Fundamental Analysis of Electrochemical Copper Deposition
for Fabrication of Submicrometer Interconnects
by
Madoka Hasegawa
Thesis submitted to Waseda University
March, 2007
Preface
Electrochemical deposition, which includes electrodeposition and electroless
deposition, is applied to fabricating micro- and nano-structured components in
electronic devices, such as interconnects in ultra large-scale integration (ULSI) and
printed circuit boards (PCB), and magnetic recording head, because it is essentially
capable of forming uniform deposits even on the complicated small structures.
Among these processes, a copper electrodeposition process for fabrication of ULSI
copper interconnects, called the “Damascene” process, is one of the most important
techniques in the semiconductor industry.
In this process, submicrometer trenches
and via-holes are filled with copper by electrodeposition, followed by chemical
mechanical polishing.
The electrodeposition process achieves the so-called
“superfilling”, which signifies filling of trenches and via-holes with copper without
producing voids.
Superfilling is achieved by small amounts of compounds used in
the plating bath as additives.
These additives allow the so-called “bottom-up” growth
to take place, which is the growth of copper deposit in the upward direction from the
bottom of trenches, resulting in superfilling.
In spite of the successful use of the
process in the interconnect fabrication technology, the behavior of these additives is
still incompletely understood because of the complicated chemistry of additives.
Therefore, it is of particular important to investigate the reaction of additives in the
electrodeposition for achieving the further improvement of this process.
In this thesis, electrodeposition for the fabrication of copper interconnects is
described.
process.
On the other hand, electroless copper deposition was investigated as a new
The author focused these studies on the effect of additives in these systems
for the mechanistic understanding of the electrochemical process.
This thesis consists of 6 chapters;
In Chapter 1, the background of this research is introduced.
The process of
the Damascene process, theoretical aspect of electrochemical deposition, and the recent
progress in the related field of research are discussed.
In Chapter 2, the effect of additives used in the Damascene electrodeposition
process was investigated based on electrochemical measurements, and the mechanism
i
of the superfilling is discussed.
In Chapter 3, the effect of additives on physical and mechanical properties of
copper electrodeposits, which are important factors determining property of
interconnects, is described.
The study was focused on the effect of additives on the
room-temperature recrystallization, or the “self-annealing”, which is known to occur in
copper electrodeposits.
Furthermore, the effect of this phenomenon on the ductility
of the deposits, which is one of the most important physical properties of the film, was
investigated.
In Chapter 4, the electrochemical copper deposition was investigated for
achieving superfilling of trenches.
Superfilling was achieved by some additives.
Moreover, the effect of other bath constituents on the performance of the additives was
investigated.
In Chapter 5, the mechanism of the superfilling observed in Chapter 4 was
investigated.
In this study, electrochemical measurements were performed to clarify
the additive effect on electroless copper deposition.
To derive the mechanistic
understanding of additive effect, the effect of the difference of the concentration of the
additive on the deposition potential, the author developed the potential measurement
apparatus.
Based on the electrochemical analysis described above, the mechanism of
superfilling was discussed.
In Chapter 6, the results obtained in the above mentioned studies were
summarized, and the future prospects of electrochemical deposition process for the
field of micro- and nano-fabrication is discussed.
ii
Contents:
Preface ......................................................................................... i
Chapter 1 General Introduction ............................................ 1
1.1 Electrodeposition process for the fabrication of copper interconnects 3
1.2 Electrochemical deposition process: Fundamental theory.................. 9
1.2.1 Electrochemical crystallization ...................................................................... 9
1.2.2 Electrodeposition......................................................................................... 14
1.2.3 Electroless Deposition ................................................................................. 15
1.2.4 Effect of additives........................................................................................ 18
1.3 Electrodeposition for fabrication of copper interconnects ................ 20
1.3.1 Copper elecrodeposition for trench-filling ................................................... 20
1.3.2 Room-temperature recrystallization of copper electrodeposits ..................... 28
1.4 Electrochemical processes for fabrication of next generation copper
interconnects ......................................................................................... 31
1.4.1 Copper electroless deposition for trench filling ............................................ 31
1.4.2 Electrochemical process for seed layer and barrier layer formation .............. 33
1.4.3 Other deposition techniques for fabrication of interconnection..................... 37
1.5 Summary ......................................................................................... 39
References............................................................................................. 40
iii
Chapter 2 An electrochemical study on the effect of bath
additives on copper electrodeposition in submicrometer
trenches..................................................................................... 47
2.1 Introduction ..................................................................................... 49
2.2 Experimental ................................................................................... 50
2.3 Effect of Cl– and PEG...................................................................... 52
2.3.1 Polarization measurements .......................................................................... 52
2.3.2 Observation of trench filling ........................................................................ 56
2.3.3 Effect of Cl and PEG on copper deposition in trenches............................... 58
2.4. Effect of SPS .................................................................................. 59
2.4.1 Polarization measurements .......................................................................... 59
2.4.2 Observation of trench filling ........................................................................ 62
2.4.3 Effect of SPS on copper deposition in trenches ............................................ 64
2.5 Effect of JGB................................................................................... 66
2.5.1 Polarization measurements .......................................................................... 66
2.5.2 Observation of trench filling ........................................................................ 69
2.5.3 Effect of JGB on copper deposition in trenches............................................ 71
2.6 Conclusions ..................................................................................... 73
References............................................................................................. 74
iv
Chapter 3 Effect of bath additives on the physical and
mechanical properties of copper electrodeposits ................... 77
3.1 Effect of bath additives on room-temperature recrystallizaiton of
copper electrodeposits ........................................................................... 79
3.1.1 Introduction................................................................................................. 79
3.1.2 Experimental ............................................................................................... 80
3.1.3 Resistivity measurements ............................................................................ 82
3.1.4 Crystallographic analysis ............................................................................. 84
3.1.5 Microstructure analysis................................................................................ 88
3.1.6 Carbon microanalysis .................................................................................. 94
3.1.7 Residual stress measurements ...................................................................... 96
3.1.8 Conclusions................................................................................................. 98
3.2 Effect of room-temperature recrystallization on ductility of copper
electrodeposits....................................................................................... 99
3.2.1 Introduction................................................................................................. 99
3.2.2 Experimental ............................................................................................. 100
3.2.3 Grain size measurements ........................................................................... 101
3.2.4 Ductility measurements ............................................................................. 103
3.2.5 Conclusions............................................................................................... 104
References........................................................................................... 105
v
Chapter 4 Void-free filling of submicrometer trenches by
electroless copper deposition..................................................107
4.1 Void-free filling of trenches with electroless copper deposits using a
combination of accelerating and inhibiting bath additives ................... 109
4.1.1 Introduction............................................................................................... 109
4.1.2 Experimental ...............................................................................111
4.1.3. Observation of trench filling ..................................................................... 113
4.1.4. Deposition rate measurements .................................................................. 118
4.1.5 Conclusions............................................................................................... 120
4.2 Void-free filling of trenches with electroless copper deposits using
glyoxylic acid as the reducing agent .................................................... 121
4.2.1 Introduction............................................................................................... 121
4.2.2 Experimental ............................................................................................. 123
4.2.3 Observation of trench filling ...................................................................... 125
4.2.4 Deposition rate measurements ................................................................... 132
4.2.5 Conclusions............................................................................................... 134
References........................................................................................... 135
vi
Chapter 5 An electrochemical study on electroless deposition
for void-free filling of submicrometer trenches ....................137
5.1 Introduction ................................................................................... 139
5.2 Experimental ................................................................................. 140
5.3 Polarization measurements ............................................................ 143
5.3 Deposition potential measurements ............................................... 146
5.4 Conclusions ................................................................................... 153
References........................................................................................... 154
Chapter 6 General conclusion.............................................155
List of achievements................................................................163
Acknowledgements .................................................................167
vii
Chapter 1 General Introduction
Chapter 1
1.1 Electrodeposition process for the fabrication of copper interconnects
Copper is widely used as interconnecting material in ultra-large scale
integration (ULSI) circuits.
Fabrication of Copper interconnection have been
achieved by the “Damascene process” [1], which is an electrodeposition process
combined with chemical mechanical polishing (CMP) (Fig. 1.1 [2]). The Damascene
process led to a remarkable change in the industry.
Most of manufacturers have now
converted to this electrodeposited copper interconnect technology.
Before the
introduction of copper to interconnects, aluminum and aluminum-copper alloy were
used as the interconnecting materials for many years. Aluminum interconnect layers
are easily fabricated by subtractive etching process (Fig. 1.2a).
In this process, the
interconnect layers are deposited by physical vapor deposition, followed by reactive
ion etching (RIE).
Aluminum is preferable as interconnects because this material
does not diffuse into SiO2 substrate and the layers adhere well to the substrate.
However, resistivity of aluminum is relatively high (2.65
cm), resulting in an
increase in resistive capacitive delay, or RC delay. The increase in RC delay hinders
the further increasing of speed in microelectronic integrated circuits.
Furthermore,
the layers suffer from the problem of electromigration, which is the gradual
displacement of the metal atoms of a conductor by the current flow.
The deterioration
of electromigration resistance leads to the electrical failure of interconnects.
With an
increase in interconnects density and a decease in the dimensions of interconnects, the
increase in RC delay and the electromigraion issue became much more critical.
Therefore, many researchers have been investigated new interconnect materials for
achieving further miniaturization of semiconductor devices.
They mainly focused on
the minimizing the problems of RC delay and electromigraion [3-10].
Copper was
expected to be a potential candidate for the interconnect material because of its lower
resistivity (1.68
cm) compared with that of aluminum (2.65
cm). The other
important advantage of copper interconnects is that copper offers much better
electromigration resistance than that of aluminum.
For similar dimension, the
time-to-failure of copper interconnect was about a hundred times as long as that of
aluminum interconnects [11-13].
Therefore, copper interconnection is able to support
higher current density to accelerate further miniaturization of interconnects.
3
Chapter 1
Figure 1.1. Cross sectional image of a 6-level copper interconnection structure
fabricated by IBM [2].
However, the application of copper to interconnection was not accomplished
for many years because the traditional process for aluminum metallization was not
applicable to copper.
As described above, aluminum interconnects was fabricated by
the subtractive etching process, in which reactive ion etching is used for patterning of
aluminum layer.
Reactive ion etching is extremely difficult to be applied to copper
interconnect fabrication because copper compound formed during the etching process
is not volatile.
Although technologies for direct subtractive etching of copper exist,
these technologies require more than 210 ºC [14-17], which is higher than the
application limit of most organic photoresist.
more expensive.
These copper etch technologies are also
Therefore, the application of copper to interconnect required a new
approach.
In 1997, IBM succeeded to apply cooper to fully integrated interconnection
by employing electrodeposition [1,11].
This process named the “Damascene process”
after the ancient inlay technique developed in the city of Damascus as early as 12th
century.
In this ancient metal work, decorative metals such as gold, silver, or copper
were hammered into grooves made in armaments or other metal objects after which the
excess material was removed by polishing.
For the interconnection fabrication, the
process begins with deposition of SiO2 or other dielectric material, followed by the
4
Chapter 1
etching of trench line or via-holes into the layer (Fig. 1.2).
Subsequently, a barrier
layer, which is essential for copper interconnects to prevent copper from diffusing into
substrate, is formed on the trenches and via-holes by dry process such as sputtering or
chemical vapor deposition.
For copper interconnects, TaN or TiN is usually
employed as a barrier material.
After the formation of a barrier layer, the trenches
and vias are completely filled with copper.
After the copper filling process, the
excessive metal deposited outside of the trenches and the vias was removed by using
chemical-mechanical polishing (CMP).
The process, in which only trenches and vias
are fabricated individual process, called “single Damascene” process (Fig. 1.3a).
On
the other hand, in “dual Damascene” process (Fig. 1.3b), two insulator levels are
patterned, filled, and planarized.
In this process, two levels of insulator and etch stop
layers are patterned as trenches and vias.
simultaneously by electrodeposition.
These trenches and vias are filled
Thus, this process reduces the number of
process step, resulting in lower cost, less requirements for equipment and space, and
potentially fewer defects.
5
Chapter 1
Figure 1.2. Comparison of subtractive etching process and damascene process.
6
Chapter 1
Figure 1.3. Cu damascene process flow options.
One of the most important issues for the Damascene process was gap-filling
of trenches and vias without producing voids, which affect a capacitance per unit
length and resistivity.
Therefore, complete filling of trenches or vias with copper
have been investigated extensively by employing several deposition techniques which
included physical vapor deposition (PVD), chemical vapor deposition (CVD),
electrodeposition,
and
electroless
deposition.
Among
these
techniques,
electrodeposition was employed for copper filling of the Damascene structure because
7
Chapter 1
this process is capable of successful filling of trenches and vias.
Furthermore, high
quality copper deposits were obtained by electrodeposition at room temperature, with
high throughput.
In the Damascene copper plating process, first, the seed layer is uniformly
deposited on the trench- and via-patterned insulator layer coated by a barrier layer.
After formation of a seed layer, copper electrodeposition is carried out to fill the
trenches and vias formed on the insulator layer.
after electrodeposition.
A copper layer covers entire surface
Therefore, the excess deposits must be removed by
planarization step such as chemical mechanical polishing (CMP).
Possible profile evolution in damascene copper plating is classified into three
types, namely, subconformal filling, conformal filling, and superfilling [1].
Figure
1.4 shows schematic illustration of these three types of profile evolution with time of
copper deposits.
Subconformal filling results from preferential flow of current at
trench openings, which is caused by geometric effect of electric field or substantial
depletion of the cupric ions inside the feature.
In subconformal deposition, copper is
deposited faster at trenche opening than at the bottom, which results in formation of
voids.
In conformal filling, copper is uniformly deposited at all point of a feature,
resulting in formation of a seam.
trenches or vias.
Superfilling is a void-free and seamless filling of
To achieve superfilling, deposition rate at the bottom is required to
be higher than those at other portions of the feature.
is called “bottom-up” growth.
using several additives.
This mode of growth of copper
The bottom-up growth of copper accomplished by
The bottom-up growth has been achieved by employing
several additives in the plating bath.
The effect of additives in the damascene copper
electrodeposition is very complicated.
These additives provide different effects on
the deposition reactions, and they also interact with each other.
The combined
addition of these additives is essential for achieving successful filling of trenches with
copper.
The purpose of the study in this thesis is to understand the effect of additives
on electrochemical deposition, and to control the feature and properties of the deposits.
The study also aims to develop new process for copper interconnects fabrication,
utilizing additive effect.
8
Chapter 1
Figure 1.4. Types of profile evolution during copper filling by electrodeposition: (a)
subconformal deposition, (b) conformal deposition, (c) superfilling.
In this chapter, fundamental aspect of electrochemical deposition process and
recent progress in the electrochemical approaches for fabrication of ULSI copper
interconnects are discussed.
1.2 Electrochemical deposition process: Fundamental theory
1.2.1 Electrochemical crystallization
Following sections, fundamental theory of electrochemical deposition of
metal is discussed.
Electrochemical deposition is the process in which metal ions are
reduced to metal by electron supplied from cathodic current (electrodeposition) or
from a reducing agent (electroless deposition).
9
The reaction occurs as a result of
Chapter 1
electron transfer at the interface between the electrode and the solution.
The deposition of metal by electrochemical deposition is much more
complicated than that by physical vapor deposition because of the presence of the
electron transfer reaction.
Furthermore, the adsorbed molecules and ions, present at
the surface of the electrode, affect the nucleation and crystal growth.
In addition to
these factors, the very high electric field applied at the surface during the deposition
should be considered in the case of electrodeosition.
In this section, the fundamental
theory of crystallization process of electrodeposition is described as an introduction.
In electrodeposition, the crystallization process is discussed as a function of the
overpotential on the surface of the electrode.
In the case of electroless deposition, the
fundamental concept of crystallization process is identical to that of electrodeposition.
In electrodeposition, crystallization is considered to occur through the
following 3 steps:
(1) the diffusion of metal ions from the bulk solution to the surface of the electrode;
(2) the reduction of metal ions into metal adatom (MZ+
Mad);
(3) the surface diffusion and crystallization of Mad (Mad
M)
In the first stage of crystallization is nucleation.
Experimental works have
shown that the rate and density of nucleation depends strongly on the deposition
current, overpotential, and bath additives.
If the initial crystal is perfect, no
deposition occurs under the critical overpotential for two-dimensional nucleation.
If
the initial crystal is not perfect and containing defects such as dislocation, and if these
defect sites are not occupied by impurity, crystal may grow at the kink site in the
dislocation.
Figure 1.5 shows the schematic illustration of the mechanism of
deposition [18].
surface (a).
In the initial stage, ions diffuse from the bulk of the solution to the
Then the ions are reduced to metal adatoms by electron transfer (b).
These adatoms diffuse along step to kink site in screw dislocation (c).
If the kink site
is not active and/or the overpotential is high enough for two-dimensional nucleation to
occur, the deposition also occurs at the kink site at surface nuclei.
10
Chapter 1
Figure 1.5. Schematic illustration of the deposition process [18].
After an eventual three-dimensional nucleation, many crystals develop at the
same time, resulting in formation of polycrystalline deposits, as is seen in every
industrial electrodeposit.
Lateral growth of the crystals may be stopped either
because of a lack of local current density or the presence of the next crystal.
Fischer
[19] proposed five main growth types of polycrystalline electrodeposits:
(1) field-oriented isolated crystals (FI) type;
(2) basis-oriented reproduction (BR) type;
(3) twining intermediate (Z) type;
(4) field-oriented texture (FT) type;
(5) unoriented dispersion (UD) type;
The FI type is usually observed at low inhibition. With an increase in current
density, whiskers, prismatic crystals, dendrites, and finally, powder deposits are
obtained.
density.
The BR type deposit is observed at moderate inhibition and/or current
Longer time of deposition rate, surface roughness increases, and the BR type
deposit can be degraded to the FI type.
The Z type deposit is considered to be the
intermediate type of the BR and the FT.
The FT type deposit is produced under fairly
strong inhibition and/or higher current density.
11
A large number of elongated crystals
Chapter 1
perpendicular to the substrate, forming a coherent deposit.
The UD type deposits is
obtained under the condition of much stronger inhibition and/or higher current density.
In the UD type deposit, a large number of small crystals are present and aggregate.
The structure of the deposits is very important factor controlling the physical
and mechanical properties.
In the chapter 3, the relationship between the
microstructure of the deposits and the behavior of the room-temperature
recrystallization of copper electrodeposits will be discussed.
12
Chapter 1
Figure 1.6. Diagram showing different types of polycrystalline deposits as the function
of current density and inhibiting intensity [18,20].
13
Chapter 1
1.2.2 Electrodeposition
During electrodeposition process, metal ions (Mz+) are reduced to metal state
(M0) by receiving electrons supplied from the cathodic current at the surface of the
electrode.
Mz+ + ze-
M0
(1.1)
The deposition rate is expressed in terms of the deposition current density, i, by [21]:
i
zF
V
(1.2)
where V, and F are the atomic volume of the deposting metal and Faraday’s constant,
respectively.
Under typical electrodeposition conditions, the relationship between the
deposition current (i) and the verpotential ( ) may be described by the Butler-Volmer
equation including cupric ion consumption [22]:
i
i0
Cx 0
exp
C bulk
zF
RT
(1.3)
The overpotential, , defines the deviation from equilibrium, the current density, i0,
describes the dynamic exchange current density that characterizes equilibrium. Cbulk
and Cx=0 is the concentrations of copper ions in the bulk solution and at the interface
between the electrode and the solution, respectively.
At high overpotentials, which are relevant to the potentials used in the
damascene copper electrodeposition, significant depletion of cupric ions is assumed.
In this case, Cx=0 may be expressed as [23]:
Cx 0
C bulk
1
i
iL
(1.4)
where iL is the limiting current density of the deposition.
14
Thus, the deposition current
Chapter 1
may be described as:
i
i0 1
i
exp
iL
zF
RT
(1.5)
At much higher potentials, the deposition current density is identical to the limiting
current density and described as:
i
iL
zFD0C bulk
(1.6)
where D0 is the diffusion coefficient of ions, and
the electrode.
is thickness of the diffusion layer at
The equation discussed above did not contain the effect of additives.
The
deposition current is affected by additives usually used in the solution to control the
feature and physical mechanical properties of the deposits.
a key technology in the additives.
The use of additive is also
They adsorb on the surface and/or form complex
with metal ions, which result in the change in the mass transfer of metal ions and in the
rate of electron transfer.
The functions of additives are discussed later.
1.2.3 Electroless Deposition
The term “electroless deposition” was originally adopted by Brenner and
Riddell [24] to describe a method of electrochemical deposition of metals without
external current source.
They invented a process of autocatalic electroless deposition
of nickel, in which the deposited metal itself acts as a catalyst for further deposition of
the metal.
In general, electroless deposition is characterized by the selective
reduction of metal ions only at the surface of a catalytic substrate.
Electroless
deposition is categorized into three types, namely, autocatalytic process, galvanic
displacement, and substrate-catalyzed process.
different from the latter two processes.
The first process is fundamentally
In the first process, thickness of the deposit
continues to increase because the deposited metal itself possesses catalytic activity.
On the other hand, in the case of the latter two processes, the deposition stops when the
substrate is covered completely with deposited metal.
The electroless deposition process used in Chapters 4 and 5 was autocatalytic
process.
In this process, metal ions receive electron from a reducing agent, resulting
15
Chapter 1
in the metal deposition.
solution.
Table 1.1 lists the components usually contained in the
Major constituents of the solution are metal ions, the reducing agent, the
complexing agents, and the pH adjusting agent.
Similar to electrodeposition, small
amount of additives are employed in electroless deposition. These additives improve
the physical and mechanical properties of the deposits and the stability of the solution.
Table 1.1. Bath composition
Component
Purpose
Metal salt
Source of metal deposit
Reducing agent
Reduction of metal ion
Complexing agent
To form metal complex
pH adjusting agent
pH adjustment
Additives
To improve bath stability
or deposit properties
Electroless deposition is usually explained by the “mixed potential theory”.
According to this theory, electroless deposition consists of cathodic and anodic partial
reactions, which occur simultaneously on the surface of the substrate.
Overall reaction and cathodic and anodic reactions are expressed as:
Overall reaction:
MLmn+ + Red
M + mL + Oxn+
(1.7)
Cathodic reaction:
MLmn+ + ne-
M + mL
(1.8)
Anodic reaction:
Red
Oxn+ + ne-
(1.9)
The two partial reactions (1.8) and (1.9) determine the potential of
electroless deposition, called mixed potential.
At the mixed potential, the cathodic
and anodic reactions occur at the same rate and net current is zero. This condition
may be described by the equation:
ic = ia = ipl
(1.10)
where ic, ia, and ipl are the cathodic current, the anodic current, and electroless plating
current at the mixed potential, respectively.
16
Chapter 1
The theory is usually applied to analyze electroless deposition processes
[25,26].
Figure 1.7 shows a schematic representation of the current–potential curves
for the mixed potential concept used to describe the principle of an electroless
deposition reaction.
If each partial reaction occurs separately, the superposition of the
curves for the two partial reactions should yield the curve for the complete electroless
bath.
Figure 1.7. Schematic diagram of the mixed potential theory [25].
More recent studies reveal that electroless deposition process is much more
complicated than represented by the simple mixed potential theory discussed above.
Copper electroless deposition system, which was employed in Chapters 4 and 5, is one
of the examples.
The details are investigated and discussed in Chapter 5.
17
Chapter 1
1.2.4 Effect of additives
The plating bath usually contains several additives.
The additives affect
nucleation and crystal growth, resulting in the significant change in the deposition rate
and the morphology and microstructure of the deposits.
The Damascene copper
deposition also employed several additives to control the filling feature of copper
deposits in trenches.
The conventional additive package is a deposition inhibitor, a
deposition accelerator, and a leveler, and chloride ions.
These additives are known to adsorb physically or chemically on the surface
of the electrode and form an adsorption layer, which changes the overpotential of the
deposition.
Some additives are also known to form a complex with metal ions, and
change the overpotential of the reaction.
Four main effects of additives on the
cathodic process may be considered for the additives in electrodeposition process [27]:
(1) Poisoning of the surface (thus, induce the reaction overpotential)
(2) Simultaneous reduction of additives, which decrease net current of the reduction of
metal ions
(3) Change in the microstructure and crystallographic texture
(4) Change in the value of the overpotential of the deposition such as the charge
transfer overpotential, the diffusion overpotential, the reaction overpotential, the
crystallization overpotential, and the true resistance overpotential.
Winand discussed the mechanism of additive effects on electrodeposition as
follows [18]:
Organic additives
The organic additives in the solution have various effects:
(1) if they adsorb at the surface of metal and have no affinity for water, they act as a
strong inhibitors;
(2) if they adsorb at the surface of metal and have affinity for water, they slightly
inhibit the deposition or activate the deposition;
(3) if they do not adsorb at the surface of metal and have affinity for water, they
slightly accelerate the deposition;
(4) if they do not adsorb at the surface of metal and have no affinity for water, they
have no effect.
Inorganic additives
Inorganic anions can change the structure of the double layer and, thus, influence the
charge transfer overpotential.
The activation anions are: Cl-, Br-, I-; Intermediate ions
are: NO3-, SO42-; and the inhibiting anions are: BF4-, NH2SO22-, ClO4-.
18
Chapter 1
The reaction mechanism of additives in electroless deposition process is
more complicated.
In this case, the additives are known to affect the anodic oxidation
of the reducing agent as well as the cathodic reduction of metal ions.
Inhibiting
additives is considered to adsorb on the surface of metal and block the anodic and/or
cathodic partial reaction.
On the other hand, most accelerating additives in electroless
deposition system are reported to enhance the anodic reaction, while it does not affect
the cathodic partial reaction.
These additives are considered to form a complex with
metal ions on the metal, resulting in the removal of the oxide or hydroxide on the
surface.
The effect of additives in both electrodeposition and electroless deposition
system is known to change when they are present together.
caused by the interactions of additives on the surface.
additive are very important in our research.
This is considered to be
The effects of interaction of
For example, polyethylene glycol in the
electroplating bath acts as a strong inhibitor in the presence of Cl-, while it slightly
inhibits copper deposition when it added alone.
As discussed in the following
chapters, the coexistence of additives affects the trench-filling, as well as the
microstructure and crystallographic of the deposits, resulting in the change in the
various properties of films.
It is of particular importance to investigate the
fundamental mechanism of additive interaction from both fundamental and practical
stand points.
19
Chapter 1
1.3 Electrodeposition for fabrication of copper interconnects
1.3.1 Copper elecrodeposition for trench-filling
The Damascene copper electrodeposition and CMP were the key
technologies which led to the implementation of copper interconnects.
Most
important requirement for success of this process is its ability to fill trenches and vias
completely without producing any defect such as void and seams.
Voids (or seams)
are undesirable because they lead to an increase in resistivity, and enhance
electromigration issues.
corrosion of copper.
Furthermore, trapped electrolyte in voids results in the
Void-free filling was accomplished by using several bath
additives in combination.
In this section, the fundamental chemistry of
electrochemical copper deposition and the effect of additives in the Damscene copper
electrodeposition are described.
Copper deposition mechanism and kinetics
Reduction of copper deposition on the cathode is considered to proceed by a
two charge-transfer step reaction [28-32]:
Cu2+ + e+
-
Cu + e
Cu+
0
Cu
E0 = 0.153 V (vs. NHE)
(1.11)
E0 = 0.52 V
(1.12)
The first reaction (the reduction of Cu2+ to Cu+) is rate determinig step
[28,32], because the rate constant of the second step may be up to three orders of
magnitude greater.
In the conventional Damascene electrodeposition bath, the
activity of cuprous ions is considered to be affected.
Chemistry of copper plating bath
An acidic copper sulfate acidic bath is employed in the damascene process
because this bath is stable, inexpensive, non-toxic, and readily available.
This bath
contains copper sulfate as the source of copper ions and sulfuric acid as the electrolyte
to provide conductivity to the solution.
This bath also contains several additives in
combination to achieve void-free filling, or “superfilling” of trenches and vias.
20
Chapter 1
The typical copper sulfate concentration in commercial use for the
interconnection fabrication is in the range from 0.2 to 0.6 mol/L.
High concentration
of cupric ions is useful when a rapid deposition is required to achieve desired
throughput.
On the other hand, low concentration of copper ions is effective to
improve uniformity of copper deposits across the wafer because in this concentration
range, charge transfer resistance is increased uniformly at all points on the wafer
surface. [33,34]
For the interconnect applications, copper sulfate concentration is
required to be sufficiently high enough to avoid the depletion of cupric ions within
high aspect-ratio features with an appropriate deposition rate.
Sulfuric acid is usually added to the plating bath to increase the conductivity
of the solution [33].
More conductive solutions minimize variation of potential
gradients within the plating bath, and thus, results in more uniform and geometry
independent interfacial kinetics [35].
Therefore, higher concentration of sulfuric acid
is preferred for the process of interconnection fabrication.
However, the solubility of
copper sulfate decreased with an increase in sulfuric acid concentration.
For example,
the maximal solubility of copper sulfate is about 0.75 mol/L in the presence of 2 mol/L
of sulfuric acid, while the value decrease to about 0.5 mol/L in the solution containing
sulfuric acid at concentration of 4 mol/L.
Therefore, the concentration of sulfuric
acid is optimized to increase conductivity of the solution while maintain the sufficient
concentration of cupric ions.
The most important issue for damascene copper filling is to achieve the
void-free filling of copper in trenches.
In ordinary deposition, voids or seams are
formed particularly in narrow and small trenches and vias because the deposition rate
at the openings tends to be higher than other portions of trenches in the absence of
additives.
This is caused by the current distribution occurred on trenches, in which
current density is concentrated at the opening of trenches for the geometrical reason.
[36]
21
Chapter 1
Figure 1.8. Current density profiles above and inside a feature [36].
Additives are known to adsorb on the cathode surface and affect the rate of
copper deposition as well as the microstructure and morphology of copper deposits.
In the damascene copper electrodeposition, void-free copper filling, or superfilling was
achieved by employing several organic bath additives in combination.
Based on their
function, these additives are categorized into three types: a deposition accelerator, a
deposition inhibitor, and a leveler [33,37,38].
The accelerator is essentially an
organic sulfur-containing compound, while the suppressor is a glycol, which decreases
the rate of copper deposition especially in the presence of chloride ions.
is generally a kind of an azo dye compound.
as leveler.
22
The leveler
This additive acts as an inhibitor as well
Chapter 1
Electrochemistry of additive reaction
A conventional additives used to accomplish “superfilling” are chloride ions
–
(Cl ),
polyethylene
glycol
(PEG),
bis(3-sulfopropyl)disulfide
(SPS)
or
3-mercapto-1-propanesulfonate (MPS), and Janus Green B (JGB).
Figure 1.9. Conventional additives used in the process of trench- filling by
electrodeposition: (a) polyethylene glycol (PEG), (b) bis(3-sulfopropyl)disulfide (SPS)
(c) 3-mercapto-1-propanesulfonate (MPS), and (d) Janus Green B (JGB)
.
The addition of chloride ions, which catalyzing the rate-determining
2+
+
Cu /Cu reaction, is known to accelerate copper deposition [31,39-42].
However, it
is also known to interact with PEG [31,43] and SPS when they are present together.
23
Chapter 1
PEG is a kind of surfactant, and is usually added to the Damascene plating
bath as a deposition inhibitor.
This additive is known to inhibit copper deposition
significantly in the presence of chloride ions, while PEG itself is a weak inhibitor.
This compound adsorbes on metal surface, and forms a polymeric PEG-Cl adlayer in
the presence of chloride ions [31,41,43]. The PEG-Cl adlayer is considred to inhibit
copper deposition strongly. Recent studies suggest that this adlayer also contains
cuprous ions.
Yokoi et al. [31] reported that this blocking layer might consist of a
complex of cuprous ion and PEG formed on specifically adsorbed chloride ions.
This
layer is considered to act as a blocking layer, and to inhibit copper deposition.
Compared with the knowledge of the effect of suppressor, the mechanism of
the accelerator such as SPS or its monomeric compound, MPS, is unclear.
Both of
them are reported to act as a strong accelerator in the presence of Cl
. For
trench filling in the Damascene process, SPS and MPS exhibit similar effects on
bottom-up growth of copper in trenches [45]. Thus, it is assumed that disulfide
linkage in SPS is cleaved in the conditions where copper electrodeposition takes place.
It is also indicated that cuprous ions and SPS form an adsorbed species and the
complex formed by this reaction stabilizes the cuprous ions [46].
2+
reduction of Cu
Because the
+
to Cu is the rate limiting step of copper electrodeposition, the
stabilization of cuprous ions leads to an increase in cuprous ions near the surface of
copper, which results in the increase in the rate of copper deposition.
Another model
for the catalytic effect of the accelerator is that SPS displaces the PEG-Cl layer, which
inhibits copper electrodeposition, and this competitive adsorption between an
accelerator and an inhibitor results in the increase in the rate of copper deposition [23].
However, the detailed reactions of these additives still remain to be unclear because of
the complexity of the interaction of additives.
The possible reactions which are
assumed to take place in the plating bath are shown in Table 1.2 [46].
regarded as a leveling agent [37,47,48].
inhibits the copper deposition.
JGB is
This additive increases the overpotential and
However, as far as the author is aware, the detailed
mechanism of the reaction of this additive has not been reported yet.
24
Chapter 1
Table 1.2. Reactions at the copper/electrolyte interface in copper plating baths
containing Cl, PEG, and SPS or MPS [46].*
*
The deprotonated MPS thiol group is indicated as ‘‘thiolate’’ in the formula.
25
Chapter 1
Mechanism of void-free filling-Effect of additives
For the application to submicrometer interconnects, it has been reported
[23,37] that the “bottom-up” growth, in which the deposition from the bottom of
trenches was accelerated relative to that in the areas of openings of the trenches, has
been shown to occur in the bath containing Cl–, PEG, and SPS or MPS, in combination.
It has been suggested that the competitive adsorption between the accelerator and the
inhibitor results in the bottom-up effect [23]. The bottom-up effect achieves the
superfilling of copper in trenches, while it also brings about the “overfill” phenomenon,
in which copper bumps are formed above the copper-filled trenches [23,48].
Early electrodeposition models for bottom-up growth [1,49-51] assumed that
the deposition reaction is controlled by mass-transfer of inhibiting additives
(transport-limited leveling model) [52-54]. This model is based on the model for
surface leveling of the electrodeposits.
The model considered based on this
traditional transport-limiting leveling model considered the effect of inhibitor and its
mass-transfer, ignoring the effect of the accelerator.
-3
Because the concentration of
-6
additive is very low (10 -10 mol/L), the adsorption of the additive is controlled by its
diffusion from the bulk of the solution into trenches.
These models predict the
decrease in the deposition rate at the opening of trenches by considering the diffusion
and consumption of inhibitor during electrodeposition.
However, these models did
not consistent with the experimental observations such as the overfill phenomenon, and
the shape evolution of copper during filling [23,55,56]. This indicated that traditional
transport-limited leveling model is not suitable for explaining the mechanism of
superfilling.
Moffat et al. [21,57]
and West et al. [58]have independently proposed
models, in which the coverage of accelerator at the bottom of trenches is assumed to
increase during the filling of trenches.
In their models, the simple competitive
adsorption between the inhibitor and the accelerator is considered.
The model
proposed by Moffat et al. [21,57]assumes that variation of the accelerator surface
coverage depends on local surface curvature, while that proposed by West et al.
[58]considered that surface site density changes during the deposition.
Their models
predict not only the superfilling but also the “overfill” phenomenon, explaining
successfully the fundamental aspects of the bottom-up growth, for instance, the
presence of an incubation period before the occurrence of the bottom-up growth
[21,42,45,55-57,59,60].
26
Chapter 1
Figure 1.10. Simulations (left) and SEM images (right) of copper deposition in
trenches with MPS 0, 0.0005, 0.005, and 0.04 mol/m3 and overpotentials: -0.097,
-0.301, -0.282, and -0.150 V [top (a) to bottom (d)], and aspect ratios: 1.8, 2.3, 3.0, 4.0,
5.6 (left to right) [59].
The overfill phenomenon itself is disadvantageous for the planarization step
(chemical-mechanical polishing, CMP), which follows the copper deposition in the
damascene process.
JGB is regarded to serve as a leveling agent, flattening these
bumps of copper deposits on the surface, or to influence the filling properties
[47,48,61].
27
Chapter 1
1.3.2 Room-temperature recrystallization of copper electrodeposits
Electrodeposited copper is known to undergo “self-annealing”, which is a
process of recrystallization occurring at room temperature.
Self-annealing leads to an
increase in the grain size by an order of magnitude, and it also brings about a decrease
in the resistivity nearly to the bulk value [62-65].
crystallographic structure [66-69], internal stress
during self-annealing.
Film properties such as
[65,70,71] also change with time
Copper deposits obtained from the bath containing the
additives for the Damascene copper electrodeposition, which include Cl-, PEG, SPS (or
MPS) [23], are also known to undergo self-annealing.
The effect of these additives
on self-annealing is also of great interest because this phenomenon affect the various
film properties which are very important to maintain the performance and reliability of
electronic devices [65,72-75].
Factors inducing self-annealing which have been discussed so far include
strain energy caused by residual stress [65,69-71,74], high internal strain energy
originated from incorporated impurities [69-71,76], defect energy originated from
dislocation [62,74], and grain boundary energy associated with fine grains [77].
Lee et al. [65]reported that two major contributions to the change in energy
of a film during grain growth are the grain boundary energy and the strain energy.
If the film did not contain defect, the grain boundary energy is determined by
the grain size.
E GB
Therefore, the grain boundary energy is described as [65]:
3
L
(1.13)
GB
EGB: Grain boundary energy per unit volume
L: Grain radius
GB:
dL
dt
Grain boundary energy per unit area
m
GB
L
m
(1.14)
GB
m: Average grain boundary mobility
: Average boundary curvature
28
Chapter 1
Grain boundary energy decreases the total energy during grain growth,
because of a reduction of the grain boundary area.
Ueno et al. [77]investigated the
effect of seed layer on the rate of self-annealing and found that the rate of
recrystallization was increased with a decrease in grain size.
This result suggested
that the large energy associated with fine grains of copper deposits is a driving force
inducing self-annealing.
Contrary to the change in grain boundary energy, the strain energy increase
as grain growth proceeds.
Strain energy associated with grain growth is given by
[65,71]:
E
Eel
1
a
2
film
1
L0
1
L
2
(1.15)
E: Young’s modulus
: poisson ratio
a: Excess volume per unit area of grain boundary caused by imperfect
packing of atoms
L0: Initial grain size
L: Final grain size
where
a2 1
E
L0
1
film
1
indicates
biaxial
modulus
of
film,
while
the
term,
2
L
, is referred to the magnitude of strain.
Equation (1.15) shows that
strain energy develops in the film with the magnitude depending on the biaxial
modulus of the film, grain boundary width, and initial grain sizes.
According to Lee et al. [65], the impact of the increase in strain energy is
much smaller that that of the increase in the grain energy.
Therefore continuous grain
growth during self-annealing is energetically feasible.
Residual stress in the deposits has been investigated by many researchers.
Some of them reported that tensile stress increased as a result of self-annealing [65,78],
whereas, others reported that stress was released after deposition [69-71].
Vas’ko et
al. [64]reported that the stress changes from compressive to slightly tensile during
29
Chapter 1
self-annealing, suggested that this stress relaxation is one factor driving self-annealing.
Several researchers suggested that crystallographic defects are a factor
inducing self-annealing [62,74].
The defect energy originated by dislocation is
expressed as follows:
Ed
Gb 2 ln
R
b
(1.16)
: constant (= 0.1)
G: shear modulus
b: Burgers vector
: dislocation density
R: Dislocation loop radius
The dislocation energy was reported to be smaller that those originated from
strain and grain boundary [71].
results.
However, other researchers have reported different
Lee et al. [62]carried out TEM observation of microstructure of the copper
film and found that the rate of self-annealing was related to dislocation density in the
as-deposited films.
The incorporation of bath additives in the deposit was also supposed to be an
important factor because bulky incorporated additives were considered to increase the
local stress at grain boundaries [71].
Furthermore, the effect of interface/surface energy is also considered to be
one of the factors contributing self-annealing.
Although many researches have been
performed on self-annealing, detailed knowledge on its mechanism is still lacking.
These factors are considered to be affected by the presence of additives, which is
known to change the microstructure of the deposit.
In addition, some of them are
reported to be incorporated in the deposits during the reaction.
However, the detailed
mechanism of the effect of additive, especially for the combined effect of additives, is
still unclear.
In this thesis, to gain further information of the factors inducing
self-annealing, the effect of each additive and the combined effect of additives on
various properties of film were investigated.
30
Chapter 1
1.4 Electrochemical processes for fabrication of next generation copper
interconnects
1.4.1 Copper electroless deposition for trench filling
As is discussed in the preceding sections, electrodeposition process is highly
useful for achieving superfilling of trenches.
However, with the shrinkage of the
dimensions of electronic circuits and the increase in the size of a wafer, a few issues,
such as the need for the deposition of a uniform seed layer of sputtered copper and the
problem of uneven current distribution on the wafer are becoming much more critical
issues for this process.
The filling of trenches and via-holes by electroless copper
deposition is expected to be an effective alternative for dealing with these problems,
because electroless deposition is, in principle, capable of forming a uniform thin film
on large substrates with superior step coverage. Furthermore, electroless deposition is
advantageous because it does not require a copper seed layer.
Copper deposition mechanism
Electroless deposition solution contains metal ions, a reducing agent, a
complexing agent, and some additives such as stabilizers and inhibitors.
Formaldehyde (HCHO) and Ethylenediaminetetraacetic acid (EDTA) are widely used
as the reducing agent and the complexing agent for electroless copper deposition,
respectively.
The overall reaction of electroless copper deposition in this solution is
[79]:
[Cu(II)-EDTA]2- + 2HCHO + 4OH-
Cu0 + 2HCOO- +2H2O + H2 + EDTA4- (1.17)
The deposition rate is described as follows:
r k [C Cu 2 ]a [CHCHO]b [COH - ]c [CEDTA 4- ]d exp
E
T
(1.18)
where k is the rate constant at a given temperature, a, b, c, and d are the
reaction orders for the reactant, E is the activation energy, and T is the temperature.
The deposition rate is affected by concentration of copper ions ([CCu2+]), formaldehyde
([CHCHO]), hydroxide ions ([COH-]), and complexing agent ([CEDTA4-]).
31
Chapter 1
Electroless deposition reactions in general have been explained based on the
mixed potential theory [80], in which the oxidation of the reducing agent and the
reduction of metal ions are assumed to occur simultaneously on the substrate.
According to this theory, superposition of polarization curves for the two partial
reactions should yield the curve for the complete electroless bath.
Based on the mixed potential theory, the anodic oxidation of the reducing
agent for Cu/formaldehyde electroless deposition system is written as [81]:
2CH2(OH)2 + 2OH-
2CHCOOH + H2 + 2H2O + 2e-
(1.19)
The cathodic reaction is the reduction of the metal complex [81]:
[Cu(II)-EDTA]2- + 2e-
Cu0 + EDTA4-
(1.20)
In alkaline aqueous solution, formaldehyde hydrated as the methylene glycol.
This compound is dissociated and exists as a glycolate anion [82].
HCHO + H2O
CH2(OH)2 + OH-
CH2(OH)2
(1.21)
CH2(OH)O-
(1.22)
The glycolate anion is reported to accelerate the cathodic reaction in copper
electroless deposition system [26,81].
At low formaldehyde concentrations, diffusion
of the methylene glycol anion from the bulk of the solution to copper surface is the
rate-determining step.
In the electroless deposition bath, formaldehyde is decomposed by the
following side reaction (Cannizzaro reaction).
2HCHO + OH-
HCOO- + CH3OH
(1.23)
This reaction is undesirable because it leads to the decomposition of the bath.
The rate of this reaction increases with increases in temperature and pH.
When
temperature is more than 70 ºC, the rate of this reaction is 3-4 times as much as that
32
Chapter 1
of the anodic oxidation of electroless deposition.
Trench filling by electroless copper deposition
Electroless copper deposition for trench filling has been investigated by a
few groups of researches.
Shacham-Diamand et al. [83,84]showed that the addition
of Triton X-100 to a plating bath improves the filling property.
Superfilling has been
reported to be achievable in an electroless copper plating bath containing small
amounts of SPS, which are used in the electrodeposition bath.
Shingubara et al.
[85-87]reported that superfilling was achieved by using SPS as an inhibiting additive.
Other researchers
[88-90] have also achieved the superfilling by using
SPS-containing baths, although they reported that SPS accelerated copper deposition at
very low concentrations.
Of particular interest concerning to these studies is that the
effect of this additive is seemed to be completely different from that in the
electrodeposition system.
In the electroless deposition system, SPS added alone has
been reported to provide bottom-up growth of copper in trenches, while the
combination of SPS, PEG and Cl- are essential in the electrodeposition system.
The
fact may reflect the complexity of the bath chemistry in the electroless copper plating
bath, such as the interaction between the additive and Cu(II)-complex or the reducing
agent.
The change in the behavior of the additive is also assumed to be due to the
difference in pH of the solution of these two deposition process.
1.4.2 Electrochemical process for seed layer and barrier layer formation
Electroless deposition for seed layer formation
A conductive seed layer is essential for electrodeposition process.
In the
conventional process, a copper seed layer is deposited prior to superfilling by
electrodeposition.
Filling property in trenches is determined by the seed layer usually
formed by physical vapor deposition (PVD). The seed layer is required to be uniform
in high-aspect-ratio and narrow trenches to maintain conductivity of the surface for
superfilling by electrodeposition.
However, the copper film deposited by PVD on
high aspect-ratio trenches tends to be discontinuous or agglomerated.
Usually, the
film at the openings of narrow trenches is thicker than that at the bottom and the side
33
Chapter 1
walls (Fig. 1.11).
The increase in thickness at the trench openings leads to the
non-uniform deposition inside the trenches, resulting in the formation of voids [33,91].
To achieve conformal deposition even on high aspect-ratio trenches, alternative
deposition techniques such as electroless deposition, atomic layer deposition (ALD),
and chemical vapor deposition (CVD) have been investigated.
Figure. 1.11. Profiles of seed layer in a trench formed by PVD (a photo on the left) and
copper in a trench electrodeposited on the PVD seed layer (a photo on the right) [33].
Among these techniques, electroless deposition is promising process for forming
seed layer because it achieves the excellent step coverage on high aspect-ratio and
narrow structures fabricated even on a large substrate.
As noted in the previous
section, electroless deposition has been investigated as the trench-filling process.
In
parallel with this, electroless deposition for seed layer fabrication has also been
investigated intensively [84,92-97]. It has been demonstrated previously that a very
thin seed layer with excellent step coverage on trenches was achieved by using
electroless deposition technique.
To achieving a thin and continuous copper seed
layer by electroless deposition on a barrier layer, the development of the catalytization
process as well as the optimization of bath composition is a key issue.
Most researchers [92-97] focused their attention on the activation process to
achieve direct deposition of a uniform and thin copper seed layer on the barrier
materials such as TiN, TaN, and WN.
In recent years, displacement deposition has
been investigated for the activation of these materials [98,99].
These activation
process employs a PdCl2/HF solution or a PdCl2/HF/HNO3 solution.
In these
solutions, two galvanic half cell reactions, which include the deposition of Pd particle
and the removal of the oxide layer on the surface of substrate, take place
34
Chapter 1
simultaneously.
Furthermore, the formation of a seed layer on these barrier metal
substrates have been investigated using vacuum technologies such as ionized cluster
beam (ICB) [95,97,100]and atomic layer deposition (ALD) [101,102].
Copper seed layer is also required to possess low resistivity and the plating bath
should be stable from the practical standpoint.
Therefore, some stabilizers such as
-
2,2’-dipyridine, CN , Neocuproine, and Rodamine, and the surfactants such as
polyethylene glycol, RE-610, and Triton-X-100 have been employed to the electroless
plating bath [84].
Direct copper electrodeposition on barrier materials
One alternative to deal with the failures of a copper seed layer, which are
responsible for many integration difficulties, is elimination of a copper seed layer.
However, copper electrodeposition on conventional barrier layers such as TaN, TiN
and WN, is very difficult because of their high resistivity and their stable oxide formed
on the surface of the layer.
Ru, and other Pt-group material such as In, Os, Rh are
expected to be a candidate as alternative barrier materials [37,103-109].
Among them,
Ru is particularly promising because bulk phase Ru and Cu are immiscible.
Furthermore, the thermal and electrical conductivity of Ru are twice as much as those
of Ta.
This is preferable because barrier materials occupy larger fraction of the
cross-sectional area of the conductors with a decrease in the dimensions of
interconnects.
Seedless superfilling of copper was demonstrated on sub-100 nm
trenches with Ru barrier layers, which is deposited by physical vapor deposition and
atomic layer deposition [110].
Electroless deposition for barrier layer formation
Electroless deposition was also applied to the formation of Cu diffusion barrier
layer by Osaka and co-workers [111-114].
In this process, Pd-catalyzed
self-assembled monolayer (SAM) is applied as the adhesion/catalysis layer, which was
the key for achieving uniform thin layers with sufficient adhesion to both SiO2 and
low-k substrates.
This electroless deposition technique is advantageous because it did
not contain any dry process such as sputtering throughout the process (All-wet
chemical method).
NiB alloy, which exhibited sufficient barrier properties against
35
Chapter 1
copper diffusion into the substrate and acceptable thermal stability, was employed as a
barrier layer.
A very conformal electroless NiB film was successfully formed on
trench-patterned substrate with excellent step coverage [111].
film was sufficiently low for copper electrodeposition.
The resistivity of the
Consequently, seedless copper
superfilling was achieved on this electroless NiB barrier layer.
Figure 1.12. (a) Schemtic illustration of process for fabrication a electroless barrier
layer, (b) SEM image of electroless NiB film deposited on trench-patterned substrate,
(c) SEM image of Cu electrodeposits on the NiB barrier layer [111,113].
An all-electroless process for the deposition of both barrier layer and
interconnecting layer is very interesting as an eventual all-wet method.
Therefore, the
study on electroless copper deposition for trench filling, the theme of Chapter 4 and 5,
is also important as a preliminary step for developing such a combined electroless
process.
36
Chapter 1
1.4.3 Other deposition techniques for fabrication of interconnection
Chemical vapor deposition (CVD) for copper filling
Superfilling of copper in trenches was also achieved by iodine-catalyzed
chemical vapor deposition (CVD) [115,116].
CVD is a chemical process often used
in the semiconductor industry for the deposition of thin films of various materials. In a
typical CVD process the substrate is exposed to one or more volatile precursors, which
react and/or decompose on the substrate surface to produce the desired deposit.
Frequently, volatile byproducts are also produced, which are removed by gas flow
through the reaction chamber.
In the process for trench filling [115,116], C2H5I iodine precursor is used to
catalyze the copper deposition.
The mechanism of superfilling is explained by the
model which includes impact of changing surface area at the bottom of trenches on the
local coverage of catalyst.
It is interesting to note that this model is almost same as
the model for superfilling by electrodepositon.
The combination of supercritical fluid and CVD for superfilling of copper is
proposed by Jason et al. [117].
In this technique Cu(I) or Cu(II) organometallic
precursors in supercritical carbon dioxide are deposited.
This technique is
advantageous because of the low viscosity and high diffusivity of supercritical fluid
whose physicochemical properties lie between those of liquid and gasses.
The
difference between this technique and the conventional CVD is the mode of precursor
transfer.
In the conventional CVD, the concentration of precursors in vapor phase is
usually low because of its low volatility.
Thus, the reaction is the mass
transfer-limited, which result in poor step coverage.
On the other hand, in this
technique, precursors in supercritical CO2 solution are up to three or orders of
magnitude greater. Therefore it is capable of fabricating highly uniform film with
excellent step coverage.
supercritical fluid.
Figure 1.13 shows the filling feature of copper by using
This image suggested that this technique is also a candidate for
the next generation techniques for filling of copper in trenches.
37
Chapter 1
Figure 1.13. SEM image of trenches filled with copper deposited by using supercritical
fluid [117].
Atomic layer deposition (ALD) for seed layer formation
Atomic layer deposition (ALD) is also a potential candidate which is capable
of forming a highly uniform seed layer in trenches.
ALD is a self-limiting, sequential
surface chemistry that deposits conformal thin-films of materials onto substrates of
varying compositions.
ALD is similar in chemistry to chemical vapor
deposition (CVD), except that the ALD reaction breaks the CVD
reaction into two half-reactions, keeping the precursor materials
separate during the reaction. ALD film growth is self-limited and based
on surface reactions, which makes achieving atomic scale deposition
control possible. By keeping the precursors separate throughout the
coating process, atomic layer control of film grown can be obtained as
fine as ~ 0.1 angstroms per monolayer.
It has been reported that
[118] ALD technique enables to
deposit copper inside narrow holes with very high aspect ratio.
Figure 1.14 SEM image of copper in trenches deposited by ALD [118].
38
Chapter 1
1.5 Summary
In this chapter, fundamental aspect of electrochemical processes and recent
development of electrochemical process in the field of fabrication of ULSI
interconnects are introduced.
The damascene process have achieved successful implementation of copper
interconnects by employing electrodeposition for trench filling.
In this
electrodeposition process, the effect of additives is very important to accomplish
superfilling.
The additives also affect the film properties.
One of the most
interesting phenomenon concerning to the properties of copper electrodeposits is
self-annealing.
For the above mentioned copper electrodeposition process, some critical
issues still remain to be addressed for further development of the process.
The issues
include the ununiformity and discontinuity of PVD seed layers in trenches, resulting in
formation of voids, and the current distribution, leading to uneven thickness of copper
deposits across a wafer.
Electroless deposition technique is highly useful for
fabrication of both copper interconnecting layer and barrier layer.
Although other
techniques such as CVD and ALD can be also candidates for the fabrication of
interconnects, the author considers that this technique is more effective because this
process is a very simple process achieving very uniform deposition on complicated
nanostructures.
Moreover, the application of electroless deposition is expected to
achieve the elimination of sputtering process, in which uniformity of the deposits tends
to be poor in trenches.
The first purpose of the author’s study is to understand the effect of additives
on the electrodeposition in submicrometer trenches (Chapter 2) and on the
self-annealing behavior of electrodeposits (Chapter 3).
Furthermore, an attempt to
control the property of copper film was performed by utilizing self-annealing.
The second objective was to investigate an electroless process for fabrication
of copper interconnects (Chapter 4).
As discussed in this chapter, the chemistry of
the electroless deposition process is more complicated than electrodeposition.
Therefore, in Chapter 5, the fundamental mechanism of the effect of additive was
discussed intensively.
39
Chapter 1
References
[1] P.C. Andricacos, C. Uzoh, J.O. Dukovic, J. Horkans, and H. Deligianni, IBM J. Res.
Dev., 42, 567 (1998).
[2] P.C. Andricacos, The Electrochemical Society Interface, 8, 32 (1999).
[3] Y.I. Ismail, E.G. Friedman, and J.L. Neves, IEEE Trans. VLSI Syst., 9, 963 (2001).
[4] A. Deutsch, G.V. Kopcsay, P.W. Coteus, C.W. Surovic, P.E. Dahlen, D.L.
Heckmann, and D.W. Duan, IEEE Trans. Electromagn. Compat., 43, 446 (2001).
[5] Y.I. Ismail, E.G. Friedman, and J.L. Neves, IEEE Trans. Circuits Syst. II-Analog
Digit Signal Process, 48, 471 (2001).
[6] A. Deutsch, P.W. Coteus, G.V. Kopcsay, H.H. Smith, C.W. Surovic, B.L. Krauter,
D.C. Edelstein, and P.J. Restle, Proceed. IEEE, 89, 529 (2001).
[7] Y.I. Ismail, and E.G. Friedman, IEEE Trans. VLSI Syst., 8, 195 (2000).
[8] Y.I. Ismail, E.G. Friedman, and J.L. Neves, IEEE Trans. Comput-Aided Des. Integr.
Circuits Syst., 19, 83 (2000).
[9] T. Sakurai, IEEE J. Solid-State Circuit, 18, 418 (1983).
[10] D.C. Edelstein, G.A. Saihalasz, and Y.J. Mii, IBM J. Res. Dev., 39, 383 (1995).
[11] Edelstein D., Heidenreich J., Goldblatt R., Cote W., Uzoh C., Lustig N., Roper P.,
McDevitt T., Motsiff W., Simon A., Dukovic J., Wachnik R., Rathore H., Schulz R.,
Su L., Luce S. and Slattery J. In Technical Digest, IEEE International Electron Devices
Meeting. pp 773 (1997).
[12] S. Venkatesan, A.V. Gelatos, V. Misra, R. Islam, B. Smith, J. Cope, B. Wilson, D.
Tuttle, R. Cardwell, I. Yang, P.V. Gilbert, R. Woodruff, R. Bajaj, S. Das, J. Farkas, D.
Watts, C. King, P. Crabtree, T. Sparks, T. Lii, C. Simpson, A. Jain, M. Herrick, C.
Capasso, S. Anderson, R. Venkatraman, S. Filipiak, B. Flordalice, K. Reid, J. Klein,
E.J. Weitzman and H. Kawasaki, in Technological Digest, IEEE International Electron
Devices Meeting. p. 769 (1997).
[13] R. Rosenberg, D.C. Edelstein, C.K. Hu, and K.P. Rodbell, Annu. Rev. Mat. Sci.,
30, 229 (2000).
[14] S.K. Lee, S.S. Chun, C.Y. Hwang, and W.J. Lee, Jpn. J. Appl. Phys. [1]-Reg. Pap.
Short notes, 36, 50 (1997).
[15] B.J. Howard, and C. Steinbruchel, Appl. Phys. Lett., 59, 914 (1991).
[16] K. Ohno, M. Sato, and Y. Arita, Jpn. J. Appl. Phys. [2] Lett., 28, L1070 (1989).
[17] G.C. Schwartz, and P.M. Schaible, J. Electrochem. Soc., 130, 1777 (1983).
[18] R. Winand, Hydrometallurgy, 29, 567 (1992).
[19] H. Fischer, Electrochim. Acta, 2, 50 (1960).
40
Chapter 1
[20] R. Winand, P. VanHam, R. Colin, and D. Milojevic, J. Electrochem. Soc., 144,
428 (1997).
[21] D. Josell, D. Wheeler, W.H. Huber, and T.P. Moffat, Phys. Rev. Lett., 8701,
(2001).
[22] A. Bard and L.R. Faulkner, Electrochemical Methods, Fundamental and
Applications, John Wiley & Sons, New York, (1980).
[23] T.P. Moffat, J.E. Bonevich, W.H. Huber, A. Stanishevsky, D.R. Kelly, G.R.
Stafford, and D. Josell, J. Electrochem. Soc., 147, 4524 (2000).
[24] A. Brenner, and G.E. Riddle, J. Research, National Bur. Standards, 37, 1 (1946).
[25] M. Paunovic, Plating, 55, 1161 (1968).
[26] Okinaka Y. and Osaka T. In Advances in Electrochemical Science and
Engineering, Vol. 3. Edited by Gerischer H. and Tobias C.W. pp 87(1994).VCH
Publishers Inc., Weinheim.
[27] Fischer H. Elektrolytische Abscheidung und Elektristallisation von Metallen. pp
729(1954).Springer, Berlin.
[28] S. Krzewska, Electrochim. Acta, 42, 3531 (1997).
[29] S. Yoon, M. Schwartz, and K. Nobe, Plat Surf Finish, 81, 65 (1994).
[30] A. Jardy, A.L. Lasallemolin, M. Keddam, and H. Takenouti, Electrochim. Acta,
37, 2195 (1992).
[31] M. Yokoi, S. Konishi, and T. Hayashi, Denki Kagaku, 52, 218 (1984).
[32] M. Yokoi, S. Konishi, and T. Hayashi, Denki Kagaku, 51, 310 (1983).
[33] J. Reid, Jpn. J. Appl. Phys. [1] Reg. Pap. Short Notes, 40, 2650 (2001).
[34] K.M. Takahashi, J. Electrochem. Soc., 147, 1414 (2000).
[35] E.K. Broadbent, E.J. McInerney, L.A. Gochberg, and R.L. Jackson, J. Vac. Sci.
Technol. B, 17, 2584 (1999).
[36] S. Soukane, S. Sen, and T.S. Cale, J. Electrochem. Soc., 149, C74 (2002).
[37] J.J. Kelly, and A.C. West, Electrochem. Solid State Lett, 2, 561 (1999).
[38] L. Mirkova, S. Rashkov, and C. Nanev, Sur. Technol., 15, 181 (1982).
[39] J.J. Kelly, and A.C. West, J. Electrochem. Soc., 145, 3472 (1998).
[40] J.J. Kelly, and A.C. West, J. Electrochem. Soc., 145, 3477 (1998).
[41] D. Stoychev, and C. Tsvetanov, J. Appl. Electrochem., 26, 741 (1996).
[42] T.P. Moffat, D. Wheeler, and D. Josell, J. Electrochem. Soc., 151, C262 (2004).
[43] J.P. Healy, D. Pletcher, and M. Goodenough, J. Electroanal. Chem., 338, 155
(1992).
[44] M. Kang, M.E. Gross, and A.A. Gewirth, J. Electrochem. Soc., 150, C292 (2003).
[45] T.P. Moffat, D. Wheeler, M.D. Edelstein, and D. Josell, IBM J. Res. Dev., 49, 19
41
Chapter 1
(2005).
[46] P.M. Vereecken, R.A. Binstead, H. Deligianni, and P.C. Andricacos, IBM J. Res.
Dev., 49, 3 (2005).
[47] S. Miura, F. Oyamada, Y. Takada, and H. Honma, Electrochemistry, 69, 773
(2001).
[48] P. Taephaisitphongse, Y. Cao, and A.C. West, J. Electrochem. Soc., 148, C492
(2001).
[49] J.O. Dukovic, IBM J. Res. Dev., 37, 125 (1993).
[50] A.C. West, J. Electrochem. Soc., 147, 227 (2000).
[51] K.M. Takahashi, and M.E. Gross, J. Electrochem. Soc., 146, 4499 (1999).
[52] C. Madore, P. Agarwal, and D. Landolt, J. Electrochem. Soc., 145, 1561 (1998).
[53] C. Madore, M. Matlosz, and D. Landolt, J. Electrochem. Soc., 143, 3927 (1996).
[54] C. Madore, and D. Landolt, J. Electrochem. Soc., 143, 3936 (1996).
[55] D. Josell, B. Baker, C. Witt, D. Wheeler, and T.P. Moffat, J. Electrochem. Soc.,
149, C637 (2002).
[56] T.P. Moffat, D. Wheeler, C. Witt, and D. Josell, Electrochem. Solid State Lett, 5,
C110 (2002).
[57] T.P. Moffat, D. Wheeler, W.H. Huber, and D. Josell, Electrochem. Solid-State
Lett., 4, C26 (2001).
[58] A.C. West, S. Mayer, and J. Reid, Electrochem. Solid-State Lett., 4, C50 (2001).
[59] D. Wheeler, D. Josell, and T.P. Moffat, J. Electrochem. Soc., 150, C302 (2003).
[60] D. Josell, T.P. Moffat, and D. Wheeler, J. Electrochem. Soc., 151, C19 (2004).
[61] T. Haba, T. Itabashi, H. Akahoshi, A. Sano, K. Kobayashi, and H. Miyazaki,
Mater Trans, 43, 1593 (2002).
[62] H. Lee, W.D. Nix, and S.S. Wong, J Vac Sci Technol B, 22, 2369 (2004).
[63] V.A. Vas'ko, I. Tabakovic, S.C. Riemer, and M.T. Kief, Microelectron. Eng., 75,
71 (2004).
[64] V.A. Vas'ko, I. Tabakovic, and S.C. Riemer, Electrochem. Solid-State Lett., 6,
C100 (2003).
[65] H. Lee, S.S. Wong, and S.D. Lopatin, J. Appl. Phys., 93, 3796 (2003).
[66] K. Pantleon, J.A.D. Jensen, and M.A.J. Somers, J. Electrochem. Soc., 151, C45
(2004).
[67] M.S. Yoon, Y.J. Park, and Y.C. Joo, Thin Solid Films, 408, 230 (2002).
[68] L.T. Koh, G.Z. You, C.Y. Li, and P.D. Foo, Microelectron. J., 33, 229 (2002).
[69] S.H. Brongersma, E. Richard, I. Vervoort, H. Bender, W. Vandervorst, S.
Lagrange, G. Beyer, and K. Maex, J. Appl. Phys., 86, 3642 (1999).
42
Chapter 1
[70] S. Lagrange, S.H. Brongersma, M. Judelewicz, A. Saerens, I. Vervoort, E.
Richard, R. Palmans, and K. Maex, Microelectron. Eng., 50, 449 (2000).
[71] C.H. Lee, and C.O. Park, Jpn. J. Appl. Phys. [1] Reg. Pap. Short Notes, 42, 4484
(2003).
[72] S.H. Brongersma, E. Kerr, I. Vervoort, A. Saerens, and K. Maex, J. Mater. Res.,
17, 582 (2002).
[73] M. Hasegawa, and Y. Hirai, J. Appl. Phys., 90, 2792 (2001).
[74] C. Lingk, and M.E. Gross, J. Appl. Phys., 84, 5547 (1998).
[75] Q.T. Jiang, M.E. Gross, G. Bersuker, B. Foran, R. Mikkola, B. Carpenter, and J.
Oemando, Mat. Res. Soc. Symp. Proc., 564, 429 (1999).
[76] M. Stangl, J. Acker, V. Dittel, W. Gruner, V. Hoffmann, and K. Wetzig,
Microelectron. Eng., 82, 189 (2005).
[77] K. Ueno, T. Ritzdorf, and S. Grace, J. Appl. Phys., 86, 4930 (1999).
[78] J.M.E. Harper, C. Cabral, P.C. Andricacos, L. Gignac, I.C. Noyan, K.P. Rodbell,
and C.K. Hu, J. Appl. Phys., 86, 2516 (1999).
[79] R.M. Lukes, Plating, 51, 1066 (1964).
[80] C. Wagner, and W. Traud, Z. Elektrochem., 44, 391 (1938).
[81] B.J. Feldman, and O.R. Melroy, J. Electrochem. Soc., 136, 640 (1989).
[82] R. Schumacher, J.J. Pesek, and O.R. Melroy, J. Phys. Chem., 89, 4338 (1985).
[83] Y. Shacham-Diamand, and S. Lopatin, Microelectron. Eng., 37-8, 77 (1997).
[84] V.M. Dubin, Y. ShachamDiamand, B. Zhao, P.K. Vasudev, and C.H. Ting, J.
Electrochem. Soc., 144, 898 (1997).
[85] Z.L. Wang, O. Yaegashi, H. Sakaue, T. Takahagi, and S. Shingubara, J.
Electrochem. Soc., 151, C781 (2004).
[86] Z.L. Wang, O. Yaegashi, H. Sakaue, T. Takahagi, and S. Shingubara, Jpn. J. Appl.
Phys. [1] Reg. Pap. Short Notes, 43, 7000 (2004).
[87] S. Shingubara, Z.L. Wang, O. Yaegashi, R. Obata, H. Sakaue, and T. Takahagi,
Electrochem. Solid-State Lett., 7, C78 (2004).
[88] J.J. Kim, S.K. Kim, and Y.S. Kim, J. Electroanal. Chem., 542, 61 (2003).
[89] C.H. Lee, S.C. Lee, and J.J. Kim, Electrochem. Solid-State Lett., 8, C110 (2005).
[90] C.H. Lee, S.C. Lee, and J.J. Kim, Electrochim. Acta, 50, 3563 (2005).
[91] J. Reid, S. Mayer, E. Broadbent, E. Klawuhn, and K. Ashtiani, Solid State Technol,
43, 86 (2000).
[92] C.H. Lee, S. Hwang, S.C. Kim, and J.J. Kim, Electrochem. Solid-State Lett., 9,
C157 (2006).
[93] S.K. Kim, S.K. Cho, J.J. Kim, and Y.S. Lee, Electrochem. Solid-State Lett., 8,
43
Chapter 1
C19 (2005).
[94] Y.S. Kim, D.L. Bae, H.C. Yang, H.S. Shin, G.W. Wang, J.J. Senkevich, and T.M.
Lu, J. Electrochem. Soc., 152, C89 (2005).
[95] Z.L. Wang, H. Sakaue, S. Shingubara, and T. Takahagi, Jpn. J. Appl. Phys. [1]
Reg. Pap. Short Notes, 42, 1843 (2003).
[96] J.J. Kim, S.K. Kim, C.H. Lee, and Y.S. Kim, J. Vac. Sci. Technol. B, 21, 33
(2003).
[97] Z. Wang, T. Ida, H. Sakaue, S. Shingubara, and T. Takahagi, Electrochem.
Solid-State Lett., 6, C38 (2003).
[98] H.H. Hsu, C.C. Hsieh, M.H. Chen, S.J. Lin, and J.W. Yeh, J. Electrochem. Soc.,
148, C590 (2001).
[99] J.C. Patterson, C. Nidheasuna, J. Barrett, T.R. Spalding, M. Oreilly, X. Jiang, and
G.M. Crean, Appl. Surf. Sci., 91, 124 (1995).
[100] Z.L. Wang, O. Yaegashi, H. Sakaue, T. Takahagi, and S. Shingubara, Jpn. J.
Appl. Phys. [2] Lett., 42, L1223 (2003).
[101] Y.S. Kim, J. Shin, J.H. Cho, G.A. Ten Eyck, D.L. Liu, S. Pimanpang, T.M. Lu,
J.J. Senkevich, and H.S. Shin, Surf Coat Technol, 200, 5760 (2006).
[102] Y. Kim, G.A. Ten Eyck, D.X. Ye, C. Jezewski, T. Karabacak, H.S. Shin, J.J.
Senkevich, and T.M. Lu, J. Electrochem. Soc., 152, C376 (2005).
[103] D. Josell, C. Witt, and T.P. Moffat, Electrochem. Solid-State Lett., 9, L3 (2006).
[104] D. Josell, J.E. Bonevich, T.P. Moffat, T. Aaltonen, M. Ritala, and M. Leskela,
Electrochem. Solid-State Lett., 9, L4 (2006).
[105] T.P. Moffat, M. Walker, P.J. Chen, J.E. Bonevich, W.F. Egelhoff, L. Richter, C.
Witt, T. Aaltonen, M. Ritala, M. Leskela, and D. Josell, J. Electrochem. Soc., 153, C37
(2006).
[106] D. Josell, C. Witt, and T.P. Moffat, Electrochem. Solid-State Lett., 9, C41
(2006).
[107] D. Josell, J.E. Bonevich, T.P. Moffat, T. Aaltonen, M. Ritala, and M. Leskela,
Electrochem. Solid-State Lett., 9, C48 (2006).
[108] M. Zheng, M. Willey, and A.C. West, Electrochem. Solid-State Lett., 8, C151
(2005).
[109] A. Radisic, Y. Cao, P. Taephaisitphongse, A.C. West, and P.C. Searson, J.
Electrochem. Soc., 150, C362 (2003).
[110] D. Josell, D. Wheeler, C. Witt, and T.P. Moffat, Electrochem. Solid-State Lett., 6,
C143 (2003).
[111] M. Yoshino, Y. Nonaka, J. Sasano, I. Matsuda, Y. Shacham-Diamand, and T.
44
Chapter 1
Osaka, Electrochim. Acta, 51, 916 (2005).
[112] T. Osaka, N. Takano, and T. Yokoshima, Surf. Coat. Technol., 169, 1 (2003).
[113] T. Osaka, N. Takano, T. Kurokawa, T. Kaneko, and K. Ueno, J. Electrochem.
Soc., 149, C573 (2002).
[114] T. Osaka, N. Takano, T. Kurokawa, and K. Ueno, Electrochem. Solid-State Lett.,
5, C7 (2002).
[115] D. Josell, S. Kim, D. Wheeler, T.P. Moffat, and S.G. Pyo, J. Electrochem. Soc.,
150, C368 (2003).
[116] S.G. Pyo, S. Kim, D. Wheeler, T.P. Moffat, and D. Josell, J. Appl. Phys., 93,
1257 (2003).
[117] J.M. Blackburn, D.P. Long, A. Cavanas, and J.J. Watkins, Science, 294, 141
(2001).
[118] Z.W. Li, A. Rahtu, and R.G. Gordon, J. Electrochem. Soc., 153, C787 (2006).
45
Chapter 2
An electrochemical study on the effect of
bath additives on copper electrodeposition
in submicrometer trenches
Chapter 2
2.1 Introduction
In this chapter, the effects of the bath additives used in the Damascene
electrodeposition process [1] were investigated by means of electrochemical
polarization measurement and cross sectional microscopy.
As described in Chapter 1,
–
several additives such as chloride ions (Cl ), polyethylene glycol (PEG),
bis(3-sulfopropyl)disulfide (SPS) or 3-mercapto-1-propanesulfonate (MPS), and Janus
Green B (JGB) are included in the bath to accomplish “superfilling” [2-4].
Investigations for understanding additive effects on the copper
electrodeposition inside submicrometer trenches have been carried out by many
researchers [3-11].
However, the behavior of these additives during the copper
electrodeposition process is still incompletely understood because of the complexity
resulting from interactions between the effects of the multiple numbers of additives
[2-5].
It has been reported [6,9,10] that the “bottom-up” growth, in which the
deposition from the bottom of trenches was accelerated relative to that in the areas of
openings of the trenches, has been shown to occur in the bath containing Cl–, PEG, and
SPS or MPS.
Many reports have shown that the addition of PEG with Cl– inhibits
copper deposition [3-6,12-18], whereas SPS accelerates the deposition when it is added
together with PEG and Cl– [19,20].
It has been suggested that the competitive
adsorption of these three additives results in the bottom-up effect [6,9,21-23], which
lead to superfilling, while it also brings about the “overfill” phenomenon, in which
copper bumps are formed above the copper-filled trenches [3,6].
The effect of these
–
additives (Cl , PEG, and SPS or MPS) is explained by the model proposed by Moffat et
al. [24,25] and West et al. [9], in which the coverage of accelerator at the bottom of
trenches is assumed to increase during the filling of trenches.
On the other hand, the
behavior of JGB, particularly its interaction with other additives is not well understood.
JGB is regarded to serve as a leveling agent, flattening these bumps of copper deposits
on the surface [3,4], or to influence the filling properties [2,8].
Chapter 2 discusses the effects of Cl–, PEG, SPS, and JGB on the filling of
submicrometer trenches to derive the mechanisms for bottom-up filling and for the
inhibition of overfill phenomenon by JGB.
In order to understand the effect of JGB
on the overfill phenomenon caused by other additives, the investigation was carried out
for the bath containing Cl–, PEG and SPS (Cl-PEG-SPS bath), and the bath containing
JGB as an additional component (Cl-PEG-SPS-JGB bath).
49
Chapter 2
2.2 Experimental
In this study, the effect of additives (Cl–, PEG, SPS, and JGB) was
investigated based on both electrochemical polarization measurements and
microscopic observation of cross sections of copper-deposited trenches.
Bath compositions
The compositions of the baths used in this study are shown in Table 1.
The
basic constituents of all baths used were 0.26 mol/L CuSO4 and 2.0 mol/L H2SO4.
Chloride ions (as HCl) and polyethylene glycol (PEG, average molecular weight 2000)
were added to make their concentrations equal to 50 ppm and 100 ppm, respectively.
The concentrations of bis(3-sulfopropyl)disulfide (SPS) and Janus Green B (JGB) was
varied as indicated in Table 1.
Bath name
Additive-free
PEG
All experiments were carried out at room temperature.
Table 2.1. Bath compositions.
CuSO4 H2SO4
PEG
Cl– (HCl)
mol L-1 mol L-1 (MW 2000)
ppm
ppm
0.26
2.0
0.26
2.0
100
Cl-PEG
0.26
2.0
100
50
Cl-PEG-SPS
Cl-PEG-SPS-JGB
0.26
0.26
2.0
2.0
100
100
50
50
SPS
ppm
JGB
ppm
1-1000
1-5
1-50
Polarization measurements
Polarization measurements were performed with a rotating disk electrode
(RDE) system (RRDE-1, Nikko Keisoku) equipped with a computer-controlled
electrochemical measuring system (HZ-3000, Hokuto Denko). A Cu disk working
electrode (0.6 cm in diameter) was polished first with a #2000 emery paper and
subsequently with 0.06- m alumina paste on a sheet of polishing cloth. A Pt counter
electrode and an Ag/AgCl reference electrode were used for the measurements.
These electrodes were placed in a three-electrode cell. The RDE was rotated at a
50
Chapter 2
speed in range of 0-2500 rpm. The measurements were performed at the potential
scan rate of 2 mV s-1. Before each experiment, the bath solution was deaerated by
bubbling N2 gas for 15 minutes.
Observation of filling of trenches
For trench-filling studies, substrates with trench arrays were employed. A
Ta barrier layer and a Cu seed layer were sputtered onto a SiO2/Si substrate. The
trenches were 200 nm wide and 500 nm deep with the aspect ratio of 2.5. The space
between the trenches was approximately equal to the trench width.
The
electrodeposition of copper was carried out galvanostatically at a current density of 10
mA cm-2 using a paddle plating system (Uyemura) with or without agitation (100 rpm).
The deposition time was varied from 15 to 90 s. After the electrodeposition, the
specimen was cleaved perpendicularly to the direction of trenches. The filling feature
of copper in trenches was examined by cross-sectional observation using a
field-emission scanning electron microscope (FE-SEM, S-4800, HITACHI).
51
Chapter 2
2.3 Effect of Cl– and PEG
2.3.1 Polarization measurements
Figure 2.1 shows cathodic polarization curves for copper electrodeposition
from the additive-free, PEG (Cl–-free), and Cl-PEG baths.
The polarization curve for
the additive-free bath (Fig. 2.1a) shows that the current density gradually increased
from the beginning of the potential scan and reached the plateau at about –0.4 V vs.
Ag/AgCl.
This plateau corresponds to the mass-transfer limiting current for the
reduction of cupric ions at potentials more negative than –0.4 V.
The continued
negative scan beyond –0.6 V led to a significant increase in current density due to
hydrogen evolution.
The addition of PEG to the bath decreased the limiting current
density (Fig. 2.1b).
In the polarization curve for the Cl-PEG bath (Fig. 2.1c), current
was essentially nil at potentials less negative than –0.1 V.
As the shape and the
limiting current density of the polarization curve for the Cl-PEG bath were identical to
those for the PEG bath, it is seen that only the potential where the current began to
increase was shifted in the negative direction by the addition of Cl–.
This result
–
suggests that the addition of both Cl and PEG greatly increased the overpotential for
copper deposition, resulting in the inhibition of copper deposition.
The decrease in the current density suggests the inhibition of copper
deposition in the coexistence of Cl with PEG. As Hearly et al. [16] and Jovic et al.
[17] reported previously, the polarization characteristic for the Cl-PEG bath depended
on the concentration of Cl
as well as the molecular weight of PEG.
concentration of Cl dramatically changed the polarization curve.
The
In the bath
containing 50 ppm Cl with PEG, the current density is very small and it increased
gradually with a change in the potential in the negative direction (Fig. 2.1). On the
other hand, at the concentrations less than 10 ppm, the current density was nil at the
potentials less negative than a certain potential, which is hereafter called a critical
potential.
The current density increases abruptly at the critical potential.
The
critical potential strongly depended on the concentration of Cl , and it shifted toward
more negative potentials with an increase in the concentration of Cl . Furthermore,
the extent of the inhibition of copper deposition was also found to be affected by the
molecular weight of PEG.
The current density decreased with an increase in the
molecular weight of PEG whereas the concentration of PEG does not exhibit distinct
change in the current density.
52
Chapter 2
Current density / mA cm-2
-100
-90
-80
-70
(a)
-60
-50
(b)
-40
-30
-20
(c)
-10
0
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
-0.7
Potential / V vs. Ag/AgCl
Figure 2.1. Cathodic polarization curves for copper electrodeposition from (a)
additive-free, (b) PEG, and (c) Cl-PEG baths. Rotation speed of RDE and potential
scan rate were 100 rpm and 2 mV s–1, respectively.
53
Chapter 2
To investigate the effect of mass transfer of these additives to the electrode,
polarization measurements were performed at the rotation speeds of 0, 100, and 2500
rpm with and without Cl– and PEG (Fig. 2.2). In both cases, the current density
increased with the rotation speed in the mass-transfer-limited region, where the
deposition rate of copper is controlled by the rate of diffusion of cupric ions.
The
limiting current densities on these cathodic polarization curves (insets of Fig. 2.2)
varied linearly with ω1/2 as predicted by the Levich equation [26].
On the other hand,
at potentials much less negative than those in the mass-transfer-limited region, the
current density was independent of the rotation speed.
54
Chapter 2
-20
ilim
(a)
Current density / mA cm-2
-15
-300
-200
-100
0
0
10 20 30
40 50
1/2
-10
0 rpm
100 rpm
2500 rpm
-5
0
0
-0.05
-0.1
-0.15
-0.2
Potential / V vs. Ag/AgCl
-15
(b)
-200
ilim
-2
Current density / mA cm
-300
-10
-100
0
0
10 20
30 40
50
1/2
0 rpm
100 rpm
2500 rpm
-5
0
0.1
0.05
0
-0.05
-0.1
-0.15
-0.2
-0.25
Potential / V vs. Ag/AgCl
Figure 2.2. Dependence of polarization characteristics for copper electrodeposition
from (a) additive-free and (b) Cl-PEG baths on the rotation speed of RDE. Rotation
speeds of RDE were 0, 100, and 2500 rpm, and the potential scan rate was 2 mV s–1.
The insets of (a) and (b) are Levich plots on limiting current density for the
additive-free and Cl-PEG baths, respectively.
55
Chapter 2
2.3.2 Observation of trench filling
Cross-sectional SEM images of trenches filled with electrodeposited
copper from the additive-free and Cl-PEG baths are shown in Fig. 2.3.
deposition was carried out under the static condition.
formation of voids in Cu-filled trenches.
spherical in shape (Fig. 2.3a).
The
The SEM images revealed the
For the additive-free bath, the voids were
On the other hand, with the Cl-PEG bath, ellipsoidal
voids with rough profiles were formed (Fig. 2.3b), while with the PEG (Cl–-free) bath,
voids formed in trenches were similar in shape to those formed with the additive-free
bath (Fig. 2.3c).
Figure 2.3. SEM images of copper electrodeposited in trenches from (a) additive-free,
and (b) Cl-PEG baths. Copper electrodeposition was performed galvanostatically at
–10 mA cm–2 for 90 s without bath agitation.
To reveal the evolution with time of the filling feature of copper in
trenches, the SEM observation of copper deposited in trenches was carried out with the
deposition time as a variable (Fig. 2.4). The Cu deposition was performed without
bath agitation.
With the additive-free bath, a thick deposit of copper was observed at
the edges of the openings of trenches in early stages of deposition process (Fig. 2.4a).
The deposition proceeded, while maintaining the feature of bulge formation around the
edges (Fig. 2.4b).
Finally, the coalescence of round edges at the trench-openings
resulted in void-formation in the Cu-filled trenches (Figs. 2.4c and 2.4d).
On the
other hand, with the Cl-PEG bath, conformal deposition was observed throughout the
filling process.
The sequential SEM images (Figs. 2.4e-2.4h) show that the surface
roughness of the merging sidewalls of trenches seems to form voids in the trenches at
56
Chapter 2
the later stages of the filling process (Figs. 2.4g and 2.4h).
To investigate the effect of mass transfer in the trenches, a series of
trench-filling experiments was also carried out with bath agitation (SEM images not
shown).
With agitation, voids still formed in copper-filled trenches obtained from
both the additive-free and the Cl-PEG baths.
Evidently, agitation brought about no
significant change in the filling feature for both baths.
Figure 2.4. Cross sectional SEM images of copper electrodeposited in trenches for
filling sequences from (a-d) the additive-free and (e-h) Cl-PEG baths. Copper
electrodeposition was performed galvanostatically at –10 mA cm–2 for 15, 30, 60, and
90 s (left to right) without bath agitation.
57
Chapter 2
2.3.3 Effect of Cl and PEG on copper deposition in trenches
A significant inhibition of copper deposition at the openings of trenches was
observed with baths containing Cl and PEG (Figs. 2.4e-2.4f). As was discussed by
Miura et al. [4], this inhibition seems to be related to the significant polarization
observed in the polarization curve for the Cl-PEG bath (Fig. 2.1c).
For the PEG bath
–
(without Cl ), the polarization curve (Fig. 2.1b) shows the suppression of the limiting
current density without any shift of potential, as compared with the curve for the
additive-free bath (Fig. 2.1a).
As PEG is considered to adsorb onto copper surface
[12,16], the decrease in the limiting current density for the PEG bath appears to be due
to adsorbed PEG molecules blocking the transport of cupric ions to the electrode
surface.
The strong inhibition of copper deposition observed with the Cl-PEG bath is
suggested to result from the formation of a blocking layer consisting of the constituents
of the bath.
It has been reported that PEG and Cl– form a complex with cuprous or
cupric ion [12-18].
As described in Chapter 1, Yokoi et al. [12] reported that the
blocking layer might consist of a complex of cuprous ion and PEG formed on
specifically adsorbed chloride ions.
58
Chapter 2
2.4. Effect of SPS
2.4.1 Polarization measurements
The polarization curves for the baths containing various concentrations of
SPS together with Cl– and PEG are shown in Fig. 2.5.
The addition of SPS to the
Cl-PEG bath brought about an increase in the slope of polarization curves, i/ E, in
the range of current density between –10 and –35 mA cm–2, from ~0.3 to ~0.5 A cm–2
V–1, indicating that copper deposition is accelerated in the Cl-PEG-SPS bath (Fig. 2.6).
It should be noted that the slope i/ E was ~0.5 A cm–2 V–1 regardless of the SPS
concentration in all polarization curves for the Cl-PEG-SPS baths. With increase in
SPS concentration, the potential where the current began to increase shifted in the
positive direction.
Such a positive shift of potential is suggestive of the greater
degree of acceleration of copper deposition at the higher SPS concentrations.
However, at the SPS concentrations higher than 1000 ppm, the dependence of the
potential shift on SPS concentration was negligible, and the polarization curves were
essentially identical to curve (e) in Fig. 2.5.
Current density / mA cm-2
-80
-70
-60
-50
(e)
(d) (c) (b) (a)
-40
-30
-20
-10
0
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
-0.7
Potential / V vs. Ag/AgCl
Figure 2.5. Effect of SPS concentration on polarization characteristics for copper
electrodeposition from Cl-PEG-SPS baths containing (a) 0, (b) 1, (c) 10, (d) 100, and
(e) 1000 ppm of SPS. Rotation speed of RDE, and potential scan rate were 100 rpm
and 2 mV s–1, respectively.
59
0.15
0.8
0.10
0.6
0.05
0.4
0
1
10
100
Slope ( i/ E)
Potential shift / V
Chapter 2
0.2
1000
Concentration of SPS / ppm
Figure 2.6. Plots of the shift of onset potential with respect to that for the SPS-free
(Cl-PEG) bath and the slope of polarization curve ( i/ E; see text) against SPS
concentration.
The rotation speed of the RDE affected the polarization characteristics for
the Cl-PEG-SPS bath (Fig. 2.7) in practically the same manner as those for the
additive-free and Cl-PEG baths shown in Fig. 2.2.
The current density in the
mass-transfer-limited region (at potentials more negative than –0.16 V) increased with
the rotation speed of RDE as predicted by the Levich equation (inset of Fig. 2.7), while
the current density was independent of the rotation speed at potentials less negative
than –0.16 V.
60
Chapter 2
-15
-400
ilim
Current density / mA cm-2
-300
-200
-100
-10
0
0
20
40
60
1/2
0 rpm
100 rpm
2500 rpm
-5
0
0.1
0.05
0
-0.05
-0.1
-0.15
-0.2
Potential / V vs. Ag/AgCl
Figure 2.7. Dependence of polarization characteristics for copper electrodeposition
from the Cl-PEG-SPS bath on the rotation speed of RDE. SPS concentration was 5
ppm. Rotation speeds of RDE were 0, 100, and 2500 rpm and the potential scan rate
was 2 mV s–1. The inset is a Levich plot on the limiting current density for the
Cl-PEG-SPS baths.
61
Chapter 2
2.4.2 Observation of trench filling
Figure 2.8 shows cross-sectional SEM images of trenches filled with
copper from the Cl-PEG-SPS baths containing various concentrations of SPS.
At
SPS concentrations between 2 and 100 ppm, no voids were observed in copper-filled
trenches, although the thickness of copper deposits in the areas above trenches
increased as compared with those obtained from the Cl-PEG bath.
The increase in
the thickness of copper deposits above trenches is considered to result from
overlapping adjacent bumps.
In the case of the wider trenches (SEM images not
shown), the presence of bumps was apparent above each trench.
This phenomenon,
i.e., an excessive deposition of copper above Cu-filled trenches, is known as “overfill”
phenomenon [3,6].
Figure 2.8. Cross sectional SEM images of copper electrodeposited in trenches from
the Cl-PEG-SPS baths. SPS concentrations were (a) 0, (b) 2, (c) 5, (d) 50, and (e) 500
ppm. Copper electrodeposition was performed galvanostatically at –10 mA cm–2 for 90
s without bath agitation.
62
Chapter 2
The sequential SEM images (Fig. 2.9) revealed that the conformal
deposition of copper proceeded during the first 30 s (Figs. 2.9a and 2.9b).
At the
deposition time of 60 s (Fig. 2.9c), the trenches were almost completely filled with
copper, whereas the increase in copper thickness at the trench opening was relatively
small.
The results suggest a greater degree of acceleration of copper deposition from
the trench bottom compared with that from the sidewalls.
These results for the
Cl-PEG-SPS bath agreed with the previous results reported by Moffat et al. [6] The
overfill phenomenon became less significant with an increase in SPS concentration and
disappeared completely with the addition of more than 500 ppm of SPS, which formed
small defects in Cu-filled trenches (Fig. 2.8e).
For the Cl-PEG-SPS baths, no
significant effect of the agitation was observed, similarly to the results obtained with
the additive-free and Cl-PEG baths.
Figure 2.9. Cross sectional SEM images of copper electrodeposited in trenches for
filling sequences from the Cl-PEG-SPS bath. SPS concentration was 2 ppm. Copper
electrodeposition was performed galvanostatically at –10 mA cm–2 for (a) 15, (b) 30,
(c) 60, and (d) 90 s without bath agitation.
63
Chapter 2
2.4.3 Effect of SPS on copper deposition in trenches
The bottom-up growth in the presence of SPS was more distinct at lower
concentrations of SPS, while it disappeared completely at high concentrations (Fig.
2.8). This SPS effect is accounted for by the acceleration of copper deposition by
SPS observed in the polarization curves for the Cl-PEG-SPS baths.
As shown in Fig.
2.5, the degree of acceleration of copper deposition increased with increasing SPS
concentration, and it tended to saturate at high SPS concentrations.
Figure 2.10
shows the relationship between SPS concentration and the current densities at –0.1,
–0.15, and –0.2 V (vs. Ag/AgCl), where copper deposition proceeded in trenches.
In
the range of low concentrations of SPS, the current density increased with increasing
SPS concentration, indicating strong dependence of the copper deposition rate on SPS
concentration.
On the other hand, the slope of the current density vs.
concentration
curves becomes smaller at higher concentrations of SPS, suggesting that the
concentration dependence of the degree of acceleration in this concentration range is
Current density / mAcm-2
small.
-60
-50
-40
-30
-0.1 V
-0.15 V
-0.2 V
-20
-10
0
0
200
400
600
800
1000
Concentration of SPS / ppm
1200
Figure 2.10. Dependence of current densities on SPS concentration at (◆) –0.1, (□)
–0.15, and (▲) –0.2 V vs. Ag/AgCl.
64
Chapter 2
This SPS concentration-dependent acceleration accounts for the observed
difference in the mode of trench filling at different SPS concentrations shown in Fig.
2.8.
These experimental results can be explained by the mechanism of the bottom-up
growth proposed by Moffat et al. [24,25].
In their model, SPS is accumulated at the
bottom of trenches because of the decrease in surface area at the bottom during the
trench-filling process.
SPS is assumed to remain at the copper-electrolyte interface
during copper deposition without being consumed.
This accumulation of SPS at the
bottom is considered to result in the bottom-up growth.
Based on this model, the
dependence of the growth mode on SPS concentration shown in Fig. 2.8 can be
interpreted as follows.
At low concentrations of SPS, the deposition rate of copper
strongly depends on SPS concentration.
Therefore, the “coverage gradient” (surface
concentration gradient) of SPS along the trench wall will provide a significant
difference in deposition rate at the bottom and at the top of the trench.
with excess SPS, the acceleration of copper deposition is saturated.
In contrast,
In this case, the
degree of acceleration of copper deposition will also be the same at all locations on the
trench wall.
The constant acceleration of copper deposition brings about the
conformal growth of copper in trenches, which leaves defects in copper-filled trenches.
65
Chapter 2
2.5 Effect of JGB
2.5.1 Polarization measurements
Polarization curves for the Cl-PEG-SPS-JGB baths containing various
concentrations of JGB are shown in Fig. 2.11.
The addition of JGB to the
Cl-PEG-SPS bath resulted in a negative shift of the potential where the current density
began to increase, which is in contrast to the effect of the addition of SPS to the
Cl-PEG bath.
The degree of the potential shift was greater at JGB concentrations
higher than 10 ppm (Fig. 2.12).
For the Cl-PEG-SPS-JGB bath, the current density in
the mass-transfer-limited region was not constant but fluctuated; nevertheless, it tended
to be lower than that for the Cl-PEG-SPS bath. The potential shift in the polarization
curves is attributable to the inhibition of copper deposition by JGB.
The slope of the
polarization curve, or i/ E, was approximately equal to 0.5 A cm–2 V–1 at JGB
concentrations between 0 and 10 ppm. The value increased to about 1.1 A cm–2 V–1
when 50 ppm of JGB was added.
Figure 2.11. Effect of JGB concentration on polarization characteristics for copper
electrodeposition. (a) Cl-PEG, (b) Cl-PEG-SPS, and (c-f) Cl-PEG-SPS-JGB baths
containing (c) 1, (d) 5, (e) 10, and (f) 50 ppm of JGB. SPS concentration was 5 ppm.
Rotation speed of RDE and potential scan rate were 100 rpm and 2 mV s–1,
respectively.
66
Chapter 2
1.2
0
Potential shift / V
0.8
-0.04
0.6
-0.06
Slope ( i/ E)
1.0
-0.02
0.4
-0.08
0.2
1
10
100
Concentration of JGB / ppm
Figure 2.12. Plots of the shift of onset potential with respect to that for the JGB-free
(Cl-PEG-SPS) bath and the slope of polarization curves ( i/ E; see text) against JGB
concentration.
67
Chapter 2
The dependence of polarization characteristics for the Cl-PEG-SPS-JGB
bath on the rotation speed is shown in Fig. 2.13.
It is of interest to note that the
electrode rotation appears to suppress the cathodic reaction at potentials less negative
than –0.17 V. This suppression was most significant at the rotation speed of 2500
rpm.
At potentials more negative than –0.17 V, where copper deposition is controlled
by the diffusion of cupric ions, the current density increased with the rotation speed of
RDE (inset of Fig. 2.13).
-15
ilim
Current density / mA cm-2
-400
-300
-200
-100
-10
0
0
20
1/2
40
60
0 rpm
100 rpm
400 rpm
900 rpm
2500 rpm
-5
0
0.1
0.05
0
-0.05
-0.1
-0.15
-0.2
Potential / V vs. Ag/AgCl
Figure 2.13. Dependence of polarization characteristics for the Cl-PEG-SPS-JGB bath
on the rotation speed of RDE. Both SPS and JGB concentrations were 5 ppm. Rotation
speeds of RDE were 0, 100 and 2500 rpm and potential scan rate was 2 mV s–1. The
inset is a Levich plot on the limiting current density for the Cl-PEG-SPS-JGB baths.
68
Chapter 2
2.5.2 Observation of trench filling
Furthermore, copper deposition in trenches from the baths containing
various concentrations of JGB with Cl–, PEG and SPS was investigated based on the
cross-sectional SEM images (Fig. 2.14).
Figure 2.14. Cross sectional SEM images of copper electrodeposited in trenches from
the Cl-PEG-SPS-JGB bath. SPS concentration was 2 ppm, and JGB concentrations
were (a) 0, (b) 1, (c) 2, and (d) 20 ppm. Copper electrodeposition was performed at –10
mA cm–2 for 90 s without bath agitation.
With the addition of 1 and 2 ppm of JGB to the Cl-PEG-SPS bath, the
thickness of copper deposits above Cu-filled trenches decreased, and the bumps above
each trench became distinctive.
The suppression of overfill phenomenon seems to
decrease the bump overlapping.
However, with the addition of 20 ppm of JGB,
defects began to appear in the trenches although the overfill phenomenon was almost
completely suppressed.
The sequential SEM images (Fig. 2.15) showed that the
conformal deposition proceeded during the first 30 s, and that the accelerated
deposition of copper from the bottom occurred at the deposition time of 60 s.
During
the first 60 s, the filling feature from the Cl-PEG-SPS-JGB bath was similar to that
from the Cl-PEG-SPS bath.
Interestingly, however, the profile of deposits from the
Cl-PEG-SPS-JGB bath was different from that formed in the Cl-PEG-SPS bath after
90 s of deposition.
The thickness of copper deposits above Cu-filled trenches from
the Cl-PEG-SPS-JGB bath was smaller than that from the Cl-PEG-SPS bath.
69
Chapter 2
Figure 2.15. Cross sectional SEM images of copper electrodeposited in trenches for
filling sequences from the Cl-PEG-SPS-JGB baths. Both SPS and JGB concentrations
were 2 ppm. Copper electrodeposition was performed galvanostatically at –10 mA
cm–2 for (a) 15, (b) 30, (c) 60, and (d) 90 s without bath agitation.
In contrast to the other baths, the effect of agitation was significant in the
JGB-containing bath (Fig. 2.16).
With agitation of the bath at the speed of 100 rpm,
the bumps above trenches became much smaller, and defect-free filling was obtained.
Figure 2.16. Cross sectional SEM images of copper deposited in trenches from the
Cl-PEG-SPS-JGB bath with agitation of the bath. Both SPS and JGB concentrations
were 2 ppm. Agitation speeds were (a) 0 and (b) 100 rpm
70
Chapter 2
2.5.3 Effect of JGB on copper deposition in trenches
Haba et al. [8] investigated the effect of JGB in the Damascene copper
electrodeposition without adding SPS.
According to their report, the addition of JGB
without SPS also achieves the bottom-up filling when there is a suitable concentration
gradient of JGB generated in a trench from the balance between the diffusion rate and
the consumption rate of JGB.
In this study, as JGB was added with SPS, the
bottom-up growth was brought about mainly by SPS, and the effect of JGB on the
filling process appeared to be insignificant.
The cross-sectional SEM images show
that the inhibition of the “overfill” phenomenon occurs at the later stages of copper
filling process (Fig. 2.14), and it is enhanced by an increase in JGB concentration (Fig.
2.15).
The inhibition of the overfill phenomenon and its dependence on JGB
concentration are related to the shift of polarization curves in the negative direction
brought about by the addition of JGB (Fig. 2.11). This negative shift is indicative of
inhibition of copper deposition by JGB.
The JGB effect appearing only at the later
stages of trench-filling is considered to be relevant to the depletion of JGB within
trenches.
In contrast to the effect of SPS (described in the preceding section), JGB is
reported to be consumed on the copper surface during the copper deposition.8 Thus,
within the trenches, where the supply of additives from the bulk of the solution is
retarded as compared with the outside of trenches, the consumption of JGB should lead
to a decrease in JGB concentration.
Although the inhibition was found to become
greater at higher concentrations of JGB, not only the overfill phenomenon but also the
bottom-up growth was inhibited at high concentrations of JGB.
It should be noted here that the effect of agitation of the bath was
noticeable only for the Cl-PEG-SPS-JGB bath.
Of particular interest is that agitation
was effective for the inhibition of the overfill phenomenon, while little effect was
observed for the bottom-up growth of copper in trenches (Fig. 2.16b).
The agitation
effect observed in the SEM images should be attributable to the enhanced transfer of
JGB to the electrode surface, in view of the results that no influence of agitation on the
filling feature of copper was observed for the other JGB-free baths.
The results of
polarization measurements performed at various rotation speeds of RDE (Figs. 2.2, 2.7
and 2.13) also support this interpretation.
The effect of RDE-rotation speed on
polarization characteristics at potentials less negative than the mass-transfer-limited
region was significant only for the Cl-PEG-SPS-JGB bath, although the limiting
current densities satisfied the Levich relation for all baths.
71
The polarization curves
Chapter 2
for the Cl-PEG-SPS-JGB baths recorded with the rotation of RDE shifted in the more
negative direction compared with the polarization curve obtained in the static bath.
This negative shift is considered to be due to the enhancement of the inhibition of
copper deposition with an increase in mass-transfer of JGB molecules to the electrode
surface.
These results suggest that the transfer of JGB molecules influences the
copper deposition rate, unlike that of the other additives.
Since agitation is
considered to affect insignificantly the fluid flow within submicrometer trenches [27],
the enhancement of JGB-transfer by agitation is assumed to be insignificant within
trenches compared with the outside of the trenches.
Thus, the significant
enhancement of JGB-transfer at the outside of trenches by agitation of the bath is
expected to bring about a strong inhibition of the overfill phenomenon without
inhibiting the bottom-up growth.
72
Chapter 2
2.6 Conclusions
Superfilling
of
copper
in
submicrometer
trenches
from
the
Cl-PEG-SPS-JGB bath was investigated from both polarization curves and
microscopic observation of trench filling.
The mechanism of superfilling in the
system of the Cl-PEG-SPS-JGB bath was discussed, particularly with a focus on the
inhibition of overfill phenomenon by JGB.
Copper electrodeposition at the opening
of the trenches is inhibited by the combination of Cl– and PEG throughout the
deposition process.
The bottom-up growth occurred in the presence of SPS and was
considered to be relevant to the SPS concentration-dependent acceleration of copper
deposition observed in polarization curves.
the later stages of deposition process.
JGB inhibits the overfill phenomenon in
However, a high concentration of JGB
inhibited the bottom-up growth, which resulted in void formation.
Bath agitation was
found to inhibit the overfill phenomenon significantly without producing voids in
copper-filled trenches.
The effect of agitation is considered to result from the
enhanced transfer of JGB at the exterior of trenches.
Therefore, the
mass-transfer-dependent inhibition by JGB is especially important for achieving the
bottom-up growth with the suppression of the overfill phenomenon.
73
Chapter 2
References
[1] P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans, and H. Deligianni, IBM J.
Res. Develop., 42, 567 (1998).
[2] J. J. Kelly and A. C. West, Electrochem. Solid-State Lett., 2, 561 (1999).
[3] P. Taephaisitphongse, Y. Cao, and A. C. West, J. Electrochem. Soc., 148, C492
(2001).
[4] S. Miura, K. Oyamada, Y. Takada, and H. Honma, Electrochemistry, 69, 773
(2001).
[5] M. Kang and A. A. Gewirth, J. Electrochem. Soc., 150, C426 (2003).
[6] T. P. Moffat, J. E. Bonevich, W. H. Huber, A. Stanishevsky, D. R. Kelly, G. R.
Stafford, and D. Josell, J. Electrochem. Soc., 147, 4524 (2000).
[7] T. P. Moffat, D. Wheeler, C. Witt, and D. Josell, Electrochem. Solid-State Lett., 5,
C110 (2002).
[8] T. Haba, T. Itabashi, H. Akahoshi, A. Sano, K. Kobayashi, and H. Miyazaki, Mater.
Trans., 43, 1593 (2002).
[9] A. C. West, S. Mayer, and J. Reid, Electrochem. Solid-State Lett., 4, C50 (2001).
[10] D. Josell, B. Baker, C. Witt, D. Wheeler, and T. P. Moffat, J. Electrochem. Soc.,
149, C637 (2002).
[11] M. Kang, M. E. Gross, and A. A. Gewirth, J. Electrochem. Soc., 150, C292 (2003).
[12] M. Yokoi, S. Konishi, and T. Hayashi, Denki Kagaku (presently Electrochemistry),
52, 218 (1984).
[13] D. Stoychev and C. Tsvetanov, J. Appl. Electrochem., 26, 741 (1996).
[14] J. J. Kelly and A. C. West, J. Electrochem. Soc., 145, 3472 (1998).
[15] J. J. Kelly and A. C. West, J. Electrochem. Soc., 145, 3477 (1998).
[16] J. P. Healy, D. Pletcher, and M. Goodenough, J. Electroanal. Chem., 338, 155
(1992).
[17] V. D. Jovic and B. M. Jovic, J. Serb. Chem. Soc., 66, 935 (2001).
[18] Z. V. Feng, X. Li, and A. A. Gewirth, J. Phys. Chem. B, 107, 9415 (2003).
[19] M. Tan and J. N. Harb, J. Electrochem. Soc., 150, C420 (2003).
[20] J. J. Kim, S.-K. Kim, and Y. S. Kim, J. Electroanal. Chem., 542, 61 (2003).
[21] Y. Cao, P. Taephaisitphongse, R. Chalupa, and A. C. West, J. Electrochem. Soc.,
148, C466 (2001).
[22] K. Kondo, N. Yamakawa, Z. Tanaka, and K. Hayashi, J. Electroanal. Chem., 559,
137 (2003).
74
Chapter 2
[23] K. Kondo, T. Matsumoto, and K. Watanabe, J. Electrochem. Soc., 151, C250
(2004).
[24] T. P. Moffat, D. Wheeler, W. H. Huber, and D. Josell, Electrochem. Solid-State
Lett., 4, C26 (2001).
[25] D. Josell, D. Wheeler, W. H. Huber, and T. P. Moffat, Phys. Rev. Lett., 87, 016102
(2001).
[26] A. J. Bard and L. R. Faulkner, Electrochemical Methods, p. 288, John Wiley &
Sons, New York (1980).
[27] K. M. Takahashi and M. E. Gross, J. Electrochem. Soc., 146, 4499 (1999).
75
Chapter 3
Effect of bath additives on the physical
and mechanical properties of copper
electrodeposits
Chapter 3
3.1 Effect of bath additives on room-temperature recrystallizaiton
of copper electrodeposits
3.1.1 Introduction
This chapter describes the effect of bath additives on the physical and
mechanical properties of electrodeposits.
The effects of additives such as chloride ions (Cl ), polyethylene glycol
(PEG), bis(3-sulfopropyl)disulfide (SPS), and Janus Green B (JGB) on the properties
of copper deposits are also of great interest for interconnect applications, because the
properties of copper are very important for maintaining the performance and reliability
of interconnections [1-5].
In this section, the effect of additives on “self-annealing”
of copper deposits was investigated.
occurring at room temperature.
Self-annealing is a process of recrystallization
This phenomenon is very important for maintaining
the properties of copper deposited in trenches because it leads to significant change in
the grain structure and resistivity in the deposit. Indeed, during self-annealing, the
grain size increases by an order of magnitude [1,6-8], and the resistivity decreases
nearly to the bulk value [9].
Several mechanisms of self-annealing have been proposed and discussed.
Factors inducing self-annealing which have been discussed so far include strain energy
caused by residual stress [1,4,10-12], high internal strain energy originated from
incorporated impurities [10-13], defect energy originated from dislocation [4], and
grain boundary energy associated with fine grains [14].
Although many researches
have been performed on self-annealing, detailed knowledge especially on the role of
additives is still lacking.
In this section, the effect of the additives (Cl . PEG, SPS and JGB) on the
self-annealing behavior was investigated. As is discussed in the previous chapter, the
effect of additives changed when they were present together.
The effect of various
combinations of the additives is considered to be also important for the self-annealing
behavior.
Thus, the author investigated the effect of various combinations of
additives based on the measurements of resistivity and residual stress, and the analysis
of crystallographic structure, and the microscopic observation of film structure.
Microanalysis of the deposits was also carried out to examine incorporation of the
additives.
79
Chapter 3
3.1.2 Experimental
In this study, resistivity, crystallographic structure and microstructure of
copper films were determined as a function of time after the electrodeposition.
Microanalysis and residual stress measurements were also performed.
Bath composition
The basic constituents of all baths used were 0.26 mol/L CuSO4 and 2.0
mol/L H2SO4. Chloride ions (as HCl), polyethylene glycol (PEG, average molecular
weight 2000), bis(3-sulfopropyl)disulfide (SPS) and Janus Green B (JGB) were added
to the baths. The concentrations of Cl , PEG, SPS and JGB were 50, 100, 10, and 10
ppm, respectively, as listed in Table 3.1. All experiments were carried out at room
temperature.
Bath name
Table 3.1. Bath compositions.
CuSO4 H2SO4 PEG (MW 2000)
mol L-1 mol L-1
ppm
Cl–
(HCl)
ppm
Additive-free
PEG
0.26
0.26
2.0
2.0
100
Cl-PEG
0.26
2.0
100
50
Cl-PEG-SPS
0.26
2.0
100
50
Cl-PEG-SPS-JGB
0.26
2.0
100
50
SPS
ppm
JGB
Ppm
10
(1-100)
10
10
Resistivity measurements and crystallographic structure analysis
For the resistivity and crystallographic structure measurements, circular glass
substrates, measuring 10 mm in diameter, with a sputtered Cu/Cr layer were employed.
Copper was electrodeposited on the substrate using a rotating disk electrode (RDE)
system equipped with a potentio/galvanostat (HABF501, HOKUTO DENKO).
-2
electrodeposition of copper was carried out galvanostatically 10 mA cm .
The
The
estimated thickness obtained was 3 m. The crystallographic structure was analyzed
by X-ray diffraction (XRD) (RAD-IC, RIGAKU), while the film resistivity was
measured by a four point probe method.
80
Chapter 3
Observation of microstructure
Microstructure of copper deposits was characterized by cross sectional
observation of deposits using a focused ion beam-scanning ion microscope (FIB-SIM,
SMI2050, SEIKO EG & G).
This experiment was performed on the copper deposited
on a Si substrate with a sputtered layer of Cu/Ti.
Copper was electrodeposited on the
substrate using the same rotating disk electrode (RDE) system as described above.
The electrodeposition of copper was carried out galvanostatically at 10 mA cm-2.
The
estimated thickness of copper obtained was 3 m..
Carbon microanalysis
Microanalysis of carbon in copper deposits was carried out by the
combustion-infrared absorption method (CS444, LECO).
For this measurement,
copper foil (99.99 ) was employed as a substrate. Copper was electrodeposited on
the substrate by using a paddle cell plating system equipped with a
potentio/galvanostat (HABF501, HOKUTO DENKO).
The electrodeposition of
copper was carried out galvanostatically at 10 mA cm-2.
performed with the deposits weighing about 1g.
81
Measurements were
Chapter 3
3.1.3 Resistivity measurements
Figure 3.1 compares the variations with time of the resistivity of copper
deposits prepared with four different baths: an additive-free bath, a bath containing Cl
and PEG (Cl-PEG bath), a bath containing Cl , PEG, and SPS (Cl-PEG-SPS bath) and
a bath containing Cl , PEG, SPS and JGB (Cl-PEG-SPS-JGB). For each specimen,
the first measurement was carried out 30 minutes after the electrodeposition current
was turned off.
The annealing effect was measured periodically during the room
temperature stand for 7 days.
The resistivity values of copper deposits from the
Cl-PEG-SPS and Cl-PEG-SPS-JGB baths decreased with time after deposition, while
those of the deposits from the additive-free and Cl-PEG baths remained constant over
the 7 day period.
The value of resistivity of copper from the Cl-PEG-SPS bath,
which was initially higher than that for the deposit obtained from the additive-free and
Cl-PEG baths, decreased continuously with time over the 7 day period after the
deposition.
For the deposits from the Cl-PEG-SPS-JGB bath, the resistivity also
decreased with time.
This result indicates that self-annealing occurred in the deposits
from the Cl-PEG-SPS-JGB bath.
However, the resistivity value of the deposits from
this bath was higher than that for the Cl-PEG-SPS bath, and even at 7 days after
deposition, it was the highest among all of the deposits from the baths examined in this
study.
The annealing behavior was found to depend strongly on the concentration
of SPS.
The resistivity value also decreased with time when SPS was added at
concentrations 10 ppm or less, although the rate and extent of self-annealing did not
depend on time.
Because self-annealing was occurred in the presence of SPS, it is
clear that SPS is a factor for inducing self-annealing.
However, such a decrease in
resistivity did not observed with the addition of 100 ppm of SPS (Fig. 3.1e).
This
result shows that excessive SPS suppress self-annealing, although SPS is essential for
self-annealing.
The surface morphology of copper deposit seemed to be dependent of
the concentration of SPS. The deposit produced in the Cl-PEG-SPS (100 ppm) bath
was dull, while that from the Cl-PEG-SPS bath containing 10 ppm of SPS was bright.
This suggests that the microstructure such as the grain size should be changed with the
concentration of SPS used in the bath.
82
Chapter 3
Figure 3.1.
Changes in resistivity of copper deposits with time of room
temperature stand.
The specimens were obtained from the (a) additive-free, (b)
Cl-PEG, (c and e) Cl-PEG-SPS, and (d) Cl-PEG-SPS-JGB baths.
concentrations, (c and d) 10 ppm and (e) 100 ppm.
83
SPS
Chapter 3
3.1.4 Crystallographic analysis
Changes in XRD profiles with time were found to correspond to the
results of resistivity measurements.
Figure 3.2 shows the XRD profiles recorded over
the period of 7 days after deposition.
The crystallographic structure of the deposit
obtained from the Cl-PEG-SPS bath (Fig. 3.2c) changed significantly with time after
the deposition while the change was not observed in the deposits from the additive-free
(Fig. 3.2a) and Cl-PEG (Fig. 3.2b) baths.
The deposits from the Cl-PEG-SPS-JGB
bath (Fig. 3.2d) clearly showed a change with time in the crystallographic structure.
Figure 3.2.
Changes in XRD profiles of copper deposits with time of room
temperature stand.
The specimens were obtained from (a) additive-free, (b) Cl-PEG,
(c) Cl-PEG-SPS, and (d) Cl-PEG-SPS-JGB baths.
84
Chapter 3
For the deposits from the Cl-PEG-SPS (Fig. 3.2c) and Cl-PEG-SPS-JGB
(Fig. 3.2d) baths, the intensities of (111) and (200) diffraction peaks increased over the
7 days after electrodeposition.
Figure 3.3 shows the variation with time of the
intensity of (111), I111, (Fig. 3.3A) and the relative intensity of (200) with respect to
(111), I200/I111, (Fig. 3.3B) obtained from the XRD profiles.
The values of I111 for
the Cl-PEG-SPS (plot c) and Cl-PEG-SPS-JGB (plot d) baths continuously increased
over the 7 day period.
On the other hand, for the deposits from the other baths (plots
a and b), such a continuous increase in restivity was not observed.
It is seen that the
changes in the peak intensity, I111, correspond well to the results of resistivity
measurements shown in Fig. 3.1.
The values of I200/I111 for the deposits from the
Cl-PEG-SPS (plot c) and Cl-PEG-SPS-JGB (plot d) bath increased with time after
deposition, while the values of I200/I111 are essentially the same for corresponding
specimens from the additive-free (plot a) and Cl-PEG (plot b) baths.
A similar
finding has also been reported by Lee et al. [1]. The change in the value of I200/I111
suggests the occurrence of anisotropic grain growth, which led to the change in
preferred crystallographic orientation upon self-annealing.
This significant increase
in (200) preferred orientation has been suggested to result from the anisotropy of
elastic constant of copper crystal [1].
Different elastic constants for different
crystallographic orientations result in a difference in strain energy for crystals with
different orientations in the film.
Because the strain energy is the lowest for (200)
orientation, the (200)-oriented grains grow preferentially during self-annealing.
The extent of the increase in I200/I111 for the deposits produced in the
Cl-PEG-SPS bath (Fig. 3.3B, plot c) was more significant than that for the deposits
from the Cl-PEG-SPS-JGB bath (Fig. 3.3B, plot d).
The value of I200/I111 at 7 days
after the deposition from the Cl-PEG-SPS bath and the Cl-PEG-SPS-JGB bath were
approximately equal to 1.5 and 0.5, respectively.
The results of the resistivity and
XRD measurements presented above show that self-annealing, which is induced by the
addition of SPS to the Cl-PEG baths, is suppressed by the addition of JGB.
It is
interesting to note in Fig. 3.3B that the increase in I200/I111 for the deposits from the
Cl-PEG-SPS and Cl-PEG-SPS-JGB baths stopped only after 3 days, while the intensity
of I111 continues to increase for at least 7 days.
These results suggest that two types of
recrystallization occurred during the period of self-annealing.
In the period of 3 days,
the value of I200/I111 in creased accompanied by the increase in the peak intensity, I111
and I200.
After the 3 day period, the intensity of diffraction peaks, I111 and I200,
increased, while the value of I200/I111 remained constant. Because the change in the
85
Chapter 3
value of I200/I111 indicates the anisotropic grain growth, it was suggested that the
anisotropic grain growth occurred in a few days after deposition, followed by the
normal grain growth for the remainder of the 7 day period.
Figure 3.3.
Changes in I111 (A) and I200/I111 (B) in XRD profiles of copper deposits
with time of room temperature stand.
The specimens were obtained from the (a)
additive-free, (b) Cl-PEG, (c) Cl-PEG-SPS, and (d) Cl-PEG-SPS-JGB baths.
86
Chapter 3
Significant change was found in the XRD profile of copper deposit from the
bath containing 100 ppm of SPS.
The intensity of (111) diffraction peak for this
deposit was much higher than those for the deposits from lower SPS concentrations.
Table 3.2 lists the intensities of Cu (111) and (200) diffraction peaks recorded with the
deposits from the bath containing various concentrations of SPS.
The intensity for
the deposits from the bath containing 50 ppm and 100 ppm was much higher than
those for the deposits from the bath containing SPS at the concentrations of 10 ppm or
less.
In addition, the intensity tends to be decreased by adding SPS at concentrations
lower than 10 ppm.
For a (200) diffraction peak, the intensity was independent of the
concentration of SPS.
Table 3.2. Intensities of Cu(111) and (200) diffraction peaks.*
SPS concentration
I111 / cps
I200 / cps
Additive-free bath
7282
1175
0 ppm (Cl-PEG bath)
9399
910
2 ppm
5585
1312
10 ppm
5881
1172
50 ppm
37900
1439
100 ppm
35917
1215
*
The data was obtained one day after electrodeposition.
The values of I200/I111 for the deposits from the bath 50 or 100 ppm of SPS
were unchanged with time.
The increase in the diffraction peaks indicates the
improvement of crystallinity of the deposit, which is supposed to result from the
elimination of crystal defect such as dislocation and vacancy.
The energy originated
from these defect or imperfection in copper crystal may also contribute to occurrence
of self-annealing.
87
Chapter 3
3.1.5 Microstructure analysis
As described above, the combination of additives was found to affect
self-annealing behavior significantly.
To consider the additive effects, the
microstructure of the deposits was investigated by an FIB-SIM.
Figure 3.4 shows
cross-sectional FIB-SIM images of copper deposit obtained from the additive-free,
Cl-PEG, Cl-PEG-SPS, and Cl-PEG-SPS-JGB baths.
The images were observed
periodically over the period of 1 month after electrodeposition.
The first
measurements for the deposits from the various baths were carried out 1 day after
deposition.
As is seen in this figure, the images of the deposits from the Cl-PEG-SPS
(Fig. 3.4c) and Cl-PEG-SPS-JGB (Fig. 3.4d) baths changed with time after deposition.
For the Cl-PEG-SPS and Cl-PEG-SPS-JGB baths (Figs. 3.4c and 3.4d), the size of
grains in the deposits observed 1 day after the deposition was uniform and very small.
The average grain size of this deposit measured in the horizontal direction from the
SIM image was 50 nm.
The size of grains in these samples increased within 2 days
after deposition, and the size of grains in the 2 day-old deposit (not shown in the
figure) was 1.5 m, which was identical to those in the 17 day- and 1 month-old
deposits. On the other hand, no such grain growth was observed in the deposits
obtained from either the additive-free bath or the Cl-PEG bath even after 1 month.
Comparison of the images obtained 1 day after deposition shows that cross sectional
structure were clearly different for the deposits obtained from the different baths.
According to Fischer’s classification [15], the deposit from the Cl-PEG-SPS bath (Fig.
3.4c) is regarded as the “unoriented dispersion (UD)” type deposit (see Chapter 1,
1.2.1).
This structure indicates that 3-dimensional nucleation occurred during
electrodeposition [16,17]. The grain structure of the specimens from the additive-free
bath (Fig. 3.4a) and the Cl-PEG bath (Fig. 3.4b) were different from that of the
specimen from the Cl-PEG-SPS bath (Fig. 3.4c). For the deposits from former two
baths, the grains were not uniform in size and shape, and rough and large grains were
distributed near the surface of the films, while the grains near the substrate were quite
small.
The sizes of grains near the surface of both the deposits were ranging from
200 nm to 1 m, while that near the substrate was approximately 48 nm. This kind of
structure is the intermediate type between the “basis reproduction (BR)” and the “field
oriented (FT)”, which may form when 2-dimensional nucleation occurs [16,17].
The
observed difference in cross-sectional structure of the 1 day-old deposits, which may
result from the difference in the mode of film growth, is attributed to the difference in
88
Chapter 3
self-annealing behavior.
Figure 3.4.
Changes in cross-sectional microstucture of copper deposits with time of
room temperature stand.
The specimens were obtained from the (a) additive-free, (b)
Cl-PEG, (c) Cl-PEG-SPS, and (d) Cl-PEG-SPS-JGB baths.
89
Chapter 3
Generally, grain growth is considered to be driven by several different kinds
of energies such as interface/surface energy, grain boundary energy, strain energy, and
defect energy [12,18]. In the present case, grain boundary energy is suggested to
mainly drive the grain growth during self-annealing.
If the grains contain no defect,
the grain boundary energy per unit volume is [12]:
3
L
E GB
(3.1)
GB
where EGB is the grain boundary energy per unit volume, L is the grain radius, and GB
is the grain boundary energy per unit area, which is 0.5 J/m2 for copper [20].
According to this equation, grain boundary energy is in inverse proportion to grain size.
The grain boundary energy increases with an increase in grain boundary area.
In the
one day-old deposits obtained from the Cl-PEG-SPS bath, the grain size was 48 nm on
an average. On the other hand, the average size of grains of the deposits obtained from
the additive-free and Cl-PEG bath were approximately 400 nm and 500 nm, although
the size of grains of the deposits from the latter two baths were not uniform.
The
grain boundary energy for the 1 day-old deposit from the Cl-PEG-SPS bath was 2.3
meV/atom.
This value was much higher than those for the deposits from the
additive-free and Cl-PEG baths, which were 0.28 and 0.22 meV/atom.
The energy in
released during grain growth is given by:
EGB
3
GB
1
Li
1
Lf
(3.2)
where initial grain size is Li and final grain size is Lf.
According to this equation,
grain growth leads to a reduction in total grain boundary energy as a result of the
decrease in grain boundary area.
The final grain size of the deposit from the
Cl-PEG-SPS bath was more than 1.5 m.
Cl-PEG-SPS bath was about 2.2 meV/atom.
Thus,
EGB for the deposit from the
The value of the released energies of strain generated by stress and that
originated from dislocation are reported to be much smaller [12]. The strain energy
released during self-annealing is reported to be 0.07 meV/atom.
The dislocation
energy in copper deposits is smaller than these energies discussed above.
90
The
Chapter 3
dislocation energy depends on the the dislocation density in copper electrodeposits, as
described in Chapter 2 (p. 30).
The dislocation density of copper electrodeposits is
14
/ m2, and that the dislocation energy is about 0.01
generally on the order of 10
meV/atom.
In comparison of these energies, the grain boundary energy for the
specimen from the C-PEG-SPS bath is very high, and, thus, it is considered that the
grain boundary energy associated by fine grains is one of the major factors contributing
the grain growth during self-annealing.
The recrystallization of copper deposits is known to occur during the process
of thermal annealing, which is usually carried out at above 400 °C.
energy produced by the heating from 20 °C to 400 °C is 35 meV/atom.
The thermal
This energy is
much higher than the grain boundary energy calculated in this study, although the
energy required for self-annealing may not the same as the energy produced by the
heating during thermal annealing.
was assumed to exist.
Thus, the other energies inducing self-annealing
One of the factors which may affect self-annealing is the local
strain energy originated from the additives incorporated in the deposits.
Lee et al.
[12] calculated the energy generated around a molecule of PEG incorporated in grain
boundary of copper deposits.
They assumed the misfitting of particle generated by
the incorporation of a large PEG molecule.
According to their model, the energy of
strain to a copper grain adjacent to the incorporated PEG molecule was 27.2 meV/atom.
The calculated vale indicates that the effect of incorporation of additives is important
for self-annealing, although they did not evaluate the amount of the incorporated
additives.
Therefore, in the next section (section 3.1.6), the amount of the additives
incorporated in a copper deposit was measured to discuss the impact of the effect of the
local strain energy generated from the additive incorporation.
The grain growth also occurred in the deposit from the Cl-PEG-SPS-JGB
bath (Fig. 3.4d).
The image of the deposit from this bath recorded 1 day after
deposition was similar to that of the deposition from the Cl-PEG-SPS bath (Fig. 3.4c),
although the former deposit contained some large grains near the substrate. The
generation of these large grains appears to be caused by JGB, although the reason for
the formation of these large grains is not understood.
It seems to be related to the
decrease in the rate and degree of self-annealing caused by JGB, which was observed
in the XRD studies.
After self-annealing, the grain size measured in the image for the
Cl-PEG-SPS-JGB bath (Fig. 3.4d) was identical to that for the Cl-PEG-SPS bath (Fig.
3.4c). The time period in which grain size was observed by FIB-SIM corresponded to
that in which anisotropic grain growth observed in XRD measurements.
91
Thus, it is
Chapter 3
suggested that the grain boundary diffusion occurs during the anisotropic grain growth.
From the results described above, it is clear that SPS exerts the most
significant effect on the occurrence of self-annealing.
However, this additive is
known to interact with Cl and PEG. To gain insight into the effect of interaction of
the additives on self-annealing, the microstructure of the deposits from the bath
containing various combinations of the Cl , SPS, and PEG was investigated.
The
FIB-SIM observation was performed for the deposits from the Cl , SPS, Cl-SPS, and
PEG-SPS baths (Fig. 5). These images reveal that self-annealing did not occur in
deposits from all baths.
The microstructure of the one day old deposits from these
baths was almost identical to that from the additive-free and Cl-PEG baths presented in
Fig. 3.4.
These results support our assumption that the initial grain size is a major
factor for inducing self-annealing.
In more precise terms, it can be stated that the size
and shape of grains in the area of the upper portion of the deposits depended on the
combination, although grains near the substrate were equally small in all one day old
deposits, as can be seen in Fig. 3.5.
92
Chapter 3
Figure 3.5.
Changes in cross-sectional microstucture of copper deposits with time of
room temperature stand. The specimens were obtained from the (a) Cl , (b) SPS, (c)
Cl-SPS, and (d) PEG-SPS baths.
93
Chapter 3
3.1.6 Carbon microanalysis
Strain energy originated from incorporated additives is also a factor for
self-annealing to be considered.
Lee et al. [12] calculated the strain energy caused by
PEG incorporation by assuming that PEG is segregated at the grain boundary, and
concluded that the stress exerted by incorporated PEG on copper atoms in the deposit
is a significant factor of self-annealing.
However, as discussed in the previous section,
they did not evaluate the amount of incorporated PEG. To investigate the effect of
incorporation of additives on self-annealing, carbon microanalysis was performed with
copper electrodeposits. Table 3.3 lists carbon contents of the deposits produced in the
baths containing various combinations of additives.
The experiments revealed that
the carbon content of copper deposited from the Cl-PEG-SPS bath was very low.
It is
interesting to note that the carbon content of copper deposit was decreased by adding
SPS to the Cl-PEG bath.
As shown in Table 3.3, carbon was not detected in the film
from the additive-free bath.
The value of carbon content was increased to 0.046 wt%
by adding Cl and PEG. Further addition of SPS decreased the carbon content of the
film and the value of carbon content of the deposit from this bath was only 0.003 wt%,
which was smaller than the reliable detection limits of 0.01 wt%. The addition of
JGB to the Cl-PEG-SPS bath increased the carbon content.
The increase in carbon
content of the film caused by adding JGB to the bath is attributed to the incorporation
of JGB in the deposit. This result corresponds to the finding reported by Haba et al.
[19] that JGB is consumed during the electrodepositon.
As mentioned earlier,
self-annealing is suppressed by adding JGB to the Cl-PEG-SPS bath.
This result
suggests that the suppression of self-annealing results from the pinning of grain growth
by incorporated JGB molecules.
For the Cl, SPS, Cl-SPS, and PEG-SPS baths, the
carbon contents were very low.
This fact indicates that the additives in these baths
were not incorporated in the deposits.
The results of carbon microanalysis suggest
that the incorporation of additives suppressed grain growth.
The results also indicate
that the strain originated from trapped additives is not an important factor for inducing
self-annealing.
94
Chapter 3
Table 3.3.
Carbon content in copper deposits.
carbon content / wt %
additive-free
not detected
Cl-PEG
0.046
Cl-PEG-SPS
0.003
Cl-PEG-SPS-JGB
0.016
Cl
0.002
SPS
0.003
Cl-SPS
0.002
PEG-SPS
0.003
95
Chapter 3
3.1.7 Residual stress measurements
Finally, the effect of residual stress in the deposits was also investigated as a
possible factor inducing self-annealing. Because the energy of stress in the deposits
would destabilize the deposit structure, self-annealing is expected to be brought about
by the relaxation of this strain energy.
Figure 3.6 is the effect of aging on tensile
stress in the copper deposits obtained from the additive-free, Cl-PEG, Cl-PEG-SPS,
and Cl-PEG-SPS-JGB baths.
If residual stress is indeed a factor inducing
self-annealing, the stress is expected to be high in the deposit from the Cl-PEG-SPS
bath and it should decrease with time during self-annealing.
not support this assumption.
However, the result did
The tensile stress in the copper deposit from the
Cl-PEG-SPS bath, which exhibited self-annealing, was the lowest among the copper
films from the baths examined in this study.
In addition, as reported earlier [7,20],
tensile stress in the deposits from the Cl-PEG-SPS bath slightly increased with time
after deposition.
The increase in tensile stress of the copper deposit produced in the
Cl-PEG-SPS bath is likely to have resulted from the grain growth during self-annealing.
An increase in the number of grains in a deposit leads to a decrease in the volume of
grain boundary, which results in a decrease in total volume of copper deposit. The
decrease in the total volume of the deposit is expected to result in the increase in
tensile stress.
Thus, the increase in tensile stress observed during self-annealing is
considered reasonable.
The result was also supported by the fact that the tensile
stress for the deposit from the Cl-PEG-SPS-JGB bath also increased with time after
deposition.
The results discussed above show that the effect of residual stress is not
important as far as self-annealing is concerned.
The decreased size of grain in the
deposits from the Cl-PEG-SPS and Cl-PEG-SPS-JGB baths results in an increased in
the area of grain boundary.
Thus, these deposits possess a large grain boundary
energy compared with the deposits from other baths.
Thus this grain boundary
energy is concluded to be a main factor for grain growth observed in the FIB-SIM
images (Figs. 3.4c and 3.4d).
96
Chapter 3
Figure 3.6.
Changes in residual stress of copper deposits with time of room
temperature stand. The specimens were obtained from the (a) additive-free, (b) Cl-PEG,
(c) Cl-PEG-SPS, and (d) Cl-PEG-SPS-JGB baths.
97
Chapter 3
3.1.8 Conclusions
The effects of bath additives on the microstructure and properties of
copper electrodeposits and self-annealing were investigated.
Self-annealing occurs in
copper deposits obtained from the bath containing Cl , PEG and SPS in combination
(the Cl-PEG-SPS and Cl-PEG-SPS-JGB baths). The grain size of the deposits before
self-annealing was very small.
The large grain-boundary energy associated with the
small grains is considered to be a cause of self-annealing.
Cl-PEG-SPS bath was found to inhibit self-annealing.
The addition of JGB to the
The inclusion of JGB in the
deposit is likely to be responsible for the inhibition of self-annealing.
in copper deposits does not induce self-annealing.
Residual stress
Furthermore, it was also found
that the behavior of self-annealing strongly depended on the concentration of SPS used
in the bath.
Significant self-annealing occurred in the deposits from the bath
containing an appropriate amount of SPS, while it was suppressed completely with an
excess SPS.
The crystallinity of grains in the deposits from the latter baths were very
high, compared with the deposits from the former bath.
Although the energy
originated from the imperfection in the crystal is reported to be much smaller than that
of the grain boundary, the result suggests that crystallinity of the grains is also
considered to be related to the behavior of self-annealing.
98
Chapter 3
3.2 Effect of room-temperature recrystallization on ductility of copper
electrodeposits
3.2.1 Introduction
Electrodeposition of copper is used for the fabrication of interconnections of
printed circuit boards (PCB), systems in package (SiP), as well as ULSI interconnects.
In these applications, mechanical properties of copper deposits are very important for
maintaining the reliability of interconnections.
One of the most important properties
of electrodeposited copper is ductility, especially for products with fine patterns.
High ductility is required to prevent the copper on PCB from cracking during soldering,
which is caused by the difference in thermal expansion coefficient between the copper
film and the substrate [21].
Recently, ductile electrodeposited copper is required for
the fabrication of interconnections in the interposer layer of SiP to absorb the stress
between PCB and semiconductor devices.
The high ductility is also advantageous in
the fabrication of copper interconnections on flexible substrates.
It is well known that the ductility of copper deposits is affected by factors
such as crystallographic structure [22,23], grain size [22-24], impurity [25-28], and
lattice strain [21,29-32].
Among them, the gain size is recognized as one of the most
important factors affecting a variety of mechanical and physical properties.
Since
grain boundaries in copper films inhibit the slippage of crystal planes and lower the
ductility of copper deposits [25], a large grained deposit with small grain boundary
volume is expected to possess a high ductility.
Ye et al. [22]reported that the ductility
of a 20 m thick electrodeposited copper film tended to increase with increasing grain
size. However, other researchers reported that the ductility of thick copper films
decreases with an increase in grain size because large grained films possess a large
internal stress, which is expected to deteriorate the ductility [23].
Although various
variables appear to influence the ductility in a complicated manner, grain size is
undoubtedly one important factor affecting the ductility.
The microstructure evolution associated with self-annealing, described in the
preceding section, is expected to enhance the ductility.
However, as far as the author
is aware, no publication is available on the effect of self-annealing on ductility.
In
this section, the effect of the above bath additives (Cl , PEG, and SPS) on the variation
of the ductility of copper electrodeposits as a result of self-annealing. The change in
ductility with time after deposition was measured and compared to variations with time
of other properties including crystallographic structure, grain size, and resistivity.
99
Chapter 3
3.2.2 Experimental
Bath conposition
The basic constituents used in this study are same as that used in the section
3.1.
For the experiment, Cl , PEG, and SPS were added as additives to the bath, in
various combinations. The concentrations of Cl , PEG, and SPS were fixed at 50,
100, and 10 ppm, respectively. All experiments were carried out at room
temperature.
Grain size measurements
The grain size of the deposit was estimated from XRD scans using Scherrer’s
formula [33]:
t
0.9
B cos
where t is grain size,
(3.3)
B
is wave length of X-ray, B is half value of width of a diffraction
peak, and
is peak position.. The details of the condition of XRD measurements were
described in the section 3.1.
Ductility measurements
The deposit ductility was evaluated with a mechanical bulge tester [34].
Copper was deposited on an 8 m-thick rolled copper foil (99.99
0.71
) with elongation of
. The thickness of the copper deposits used for the ductility measurement was
30 m. The current density used for the electrodeposition was 30 mA cm-2. Because
of the fact that the thickness and the ductility of the copper foil substrate were both
much smaller than those of the copper deposits, it was assumed that the ductility
measured with the substrate-deposit composite film is equal to the ductility of the
deposit itself.
100
Chapter 3
3.2.3 Grain size measurements
To estimate the size of grain in copper deposits, XRD profiles recorded with
copper films prepared in an additive-free bath, a bath containing Cl and PEG (Cl-PEG
bath), and a bath containing Cl , PEG, and SPS (Cl-PEG-SPS bath). The profiles of
XRD scan are shown in Fig. 3.2 (section 3.1).
The grain size of each deposit calculated from the XRD peaks for the (111)
and (200) orientations using Scherer’s formula is shown in Fig. 3.7.
For the
additive-free bath, the grain sizes calculated from both (111) and (200) peaks did not
change within the period of this investigation.
For the Cl-PEG and Cl-PEG-SPS
baths, the grain sizes calculated from both (111) and (200) orientations in the
as-deposited state were much smaller than those for the additive-free bath.
The grain
size of the deposit from the Cl-PEG bath increased during the first day after deposition.
However, the grain size stopped to incrrease 2 days after the deposition.
Unlike the
deposit from the Cl-PEG bath, the sizes of both (111) and (200)-oriented grains of the
deposit obtained from the Cl-PEG-SPS bath continued to increase with time over the
period of 7 days after the deposition, and they were greater by ~50 % for
(111)-oriented grains and ~70 % for (200)-oriented grains than those measured 7 days
after deposition from the additive-free bath.
101
Chapter 3
Figure 3.7.
Change in grain size of copper deposits with time at room temperature.
The specimens were obtained from the additive-free ( ◇ ), Cl-PEG ( ▲ ), and
Cl-PEG-SPS baths (○). The grain size was calculated from XRD peaks for (111) (a),
and (200) orientations (b) using Scherer’s formula.
102
Chapter 3
3.2.4 Ductility measurements
Table 3.4 compares the ductility values in
elongation of the electrodeposited
copper films obtained from the additive-free, Cl-PEG, Cl-PEG-SPS baths. The
ductility of each film was measured 15 days after deposition.
film from the Cl-PEG-SPS was 3.19
The elongation of the
, which is 1.54 times larger than that of the
specimen obtained from the additive-free bath (2.07
).
In contrast, the elongation of
the deposit produced in the Cl-PEG (SPS-free) bath was as low as 1.22 . The
higher ductility of the deposit formed in the Cl-PEG-SPS bath is attributed to a change
in microstructure of the film that took place during the self-annealing.
Table 3.4. Comparison of ductility values in % elongation of copper deposits obtained
from additive-free, Cl-PEG, and Cl-PEG-SPS baths.*
Elongation / %
2.07
1.22
3.19
Additive-free
Cl-PEG
Cl-PEG-SPS
*
The ductility of each specimen was measured 15 days after deposition. The values
reported are the averages of several independent measurements.
To clarify the relation between the ductility and the change in film
microstructure during self-annealing, the variation of ductility with the time elapsed
after deposition was investigated (Fig. 3.8).
The ductility of the deposit formed in the
additive-free bath remained constant up to 15 days after a slight increase during the
first 2 days.
On the other hand, the ductility of the deposit from the Cl-PEG-SPS bath
continued to increase with time up to 11 days, after which it remained constant for at
least 15 days.
This improvement of ductility observed for the Cl-PEG-SPS bath is
relevant to the increase with time of the grain size observed in this study.
Thus, it
was concluded that the improvement of ductility of copper was brought about by the
recrystallization which took place as a result of self-annealing.
103
Chapter 3
Figure 3.8.
Change in elongation of copper deposits with time at room temperature.
The specimens were obtained from the additive-free (◇) and Cl-PEG-SPS baths (○).
3.2.5 Conclusions
The effects of bath additives on the ductility of copper electrodeposits were
investigated.
The ductility was improved by the addition of SPS to the bath
containing Cl , PEG in combination. The ductility of the deposits from this bath was
also found to increase continuously with time after deposition. This change in the
ductility with time was shown to be correlated with the change in grain size of the
deposits resulted from room-temperature recrystallization.
It is thus concluded that
the emhancement of the ductility resulted from the change in microstructure of the film
which took place upon self-annealing.
104
Chapter 3
References
[1] H. Lee, S.S. Wong, and S.D. Lopatin, J. Appl. Phys., 93, 3796 (2003).
[2] S.H. Brongersma, E. Kerr, I. Vervoort, A. Saerens, and K. Maex, J Mater Res, 17,
582 (2002).
[3] M. Hasegawa, and Y. Hirai, J. Appl. Phys., 90, 2792 (2001).
[4] C. Lingk, and M.E. Gross, J. Appl. Phys., 84, 5547 (1998).
[5] Q.T. Jiang, M.E. Gross, G. Bersuker, B. Foran, R. Mikkola, B. Carpenter, and J.
Oemando, Mat. Res. Soc. Symp. Proc., 564, 429 (1999).
[6] H. Lee, W.D. Nix, and S.S. Wong, J. Vac. Sci. Technol. B, 22, 2369 (2004).
[7] V.A. Vas'ko, I. Tabakovic, S.C. Riemer, and M.T. Kief, Microelectron. Eng., 75,
71 (2004).
[8] V.A. Vas'ko, I. Tabakovic, and S.C. Riemer, Electrochem. Solid-State Lett., 6,
C100 (2003).
[9] T.P. Moffat, J.E. Bonevich, W.H. Huber, A. Stanishevsky, D.R. Kelly, G.R.
Stafford, and D. Josell, J. Electrochem. Soc., 147, 4524 (2000).
[10] S. Lagrange, S.H. Brongersma, M. Judelewicz, A. Saerens, I. Vervoort, E.
Richard, R. Palmans, and K. Maex, Microelectron. Eng., 50, 449 (2000).
[11] S.H. Brongersma, E. Richard, I. Vervoort, H. Bender, W. Vandervorst, S.
Lagrange, G. Beyer, and K. Maex, J. Appl. Phys., 86, 3642 (1999).
[12] C.H. Lee, and C.O. Park, Jpn. J. Appl. Phys. [1] Reg. Pap. Short Notes, 42, 4484
(2003).
[13] M. Stangl, J. Acker, V. Dittel, W. Gruner, V. Hoffmann, and K. Wetzig,
Microelectron. Eng., 82, 189 (2005).
[14] K. Ueno, T. Ritzdorf, and S. Grace, J. Appl. Phys., 86, 4930 (1999).
[15] H. Fischer, Electrochim. Acta, 2, 50 (1960).
[16] R. Winand, P. VanHam, R. Colin, and D. Milojevic, J. Electrochem. Soc., 144,
428 (1997).
[17] R. Winand, Hydrometallurgy, 29, 567 (1992).
[18] C.V. Thompson, Annu. Rev. Mat. Sci., 30, 159 (2000).
[19] T. Haba, T. Itabashi, H. Akahoshi, A. Sano, K. Kobayashi, and H. Miyazaki,
Mater. Trans., 43, 1593 (2002).
[20] J.M.E. Harper, C. Cabral, P.C. Andricacos, L. Gignac, I.C. Noyan, K.P. Rodbell,
and C.K. Hu, J. Appl. Phys., 86, 2516 (1999).
[21] L. Zakraysek, Matal Finish., 85, 29 (1987).
[22] X.P. Ye, M. Debonte, J.P. Celis, and J.R. Roos, J. Electrochem. Soc., 139, 1592
105
Chapter 3
(1992).
[23] M. Matsuoka, S. Ito, and T. Hayashi, Kinzoku Hyomen Gijutu, 33, 385 (1982).
[24] L.L. Shreir, and J.W. Smith, J. Electrochem. Soc., 98, 193 (1951).
[25] S. Nakahara, Y. Okinaka, and H.K. Straschil, J. Electrochem. Soc., 136, 1120
(1989).
[26] Y. Okinaka, and H.K. Straschil, J. Electrochem. Soc., 133, 2608 (1986).
[27] J.E. Graebner, and Y. Okinaka, J. Appl. Phys., 60, 36 (1986).
[28] D. Anderson, R. Haak, C. Ogden, D. Tench, and J. White, J. Appl. Electrochem.,
15, 631 (1985).
[29] F. Barlat, and O. Richmond, Mater. Sci. Eng., 95, 15 (1987).
[30] F. Barlat, Mater. Sci. Eng., 91, 55 (1987).
[31] W. Riedel, Galvanotechnik, 71, 1207 (1980).
[32] H.J. Read, Plating, 49, 602 (1962).
[33] B.D. Culity, Elements of X-ray Diffraction, Addison-Wesley Pub., Massachusetts,
(1978).
[34] S. Nakahara, Y. Okinaka, and D.R. Turner, J. Test. Eval., 5, 178 (1977).
106
Chapter 4
Void-free
filling
of
submicrometer
trenches by electroless copper deposition
Chapter 4
4.1 Void-free filling of trenches with electroless copper deposits
using a combination of accelerating and inhibiting bath additives
4.1.1 Introduction
In Chapter 4, the result of the investigation on electroless copper deposition
for filling of damascene trenches was described.
The successful void-free filling, or
“superfilling”, has already been achieved by electrodeposition employing several bath
additives in combination.
These additives are essential for the bottom-up growth of
copper in trenches, which leads to superfilling.
As discussed in Chapter 1, the
bottom-up growth results from the combined effect of accelerating and inhibiting
additives.
Although this process achieved implementation of copper interconnects, a
few critical issues remain to be addressed, which include the need for a uniform
suputtered copper seed layer, and the problem of non-uniform current distribution on
the wafer.
Electroless copper deposition is expected to be useful for the fabrication of
ULSI copper interconnections as an alternative to electrodeposition because of its
excellent step coverage capability expected on high aspect-ratio and nanometer-scaled
3D structures and a good copper thickness uniformity achievable on large substrates.
The bottom-up growth of copper by electroless deposition has been demonstrated with
a plating bath containing some deposition inhibitors.
Shacham-Diamand et al.
[1,2]showed that the addition of Triton X-100 to a plating bath improves the filling
property.
Recently, Shingubara et al. [3-5]achieved void-free filling of high aspect
ratio via-holes with the addition of bis(3-sulfopropyl)disulfide (SPS).
The latter
authors reported that SPS acts as an inhibitor, and assumed that its consumption inside
the trenches brings about a relative decrease in the deposition rate at trench openings
compared to that at the trench bottom [3].
It is of interesting to note that the effect of
SPS on copper deposition in the electroless system is completely different from that in
the electrodeposition system.
The difference in the performance of this additive may
result from the effect of the reducing agent and complexing agent.
However, the
effect of SPS additive has been investigated also by other researchers [6,7], who
reported a different effect of this additive.
Lee et al. [7]reported that SPS accelerates
the deposition rate at very low concentrations.
Although void-free filling was
achieved with SPS, the deposition rate of their plating system was very low [3,7]. To
109
Chapter 4
resolve this problem while maintaining good filling property, we examined the concept
of combining accelerator and inhibitor, which is effective in the case of the
electrodesposition system.
Additive effects in electroless copper deposition have been investigated for
many years.
The investigations have been focused mainly on improving deposition
rate and bath stability.
Some additives [8-10] have been reported to increase the
deposition rate as well as stabilize the bath, while many stabilizers such as
2,2’-dipyridyl [11], thiourea [12], and some inorganic cyanide compounds
decrease
the
deposition
rate.
Among
them,
the
author
[13]
employed
8-hydroxy-7-iodo-5-quinoline sulfonic acid (HIQSA, Fig. 4.1) as an accelerator for
electroless copper deposition.
As was reported by Schoenberg [9], this additive
enhanced the deposition rate by as much as 50 percent, while it stabilized the bath.
The present study also employed polyethylene glycol (PEG) as a deposition inhibitor.
In this section, these additives were used in the conventional Cu(II)-EDTA
electroless plating bath, which contains formaldehyde as the reducing agent.
The
effect of combining HIQSA and PEG on the filling of trenches by electroless copper
deposition was investigated by microscopic observation of cross sections of
trench-patterned substrates and by deposition rate measurements on unpatterned
substrates.
Figure 4.1. Structure of 8-hydroxy-7-iodo-5-quinoline sulfonic acid (HIQSA).
110
Chapter 4
4.1.2 Experimental
The effect of the additives on electroless copper deposition in trenches was
investigated by observing trench-filling and by measuring the deposition rate on
unpatterned, flat substrates.
Bath composition
Table 4.1 lists the bath compositions used in this study.
All baths contained
0.04 mol/L of CuSO4, 0.04 mol/L of formaldehyde (HCHO), added as
paraformaldehyde,
0.08
mol/L
(Na4EDTA·4H2O), and additives.
of
tetrasodium
ethylenediamine
tetraacetate
The additives investigated in this study were SPS,
HIQSA, and PEG with an average molecular weight of 4000.
The baths were
◦
operated at 70 C and a pH of 12.5 with agitation provided by magnetic stirring and air
bubbling.
Table 4.1.
Bath compositions.
Chemical
Concentration
CuSO4・5H2O
0.04 mol L-1
HCHO
0.04 mol L-1
Na4EDTA·4H2O
0.08 mol L-1
SPS
0-1 ppm
PEG
0-2 ppm
HIQSA
0-5 ppm
Observation of trench filling
For trench-filling studies, SiO2/Si substrates with trench arrays were
employed.
The trenches were 200 nm wide and 500 nm deep.
A layer of tantalum
metal and a copper seed layer were sputtered onto the substrates.
The patterned
substrates were pretreated with 5 vol of H2SO4, rinsed in ultra pure water, and
immersed into a plating bath. After the deposition, the specimens were either cleaved
in liquid nitrogen or cut across the thickness using focused ion beam (FIB, FB-2100,
Hitachi).
The filling feature in cross-sectioned trenches was examined by a field
111
Chapter 4
emission scanning electron microscope (FE-SEM, S-4800, Hitachi).
Deposition rate measurements
The deposition rate of copper from these baths was measured gravimetrically
on a rolled copper foil (99.99
).
This substrate was activated by immersing into 5
vol of H2SO4 and rinsing with ultra pure water. Electroless copper deposition was
carried out for 15min.
112
Chapter 4
4.1.3. Observation of trench filling
Trench filling studies and deposition rate measurements were performed
with electroless copper plating baths containing various concentrations of additives.
First, the effect of SPS, which has been reported to achieve void-free filling [3-7], was
examined.
Then, the effects of HIQSA and PEG were investigated.
Effect of SPS
Figure 4.2 shows cross-sectional SEM images of copper deposits produced
in a bath without additives (additive-free bath), and a bath containing SPS (SPS bath).
The two specimens that yielded the images on the left were cleaved in liquid nitrogen,
while those that produced the images on the right were cut by FIB.
Both techniques
for the preparation of cross sections revealed the presence of voids within the trenches
of specimens plated in the additive-free bath (Fig. 4.2a).
Furthermore, on the surface
of copper deposits formed in the additive-free bath, large and small protrusions were
found to be distributed randomly.
The formation of voids in this bath is attributed to
the conformal deposition, which is one of the characteristics of electroless deposition,
and also to the surface roughness of the deposits.
As can be seen in the figure, the
voids observed with the cleaved specimen were larger than those found with the FIB
processed specimen.
The size of voids in the cleaved specimen was larger than that
in the FIB processed specimen also for the other baths investigated in this study.
One
presumable cause of this difference is that the copper deposit tends to rupture at
structural discontinuities such as voids and grain boundaries when it is cleaved.
The
high ductility of the copper deposit may also be relevant.
For the SPS bath, voids were also observed in the images of both cleaved
and FIB processed specimens (Fig. 4.2b).
The size of voids in the trenches was
relatively large compared with that for the additive-free bath.
voids existed also in the exterior of trenches.
Furthermore, some
The voids in the exterior of the trenches
are likely to be trapped hydrogen bubbles generated in the deposition reaction. Thus,
SPS did not achieve void-free filling under the experimental conditions used in this
study.
113
Chapter 4
(a)
(b)
300 nm
Figure 4.2. Cross sectional SEM images of copper deposited in trenches from
additive-free (a), and SPS (b) baths.
Concentration of SPS, 0.3 ppm.
Specimens
were cleaved in liquid N2 (photos on the left) or cut by FIB (photos on the right).
114
Chapter 4
Effect of HIQSA and PEG
Figure 4.3 shows cross-sectional SEM images of copper deposits obtained
from baths containing HIQSA (HIQSA bath), PEG (PEG bath), and both HIQSA and
PEG (HIQSA-PEG bath).
The filling feature in the image for the sample obtained
from the HIQSA bath (Fig. 4.3a) showed that voids were not eliminated by adding
HIQSA alone.
Voids were present in the mid portion of all trenches in the images of
both cleaved and FIB processed specimens.
The voids formed in the HIQSA bath
were identical in size to those formed in the additive-free bath.
HIQSA did not affect the filling property.
The concentration of
These results show that HIQSA, when
added alone, does not affect the filling property.
On the other hand, the filling property tended to be improved by the
addition of 2 ppm of PEG (Fig. 4.3b).
The number of voids found in the
cross-section of the specimen prepared with the PEG bath was less than that observed
with the specimen plated in the additive-free bath, although the difference was small.
The formation of protrusions on the surface was also suppressed by 2 ppm of PEG.
However, the filling feature deteriorated at higher PEG concentrations.
present in all trenches at 100 ppm of PEG (not shown in the figure).
Voids were
One possible
explanation for the improvement of filling property by PEG is that the decrease in PEG
concentration in the interior of the trenches results in a higher deposition rate inside the
trenches than at the trench openings.
It should be noted here that the substrate
pretreatment procedure carried out before the electroless deposition was the rinse in
ultra pure water.
In these small trenches employed in this study, the trenches must be
filled with rinse water. In the following process, in which the substrate is immersed
in the electroless plating bath, the water remaining in trenches should be displaced by
the electroless plating solution.
Because trenches were very small in diameter, the
mass-transfer from the bulk of the plating solution into the trenches must have been
greatly restricted.
Consequently, the concentration of all bath constituents were
considered to be essentially nil inside the trenches at the initial stages of filling process.
The increase in the concentration of PEG is assumed to be slower than that for the
other bath constituents, because this additive is a large polymer molecule.
Thus, the
concentration of PEG at the trench bottom must have been lower that that at the
opening during trench filling.
This assumption also explains the deterioration of
filling property at high concentrations of PEG.
Because the influence of
concentration gradient is considered to be insignificant at a high concentration of PEG,
115
Chapter 4
the conformal deposition, which results in void formation, can occur in trenches.
The addition of HIQSA to the PEG-containing bath completely suppressed
the formation of voids in the trenches (Fig. 4.3c).
No voids were observed in the
image of either cleaved or FIB processed specimen.
Furthermore, the surface of the
copper deposit formed in the HIQSA-PEG bath was also significantly smoother than
that of the deposit from the additive-free bath.
Because the addition of neither
HIQSA alone nor PEG alone achieved void-free filling, it is clear that the combined
effect of HIQSA and PEG was responsible for the observed improvement of the filling
property.
116
Chapter 4
(a)
(b)
(c)
300 nm
Figure 4.3. Cross sectional SEM images of copper deposited in trenches from HIQSA
(a), PEG (b), and HIQSA-PEG (c) baths.
and 3 ppm, respectively.
Concentrations of PEG and HIQSA were 2
Specimens were cleaved in liquid N2 (photos on the left) or
cut by FIB (photos on the right).
117
Chapter 4
4.1.4. Deposition rate measurements
To gain further insight into the combined effect of HIQSA and PEG, the
copper deposition rates of the baths containing various concentrations of HIQSA alone,
PEG alone, and both HIQSA and PEG were determined by performing blanket plating
of the substrate.
For the HIQSA (PEG-free) bath, the deposition rate was higher than
that of the additive-free bath at all HIQSA concentrations used in this study (Fig. 4.4).
The deposition rate was independent of the HIQSA concentration in the range of 1 to 5
ppm.
These results correspond to the observation that the effect of HIQSA
concentration was insignificant in the trench filling process.
The deposition rate
decreased significantly upon further addition of PEG to the HIQSA bath, although it
still tended to be higher than that of the PEG (HIQSA-free) bath (Fig. 4.5).
It is
important to note that copper deposition was accelerated by HIQSA only at very low
concentrations of PEG (lower than 1 ppm).
The deposition rate of the bath containing
3 ppm of HIQSA without PEG was 10.1 m/hr, which was higher than that of the
additive-free bath (6.48 m/hr). However, the deposition rate abruptly decreased
when a few ppm of PEG was added to the HIQSA bath. The result shows that the
acceleration of copper deposition by HIQSA is significant only at PEG concentrations
lower than 1 ppm, and it becomes insignificant in the presence of more than 1 ppm of
PEG.
On the other hand, for the bath without HIQSA, the extent of the decrease in
deposition rate with an increase in PEG concentration was relatively small.
These
results suggest that the improvement of the filling property observed in the presence of
both HIQSA and PEG is relevant to the observed acceleration of copper deposition
brought about by HIQSA at very low concentrations of PEG.
As noted earlier, the
diffusion rate of PEG, which is a large polymer molecule, is most likely to be much
smaller than that of HIQSA.
The slow diffusion rate should lead to a decrease in the
concentration of PEG in the interior of trenches, where the diffusion rate must be lower
than that at the opening.
Because the acceleration effect of HIQSA is significant only
at very low concentrations of PEG, the improvement of the filling property observed
with the addition of both HIQSA and PEG must be due to both the acceleration effect
of HIQSA at the bottom of trenches, where the concentration of PEG is low, and the
inhibition effect of PEG at trench openings.
From the results obtained in this study, it is concluded that the
combination of an accelerator and an inhibitor is highly effective for void-free filling
of trenches.
118
Chapter 4
Deposition rate / m hr
1
12
10
8
6
4
2
0
0
1
2
3
4
5
6
HIQSA concentration / ppm
Figure 4.4. Effect of HIQSA concentration on the rate of electroless copper deposition.
Deposition rate / m hr
1
12
10
8
6
4
2
0
0.5
1
1.5
2
2.5
PEG concentration / ppm
Figure 4.5. Effect of the combination of HIQSA and PEG on the rate of electroless
copper deposition.
Deposition rates were measured for the baths containing various
concentrations of PEG without HIQSA (■), and for the baths containing various
concentrations of PEG with 3 ppm of HIQSA (◯).
119
Chapter 4
4.1.5 Conclusions
The effect of HIQSA as an accelerating additive on the electroless copper
deposition in submicrometer trenches was investigated.
It was demonstrated that
adding HIQSA together with PEG to the plating bath improved trench filling properties.
Deposition rate measurements performed with the bath containing various
concentrations of HIQSA and PEG showed that the acceleration of copper deposition
by HIQSA was significant only at PEG concentrations lower than 1 ppm.
The
void-free filling achieved in the simultaneous presence of both additives is attributed to
the acceleration effect of HIQSA at the trench bottom combined with the inhibition
effect of PEG at the trench openings, the latter effect being due to the low diffusion
rate of PEG molecules into the narrow trenches.
120
Chapter 4
4.2 Void-free filling of trenches with electroless copper deposits
using glyoxylic acid as the reducing agent
4.2.1 Introduction
In the previous section (4.1), the author investigated superfilling with a
conventional electroless plating bath containing formaldehyde as the reducing agent as
a primary step of this study, and demonstrated that superfilling was achievable by
using the combination of 8-hydroxy-7-iodo-5-quinoline sulfonic acid (HIQSA) as an
accelerating additive and polyethylene glycol (PEG) as an inhibiting additive.
Although superfilling was achieved in this bath, some modification of the bath
composition was required for practical reasons.
Namely, this bath contained
formaldehyde as the reducing agent, which is undesirable for environmental reasons,
and it also contained a large amount of sodium ions from the pH adjusting agent added
as sodium hydroxide (NaOH), which is known to adversely affect properties of
transistors.
The present investigation was performed with a bath containing glyoxylic
acid [14,15] as an alternative to formaldehyde.
The all over all reaction of copper
electroless bath containing glyoxylic acid correspond to that contaning formaldehyde
[14,15].
Over all reactions of these baths are:
Glyoxiric acid bath:
Cu2+ + 2CHOCOOH + 4OH
Cu0 + 2HC2O4 +2H2O + H2
(4.1)
Formaldehyde bath:
Cu2+ + 2HCOH + 4OH
Cu0 + 2HCOO +2H2O + H2
(4.2)
Partial anodic oxidation reactions are:
Glyoxiric acid bath:
CHOCOOH + 3OH
HC2O4 +2H2O + 2e (E0 = +1.01 V)
Formaldehyde bath:
HCOH + 4OH
HCOO +2H2O + 2e (E0 = +1.07 V)
121
(4.4)
(4.3)
Chapter 4
In addition, the present study employed tetramethyl ammonium hydroxide
(TMAH) [16]as the pH adjusting agent instead of NaOH.
It was expected that the use of the different reducing agent might modify the
additive effects, because both anodic and cathodic partial reactions comprising a full
electroless deposition reaction are generally known to be affected by bath additives
[8,12].
Furthermore, other bath constituents might also affect partial as well as
overall electroless deposition reactions.
For example, Nakano et al. [16]reported that
the ease of electroless nickel deposition was strongly affected by replacing the alkaline
salts in the bath with ammonium salts.
Thus, the use of an alternative pH adjusting
agent was also thought to change the effect of additives on trench-filling in the present
study.
To clarify effect of these bath constituents on the bath performance, the
effects of PEG and HIQSA on the trench-filling of patterned substrates and the
deposition rate on unpatterned substrates were investigated.
The results were
compared with those obtained in the bath containing formaldehyde and sodium ions.
Figure 4.6. Structures of (a) glyoxylic acid and (b) formaldehyde.
122
Chapter 4
4.2.2 Experimental
The effect of PEG and HIQSA additives on electroless copper deposition in
trenches was investigated by observing trench-filling and by measuring the deposition
rate on unpatterned substrates.
Bath composition
Table 4.2 lists the compositions of three different baths, Baths A, B and C, used
in this study.
All baths were operated at 70 ºC and a pH of 12.5.
Table 4.2. Bath compositions.
Bath A
Bath B
Bath C
CuSO4・5H2O
0.04 mol L-1
0.04 mol L-1
0.04 mol L-1
Formaldehyde (HCHO)
0.04 mol L-1
0.04 mol L-1
-
-
-
0.08 mol L-1
0.08 mol L-1
0.08 mol L-1
0.08 mol L-1
12.5
12.5
12.5
Glyoxylic acid
EDTA*
pH
TMAH**
pH adjusting agent
NaOH
PEG*** (Mw 4000)
1 ppm
1 ppm
1 ppm
****
3 ppm
3 ppm
3 ppm
HIQSA
Bath temperature:
*
70 ºC
Ethylenediaminetetraacetic acid
**
***
****
TMAH
**
Tetramethyl ammonium hydroxide
Polyethylene glycol
8-Hydeoxy-7-iodo-5-quinoline sulfonic acid
For trench-filling studies, SiO2/Si substrates with trench arrays were employed.
The trenches were 130 nm wide and 350 nm deep.
A layer of tantalum metal and a
copper catalytic layer were sputtered onto the substrates.
The thickness of this Cu/Ti
layer on the side walls of trenches was approximately 35 nm.
The patterned
substrates were pretreated with 5 vol% H2SO4, rinsed in ultra pure water, and
immersed into the plating bath.
After copper deposition, the specimens were cut
across the thickness using focused ion beam (FIB, FB-2100, Hitachi).
123
The filling
Chapter 4
feature in cross-sectioned trenches was examined by a field emission scanning electron
microscope (FE-SEM, S-4800, Hitachi).
The deposition rate of copper from these baths was measured gravimetrically
on an unpatterned substrate with a Cu catalytic layer on a Ti adhesion layer, both of
which were formed by physical vapor deposition.
This substrate was activated by
immersing into 5 vol of H2SO4 and rinsing with ultra pure water before the
electroless copper deposition was carried out.
124
Chapter 4
4.2.3 Observation of trench filling
Trench filling studies were performed with Baths A, B and C without additives,
with PEG, and with both PEG and HIQSA.
The effects of PEG and HIQSA in these
three baths were examined by deposition rate measurement, which was carried out with
unpatterned substrates.
First experiment was carried out with Bath A containing PEG alone, HIQSA
alone, and PEG and HIQSA in combination.
The composition of Bath A is same as
that of the bath used in the previous study (section 4.1). To examine the efficiency of
these additives, trenches with 100 nm width and with aspect-ratio of 3, which is smaller
in size than those used in the section 4.1, were employed.
Figure 4.7 shows cross-sectional SEM images of copper deposits obtained from
baths without additives (additive-free bath, Fig. 4.7a), and with PEG alone (PEG bath,
Fig. 4.7b), and with both PEG and HIQSA (PEG-HIQSA bath, Fig. 4.7c).
Void-free
filling was achieved in the PEG-HIQSA bath (Fig. 4.7c), while voids were clearly
present in the specimens prepared in the additive-free (Fig. 4.7a) and PEG (Fig. 4.7b)
baths.
The results indicate that the combined addition of PEG and HIQSA is highly
effective for void-free filling of trenches.
The addition of neither PEG nor HIQSA
alone achieved void-free filling of trenches.
It is apparent that the combined effect of
PEG and HIQSA was necessary for the void-free filling.
Figure 4.7. Cross-sectional SEM images of copper deposited in trenches from (a)
additive-free, (b) PEG, and (c) PEG-HIQSA baths.
that of Bath A.
The basic bath composition was
Concentrations of PEG and HIQSA were 1 and 3 ppm, respectively.
Deposition time, 3min.
125
Chapter 4
Effect of pH adjusting agent
To examine the effect of pH adjusting agent, a trench-filling study was
performed with Bath B, which contained HCHO as the reducing agent and TMAH as
the pH adjusting agent.
The effects of PEG and HIQSA in this bath (Fig. 4.8) are
similar to those in Bath A (Fig. 4.7).
As can be seen in the images, voids were present
in the specimens from the additive-free (Fig. 4.8a) and PEG (Fig. 4.8b) baths.
On the
other hand, voids were not observed in the specimen prepared in the PEG-HIQSA bath
(Fig. 4.8c).
These results show that the combined effect of PEG and HIQSA appears
irrespective of which pH adjusting agent is used.
Figure 4.8. Cross-sectional SEM images of copper deposited in trenches from (a)
additive-free, (b) PEG, and (c) PEG-HIQSA baths.
that of Bath B.
The basic bath composition was
Concentrations of PEG and HIQSA were 1 and 3 ppm, respectively.
Deposition time, 3min.
126
Chapter 4
Effect of reducing agent
In contrast, the effect of the additives was found to depend strongly on the
reducing agent used in the bath. The cross sectional images of copper-filled trenches
formed in Bath C are shown in Fig. 4.9.
This bath contained glyoxylic acid as the
reducing agent and TMAH as the pH adjusting agent.
With this bath, superfilling was
achieved when PEG was added alone (Fig. 4.9b).
The effect of HIQSA was also
found to depend on the reducing agent.
For the PEG-HIQSA bath, trench filling was
incomplete after the deposition time of 3 min (Fig. 4.9c), whereas the trenches were
completely filled with copper after the same deposition time in the PEG bath (Fig.
4.9b).
Figure 4.9. Cross-sectional SEM images of copper deposited in trenches from (a)
additive-free, (b) PEG, and (c) PEG-HIQSA baths.
that of Bath C.
The basic bath composition was
Concentrations of PEG and HIQSA were1 and 3 ppm, respectively.
Deposition time, 3min.
127
Chapter 4
Effect of PEG concentreation on bottom-up growth (Bath C)
Copper was deposited in trenches with Bath C containing various
concentrations of PEG (Fig. 4.10). It is seen that the effectiveness of trench-filling
strongly depends on the concentration of PEG.
concentrations of PEG.
It deteriorated at the higher
As shown in Figs. 4.10c and 4.10d, voids were present in all
trenches filled with copper deposits from the bath containing more than 5 ppm of PEG.
Figure 4.10. Cross-sectional SEM images of copper deposited in trenches from (a)
additive-free bath, and (b-d) PEG bath. Concentrations of PEG were (a) 0, (b) 1, (c) 5,
and (d) 10 ppm, respectively. Deposition time, (a and b) 3 min, and (c and d) 15 min.
128
Chapter 4
Deposition rate on patterned substrates
Figure 4.11 shows the evolution with time of the filling feature in trenches
obtained in the bath containing 1 ppm of PEG. The result clearly shows that the
deposition at trench openings was very slow, while that at the bottom was much faster.
A significant change in thickness of copper deposit was found to occur
during the period between 30 sec and 1 min, while the extent of increase in thickness
of copper was very small during the first 30 sec.
Copper grew rapidly from the trench
bottom between 30 sec and 1 min (Figs. 4.11a and 4.11b), and the copper filling was
almost completed at 5 min (Fig. 4.11e).
After the trenches were filled with copper,
the extent of the increase in thickness of copper deposit with time became much
smaller (5-10 min).
Figure 4.11. SEM images showing the change with time of cross-sectional
characteristics of copper deposited in trenches from PEG bath. Concentration of PEG
was 1 ppm.
Deposition time, (a) 0.5 min, (b) 1 min, (c) 2 min, (d) 3 min, (e) 5 min,
and (f) 10 min, respectively.
129
Chapter 4
To estimate the difference in deposition rate at the trench opening and at the
bottom, the thickness of copper deposit at the trench opening (Topening) and that at the
bottom (Tbottom) were measured from cross sections shown in Fig. 4.11, and the results
were plotted against deposition time (Fig. 4.12a). Furthermore, the deposition rate at
the bottom was calculated from the slope of the deposit thickness vs. time curve shown
in Fig. 4.12a (Fig. 4.12b).
The plot shown in Fig 4.12b clearly shows that the deposition rate at the
bottom increased significantly during the period of 30sec to 1min.
The deposition rate
at 1 min was 16.9 m/hr. It is of particular interest to note that this deposition rate was
much higher than that for the additive-free bath measured on an unpatterned substrate
(11.0 m/hr). The rate of deposition at the bottom decreased gradually after 1 min and
continued to decrease until the trenches were completely filled with copper (5 min).
After that, the deposition rate remained constant at least for 5min, and the value of the
deposition rate during this period nearly the same as that of the PEG bath measured on
an unpatterned substrate (1.95 m/hr).
Contrary to the deposition behavior at the bottom, the increase in the
thickness of deposit at the trench openings (Topening) was essentially nil during the filling
process (Fig. 4.12a, 0-5 min).
After the trenches were filled with copper, the thickness
of copper at the openings started to increase.
During the period of 5 to 10 min, the
deposition rate at the opening was similar to that at bottom, and it was identical to the
deposition rate measured on an unpatterned substrate, which was 1.95 m/hr.
As described above, changes in deposition rate at the bottom and at the
opening of trenches during the filling process observed in this study were completely
different from each other.
This is considered to be brought about by the decrease in
concentration of PEG at the bottom of trenches, which is caused by the rinse water
remaining in trenches after the pretreatment process, as is discussed earlier (see section
4.1).
Because PEG is a large polymer compound, the diffusion rate of this molecule
must be low.
Therefore, the concentration of PEG at the trench bottom should be
lower than that at the opening.
We believe that the observed difference in the change
in deposition rate with time at the bottom and at the opening is due to a decrease in
concentration of PEG inside trenches where the mass transfer of species from the bulk
of the solution into trenches is greatly restricted.
130
Chapter 4
Figure 4.12. (a) Change with time of the thickness of copper deposits at bottom (Tbottom)
and at opening (Topening) of trenches, and (b) change with time of deposition rate at the
bottom of trenches.
The data were obtained from the SEM images.
131
Chapter 4
4.2.4 Deposition rate measurements
To gain insight into the effect of those additives on copper deposition,
deposition rate measurements were carried out on the unpatterned substrate.
Effect of pH adjusting agent
The rate of copper deposition measured in Bath A without additives
(additive-free bath) was 6.48 m hr-1, and it decreased by 15
when PEG was added,
while the addition of HIQSA increased the rate by more than 50
. The deposition
-1
rate for the PEG-HIQSA bath was 5.75 m hr , which was almost the same as that of
the PEG bath, which was 5.65 m hr-1. We reported previously that the acceleration
effect of HIQSA in the PEG-HIQSA bath was significant only at very low
concentrations, and that this concentration dependence of deposition rate is considered
responsible for achieving the void-free filling.
HIQSA is similar to that in Bath A.
In Bath B, the effect of PEG and
This result agreed well with the result of trench
filling studies.
Effect of reducing agent
For Bath C, the effects of these additives on the rate of copper deposition were
quite different from those in Baths A and B.
As shown in Fig. 4.13a, the inhibition
effect of PEG in Bath C was more significant than that in Baths A and B.
The rate of
copper deposition for Bath C containing PEG was 1.08 m/hr, which is as much as 85
lower than that of the additive-free bath (7.45 m/hr). It should be noted that this
deposition rate was nearly the same as that at the bottom and at the opening of trenches
measured in the PEG bath after the trench was filled with copper (Fig.4.12, 5-10 min).
The effect of HIQSA on the deposition rate of this bath was also different from that of
Baths A and B. The rate of deposition decreased by 18 with the addition of 3 ppm
of HIQSA, and it continued to decrease with an increase in HIQSA concentration.
The results of deposition rate measurements described above revealed that the effects
of both PEG and HIQSA depend strongly on the reducing agent, indicating that both
additives affect the reaction of the reducing agent.
The dependency of the additive effect on the reducing agent described above
132
Chapter 4
explains the results of trench filling studies.
Because the inhibition effect of PEG in
Bath C is more significant than that in Baths A and B, the difference in the rate of
copper deposition between the trench opening and the trench bottom, which is caused
by the decrease in PEG concentration at the bottom brought about by the slow
diffusion of PEG, should be more significant.
As a result, the sufficient bottom-up
growth was provided by adding PEG alone to Bath C.
not accelerate but decelerate the deposition.
On the other hand, HIQSA did
Consequently, the filling property was
not improved by the addition of HIQSA to the PEG bath in the case of Bath C, and the
rate of copper deposition within the trenches was decreased in this bath.
Figure 4.13. Effect of HIQSA and PEG on the rate of electroless copper deposition in
Bath C.
Deposition rate was measured for the baths containing various
concentrations of (a) PEG without HIQSA (solid line), and (b) PEG without HIQSA.
Open circles connected with a dashed line in Fig. 4.13a show the deposition rates for
Bath A.
133
Chapter 4
4.2.5 Conclusions
The effect of PEG and HIQSA on the electroless copper deposition in
trenches was investigated.
Effects of reducing agent and pH adjusting agent on the
behavior of the baths were also examined.
depended on the reducing agent.
The effect of the above additives strongly
Void-free filling was achieved by adding PEG alone
to the bath containing glyoxylic acid as the reducing agent, while the addition of both
PEG and HIQSA was necessary in the bath containing formaldehyde as the reducing
agent.
In the former bath, the significant bottom-up growth occurred after a few tens
of seconds of the incubation period.
The presence of the incubation period during the
trench filling process by electroless deposition is similar to the deposition behavior
which is known to be observed during the superfilling by electrodeposition.
After the
incubation period, the deposition rate at the bottom of trenches increased significantly.
On the other hand, the thickness at the opening was unchanged during filling process.
The deposition rate measurements carried out with unpatterned substrates
revealed that the performance of the bath depend strongly on the reducing agent used
in the bath, while the effect of pH adjusting agent is very small.
The inhibiting effect
of PEG in the bath containing glyoxylic acid as the reducing agent was much stronger
than that in the bath containing formaldehyde.
The significant inhibiting effect of
PEG in the former bath is attributed to the superfilling observed in this bath containing
PEG alone.
Furthermore, HIQSA, which increase the rate of copper deposition, did
not act as an accelerator in the glyoxylic acid bath.
result of trench filling.
134
The results also correspond to the
Chapter 4
References
[1] Y. Shacham-Diamand, and S. Lopatin, Microelectron Eng, 37-8, 77 (1997).
[2] V.M. Dubin, Y. ShachamDiamand, B. Zhao, P.K. Vasudev, and C.H. Ting, J.
Electrochem. Soc., 144, 898 (1997).
[3] Z.L. Wang, O. Yaegashi, H. Sakaue, T. Takahagi, and S. Shingubara, J.
Electrochem. Soc., 151, C781 (2004).
[4] Z.L. Wang, O. Yaegashi, H. Sakaue, T. Takahagi, and S. Shingubara, Jpn. J. Appl.
Phys. [1] Reg. Pap. Short Notes, 43, 7000 (2004).
[5] S. Shingubara, Z.L. Wang, O. Yaegashi, R. Obata, H. Sakaue, and T. Takahagi,
Electrochem. Solid-State Lett., 7, C78 (2004).
[6] C.H. Lee, S.C. Lee, and J.J. Kim, Electrochem. Solid-State Lett., 8, C110 (2005).
[7] C.H. Lee, S.C. Lee, and J.J. Kim, Electrochim. Acta, 50, 3563 (2005).
[8] M. Paunovic, and R. Arndt, J. Electrochem. Soc., 130, 794 (1983).
[9] L.N. Schoenberg, J. Electrochem. Soc., 119, 1491 (1972).
[10] F. Hsnns, Z.A. Hamid, and A.A. Aal, Mater. Lett., 58, 104 (2003).
[11] M. Oita, M. Matsuoka, and C. Iwakura, Electrochim. Acta, 42, 1435 (1997).
[12] A. Hung, J. Electrochem. Soc., 132, 1047 (1985).
[13] M. Saito, and H. Honma, Kinzoku Hyoumen Gijutsu, 29, 190 (1978).
[14] H. Honma, and T. Kobayashi, J. Electrochem. Soc., 141, 730 (1994).
[15] J. Darken, Trans. Inst. Metal Finish., 69, 66 (1991).
[16] H. Nakano, T. Itabashi, and H. Akahoshi, J. Electrochem. Soc., 152, C163 (2005).
135
Chapter 5
An electrochemical study on electroless
deposition
for
void-free
submicrometer trenches
filling
of
Chapter 5
5.1 Introduction
In Chapter 4, electroless copper deposition was studied, and it was found that
superfilling was achieved with PEG in the bath containing glyoxylic acid as the
reducing agent, while the combed addition of HIQSA and PEG was necessary in the
bath containing formaldehyde bath.
For the former bath, the deposition behavior at
the bottom and at the opening of trenches was found to be completely different during
the trench filling.
In this bath, copper deposition occurred significantly at the bottom
of trenches during trench filling process, while the deposition rate at the opening was
nil.
This interesting phenomenon is attributed to the difference in PEG concentration
at the bottom and at the opening of trenches.
In this chapter, to gain further insight into the mechanistic understanding of
PEG effect on copper electroless deposition in trenches, electrochemical measurements
such as a polarization measurement and a deposition potential measurement were
carried out.
The author especially focused the investigation on the deposition
potential during trench filling, and found that the deposition potential affect is very
important factor inducing the bottom-up growth.
139
Chapter 5
5.2 Experimental
The effect of PEG on electroless copper deposition was examined by
polarization measurements.
Deposition potential measurement was also performed.
The latter experiment was performed on both patterned and unpatterned substrates. To
obtain understanding of the mechanism of the effect of PEG on trench-filling, further
investigation was carried out by using the potential measuring apparatus with a Cu strip
electrode (Fig. 5.1).
Bath composition
Table 5.1 lists the compositions of the Baths, used in this study.
The basic
composition of the baths is identical to that of Bath C, described in the section 4.2
(Chapter 4). The baths were operated at 70 ºC and a pH of 12.5.
Table 5.1. Bath compositions.
Chemical
Concentration
CuSO4・5H2O
0.04 mol L-1
Glyoxylic acid
0.08 mol L-1
EDTA*
0.08 mol L-1
pH
12.5
pH adjusting agent
TMAH**
PEG*** (Mw 4000)
1 ppm
(0-1 ppm)
Bath temperature:
*
70 ºC
Ethylenediaminetetraacetic acid
**
***
Tetramethyl ammonium hydroxide
Polyethylene glycol
140
Chapter 5
Polarization measurement
Electrochemical polarization measurements were performed with the complete
electroless plating baths containing various concentrations of PEG.
Polarization
curves were recorded with a copper rotating disk electrode system (RRDE-1, Nikko
Keisoku) equipped with a computer-controlled measuring system (HZ-3000, Hokuto
Denko).
A Cu disk working electrode measuring 0.6 cm in diameter was polished
first with #2000 emery paper and subsequently with 0.06- m alumina paste on a sheet
of polishing cloth. A Pt counter electrode and an Ag/AgCl reference electrode were
placed with the working electrode in a three-electrode cell. The RDE was rotated at a
speed of 100 rpm.
-1
mV s .
The measurements were performed at the potential scan rate of 10
Before each experiment, the bath solution was aerated to maintain the bath
stability.
Deposition potential measurement
The effect of PEG on the deposition potential was investigated on both
patterned and unpatterned substrates.
The potential during immersion of the substrate
into the bath was measured with the same computer-controlled measuring system as
described above.
The results on patterned and unpatterned substrates were compared
and discussed.
To consider the effect of concentration distribution of PEG, which was
expected to be produced during trench filling, we also conducted another set of
experiments consisting of potential measurements.
set of experiments is illustrated in Fig. 5.1.
The apparatus constructed for this
Two electroless copper plating baths
(Baths 1 and 2) were prepared, placed in separate vessels, and connected to each other
through a salt bridge, which was a glass capillary filled with the basic bath solution
without additives.
These baths were also connected to an Ag/AgCl reference
electrode (Fig. 5.1). In this study, two kinds of setups, namely, Setups A and B, were
employed to examine the effect of location of the salt bridges.
In Setup A (Fig. 5.1a),
an Ag/AgCl reference electrode was connected to Bath 2 through the salt bridge.
On
the other hand, in Setup B (Fig. 5.1b), the reference electrode was connected to Bath 1.
A strip of Cu foil was employed as the working electrode. The two ends of this Cu
foil were immersed into Baths 1 and 2, respectively.
141
Measured amounts of an
Chapter 5
aqueous solution of PEG were added to these baths, while the potential of the Cu strip
electrode was monitored during the process.
(a)
(b)
Electrometer
Electrometer
Electrometer
Electrometer
Cu electrode
Cu electrode
RE
Bath 1
RE
Bath 2
Bath 1
Water bath (70 ºC)
Bath 2
Water bath (70 ºC)
Figure 5.1. Schematic illustration of the setups used for model experiments involving
potential measurements in electroless copper baths with different compositions and
activities: (a) Setup A and (b) Setup B.
142
Chapter 5
5.3 Polarization measurements
Electrochemical polarization measurements were carried out to investigate
the effect of PEG on the kinetics of eletroless copper deposition.
Electroless
deposition reactions in general have been investigated based on the mixed potential
theory [1], in which the oxidation of the reducing agent and the reduction of metal ions
are assumed to occur simultaneously on the substrate.
According to this theory,
superposition of polarization curves for the two partial reactions should yield the curve
for the complete electroless bath.
However, in the case of the Cu(II)-EDTA
electroless plating system with formaldehyde as the reducing agent, it has been
reported that the polarization curve obtained with the complete bath did not correspond
to the sum of the curves obtained separately for the anodic and cathodic partial
reactions [2-4].
Okinaka [2] compared the polarization curve for the complete bath
with the curve for the solution in the absence of formaldehyde and with that for the
solution in the absence of cupric ions.
It was found that the reduction of
Cu(II)-EDTA complex was greatly accelerated by the presence of formaldehyde, which
suggested the existence of a strong interaction between formaldehyde and the copper
complex.
Because glyoxylic acid undergoes an oxidation reaction similar to that of
formaldehyde [5], the oxidation of the former reducing agent is also expected to
interact with the reduction reaction of the Cu(II)-EDTA complex.
To investigate the
effect of the interaction between the reducing agent and the copper complex on
electroless copper deposition, electrochemical polarization measurements were
performed with a complete bath containing both glyoxylic acid and Cu(II)-EDTA
complex (Fig. 5.2, curve 1), a bath containing no Cu(II) complex (Fig. 5.2, curve 2),
and a bath containing no glyoxylic acid (Fig. 5.2, curve 3). These curves clearly
showed that glyoxylic acid greatly accelerated the reduction of Cu(II)-EDTA complex.
Furthermore, the anodic current peak at -0.5 V was also found to increase in the
presence of Cu(II)-EDTA complex.
These curves show that the reduction of
Cu(II)-EDTA complex is accelerated by glyoxylic acid, and that the oxidation of
glyoxylic acid is also accelerated in the presence of Cu(II)-EDTA complex.
In the
next experiment, therefore, the effect of PEG on electroless deposition was investigated
with the complete electoless bath in the presence of both the reducing agent and
Cu(II)-EDTA.
143
Chapter 5
Figure 5.2. Polarization curves for (1) the complete electroless copper bath containing
both glyoxylic acid and cupric ions, for (2) the bath containing no cupric ions, and for
(3) the bath containing no glyoxylic acid.
Figure 5.3a shows the polarization curves recorded in the complete baths
containing various concentrations of PEG.
The anodic and cathodic current peaks at
-0.45 V and -0.90 V vs. Ag/AgCl, are attributed to the oxidation of glyoxylic acid and
the reduction of Cu(II)-EDTA complex, respectively.
The current densities at both
peaks decreased with an increase in PEG concentration.
This indicates that both the
anodic and cathodic reactions were suppressed by PEG. The addition of PEG was also
found to affect the deposition potential. Figure 5.3b shows a plot of copper deposition
potential, which is a zero current potential in the polarization curve, against PEG
concentration.
The deposition potential in the additive-free (PEG-free) bath was
-0.67 V vs. Ag/AgCl.
It shifted in the negative direction with an increase in PEG
concentration in the range of lower than 0.5 ppm.
The deposition potential was
almost constant in the presence of more than 0.5 ppm of PEG, and the potential
measured in the bath containing 1 ppm of PEG was -0.75 V.
144
Chapter 5
Figure 5.3. (a) Polarization curves for Bath C with additions of (1) 0, (2) 0.1, (3) 0.3, (4)
0.5, and (5) 1 ppm of PEG.
(b) Effect of PEG on the deposition potential obtained
from polarization curves.
145
Chapter 5
5.3 Deposition potential measurements
The polarization measurement showed that the addition of a greater amount of
PEG shifted the potential for electroless copper deposition to a more negative value.
For trench filling, the concentration of PEG at the bottom is expected to be lower than
that at the opening.
Thus, the effect of PEG on the deposition potential within the
trenches is expected to be different from that observed on an unpatterned substrate.
Therefore, the potential of a trench-patterned substrate immersed into the additive-free
and PEG baths was measured (Fig. 5.4), and the result was compared with that
measured on an unpatterned substrate.
measured with a patterned substrate.
Curves a and b in Fig. 5.4 show the potential
For this measurement, the substrate was
immersed into the bath (0 sec), and its potential was measured continuously during the
period of copper deposition.
For the first 12 sec, the potential measured in the
additive-free bath (curve a) was less negative than -0.5 V, where copper deposition did
not proceed, while it shifted to more negative values with time during this period.
12 sec the potential abruptly shifted from -0.47 V to -0.67 V.
At
This initial period of 12
sec was regarded as the induction period involved in the overall electroless copper
deposition.
After the induction period was over, the deposition potential became
almost constant at -0.68 V, for at least 2 min.
Compared to this result found with the
additive-free bath, the potential-time curve obtained with a patterned substrate in the
bath containing 1 ppm of PEG (curve b) was quite different, although the results were
similar to each other for the initial induction period of 12 sec.
The potential of the
PEG bath also shifted significantly at 12 sec in the negative direction.
After the
induction period was over, the deposition potential for the PEG (1 ppm) bath shifted to
-0.7 V (curve b, 12 sec), which was similar to the deposition potential of the
unpatterned substrate in the additive-free (PEG-free) bath (curve c).
Unlike for the
additive-free bath, the potential for the PEG bath (curve b) continued to shift gradually
from -0.70 V to -0.76 V during the period of 12 to 70 sec, at which time it became
constant.
The potential measured with the patterned substrate after 70 sec was -0.76
V (curve b).
It should be noted that this potential was essentially identical to that
measured on the unpatterned substrate in the PEG (1 ppm) bath (curve d).
The
time-dependence of the potential of the unpatterned substrate (curves c and d) in both
additive-free and PEG baths was different from that of the patterned substrate (curves a
and b).
The induction period of initial 12 sec observed with the patterned substrate
was not observed with the unpatterned substrate.
The potential of the unpatterned
substrate in the additive-free bath (curve c) and that in the PEG (curve d) bath shifted
146
Chapter 5
significantly in the negative direction within 1.5 sec.
After this period of time, the
potential values in the two baths reached -0.7 V and -0.76 V, respectively, after which
these values remained steady and independent of time.
It is important to note that the
potential shift after the induction period observed with the patterned substrate (curve b,
12-70 sec), was not found with the unpatterned substrate.
Both the presence of
induction period and the negative shift of the potential for the PEG bath observed after
the induction period were characteristic of the patterned substrate.
Thus, these
phenomena are attributable to the difference between the conditions of mass transport
from the bulk of the solution to the inside of trenches and those to the surface of the
unpatterned substrate.
The observed change in the potential in the PEG (1 ppm) bath measured with
the patterned substrate is attributed to the change in PEG concentration in trenches
during the initial diffusion process of PEG as discussed in the preceding chapter
(Chapter 4).
For the experiments, the patterned substrate, which had just been
activated with H2SO4 and rinsed with ultrapure water, was immersed into the bath.
It
is important to noted here that the substrate-pretreatment procedure immediately before
the electroless deposition was the rinse in water. Because the trenches employed in
this study were very small (130 nm-wide and 350 nm-deep), the trenches must have
been filled with rinse water at the initial stages.
The rinse water in the trenches must
have been replaced by the electroless plating solution before the electroless plating
began to take place.
The increase in the concentration of PEG, which is a large
polymer molecule, is assumed to be slow inside the trenches, where the mass transfer
of bath constituent species is greatly restricted. Thus, the concentration of PEG at the
trench bottom must have been lower than that at the opening during trench filling.
Therefore, the observed difference in the deposition behavior at the bottom and at the
opening (Fig. 4.12) is believed to be due to the difference in PEG concentration at the
two different locations.
147
Chapter 5
Figure 5.4. Change with time of the potential of (a and b) a patterned substrate and (c
and d) an unpatterned substrate during electroless copper deposition.
Measurements
were carried out in (a and c) additive-free bath, and (b and d) PEG bath.
Concentration of PEG, 1 ppm.
To examine the effect of the rinse water remaining in trenches on electroless
deposition, the potential measurements were performed with the substrates pretreated
with the additive-free bath at 20 °C after the normal pretreatment in H2SO4 and water
rinse.
Figure 5.5 compares the potential-time curves measured in the additive-free
bath with patterned substrates which were subjected to the different pretreatment
148
Chapter 5
process of immersion in the additive-free bath at 20 °C, as an intermediate extra step,
between the water rise and the immersion in the additive-free bath at 70 °C.
The
immersion times in the 20 °C additive-free bath were changed from 0 to 60 sec (curves
a-c), respectively.
It should be noted that no copper deposition took place during this
intermediate extra step because the temperature of the bath is too low to start
electroless copper deposition.
The results clearly show that the length of the
induction period involved in the overall electroless copper deposition decreased with
an increase in the time of immersion of the patterned substrate in the 20 °C bath.
This
result shows that bath constituents must first diffuse into the water-filled trenches
before the copper deposition begins to proceed. The longer induction period observed
in the potential measurement with the patterned substrate (Fig. 5.4), as compared with
that for the unpatterned substrate, is attributed to the presence of rinse water remaining
in the trenches of the patterned substrate after the pretreatment process.
Figure 5.5. Effect of rinse water on the induction period in electroless copper deposition
in trenches. The immersion times in the additive-free bath (20 °C) before the
measurement: (a) 0 sec, (b) 20 sec, and (c) 60 sec.
149
Chapter 5
To understand the effect of concentration distribution of PEG on a substrate on the
copper deposition potential in more detail, we carried out a series of simple model
experiments using the apparatus illustrated in Fig. 5.1a (Setup A), where the reference
electrode was connected to Bath 2 through a salt bridge.
1 and 2) were prepared for the experiment.
Two PEG-free baths (Baths
Each end of a Cu foil electrode was
immersed into either of the two baths (0 sec), and the potential of each end of the foil
was measured against the reference electrode (The duration in which these potentials
were measured is designated as the “first period”). For the second period, an aqueous
solution of PEG was added to Bath 1 (45 sec) to make the PEG concentration in the
bath equal to 1 ppm.
For the third period, the same amount of the PEG solution was
added to Bath 2 (70 sec).
The variations of electrode potential during these periods
are shown in Figure 5.6a.
Figure 5.6. Change with time of the potential of a copper strip electrode with two ends
immersed into Baths 1 and 2, respectively.
Measurements were performed with (a)
Setup A and (b) Setup B (see Figs. 5.1a and 5.1b).
150
Chapter 5
When neither bath contained PEG (first period), the deposition potential was
approximately equal to -0.70 V, which was identical to the deposition potential for the
additive-free bath (Fig. 5.4c).
sec.
The PEG solution was then added only to Bath 1 at 45
In view of the observed effect of PEG on the deposition potential shown in Fig.
5.3b, the potential of the Cu foil electrode was expected to change upon the addition of
PEG to Bath 1.
On the contrary, however, the potential remained unchanged when
PEG was added only to Bath 1 (the second period).
On the other hand, when PEG
was added also to Bath 2 (70 sec), the deposition potential gradually shifted in the
negative direction to -0.76 V, which was almost identical to the deposition potential
found with the bath containing 1 ppm of PEG. In the experiment performed with
Setup A, described above, the potential of Cu strip electrode was measured against the
reference electrode, which was connected to Bath 2 through the salt bridge.
To
examine the effect of the location the salt bridge, the same measurement was
performed with Setup B, where the reference electrode was connected to Bath 1 (Fig.
5.1b).
The procedure of the potential measurement was almost the same as that
performed with Setup A (Fig. 5.6a).
For this measurement with Setup B, two
PEG-free baths (Baths 1 and 2) were prepared, and each end of a Cu foil electrode was
immersed into either of the two baths (0-85 sec, the “first period” in Fig. 5.6b).
At 85
sec, PEG aqueous solution was added to Bath 1 (85 sec-130 sec, the “second period” in
Fig. 5.6b).
The result of this measurement (Fig. 5.6b) clearly shows that the potential
was unchanged with the addition of PEG to Bath 1, and that the potential was kept at
-0.7 V, which was identical to the deposition potential measured in the additive-free
bath with an unpatterned substrate.
From the results, the potential of the Cu strip
electrode was unaffected by the location of the salt bridge.
Therefore, it is considered
that the copper deposition at an electrode with two distinctly separated areas, i.e., with
one area contacting the PEG-free bath and the other area contacting the
PEG-containing bath was controlled by the reaction which occurs at the interface
between the copper surface and the PEG-free bath.
This result also suggests that the
deposition occurs at the potential corresponding to the deposition potential assumed by
that portion of the substrate where the deposition potential was less negative.
The shift of potential observed on a patterned substrate during the progress
of the electroless deposition reaction is considered to result from the decrease in PEG
concentration within the trenches, which, in turn, is caused by the slow diffusion of
these species from the bulk of the solution.
The concentration of PEG, which is a
large polymer molecule, is expected to decrease toward the trench bottom, where the
151
Chapter 5
diffusion of PEG from the bulk of the solution is restricted for geometrical reasons.
As suggested by the results shown in Fig. 5.6, the deposition potential is controlled by
the concentration of PEG at that portion of the electrode which is in contact with the
solution where PEG concentration is the lowest. If the concentration gradient of PEG
thus produced in trenches is taken into consideration, the deposition potential during
trench filling is expected to be determined by the PEG concentration at the bottom of
the trench.
The diffusion of PEG from the bulk of the solution into trenches and the
change in the feature of trench filling during the copper deposition appear to lead to an
increase in the concentration of PEG at the bottom.
As a result, the difference in
deposition rate at the bottom and at the opening should become less significant with
time.
The effect of PEG on trench filling is discussed based on the results of potential
measurements as follows.
If the concentration of PEG at the bottom is nil, while that
at the opening is 1 ppm, the deposition potential during trench filling should correspond
to that of the additive-free (PEG-free) bath, which is approximately -0.7 V.
Because
this potential is less negative than that measured on an unpatterned substrate in the bath
containing 1 ppm of PEG (-0.75 V), the deposition rate at the trench openings in the
PEG (1 ppm) bath should be lower than that on an unpatterned substrate in the same
bath.
Indeed, the deposition rate at this potential measured in the PEG bath was 0.72
-1
m hr , which was lower than that at the deposition potential of the PEG (1 ppm) bath,
which was 1.08 m hr-1. This assumption was supported by the fact that the effect of
PEG on the filling properties strongly depends on its concentration in the bath, which is
shown in Fig. 4.10 (Chapter 4).
The optimum concentration of PEG which achieved
the superfilling was 1 ppm, which was in the range where the PEG concentration
dependence of its inhibition effect and its effect on deposition potential was most
significant.
Thus, it is concluded that the inhibition effect of PEG and the shift of
potential, determined by the local PEG concentration at the trench bottom, are the most
important factors controlling the achievement of superfilling.
152
Chapter 5
5.4 Conclusions
Electrochemical measurements were performed with the bath containing
glyoxylic acid to investigate the effect of PEG on trench filling.
The addition of PEG
shifted the deposition potential in the negative direction, and it decreased the
deposition rate significantly.
The higher deposition rate at the bottom of trenches
than that at the opening, observed during copper filling in the bath containing PEG,
was attributed to the decrease in PEG concentration at the bottom, which results from
the rinse water remaining in the trenches before the substrate was immersed in the bath.
The effect of PEG on the deposition potential during trench filling was discussed by
taking into account the local concentration distribution inside the trenches.
The
potential measurement carried out with a Cu strip electrode, each end of which was
immersed in either of the two different baths with and without PEG connected through
a salt bridge, revealed that the measured potential was identical to that assumed by a
piece of copper immersed in a separate, additive-free bath.
These results suggest that
the deposition potential measured with a copper electrode consisting of portions
immersed in a solution with locally different activities, assumes the value
corresponding to the deposition potential of a copper electrode immersed in the
solution with the highest activity.
Under the conditions of trench filling, the potential
of copper deposition is controlled by the local PEG concentration at the trench bottom,
where the concentration of PEG is expected to be the lowest.
It is concluded that the
superfilling achieved by the addition of PEG is brought about by the shift of deposition
potential resulting from the low PEG concentration at the trench bottom, combined
with the inhibition effect of PEG at the trench opening.
153
Chapter 5
References
[1] C. Wagner, and W. Traud, Z. Elektrochem., 44, 391 (1938).
[2] Y. Okinaka. and T. Osaka, in Advances in Electrochemical Science and
Engineering, Vol. 3, Edited by H. Gerischer and C.W. Tobias, pp 87, VCH Publishers
Inc., York, NY, (1994).
[3] B.J. Feldman, and O.R. Melroy, J. Electrochem. Soc., 136, 640 (1989).
[4] H. Wiese, and K.G. Weil, Ber. Bunsen-Ges. Phys. Chem. Chem. Phys., 91, 619
(1987).
[5] H. Honma, and T. Kobayashi, J. Electrochem. Soc., 141, 730 (1994).
154
Chapter 6 General conclusion
Chapter 6
In Chapter 6, the results obtained in this research were summarized and the
prospect of the electrochemical deposition process for micro- and nano-fabrication is
discussed.
In this thesis, the electrochemical deposition (electrodeposition and
electroless deposition) of copper for the fabrication of ULSI interconnects was
investigated.
The effect of bath additives successfully achieves the precise control of
feature of copper growth in submicrometer trenches both in electrodeposition and
electroless deposition systems.
The additives were also found to provide significant
effects on the physical and mechanical properties of films.
The results obtained from
the research described in Chapters 2, 3, 4, and 5 are summarized.
I. An electrochemical study on the effect of bath additives on copper electrodeposition
in submicrometer trenches (Chapter 2)
Effects of the conventional bath additives (chloride ions (Cl–), polyethylene
glycol (PEG), bis(3-sulfopropyl)disulfide (SPS), and Janus Green B (JGB)) used in the
Damascene process on the filling of submicrometer trenches with electrodeposited
copper were investigated by means of electrochemical polarization measurement and
cross sectional microscopy. The addition of Cl–, PEG, and SPS in combination provide
bottom-up growth, while JGB inhibit the overfill copper-bumps.
The combination of Cl– and PEG inhibited copper deposition in the areas of
opening of the trenches, while SPS accelerated it at the bottom. Polarization curves
showed that the degree of acceleration of copper deposition by SPS increases with the
concentration of SPS. This SPS concentration-dependent acceleration accounts for the
observed bottom-up growth.
The addition of JGB was found to inhibit copper deposition at the later
stages of the filling process, leading to the suppression of overfill phenomenon,
although the bottom-up growth was also inhibited at high JGB concentrations. Bath
agitation significantly enhanced the inhibition effect of JGB on the overfill
phenomenon, without disturbing the bottom-up growth. These results suggest that the
dependence of the inhibition effect of JGB on its mass-transfer condition is responsible
for the significant inhibition of copper deposition only at the outside of trenches,
resulting in the leveling of overfill bumps without disturbing bottom-up growth.
157
Chapter 6
II. Effect of bath additives on the physical and mechanical properties of copper
electrodeposits (Chapter 3)
(i) Effect of bath additives on room-temperature recrystallization of copper
electrodeposits
The effect of Cl , PEG, SPS, and JGB on the room-temperature
recrystallization, or “self-annealing”, of copper electrodeposits was investigated.
Self-annealing occurred in the deposit obtained from the bath containing Cl PEG, and
SPS in combination. The deposits in which self-annealing was found to occur
consisted of very small grains in the as-deposited state, while those which did not
undergo self-annealing consisted of coarse large grains.
The small grain size was
considered to be a major factor inducing self-annealing.
The carbon analysis of the
deposit obtained in the presence of JGB indicated that this compound or its
decomposition product is incorporated in the deposit, and that self-annealing is
inhibited by this incorporated compound.
(ii) Effect of room-temperature recrystallization on ductility of copper electrodeposits
The effect of the microstructure evolution, which takes place upon
self-annealing was investigated.
During the period of self-annealing, the ductility
was found to increase by a factor of 1.5. The increase in ductility is shown to be
relevant to a change in microstructure of the copper deposits.
III Void-free filling of submicrometer trenches by electroless copper deposition
(Chapter 4)
(i) Void-free trench filling of trenches with electroless copper deposits using a
combination of accelerating and inhibiting additives
Electroless copper deposition was performed on submicrometer-trench
patterned substrates with a bath containing formaldehyde as the reducing agent, using
some additives.
Void-free copper filling of trenches was achieved by the addition of
both 8-hydroxy-7-iodo-5-quinoline sulfonic acid (HIQSA) as an accelerating additive
and polyethylene glycol (PEG) as an inhibiting additive at specific concentrations.
158
Chapter 6
Copper deposition rate measurements revealed that HIQSA accelerated the deposition
only when it was added together with a very low concentration of PEG.
The void-free
filling is considered to have resulted from the significant acceleration brought about by
HIQSA at the trench bottom, where the concentration of PEG is low.
The decrease in
the concentration of PEG inside trenches is considered to result from the rinse water
remaining in the trenches before the substrate was immersed in the electroless plating
bath.
(ii) Void-free trench filling of trenches with electroless copper deposits using glyoxylic
acid as the reducing agent
Filling of trenches by electroless copper deposition was investigated in the
bath containing glyoxylic acid as the reducing agent.
The additive effect was found
to depend strongly on the reducing agent used in the bath.
Void-free trench-filling
was achieved by using polyethylene glycol (PEG) as an inhibiting additive in the bath
containing glyoxylic acid as the reducing agent, while the combined addition of
8-hydroxy-7-iodo-5-quinoline sulfonic acid (HIQSA) and PEG was necessary for
achieving void-free filling in the bath containing formaldehyde as the reducing agent.
IV An electrochemical study on electroless deposition for void-free filling of
submicrometer trenches (Chapter 5)
The effect of PEG on trench filling was studied in detail based on deposition
potential and electrochemical polarization measurements.
The addition of PEG shifted the deposition potential in the negative direction.
It is suggested that the deposition potential depends on the local concentration of PEG
at the trench bottom where it is expected to be low.
The deposition potential during
trench filling was found to depend on this PEG-concentration gradient in trenches, and
the effect of PEG on the deposition potential is attributed to the significant inhibition at
the opening of trenches.
The additives are one of the most important factors controlling the feature
and kinetics of deposition, and the microstructure and various properties of film.
For
the filling of trenches of copper by electrodeposition (Chapter 2), the diffusion of
159
Chapter 6
additives from the bulk of the solution into the trenches and/or the change in the
surface area at the trench bottom during deposition is found to be very important to
consider the bottom-up growth.
The effect of these factors will be much more
important to fill the further miniaturized structures.
The diffusion of species must be
a critical issue when the size of trenches reaches at several nanometers, which is
similar to the molecular size of the bath constituents. The impact of the change in the
surface area at the trench bottom is also expected to become more significant with an
increase in the aspect ratio.
However, as was noted in Chapter 1, the most critical
issue for miniaturization is the discontinuity and non-uniformity of barrier and seed
layers, sputtered onto the trenches.
Therefore the author believes that electroless
deposition is a potential candidate for the alternative to the current process (Chapter 4).
It was demonstrated that superfilling by electroless copper deposition was achieved by
adding PEG.
The particularly interesting facts found in this study were that the effect
of additives depends strongly on the reducing agent used in the bath, and that the
deposition behaviour at the bottom and at the opening of trenches during superfilling
was completely different.
The detailed analysis for these phenomena will contribute
to the mechanistic understanding of electroless deposition.
The results of potential
measurements in Chapter 5, shows an interesting aspect of electroless deposition.
Model potential measurements using a Cu strip electrode, which was performed to
examine the effect of PEG concentration gradient on deposition potential, revealed that
the deposition potential is determined by the deposition potential on the portion of
substrate where the local PEG concentration is the lowest when the difference in the
concentration of PEG exists on the surface of the substrate. The author considers that
the findings provide the important knowledge for understanding the fundamental
mechanism of electroless deposition and that the method of potential measurement is
useful for analyzing the effect of additives on electroless deposition in trenches.
The author believes that electrochemical deposition is more useful for
fabricating complicated nanostructures than the other techniques such as CVD and
ALD, because the former technique achieves a simple, low-cost, and high-throughput
process.
However, the author also considers that combination of the electrochemical
process and these techniques or their concepts will yield a new effective process,
which may expand the capability of electrochemical deposition process.
Moreover, electrochemical deposition is applied to the processes of the
fabrication of micro electromechanical systems (MEMS), intensively, in these days.
160
Chapter 6
In these processes, the desired size of the structure is ranges from submicrometer to
some millimetres.
Therefore, the further development of electrochemical techniques
will be essential in these fields.
The fundamental understanding of electrochemical
deposition reaction is also expected to be much important to control the deposition,
simultaneously.
For the further understanding of the system, the development of
analysis method is needed.
Therefore, the author considered that the potential
measurement system developed in the present study (shown in Fig. 5.1, Chapter 5) is a
significant challenge for analysing the electroless deposition system.
Electrochemical techniques such as electrodeposition and electroless
deposition using bath additives are highly effective for fabricating metal nanostructures
such as interconnects in ULSI.
To control the deposition behaviour, the fundamental
understanding and analysis method of the mechanism of the additive effect is very
important. The author believes that the mechanistic understanding of electrochemical
reaction which includes crystallization phenomenon during deposition and molecular
interactions of additives and other compounds will provide the further improvement of
micro and nanofabrication process by electrochemical deposition as well as the
contribution to the development of theory of electrochemistry.
161
List of achievements
List of achievements
1. Original articles
“Evidence for “superfilling” of submicrometer trenches with electroless copper
deposit”
M. Hasegawa, N. Yamachika, Y. Shacham-Diamand, Y. Okinaka, T. Osaka
Appl. Phys. Lett., 90, 101916 (2007).
“An electrochemical investigation of additive effect observed in trench-filling of ULSI
interconnects by electroless copper deposition”
M. Hasegawa, N. Yamachika, Y. Okinaka, Y. Shacham-Diamand, T. Osaka
Electrochemistry, accepted for publication.
“Void-free trench-filling by electroless copper deposition using the combination of
accelerating and inhibiting additives”
M. Hasegawa, Y. Okinaka, Y. Shacham-Diamand, T. Osaka
Electrochem. Solid-State Lett., 9, C138-C140 (2006).
“Enhancement of the ductility of electrodeposited copper films by room-temperature
recrystallization”
M. Hasegawa, Y. Nonaka, Y. Negishi, Y. Okinaka, T. Osaka
J. Electrochem. Soc., 153, C117-C120 (2006).
“Effects of additives on copper electrodeposition in submicrometer trenches”
M. Hasegawa, Y. Negishi, T. Nakanishi, T. Osaka
J. Electrochem. Soc., 152, C221-C228 (2005).
163
List of achievements
“Effect of bath additives on the room-temperature recrystallization of copper
electrodeposits”
M. Hasegawa, Y. Negishi, Y. Okinaka, T. Osaka
J. Electrochem. Soc., submitted.
“Application of electroless plating technique in the fabrication of ULSI interconnects”
T. Osaka, M. Yoshino, M. Hasegawa
Electrochemical Society Transactions, submitted.
2. Oral Presentations
A. International Conference (Presented by M. Hasegawa et al.)
“Effect of polyethylene glycol as an inhibiting additive on electroless copper
deposition in submicrometer trenches”
The 57th Annual Meeting of the International Society of Electrochemistry (ISE),
Edinburgh, UK, September, 2006.
“Effect of additives on room-temperature recrystallization of copper electrodeposits”
Electrochemical Micro and Nano Systems Technologies 2006, Germany, August 2006.
“Effect of additives on microstructure and self-annealing of copper electrodeposits”
2006 Gordon Research Conference on Electrodeposition, New London, USA, July
2006.
“Effect of the combination of PEG and HIQSA additives on electroless copper
deposition for ULSI interconnection”
Advanced Metallization Conference 2005: 15th Asian Session (ADMETA 2005),
Tokyo, Japan, October, 2005.
164
List of achievements
“Ductility of electrodeposited copper films: Effects of bath additives and current
waveform”
The 56th Annual Meeting of the International Society of Electrochemistry (ISE),
Busan, Korea, September, 2005.
“Effect of accelerating additive on electroless Cu deposition in submicrometer
trenches”
The 56th Annual Meeting of the International Society of Electrochemistry (ISE),
Busan, Korea, September, 2005.
“Effect of additives on copper electrodeposition of for ULSI interconnection”
The 2nd 21COE International Symposium on 'Practical Nano-Chemistry', Tokyo,
Japan, December 2004.
“Effect of accelerator on copper electrodeposition kinetics”
The 5th International Symposium on Electrochemical Micro & Nano System
Technologies, Tokyo, Japan, September 2004.
“Mass-transfer effect of Janus Green B on copper electrodeposition in submicrometer
trenches”
206th Meeting of The Electrochemical Society, Honolulu, USA October 2004.
“Effects of additives on copper electrodeposition in submicrometer trenches”
204th Meeting of The Electrochemical Society, Orland, USA, October 2003.
B. Domestic Conference (Presented by M. Hasegawa et al.)
“Effect of additives on self-annealing of copper electrodeposits”
The 115th Conference of The Surface Finishing Society of Japan, October 2006.
165
List of achievements
“Electroless copper deposition in submicrometer trenches ―An electrochemical study
of additive effects”
The 72th Conference of The Electrochemical Society of Japan, April 2005.
“Effect of additives on copper superfilling of submicrometer trenches”
International Symposium on Molecular Nano-Engineering and its Development into
Microsystems: Semiconductor Nanotechnology, December 2004.
“Effect of additives on trench filling of copper electrodeposits for ULSI applications”
The 106th Conference of The Surface Finishing Society of Japan, September 2006.
“Fundamental
analysis
on
copper
filling
of
submicrometer
trenches
by
electrodeposition”
The 105th Conference of The Surface Finishing Society of Japan, March 2002.
3. Patent
Application No. 2006-222021
“Electroless copper plating bath, and the processes of electroless copper deposition and
fabrication of copper interconnects”
T. Osaka, M. Hasegawa
Waseda University, August 2006.
Application No. 2005-323127
“Electroless copper plating bath, and the processes of electroless copper deposition and
fabrication of copper interconnects”
T. Osaka, M. Hasegawa
Waseda University, November 2005.
166
Acknowledgements
The present study was carried out at the Department of Applied Chemistry,
Graduate School of Science and Engineering, Waseda University.
The work of this thesis could not have been accomplished without the
supported and help of friends and colleague in Waseda University, to whom I should
like to express my gratitude.
Sincere thanks go to my supervisor, Professor Dr. Tetsuya Osaka, for his
invaluable help, guidance, encouragement and a great deal of tolerance during all these
days.
I should also like to express my sincere gratitude to Professor Dr. Takayuki
Homma for his valuable suggestion and encouragement.
I offer Dr. Okinaka my
deepest thanks for his kind help, continuous encouragement, and useful advice.
I
wish to express my sincere acknowledgement to Professor Dr. Yosi Shacham-Diamand
for his help and useful advice.
for his kind help and advice.
Many thanks are due to Professor Dr. Itsuaki Matsuda
I am grateful to Professors Dr. Kazuyuki Kuroda and Dr.
Yoshiyuki Sugahata for their kind advice.
I should also express my sincere gratitude to Associate Professor Dr. Takuya
Nakanishi and Dr. Tokihiko Yokoshima for their inspiring and fruitful discussion.
I
also wish to thank Professor Dr. Jiro Hokkyo, Professor Dr. Toru Asahi, and Associate
Professor Dr. Toshiyuki Momma for their encouragement and valuable comments.
Special thanks are due to Dr. Junji Sasano, Dr. Daisuke Niwa, Dr. Jun Kawaji, Dr.
Hirotaka Sato, Dr. Shinji Motokawa, Dr. Takahiro Shimizu, Dr. Junichi Sayama, and
Dr. Hitomi Mukaibo for their kind help and advice.
Also, deep and heartfelt gratitude
is offered to my senior and junior colleagues, Mr. Seiichi Nakamura, Mr. Yuichi
Nonaka, Mr. Yoshinori Negishi, Ms. Sawako Taki, Mr. Yuji Ikura, Ms. Izumi
Kawakita, Ms. Emiko Kazuma, Mr. Noriyuki Yamachika for their millions of
encouragements and assistance.
Many thanks are due to all members of Professor Osaka and Professor
Homma’s laboratory for their kind help during the course of this thesis.
Finally I express greatest thanks to my parents, Mr. Yoichi Hasegawa and
Mrs. Shoko Hasegawa, for deep affection and constant financial support.
Madoka Hasegawa
167