Chapter 1 Introduction
1.2 Thesis Outline
In this study, we studied the effect of CuBTA layer on surface leakage. We would compare
Cu(Ⅰ)-BTA layer with Cu oxide on Cu surface from the view of surface leakage. Cu oxide on Cu
surface could be used to describe the surface condition after conventional CMP.
In chapter 2, three types of chelotors would be compared their cleaning efficiency and corrosion
effect in this study. In additional, we would explore the influence of pH on chelating Cu ions.
Furthermore, wetting ability and corrosion of Cu lines in the cleaning solutions also were discussed.
In chapter 3, we would compare Cu(Ⅰ)-BTA layer with Cu oxide on Cu surface fromed the view of
surface leakage. To build the condition of Cu oxide on Cu surface, buffing with KOH also was discussed.
In additional, the thermal stability of CuBTA layer would be discussed.
Finally, conclusions were given in chapter 4.
Chapter 2
Cleaning Efficiency of Chelator Solutions 2.1 Introduction
Although Cu CMP has being the enabling technology for multilevel Cu interconnect manufacturing,
there were several challenges to its implementation. One of the serious challenges was post-Cu CMP
cleaning. After Cu CMP, Cu contamination presented on the surface in the form of homogeneous film
[16]. Cu diffused quickly both in the silicon wafer and in deposited dielectric films. Cu would formed
several deep levels in the silicon band gap and acted as recombination centers, which reduced minority
carrier lifetime [5]. In additional, Cu residuals on dielectric would form a leaky path and were likely to
lead short between metal lines [2][3]. Hence, Cu was considered as a very serious contamination for
silicon device and must need to be removed from interlevel dielectrics surface after Cu CMP.
It was a challenge to reduce Cu residuals on the dielectrics to a level less than 5×1010 atoms/cm2
without causing corrosion to Cu lines [16]. Figure.1-4 showed the Pourbaix diagram of Cu-H2O system. It
indicated that acidic solutions(pH<5), Cu oxides did not form and Cu dissolves as Cu＋ at noble (high)
potential. On the other hand, in highly alkaline solutions at pH>13, Cu would form CuO2－ at noble
potential. Because Cu was corroded in acidic and alkaline solutions easily, ammonium hydroxide
(NH4OH) and hydrofluoric acid (HF) which were used for conventional post-oxide CMP cleaning would
be not suitable for post-Cu CMP cleaning [17]. Especially for NH4OH, it formed soluble and stable
complex compounds with Cu, which would corrode the Cu lines seriously as shown in Figure.1-5.
In this study, several metal chelators would be used to remove Cu ions from the interlevel dielectrics
surface. Metal chelators were known to form stable complexes with Cu ions [18]. Because metal
chelators had one or several dentates, they would react as electron-pair acceptors to form coordination
compounds or complex ions with metal ions. The metal chelators in solution would form uncharged
Cu-chelator complexes by coordination with metal ions. The removal of Cu contamination from wafer
surface by metal chelators could be understood by the distribution equilibrium [19]. The distribution
equilibrium could be expressed by the following equations. Metal ions deposited on the wafer surface
were dissolved into the aqueous phase (Eq.2-1). The complex reaction occured between metal ion and
metal chelator molecule (Eq.2-2). Metal-chelator complex might absorb again on the wafer surface
(Eq.2-3). Metal chelator molecule might adsorb on the wafer surface (Eq.2-4).
(Eq.2-1)
(aqueous Laqueous MLaqueous
M + ↔ (Eq.2-2)
phase, (solid)= adsorb on the surface.
Three types of chelotors with different structures would be compared from their cleaning efficiency
and corrosion effect in this study. Table.2-1 listed the three types of chelator solutions. In the type1, citric
acid and ADPA-60 had two dentates. In the type2, EDTA and 422-25S had six dentates. The catechol and
TBC were aromatic compounds with a cyclobenzeze.
The pH of chelaors also played important role for chelating capability. We would research the
influence of pH on chelating Cu ions. Furthermore, good wetting ability would ensure whole wafer
surface would be coved with chelator chemicals, which make Cu ions cleaning uniformly. We would
discuss the wetting ability of chelator solutions in this study.
2.2 Experimental
2.2.1 Wetting Ability Test of Chelator Solutions
The substrates were standard 6-inch diameter p-type silicon, (100) orientation wafers. After the
standard RCA, about 5500A thick SiO2 was thermally grown from the silicon substrate in the furnace.
Contact angle test was carried out to decide the wetting ability of chelator solutions. In this study,
the concentration of the solutions is 1E-2 and each drop was fixed at 3ml.
2.2.2 Corrosion Test
The blanket Cu test wafers were stacked Cu/Ta layer structure with a combination thickness of
1000/50 nm which were sputtering deposited onto the p-type, (100) oriented, 6-inch bare silicon wafers
with 200 nm thick oxide deposited by PECVD. The PECVD system was STS multiplex cluster system
and the sputter system was ULVAC SBH-3308 RDE. The under layer of 50nm Ta was used as an
adhesion promoter for the Cu deposition, since Cu did not adhere well on the thermal oxide. It was also
used as a diffusion barrier, because Cu was very easy to diffuse into oxide with high diffusivity.
The blanket wafer was immersed into the cleaning chelator solutions for 3 minutes. The
concentration of chelator solutions was 1E-2 M. Four point measurement was performed to measure the
thickness of Cu films before and after etch respectively and calculate the etch rate, describing in chapter
2.3.
2.2.3 Cleaning Test
2.2.3.1 Sample Preparation
The experiment flow was shown in Figure.2-1. The substrates were standard 6-inch diameter p-type
silicon, (100) orientation wafers. After the standard RCA, about 5500A thick SiO2 was thermally grown
from the silicon substrate in the furnace. To make the Cu ions bonded on the oxide layer uniformly
without agglomeration, CMP process was used before immersing in 1M CuSO4 for 2 minutes. After CMP
process, the surfaces of oxide layers would become fresher and bond with Cu ions easily. The CMP setup
is described later in chapter 2.2.3.2. Then, the blanket wafers were cleaned using D. I. Water ( DIW ) by
the post-CMP cleaner of Solid State Equipment Corporation MODE 50 (SSEC-M50). The duration of
cleaning was 7 cycles (15 cycles/min) and the rotation rate of wafer was 800 rpm. Then, blank wafers
were dry spun at the rotation rate of 2500 rpm.
After preparing the Cu contaminated wafer, the first TXRF analysis was carried out to calculator the
amount of Cu ions. Three types of chelator solutions cleaning were performed on the SSEC-M50 cleaner,
and following with the second TXRF analysis. Table.2-2 listed the cleaning steps and parameters of
SSEC-M50. Then the second TXRF analysis was executed to calculate the cleaning efficiency, describing
in chapter 2.3.
2.2.3.2 CMP Process Polisher Setup
A Westech Model 372M CMP processor (Figure.2-2), consisting of a wafer carrier and a primary
circular polishing table mounted with Rodel IC 1400TM grooved (made of polyurethane impregnated
polyester) pad and a secondary buffing table mounted with an Rodel Politex Regular E.TM pad, a carrier to
hold wafers against the pad, and a Rodel R200-T3 carrier film to provide buff between the carrier and
wafer was used for CMP experiments. Recesses in the carrier template mechanically constrain a single
6-inch wafer, preventing it from sliding out from under the carrier during polishing. A polymeric film
placed in the recess brought the wafer slightly above the surrounding template surface. When the film
was wetted, it provided sufficient surface tension to hold the wafer while it is being positioned over the
polishing table. The teflon retaining ring was recessed from the wafer surface about 7 miles. The slurry,
pumped out from a reservoir at a controlled rate, was dispensed onto the center of the table. The table and
the carrier were both motor driven spindles, rotated independently at constant angular velocities (rpm).
The arm was oscillated about their position at half radius of the table to utilize more pad area and to
reduce pad wear [20]. Pressure at the wafer-slurry-pad interface was controlled via an overhead
mechanism, which allowed pressure to be applied onto the wafer carrier.
Pad Prewet & Pad Conditioning
Pad prewet was performed before the start of each polishing action. The prewet slurry flow rate was
at 300 ml/min and the prewet time was fixed at 20 seconds.
Pad conditioning was employed to resurface the pad in order to maintain the removal rate without
sacrificing uniformity. The purpose of pad conditioning was to clean the slurry residuals and to lift the
pad fiber for further processing. Without this procedure, the polishing rate decreased substantially after
several polishing cycles. In our experiments, pad conditioning was done by brush artificially. Pad
conditioning was performed before and between each wafer, and polishing was terminated before pad
glazing could cause significant reduction in removal rate.
Polishing Recipes & Slurry Formations
The polishing recipes and slurry formulations in cleaning experiment were all listed in the Table.2-3.
The commercial SS-25 slurry was colloidal silica abrasive with the size of 30-50 nm approximately. In
the phase1, oxide layers were polished to establish the fresher surface and would bond with Cu ions
easily. Phase2 is to remove the residual slurry from wafer surface.
2.3 The Performances of Chelator Solutions Wetting Ability
Information obtained from contact angle provided the fundamental understanding of solid-solid and
solid-liquid intermolecular interactions (ex: van-der Waals, acid/base type interations, and electrostatic
interations). Considered the drop of a liquid rested on a solid surface. The drop of liquid forming an angle
might be considered as resting in equilibrium by balancing the three forces involved. Namely, the
interfacial tensions between solid and liquid (γSL), that between solid and vapor (γS) and that between
liquid and vapor (γL) interface. The equilibrium of three forces and the resulting contact angle was given
by the well-known Young＇s equation (Figure.2-3) : θ γ γ
γSL = S − Lcos (Eq.2-5)
where θ was contact angle [21].
Good wetting ability would ensure the whole wafer surface would be coved with cleaning chemicals,
which make Cu ions cleaning uniformly. In chelator wetting ability test, contact angle system was KRuSS
GmbH and each drop is fixed at 3ml. The concentrations of chelator solutions were 1E－2.
Cu Film Thickness Measurement
In this study, thickness of the Cu film was calculated by dividing the film resistivity with its
measured sheet resistance. The relation between thickness and resistivity was given by
resistivity of Cu film in our experiment was in the range from 1.8 µΩ‧cm/□ to 2.3 µΩ‧cm/□.
Four point probe system ( Napson RT-80/RG-80 ) was used to measure sheet resistance. For a thin
wafer with thickness T much smaller than either a or d, the sheet resistance Rs was given by
I CF V
Rs = ⋅ (Eq.2-7)
where CF was the correction factor (Figure.2-4). In the limit when d»S, where S was the probe spacing,
the correction factor becomes (π/ln 2)=4.54 [22].
The etch rate of blanket Cu films were calculated by following formula:
The total reflection X-ray fluorescence spectrometry (TXRF) could sensitively detect the metallic
impurities on surface. In this study, the Cu contamination on the dielectric surface was detected using an
ATOMIKA 8030W TXRF system. TXRF was based on the photoelectric effect. When an atom irradiated
with highly energetic photons, an electron from one of the inner shells might be ejected. As the vacant
place was filled by an electron from an outer shell, a photon whose energy was characteristic of the atom
was released. This radiation was called fluorescent radiation and detected by an energy dispersive
detector. TXRF made use of total reflection of the primary X-ray beam at grazing incidence (Figure.2-5)
[23] [24]. The high reflectivity in the total reflection mode resulted in an extremely low energy transfer
from the incident beam into the irradiated substrate, because most of the energy was reflected and does
not penetrate through the interface.
In this study, TXRF was used to decide the amount of Cu ions. The cleaning efficiency was gave by
Cleaning Efficiency=
[amount of Cu ions after chelators cleaning]
[amount of Cu ions before chelators cleaning]
(Eq.2-9)
Electrochemical Analysis
Figure.2-6 depicted a typical electrochemical corrosion test cell consisting of three electrodes
submerged in an electrolyte. Electrical current form a potentiostat changed the test electrode potential
from its open circuit potential (OCP), to a potential value that was determined by the magnitude of
potentiostat current. Test electrode polarization was measured as a potential difference between reference
and test electrodes. No electrical current flowed between a potentiostat and reference electrodes, so it
remained at its OCP and provided a fixed reference point for corrosion measurement. The reference
electrode was also used provide feedback to the potentiostat, so that test electrode potential could be
monitored and adjusted to a desired level [25].
In this study, electrochemical analysis was used to describe the corrosion behavior of chlators with
different pH. All electrochemical analyses were carried out in a conventional three-electrode system at
room temperature. A platinum electrode was used as the counter electrode, and an Ag/AgCl was
employed as the reference electrode. A Cu cylinder was used as the working electrode, and area of cross
section is 0.5 cm2.
2.4 Results and Discussions
2.4.1 Wetting Ability and Corrosion of Chelotor Solutions
The wetting ability of metal chlators was investigated by contact angle measurement. The results
were shown in Table.2-4. The concentration of the solutions is 1E-2 and each drop was fixed at 3ml.
Table.2-4 indicated that all three types of chelator solutions had low contact angles, which implied good
wetting ability. Good wetting ability ensured whole wafer surface would be coved with chelator
chemicals, which make Cu ions cleaning uniformly around the wafer..
Figure.2-7 showed the corrosion effect of chelator solutions for Cu films and the formula was shown
in Eq.2-8. The concentration of the solutions is 1E-2 M and etch time was 3 minutes. It indicated that all
three types of chelater solutions had low corrosion rate, even in high concentration.
2.4.2 Cleaning Efficiency of Chelator Solutions
Figure.2-8 and Figure.2-9 showed the results of cleaning efficiency. The calculating formula was
showed in Eq.2-9, which the amount of Cu ions before chelators cleaning ranged about from 170×1010 to
200×1010 atoms/cm2. As shown in Figure.2-8, the cleaning efficiency would be saturated after 15 cycles
(15 cycles/min) cleaning time for all kinds of cleaning solution. EDTA had six strong potential sites for
bounding with Cu ions: the four carboxyl groups and the two amino groups, hence EDTA had the best
cleaning efficiency. On the contrary, Catechol and TBC had only two dentates and exhibited the worse
cleaning efficiency. The cleaning efficiency was strongly dependent on the numbers of chelating sites.
Figure.2-9 indicated the concentration of chelators only had little influence on cleaning efficiency.
The low concentration of chelators had enough ability to cleaning the most Cu ions. The other Cu ions
could not be chelated, even in high concentration of chelators. This indicated chelators could be used for
cleaning with low concentration to decrease budget.
2.4.3 pH Effect on Cleaning Efficiency
The pH of cleaning solutions would influence chelating capability owing to the varying activing of
protonated or deprotonated functional groups. For a metal chelator, Y, the chelating reaction for a metal
ion M could be represented as [26]:
MY
The concentration of Y in the solution was given by:
]
The proton equilibrium-constant was given by
n
Substituting Eq.2-14 into Eq.2-12, the expression ofαY could now be written as:
∑
=according to Eq.2-14, at a high pH of cleaning solutions.
In this study, type1 of chelators was taken for experiment and KOH was used to modify the pH of
chelator solutions. Figure.2-10 and Figure.2-11 showed the cleaning efficiency of citric acid with various
cleaning times and various concentration, respectively. Both Figure.2-10 and Figure.2-11 showed citric
acid in the acidic environment had better cleaning efficiency than in the alkaline environment, but it did
not agreed with Eq.2-14. ADPA-60 had the same result with citric acid as shown in Figure.2-12 and
Figure.2-13. It might be due to that Cu was oxidized to Cu oxide in the alkaline environment. Chelators
could not chelate those Cu oxide, hence the cleaning efficiency reduce in the alkaline environment. The
Pourbaix diagram indicated that CuO formed between pH=7 to pH=13 at noble (high) potential and Cu2O
formed between pH=5 to pH=15 at active (low) potential as shown in Figure.1-4. Figure.2-14 indicated
the etch rate of citric acid in the alkaline environment lower than in the acidic environment, which also
implied the passivation appeared in the alkaline environment. In additional, Tafel diagram as shown in
Figure.2-15 indicated that surface passivationon Cu formed in the alkaline environment. ADPA-60 had
the same phenomenon with citric acid in the alkaline environment as shown in Figure.2-16 and
Figure.2-17. It was reasonable to presume that all three types of chelators had the lower cleaning
efficiency in the alkaline environment than in the acidic environment, because Cu was oxidized to Cu
oxide in the alkaline environment.
2.5 Summary
In this study, we discussed several effects on cleaning efficiency of three types of clelators. All three
types of chelator solutions had low contact angles, which implied good wetting ability. Good wetting
ability ensured whole wafer surface would be coved with chelator chemicals, which made Cu ions
cleaning uniformly around the wafer. In additional, chelater solutions had low corrosion rate for Cu lines,
even in high concentration.
The cleaning efficiency was strongly dependent on the numbers of chelating sites. EDTA had six
strong potential sites for bounding with Cu ions: the four carboxyl groups and the two amino groups,
hence EDTA had the best cleaning efficiency. On the contrary, Catechol and TBC had the fewest dentates
and showed the worse cleaning efficiency. Besides, chelator solutions in the alkaline environment were
improper for cleanig, because Cu was oxidized to Cu oxide in the alkaline environment. Chelators could
not chelate those Cu oxide, hence the cleaning efficiency reduced in the alkaline environment.
Chapter 3
Effect of CuBTA Layer on Surface Leakage 3.1 Introduction
The damascene process was regarded to be the an essential and critical step for manufacturing Cu
interconnect, and the chemical mechanical polishing of Cu and barrier metal was the key to enable this
process. There were several CMP issues which should be taken into account for implement metal
polishing—non-uniformity, rounding, dishing, and erosion—had also been addressed. In order to provide
damascene metal lines with high accuracy and yield, a two step CMP had been introduced to achieve a
large removal rate while suppressing metal dishing (Figure.3-1). The performance of the second step
polishing was to remove the barrier metal selectively. Whatever the degree of dishing during the first step
was, if the removal rates of tantalum (barrier metal) and oxide (interlevel dielectric) higher than that of
Cu, it was able to reduce both the dishing and oxide erosion within the accepted range in the second step.
It had been demonstrated that the slurry composed of colloidal silica abrasive and H2O2 could satisfy the
demands of the second step polishing.
During the CMP process, Cu would be oxidized to form Cu oxides (CuO or Cu2O) and Cu
hydroxides (Cu(OH)2) passivation on Cu surface[13]. Then, these passivation on high feature would be
polished to reach global planarization, while low feature would not be polished. As polishing with
colloidal silica based slurry, it shows the strong absorption of colloidal silica on Cu surface. This might
be related to that the colloidal silica chemisorbed on the Cu oxide layer by means of oxygen bridging
bonding [14]. It was difficult to remove colloidal silica by conventional chemical clean. In additional,
Several papers also indicated this Cu oxide layer was a source of surface leakage [2] [3].
A novel process, which was buffing with Nitric acid (HNO3) and 1H-benzotriazole (1H-BTA,
C6H4N3H), could remove colloidal silica abrasives from Cu surface [1]. HNO3 would dissolve Cu oxide
C6H4N3H), could remove colloidal silica abrasives from Cu surface [1]. HNO3 would dissolve Cu oxide