Synergic effect of chelating agent and oxidant on chemical mechanical planarization

    Corresponding author: Weijuan Liu, lwjcpf@163.com
  • Institute of Microelectronics, Hebei University of Technology, Tianjin 300130, China

Key words: CMPtheoretical modelsynergic ratiosdishingstatic etch rates

Abstract: Chemically dominant alkaline slurry, which is free of BTA (benzotriazole) and other inhibitors, was investigated. The synergic effect of the chelating agent and oxidant on the chemical mechanical planarization (CMP) was taken into consideration. Copper CMP slurry is mainly composed of an oxidizer, nonionic surfactant, chelating agent and abrasive particles. The effect of different synergic ratios of oxidant with chelating agent on the polishing removal rate, static etch rate and planarization were detected. The planarization results reveal that with the increase of oxidant concentration, the dishing value firstly diminished and then increased again. When the synergic ratios is 3, the dishing increases the least. A theoretical model combined with chemical-mechanical kinetics process was proposed in the investigation, which can explain this phenomenon. Based on the theoretical model, the effect of synergic ratios of oxidant with chelating agent on velocity D-value (convex removal rate minus recessed removal rate) was analyzed. The results illustrate that when the synergic ratio is between 2.5—3.5, the velocity D-value is relatively higher, thereby good planarization can be achieved in this interval. This investigation provides a new guide to analyze and study copper line corrosion in the recessed region during copper clearing polishing.

    HTML

1.   Introduction
  • At the moment, international and domestic giga scale integration circuits (GSIC) have begun to develop 7 and 28 nm technology nodes respectively. Improving integration and increasing interconnect layers have become inevitable trends in the development of microelectronics[1]. However, the development of a three-dimensional interconnect structure for planarization presents a great challenge to the CMP process. With the development of microelectronic materials and technologies, currently the international community has proposed three requirements, that there is low pH, low abrasive concentration and low down-pressure. Since conventional copper slurries contain BTA, and Cu-BTA monomolecular film is mechanically stronger and more corrosion resistant[2, 3], CMP is mainly mechanical dominant. While porous low-$k$ dielectric materials is incorporated in the interconnect structures, achieving planarization at low down-pressure is inevitable[1, 2]. So it is necessary to develop a chemically dominant alkaline slurry substituting mechanically dominant acidic slurry[3, 4]. However, in chemically dominant CMP, corrosion of copper line highlights in the recessed region during the copper clearing polishing. Based on the kinetic control process, the impact of synergic ratios of oxidant with chelating agent on the planarization was analyzed and verified[5, 6, 7, 8, 9, 10, 11]. In this investigation, effectively removing residual copper at low down-pressure was achieved. Optimizing the synergic ratios of the oxidant with the chelating agent provides a new guide to solve the copper line corrosion in the recessed region.

2.   Experiment
  • The experiments were performed on the Alpsitec E460E platform, which is equipped with a diamond conditioner. The polishing pads used in the experiment were Rohm Haas IC1000. An XP-300 surface profiler was used to measure the thickness of the copper film, which was provided by U.S. KLA Tencor/AMBIOS Technology. The static etch rate was performed using a CL-4A type thermostatic heating magnetic stirrer. The novel alkaline copper slurry used in the experiment consists of an abrasive particle (colloidal silica), a nonionic surfactant, a chelating agent and an oxidizer. The abrasive particle size is 60-70 nm and the mass fraction is 40 wt %. The FA/OVI type chelating agent was developed by the Institute of Microelectronics, Hebei University of Technology. The FA/OI type nonionic surfactant is mainly used for reducing the surface tension, which is conducive to being preferentially adsorbed on the reactant surface. The mass fraction of hydrogen peroxide (oxidant) is 30 wt %. The effect of different synergic ratios of the oxidant with the chelating agent on the polishing removal rate, static etch rate and planarization were investigated. The process conditions used in the experiment are summarized in Table 1.

  • 2.1.   Polishing removal rate

  • The removal rate was calculated by the weight loss of 3~inch copper (99.99 % purity), which was measured by analytical balance (accuracy $\pm$ 0.1 mg, Mettle Toledo AB204-N). The formula can be expressed as[19]: CuRR $=$ $\Delta m/\left( \rho \pi R^2t \right)$. In the formula, CuRR is the removal rate, $\Delta m$ is mass loss, $R$ is radius ($R$ $=$ 3.81 cm), $\rho $ is the density, and $t$ is the polishing time.

  • 2.2.   Static etch rate

  • In the experiment, the 1 $\times$ 5 cm$^{2}$ copper wafer (initial thickness $\sim $1.2 $\mu$m) samples were used to measure the static etch rate of the slurries. Static dissolution experiments were carried out in a 1000 ml glass beaker with a 1000 ml etchant solution. A CL-4A type thermostatic heating magnetic stirrer was used. The static etch rate (SER) was calculated by the thickness decrease of the copper wafer, which was measured by step height analysis. The formula can be expressed as: SER $=$ (SH$_{\rm pre}$ $-$ SH$_{\rm post})$ / $t$, where SH$_{\rm pre}$ is the thickness of the wafer before static etching and SH$_{\rm post}$ is the remaining thickness of the wafer after static etching.

  • 2.3.   Planarization test

  • Achieving planarization is the major purpose of the CMP. In this paper, the slurries' planarization tests were monitored on the TM2 pattern wafer. In order to characterize the passivation ability of the different synergic ratios on the recessed region during copper clearing polishing, we defined the over-polishing (op) time for 60 s and counted the increased value of the dishing. All of the wafers were dealt with in the same process condition before over-polishing. The dishing value can be measured by the XP-300 surface profiler.

3.   Results and discussion

    3.1.   Synergic effects of oxidant and chelating agent on removal rate and static etch rate

  • In this paper, the polishing removal rate and the static etch rate were investigated with synergic ratios (the volume ratios of oxidant and chelating agent) fixed at 0.5, 1.0, 1.5, 2, 2.5 respectively. The polishing removal rate and the static etch rate were used to simulate the velocity of the convex region and the recessed region of the pattern wafer. Velocity D-value is the removal rate of the convex region minus the removal rate of the recessed region. In Figure 1, the chelating agent concentration of slurry A, slurry B, slurry C were fixed at 0.4 vol %, 0.3~vol %, 0.2 vol % respectively, realizing horizontal comparison. The corresponding polishing rate, static etch rate and velocity D-value are shown in Figure 1. Referring to Figure 1(a), the polishing removal rate firstly increases and then shows a slight decrease approaching a constant, while the static corrosion rate curve reaches a minimum point and then gradually increases again. Figure 1(b) illustrates the max extremum velocity D-value and vividly describes that with the concentration of the chelating agent increase, the max extremum velocity D-value gradually shifted to the right. It is clear that with the increase of the chelating content, more oxidant was needed to achieve stronger passivation layers in the recessed region.

  • 3.2.   Effect of different synergic ratios on planarization

  • The effect of synergic ratios on copper passivation in the recessed region was investigated. As can be seen from Figure 2, before the over-polishing test, the residual copper on the pattern wafer was removed completely. All of the sample wafers were dealt with in the same process condition, so that we can compare the experiment results exactly. Referring to Figure 3, the relationship between the passivation in the recessed region and the synergic ratios was achieved. The concentration of oxidant varies from 0.6 vol % to 1.5 vol % with the concentration of chelating agent fixed at 0.3 vol % (Figures 2(a) and 2(b)). In Figures 2(c) and 2(d), the concentration of the chelating agent was fixed at 0.4 vol % and the concentration of the oxidant varies from 0.8 vol % to 1.6 vol %. In Figure 3, pre op refers to the dishing value before over-polishing (op), and the planarization results under different synergic ratios are given in the figures. The TM2 pattern wafer results reveal that with the increase of the synergic ratios, the dishing decreases towards the minimum and then slightly increases again. Simultaneously, we can conclude that when the synergic ratio is 3, namely the oxidant is three times higher than the cheating agent, the dishing increases the least.

    Referring to Figure 2(b), the electron microscope diagram of the TM2 pattern wafer after 60 s over-polishing (op) is given. Compared with Figure 2(a), to some extent, the color of the pattern wafer after op is darker than before op. However, the low-$k$ was not observed, that is to say the barrier layer was slightly worn by copper clearing cleaning for the long time of over-polishing. Combined with the specific dishing values after op shown in Figure 3, all the results powerfully prove that the planarization efficiency is good enough to achieve a high efficiency of removal of residual copper, avoiding a substantial increase of the dishing value.

  • 3.3.   Theoretical model and results analysis

  • Low down-pressure is one of the urgent requirements in CMP, so that chemically dominant alkaline slurry is the new trend in planarization[12]. However, there exists isotropic corrosion in the chemically dominant CMP. The velocity D-value of the convex region and the recessed region decreases. A long duration of over-polishing will lead to an increased loss of copper lines in the recessed region. Meanwhile, as chemical dissolution is the predominant factor in the alkaline slurry[13, 14], the threshold reaction barrier can be achieved under low pressure. The copper in the recessed region can be easily dissolved, which contributes to the increase of dishing. Thus, a theoretical model was proposed to find out the synergic ratios to gain the maximum D-value and achieve a dense passivation layer.

    The chemical mechanical planarization includes a couple effects of chemical dissolution and mechanical wear dissolution[15]. The mechanical action mainly includes the relative rotation, pressure and abrasive. In copper CMP, abrasives play the role of removing the passivation layers and activating the chemical reaction between the copper and the slurry[7]. The chemical dissolution is mainly controlled by the oxidant and the chelating agent. The oxidant reacts with the copper surface to form Cu$^{+}$ or Cu$^{2+}$, and the chelating agent reacts with Cu$^{+}$ or Cu$^{2+}$ species to form a soluble species[3]. The reaction equations are as follows[18]:

    The removal rate is faster in the convex region than in the recessed region, which can be explained from chemical and mechanical aspects respectively. From the view of chemical action, although the chemical action is isotropic, the reaction in the convex region is faster. This is because CMP is a chemical-mechanical synergic effect, so the stronger mechanical action in the convex region can break the reaction equilibrium more easily and promote the chemical dissolution. From the point of mechanical action, we can explain it from the law of mass action. The law of mass action can be expressed as:
    ${\rm RR}=K C_{\rm A} C_{\rm B}, \quad K=-\Delta E/(RT), $(1)
    where $C_{\rm A}$, $C_{\rm B}$ represents the concentration of the solution, $\Delta E$ is the activation energy, $R$ is a constant, and $T$ is the temperature. With regard to the same slurry, $C_{\rm A}$, $C_{\rm B}$ are constant parameters, which stand for the concentration of the chelating agent and the oxidant respectively. The removal rate can be linked to the coefficient $K$. As can be seen from Equation (1), the removal rate is proportional to the temperature. The pressure in the convex region is greater than in the recessed region, thus more friction is converted into internal energy, causing the temperature rise in the convex region. We can conclude that the removal rate in the convex region is faster than in the recessed region. Chemical-mechanical equilibrium was firstly achieved in the recessed region; thus the theoretical model (shown in Figure 4) was proposed. The planarization results can be explained by referring to Figure 4.

    Referring to Figure 4, when the synergic ratio of the oxidant and chelating agent is less than 2.5, the concentration of Cu$^{+}$, Cu$^{2 +}$ concentration is less than the concentration of R(NH$_2$)$_2$ provided by the chelating agent. The removal rate is mainly controlled by the oxidant. The convex and recessed region in region I are the oxidative reaction controlled process. The film of copper oxides and hydroxides is relatively porous, which is easily mechanically broken. The chelating agent can easily form a complex with the Cu$^{+}$, which is predominantly on the copper surface at a lower oxidant concentration[15]. Although the mechanical is weak in the recessed region, the passivation layers can also be worn by the abrasive for the fragile passivation layer.

    With the increase of oxidant, the synergic ratios reach the interval between 2.5 and 3.5 (Figure 4, region II). In region II, Cu$^{+}$ converts to Cu$^{2 +}$, which is more stable under CMP conditions[15]. That is to say, the film formed by copper ions becomes thicker, harder and more anticorrosive[2, 3]. CMP is the chemical-mechanical synergic behavior, where chemical actions convert strongly bonded surface molecules to weakly bonded molecular species and the mechanical actions deliver the energy from the applied pressure and relative rotation, breaking the molecular bonds and debonding the surface molecules[6, 16]. Due to the mechanical action being weaker in the recessed region, the mass transfer is relatively slower than in the convex region[17]. The reaction barrier is not easily broken with the increase of film thickness in the recessed region. When the oxide generation rate is equal to the chelating agent etch rate of Cu$^{+}$ and Cu$^{2+}$, an equilibrium is reached firstly in the recessed region and the maximal removal rate can be achieved[3, 7]. At this time, the removal rate of the convex region increases more than the velocity of the recessed region. The D-value increases and the passivation in the recessed region is good. Planarization is more easily achieved in this interval.

    When the oxidant concentration continues to increase, the synergic ratio is higher than 3.5 (Figure 4, region III). The removal rate increases as the oxidant increases and then approaches a constant with a further increase of oxidant concentration[3] (Figure 1(a)). The copper ion concentration provided by the oxidant is greater than the chelating molecules provided by the chelating agent. The chelating agent etch rate of [Cu$^{+, 2+}$] ions is the controlled process in region III. The film becomes thicker for excess oxidant, which prolongs the polishing time to remove the copper film, because [Cu$^{+, 2+}$] are easier to remove than fresh copper. The existence of the [Cu$^{+, 2+}$] will affect the endpoint detection. The endpoint of copper clearing will last until there are no copper ions detected in the slurry. So the excess film will be worn by the abrasive, until fresh copper is exposed and no [Cu$^{+, 2+}$] is detected. Long-time copper clearing polishing severely increases the dishing form for removing the thicker film. Thus, the loss of recessed copper lines gradually increases again.

    In summary, the dishing of the recessed region diminishes with the increase of oxidant, forming thicker anticorrosion film. Then, more oxidant formed much thicker film. The thicker film in the recessed region must be removed to reach the copper clearing endpoint. Hence, the loss of copper line in the recessed region rises again. Namely, the dishing gradually increases again. This conclusion can guide our work in the dishing control during copper clearing polishing.

4.   Conclusion
  • In this paper, the effect of synergic ratios on the polishing removal rate, static etch rate and the planarization have been investigated. Combined with the chemical-mechanical kinetics theory, a theoretical model was proposed and the effect of synergic ratios on planarization was analyzed. When the oxidant is three times the chelating agent, the recessed region can be well passivated with the combined chemical and mechanical interaction[12]. This research is very significant in improving the passivation of copper lines in the recessed region during copper clearing polishing theoretically and practically for alkaline slurry, which is chemically predominant and has isotropic characteristics.

Figure (4)  Table (2) Reference (19) Relative (20)

Journal of Semiconductors © 2017 All Rights Reserved