), low resistivity ($\sim$7 $\mu$$\Omega \cdot$cm), and negligible immiscibility with copper (Cu) even at 900 C^[8]. In addition, Ru is used as a bottom electrode in next-generation memory capacitors because it has an excellent electrical performance such as high breakdown voltage, low leakage current, and availability of MOCVD precursors.

However, Ru, being a noble and hard metal, is difficult to polish and planarize. The Ru RR was almost zero in the colloidal-silica based slurry without oxidizers because of the much lower mechanical hardness of SiO$_{2}$ (9.05 GPa) than that of Ru (14.54 GPa)^[14]. One way to enhance the removal rate is by modifying its chemical and mechanical properties using oxidizers and complexing agents^[9]. So far, several oxidizers^{[3, 5, 10, 11, 12, 13, 14, 15]} have been investigated. However, H$_{2}$O$_{2}$ was preferentially chosen as an oxidizer in Ru polishing slurry because of its powerful oxidizing capability, and metal-ion-free state.

In this paper, the effect of organic-alkali hydroxide multi-amine (FA/O) and H$_{2}$O$_{2}$ on Ru removal rate (RR) and static etching rate (SER) was firstly investigated. We also measured the potentiodynamic polarization curves of Ru electrodes. As a result, it was indicated that the added FA/O could improve Ru RR and decrease corrosion voltage. Then the Ru removal mechanism with H$_{2}$O$_{2}$ and FA/O was revealed. Finally, the effect of non-ionic surfactant AD on Ru roughness was studied, and experimental results showed that the added AD could achieve satisfactory Ru surface quality.

2.Experimental details

The three-inch coupons (99.99% purity) were polished using an E460E CMP polisher of Alpsitec Company and a Rohmand Haas IC 1000TM pad at 2 psi down pressure, the carrier/platen rotational speed 55/60 rpm, the slurry flow rate 150~mL/min and the polishing time 9 min. All experiments were performed at pH 9.0 where toxic RuO$_{4}$ is not formed^{[16, 17]}. Before the experiment, the ex-situ pad conditioning was conducted with a TBW diamond conditioner for 5~min. Ru RR was determined by measuring the weight loss of before and after polishing with an analytical balance (Mettler Toledo AB204-N, 0.001 mg resolution). Each experiment was repeated three times. The slurries were composed of de-ionized (DI) water, colloidal silica (mean particle size 50 nm), oxidizer H$_{2}$O$_{2}$, complexing agent FA/O and non-ionic surfactant AD. In order to investigate the slurries' static etching rate, coupons were carried out at room temperature by dipping into the tested slurries (without silica particles). Dip time was set as 30 min to improve the accuracy of the measured rates, coupons before and after dipping were measured using an analytical balance.

Potentiodynamic polarization (PDP) is a type of potentiodynamic electrochemical technique in which the potential of the working electrode versus the reference electrode is scanned over a range and the current is measured simultaneously. Potentiodynamic polarization tests use an LK2005B electrochemical work station. For these experiments, we used a three-electrode configuration with platinum as the counter electrode, a saturated calomel electrode (SCE) as the reference electrode, and Ru as the working electrode. The measurement area was 1~cm$^{2}$. The potential was varied between working and reference electrodes and the current flow between working and counter electrodes was measured. The voltage was scanned at a rate of 0.005 V/s in the range of $-1.0$ to $+2.0$ V. Corrosion current ($i_{\rm corr})$ was obtained from the Tafel plots.

Three-inch coupons of Ru were first pretreated with polishing for 9 min. Then the contact angles of a de-ionized water drop on these coupons were measured using a JC2000D goniometer assembly on a vibration-free optical table. The reported contact angle is an average of 5 contact angles measured at different locations on the wafer.

Averaged (root mean square, RMS) surface roughness was measured by Agilent 5600LS atomic force microscopy (AFM), typically over a 10 $\times$ 10 $\mu$m$^{2}$ area, using a non-contact mode.

3.Results and discussion

3.1.Effect of H$\boldsymbol{_{2}}$O$\boldsymbol{_{2}}$ and FA/O on Ru RR and SER

Figure 1 shows the Ru RR in the colloidal silica-based slurry as a function of the type of complexing agents at the same concentration of 0.05 mol/L at a pH 9. Many complexing agents including citric acid (C$_{6}$H$_{8}$O$_{7})$, glutamic acid (C$_{5}$H$_{9}$NO$_{4})$, sodium oxalate (Na$_{2}$C$_{2}$O$_{4})$, organic-alkali hydroxide multi-amine II (FA/O II), and glycine (C$_{2}$H$_{5}$NO$_{2})$ also demonstrated a low Ru RR. The Ru only had a high RR with organic-alkali hydroxide multi-amine (FA/O). The Ru had an RR of over 25 nm/min. These results indicated that only limited numbers of complexing agents were effective in polishing Ru.

We measured the potentiodynamic polarization curves of Ru electrodes with complexing agents, as shown in Figure 2(a). Ru electrodes displayed different corrosion behaviors with different complexing agents, and we calculated the corresponding corrosion potential ($E_{\rm corr})$ and the corrosion current density ($i_{\rm corr})$, as shown in Figure 2(b). In addition, $i_{\rm corr}$ had three peaks at C$_{6}$H$_{8}$O$_{7}$, Na$_{2}$C$_{2}$O$_{4}$ and FA/O where the Ru film had an observable RR. Hence, the corrosion behavior of the Ru surface seemed to have a certain influence on the RR. However, the magnitude of $E_{\rm corr}$ and $i_{\rm corr}$ does not strictly match the RR because FA/O had relatively high $i_{\rm corr}$ and considerably low $E_{\rm corr}$. Thus, FA/O had significantly higher $i_{\rm corr}$ than the other complexing agents, which would have a direct effect on the Ru RR.

The synergetic effects of H$_{2}$O$_{2}$ and FA/O on the Ru RR were shown in Figure 3(a). The slurries were composed of 20.0% colloidal silica, different concentration of H$_{2}$O$_{2}$ and FA/O and with pH 9.0. It could be seen that the Ru RR first linearly increased with the addition of 9 mL/L H$_{2}$O$_{2}$ probably due to the formation of a porous Ru oxides layer on the surface during polishing. Then the Ru RR slowly decreases with further increasing H$_{2}$O$_{2}$, which could be attributed to the fact that high concentration of H$_{2}$O$_{2}$ can accelerate the formation of uniform and tight passive film, especially the Ru oxides like RuO$_{2}$ on the surface. Meanwhile, the Ru RR increased with increasing FA/O concentration. Furthermore, the Ru RR was 24.46 nm/min of the peak value at 9 mL/L. Here FA/O was used as an effective Ru complexing agent, which could chelate Ru oxides (such as (RuO$_{4})$$^{2-}$ and RuO$_{4}$$^{-})$ and enhance Ru RR. The effect of H$_{2}$O$_{2}$ and FA/O on the Ru SER is shown in Figure 3(b). It could be seen that with the increasing concentration of H$_{2}$O$_{2}$ and FA/O, the Ru SER gradually increased, while the Ru SER almost maintains 0-2.2 nm/min. For all above, the results indicated that the synergetic work of H$_{2}$O$_{2}$ and FA/O can enhance the Ru RR.

Figure 4(a) shows the potentiodynamic voltage scan of Ru as a function of H$_{2}$O$_{2}$ concentration at pH 9. It revealed that the anodic current density first rapidly increased when the H$_{2}$O$_{2}$ concentration increased to 9 mL/L, and then decreased when the H$_{2}$O$_{2}$ concentration further increased to 12 mL/L, which was similar to the trend Ru RR and SER observed in Figure~3. The increase in the corrosion current density indicated more Ru was oxidized and thus the enhancement in the RR. However, the high concentration of H$_{2}$O$_{2}$ could accelerate the formation of uniform and tight Ru oxides, and then resulted in the decrease of Ru RR. The trend of the anodic current density revealed that the addition of low concentration of H$_{2}$O$_{2}$ could facilitate the formation of Ru oxides on the Ru surface. Figure 4(b) shows the potentiodynamic voltage scan of Ru as a function of FA/O concentration at pH 9. It revealed that the corrosion current density increased with increase in the concentration and saturated at 9 mL/L FA/O, the anodic current density revealed that the addition of FA/O could promote the increase of Ru dissolution. More importantly, Figure 4(c) shows the addition of FA/O could decrease the difference in the corrosion potentials and galvanic corrosion at the interface of Cu and Ru^[16]. Hence, this can be a big facilitator for the application of this slurry.

3.2 Ru removal mechanism with H₂O₂ and FA/O

Figure 5 shows schematic diagram of the Ru polishing process. H$_{2}$O$_{2}$ can oxidize the Ru surface to form RuO$_{2}$ as shown in Equation (1). In addition, the dissociation of H$_{2}$O$_{2}$ can increase hydroxyl ion concentration as shown in Equation (2).

$${\rm Ru}+2{\rm H}_2 {\rm O}_2 \to {\rm RuO}_2 +2{\rm H}_2 {\rm O},$$

(1)

$${\rm H}_2 {\rm O}_2 \to 2\left( {\ast {\rm OH}} \right).$$

(2)

Ru surfaces can interact with hydroxyl ions to form various Ru oxide species, such as Ru(OH)$_{3}$, Ru(OH)$_{4}$, RuO$_{2}$, RuO$_{4}$$^{-}$ and RuO$_{4}$$^{2-}$, the possible chemical reactions were shown in the following reactions (3)-(8):

${\rm Ru}+3{\rm OH}^-\to {\rm Ru}\left( {{\rm OH}} \right)_3 +3{\rm e}^-,$

(3)

${\rm Ru}+4{\rm OH}^-\to {\rm Ru}\left( {{\rm OH}} \right)_4 +4{\rm e}^-,$

(4)

${\rm Ru}+8\left( {{\rm OH}} \right)^-\to {\rm RuO}_4 ^-+4{\rm H}_2{\rm O}+7{\rm e}^-,$

(5)

${\rm RuO}_4 ^-+2\left( {{\rm OH}} \right)^-\to \left( {{\rm RuO}_4 } \right)^{2-}+{\rm H}_2 {\rm O}+\frac{1}{2}{\rm O}_2 +{\rm e}^-,$

(6)

${\rm Ru}\left( {{\rm OH}} \right)_4 \to {\rm RuO}_2 +2{\rm H}_2{\rm O},$

(7)

${\rm RuO}_2 +4\left( {{\rm OH}} \right)^-\to \left( {{\rm RuO}_4 } \right)^{2-}+{\rm H}_2 {\rm O}+2{\rm e}^-.$

(8)

These RuO$_{4}$$^{2-}$ and the FA/O complexing agents can be changed into soluble amine salts as shown in Equations (9) and (10). The molecular volume of amine salts was easier to separate from the surface under the friction of the abrasive and the polishing pad.

$${\rm NH}_2 -{\rm R}-{\rm NH}_2 +2{\rm H}_2 {\rm O}\to {\rm NH}_3 ^+-{\rm R}-{\rm NH}_3 ^++2{\rm OH}^-,$$

(9)

$$\begin{split} \label{eq10} 2{}& {\rm NH}_3 ^+-{\rm R}-{\rm NH}_3 ^++2\left( {{\rm RuO}_4 } \right)^{2-}\to \\& {\rm NH}_3 -{\rm R}-{\rm NH}_3 -{\rm RuO}_4 -{\rm NH}_3 -{\rm R}-{\rm NH}_3 -{\rm RuO}_4 . \end{split}$$

(10)

3.3 Effect of non-ionic surfactant on Ru roughness

The contact angles between the polished copper wafers and DI water are shown in Figure 6. Three-inch coupons of, Ru were first polished for 9 min using the desired solutions (The slurries were composed of 20.0% colloidal silica, 9 mL/L H$_{2}$O$_{2}$, 9 mL/L FA/O, different concentrations of non-ionic surfactant AD and with pH 9.0.). It revealed that the contact angle significantly decreases from 71.15$^\circ$ to 33.56$^\circ$ when the concentration of AD increased from 0 to 10 mL/L, and then slowly increased to 42.35$^\circ$ when the concentration of AD further increased to 25 mL/L, which indicated that the addition of AD could decrease the surface tension and improve the flow behavior of slurries, and thus improve the mass transmission of reactants and the resultants (because the hydrophilic group of non-ionic surfactant AD nearly accounts for 2/3-4/5, its nonpolar carbon-hydrogen group contacts solid, but the hydrophilic polyethylene chain stretches to water mostly to form a thicker protected layer which is a benefit to improve the removal rate of resultants^[18]})$ and ameliorate the nonuniform situation. Meantime, Ru surface morphology and surface quality data are shown in Figure 7. It revealed from Figures 7(a)-7(f) that, with the addition of non-ionic surfactant AD, the Ru surface became quite smooth compared with that without addition of AD. Also it could be seen that the addition of AD could lead to significant improvement of the surface roughness RMS.

4. Conclusions

In this paper, the mechanism of Ru CMP depends on H$_{2}$O$_{2}$ from the corrosion and oxidation processes. With the concentration of H$_{2}$O$_{2}$ increased to 9 mL/L, the Ru RR and SER decrease due to the formation of compact oxides layer on the Ru surface. In addition, FA/O is used as an effective Ru complexing agent, which can chelate Ru oxides (such as (RuO$_{4})$$^{2-}$ and RuO$_{4}$$^{-}$ change into soluble amine salts [R(NH$_{3})$$_{4}$] (RuO$_{4})$$_{2})$. Hence, the results indicate that the synergetic work of H$_{2}$O$_{2}$ and FA/O can enhance the Ru RR. Finally, the non-ionic surfactant AD is used to improve the Ru surface roughness in H$_{2}$O$_{2}$-based Ru barrier slurries. It is revealed that the Ru surface becomes quite smooth compared with that without addition of AD. Also it can be seen that the addition of AD can lead to significant improvement of the surface roughness.