Localized corrosion effects and modifications of acidic and
alkaline slurries on copper chemical mechanical polishing
Tzu-Hsuan Tsai, Shi-Chern Yen
*Department of Chemical Engineering, National Taiwan University, Taipei 106-17, Taiwan Received 14 May 2002; received in revised form 14 May 2002; accepted 1 November 2002
Abstract
This study demonstrates the CMP performance can be enhanced by modifying the corrosion effects of acidic and alkaline slurries on copper. A corrosion test-cell with a polishing platform is connected with the potentiostat to investigate the corrosion behaviors of copper CMP in various alumina slurries. Experiments show that the slurry needs to be maintained in acidic pH < 4:56 or alkaline pH > 9:05 surroundings and thus better dispersion of alumina particles and less residual contaminant on copper surface can be obtained. The surface defects after copper CMP using acidic and alkaline slurries are described by pitting corrosion mechanisms, and these mechanisms can be regarded as a basis to modify their corrosion effects. Experimental results indicate that it is necessary to modify the dissolution of HNO3 and oxidization of NH4OH for copper CMP slurries. Consequently, the slurries of 5 wt.% HNO3by adding 0.1 wt.% BTA or 5 wt.% KNO3by adding 1 wt.% NH4OH achieve good CMP performance for copper with higher CMP efficiency factor (CMPEF), 1460 and 486, and lower surface roughness (Rq), 4.019 and 3.971 nm, respectively. It is found that AFM micrographs can support the effectiveness of corrosion modifications for copper CMP in various slurry chemistries.
# 2003 Elsevier Science B.V. All rights reserved.
Keywords: Chemical mechanical polishing; Copper; Pitting corrosion; Alumina slurry; Potentiodynamic curve; Surface roughness
1. Introduction
Copper has been the material favored for integrated circuit manufacturers, owing to its excellent electro-migration resistance and low electrical resistance. One feasible means of integrating copper technology is by using dual damascene procedure in which chemical mechanical polishing (CMP) is employed as a viable method for planarizing the surface globally[1]. The CMP has been successfully applied in semiconductor production to create tungsten interconnects [2], and
previous literature has confirmed that the CMP mechanism of tungsten metal comprises the cyclic chemical formation of a metal passive film and its subsequent mechanical removal[3]. However, copper is softer and more sensitive to corrosion than tungsten, and the slurry composition and corrosion mechanisms of copper during CMP may be more complicated. Therefore, more thoroughly understanding the corro-sion effects on removal rates and surface finishing is critical to achieving copper metallization.
A commonly used CMP utilizes abrasive particles that are suspended in colloidal slurry. Different addi-tives or reagents can be combined with the slurry to provide high removal rates, good planarity and high selectivity [4–6], and to improve the dispersion of
Applied Surface Science 210 (2003) 190–205
*Corresponding author. Tel.:þ886-2-23630397;
fax:þ886-2-23630397.
E-mail address: scyen@ccms.ntu.edu.tw (S.-C. Yen).
0169-4332/03/$ – see front matter # 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0169-4332(02)01224-2
abrasive particles and rheological properties of the slurry [7,8]. As reported, copper CMP can be per-formed in either acidic or neutral or alkaline media with different function additives. For examples, an oxidizer is generally added to accelerate oxidization actions during copper CMP, while a film-forming agent inhibits the dissolution of surface being polish-ing [9]. Still dispersing agents or surfactants can be used to improve the uniformity and reduce the wafer defects and residual particles[10]. However, the vari-ety of polymer surfactants added into slurries results in more complicated problems of post-cleaning of CMP. In order to avoid coagulating of particles, choosing the slurry where the materials carry the same charge to repel each other is thinkable[11]. On the other hand, having corrosion behaviors between the copper sur-face and the slurry in hand is thus important, and modifying corrosion attack in various slurries must be studied by use of in situ electrochemical measure-ments[12]. Herein we try to illustrate the localized attack on copper surface by using the pitting mechan-isms. As mentioned above, the planarization of metal CMP is achieved by the cyclic chemical formation of a metal passive film and its subsequent mechanical removal. During this removal process, it can be regarded as a forced non-uniform breakdown of the surface layer or passive film. The breakdown phenom-enon results in the localized corrosion and initiates the pitting behaviors[13]. Thus, it is important to control the pitting actions of copper CMP by modifying the corrosion environment. If the pitting propagation does not be inhibited, then the pit is formed seriously. Hence the global planarization of CMP cannot be achieved. However, few studies have examined the corrosion mechanisms of copper CMP by pitting. Still little attention has been given to the corrosion
modification of the relation between mechanical and chemical actions.
This work investigated the localized corrosion characteristics of copper CMP in acidic and alkaline slurries. The defined indication of modifications has been proven to consist with the average roughness of copper surface and reasonably to explain the re-passivation ability. One is CMP efficiency factor (CMPEF), the ratio of removal rates (RR) to corrosion rates (CR); the other is the differentiation of RR with respect to the pressure. It is found the AFM micrographs agree well with the localized corrosion mechanisms constructed by using the electrochemical results and the defined indications.
2. Experimental
To investigate the corrosion effects of various che-mical compositions on copper CMP, the slurries were prepared by use of analytical grade reagents. The polishing slurry was prepared comprising only one abrasive, one corrosion enhancer or one film-forming agent, as listed in Table 1, where the abrasive was 50 nm a-alumina particle, and 5 wt.% for all measure-ments. The colloid slurry was prepared from a high sear stress condition in a mixer with high rotation speed to maintain a good dispersion of abrasive par-ticles and their pHs were adjusted by the addition of HNO3 or NH4OH. An acoustic spectrometer model
DT-1200 was used to determine the zeta potentials of materials.
The copper CMP was simulated in a corrosion test-cell with the specially designed base as the polishing platform.Fig. 1schematically displays this polishing setup. Electroplated copper films with a thickness of
Table 1
Specifications of slurry design for copper CMP Specification Slurry type
DI water 1–10 wt.% HNO3 1–10 wt.% HNO3þ 0.1 wt.% BTA 1–5 wt.% NH4OH 1–5 wt.% NH4OHþ 5 wt.% KNO3 pH range 6.98 1.0 1.0 12.0 11.5
Abrasive 5 wt.% a-Al2O3 a-Al2O3 a-Al2O3 a-Al2O3 a-Al2O3
Corrosion enhancer – HNO3 HNO3 NH4OH KNO3,NH4OH
Film-forming condition – – BTA Alkaline solution Alkaline solution
around 4 mm as the working electrode were prepared by use of a CuSO4/H2SO4-based system. Meanwhile,
a platinum counter electrode and Ag/AgCl reference electrode were used for the potentiodynamic measure-ments. The reference electrode was inserted into the corrosion cell through a Luggin tube, the tip of which was around 2 mm from the working electrode. The area abraded by the polishing cloth was 0.197 cm2. The downward force during polishing was varied from 0 to 68947.6 N/m2, and was measured using a weigh-ing balance supportweigh-ing the entire electrochemical cell. In this configuration, the copper-plated film rotated on the pad under pressure. A long napped synthetic rayon cloth (Buehler Microcloth) was used as the polishing pad for the corrosion experiments herein.
The electrochemical experiments were conducted by use of a potentiostat/galvanostat (EG&G model 273), while the corrosion software EG&G model 352 was adopted for potentiodynamic measurements. The elec-trochemical method of chronopotentiometry was employed before and after CMP to measure the thick-ness of the copper deposited film, allowing the copper removal rate to be calculated. The potentiodynamic data obtained with a voltage scan rate of 5 mV/s were used to measure the corrosion current densities and potential. Meanwhile, the corrosion current density was converted to the corresponding electrochemical corro-sion rate in namometer per minute by use of Faraday’s
law. Moreover, according to corrosion theory[13], the shift trend of potentiodynamic curves with various slurry concentrations should provide information regarding the corrosion mechanisms of copper CMP.
In order to measure the surface profile, copper was punched to disk of 3 in. in diameter for polisher operating at 34473.80 N/m2 and 8.376 rad/s with IC1000/SUBA IV pad. An atomic force microscopy (AFM) model Nanoscope IIIa with tapping mode was used to characterize the surface morphology and average roughness of samples before and after CMP using various slurries.
3. Results and discussion 3.1. Particulate contamination
The zeta potentials of materials of interest to copper CMP are shown in Fig. 2. The corresponding iso-electric points (IEP), i.e. the pH with zero charge on the surface, were measured to be 9.05, 4.56 and 7.64 for a-Al2O3, copper and cuprous oxide, respectively.
In order to minimize the particulate contamination on the copper surface, the same sign of surface charges between materials and copper to repel each other is necessary. As shown in Fig. 2, at acidic pH < 4:56 and alkaline pH > 9:05, the copper always
Fig. 1. The schematic representation of a corrosion test-cell used to measure the corrosion behaviors of copper films during CMP. 192 T.-H. Tsai, S.-C. Yen / Applied Surface Science 210 (2003) 190–205
had the repulsive electrostatic force with a-Al2O3and
cuprous oxide. However, a-Al2O3 and copper with
opposite charges at pH 4.56–7.64 would result in alumina contamination. Then at pH 7.64–9.05, i.e. near a neutral solution, both Al2O3and cuprous oxide
would be remained on copper surface. In acidic and alkaline slurries, therefore, the dispersion of a-Al2O3
particles in slurries is more stable, and the adhesion of residual materials on copper surface is avoidable. Next, the corrosion effects of acidic and alkaline slurries are investigated, respectively.
3.2. Corrosion effects of acidic and alkaline solutions
3.2.1. HNO3slurry
Fig. 3shows the effects of HNO3concentrations on
the polarization curves of copper and the corrosion parameters are listed in Table 2. It displays that the cathode polarization curve moves right with increasing HNO3concentration and the anodic region
does not almost shift. In addition, the corrosion poten-tial and corrosion current density are determined by the intersections of cathodic and anodic curves.
Therefore, both of these parameters would increase with increasing HNO3 concentration. According to
corrosion theory[13], the rightward shift of the cath-ode polarization curves reveal that corrosion is mainly accelerated by cathode reactions. Hence, HNO3is a
strong copper oxidizer capable of rapidly attacking copper. Additionally, the potentiodynamic curves in
Fig. 3 exhibit no steep slope in the anodic range, meaning that no passive films are formed on the copper surface. Consequently, copper may directly dissolve in HNO3solutions.
The corrosion potentials and pH in HNO3solutions
are listed inTable 2. The corresponding range and the equilibrium components of Pourbaix diagram are drawn on Fig. 4 [14]. It indicates that copper is corroded to Cu2þin HNO3 solutions, and no oxide
film is formed to protect the surface from corrosion. Copper dissolution is thus expected to be the dominant reaction in HNO3slurries. As mentioned above,
cop-per CMP in HNO3-based slurries contains the
follow-ing electrochemical reactions: Anode:
Cu! Cu2þþ 2e; E0¼ 0:337 V (1)
Fig. 2. Zeta potentials of alumina, copper and cuprous oxide as a function of pH. The vertical lines represent the isoelectric points of a-alumna (9.05), copper (4.56) and cuprous oxide (7.64).
Cathode: NO3þ 3Hþþ 2e! HNO2þ H2O; E0¼ 0:934 V (2) NO3þ 4Hþþ 3e! NO þ 2H2O; E0¼ 0:957 V (3) O2þ 4Hþþ 4e! 2H2O; E0¼ 1:229 V (4)
Even though the high removal rate for copper CMP in the slurry of 5 wt.% HNO3 can be attained from
the experiment, as listed in Table 2, the over high corrosion rate which is calculated from the measured corrosion current density with abrasion, is contained in this slurry. Consequently, the corrosion condi-tion can be described by the phenomenon of copper pitting as shown inFig. 5. It illustrates this schematic localized corrosion mechanism of copper CMP in
Fig. 3. Potentiodynamic curves of copper in the slurries of various HNO3concentrations without abrasion.
Table 2
Summary of pH, corrosion potentials (Ecorr), corrosion current densities (icorr), corrosion rates (CR) and removal rates (RR) of copper in HNO3
and NH4OH slurries
Solution pH Ecorr(mV) vs. Ag/AgCl icorr(A/cm2) RRa(nm/min) CRa(nm/min)
HNO3 1 wt.% 1.27 81.2 8.2 104 467 26.46 3 wt.% 1.15 107.1 2.6 103 525 46.30 5 wt.% 0.86 119.3 5.8 103 589 37.48 NH4OH 1 wt.% 11.2 451.11 5.6 105 275 0.140 3 wt.% 11.7 488.6 7.0 105 336 0.036 5 wt.% 12.3 534.4 9.1 105 270 0.080
aAbrasion at 34473.80 N/m2and 8.376 rad/s.
HNO3 slurries. Localized anodes are formed within
pits produced by the polishing of the particles, and cathodes are formed on the surrounding surface to support the dissolution reaction of copper. If no pro-tective films cover the copper surface to reduce the high corrosion rate, then the pitting propagation destroys the surface and deepens the pit so that the roughness is induced. Hence it was observed that the high corrosion rate worsens the smoothness of the copper surface in HNO3slurries.
3.2.2. NH4OH slurry
Fig. 6presents the potentiodynamic curves of cop-per in NH4OH slurries with no abrasion, and the
corrosion parameters are listed inTable 2. As shown in Fig. 6, the anodic passive region is prominent.
Meantime, this passivation can effectively reduce the copper corrosion rate. Then the corresponding potentials and pH ranges are drawn on Fig. 4. It indicates that copper is expected to form an oxide film in an alkaline solution. When the pH falls in the range 7–13, the potential of Cu2O formation is low.
However, CuO forms on copper at more positive potential. Therefore, the range listed inTable 2reveals that the passive film should be Cu2O in 1–5 wt.%
NH4OH solutions. On the other hand, for copper–NH3
system, the formation of stable copper–ammonia complex is rapid and the equilibrium components of Pourbaix diagram are illustrated in Fig. 4 [14,15]. Therefore, the anodic current density due to copper oxidation increases in spite of the pas-sivation and the corrosion potential decreases with
Fig. 4. Potential–pH equilibrium diagram for the system, copper–water, at 25 8C[14]. The dash regions and the equilibrium components corresponding to the Pourbaix diagram in HNO3and NH4OH solutions are shown[14,15].
increasing NH4OH concentration. It can be deduced
that the anodic current density increases because of the formation of stable copper–ammonia complex ion. The decreased corrosion potential reveals that the concentration of the copper ions is reduced according to the following Nernst relationship[13]:
DE/ ln aCu2þ (5)
where DE is the difference in potential and standard potential of an electrode, and aCu2þ is the activity of
Cu2þ. The decreased corrosion potential also means that the copper–ammonia complex ion forms in the solution instead of Cu2þ. Therefore the un-complex copper ion reduces with increasing NH4OH
concen-tration, and the corrosion potential is also decreased. To sum up, the electrochemical reactions of copper CMP in NH4OH-based slurries can be described as
follows: Anode: 2Cuþ H2O! Cu2Oþ 2Hþþ 2e; E0¼ 0:471 V (6) Cu2Oþ 2Hþ! 2Cu2þþ H2Oþ 2e; E0¼ 0:203 V (7)
Reaction and complex: Cu2þþ 4NH3! CuðNH3Þ4 2þ (8) Cu2Oþ 2Hþþ 2NH3 ! 2CuðNH3Þ2 þþ H 2O (9) Cathode: Cu2þþ 2NH3þ e! CuðNH3Þ2 þ; E0¼ 0:457 V (10) O2þ 2H2Oþ 4e! 4OH; E0¼ 0:401 V (11)
Fig. 7illustrates the schematic localized corrosion mechanisms of copper CMP in NH4OH-based slurries.
On the convex areas, the oxide film is mechanically abraded from the surface, and the dislodged materials are then dissolved by complex reactions. However, the oxide film can cover recessed areas and protect the underlying copper. Thus, the re-exposed copper region receives less mechanical abrasion, in comparison with the un-breakdown passive film, and then this region would be re-passivated. Due to this situation, the pitting propagation is inhibited. Then above-mentioned actions could be applied to each point of copper surface uni-formly to produce global planarization.
Fig. 5. The schematic localized corrosion mechanisms of copper CMP in HNO3-based slurries.
The CMP results of NH4OH slurries, listed in
Table 2, indicate that all of the removal rates and corrosion rates in NH4OH slurries are lower than those
in HNO3slurries owing to the formation of passive
films. Hence, the NH4OH-based slurries with low
corrosion rate can perform better than HNO3-based
slurries from the perspective of planarization. As mentioned above, the reactions of chelating copper ions are expected to increase corrosion rates, while the passive reactions decrease the corrosion rate. How-ever, the corrosion rate obtained from our experiments does not increase obviously with increasing NH4OH
concentrations. As a result, the removal rate is still so low in NH4OH slurries that the throughput of CMP is
limited. Thus, it should be concluded the rate-deter-mining step among the mechanisms of copper in NH4OH slurries is the oxidation reaction of the copper
surface film but chelating reactions. In addition, the slow formation rate and the rapid chelating rate of passive films induce the pitting propagation in the breakdown region before initiating another breakdown of passive film. Therefore, the removal rate of copper
CMP needs to be enhanced by adding an oxidant, such as KNO3, to NH4OH-based slurries so that the copper
oxidation can be accelerated.
3.3. Corrosion modifications of acidic and alkaline slurries
Since the over high dissolution rate worsen the smoothness of the copper surface in acidic HNO3
slurries and the over low removal rate limits the application for the need of high throughput in alkaline NH4OH slurries, some modifications of corrosion
effects to apply the acidic and alkaline slurries are required.
3.3.1. Dissolution modification
As mentioned above, the alumina slurry needs to be maintained in either acidic or alkaline surroundings. Meanwhile, it is considered to reduce the corrosion rate in acidic HNO3 slurries. So a well-known
oxide resist for copper, benzotriazole (BTA), has thus been adopted herein as an additive to the slurries.
Fig. 6. Potentiodynamic curves of copper in the slurries of various NH4OH concentrations without abrasion.
The modified effects of adding BTA are presented in
Fig. 8. Increase in corrosion rates with HNO3
con-centrations indicate once again that no passive film is formed on the copper surface. In addition, Fig. 8
reveals that the corrosion rate in 5 wt.% HNO3
declined two orders of magnitude after 0.1 wt.% BTA was added. The copper surface is thus clearly anti-corroded by the BTA. It shows that the adding of an inhibitor, BTA, can form an adsorbed BTA of non-native film on copper surface (Cu-BTA film) and this non-native film can protect copper against pitting propagation. Furthermore, the dependence of copper corrosion rates on BTA concentrations in 5 wt.% HNO3solution is also displayed inFig. 8(top axis).
It is obvious that copper corrosion rate decreased with increasing BTA concentrations. However, the non-native passive effect of BTA was weakened when the BTA concentrations exceed 0.1 in 5 wt.% HNO3
slurries. It seems that the adsorption of 0.1 wt.% BTA on a 0.197 cm2copper surface is saturated. Moreover, to investigate the effect of BTA on the polishing, potentiodynamic measurements were conducted in 5 wt.% HNO3 slurries, and their potentiodynamic
curves were presented inFig. 9. From the figure, add-ing 0.1 wt.% BTA produces a significant difference in
corrosion potential (DEd) between abrasion and no
abrasion, 145.8 and 0.1 mV, respectively, listed in
Table 3. Nevertheless, this difference, on the contrary, would not been observed from the curves without BTA. Hence, this difference can be explained by the sensibility of surface for the abrasion. When 0.1 wt.% BTA is added, the Cu-BTA film is formed stably on the surface with no abrasion, and the cor-rosion potential approached near to zero. However, Cu-BTA film is removed from copper with abrasion to create a corrosion interface of copper and HNO3, so
the reaction potential is higher. As observed, the corrosion potential in the electrochemical measure-ments varied with the mechanical polishing, but that did not occur when no BTA is added, i.e. in only 5 wt.% HNO3 slurries. This phenomenon suggests
that BTA can effectively protect the recessed areas from undesired copper pitting propagation and can be included in the acidic slurries of copper CMP, such as HNO3-based slurries.
3.3.2. Oxidization modifications
Corrosion effects of alkaline slurries require to be modified from the view of enhancing oxidation rate of copper as mentioned before. Potassium nitrate is a
Fig. 7. The schematic localized corrosion mechanisms for copper CMP in NH4OH-based slurries.
strong oxidizer owing to the oxidation power of nitrate, and thus can be considered to add into alkaline slurries. Moreover, the pH of slurries would not be varied with the added KNO3, so the dispersion of
alumina abrasives can be maintained.Fig. 10displays all of the potentiodynamic curves, including 5 wt.% KNO3, 5 wt.% KNO3þ 1 wt.% NH4OH and 1 wt.%
NH4OH slurries and the corrosion parameters are
summarized in Table 3. In comparison with these curves of abrasion and no abrasion, it reveals that the anodic passive regions of potentiodynamic curves are quite prominent when measured without abrasion, indicating that passive actions exist on copper surfaces no matter in 5 wt.% KNO3, 5 wt.% KNO3þ 1 wt.%
NH4OH or 1 wt.% NH4OH slurries. Among these
curves of Fig. 10, the potentiodynamic curves for
Fig. 8. Copper corrosion rates versus HNO3 concentrations (bottom axis) with and without 0.1 wt.% BTA (* and &) and vs. BTA
concentrations (top axis) in 5 wt.% HNO3solutions (~).
Table 3
Summary of corrosion potentials, corrosion current densities and the drop of corrosion potentials with abrasion (DEd) of copper in various
slurries
Slurry Ecorr(mV) vs. Ag/AgCl icorr(A/cm
2
) DEd
a
(mV) No abrasion Abrasion No abrasion Abrasion
5 wt.% HNO3 119.3 119.0 5.8 103 1.7 103 0.3 5 wt.% HNO3þ 0.1 wt.% BTA 0.1 145.8 3.7 106 1.5 105 145.7 5 wt.% KNO3 4.9 180.5 5.3 106 7.9 106 175.6 1 wt.% NH4OH 370.1 410.2 5.6 105 6.5 106 40.1 5 wt.% KNO3þ 1 wt.% NH4OH 385.9 422.1 1.2 105 3.4 105 36.2 aDE
d¼ Ecorrðno abrasionÞ EcorrðabrasionÞ.
the 5 wt.% KNO3 slurries have the largest drop of
corrosion potentials with abrasion (DEd); meanwhile,
an increment in the corrosion current density occurs for the abraded sample. This phenomenon indicates that the strong oxidization of copper surface results in the formation of copper oxide in 5 wt.% KNO3
slur-ries. Therefore, the corrosion reaction takes place at higher potential with no abrasion. When fresh copper surface is exposed again to the 5 wt.% KNO3slurry by
removing passive films with abrasion, the copper could be oxidized progressively at lower potential. As adding 5 wt.% KNO3 into 1 wt.% NH4OH, the
drop of corrosion potentials with abrasion (DEd) is
lessened to 36.2 mV as listed in Table 3. This value approached to DEdin 1 wt.% NH4OH slurry. It reveals
that chemical complexation as well as mechanical polishing achieves the removal of passive films. When NH4OH is added during copper CMP, complexation
by NH3 helps to dissolve the passive film and
dis-lodged materials. The corrosion potential is thus less sensitive to abrasion, and less difference in potentials could be observed in KNO3–NH4OH slurries.
Table 4lists the experimental polishing results from several slurries. A high removal rate, which is equiva-lent to high process throughput, is always essential for CMP, but a high corrosion rate is generally not accep-table due to the induced surface roughness. The ratio of removal rate to corrosion rate is defined as the CMP efficiency factor. Hence, a high CMPEF means high mechanical removal ability and a well planarity. However, usual slurry has two conflicting functions in the mechanism of metal CMP: one is to dissolve materials dislodged from the surface and carry them away to avoid re-deposition, while the other is to assure planarity by forming a surface film to avoid the pit-ting propagation of breakdown areas on the surface.
Fig. 9. Potentiodynamic curves of copper in 5 wt.% HNO3 and 5 wt.% HNO3þ 0.1 wt.% BTA slurries with no abrasion and abrasion
(34473.80 N/m2/8.376 rad/s).
As a result, both high removal rate and high CMPEF must be achieved herein. As presented in Table 4, CMPEF in 5 wt.% KNO3is high enough, but the removal
rate in 5 wt.% KNO3(253 nm/min) is so low that KNO3
slurry has limited applications for the copper CMP process. The measured results also show that the oxidi-zer 5 wt.% KNO3can significantly increase the removal
rate from 253 nm/min in the absence of NH4OH, to
360 nm/min with 1 wt.% NH4OH added. Furthermore,
although all the cases of HNO3-based slurries have
higher removal rates according toTable 4, CMPEF in
pure HNO3slurry is so low that the pitting propagation
cannot be inhibited to make appropriate slurry for copper. Therefore, in view of higher CMPEF and higher RR, pure HNO3, pure KNO3or pure NH4OH
slurry is not suitable for copper CMP, but the modified slurries can improve their performances.
3.4. Corrosion effects on mechanical action
The mechanical action is also controlled by corro-sion behaviors between copper and slurries. It is
Fig. 10. Potentiodynamic curves of copper in 5 wt.% KNO3, 5 wt.% KNO3þ 1 wt.% NH4OH and 1 wt.% NH4OH slurries with no abrasion
and abrasion (34473.80 N/m2/8.376 rad/s).
Table 4
Removal rates (RR), corrosion rates (CR), their ratios (CMPEF), the average value of dRR/dP and the surface roughness (Rq) of copper for
CMP with various slurries at 34473.80 N/m2and 8.376 rad/s
Solution RR (nm/min) CR (nm/min) CMPFEa dRR/dP Rq(nm)
H2O 73.6 0.014 5.26 103 – 29.345 5 wt.% HNO3 589 37.48 1.51 101 8.0 32.024 5 wt.% HNO3þ 0.1 wt.% BTA 484 0.332 1.46 103 17.5 4.019 1 wt.% KNO3 253 0.174 1.45 10 3 – – 1 wt.% NH4OH 275 0.140 1.96 10 3 4.4 14.791 5 wt.% KNO3þ 1 wt.% NH4OH 360 0.742 4.86 10 2 9.9 3.971 aCMPEF¼ RR=CR.
known that the simplest relation describing the mechanical action of copper polishing is the Preston equation [16], namely the removal rate is directly proportional to the polishing pressure and the relative speed. Herein, the removal rate is a function of both
downward force and relative rotating speed, as dis-played inFig. 11, for various slurries. Combined with the corrosion mechanisms above, this relation may be used to explain the removal condition on copper surfaces. For a constant rotation speed, the removal
Fig. 11. Copper removal rates versus pressures at 8.376 rad/s in various slurries: (a) 5 wt.% HNO3and 5 wt.% HNO3þ 0.1 wt.% BTA
slurries; (b) 5 wt.% KNO3þ 1 wt.% NH4OH and 1 wt.% NH4OH slurries.
rate mainly relies on the polishing pressure (P), and its dependency can be described by the differentiation of RR with respect to the P, dRR/dP, as listed inTable 4. It indicates that the measured results of dRR/dP in 5 wt.% HNO3 þ 0.1 wt.% BTA (17.5) and 5 wt.%
KNO3þ 1 wt.% NH4OH (9.9) slurries are higher than
those in 5 wt.% HNO3(8.0) and 1 wt.% NH4OH (4.4),
respectively, found fromFig. 11a and band listed in
Table 4. The higher dRR/dP indicates that more passive films are removed, that is, more films are formed on copper surfaces by the chemical adsorption of BTA or oxidation by nitrate. Meanwhile, higher
Fig. 12. AFM micrographs showing the morphology of copper surface: (a) before polishing and after polishing in slurries of (b) DIW, (c) 5 wt.% HNO3, (d) 5 wt.% HNO3þ 0.1 wt.% BTA, (e) 1 wt.% NH4OH and (f) 5 wt.% KNO3þ 1 wt.% NH4OH.
dRR/dP also means the removal mechanisms are sensitive to mechanical polishing, representing higher CMPEF and better planarity. The results of dRR/dP in
Table 4indicate that BTA or KNO3has a good ability
to passivate copper surfaces and protect copper from pitting propagation.
3.5. Surface morphology of copper
Fig. 12 show the AFM micrographs of copper surfaces before and after CMP with various slurries and the average surface roughness is listed inTable 4. Before CMP, the image of the copper surface with the calculated surface roughness, 62.379 nm, for the scan size 5 mm 5 mm is shown in Fig. 12a. Then the roughness was reduced to 29.345 nm after CMP using only 5 wt.% a-Al2O3aqueous slurries without adding
any other chemical reagent. However, the surface roughness as shown inFig. 12cis increased contrarily after CMP using 5 wt.% HNO3without any corrosion
modifications. It indicates that serious dissolution on copper surface results in depth pitting corrosion, and the pitting propagation for copper proceeds to destroy the planarization. Thus, the pits of copper surface are obvious as indicated inFig. 12c. As 0.1 wt.% BTA was added into 5 wt.% HNO3slurries, shown inFig. 12d, it
can be seen that the polished copper surface appears relatively smooth, and what is more, the surface roughness decreases from 32.024 to 4.019 nm. On the other hand, after copper CMP using 1 wt.% NH4OH slurry, the surface roughness decreases to
14.791 nm that is lower than it in 5 wt.% HNO3
slurries. Several smaller pits of copper surface, how-ever, are still found as indicated inFig. 12d. Above results respond to the mechanisms of pitting corrosion as described inFigs. 6 and 7, i.e. the pitting propaga-tion of breakdown regions may occur in HNO3and
NH4OH slurries. Furthermore, the surface roughness
is the lowest after CMP using 5 wt.% KNO3þ 1 wt.%
NH4OH slurry. All of the AFM results are consistent
with above corrosion actions. As was noted pre-viously, different corrosion modifications of acidic and alkaline can enhance the CMP performance and achieve better planarization. Thus, it is supposed that copper CMP using 5 wt.% HNO3with the dissolution
modification by adding 0.1 wt.% BTA or 1 wt.% NH4OH with the oxidization modification by adding
1 wt.% KNO3can provide a smooth and homogeneous
surface, which is well adequate for delineating copper patterns in the deep sub-micron integrated circuits.
4. Conclusion
In acidic pH < 4:56 and alkaline pH > 9:05 slur-ries, the copper always had the same charge with a-Al2O3and Cu2O. Thus, the contamination of residual
particles on the copper surface is minimized and the dispersion of a-Al2O3particles in slurries is improved
for these pH ranges. The surface layer of copper experiences, however, pitting initiation and propaga-tion in HNO3and NH4OH slurries. These phenomena
would worsen the planarization. Thus, some corrosion modifications are required to enhance the CMP effi-ciency. In a general copper CMP process, both removal rate and CMPEF must be high. The addition of 0.1 wt.% BTA modifies 5 wt.% HNO3slurries to
achieve effective corrosion inhibition for copper films and to obtain a high CMPEF. On the other hand, adding an extra oxidizer, KNO3, modifies 5 wt.%
NH4OH to increase the copper anodic oxidation on
the copper surface, and to accelerate the removal rate as well as the CMPEF. Moreover, the higher slopes of removal rate vs. pressure (dRR/dP) for 5 wt.% HNO3
þ 0.1 wt.% BTA and 5 wt.% KNO3þ 1 wt.% NH4OH
slurries indicate that corrosion modifications can enhance the surface sensitivity to mechanical polish-ing, i.e. CMPEF. Above all, corrosion measurements and pitting mechanisms used herein are the most viable means of evaluating the performance of slurries and the direction of modifications for metal CMP.
Acknowledgements
The authors thank the National Science Council of Taiwan, Republic of China, for financially supporting this research under Contract no. NSC 89-2214-E-002-030.
References
[1] D. Edelstein, J. Heidenreich, R. Goldblatt, W. Cote, C. Uzoh, N. Lustig, P. Roper, T. McDevitt, W. Motsiff, A. Simon, J. Dukovic, R. Wachnik, H. Rathore, R. Schulz, S. Su, S. Luce, J. Diggest, in: Proceedings of the International Conference 204 T.-H. Tsai, S.-C. Yen / Applied Surface Science 210 (2003) 190–205
IEEE International Electronic Devices Meeting, Washington, DC, 1997, p. 773.
[2] M.T. Wang, W.K. Yeh, M.S. Tsai, W.T. Tseng, T.C. Chang, M.C. Chen, Mater. Chem. Phys. 51 (1997) 75–79.
[3] F.B. Kaufman, D.B. Thompson, R.E. Broadie, M.A. Jaso, W.L. Guthrie, D.J. Pearson, M.B. Small, J. Electrochem. Soc. 138 (11) (1991) 3460–3465.
[4] Y. Gotkis, S. Alamgir, L. Yang, F. Dai, F. Mitchell, J. Nguyen, L. Shumway, L.R. Walesa, J. Yang, P. Nunan, K. Holland, J. Vacuum Soc. Technol. B 17 (5) (1999) 2269–2271. [5] P. Wrschka, J. Hernandez, G.S. Oehrlein, J.A. Negrych, G.
Haag, P. Rat, J.E. Currie, J. Electrochem. Soc. 148 (6) (2001) G321–G325.
[6] Y. Li, S.V. Babu, Electrochem. Solid-State Lett. 4 (2) (2001) G20–G22.
[7] H.G. Kim, Y.M. An, D.K. Moon, J.G. Park, Jpn. J. Appl. Phys. 39 (1) 3A (2000) 1085–1090.
[8] W.C. Chen, C.T. Yen, J. Vacuum Soc. Technol. B 18 (1) (2000) 201–207.
[9] S. Kondo, N. Sakuma, Y. Homma, N. Ohashi, Jpn. J. Appl. Phys. 39 (1) 11 (2000) 6216–6222.
[10] M. Bielmann, U. Mahajan, R.K. Singh, D.O. Shah, B.J. Palla, Electrochem. Solid-State Lett. 2 (3) (1999) 148–150. [11] L. Zang, S. Raghavan, M. Weling, J. Vacuum Soc. Technol. B
17 (5) (1999) 2248–2255.
[12] E.A. Kneer, C. Raghunath, V. Mathew, S. Raghavan, J. Electrochem. Soc. 144 (9) (1997) 3041–3049.
[13] D.A. Jones, Principles and Prevention of Corrosion, second ed., Prentice Hall, Upper Saddle River, NJ, 1983.
[14] M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solutions, NACE, Houston, TX, 1975.
[15] H.E. Johnson, J. Leja, J. Electrochem. Soc. 112 (1965) 638.
[16] F.W. Preston, J. Soc. Glass Technol. 11 (1927) 214. T.-H. Tsai, S.-C. Yen / Applied Surface Science 210 (2003) 190–205 205