NOVEL ALKALINE COPPER
ELECTROPLATING PROCESSES FOR
APPLICATIONS IN INTERCONNECT
METALLIZATION
by
ANIRUDDHA AHMED JOI
Submitted in partial fulfillment of the requirements for the
degree of Doctor of Philosophy
Dissertation Adviser: Professor Uziel Landau
Department of Chemical Engineering
CASE WESTERN RESERVE UNIVERSITY
August, 2013
CASE WESTERN RESERVE UNIVERSITY
SCHOOL OF GRADUATE STUDIES
We hereby approve the thesis/dissertation of
Aniruddha Ahmed Joi
Candidate for the
Doctor of Philosophy
Professor Uziel Landau
degree*.
.
Professor Rohan Akolkar
.
Professor Robert Savinell
.
Professor Christian Zorman
(date)
.
May 20, 2013
.
.
*We also certify that written approval has been obtained for any proprietary
material contained therein.
2
To my parents, Zahirul Haque and Farida Yasmin for their tireless support. I would also
like to acknowledge Prof. Landau and Prof. Akolkar for their guidance and mentorship
during the last four years. To my brothers, Anupam and Aditya, who although were of
no help whatsoever, have always been there. Finally, to all my friends and colleagues in
Cleveland for the occasional distractions from work and keeping things in perspective.
3
Table of Contents:
List of Tables……………………………………………………………………….7
List of Figures…………………………………………………………………...….8
Abstract……………………………………………………………………………19
Chapter 1: Introduction……………………………………………………………21
Chapter 2: Experimental Methodology……………………………………………29
2-1: General Electrochemical Setup……………………………………….29
2-2: Direct Deposition of Cu onto Ru……………………………………...29
2-3: Additives Characterization in Alkaline Media………………………..30
2-4: Bottom-up Fill in Alkaline Media in the presence of Additives….…..31
2-5: Galvanostatic Pulse Plating of Copper Germanide…………………...32
2-6: Materials Characterization of Cu3Ge………………………………….33
2-6.1: XEDS……………………………………………………......33
2-6.2: XRD………………………………………………………….34
2-6.3: Four Point Probe……………………………………………..36
Chapter 3: Direct Deposition of Cu onto Ru………………………………………..39
3-1: Nucleation Mechanism onto a Foreign Substrate……………………...40
3-2: Nucleation Density and Overpotential………………………………...44
3-3: Alkaline Copper-Complex Chemistry…………………………………46
3-4: Galvanostatic Pulse Waveform………………………………………...49
3-5: Results & Discussion………………………………………………….52
3-6: Conclusions……………………………...……………………………..55
Chapter 4: Additives’ Characterization in Alkaline Medium……………..………..58
4-1: Polarization Behavior of Polyether Additives in Alkaline
Complexed Medium…………………………………………………...62
4
4-2: Polarization Behavior of SPS in Alkaline Complexed Medium…….…..66
4-3: Identification and Characterization of an Anti-suppressor in the
Alkaline Medium……………………………………………………….76
4-4: Discussion………………………………………………………………81
4-5: Conclusions……………………………………………………………..83
Chapter 5: Bottom-up Metallization by Copper Plating from Alkaline Media in the
Presence of Additives…………..………………………………………...86
5-1: Background……………………………………………………………...86
5-2: Results…………………………………………………………………...90
5-3: Discussions………………………………………………………………95
Chapter 6: Pulse Plating of Copper Germanide………………………………………98
6-1: Challenges for Electroplating Cu3Ge…………………………..……….100
6-2: Cu3Ge Electrodeposition by Pulsing……………..……………………..103
6-3: Materials Characterization………………………..…………………….107
6-3.1: Cross-sectional SEM……………………..…………………...107
6-3.2: XPS……………………………………………………………109
6-3.3: XRD and Resistivity…………………………………………..110
6-4: Conclusions……………………………………..……………………...112
Chapter 7: Characterization of the Tartrate Plating Bath with Additives…………….115
7-1: Current Efficiency……………………………………………………….115
7-2: Bath Stability…………………………………………………………….123
Chapter 8: Mass Transfer and Kinetics Consideration in Pulse Plating of Cu3Ge…...124
8-1: Development of the Steady-State Galvanostatic Pulse Plating Model….129
8-2: Determination of Transport and Kinetics Parameters……..….…………134
8-2.1: Determination of the kinetic parameters……………………….134
5
8-2.2: Determination of the transport parameters……………………..137
8-3: Effects of mass-transfer and kinetics in Cu3Ge pulse plating…………...141
8-4: supplementary data.……………………………..……………………...146
Chapter 9: Conclusions and Future Work…………………………………………...150
Chapter 10: Appendix: Additional work-Electrodeposition of Cu-Mn ……..….…...154
10-1: Introduction……………………………………………………………154
10-2: Experimental…………….……………..…………….………..………156
10-3: Results and Discussions.…………………………..……………..……157
10-4: Conclusions:……………………………….………..…………………166
List of Symbols……….……….……..………………………………………………169
6
List of Tables
Table 4-1: Electrolyte composition, pH and overpotential (Δη) at the electrode for the
injection studies in Figure 4-6. …………………………………………………………..69
Table 4-2: A summary of the additives combination needed for bottom-up fill in the
acidic and alkaline complexed copper electrolytes……………………………………....83
Table 7-1: Fraction of the total current contributed by H2 evolution during copper plating
with additives at -1.25 V vs. SHE. The calibration curve was used to convert height
change to total H2 current……………………………………………………………....120
Table 8-1: Kinetic parameters for copper, germanium and hydrogen obtained via least
square fitting of the polarization curves in figure 8-1………………………………......135
7
List of Figures
Figure 1-1: Process steps for fabrication of dual damascene structures with copper
interconnects- (1) SiO2 or other low-k dielectric deposition (2) trench and via definition
by etching (3) Barrier and seed layer deposition (4) Feature fill by copper electroplating.
Figure adapted from Andricacos et al 2……………………………………………….…22
Figure 1-2: SEM view of the copper wires in an integrated circuit chip shown here after
dissolving the silicon. Photograph courtesy of IBM Corporation13 (1997)…………..…22
Figure 1-3: (a) Current process for metallizing interconnects with copper. Cu seed layer
forms overhang preventing bottom-up fill at smaller geometries (b) proposed scheme for
future generation interconnects. Electro-nucleation of a uniform Cu seed layer on Ru
eliminates the vapor deposited Cu seed layer…………………………………………....24
Figure 1-4: Molecular structures of typical additives used in bottom-up fill; polyethylene
glycol (PEG), bis-(3-sulfopropyl) disulfide (SPS) and polyethyleneimine (PEI)…….....25
Figure 2-1: Schematic of the injection study used to investigate the transient and steady
state additives effects………………………………………………………………….…31
Figure 2-2: Schematic of the pulse plating waveform used to deposit Cu-Ge in a 3:1
stoichiometric ratio………………………………………………………………………33
Figure 2-3: Schematic of the Bragg’s law and x-ray diffraction phenomenon…………36
Figure 2-4: Schematic of the four-point probe measurement5. Current is applied between
terminals 1 and 4 which induces a voltage drop between terminals 2 and 3…………….37
8
Figure 3-1: Schematic representation of the three different growth modes (a) ‘VolmerWeber’ 3-D island growth mode (b) ‘Frank-van der Merwe’ layer-by-layer growth (c)
‘Stranski-Krastanov’ 2-D layer growth followed by 3-D island growth……………...…42
Figure 3-2: A top and side view schematic of nucleation on a foreign substrate of
hemispherical nuclei for homogeneous nucleation and isotropic growth……………......43
Figure 3-3: Change in Gibbs free energy for formation of a nucleus with n atoms at
different overpotential values…………………………………………………………....45
Figure 3-4: A copper-tartrate molecule. Each copper molecule binds with two tartrate
molecules. Figure adapted from ref. 19…………………………………………... ……47
Figure 3-5: Polarization curves for copper-tartrate and acidic CuSO4 solutions.
Concentrations and pH of the electrolytes are described in section 2-1 and 2-2. High
deposition overpotential is observed in the Cu-tartrate electrolyte……………………...48
Figure 3-6: Pulse current densities and respective critical pulse widths, τ. CB,Cu=0.1 M
Cu and DCu=5  10-6 cm2/s. The shaded region represents the ranges of ipulse and ton that
can be used without completely depleting copper at the electrode…………………..…..51
Figure 3-7: Cu nucleation on Ru from (a) non-complexed acidic bath using DC current.
10 mA/cm2 was applied for 5s (b) alkaline copper-tartrate bath using DC current. 10
mA/cm2 was applied for 5s (c) acidic bath using galvanostatic pulse waveform. Each
pulse was 0.5 A/cm2 in amplitude and 1 ms in duration. Off time was 49 ms (d) alkaline
copper tartrate bath using galvanostatic pulse waveform. Each pulse was 2.5 A/cm2 in
amplitude and 0.2 ms in duration. Off time was 49.8 ms. Total charge applied for all
experiments were 50 mC/cm2. Average current density during each pulse period was 10
mA/cm2 ………………………………………………………………………………..…53
9
Figure 3-8: Copper nucleation on ruthenium from an alkaline Cu-EDTA electrolyte
using high current density pulsing. Galvanostatic pulse waveform was identical to the
one used in Figure 3-6d………………………………………………………..…………54
Figure 4-1: Transient and steady state polarization behavior of PEG and SPS in the
acidic copper plating bath containing 70 ppm chloride. Injection of 100 ppm PEG
strongly polarizes the electrode. Injection of 50 ppm SPS into the PEG, Cl- containing
bath depolarizes the electrode. Average current density was 10 mA/cm2 and a rotation
speed of 400 rpm was used. Figure adapted from ref [14]…………………………..…..60
Figure 4-2: PEG and PLE injection studies in the copper tartrate electrolyte containing
chloride ions. Significant passivation on the electrode is observed with PLE at 1100 ppm
concentration. Plating current density was 20 mA/cm2 and the rotation speed was 200
rpm…………………………………………………………………………………..…...63
Figure 4-3: PEG and PLE injection studies in the copper EDTA electrolyte containing
chloride ions.
Less than 50 mV of passivation is observed even at 1100 ppm
concentration of additives. Plating current density was 20 mA/cm2 and the rotation speed
was 200 rpm………………………………………………………………………...……64
Figure 4-4: SPS injection study in copper tartrate electrolytes having pH values of 10
and 12.5. A sharp transition is observed at the high pH value, after SPS injection.
Injection studies were performed at a current density of 20 mA/cm2 with a rotation speed
of 200 rpm…………………………………………………………………………...…...67
Figure 4-5: SPS saturation at the copper electrode in the copper tartrate electrolyte
(pH=12.5). SPS was injected in 1-2 ppm increments until a total concentration of 20 ppm
in solution is achieved. Passivation at the electrode reaches saturation at 6 ppm of SPS
10
concentration. Addition of more SPS had no further effect on the electrode polarization.
Current density and rotation speed were 20 mA/cm2 and 200 rpm, respectively………..68
Figure 4-6: SPS injection study in citrate, EDTA, DTPA and tartrate complexed
electrolytes. Molar ratios of copper to the complexing agents in solution are provided in
table 4-1.
SPS demonstrates significant inhibition only in the tartrate complexed
electrolyte. Plating conditions were identical to that used in Figure 4-5………………..70
Figure 4-7: SPS injection study in 0.2, 0.4 and 0.5 M tartrate solutions.
Copper
concentration was 0.1 M in all three solutions. 20 ppm SPS was injected. Overpotential
changes were approximately the same in all three solutions. Experimental conditions
were identical to that used in Figure 4-5…………………………………………………71
Figure 4-8: SPS injection studies in 0, 0.05, 0.11 and 0.17 M EDTA solutions. Copper
concentration was 0.1 M in all four solutions. 20 ppm SPS was injected at the 200 s
mark. Maximum passivation was observed when no EDTA was present in solution .
Experimental conditions were identical to that used in Figure 4-5…………………...…73
Figure 4-9: Dimensionless suppression for the electrolytes in Figure 4-8 after 20 ppm
SPS injection. max was taken as the suppression observed in the 0 M EDTA and 0.5 M
tartrate electrolyte after SPS injection. Fraction of total copper complexed to tartrate in
[ EDTA]
2
…………………………….....74
[Cu 2 ]
[Cu 2 ] 
solution (Cu-T) was calculated using
Figure 4-10: Injection of 15 ppm SPS into the alkaline copper tartrate electrolyte
followed by 40 ppm and 80 ppm PEI injection. PEI partially depolarizes the electrode,
neutralizing the suppressing effect of SPS. Current density and rotation speed were 20
mA/cm2 and 200 rpm, respectively……………………………………………………....78
11
Figure 4-11: Injection of 40 ppm PEI in the SPS free copper tartrate electrolyte after 200
s of plating. PEI does not exhibit any steady state polarization at the copper electrode.
Current density and rotation speed were 20 mA/cm2 and 200 rpm, respectively………..79
Figure 4-12: (a) simultaneous injection of a pre-mixed 15 ppm SPS and 40 ppm PEI
solution in the alkaline copper-tartrate electrolyte (b) simultaneous injection of pre-mixed
70 ppm Cl-, 150 ppm PEG and 50 ppm SPS into the acidic copper bath. (b) was adapted
from ref. [13]. A fast suppression, followed by slow depolarization is observed in both
cases…………………………………………………………………………………...…80
Figure 4-13: Polarization curves for copper deposition from an alkaline copper-tartrate
electrolyte containing 50 ppm SPS, 100 ppm PEI and no additives. Significant steady
state polarization is observed in the SPS saturated electrolyte. PEI saturated electrolyte
does not exhibit any polarization relative to the fresh electrolyte. Rotation speed of 200
rpm was used for all experiments……………………………………………………......82
Figure 5-1: Illustration of the ‘CEAC’ mechanism. Concentration of the accelerator at
the bottom increases as surface area contracts due to plating. After growth to a certain
distance, accelerator coverage reaches 100%....................................................................89
Figure 5-2: Bottom-up fill from the additive containing alkaline bath under galvanostatic
deposition (25 mA/cm2) for 45s.
Note the accelerated growth at the bottom corners
giving rise to a triangular shape. Aspect ratio of the feature is 1:1 and trench width is
approximately 1 µm. Rotation speed was 200 rpm…………………………………..…91
Figure 5-3: Bottom-up fill from the additive containing alkaline bath under identical
conditions as figure 5-2.
Note the suppressed growth on the sidewalls and bottom-up
growth. Aspect ratio of the feature is 2.5:1 and trench width is approximately 500 nm..92
12
Figure 5-4: Partial-fill profiles for Cu electrodeposition from an alkaline, tartratecomplexed Cu electrolyte containing 15 ppm SPS and 40 ppm PEI. The plating current
density (on the flat region of the wafer segment) is 25 mA/cm2 and the plating times are
20 s, 30 s, 45 s and 75 s. Note the accelerated growth at the trench bottom corners at
short times (20 s), progression of flat bottom-up fill at 30-45 s, and complete fill after 75
s of plating. Plating conditions are the same as in Figure 2……………………….…….93
Figure 5-5: Partial-fill profiles after Cu electrodeposition within 500 nm wide trenches
from an alkaline, tartrate-complexed Cu electrolyte at 25 mA/cm2 for 45 s.
The
electrolyte contained (a) 15 ppm SPS; (b) 40 ppm PEI; and (c) 15 ppm SPS and 40 ppm
PEI. Conformal fill is observed in electrolytes containing a single additive, leading to
void formation. Bottom-up fill is observed in electrolyte containing both SPS and PEI,
indicating the role of antagonistic interactions between these additives.
Plating
conditions: i=25 mA/cm2, pH=12.5, [Cu]=0.1 M, [Tartrate]=0.5 M, rotation speed=200
rpm……………………………………………………………………………………….94
Figure 6-1: Polarization behavior (blue, left ordinate) and deposit composition (red, right
ordinate) in Cu and Ge co-deposition from an alkaline tartrate complexed electrolyte
under DC. Electrolyte contains 25 mM Cu, 96 mM Ge, 300 mM tartrate and 325 mM
NaOH. Films obtained using DC plating do not meet the target Cu:Ge composition of
3:1 required for Cu3Ge formation…………………………………………………...….102
Figure 6-2: Surface roughness profiles of electroplated Cu using DC (black dashed line)
and pulse (blue solid line) plating. The average current density was 15 mA/cm2 for both
cases. The rms roughness of the DC plated film was 334nm while that of the pulse plated
13
film was 73 nm.
Both experiments corresponded to a total charge of 2.85
coulombs………………………………………………………………………………..104
Figure 6-3: Ge content in electrodeposited Cu-Ge films as a function of the instantaneous
current density (ion). Target Ge content of 25 at. % is achieved when ion exceeds 200
mA/cm2. The insert shows a schematic of the square pulse waveform applied during
pulse plating. ton depends on ion according to Eq. 3-8……………………………….....105
Figure 6-4: EDS spectra of Cu-Ge plated at a pulse current density of 205 mA/cm2 with
ton=2.2 ms. EDS indicates near stoichiometric ratio of 3.3:1 of Cu:Ge……………..…106
Figure 6-5: (a) Cross-section of the electroplated Cu-Ge film on a Ru substrate. Pulse
plating conditions: ion= 205 mA/cm2, ton=2.2 ms and toff=28 ms. Total plating time is 7
minutes. (b) Surface morphology of the plated film…………………………………....108
Figure 6-6: Composition depth profile of a 150 nm Cu3Ge film deposited using the same
waveform. A uniform Cu and Ge composition is observed…………………………....109
Figure 6-7: XRD spectra of as-deposited (a) and annealed (b) films showing ε-Cu3Ge
formation. All the diffraction peaks associated with the monoclinic ε-Cu3Ge16 are shown
in (c). Pulse plating conditions: ion=205 mA/cm2, ton=2.2 ms and toff=28 ms. Total
plating time =14 minutes. Calculated film thickness was ~0.9 µm……………………111
Figure 7-1: Faradaic efficiency for copper deposition from the alkaline tartrate
electrolyte, alkaline tartrate electrolyte containing 15 ppm SPS and 15 ppm SPS+40 ppm
PEI. Copper was plated under galvanostatic condition (20 mA/cm2 for 30 mins, ω=200
rpm). Additives lower the current efficiency by 40-50 %..............................................116
14
Figure 7-2: A schematic of the experimental setup used to quantitatively determine H2
evolution………………………………………………………………………………..118
Figure 7-3: Calibration curve for H2 evolution using the setup shown in figure 7-2. H2
was evolved at a constant potential of -1.25 V vs. SHE from an aqueous electrolyte with
pH=12.5. Slope of the linear plot was 3728.5 mC/cm, assuming 100 % efficiency for the
H2 evolution reaction…………………………………………………………………...119
Figure 7-4: Current efficiency for copper plating from the tartrate solution with different
concentrations of SPS in solution. Efficiency decreased as SPS concentration in solution
increased. Experimental conditions were identical to that used in figure 7-1………....121
Figure 8-1: Governing equation and boundary conditions during the pulse ‘on’ and ‘off’
cycle. The electrode is located at x=0……………………………………………….…126
Figure 8-2: Schematic of the concentration profile at the electrode during galvanostatic
pulsing when toff is comparable to ton. A build-up of concentration gradient occurs within
the diffusion layer.
The duration of this unsteady state depends on the prevailing
hydrodynamic conditions of the system. For typical values, the duration of this unsteady
state is calculated to be about 10 s…………………………………………………..….131
Figure 8-3: A schematic of the duplex boundary layer model.
The concentration
fluctuation occurs within δp while gradient in the intermediate region (δ-δp) remains
stationary. The gradient within the stationary region depends on the hydrodynamics of
the system………………………………………………………………………..……...132
Figure 8-4: Polarization curves of (a) hydrogen (b) germanium and (c) copper on a
copper rotating disk electrode from the copper and germanium containing electrolyte at
15
1000 rpm. Least squares fitted curves are shown as well as the best fit values for the
kinetic parameters…………………………………………………………………...….136
Figure 8-5: Cyclic voltammograms on a steel rotating disk electrode from the copperfree alkaline tartrate electrolyte. Potential was scanned from -0.42 V (vs. SCE) to -2.4 V
(vs. SCE) in the cathodic direction and to +0.1 V (vs. SCE) in the anodic direction. The
figure above has been magnified to better illustrate the peak currents………………....138
Figure 8-6: A plot of peak current densities vs. scan rates, obtained from Figure 8-4.
Diffusion coefficient can be calculated from slope of the linear plot. Slope was obtained
by fitting a straight line using least squares method in origin……………………….....139
Figure 8-7: Atomic concentration of germanium in the alloy at different pulse current
densities. A pulse ‘on’ time (ton) = 2.2 ms and duty cycle = 0.02 were used in the steadystate model, same as in the experiments.
Diffusion boundary layer of copper and
germanium were assumed to be 20 µm and 4 µm, respectively…………………...…...142
 C 
Figure 8-8: Dimensionless concentration of copper and germanium  s  at different
 Cbulk 
pulse current densities. Simulation and experimental conditions were same as in Figure
8-6……………………………………………………………………………...……….143
Figure 8-9: Faradaic efficiency of Cu3Ge pulse plating at different applied current
densities calculated using the steady-state model.
Identical parameters as Figure 8-6
were used………………………………………………………………...……………..144
Figure 8-10: Faradaic efficiency of Cu3Ge pulse plating at various total plating times.
Experimental condition: i=205 mA/cm2, ton=2.2 ms, toff=100 ms…………………..….145
16
Figure 10-1: Linear sweep voltammetry on a Cu electrode from 1 M (NH4)2SO4 + 0.59
M MnSO4.H2O electrolyte and the complete electrolyte (0.05 M CuSO4.5H2O + 0.052 M
EDTA + 0.59 M MnSO4.H2O + 1 M (NH4)2SO4). Scan rate and rotation speed were 50
mV/s and 300 rpm, respectively. A magnified image of the area of interest is shown in
the inset in bottom right corner……………………………………………...………….159
Figure 10-2: Co-deposition of copper and manganese on a copper RDE using DC
waveform. Potential transients, EDS spectra and morphologies of the films deposited at
20 and 200 mA/cm2 are shown. Rotation speed was 300 rpm in all cases………….....160
Figure 10-3: Schematic of the pulse plating waveform used to deposit Cu-Mn in a 49:1
stoichiometric ratio. A 300 mA/cm2 pulse with a 0.167 s ‘on’ time was followed by a 5
mA/cm2 pulse with a 5 s ‘on’ time. A pulse train consisting of 109 pulses was applied for
a total charge density of 8.2 C/cm2………………………………………………...…...162
Figure 10-4: (a) Cross-section of the Cu-Mn film electrodeposited on a Cu seeded wafer
using the pulse waveform shown in Figure 1.
A protective layer of platinum was
sputtered on top of the deposit prior to milling (b) a top-down micrograph of the plated
Cu-Mn film without the Pt cap (b) EDS spectra of the Cu-Mn film indicating codeposition of Cu and Mn in a 49:1 ratio…………………………………………...…...163
Figure 10-5: Surface roughness profiles of electroplated Cu-Mn using DC (black dashed
line) and pulse (red solid line) plating. Both experiments corresponded to a total charge
density of 8.2 Cb/cm2. The rms roughness of the DC plated film was 850 nm, while that
of the pulse plated film was 372 nm……………………………………………………164
17
Figure 10-6: (a) XPS spectra of Mn 3p at the surface of the electroplated Cu-Mn film
before and after annealing (b) a depth profile of the manganese content in the alloy before
and after annealing that indicates segregation of Mn at the interface……………….….166
18
Novel Alkaline Copper Electroplating Processes for Applications in Interconnect
Metallization
Abstract
by
ANIRUDDHA AHMED JOI
Electrical interconnects in microchips, present in just about every electronic
device, are currently formed by bottom-up electroplating of copper within etched nanoscale features. The copper is plated from an acidified copper plating bath in the presence
of special additives that provide the current distribution essential bottom-up fill.
However, as the copper interconnects will become smaller in future generation
microprocessors, a drastic redesign of the current process steps is required. Achieving a
uniform copper seed layer by vapor deposition techniques becomes increasingly difficult
at geometries that are less than about 20 nanometers wide. A proposed alternative is the
bottom-up fill of features by directly plating copper onto a vapor deposited ruthenium
barrier layer that can be deposited uniformly and provides the requisite conductivity.
However, the conventional acidic copper plating solution cannot provide sufficient
adhesion and uniformity on top of ruthenium. A chemistry that provides high nucleation
density of Cu on Ru needs to be designed. Furthermore, bottom-up fill capability from
this chemistry is also desired.
We describe here a novel alkaline, tartrate-complexed copper electrolyte that
provides high nucleation density on ruthenium and in the presence of special additives,
also provides bottom-up fill. High nucleation density (> 1011 nuclei/cm2) on ruthenium is
obtained by current pulsing in the alkaline tartrate-complexed electrolyte. Bottom-up fill
19
is achieved using a mixture of two additives: bis-(3-sulfopropyl) disulfide (SPS) and
polyethyleneimine (PEI). Chronopotentiometric studies indicate that, unlike in
conventional acidic electrolytes, SPS acts as a ‘suppressor’ and PEI acts as an ‘antisuppressor’ in the alkaline media. Partial-fill experiments on patterned structures confirm
the required SPS-PEI interaction leading to bottom-up fill from the tartrate-complexed
copper electrolyte.
Furthermore, we extend the versatility of these alkaline complexed electrolytes by
reporting electrodeposition of copper germanide (Cu3Ge) thin films from an alkaline
tartrate-complexed electrolyte. Using current pulsing, co-deposition of copper and
germanium in the stoichiometric ratio Cu:Ge=3:1 is achieved while maintaining a smooth
and compact electrodeposit. X-ray diffraction confirms the presence of ε-Cu3Ge phase
with a monoclinic crystal structure in the as-deposited films. Effects of mass-transfer and
kinetics in Cu3Ge pulse plating are modeled providing good agreement with the
experimental data.
20
Chapter 1
Introduction
Miniaturization of interconnect dimensions for extension of Moore’s law into sub20 nm features require modification of current metallization schemes. Presently, copper
metallization in electrical interconnect fabrication is accomplished using a process called
damascene or dual damascene process1 (Figure 1-1). In this technology, features (trench
lines or via-holes) are first etched in SiO2 or other low-k dielectric materials. This is
followed by deposition of a barrier layer such as TaN or TiN onto the patterned features
by sputtering (PVD) or chemical vapor deposition (CVD). This barrier layer prevents
copper from diffusing into silicon and causing device failure. A uniform copper seed
layer is then deposited onto the barrier layer by physical vapor deposition. Patterned
features with the barrier and seed layers are subsequently filled with copper by
electrodeposition.
21
SiO2
Trench
1
2
4
3
Plated Cu
Figure 1-1: Process steps for fabrication of dual damascene structures with copper
interconnects- (1) SiO2 or other low-k dielectric deposition (2) trench and via definition
by etching (3) Barrier and seed layer deposition (4) Feature fill by copper electroplating.
Figure adapted from Andricacos et al 2.
Figure 1-2: SEM view of the copper wires in an integrated circuit chip shown here after
dissolving the silicon. Photograph courtesy of IBM Corporation13 (1997).
22
However, as feature sizes become smaller, obtaining a uniform and defect free Cu
seed layer by vapor deposition techniques has become increasingly challenging due to
overhang formation at the via rims brought about by the preferential accessibility of this
region3,4 (Figure 1-3a). Moreover, the seed layer occupies a significant fraction of the
feature cavity at these narrow geometries, making scaling difficult. A suggested
alternative has been to replace the two underlayers (of Ti and Cu) by a single layer that
can provide the dual functionality of barrier to diffusion and a substrate which can be
directly plated by Cu electrodeposition (Figure 1-3b). Ru has been identified as a
candidate material offering the requisite thermal and electrical properties5 and
impediment to Cu diffusivity6. However, when plating copper onto Ru, the coalescence
thickness of the deposited film at which the reduction in substrate resistance is realized
depends upon the prevailing nucleation mechanism7. Conventional acid copper
chemistry with organic plating additives results in poor nucleation and uniformity on the
barrier layer. Therefore, a novel electrodeposition chemistry is required that enables
adequate Cu nucleation and uniformity on Ru to eliminate the limitations of vapor
deposited seed layers for future generation interconnects.
23
Electro-nucleated
Cu seed
Ta/TaN/Ru
SiO2
(a)
(b)
Figure 1-3: (a) Current process for metallizing interconnects with copper. Cu seed layer
forms overhang preventing bottom-up fill at smaller geometries (b) proposed scheme for
future generation interconnects. Electro-nucleation of a uniform Cu seed layer on Ru
eliminates the vapor deposited Cu seed layer.
There is, however, one limitation to the above stated scheme of directly plating
Cu on to Ru. Two different process steps and electrolytes are required: a new chemistry
and process for copper seed deposition on Ru, followed by bottom-up fill from additivescontaining conventional acid copper bath.
Additive assisted bottom-up electroplating was first reported by IBM researchers
in the early 90’s2. Without special additives, plating in high aspect ratio features results
in voids within the plated features. Properly selected additives mixtures can be used to
modify the deposition kinetics of copper within the features enabling preferential plating
at the bottom corners of the features while inhibiting deposition on the flat top and the
sidewalls 2,8-10. There are three families of additives that are typically used in feature fill:
suppressors, anti-suppressors or accelerators and levelers. Suppressors are usually a
24
polyether or polyether derivative, e.g. polyethylene glycol (PEG, Figure 1-4). These
molecules are diffusion-limited and adsorbs preferentially on the sidewalls and rims,
inhibiting the copper deposition. The anti-suppressor is typically a sulfonate derivative
such as bis-(3-sulfopropyl) disulfide (SPS, Figure 1-4). These molecules are fast
diffusing and adsorb preferentially at the feature bottom preventing the inhibitor from
adsorbing at those sites, and thus maintaining non-inhibited copper deposition at the
bottom of features. PEG requires the presence of chloride ions to be functional. It has
been shown that it is the differential transport-adsorption of these additives8 and the
acceleration of the growing bottom corners due to local SPS concentration9,10 that makes
bottom-up fill possible. A third type of additive called ‘leveler’ is often used to eliminate
the bump formation over completely filled features. Levelers are typically nitrogen
containing compounds such as polyethyleneimine (PEI, Figure 1-4) or dies e.g., Janus
Green.
PEG
SPS
PEI
Figure 1-4: Molecular structures of typical additives used in bottom-up fill; polyethylene
glycol (PEG), bis-(3-sulfopropyl) disulfide (SPS) and polyethyleneimine (PEI).
25
Another roadblock to interconnect scaling is that pure Cu exhibits diminished
material properties in extremely small features, e.g., as dimensions approach the electron
mean free path in Cu (~39 nm), the rise in Cu resistivity introduces a time delay that
deteriorates device performance. Furthermore, the reduced electromigration (EM)
lifetime of Cu in smaller geometries, due to increased grain boundary density and stressinduced voiding, contributes to poor reliability11. A possible means to circumvent these
adverse effects is to utilize Cu-alloys with improved material properties. Cu3Ge had
recently been suggested for damascene integration due to its low room temperature
resistivity and high oxidation resistance12. EM reliability of Cu can be significantly
improved by ‘doping’ it with impurities such as Mn11. Traditionally, these Cu-alloys are
fabricated using PVD or CVD techniques which are not scalable. Hence, electrochemical
route for depositing these alloys is highly desirable.
For ease of process integration it is of practical interest to have high nucleation,
bottom-up metallization and Cu-alloy plating capability from a single electroplating bath
since this would eliminate the need for multiple process steps. Therefore, the objectives
of this work are to develop and characterize a versatile copper plating chemistry and
additives having the following multiple functionalities:
(i) Direct plating of Cu onto Ru with high nucleation density
(ii) Bottom-up metallization of interconnects
(iii) Electrodeposition of Cu-alloys, e.g., Cu3Ge, Cu-Mn, etc.
26
The theses structure is as follows:
 In Chapter 2, Experimental methodologies employed in this work are outlined.
 In chapter 3, a novel process for directly plating copper onto ruthenium is
described. A brief theoretical overview of the nucleation phenomena is also
provided emphasizing the relevant parameters.
 In chapter 4, behaviors of conventional additives in the novel copper
chemistry are characterized and a unique suppressor-accelerator combination
is identified.
 In chapter 5, bottom-up fill is demonstrated in the non-conventional chemistry
with additives identified in chapter 4. A qualitative comparison is made to
bottom-up fill mechanism in conventional acidic media with PEG and SPS.
 In chapter 6, Electrodeposition of Cu3Ge using a pulse waveform is described.
A similar plating bath used in chapters 3-5 is also used for electroplating
Cu3Ge.
 In chapter 8, Theoretical aspects of pulse plating is described emphasizing the
mass transfer phenomena during pulsing. A steady-state model is developed
and compared to data obtained from pulse plating of Cu3Ge.
27
References:
1. Y. Sacham-Diamand, T. Osaka, M. Datta, T. Ohba, Advanced Nanoscale ULSI
Interconnects: Fundamentals and Application, Springer, New York (2009).
2. P.C. Andricacos, C. Uzoh, J.O. Dukovic, J. Horkans and H. Deligianni, IBM J. of
Res. and Dev., 42, 567 (1998)
3. A. Radisic, Y.Cao, P. Taephaisitphongse, A. C. West and P. C. Seasrson, J.
Electrochem. Soc., 150, C362 (2003).
4. S. Armini, Z. tokei, H. Volders, Z. El-Mekki, A. Radisic, G. Beyer, W.
Ruythooren, P. M. Vereecken, Microelectron. Eng., 88, 754 (2011).
5. M. W. Lane, C.E.Murray, F. R. McFeely, P. M. Vereecken, and R. Rosenberg,
Appl. Phys. Lett., 83, 2330 (2003).
6. R. Chan, T. N. Arunagiri, Y. Zhang, O. Chyan, R. M. Wallace, M. J. Kim and T.
Q. Hurd, J. Electrochem. Soc.,7, G154 (2004).
7. M. J. Willey and A. C. West, Electrochim. Acta, 52, 6484 (2007).
8. R. Akolkar and U. Landau, J. Electrochem. Soc., 151, C702 (2004).
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. A.C. West, S. Mayer and J. Reid, Electrochem. Solid-State Lett., 4, C50 (2001).
11. L. Krusin-Elbaum and M. O. Aboelfotoh, Appl. Phys. Lett. 58, 1341 (1991)
28
12. J. P. Gambino, “Improved Reliability of Copper Interconnects Using Alloying”,
Proc. 17th IEEE IPFA, pp. 1-7, 2010.
13. IBM 100: Icons of Progress, “Copper Interconnects: The Evolution of
Microprocessors” (http://www-03.ibm.com/ibm/history/ibm100/us/en/).
29
Chapter 2
Experimental Methodology
2-1. General Electrochemical Setup
A three electrode setup was utilized for all electrochemical experiments. The
reference electrode was a Cu/Cu2+ redox couple (+0.34 V vs. SHE), placed 2 cm away
from the cathode so that the ohmic drop was constant. In all cases, potentials are,
however, reported versus a standard hydrogen electrode (SHE). A platinized titanium
mesh disk with a diameter of 5.5 cm placed at the bottom of the beaker served as the
anode. All electrochemical experiments were performed with a potentiostat/galvanostat
(VSP, Princeton Applied Research) and data were collected using the EC-Lab software.
Electrochemical polarization measurements were conducted on a Cu disk electrode with
an active area of 0.325 cm2 unless otherwise specified. The conventional acidic Cu
sulfate solution used for comparison purposes consisted of 0.1 M CuSO4.5H2O at pH = 2.
The pH was adjusted by the addition of concentrated H2SO4 (Fisher, Certified ACS).
2-2. Direct Deposition of Cu onto Ru
The electrolyte used for directly plating Cu onto Ru consisted of 0.1 M
CuSO4.5H2O (Fisher, Certified ACS) and 0.5 M C4H4O6KNa·4H2O (Tartrate, Acros
organics) with pH adjusted to 12.5 by NaOH (Fisher, Certified ACS) addition. Cu was
nucleated on silicon wafer segments coated by a PVD ruthenium layer (100 Å Ru/800 Å
Ta/1 k Å SiO2). The wafer segments consisted of 0.1-0.5 cm2 pieces which were
pretreated by applying cathodic current at a constant potential of -0.493V vs. Cu/ Cu2+
reference electrode in an acid electrolyte for 1 minute. The acid electrolyte comprised of
30
1.8 M H2SO4 and 1 mM NaCl. This potential is close to the H2 evolution potential and
reduces the spontaneously forming oxide layer that is known to be detrimental toward Cu
nucleation on Ru1. Electronucleation of Cu on Ru was performed in a 100 ml beaker
containing 50 ml electrolytic solution with no stirring.
A square wave pulse train consisting of an extremely short (1 ms) very high
current density (2.5 A/cm2) square cathodic current pulse, followed by an off period of 49
ms was applied. The average current density and total charge were kept at 10 mA/cm2
and 50 mC/cm2, respectively. Schematic of the similar pulse train is illustrated in Figure
2-2. Cu nucleation on Ru was observed ex situ in a Hitachi S4500 scanning electron
microscope (SEM) and nuclei were counted manually to obtain nucleation density.
2-3. Additives Characterization in Alkaline Media
A Cu rotating disk electrode (RDE) with an active area of 0.32 cm2 and rotating at
200 rpm was used for additives adsorption studies. The electrolyte consisted of 0.1 M
CuSO4 + 0.5 M tartrate at pH =12.5, as described in the previous section. Additives that
were studied include: PEG (M. W. =104, Sigma Aldrich), SPS (Raschig) and PEI (M. W.
=600, Alfa Aesar). Highly concentrated stock solutions (10,000 ppm) of the additives
were prepared and aliquots of the stock solution were injected into 50 ml of the
electrolyte. Volume of the aliquot to be injected was calculated based on the desired final
concentration of the additives in the electrolyte. Transient additives effects were
examined by injecting the additives into the electrolyte under galvanostatic conditions
(i=20 mA/cm2) and recording the chronopotentiometric response at the electrode, as
31
described by Akolkar and Landau2. The schematic shown in Figure 2-1 illustrates the
injection study setup used for additives’ charaterization.
Figure 2-1: Schematic of the injection study used to investigate the transient and steady
state additives effects.
2-4. Bottom-up Fill in Alkaline Media in the presence of Additives
Partial via-fill and full-fill studies were performed on patterned test coupons
mounted on the rotating disc electrode. These test coupons had trench structures ranging
from 500 nm to 1 µm in width. All test coupons were pre-coated with a thin PVD barrier
and a Cu-seed layer. Feature fill studies were performed in the Cu-tartrate electrolyte
that was pre-mixed with 15 ppm SPS and 40 ppm PEI. An average current density of 25
mA/cm2 (calculated based on the projected wafer area) was used for feature fill. After
plating, wafers were cleaved perpendicular to the plated features and bottom-up fill was
32
determined by examining the SEM images of cross-sections of the plated samples in a
Hitachi S4500 scanning electron microscope.
2-5. Galvanostatic Pulse Plating of Copper Germanide
A similar three electrode setup to that described earlier was used. However,
instead of a Cu/Cu2+ redox couple, a saturated calomel electrode (SCE, +0.244 vs. SHE)
served as the reference electrode. Electrochemical polarization measurements were
conducted on a Cu disk electrode with an active area of 0.32 cm2 rotating at 1000 rpm.
The electrolyte contained 25 mM CuSO4.5H2O (Fisher, Certified ACS), 95 mM GeO2
(Alfa Aesar), 300 mM sodium potassium tartrate (Acros organics) and 325 mM NaOH
(Fisher, Certified ACS). Addition of the prescribed amount of NaOH resulted in a pH 13
solution.
A pulse train consisting of an instantaneous current density of 205 mA/cm2
applied for 2.2 ms followed by a rest period of 28 ms was used for depositing Cu3Ge. A
schematic of the pulse train is shown in Figure 2-2. Average current density for one
pulse period (ton+toff) was 15 mA/cm2.
33
Pulse Current Density (mA/cm2)
ion=205 mA/cm2
ton=2.2 ms
200
toff=28 ms
150
100
50
iavg=15 mA/cm2
0
0
10
20
30
40
50
60
Time (ms)
Figure 2-2: Schematic of the pulse plating waveform used to deposit Cu-Ge in a 3:1
stoichiometric ratio.
2-6. Materials Characterization of Cu3Ge
For thin-film characterization, Cu3Ge films were electrodeposited onto silicon
wafer coupons pre-coated with a 4 nm Ru layer. This was followed by annealing of the
electrodeposited Cu3Ge films in a nitrogen environment (< 2 ppm O2) of a PureLab HE
4GB 2500 glove box at ~300 oC for 30-60 minutes. Prepared films were then subjected
to various materials characterization.
XEDS (X-ray Energy Dispersive Spectroscopy) and FIB (Focused Ion Beam)
Elemental composition of the plated film was studied using XEDS. In this
technique, the surface of the film is bombarded with a high energy electron beam. This
high energy beam may excite an electron in the inner shell of the atom, eventually
34
ejecting it from its shell creating an electron hole. Electron from an outer, higher energy
shell then fills the hole releasing energy in the form of x-ray radiation. A spectrometer
records the number and energy of the x-rays that are emitted. Released x-rays are
characteristic of the difference in energy levels and the atomic structure of the element
from which it originates, providing measurement of elemental composition of the film.
XEDS is instrumental in measuring composition of bulk films (0.5-1 µm). Analysis of
the plated Cu3Ge film was performed in a Hitachi S4500 SEM equipped with a Noran
Energy Dispersive Spectrometer (EDS). A 20 kV electron beam was used.
Elemental XEDS mapping was also utilized to measure the film thickness in
cross-sections prepared by FIB milling. Unlike SEM, which utilizes electron beam, FIB
uses a focused ion beam that can be used to mill the film. A thin platinum layer was
sputtered onto Cu3Ge prior to milling to protect the underlying plated sample. FEI Nova
Nanolab 200 SEM equipped with a gallium ion source was used to etch the surface and
XEDS elemental mapping was performed in-situ.
XRD (X-Ray Diffractometry)
X-ray diffractometry was used to characterize the crystal structure and particle
size in the plated film. X-ray diffraction analysis utilizes the Bragg’s law:
n  2d sin 
(2-1)
Where n is any integer (1, 2, 3,…), λ is the x-ray wavelength, d is the spacing between
parallel crystal planes and θ is the Bragg angle. When this relationship is satisfied, i.e.
35
the incident angle is the Bragg angle; diffracted waves are in phase and are captured in
the radiation detector. This is shown by the schematics in figure 4-3. These in phase
diffractions at the Bragg angle show up on the diffraction spectra as high intensity peaks
for a specific crystal plane. The diffraction spectra of the sample can then be compared
to the spectra of a known standard from the database to determine its structure and crystal
planes present.
XRD can also be used to measure the particle size in the film. When the particle
size is less than 100 nm, broadening of the diffraction line occurs and the extent of this
line broadening is given by the Scherrer equation3:
t
0.9
 cos 
(2-2)
Where t is the diameter of the crystal particle, β is the broadening of the diffraction line
measured at half of its maximum intensity, θ is the Bragg angle and λ is incoming x-ray
wavelength. The XRD studies were performed in a Bruker Discover D8 X-ray
diffractometer using a cobalt Kα X-ray source (λ=1.79Å).
36
Figure 2-3: Schematic of the Bragg’s law and x-ray diffraction phenomenon.
Four Point Probe Measurement
Electrical resistivity of the Cu3Ge thin films were measured using four-point
probe technique. The four-point probe is an instrument that calculates the resistance
across a film by measuring the voltage drop for an applied electrical current. This is
illustrated by the schematics in figure 4-4. Potential difference (V) between wires 2 and 3
is measured as an electrical current (I) is applied between wires 1 and 4. Since only
negligible current flows in wires 2 and 3, the contact resistance of these wires is very low.
For s>>t, where s is the distance between wires/tips (Fig. 4-4) and t is the film thickness,
the sheet resistance (Rs) is given by:
Rs  k 
V
I
(2-3)
Where k is a geometric factor. For a semi-infinite thin sheet, value of k is 4.53. The
resistivity is then calculated by multiplying the sheet resistance by the film thickness.
The film thickness was calculated based on weight measurements using Faradays law.
37
Electrical resistivity measurements were performed using a Lucas Labs 320 four-point
resistivity probe equipped with a Keithly 2400 source meter.
1
3
2
4
S
t
Figure 2-4: Schematic of the four-point probe measurement5. Current is applied between
terminals 1 and 4 which induces a voltage drop between terminals 2 and 3.
In addition to the techniques described above, several other instruments were
utilized for materials characterization. They were used sparingly so as not to warrant
detailed discussions. A KLA-Tencor P-6 Stylus Profilometer was used to characterize
the roughness of plated films. A PHI VersaProbe XPS Microprobe was used to measure
the depth profile of electroplated Cu3Ge.
38
References:
1. L. D. Burke, N. S. Naser and R. Sharna, J. Appl. Electrochem., 38, 377 (2008).
2. R. Akolkar and U. Landau, J. Electrochem. Soc., 151, C702 (2004).
3. B. D. Cullity, Elements of X-ray Diffraction, Addison-Wesley Publishing
Company, Massachusetts (1967), p. 262.
4. W. D. Callister and D. G. Rethwisch, Materials Science and Engineering: An
Introduction, 8th Ed., Wiley, New York (2010), Figure 3.37.
5.
http://www-inst.eecs.berkeley.edu/~ee143/fa10/lab/four_point_probe.pdf
39
Chapter 3
Direct Deposition of Copper onto Ruthenium
Nucleation of copper on barrier layer materials, e.g. TaN1, TiN2 and Ru3-5 has
been studied and nucleation densities have been reported. Reported nucleation densities
of Cu on Ru from conventional acidic plating bath are in the range of 108-109 nuclei/cm2.
The reported nucleation densities correspond to a coalescence thickness of 160 nm
(Figure 3-2 and Equation 3-1); too large to be of any practical use in future generation
interconnects. This low nucleation density is attributed to the low deposition
overpotential of traditional acidic copper plating bath, typically in the range of 80-100
mV at a current density of 10 mA/cm2. Effect of additives, e.g. PEG, SPS and Cl- on
nucleation of Cu has also been investigated5-7. Incorporation of additives (specifically
PEG) in the plating bath increased the nucleation density only by a small factor (between
4 to 8). West et al.8 demonstrated the application of a pulse waveform to significantly
enhance Cu coverage on Ru from a PEG containing plating bath. While nucleation
density data were not reported, pulse plating was shown to have a significant impact on
the nucleation density. Alkaline complexed-Cu electrolytes have also been developed to
achieve high Cu nucleation density and uniform plating 9-12. In addition to their
application in direct Cu plating, these electrolytes have been shown to be advantageous
even in conventional interconnect metallization due to the lower susceptibility for seed
layer dissolution in an alkaline medium13. In the following section, we demonstrate
enhanced nucleation of Cu on Ru using:
(i)
An alkaline copper-complex chemistry and
(ii)
High current density pulse waveform.
40
3-1. Electrochemical Nucleation Mechanism onto a Foreign Substrate
Electrochemical nucleation and subsequent crystal growth of metals (Me) occurs
at the substrate (S) and conducting electrolyte interface, generally in three stages14: (i)
formation of adatoms on a substrate (ii) two-dimensional (2-D) or three-dimensional (3D) phase formation via nucleation and cluster growth and (iii) growth of the threedimensional bulk phase. The mechanism of nucleation and growth in
electrocrystallization and whether a 2-D or 3-D phase will form depend mainly on two
parameters; the binding energy of the metal atoms on a foreign substrate or the adhesive
force (  Me ,ads  S ) and the binding energy of the metal atoms on a native substrate or the
cohesive force (  Me ,ads  Me ). Neglecting kinetic influences and any Me-S alloy formation,
three different types of nucleation modes, based on the aforementioned parameters, can
be distinguished14:
(a) ‘Volmer-Weber’ 3-D Island Growth: If  Me, ads  s   Me ,ads  Me , namely the cohesive
force is much stronger than the adhesive force, then phase formation will occur via 3D island growth mechanism. In this case, metal deposition will preferentially occur
on already deposited Meads resulting in island growth (Figure 3-1a).
(b) ‘Frank- van der Merwe’ Layer-by-layer Growth: If  Me, ads  s   Me ,ads  Me , namely
the adhesive force is much stronger than the cohesive force, then phase formation will
occur via 2-D layer-by-layer growth mechanism (Figure 3-1b). This is the case of
metal underpotential deposition.
41
(c) ‘Stranski-Krastanov’ Growth Mode: This is an intermediary process that includes
both 2-D layer and 3-D island growth mechanisms. The growth mode transitions
from a 2-D layer-by-layer growth to a 3-D island growth at a critical thickness that
depends on the surface energies and lattice strains (Figure 3-1c).
From the discussion above, it is apparent that a 2-D layer-by-layer growth mechanism
would be ideally suited for direct deposition of Cu onto Ru. Uniform, monolayers thick
Cu film could be deposited that might serve as conductive seed layers for subsequent
bottom-up electroplating. However, it has been shown that Cu nucleation on Ru follows
the Volmer-Weber 3-D island growth mode5,15. In island growth mechanism, nucleation
density plays a key role because it determines the minimum island coalescence thickness,
i.e. thickness at which the film can be considered uniform and substrate induced
resistivity effects are drastically reduced. For hemispherical shaped islands, the
minimum film thickness, dcrit, can be calculated if one assumes homogeneous nucleation
and isotropic growth of the nuclei (Figure 3-2) and Equation 3-1 can be derived16.
d crit 
1
2 N
(3-1)
42
Figure 3-1: Schematic representation of the three different growth modes (a) ‘VolmerWeber’ 3-D island growth mode (b) ‘Frank-van der Merwe’ layer-by-layer growth (c)
‘Stranski-Krastanov’ 2-D layer growth followed by 3-D island growth.
43
Figure 3-2: A top and side view schematic of nucleation on a foreign substrate of
hemispherical nuclei for homogeneous nucleation and isotropic growth.
It follows from Equation 3-1 that in order to deposit a 10-nm thick continuous film, a
nucleation density of 2.5x1011/cm2 is required, if island growth is assumed isotropic.
44
3-2. Nucleation Density and Overpotential
The phenomenon of electrochemical nucleation has been treated from a
thermodynamics point of view by Budevski et al17. The free energy of formation for a
nucleus (N), containing ‘n’ atoms can be described by contributions from overpotential
(η) and surface energy (Φ) :
G (n)  nzF    (n)
(3-2)
Here, z is the number of electrons transferred and F is Faradays constant. The driving
force or overpotential term is always negative and lowers the free energy of cluster
formation. Excess surface energy term is positive and takes into account the increase in
free energy due to formation of a new phase, i.e. formation of new nuclei from a bulk
phase. Effects of these competing forces on free energy of formation are illustrated in
Figure 3-3. The surface excess energy was taken proportional to n2/3 and the
proportionality constant set arbitrarily to 96.5 KJ14. Two conclusions can be readily
reached by observing Figure 3-3. Firstly, high overpotential lowers the barrier for
formation of stable nuclei. Secondly, size of critical nuclei, i.e. number of atoms required
for a cluster or nucleus to be stable, also decreases at high overpotential. A stable
nucleus is defined as the n at which
G
 0 . The rate of nucleation then can be defined
n
by an Arrhenius rate relationship:
J
N
G
 AJ exp(
)
t
kT
(3-3)
Where AJ is a proportionality constant. High nucleation rate and nucleation density
therefore can be obtained by application of high overpotential as free energy barrier for
45
stable nuclei formation decreases. This can be achieved by selection of appropriate
plating chemistry and waveform.
Figure 3-3: Change in Gibbs free energy for formation of a nucleus with n atoms at
different overpotential values.
46
3-3. Alkaline Copper-Complex Chemistry
High deposition overpotential facilitating a high nucleation rate can be achieved
by strongly complexing the copper in solution. The chelation effect between copper and
the complexing agent greatly reduces the concentration of free Cu ions in solution. This
is described by the stability constant of the copper complex in solution:
 [Cu x y  M n m ]xy  mn 
pKCux  M n  log 
y x
m n 
 [Cu ] [ M ] 
(3-4)
Where Cu x  M n is the copper complex molecule. Reduction in the concentration of free
copper ion in solution shifts the equilibrium reduction potential to a more cathodic
direction. This is illustrated by the Nernst equation:
Eeq  E0 
RT aCu 
ln
zF
aCu
(3-5)
Where Eeq is the equilibrium reduction potential, R is the ideal gas constant, T is
temperature of the solution and a is the chemical activity of the species, which correlates
closely to its concentration in solution. E0 was taken as the open circuit potential of a Cu
electrode in the copper-tartrate electrolyte. The complexing agent that we have used to
enable direct deposition of Cu onto Ru is tartrate ( pK Cu Tartrate  5.6 24, Figure 3-4). The
copper-tartrate electrolyte has been well characterized and shown to be capable of
directly plating Cu onto Zn with high nucleation density18. From the polarization curve
in Figure 3-5, Eeq of Cu from tartrate is observed to be -0.14V (vs. SHE), significantly
more cathodic than the equilibrium reduction potential of Cu from acidic non-complexed
electrolyte (0.3 V vs. SHE).
47
Cu
Figure 3-4: A copper-tartrate molecule. Each copper molecule binds with two tartrate
molecules. Figure adapted from ref. 19.
The deposition overpotential, η, can then be defined as:
  V  Eeq  
(3-6)
Where V is the applied potential and φ is electrostatic potential measured at the interface
of double layer and electrolyte. The ohmic potential drop (R) in solution is calculated
using Newman’s equation20 for a measured solution resistance at a disk electrode. In
most cases, the ‘iR drop’ in solution is less than 30 mV for the given disk geometry used
here. Copper tartrate electrolyte demonstrates significantly larger polarization for copper
deposition compared to acidic CuSO4 electrolyte. For example, at a cathodic current
density of 10 mA/cm2, overpotential for copper deposition from the copper-tartrate
electrolyte is approximately 795 mV while for the acidic CuSO4 electrolyte the
overpotential is only 4 mV (Figure 3-5).
48
0
-10
i (mA/cm2)
-20
-30
Cu-Tartrate
-40
-50
CuSO4
-60
-1.5
-1.0
-0.5
0.0
0.5
Applied Potential (V vs. SHE)
Figure 3-5: Polarization curves for copper-tartrate and acidic CuSO4 solutions.
Concentrations and pH of the electrolytes are described in section 2-1 and 2-2. High
deposition overpotential is observed in the Cu-tartrate electrolyte.
49
3-4. Galvanostatic Pulse Waveform
In our work, we further increase the overpotential for Cu deposition by employing
a galvanostatic pulse waveform instead of DC current. In this scheme, extremely high
current density/overpotential pulses can be applied because the transient copper depletion
at the electrode is time-dependent. The pulse width (ton) can be set such that the reducing
species (Cu) does not deplete to zero at the electrode for a given pulse current density
reaching mass transfer limitation. The non-steady state diffusion at the cathode is
governed by Fick’s 2nd law:
CCu
 2CCu
 DCu
t
x 2
Where DCu is the copper diffusion coefficient. For a galvanostatic pulse and a stagnant
electrolyte, the following boundary conditions apply:
CCu  CB ,Cu  x  0, t  0
CCu  CB ,Cu  x  , t  0
CCu
i
 Pulse  x  0, t  0
x
nFDCu
Where CB,Cu is the bulk copper concentration, ipulse is the pulse current density. Since
nucleation occurs at a time scale that is much smaller (<1 s) than the time it requires to
establish the hydrodynamic boundary layer (>10 s) the assumption of no convection is
appropriate. Migration effects are also assumed to be negligible in a well-supported
electrolyte21. Double layer charging time for a typical capacitance of 50 µF/cm2 is in the
range of 1-10 µs22, orders of magnitude smaller than the applied pulse width. Therefore,
damping of the applied current due to the capacitive components can also be neglected in
50
this analysis. A more detailed analysis and review of the mass transfer effects in pulse
plating is presented in chapter 8 of the dissertation. Integration of Fick’s non-steady state
diffusion equation with the boundary conditions yields Sand’s equation23 for CCu at the
electrode (x=0):
1/2
CCu
 t 
2i
 CB ,Cu  Pulse  on 
nF   DCu 
(3-7)
In equation 3-7, CCu can be set to zero and a critical time parameter, defined as τ, can be
obtained:

CB ,Cu 2 n 2 F 2 DCu
4iPulse 2
(3-8)
For a given pulse current density, concentration of copper at the electrode reaches zero
after a time period τ, i.e. the mass transfer limit has been reached. This is illustrated in
Figure 3-6. The shaded region represents the range of ipulse and ton values that can be used
without completely depleting copper at the electrode during nucleation.
51
Critical Time (s)
100
10-1
10-2
ton < 
10-3
10-4
0.5
1.0
1.5
2.0
2.5
3.0
Pulse Current Density (A/cm2)
Figure 3-6: Pulse current densities and respective critical pulse widths, τ. CB,Cu=0.1 M
Cu and DCu=5  10-6 cm2/s. The shaded region represents the ranges of ipulse and ton that
can be used without completely depleting copper at the electrode.
52
3-5. Results and Discussion
Nucleation densities of Cu on Ru from the acidic and alkaline electrolytes under
DC and non-DC waveforms are reported in Figure 3-7. An improvement by 105/cm2 in
nucleation density of Cu on Ru is observed as plating bath and waveform are switched
from deposition under DC cathodic current applied to non-complexed, acidic electrolyte
to deposition under very short high current density pulses (ms) from complexed, alkaline
electrolyte. Volmer-Weber type island growth is observed when Cu is electrodeposited
from non-complexed acidic bath at low current density (Figure 3-7a). Large nuclei and
preferential nucleation of Cu upon itself is observed that give rise to large copper islands.
This is in agreement with the claim made in chapter 3-2, that a stable nuclei will be large
(>100 nm) at low overpotential conditions (η~0.1 V in Figure 3-7a). Nuclei density was
counted to be ~106 nuclei/cm2, similar to what has been reported for acidic copper
electrolyte. Nuclei density increases only five times when the alkaline copper-tartrate
bath is used (Figure 3-7b). Similar Volmer-Weber type growth for Cu nucleation on Ru
is also observed from the copper-tartrate electrolyte using a DC waveform (η~0.4 V).
Pulsing significantly affects the nucleation mode and increases Cu nuclei density on Ru
(Figure 3-7c). Average nuclei sizes are much smaller and less three-dimensional islands
are observed compared to the DC plating condition (η~1 V in Figure 3-7c). As can be
seen in Figure 3-7d, Nuclei density in the order of 1011 nuclei/cm2 is obtained when high
current density pulses are used to nucleate Cu from the copper-tartrate electrolyte. Nuclei
sizes are 10-15 nm in diameter and no large three-dimensional islands are observed
indicating high nucleation rate at high plating overpotential (η~3.9 V).
53
(a)
(b)
6
2
N~1.2 x 10 /cm
6
2
N~6 x 10 /cm
(d)
(c)
7
2
N~2.5 x 10 /cm
11
2
N~1.6 x 10 /cm
Figure 3-7: Cu nucleation on Ru from (a) non-complexed acidic bath using DC current.
10 mA/cm2 was applied for 5s (b) alkaline copper-tartrate bath using DC current. 10
mA/cm2 was applied for 5s (c) acidic bath using galvanostatic pulse waveform. Each
pulse was 0.5 A/cm2 in amplitude and 1 ms in duration. Off time was 49 ms (d) alkaline
copper tartrate bath using galvanostatic pulse waveform. Each pulse was 2.5 A/cm2 in
amplitude and 0.2 ms in duration. Off time was 49.8 ms. Total charge applied for all
experiments were 50 mC/cm2. Average current density during each pulse period was 10
mA/cm2.
In addition to tartrate, copper was nucleated on ruthenium also from an
ethylenediaminetetraacetate (EDTA) complexed electrolyte ( pK Cu  EDTA  18.7 24). The
electrolyte comprised of 0.1 M Cu and 0.2 M EDTA with pH adjusted to 12 by addition
54
of NaOH. Observed nuclei density of Cu on Ru from the EDTA complexed electrolyte
was of the same order of magnitude as that of tartrate complexed electrolyte. Under
experimental conditions identical to Figure 3-7d, nuclei density of 1.2  1011/cm2 was
obtained (Figure 3-8). Note that the SEM micrograph in Figure 3-8 has been magnified
by 4  , compared to Figure 3-7, for a better visualization of individual nuclei.
Figure 3-8: Copper nucleation on ruthenium from an alkaline Cu-EDTA electrolyte
using high current density pulsing. Galvanostatic pulse waveform was identical to the
one used in Figure 3-7d.
55
3-6. Conclusions
Novel alkaline copper-complex chemistry has been developed that provides high
polarization for copper deposition, enhancing the nucleation density of Cu on noble metal
substrate such as ruthenium. Nucleation densities in the order of 1011/cm2 were obtained
from the Cu-tartrate and Cu-EDTA electrolytes using high current density pulsing,
significantly higher than that from acidic copper solution with DC current. The alkaline
chemistry combined with pulsing can enable direct plating of Cu on Ru, eliminating the
need for a vapor deposited copper seed layer.
56
References:
1. A. Radisic, Y.Cao, P. Taephaisitphongse, A. C. West and P. C. Seasrson, J.
Electrochem. Soc., 150, C362 (2003).
2. A. Radisic, J.G. Long, P. M. Hoffman and P. C. Searson, J. Electrochem.
Soc.,148, C41 (2001).
3. J. Kelber, S. Rudenja, C. Bjelkevig, Electrochim. Acta, 51, 3086 (2006).
4. O. Chyan, T.N. Arunagiri and T. Ponnuswamy, J. Electrochem. Soc., 150 C347
(2003).
5. M. Zheng, M.Willey and A.C. West, Electrochem. Solid-State Lett., 8, C151
(2005).
6. L. Guo, A.Radisic and P. C. Searson, J. Electrochem. Soc., 153, C840 (2006).
7. U. Emekli and A.C. West, J. Electrochem. Soc., 157, D257 (2010).
8. U. Emekli and A.C. West, Electrochim. Acta,. 54, 1177 (2009).
9. S. Armini, J. Electrochem. Soc., 158, D390 (2011).
10. S. Armini, Z. El-Mekki, K. Vandersmissen, H. Philipsen, S. Rodet, M. Honore, A.
Radisic, Y. Civale, E. Beyne and L. Leunissen, J. Electrochem. Soc., 158, H160
(2011).
11. R. Akolkar, T. Indukuri, J. Clarke, T. Ponnuswamy, J. Reid, A. J. McKerrow, S.
Varadarajan, “Direct seed electroplating of copper on ruthenium liners”,
Interconnect Technology Conference and 2011 Materials for Advanced
Metallization (IITC/MAM), 2011 IEEE International , pp.1-3, 8-12 May 2011.
57
12. A. Joi and U. Landau, “An Alkaline Copper Plating Process Providing High
Nucleation Density on Ru and Bottom-up Fill”, Abstract #1944, 220th ECS
Meeting, 9-14 October 2011, Boston, MA.
13. L. Boehme, J. Wu, X. Kang, R. Preisser and U. Landau, “The Impact of
Electrolyte Acidity on Bottom-up Metallization of Copper Interconnects”,
Abstract #2734, 222nd ECS Meeting, 7-12 October 2012, Honolulu, HI.
14. E. Budevski, G.Staikov and W.J. Lorenz, Electrochemical Phase Formation and
Growth: An Introduction to the Initial Stages of Metal Deposition. Vol. 5, VCH,
New York (1997).
15. Z. W. Sun, R. He and J. O. Dukovic, “Direct Plating of Cu on Ru: Nucleation
Kinetics and Gapfill Chemistry”, Advanced Metallization Conference (2004), pp.
531-537.
16. L. Guo and P. C. Searson, Langmuir, 24, 10557 (2008).
17. E. Budevski, G. Staikov and W. J. Lorenz, Electrochim. Acta, 45, 2559 (2000).
18. Chi-Hong Liao, “An Environmentally Friendly Electroplating Process of Copper
from an Alkaline Solution”, Dissertation (2012).
19. http://en.wikipedia.org/wiki/File:Copper-tartrate-complex-3D-balls.png
20. J. Newman, J. Electrochem. Soc., 113, 501 (1966).
21. J. Newman, Electrochemical Systems, 3rd Ed., Wiley, New York (2004).
22. N. Ibl, Surf. Technol., 10, 81 (1980).
23. H. J. S. Sand, Philos. Mag., 1, 45 (1901).
24. Martrell, A. R. S., Critical Stability Constants. Vol. 2, Plelum press, New York
(1974).
58
Chapter 4
Additives’ Characterization in Alkaline Medium
In order for bottom-up fill to occur in features plated from any copper electrolyte,
including the alkaline complexed copper electrolyte, special plating additives are
required. A suppressor that inhibits copper deposition at the feature top surface and rim,
and an anti-suppressor that provides uninhibited growth at the feature bottom (and
subsequently displaces the suppressor) are needed to facilitate bottom-up growth.
Conventional additives used for bottom-up fill in acidic copper medium, e.g.
Polyethylene glycol (‘PEG’) as suppressor and bis-sulfopropyl sulfonate (‘SPS’) as antisuppressor might not function as intended in the alkaline complexed medium. A
convenient and widely used diagnostic tool for characterizing the suppressing and antisuppressing characteristics of additives is injection method1. In this technique, a rotating
disk electrode having well defined transport (rotation speed) is operated at a galvanostatic
or potentiostatic mode. Once a steady state is reached, a mixture of additives having a
predetermined composition and volume is injected into the solution using a syringe. The
current or potential response is recorded until a new steady state is reached. Typical
transient and steady state polarization behaviors of PEG and SPS in a conventional acidic
copper plating bath are shown in figure 4-1. The plating bath must also contain chloride
ions (~50 ppm) which are required to activate the PEG so that it will provide the expected
polarization2. The increased cathodic polarization of the electrode (by almost 250 mV)
after PEG injection, as noted in figure 4-1, indicates highly passivated plating in the
presence of PEG. This has been attributed to a Cu-Cl-PEG layer formation on the
electrode that hinders copper deposition3. Once a steady state is attained, injection of
59
SPS into the PEG and chloride containing electrolyte slowly depolarizes the electrode.
The SPS depolarization occurs over a much larger time scale (> 60 s) than PEG
polarization, which reaches saturation about a second or less (Figure 4-1). While the
molecular level mechanism for this process is not understood, some theories have been
proposed. Moffat et al. theorized that laterally mobile and structurally intact SPS adsorbs
on the chloride adlayer, disrupting the PEG-Cl complex formation4. Broekmann et al.
proposed a two-step mechanism where SPS adsorption on the chloride adlayer is
followed by SPS cleavage (into MPS) and a surface confined reaction involving chloride
that disrupts the Cu-Cl-PEG layer5. Whatever the molecular level mechanism may be, it
is evident from injection studies that SPS partially eliminates the polarization of PEG and
is an effective anti-suppressor. For bottom-up fill to occur, an additives combination of a
suppressor and anti-suppressor with the appropriate adsorption and transport properties
(such as PEG and SPS in acidic solutions) needs to be identified, demonstrating similar
polarization behavior in the alkaline complexed copper electrolyte.
60
0.00
Overpotential (V)
-0.05
70 ppm
Cl-
100 ppm PEG 4000 injected
-0.10
-0.15
-0.20
pH~2
-0.25
50 ppm SPS injected
-0.30
100
200
300
400
500
600
700
Time (s)
Figure 4-1: Transient polarization behavior of PEG and SPS in the acidic copper plating
bath containing 70 ppm chloride. Injection of 100 ppm PEG (MW=4000 g/mol) strongly
polarizes the electrode. Injection of 50 ppm SPS into the PEG + Cl- containing bath
depolarizes the electrode. Average current density was 10 mA/cm2 in copper plating
from a 0.5 M acidified (pH=2) copper sulfate solution. The disk cathode was rotated at
400 rpm. Figure was adapted from ref [14].
Injection studies in the alkaline complexed electrolyte were performed at a
constant current density of 20 mA/cm2. During bottom-up plating, current densities at
the bottom of the feature can be as high as 100 mA/cm2 while current densities at the top
of the features can be as low as 2 mA/cm2. Therefore a galvanostatic injection
experiment on a flat electrode does not accurately simulate the environment present
during bottom-up growth. This technique is adequate, however, for characterizing
61
qualitatively the suppressive and anti-suppressive qualities of additives at a given average
plating current density. Operating the injection experiment in potentiostatic mode
resulted in data that were too noisy, the reason for which is not clearly understood.
Hence, galvanostatic injection studies were utilized to characterize additives in the
alkaline medium.
62
4-1. Polarization Behavior of Polyether Additives in Alkaline Complexed Medium
Polarization behaviors of two polyether additives were characterized in the
alkaline complexed Cu electrolytes using injection studies. These additives were
polyethylene glycol (PEG) and Polyoxyethylene Lauryl Ether (PLE). PLE has been
shown to provide significantly higher polarization than PEG in the acidic electrolyte6.
The results of PEG and PLE injections into the copper-tartrate electrolyte containing 70
ppm chloride at pH=13 are shown in figure 4-2. Additives were injected into the plating
bath in increments until a final concentration of 1100 ppm in solution is reached,
ensuring complete saturation. The polarization generated by PEG on the electrode is
minimal in the explored range, reaching only 50 mV. Similar polarization behavior on
the electrode is observed in PLE injection studies up to a concentration of 500 ppm.
However, Substantial polarization (500 mV) is eventually observed when injecting much
higher dosage of PLE, exceeding 1100 ppm. As shown in figure 4-2, the onset of this
polarization is relatively slow, rendering it impractical for dual damascene application,
where the entire via fill is completed in just a few seconds.
63
Figure 4-2: PEG and PLE injection studies in the copper tartrate electrolyte in the
presence of chloride ions. Significant polarization on the electrode is observed with PLE
at 1100 ppm concentration. Plating current density was 20 mA/cm2 and the rotation
speed was 200 rpm. Plating solution comprised of 0.1 M copper sulfate, 0.5 M tartrate, at
pH=12.5. Chloride concentration was 50 ppm.
Similar injection studies with PEG and PLE were carried out in a copper-EDTA
electrolyte as well (Figure 4-3). PEG and PLE demonstrated less passivation in copperEDTA than in the copper-tartrate. PEG only polarizes ~30 mV at 1100 ppm in the EDTA
solution while PLE polarization is slightly higher, ~50 mV at the same concentration.
64
-0.80
600 ppm
Overpotential (V)
-0.85
PEG
100 ppm
-0.90
-0.95
-1.00
200
PLE
400 ppm
Cu-EDTA + Cl- solution
300
400
500
600
700
800
900
Time (s)
Figure 4-3: PEG and PLE injection studies in the copper EDTA electrolyte. 50 ppm
chloride ions were present. Less than 50 mV of polarization is observed even with 1100
ppm of additives. Plating current density was 20 mA/cm2 and the rotation speed was 200
rpm. Plating solution comprised of 0.1 M copper sulfate, 0.2 M EDTA, at pH=12.5.
Use of such high concentrations of polyether suppressors during plating is not
practical. Typical bulk suppressor concentration used in feature fill is approximately 50100 ppm, significantly smaller than the concentration needed to suppress copper
deposition in the alkaline medium using PLE or PEG. Using such a high concentration of
additives might also introduce significant amount of impurity in the plated copper
increasing the copper resistivity. The lack of polarization in the alkaline medium is likely
due to the absence of adsorbed chloride on the copper surface at pH~12. At such a high
65
pH, hydroxide adsorption is likely to dominate chloride adsorption, similarly to
observations made by Lipkowski et al. on Au electrode7.
66
4-2: Polarization Behavior of SPS in Alkaline Complexed Medium
Polarization behavior of SPS, conventionally an accelerator or anti-suppressor in
the acidic medium (chloride has little effect on SPS activity in acidic solutions), has also
been characterized in the alkaline copper tartrate electrolyte using injection studies. No
chloride was present in the electrolytic solution since hydroxide adsorption takes
precedent over chloride adsorption in alkaline medium, rendering its effects in the
alkaline medium insignificant. This is supported by experimental evidence indicating
similar polarization characteristics with and without the presence of chloride ions in the
alkaline solution. Transient and steady-state overpotential at the electrode after SPS
injection into the alkaline copper tartrate electrolyte are reported in figure 4-4. Injection
of 15 ppm SPS into the copper tartrate electrolyte (pH=12.5) rapidly (within 10 s)
produces strong suppression, almost 400 mV, at the electrode. This is markedly larger
that the polarization observed in the acidic copper medium in the presence of PEG and
PLE. Effect of SPS inhibition at a lower pH (pH=10) is slightly diminished. Higher SPS
concentration is required to produce the maximum inhibition and the time to reach
saturation is longer, i.e. the sharp transition seen at a higher pH is not observed. This is
possibly due to a different configuration of copper tartrate complexes in solution at
higher pH values. At pH<12, copper in solution is complexed primarily as CuT24 .
However, at pH>12, formation of CuT2 (OH ) 62 complexes8 might alter the SPS
adsorption process on the copper electrode resulting in the sharp transition observed at
pH=12.5.
67
-0.6
Overpotential (V)
-0.7
15 ppm SPS injected
5 ppm SPS injected
-0.8
-0.9
-1.0
-1.1
400 mV
pH=10
pH=12.5
-1.2
-1.3
100
150
200
250
300
350
400
Time (s)
Figure 4-4: SPS injection study in copper tartrate electrolytes having pH values of 10
and 12.5. A sharp transition is observed at the high pH value, after SPS injection.
Injection studies were performed at a current density of 20 mA/cm2 with a rotation speed
of 200 rpm. Solution composition was identical to that in Figure 4-2 without chloride
ions.
SPS saturation study was also performed to determine the concentration of SPS at
which the polarization at the copper electrode becomes saturated, i.e. does not change
anymore with the addition of SPS. SPS was injected into the copper tartrate electrolyte in
1-2 ppm increments until no change in the steady state potential was observed (Figure 45). Electrode polarization reaches saturation at ~6 ppm SPS and no further drop in
electrode potential was observed upon adding more SPS into the solution (final total
concentration of SPS in solution was 20 ppm).
68
-0.6
-0.7
Overpotential (V)
-0.8
1 ppm
2 ppm
-0.9
-1.0
1 ppm
2 ppm
2 ppm 2 ppm
-1.1
10 ppm
-1.2
-1.3
100
150
200
250
300
350
400
450
500
550
600
Time (s)
Figure 4-5: Saturation of the polarization effect at a copper electrode in the copper
tartrate electrolyte (pH=12.5), upon addition of SPS. SPS was injected in 1-2 ppm
increments until a total concentration of 20 ppm in solution is achieved. Polarization
reaches saturation at 6 ppm of SPS. Addition of more SPS had no further effect on the
electrode polarization. Current density and rotation speed were 20 mA/cm2 and 200 rpm,
respectively. Solution composition was identical to that in Figure 4-4.
The SPS inhibition phenomenon was found to be unique to the tartrate complexed
solution. SPS injection studies were performed in solutions where copper was
complexed with different complexing agents (Figure 4-6); trisodium citrate ( Na3C6 H 5O7
), diethylenetriamine pentaacetic acid (DTPA, C14 H 23 N 3O10 ) and tetrasodium
etheylenediamine tetraacetate (EDTA, 4NaC10 H12 N 2O8 ). The molar concentration ratios
69
of copper to the complexing agents in solution and pH of the electrolytes are provided in
table 4-1. Injection of 15 ppm SPS into the citrate, DTPA and EDTA complexed
electrolytes produced negligible polarization at the electrode (20-60 mV). Significant
polarization was only observed in the copper-tartrate electrolyte, producing almost 400
mV of suppression. This indicates that the SPS inhibition phenomenon is specific to the
tartrate anion or to the tartrate complexed copper. Overpotential values were recorded at
a constant current density of 20 mA/cm2. With the exception of tartrate, neither alkalinity
nor generic complexation appear to cause significant polarization at the electrode due to
SPS.
Table 4-1: Electrolyte composition, pH and overpotential (Δη) at the electrode for the
injection studies shown in Figure 4-6. All overpotential values correspond to a current
density of 20 mA/cm2.
Complexing
Agent (CA)
Citrate
Cu : CA (molar)
in solution
1:5
5.5
Overpotential
Δη (mV)
20
DTPA
1:2
12.5
50
EDTA
1:2
13.0
60
Tartrate
1:5
12.5
400
pH
70
-0.6
-0.7
Cu-Citrate, pH=5.5
Overpotential (V)
-0.8
20 ppm SPS
Cu-DTPA, pH=13
-0.9
10 ppm SPS
Cu-EDTA, pH=12.5
-1.0
-1.1
Cu-Tartrate, pH=12.5
-1.2
-1.3
140
160
180
200
220
240
Time (s)
Figure 4-6: SPS injection effect on cathodic copper polarization in citrate, EDTA, DTPA
and tartrate complexed alkaline copper electrolytes. Molar ratios of copper to the
complexing agents in solution are provided in table 4-1. SPS exhibits significant
inhibition only in the tartrate complexed electrolyte. Plating conditions were identical to
those in Figure 4-5.
The inhibitive effect of SPS on copper deposition from the tartrate complexed
electrolyte was further characterized by injection studies. SPS ( to generate 20 ppm in
solution) was injected into three different electrolytes having varied concentrations of
tartrate in solution (Figure 4-7). In the low concentration solution (0.2 M tartrate),
tartrate was stoichiometrically complexed to 0.1 M copper and no free tartrate ion was
present in solution. In the high concentration solutions (0.4 and 0.5 M tartrate), excess
tartrate ions, 0.2 M and 0.3 M tartrate, respectively, were present in solution. It was
71
observed that SPS exhibited similar polarization in all three electrolytes, regardless of the
concentration of tartrate in solution with overpotential between 350 to 400 mV for all
three cases. This indicates that the SPS passivation effect is independent of the tartrate
concentration in solution and the presence of excess free tartrate anions does not affect
the polarization at the electrode due to SPS adsorption. This also suggests that the
observed behavior relates to surface modification due to additives adsorption rather than
to bulk chemistry effects.
-0.6
-0.7
0.2 M Tartrate
Overpotential (V)
-0.8
Δη=350-400 mV
-0.9
-1.0
-1.1
20 PPM
SPS
0.5 M Tartrate
0.4 M Tartrate
-1.2
-1.3
180
200
220
240
260
280
300
Time (s)
Figure 4-7: Polarization associated with SPS injection into 0.2, 0.4 and 0.5 M tartrate
solutions. Copper concentration was 0.1 M in all three solutions. pH in all three
solutions was 12.5. SPS was injected to generate a 20 ppm solution. Overpotential
changes were approximately the same in all three solutions. Experimental conditions
were identical to those in Figure 4-5.
72
It is evident that the copper-tartrate complex plays an essential role in the SPS
inhibition effect observed at the electrode. This is further illustrated in figure 4-8 and 49. To investigate the specific role of the copper-tartrate complex in inhibiting copper
deposition, injection studies (20 ppm SPS injection) were performed in electrolytes
containing several different concentrations of tartrate and EDTA complexant. Since
copper forms a much stronger complex with EDTA (pKa=18.6) than with tartrate
(pKa=5.8), relative amount of the copper-tartrate complex can be varied by modifying the
EDTA concentration in solution. For example, in a 0.1 M Cu electrolyte containing 0.17
M EDTA and 0.05 M tartrate, 0.085 M Cu will be complexed via the copper-EDTA
complex (1:2 stoichiometric ratio) while the rest will be complexed via tartrate. It is
clearly observed in Figure 4-8 that passivation at the electrode increases markedly with
an increase in copper-tartrate complex in solution and reaches saturation when 100% of
the copper in solution is complexed via tartrate, as opposed to EDTA (Figure 4-9).
73
-0.6
-0.7
Overpotential (V)
-0.8
0.11 M EDTA + 0.11 M Tartrate
0.17 M EDTA + 0.05 M Tartrate
-0.9
-1.0
-1.1
0.05 M EDTA + 0.17 M Tartrate
-1.2
0 M EDTA + 0.5 M Tartrate
-1.3
180
200
220
240
260
280
Time (s)
Figure 4-8: Polarization response to SPS injection into 0, 0.05, 0.11 and 0.17 M EDTA
solutions. Copper concentration was 0.1 M in all four solutions. 20 ppm SPS was
injected at 200 s. Maximum polarization was observed when no EDTA was present in
solution . Experimental conditions were identical to those in Figure 4-5.
74
Relative Suppression (η/ηmax)
Figure 4-9: Relative suppression effect for the electrolytes in Figure 4-8 after 20 ppm
SPS injection. max was taken as the suppression observed in the 0 M EDTA and 0.5 M
tartrate electrolyte after SPS injection. Fraction of total copper complexed to tartrate in
[ EDTA]
2
.
2
[Cu ]
[Cu 2 ] 
solution (Cu-T) was calculated using
The inhibition effect of SPS on copper deposition is, however, not unique. It is
known that, in the absence of chloride ions, SPS provides mild inhibition in an acidic
media9. However, in the alkaline medium, in the presence of tartrate, this effect is much
more pronounced. While the molecular level mechanism of the SPS induced inhibition
process in the alkaline tartrate medium is not currently known, a hypothesis may be
proposed based on the injection studies described earlier. The pronounced inhibition
75
might be due to the interaction of adsorbed SPS moieties with the CuT24 . This could
lead to formation of a surface passive film which sterically hinders Cu electrodeposition,
a mechanism similar to the SPS surface reaction mechanism propsed by Broekmann et al
in the acidic medium5.
76
4-3: Identification and Characterization of an Anti-suppressor in the Alkaline
Medium
Now that SPS has been identified and characterized as an effective suppressor in
the alkaline copper tartrate electrolyte, an anti-suppressor is required for the alkaline
electrolyte so that it could negate the inhibiting effect of SPS on copper deposition. This
would provide an analogous additives pairing in the alkaline media, similar to that of
PEG/SPS for bottom-up fill in the acidic electrolyte. The desired traits of an effective
anti-suppressor can be identified and characterized via injection studies. Injection of an
effective anti-suppressor into a SPS containing electrolyte will slowly depolarize the
electrode, negating the suppression provided by SPS. This is essential for displacing the
inhibiting additive as the deposit is progressing from the bottom towards the via rim.
Transient and steady state polarization should appear similar to the shape observed in
Figure 4-1 where 50 ppm SPS had been injected into the PEG containing acidic copper
electrolyte. In addition to deactivating the SPS, the anti-suppressor must also exhibit
minimal polarization on a bare copper electrode in a fresh electrolyte (no SPS), so as to
facilitate unhindered copper deposition at bottom of the features.
In conventional acidic process, nitrogen containing molecules such as
polyethyleneimine (PEI), typically known as ‘levelers’, are used to negate the
accelerating effect of SPS and prevent bump formation after feature fill. This
‘neutralizing’ effect has been attributed to an interfacial anion/cation ion pairing
interaction between cationic PEI and anionic sulfonate group of the SPS in the acidic
medium10. These levelers are also strong suppressors, however unlike the PEG-Cu-Cl
77
ensemble, PEI forms an irreversible adduct at the interface that cannot be deactivated by
SPS11.
It might be possible to exploit the electrostatic interaction phenomenon which
occurs between SPS and PEI to achieve the desired antagonistic behavior required for
bottom-up growth. If PEI behaves similarly in alkaline medium, as it does in the acidic
one, injection of PEI into a SPS containing alkaline tartrate electrolyte will slowly
depolarize the electrode. This hypothesis was experimentally verified in figure 4-10,
showing different concentrations of PEI (40 and 80 ppm) injected into the electrolyte into
which 15 ppm SPS had been injected earlier. PEI indeed depolarizes the inhibition
provided by SPS, albeit partially. At higher PEI concentration (80 ppm), more electrode
depolarization is observed indicating a higher degree of SPS deactivation. Therefore, the
rate and the extent of SPS deactivation can be tuned by choosing a suitable concentration
PEI for a given amount of SPS.
78
-0.6
-0.7
Overpotential (V)
15 ppm SPS
-0.8
-0.9
80 ppm PEI
-1.0
-1.1
-1.2
40 ppm PEI
-1.3
200
220
240
260
280
300
320
340
360
Time (s)
Figure 4-10: Electrode polarization response to injection of 15 ppm SPS into the
alkaline copper tartrate electrolyte followed by 40 ppm and 80 ppm PEI injection. PEI
partially depolarizes the electrode, neutralizing the suppressing effect of SPS. Current
density and rotation speed were 20 mA/cm2 and 200 rpm, respectively. The electrolyte
composition was identical to those in figure 4-5.
A secondary requirement for an effective anti-suppressor is that it must not by
itself inhibit copper deposition, at least not to the same extent as the suppressor for
bottom-up growth to occur. In the acidic medium, PEI and PEG exhibits comparable
suppression (  PEG  212 mV vs.  PEI  171 mV)11. Therefore, it is of importance to
characterize the polarization behavior of PEI on a bare copper electrode in SPS-free
alkaline tartrate electrolyte. This is shown in figure 4-11. 40 ppm of PEI was injected
into SPS-free electrolyte and transient and steady state polarization at the electrode were
79
recorded. No steady state polarization of the copper electrode was observed, confirming
suitability of PEI as an anti-suppressor in the alkaline medium.
-0.6
-0.7
Overpotential (V)
-0.8
40 ppm PEI
-0.9
-1.0
-1.1
-1.2
-1.3
200
240
280
320
360
400
Time(s)
Figure 4-11: Polarization following the injection of 40 ppm PEI into SPS free copper
tartrate electrolyte after 200 s of plating. PEI does not produce any steady-state
polarization at the copper electrode. Current density and rotation speed were 20 mA/cm2
and 200 rpm, respectively. The electrolyte composition was identical those in figure 410.
Fill times for dual damascene interconnect features (trenches or vias) from
conventional copper chemistry are typically less than a minute. Therefore, a sequential
injection of the suppressor and anti-suppressor during plating is not a practical scenario.
In industrial processes, additives (organic) are pre-mixed in additives-free make-up
solution (‘VMS’, copper sulfate electrolyte containing sulfuric acid and chloride) from
80
which bottom-up fill is accomplished. Hence, a more representative simulation of the
actual plating conditions can be obtained by simultaneous injection study. Predetermined concentrations of SPS and PEI are mixed and the resulting solution is injected
into the alkaline copper plating bath. The transient and steady state overpotential at the
electrode after the simultaneous injection is shown in figure 4-12a. An initial suppression
due to fast SPS adsorption is observed which is slowly depolarized by the ‘neutralizing’
effect of PEI. A similar interaction is observed in the acidic medium with PEG-SPS
(Figure 4-12b), indicating further the suitability of SPS and PEI for bottom-up fill in an
alkaline tartrate medium.
-0.6
(a)
Overpotential (V)
-0.7
15 ppm SPS + 40 ppm PEI
-0.8
-0.9
-1.0
-1.1
-1.2
-1.3
180
200
220
240
260
280
300
Time (s)
(b)
Figure 4-12: (a) Polarization transient response to the injection of a pre-mixed 15 ppm
SPS and 40 ppm PEI solution in the alkaline copper-tartrate electrolyte (b) Polarization
81
response to injection of pre-mixed 70 ppm Cl-, 150 ppm PEG and 50 ppm SPS into the
acidic copper bath. (b) was adapted from ref. [13]. A fast suppression, followed by slow
depolarization is observed in both cases.
4-4. Discussion
To achieve bottom-up fill, the current density at the bottom of the features ( iB )
must be significantly larger than the current density on the sidewalls ( iSW ), exceeding the
aspect ratio of the features. For bottom-up fill in trenches, this can be written as:
iB
L

iSW W
(4-1)
Where L and W are the depth and width of the trench, respectively. Accordingly, the
bottom-up fill ratio can be approximated as the ratio iB/iSW 6. Assuming that deposition on
the sidewall occurs under SPS controlled kinetics and deposition at the bottom follows
PEI kinetics, current densities, iB and iSW , can be approximated for a given overpotential
using the respective polarization curves. Polarization curves of copper deposition from
SPS and PEI saturated copper tartrate electrolytes are shown in figure 4-14. SPS
saturated electrolyte exhibits significant polarization at the electrode in the Tafel range,
as observed earlier in the injection studies. Both PEI saturated electrolyte and no
additives containing electrolytes exhibit essentially similar polarization behavior. If the
plating rate at the bottom is given by the SPS polarization curve and the rate on the
sidewall by the PEI polarization curve, then at a nominal cathodic plating overpotential of
-1.1 V, a bottom-up fill ratio of ~11 is obtained (Figure 4-14). This is within the range of
values reported in the acidic medium with PEG-SPS 6,14.
82
In this simplified analysis, the intricate time dependent transport-adsorption
phenomenon of additives that is essential for bottom-up fill in small geometries is
neglected. This has been described in detail by Akolkar and Landau1 and implication of
the diffusion-adsorption mechanism for bottom-up fill in the alkaline medium with SPS
and PEI is discussed in the following chapter.
50
Current Density (mA/cm2)
40
No Additives
30
iB
100 ppm
PEI
20
10
iSW
50 ppm
SPS
0
0.6
0.8
1.0
1.2
1.4
1.6
Overpotential (V)
Figure 4-13: Polarization curves for copper deposition from an alkaline copper-tartrate
electrolyte containing either 50 ppm SPS, 100 ppm PEI and no additives. Significant
steady state polarization is observed in the SPS saturated electrolyte. PEI saturated
electrolyte does not exhibit any polarization relative to the fresh electrolyte. Rotation
speed of 200 rpm was used for all experiments. Electrolyte composition was identical to
those in figure 4-10.
83
4-5. Conclusions
An alkaline, tartrate-complexed copper electrolyte containing SPS and PEI as
additives exhibiting the required suppressor/antisuppressor interaction essential for
bottom-up fill is described. In contrast to their role in acidic electrolytes, in alkaline
medium, SPS acts as a suppressor and PEI acts as an anti-suppressor. The interactive
behavior of these additives is characterized through injection studies. While the SPS
suppresses the electrode surface rapidly, the PEI deactivates the SPS slowly –
antagonistic effects quite similar to the PEG-SPS interactions which are known to
provide bottom-up fill in conventional acid-copper electrolytes. This is summarized in
table 4-1.
Table 4-2: A summary of the additives combination needed for bottom-up fill in the
acidic and alkaline complexed copper electrolytes
Electrolyte Type
Suppressor
Anti-suppressor
Leveler
Acid-copper
PEG
SPS
PEI
Alkaline tartrate-copper
SPS
PEI
–
.
84
References:
1. R. Akolkar and U. Landau, J. Electrochem. Soc., 151, C702 (2004)
2. J. J. Kelly and A. C. West, J. Electrochem. Soc., 145, 3477 (1998)
3. Z.V. Feng, X. Li, A. A. Gewirth, J. Phys. Chem. B, 9415 (2003)
4. T.P. Moffat, L.Y.O. Yang, J. Electrochem. Soc., 157, D228 (2010)
5. N. T. M. Hai, T. T. M. Huynh, A. Fluegel, M. Arnold, D. Mayer, W. Reckien, T.
Bredow and P. Broekmann, Electrochim. Acta, 70, 286 (2012)
6. J. Mendez, R. Akolkar and U. Landau, J. Electrochem. Soc., 156, D474 (2009)
7. A. Chen and J. Lipkowski, J. Phys. Chem. B, 103, 682 (1999)
8. E. Norkus, A. Vaskelis, I. Zakaite and J. Reklaitis, Chemija, 2, 16 (1997)
9. W. -P. Dow, H. –S. Huang, M. –Y. Yen and H. –H. Chen, J. Electrochem.
Soc.,152, C77 (2005)
10. T.P. Moffat, D. Wheeler, S.K. Kim, D. Josell, J. Electrochem. Soc., 153,
C127 (2006)
11. N. T. M. Hai, K. W. Kramer, A. Fluegel, M. Arnold, D. Mayer, and P.
Broekmann, Electrochim. Acta,83, 367 (2012)
12. P. Broekmann, A. Fluegel, C. Emnet, M. Arnold, C. Roger-Goepfert, A. Wagner,
N. T.M. Hai and D. Mayer, Electrochim. Acta, 56, 4724 (2011)
13. James Adolf, ‘Modeling the Role of Plating Additives in the Metallization of
Semiconductor Interconnects: From Dual Damascene to Through Silicon Vias’,
PhD dissertation (2010).
85
14. L. Boehme, J. Wu, X. Kang, R. Preisser and U. Landau, “The Impact of
Electrolyte Acidity on Bottom-up Metallization of Copper Interconnects”,
Abstract #2734, 222nd ECS Meeting, 7-12 October 2012, Honolulu, HI
86
Chapter 5
Bottom-up Metallization by Copper Plating from Alkaline Media in the
Presence of Additives
5-1. Background
It has been stated earlier that the local current densities within the feature vary
during bottom-up plating. This variation is due to the position and time-dependent
additives coverage within the feature that modify the copper deposition kinetics. The
unsteady state additives surface coverage and the subsequent bottom-up fill due to
additives’ interaction has been modeled extensively for the PEG-SPS-Cl system in acidic
medium1-5. These models are discussed briefly herein so that a comparison between the
bottom-up fill mechanisms in acidic and alkaline media can be made. It can be gathered
from the injection studies that deactivation or displacement of the suppressor by the
antisuppressor occurs over a much longer time-scale (> 100 s) than the initial suppressor
adsorption (< 10 s). Based on that observation, two distinct regimes during feature fill
can be identified1,2: the initial competitive additives adsoprtion on bare copper until
surface becomes saturated and the following additives interaction that results in bottomup fill.
The first regime is dominated by unsteady-state transport and adsorption
characteristics of the additives (PEG and SPS) and occurs before any significant copper
growth takes place. PEG and SPS have significantly different transport and adsorption
properties. The anti-suppressor (SPS) is a relatively small, fast-diffusing molecule that
adsorbs relatively slowly on copper. As the wafer is immersed into the solution, anti-
87
suppressor diffuses into the features with little adsorption on the sidewalls and the flat
top, since its adsorption rate is relatively slow compared to its diffusion rate. The
suppressor (PEG), on the other hand, is a slowly diffusing, large organic molecule (M.W.
> 1000 g/mol) with extremely fast adsorption kinetics. Therefore, unlike SPS, as PEG
diffuses into the feature it immediately adsorbs onto the first available free copper sites
that aren’t occupied by SPS. Since PEG is the diffusion limited species, the time-scale
for this regime can be estimated as the time required for PEG to diffuse to bottom of the
features.
Actual feature fill occurs in the second regime and this process is dominated by
the unsteady state additives’ interaction. The additives’ interaction is accompanied by a
redistribution of additives on the surface due to two combined mechanisms: the area
reduction during plating which enhances SPS concentration at the bottom4,5 and
displacement of PEG by the stronger binding SPS which adsorbs from solution1,3. The
enhancement of SPS concentration due to the contracting surface, termed ‘CEAC
mechanism’, is based on the observation that SPS neither desorbs nor incorporates into
the deposit6. Instead, it stays preferentially bound to the bottom surface, whose area is
shrinking. This is illustrated schematically in figure 5-1. The surface concentration is
further increased by the displacement of the weaker bonding PEG by SPS as the bottomup growth continues. This displacement effect is observed in the injection studies as the
slow depolarization of the planar disk electrode. The relative importance of the two
mechanism (‘CEAC’ and ‘adsorption-interaction’) depends on the geometry of the
feature. Since transport and adsorption time-scale is inversely related to feature width or
radius1,2, the transient transport-adsorption mechanism for bottom-up fill becomes
88
prominent in high aspect ratio features (
1
L
 20 ) while its effects in larger features (
W
L
 10 ) can be negligible1.
W
Since all implications of these intricate additives’ adsorption-interaction
mechanism that are instrumental in achieving bottom-up fill cannot be captured by
injection studies on a planar electrode, plating tests were conducted on patterned wafer
coupons from the alkaline electrolyte containing SPS and PEI. In addition to providing a
direct evidence of bottom-up fill capability of the SPS-PEI system, feature fill studies
also contribute insight into the prevailing bottom-up fill mechanism. By varying the
geometry, i.e. the feature width, and observing the growth of the copper surface, the
nature of the SPS and PEI interaction may possibly discerned and the mechanism mode
can be characterized. For example, in low aspect ratio features where transport effects
are negligible, a preferential deposit growth at the bottom corners is indicative of the
localized corner acceleration effect described by the CEAC mechanism5. A flat bottomup growth that is characteristic of the effective bottom-up fill1,6 is also desired.
89
Figure 5-1: A schematic illustration of the ‘CEAC’ mechanism. Concentration of the
accelerator at the bottom increases as surface area contracts due to plating. After growth
to a certain distance, accelerator coverage reaches 100%.
90
5-2. Results
To characterize the role of antagonistic SPS-PEI interactions in generating the
bottom-up fill, plating studies were performed on patterned wafers from the alkaline
tartrate electrolyte containing SPS and PEI. Additives concentrations used in bottom-up
fill were chosen to be similar to those used in injection studies; 15-20 ppm SPS and 40
ppm PEI. No chloride ions were added as it had no effect on the additives’ adsorption
and interaction. The patterned coupons contained a wide range of feature sizes. This
enabled direct observation of different stages of copper bottom-up growth. In figure 5-2,
bottom-up growth in low aspect ratio feature (AR~1) from the alkaline tartrate electrolyte
with SPS and PEI is shown. Copper was plated at an average current density of 25
mA/cm2 for 45 s. One of the key characteristics of bottom-up fill is observed. The
triangular shape evolving at the bottom corners is indicative of the accelerated deposition
rate relative to rest of the surface. The enhanced deposition in the bottom corners is more
than what would be expected in geometric plating without additives or just suppressor.
The accelerated growth is due to the differential kinetics of SPS and PEI and the ‘CEAC
effect’, as discussed earlier. Similar shape evolution has also been observed by West et
al. in the acidic copper medium with PEG and SPS. While copper deposition rates in the
corners are enhanced, growth of copper on the sidewall and via-top remain strongly
suppressed due to inhibition by SPS (Figure 5-3). Copper deposition is observed to
initiate at the bottom and advances upward with negligible growth on sidewalls and rims
indicating a highly effective bottom-up fill.
91
Figure 5-2: Bottom-up fill from the additive containing alkaline bath under galvanostatic
deposition (25 mA/cm2) for 45s. Note the accelerated growth at the bottom corners
giving rise to a triangular shape. Aspect ratio of the feature is 1:1 and trench width is
approximately 1 µm. Rotation speed was 200 rpm. The electrolyte composition was 0.1
M copper sulfate, 0.5 M tartrate at pH=12.5. SPS and PEI concentrations were 15 and 40
ppm, respectively.
92
Figure 5-3: Bottom-up fill from the additive containing alkaline bath under identical
conditions as figure 5-2. Note the suppressed growth on the sidewalls and bottom-up
growth. Aspect ratio of the feature is 2.5:1 and trench width is approximately 500 nm.
Composition of the electrolyte was identical to those in figure 5-2.
Progressive-fill studies were conducted in electrolytes containing a mixture of 15
ppm SPS and 40 ppm PEI. Figure 5-4 shows the partial-fill results for increasingly
longer plating times: 20 s, 30 s, 45 s and 75 s. At short times (20 s), a localized
acceleration at the trench bottom corners is observed identical to figure 5-2. At moderate
time-scales (30-45 s), a flat bottom develops, which progressively advances upwards.
This flat bottom is again a known characteristic of the bottom-up fill process in the acidic
medium. At an average current density of 25 mA/cm2 used in this experiment, complete
trench fill was achieved in 75 s. During the plating process, trench sidewalls and the
wafer coupon ‘field’ regions remained completely suppressed.
93
Figure 5-4: Partial-fill profiles for Cu electrodeposition from an alkaline, tartratecomplexed Cu electrolyte containing 15 ppm SPS and 40 ppm PEI. The plating current
density (on the flat region of the wafer segment) is 25 mA/cm2 and the plating times are
20 s, 30 s, 45 s and 75 s. Note the accelerated growth at the trench bottom corners at
short times (20 s), progression of flat bottom-up fill at 30-45 s, and complete fill after 75
s of plating. Plating conditions are the same as in Figure 2.
To verify the role of antagonisitic SPS-PEI interactions in the bottom-up fill,
feature fill was also performed in electrolytes containing a single additive (SPS or PEI).
As seen in Figure 5-5 (a) and Figure 5-5 (b), electrolytes containing a single additive
(SPS or PEI) do not provide bottom-up fill, but give rise instead to conformal or
geometric plating, leading eventually to void trapping. Bottom-up fill was achieved only
94
from an electrolyte containing both SPS and PEI [Figure 5-5 (c)], indicating that SPS-PEI
interactions are essential for generating bottom-up fill.
Figure 5-5: Deposit profiles after Cu electrodeposition within 500 nm wide trenches
from an alkaline, tartrate-complexed Cu electrolyte at 25 mA/cm2 for 45 s. The
electrolyte contained (a) 15 ppm SPS; (b) 40 ppm PEI; and (c) 15 ppm SPS and 40 ppm
PEI. Conformal fill is observed in electrolytes containing a single additive, leading to
void formation. Bottom-up fill is observed in electrolyte containing both SPS and PEI,
indicating the critical role of antagonistic interactions between these additives. Plating
conditions: i=25 mA/cm2, pH=12.5, [Cu]=0.1 M, [Tartrate]=0.5 M, rotation speed=200
rpm.
95
5-3. Discussion
In conventional acidic electrolytes, a transient diffusion-adsorption mechanism is
known to play a critical role in the bottom-up fill. This diffusion-adsorption mechanism is
aided by the slow diffusion and strong adsorption of the suppressor (PEG) in comparison
to the rapid diffusion weak adsorption of the anti-suppressor (SPS). This mechanism
plays a prominent role in bottom-up fill of features having an aspect ratio greater than 10.
The reason for this lies in the fact that the time it takes for PEG to saturate the surface is
inversely related to the width or radius of the features. This can be described by an
equation as follows2:
 PEG 
 PEG L2
B
DPEGCPEG
W
(5-1)
Here τ is the PEG saturation time, Γ is the maximum surface coverage, D is the
diffusion coefficient, CB is bulk concentration, and L and W are length and width of the
trench, respectively. In trenches with very small width, this saturation time becomes
comparable to the feature fill time. Therefore, the adsorption and diffusivity of
respective additive species need to be taken into account in addition to the CEAC
mechanism to explain the bottom-up fill phenomenon. In the feature fill studies from the
alkaline medium described in section 5-2, aspect ratios of the features were less than 3.
Hence, diffusion characteristics of SPS and PEI were not instrumental and bottom-up fill
was mainly accomplished by the rapid adsorption of SPS on the sidewalls providing
strong inhibition and the CEAC mechanism accelerating copper growth at the bottom
corners.
In smaller features, e.g., sub-100 nm trenches, the diffusion and adsorption
characteristics of SPS and PEI play a prominent role in bottom-up fill. The adsorption
96
constant of SPS on a Cu substrate in the alkaline medium may be assumed analogous to
that of PEG in the acidic electrolyte. The reasoning for this is that PEG and SPS both
exhibit similar time constants for adsorption in their respective injection studies described
in chapter 4. Unlike the adsorption characteristics, the diffusive properties of PEG and
SPS are quite different. In the acidic medium, PEG, the suppressor, is the diffusion
limited species having a diffusion constant in the range of 10-7-10-6 cm2/s, depending on
the molecular weight7. SPS, the accelerator, is the fast diffusing species with diffusivity
between 10-6 to 10-5 cm2/s. However, in the alkaline medium discussed herein, the
suppressor (SPS) diffusion is faster than the anti-suppressor (PEI) since the molecular
weight and consequently the size of SPS is much smaller than that of PEG. The linear
branched PEI has a molecular weight that is twice that of the SPS. Therefore, in the
alkaline tartrate medium, PEI (anti-suppressor) is actually the diffusion limited species.
While this may be deleterious to the bottom-up fill particularly in narrow sub-100 nm
features, we do not observe its ill effects in the feature-fill studies reported above owing
to the large feature widths of 500 nm. In future studies involving more aggressive
structures, we believe that optimizing the suppressor diffusion rates in our alkaline
system will be important. This may be accomplished in two ways. The suppressor bulk
concentration in the electrolyte may be lowered to counter the adverse effects of
decreasing feature size, as shown in equation 5-1. Relative diffusion rates can also be
optimized by designing higher molecular weight ‘SPS-like’ suppressors with lower
diffusion coefficients or by incorporating lower molecular weight anti-suppressor with
higher diffusion coefficients so that DSPS << DPEI. The diffusion of SPS can therefore be
97
controlled by tuning one or both of these parameters to obtain bottom-up fill in higher
aspect ratio features.
References:
1. R. Akolkar and U. Landau, J. Electrochem. Soc., 156, D351 (2009).
2. J. Adolf and U. Landau, J. Electrochem. Soc., 158, D469 (2011).
3. R. Akolkar and U. Landau, J. Electrochem. Soc., 151, C702 (2004).
4. A.C. West, S. Mayer and J. Reid, Electrochem. Solid-State Lett., 4, C50 (2001).
5. 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).
6. R. Akolkar and V. Dubin, Electrochem. Solid-State Lett., 10, D55 (2007).
7. J. Adolf, “Modeling the Role of Plating Additives in the Metallization of
Semiconductor Interconnects: From Dual Damascene to TSV”, Dissertation
thesis (2011).
98
Chapter 6
Pulse Plating of Copper Germanide
Moore’s law reflects the miniaturization of interconnect dimensions in logic and
memory devices. In current generation devices, these interconnects are fabricated using
electrodeposited copper (Cu). However, as interconnect dimensions become extremely
smaller, the electrical resistivity of Cu increases markedly due to grain boundary and
interface scattering1. This becomes particularly noticed as interconnect dimensions
approach the electron mean free path in Cu (λCu~39 nm). The increase in Cu resistivity
deteriorates device performance parameters such as the RC delay.
Future generation interconnects require different materials with reduced electrical
resistivity. A promising candidate material proposed by researchers at IBM, is copper
germanide (ε-Cu3Ge), with an electrical resistivity of 10 µΩ-cm at room temperature2,3.
Additionally, Cu3Ge is remarkably oxidation-resistant making it an excellent candidate
for damascene integration4,5. To date, the established method for fabricating Cu3Ge thin
films has been the sequential e-beam evaporation of Cu and Ge followed by annealing1-6.
While this method is useful for studying Cu3Ge material properties, it cannot be
integrated into the established dual-damascene process flow used for nano-scale
interconnect fabrication. Since electrodeposition is the preferred method for fabricating
conventional Cu interconnects, identifying an electrodeposition route for fabricating
Cu3Ge is highly desirable. In addition to its low cost of ownership, electrodeposition
offers numerous advantages over dry processes including high deposition rate, operation
at ambient temperature and pressure, and the ability to generate bottom-up feaure fill7.
99
Described here is the electrodeposition of Cu3Ge from an alkaline tartratecomplexed electrolyte. The choice of this electrolyte is based on our prior work
demonstrating high nucleation density in direct plating of Cu onto ruthenium (Ru)11 and
‘bottom-up’ fill12 from such an alkaline complexed electrolyte. Additionally, high
current density pulses are shown to enable co-deposition of Cu and Ge in the
stoichiometric ratio (Cu:Ge=3:1) while minimizing roughness evolution and providing
smooth and compact electrodeposits. Furthermore, the electrical resistivity, crystal
structure, grain size and impurity content of the electrodeposited Cu3Ge thin films are
reported.
100
6-1. Challenges for Electroplating Cu3Ge
The reduction reaction and standard reduction potential for copper and
germanium are:
Cu 2+ +2e-  Cu 0
E 0 =+0.34 V
Ge4+ +4e-  Ge0
E 0 =+0.12 V
The deposition potential of germanium is more cathodic than copper and there exists
a potential difference of 0.22 V that needs to be bridged to initiate co-deposition of
copper and germanium. Additionally, due to extremely fast hydrogen evolution kinetics,
more than a monolayer thick pure Ge cannot be deposited from an aqueous electrolyte. It
can only be co-deposited as an alloy8. Only two articles, dating back to the 1950s, that
discuss co-deposition of Cu and Ge from an aqueous electrolyte were found: (i) Fink and
Dokras described electrodeposition of Cu3Ge from a cyanide electrolyte8; and (ii) Gupta
et al. reported co-deposition of Cu and Ge from complexed electrolytes9. The cyanide
solution employed by Fink and Dokras8 is highly regulated in the United States and thus
undesirable for large-scale fabrication. The study by Gupta et al.9 employed dilute
electrolytes and large DC currents (~30 mA/cm2), leading to rough and porous
electrodeposits10 that are unacceptable in interconnect fabrication. Neither of the two
aforementioned studies reported material properties of electroplated Cu3Ge relevant to
semiconductor applications, such as electrical resistivity, grain size and crystal structure.
The limitation of DC plating is further illustrated in the polarization curve in
Figure 6-1, obtained during co-deposition of Cu and Ge from the alkaline tartrate
complexed electrolyte. It is observed that appreciable deposition rates (>10 mA/cm2) are
achieved only at electrode potentials more cathodic than -1 V (vs. SHE). The open
101
circuit potentials for the Cu+2/Cu and the Ge+4/Ge couples measured in the alkaline
tartrate complexed electrolyte were -0.146 V and -0.18 V (vs. SHE), respectively. Thus,
the net overpotential measured at a current density of 10 mA/cm2 was about 0.95 V,
indicating significant electrode polarization, which is typical of complexed electrolytes13.
The polarization curve approaches a plateau in the current density range of 20-25
mA/cm2. This plateau is attributed to the limiting current density of Cu
electrodeposition, which is estimated in our system to be about 21 mA/cm2 following the
Levich equation14. Film composition, i.e., the atomic ratio of Cu:Ge, as a function of the
electrode potential is also shown in Figure 1. It is observed that the Ge content in the
film increases as the deposition potential becomes more cathodic. While the target ratio
of Cu:Ge for Cu3Ge formation is 3, none of the co-deposited films obtained below the
limiting current density of Cu meet this target. At a current density of 20 mA/cm2, i.e.,
just below the limiting current density of Cu, the Cu:Ge ratio is about 7 indicating
excessively Cu-rich deposits. By applying large DC currents, e.g., 44 mA/cm2, well
above the limiting current of Cu, films with Cu:Ge as low as 4.5 (still above the target)
were obtained; however, these films were visibly rough and not suitable for the intended
application of interconnect fabrication. Clearly, in order to achieve the target
composition, significantly larger current densities are required, yet this range is barred
due to the non-acceptable deposit texture.
102
Figure 6-1: Polarization behavior (blue, left ordinate) and deposit composition (red, right
ordinate) in Cu and Ge co-deposition from an alkaline tartrate complexed electrolyte
under DC. Electrolyte contains 25 mM Cu, 96 mM Ge, 300 mM tartrate and 325 mM
NaOH. Films obtained using DC plating do not meet the target Cu:Ge composition of
3:1 required for Cu3Ge formation.
103
6-2. Cu3Ge Electrodeposition by Pulsing
Given the aforementioned limitations of galvanostatic (DC) plating in achieving
the target composition without compromising film quality, pulse plating of Cu-Ge was
employed. Current pulsing has been shown to enable electrodeposition at instantaneous
current densities well above the limiting current density while suppressing the evolution
of roughness15. This was verified this for the studied system by comparing the surface
roughness of electroplated Cu films obtained using DC and pulse waveforms at the same
average current density. For comparison, a DC current density of 15 mA/cm2 and a
square pulse waveform with instantaneous current density (ion) = 205 mA/cm2, ‘on’ time
(ton) = 2.2 ms and ‘off’ time (toff) = 28 ms (corresponding to an average current density of
15 mA/cm2) were chosen. Pulsing significantly reduced the surface roughness of
electroplated Cu (Figure 6-2). The root mean squared (rms) roughness of the pulse plated
Cu was 74 nm, substantially lower than that of the DC plated Cu (334 nm).
104
Figure 6-2: Surface roughness profiles of electroplated Cu using DC (black dashed line)
and pulse (blue solid line) plating. The average current density was 15 mA/cm2 for both
cases. The rms roughness of the DC plated film was 334nm while that of the pulse plated
film was 73 nm. Both experiments corresponded to a total charge of 2.85 coulombs.
In order to achieve the target stoichiometric ratio of Cu:Ge (~3:1) in the film,
several pulse waveforms were attempted. It was observed that the magnitude of the
instantaneous current density (ion) during pulsing modulates the Ge content in the plated
Cu-Ge films. Specifically, as ion is increased, the Ge content in the film also increases.
Ge content in the film is less that 15 at. % when ion<100 mA/cm2. For ion>200 mA/cm2,
Ge content approaches the target value of 25 at. % and a bright, silvery film of Cu3Ge is
observed8. Schematic of a typical pulse waveform used to deposit Cu-Ge in a 3:1
stoichiometric ratio is shown in Figure 2. Relative plating rate of Ge increases at higher
105
current densities. This is due to the fact that during high current pulses Cu deposition
occurs close to its mass transport limit while Ge deposition is not hindered by transport
effects, owing to its high concentration in the bulk. Ge deposition is controlled by its
kinetics, i.e., the rate at which Ge ions are reduced at the electrode. This enables
deposition of higher Ge alloys without compromising roughness.
Figure 6-3: Ge content in electrodeposited Cu-Ge films as a function of the instantaneous
current density (ion). Target Ge content of 25 at. % is achieved when ion exceeds 200
mA/cm2. Pulse time depends on the instantaneous current density according to Eqn. 3-8.
Average current density was kept at 15 mA/cm2 for all pulse waveforms.
106
Cu
2000
Counts
1500
Cu- 77 %
Ge- 23 %
1000
Ge
500
O
0
1
2
3
4
KeV
Figure 6-4: EDS spectra of Cu-Ge plated at a pulse current density of 205 mA/cm2 with
ton=2.2 ms. EDS indicates near stoichiometric ratio of 3.3:1 of Cu:Ge.
107
6-3. Materials Characterization
I.
Cross-sectional SEM
The Cu-Ge films electroplated using the pulse waveform (ion=205 mA/cm2,
ton=2.2 ms and toff=28 ms) were characterized after annealing them under N2 atmosphere
at 300 oC for 30 minutes. The films were first capped with sputtered platinum and crosssections were prepared using FIB. SEM image of a film cross-section is shown in Figure
6-5(a) along with surface morphology of an uncapped film (Figure 6-5(b)). The
thickness of the Cu-Ge film was about 500 nm. Thus, the faradaic efficiency of Cu-Ge
electrodeposition was estimated to be approximately 25 %. The low faradaic efficiency
is attributed to hydrogen co-evolution. This is expected since the electrode potential of 2.5 V (vs. SHE) reached during the ‘on’ cycle is significantly cathodic compared to the
onset potential for hydrogen evolution of -0.77 V (vs. SHE) at pH=13. The growth rate
of the film using the aforementioned pulse waveform was calculated to be about 0.4
Å/pulse. At higher instantaneous current densities, the growth rate increased non-linearly
due to increased hydrogen co-evolution.
108
Figure 6-5: (a) Cross-section of the electroplated Cu-Ge film on a Ru substrate. Pulse
plating conditions: ion= 205 mA/cm2, ton=2.2 ms and toff=28 ms. Total plating time is 7
minutes. (b) Surface morphology of the plated film.
109
II.
XPS
The Cu and Ge concentration throughout the plated film was uniform, as
measured by XPS depth profiling (Figure 6-6). XPS shows near stoichiometric Cu:Ge
ratio of 3.3:1 throughout the film, consistent with earlier EDS findings. XPS also
confirmed the inclusion of oxygen (~3 at. %) and carbon (~1 at. %) impurities in the film.
Figure 6-6: Composition depth profile of a 150 nm Cu3Ge film deposited using the same
waveform. A uniform Cu and Ge composition is observed.
110
III.
XRD and Resistivity
The plated Cu-Ge films were characterized by XRD. As shown in Figure 6-7,
XRD confirms the presence of crystalline monoclinic ε-Cu3Ge in the as-deposited film16.
The original crystal structure was conserved post annealing (Figure 6-7). The film
exhibited strong <111> texture similar to that observed in Cu3Ge films deposited by dry
techniques4,18. The effect of annealing on the Cu3Ge grain size could also be assessed by
XRD, as reported previously18. In the films studied here, the grain size, as estimated
from the full width at half maximum (FWHM) of the Cu3Ge <111> peak using the
Scherrer equation19, increased from 16 nm (pre-anneal) to 24 nm (post-anneal). Grain
growth due to annealing was accompanied by a reduction in the electrical resistivity of
Cu3Ge from 95 µΩ-cm (pre-anneal) to 45 µΩ-cm (post-anneal). Annealing for an
additional 30 minutes did not reduce the resistivity and the grain size any further. The
resistivity of the electrodeposited Cu3Ge films was higher than that reported in the
literature2-4,6 for e-beam evaporated films (~10 µΩ-cm). It is likely that the higher
resistivity of the electrodeposited Cu3Ge is due to relatively small grain size20 (~24 nm)
as compared to that observed in e-beam evaporated films (100-500 nm)3,6. The smaller
grain size of the electrodeposited Cu3Ge films is likely due to impurities incorporated in
the film during electrodeposition, which inhibit grain growth21. As shown in Figure 3,
the films contain impurities in the form of carbon (~1 at. %) and oxygen (~3 at. %).
Furthermore, the inhibition of grain growth in Cu3Ge due to impurities is indirectly
evidenced by the fact that resistivity does not change as the annealing time is increased
from 30 minutes to 1 hour indicating that grain growth terminates at ~24 nm. Additional
111
work is needed to understand and eliminate the source of impurity incorporation during
Cu3Ge electrodeposition from the studied electrolyte.
Figure 6-7: XRD spectra of as-deposited (a) and annealed (b) films showing ε-Cu3Ge
formation. All the diffraction peaks associated with the monoclinic ε-Cu3Ge16 are shown
in (c). Pulse plating conditions: ion=205 mA/cm2, ton=2.2 ms and toff=28 ms. Total
plating time =14 minutes. Calculated film thickness was ~0.9 µm.
112
6-4. Conclusions
In summary, ε-Cu3Ge thin films with a monoclinic crystal structure were
fabricated using pulse electrodeposition from an alkaline tartrate complexed electrolyte.
This unique electrodeposition process, unlike previous dry techniques used to deposit
Cu3Ge, is easy to integrate in conventional dual damascene metallization. At the present
time, the films electrodeposited here show higher than expected bulk electrical resistivity
(45 µΩ-cm), which we believe is due to a combination of grain boundary and impurity
induced scattering.
113
References:
1. T. S. Kuan, C. K. Inoki, G. S. Oehrlein, K. Rose, Y. –P. Zhao, G. –C. Wang, S.
M. Rossnagel and C. Cabral , “Fabrication and Performance Limits of Sub-0.1
µm Cu Interconnects”, MRS Proceedings, 612 (2000)
2. L. Krusin-Elbaum and M. O. Aboelfotoh, Appl. Phys. Lett. 58, 1341 (1991)
3. M. O. Aboelfotoh, and H. M. Tawancy, J. Appl. Phys. 75, 2441 (1994)
4.
Y-. Chao, Y. Xu, R. Scholz, and J. C. S. Woo, IEEE Electron Device Lett. 27, 549
(2006)
5.
H. K. Liou, J. S. Huang and K.N. Tu, J. Appl. Phys. 77, 5443 (1995)
6. M. O. Aboelfotoh, K. N. Tu, F. Nava, and M. Michelini, J. Appl. Phys. 75, 1616
(1994)
7. P.C. Andricacos, C. Uzoh, J.O. Dukovic, J. Horkans and H. Deligianni, IBM J. of
Res. and Dev. 42, 567 (1998)
8. C. G. Fink and V. M. Dokras, Trans. Electrochem. Soc. 95, 80 (1949)
9. P. R. Subbaraman and J. Gupta, J. Sci. Ind. Research (India) 15B, 306 (1956)
10. A. Brenner, Electrodeposition of Alloys: Principles and Practice-Vol.2, Academic
Press, New York (1963), p.137.
11. A. Joi and U. Landau, An Alkaline Copper Plating Process Providing High
Nucleation Density on Ru and Bottom-up Fill, Abstract #1944, 220th ECS
Meeting, 9-14 October 2011, Boston, MA.
12. A. Joi, R. Akolkar, U. Landau, J. Electrochem. Soc. 160, D3001 (2013).
114
13. R. Akolkar, T. Indukuri, J. Clarke, T. Ponnuswamy, J. Reid, A. J. McKerrow, S.
Varadarajan, “Direct seed electroplating of copper on ruthenium liners”,
Interconnect Technology Conference and 2011 Materials for Advanced
Metallization (IITC/MAM), 2011 IEEE International , pp.1-3, 8-12 May 2011.
14. V. G. Levich, Physiochemical Hydrodynamics, Prentice-Hall, New Jersey (1962),
p.60.
15. K. I. Popov, M.D. Maksimovic, B. M. Ocokoljic and B. J. Lazarevic, Surf.
Technol. 11, 99 (1980)
16. ICDD (2004), ‘PDF 06-0693, PDF 36-1134 (Database), International Centre for
Diffraction Data, Newtown Square, PA, USA.
17. F. M. d’Heurle and J. Gupta, Appl. Surf. Sci. 73, 214 (1993)
18. K. A. Darling, R. K. Guduru, C. L. Reynolds Jr., V. M. Bhosle, R. N. Chan, R. O.
Scattergood, C. C. Koch, J. Narayan and M. O. Aboelfotoh, Intermetallics 16, 378
(2008)
19. B. D. Cullity, Elements of X-ray Diffraction, Addison-Wesley Publishing
Company, Massachusetts (1967), p. 262.
20. W. Wu, S. H. Brongersma, M. Van Hove and K. Maex, Appl. Phys. Lett. 84,
2838 (2004)
21. 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).
115
Chapter 7
Characterization of the Tartrate Plating Bath with Additives
Practical aspects of the copper plating process, e.g. current efficiency and stability
of the bath are important parameters that need to be characterized for commercial
implementation. Current efficiency and bath stability of the copper tartrate complex
electrolyte have been discussed by Liao et al1. However, incorporation of additives such
as SPS and PEI into the tartrate complexed electrolyte might significantly alter its
properties. These additives are essential for achieving copper metallization from the
tartrate complexed electrolyte. Therefore, the effect of SPS and PEI on current efficiency
and bath stability need to be characterized.
7-1. Current Efficiency
Faradaic efficiency of the copper plating process was calculated using the
following equation:
 f (%) 
Wexp
WTheo
100
(7-1)
Here, Wexp is the experimentally observed weight and WTheo is the theoretical
weight of the deposit calculated using Faraday’s law. Faraday’s law states that the
weight of the plated film is directly proportional to the current and plating time which can
be expressed as follows:
WTheo 
ItM
nF
(7-2)
116
Here, I is the current, t is the plating time, M is molecular weight of copper, F is
the Faraday’s constant and n=2 for reduction of copper. Plating experiments were
conducted at a current of 6.33 mA (20 mA/cm2) and room temperature for 20 minutes.
Weights of the deposit were calculated by measuring the weight of the electrode before
and after experiments. Current efficiencies for copper plating from the tartrate
complexed electrolyte with and without additives are shown in figure 7-1. It was
observed that the additives lower the current efficiency by almost 40-50 %. Current
efficiency without additives was found to be ~95 %, as reported by Liao et al. Copper
plated under potentiostatic conditions yielded similar efficiency values as well.
Figure 7-1: Faradaic efficiency for copper deposition from the alkaline tartrate
electrolyte, alkaline tartrate electrolyte containing 15 ppm SPS and 15 ppm SPS+40 ppm
PEI. Copper was plated under galvanostatic condition (20 mA/cm2 for 30 mins, ω=200
rpm). Additives lower the current efficiency by 40-50 %.
117
Initially, it was thought that the 30-40 % inefficiency might be due to hydrogen
evolution. The hydrogen evolution potential at pH 12.5 was calculated to be -0.74 V vs.
SHE. The potential at steady state for copper plating at 20 mA/cm2 from the 15 ppm SPS
and 15 ppm SPS + 40 ppm PEI solutions were -1.25 V vs. SHE and -1.05 V vs. SHE,
respectively. If hydrogen evolution was the primary contributor to the inefficiency,
efficiency would have been much lower in the former solution than the latter (-200 mV
difference in steady state potential). However, efficiencies in both solutions, 15 ppm SPS
and 15 ppm SPS + 40 ppm PEI, were observed to be similar. To verify the above
observation, an experiment was designed to quantitatively measure the hydrogen
evolution during plating in the presence of additives. A schematic of the experimental
setup is shown in figure 7-2. An inverted test tube was used for collecting the hydrogen
that was generated during plating. The evolved hydrogen was quantified by measuring
the difference in heights of the solution level in the inverted tube before and after plating.
Working electrode was a piece of copper tape that was wrapped around the negative
copper wire electrode submerged in solution in the inverted beaker. The solution was
mechanically stirred using a stir bar. Surface area of the working electrode could not be
measured accurately. Hence, H2 evolution experiments were performed in a
potentiostatic mode (-1.25 V vs. SHE). First, a calibration curve was constructed by
evolving pure H2 on the copper electrode from an aqueous 12.5 pH solution. Height
changes (Δl) were recorded for a total amount of charges (mC) passed and assuming no
other side reaction occurred, a calibration plot for Δl vs. mC was obtained. The
calibration plot for H2 evolution from the 12.5 pH solution is shown in figure 7-3. The
118
height change due to H2 evolution was found to be linearly proportional to the charge
passed with a proportionality constant of 3728.5 mC/cm.
Figure 7-2: A schematic of the experimental setup used to quantitatively determine H2
evolution.
119
Figure 7-3: Calibration curve for H2 evolution using the setup shown in figure 7-2. H2
was evolved at a constant potential of -1.25 V vs. SHE from an aqueous electrolyte with
pH=12.5. Slope of the linear plot was 3728.5 mC/cm, assuming 100 % efficiency for the
H2 evolution reaction.
Once the calibration plot was obtained, copper plating was performed at -1.25 V
vs. SHE from the copper tartrate solution containing 15 ppm SPS and 40 ppm PEI. An
identical setup as the one described earlier was used (figure 7-2). The height change was
recorded and fraction of the total charge that contributed to hydrogen evolution was
calculated using the calibration curve. The data is shown in table 7-1. It was found that
H2 evolution contribution is only 5-6 % of the total current.
120
Table 7-1: Fraction of the total current contributed by H2 evolution during copper plating
with additives at -1.25 V vs. SHE. The calibration curve was used to convert height
change to total H2 current.
Height Change (cm) H2 charge (mC) Total Charge (mC) Fraction (%)
0.07
261.00
6200.30
4.21
0.09
335.57
7802.79
4.30
0.28
1043.98
12160.21
8.59
5.7
Average =
The 35-45 % loss in current efficiency that is unaccounted for must be due to side
reaction(s) induced by the presence of additives. Since the additives are present in such a
small amount (parts per million), the 35-45 % loss in current cannot be due to reduction
of the additives themselves. From figure 7-1, it is evident that the decrease in efficiency
is mainly due to the presence of SPS in solution. This was further verified by performing
plating experiments with different concentration of SPS in solution. It was observed that
current efficiency decreased markedly as concentration of SPS in solution increased
(figure 7-4). Current efficiency was as low as 15 % when SPS concentration in solution
exceeded 70 ppm.
121
Figure 7-4: Current efficiency for copper plating from the tartrate solution with different
concentrations of SPS in solution. Efficiency decreased as SPS concentration in solution
increased. Experimental conditions were identical to that used in figure 7-1.
Based on the above observations, it can be proposed that the inefficiency is
possibly due to SPS catalyzed side reaction(s). Time-of-Flight Secondary Ion Mass
Spectrometry (figure not shown) revealed the plated copper to be 97 % pure (3 at. %
impurity: 2 at. % O, 0.5 at. % C and 0.5 at. % S). Therefore, any side deposition reaction,
in addition to copper, cannot be responsible for the observed ~50 % inefficiency. There
are two possible reactions that might be occurring which could explain the observed
efficiency. One possible reaction is the decomposition of tartrate in solution to a lower
oxidation state. Another possible reaction might be the reduction of Cu(II) to Cu(I) that
is either stabilized in solution via complexation with tartrate or decomposed as Cu2O.
Significant amount of red precipitates are observed in solution after ~12 hours of plating
122
in the presence of SPS. This precipitate was identified as Cu2O by Liao et al. The
observed inefficiency, therefore, could be attributed to possible SPS catalyzed side
reactions involving reduction of Cu2+ to Cu+ in solution. A detailed investigation into the
mechanism of copper reduction in the presence of additives fell outside the scope of this
work and was not researched further. However, for large scale commercial
implementation, this phenomenon deserves further studies and characterizations.
123
7-2. Bath Stability
Stability of the copper-tartrate bath was investigated by Liao et al. The
electrolyte was found to be less stable at a lower pH values (pH <13). This is due to a
different copper-tartrate complex formation at higher pH (pH > 13) that is more stable
than its low pH counterpart. Similar behavior was observed in the tartrate plating bath
containing additives. The bath at lower pH containing SPS and PEI decomposed within
24 h whereas bath at higher pH remained stable longer than 200 h. Electroplating at 20
mA/cm2 from a low pH bath containing 15 ppm SPS resulted in decomposition of the
solution after ~5 h of plating whereas high pH bath remained soluble even after 12 h of
continuous plating.
References:
1. C-H. Liao, PhD Dissertation, Case Western Reserve University, Cleveland, OH
(2012).
124
Chapter 8
Mass Transfer and Kinetics Consideration in Pulse Plating of Cu3Ge
Metal and alloy deposition by pulse plating may offer some advantages over DC
plating. In pulse plating, there are additional ‘control knobs’, e.g. frequency, duty cycle,
periodic reverse or anodic current and waveform shape, that can be adjusted to modify
film morphology and composition. In chapter 6, pulse plating was shown to enable
electrodeposition of smooth Cu3Ge film in a 3:1 ratio that had not been achievable by DC
plating. Several approaches have been used over the years to analyze the effects of mass
transfer and electrode kinetics in pulse plating of metals. Pioneering work in this field
was done by Rosebrugh and Miller1. They analytically solved for the concentration
profile of the reactant at the electrode for several different current waveforms by using an
infinite Fourier series method. Cheh used the solution provided by Rosebrugh and Miller
and demonstrated that the average plating current during pulsing can never exceed the
limiting direct current density under the same hydrodynamic conditions2. Ibl et al.
proposed a simplified duplex boundary layer model consisting of a stationary outer
diffusion layer and a dynamic inner boundary layer3. The concentration profile within
the inner diffusion layer fluctuates with the frequency of the pulse current density. A
linear concentration profile was assumed in the stationary boundary layer and this model
was shown be reasonably accurate for small duty cycles4. Controlled potential pulse
plating model has also been developed5,6 and it was demonstrated that one can control the
thickness of the diffusion layer by modifying the pulse frequency of the potential pulses,
whereas for controlled current pulses, diffusion layer thickness grows with time and it is
125
eventually limited only by convection. The analytical approaches described above,
however, apply only to for single metal deposition. Therefore, the kinetics of the
electrode reaction is neglected in these analyses. For multiple electrode reactions, such
as in the case of alloy plating, kinetic rates at the electrode of respective reactions must
be taken into account since relative reaction rates at the electrode must be adjusted to
accommodate the total applied current density. Therefore, the concentration and reaction
rate of each species at the electrode must be solved simultaneously to obtain an average
composition of the alloy. A general outline of the process is given below. Several
assumptions are made in this analysis(1) There is no interaction between the ionic species in solution
(2) No homogeneous chemical reaction occurs in solution
(3) Double-layer charging effects are neglected
(4) The alloy does not dissolve during the off-period
(5) The effect of migration is neglected in the presence of excess supporting
electrolyte
(6) A flat electrode is assumed and geometry effects on the current distribution
are not considered
Applying these assumptions to the convective-diffusion equation, the governing equation
for one-dimensional diffusion can be written as:
c j
t
 Dj
 2c j
x 2
(8-1)
With the following initial and boundary conditions126
t  0, C j  CBulk , j
(8-2)
x   , C j  CBulk , j
(8-3)
x  0,
x  0,
C j
x
x 0 
C j
x
x 0
ij
nFD j
for t  ton
 0 for t  toff
(8-4)
(8-5)
Here, Cj is the concentration of the ionic species j, Dj is its diffusion coefficient and x is
the distance from the electrode into the electrolyte, where x=0 is the electrode surface. δ
is the diffusion layer thickness. This is illustrated schematically in Figure 8-1.
Figure 8-1: Governing equation and boundary conditions during the pulse ‘on’ and ‘off’
cycle. The electrode is located at x=0.
127
The concentration of each species is coupled to its reaction rate via mass-transfer
controlled Butler-Volmer kinetics:
 Cs , j
i j  i0, j 
 Cbulk , j

  jnjF 
 j 
RT

 
 e

 Cs , j
 i0, j 
 Cbulk , j

  b j  E  E0 , j   
 e

(8-6)
Here, i0,j and bj are exchange current density and inverse Tafel slope, respectively, for the
deposition of species j from solution. Cs,j is the concentration at the electrode and Cbulk,j
is the bulk concentration. ηΩ is the ohmic overpotential and E0,j is the equilibrium
reduction potential for species j. The reaction rate of species j is given by its partial
current, ij.
For galvanostatic pulses, the sum of the partial current densities must be equal to
the total applied current density:
i
j
 iapplied
(8-7)
Equations (8-1), (8-6) and (8-7) must be solved simultaneously for each species in
the system to obtain Cj, ij and E (electrode potential). This complicated system of
equations has been solved numerically by various authors. Yin and White7 employed a
finite difference method to model the pulse plating of a Ni-Cr alloy. This model
considered the effect of migration and they found that after 2-3 cycles, the solution
approaches a periodic steady-state. More recently, Liu and West8 modeled galvanostatic
pulse plating of Au taking into account H2 evolution. Ruffoni and Landolt9 employed a
superposition method to simulate the pulse plating of a Au-Cu-Cd alloy under pure
128
diffusion (stagnant solution) conditions. Verbrugge and Tobias13 utilized a series
solution to the current-step problem at long and short times and combined it with the
superposition principle to model the mass transfer during pulse plating of a
multicomponent alloys. The results agreed very well to those obtained by numerically
solving Equations (8-1)-(8-5). A DC plating model predicting current distribution in a
two-dimensional cell in the presence of multiple electrode reactions was also provided by
Menon and Landau14.
Solving this set of equations with the stipulation of the stated boundary/initial
conditions for a system involving more than two species over tens of thousands of pulses
is very cumbersome and requires significant amount of computation. Since we are only
interested in the bulk or average quantities (after > 10 minutes of plating) such as
composition of the film and faradaic efficiency of the process, the unsteady-state effects
during the initial stages of plating can be neglected and the equations may be solved for
one representative, steady state pulse that yields the average quantities for the film
composition and the plating process. Therefore, in the present work, we employ a steadystate approach to solve the above set of equations. The simplified and approximate
steady-state model for pulse plating is outlined in section 8-1.
129
8-1. Development of the Steady-State Galvanostatic Pulse Plating Model
The mass transfer phenomena in galvanostatic pulse plating can be categorized
according to duty cycle of the pulse scheme. In the case of infinitely small duty cycle (toff
is significantly larger than ton), the analysis of mass transfer effects in pulsing becomes
simpler. During the ‘on’ cycle, the concentration of the respective species at the
electrode is depleted as compared to its bulk value. This depletion is dictated by the
magnitude of the applied current density and reaction rate at the electrode of the ions
being reduced. Since toff is significantly larger than ton in this scenario, it is assumed that
diffusion during the rest period will able to relax the surface concentration back to its
bulk value. Thus, there is no concentration gradient within the diffusion boundary layer
and the next ‘on’ pulse initiates from its original bulk condition (see Figure 8-2). In this
situation, every pulse is identical to the ones that follows and precedes it if effects of ion
consumption during plating and deposited film induced changes in the kinetic parameters
are neglected. The system is always subject to ‘steady-state’ and intensive properties, i.e.
film composition and process efficiency can be obtained by solving equations (8-1)-(8-4)
for a single ‘on’ pulse. The analytical solution for this is given by Rosebrugh and Miller:
C j  Cbulk , j

i j 
8

1

n j FD j   2


p  odd
1
e
p2
   2 D t  
j
  p2

2 
 
4

 





(8-8)
It has been shown12 that for very small values of ton and large values of δ (Dton/δ2 < 0.1),
1/2
the transient term in the parenthesis can be replaced with
2  Dton 
 1/2   2 
. The above
analytical expression must be coupled with equations (8-6) and (8-7) and solved
130
simultaneously. However, the computation is much simpler in this case as the system of
equations reduces to a set of non-linear algebraic equations that can be solved for each
species in the system by iteration.
In the second case when toff is not significantly larger than ton (0.01< duty cycle <
1), diffusion during the rest period cannot relax the surface concentrations back to its
bulk values. Therefore, the following ‘on’ pulse initiates from values that are slightly
lower than the original bulk values. A build-up of concentration gradient occurs near the
electrode as diffusion layer continues to grow with each pulse cycle. This build-up lasts
until the convection in the system halts further growth of the diffusion boundary layer.
At this point, system reaches a ‘quasi’ steady-state where the concentration profiles
oscillate between two fixed points. This is illustrated schematically in Figure 8-2. The
time that it takes to reach the ‘quasi’ steady-state can be approximated by diffusion layer
build-up time (tsteady-state~ δ2/Dj). For a typical diffusion coefficient of 10-6 cm2/s and
diffusion layer thickness of 30 µm, the time to reach ‘quasi’ steady-state is calculated to
be ~9 s. This is negligible compared to the total plating time which is larger than 10
minutes. Thus, the bulk intensive properties of the film can be obtained by modeling the
mass-transfer at this ‘quasi’ steady-state. The solution for the concentration profile at the
electrode can be approximated by shifting the solution of Rosebrugh and Miller (Eqn. 88) by a steady state depletion term (Figure 8-2).
131
Figure 8-2: Schematic of the concentration profile at the electrode during galvanostatic
pulsing when toff is comparable to ton. A build-up of concentration gradient occurs within
the diffusion layer. The duration of this unsteady state depends on the prevailing
hydrodynamic conditions of the system. For typical values, the duration of this unsteady
state is calculated to be about 10 s.
The steady state depletion term can be calculated by considering a duplex
diffusion layer model at ‘quasi’ steady state as described by Ibl3. In this model, one
envisions existence of two boundary layers near the electrode (Figure 8-3). The pulse
boundary layer (δp), which exists during the pulse ‘on’ time and the diffusion boundary
layer (δ), which is characterized by the hydrodynamic of the system. The length of the
pulse boundary layer is given approximately by δ ~ (Djton)1/2. This is extremely small
compared to δ. If one assumes a linear concentration profile within the stationary region
(δ-δp), the gradient of which is related to the ‘average’ current density of the pulse
waveform, it is possible to calculate the steady state depletion term by performing a
simple mass balance. The ‘average’ current density is defined as:
iave  i pulse  
(8-9)
132
Where γ is the duty cycle.
Cs,j
0
Figure 8-3: A schematic of the duplex boundary layer model. The concentration
fluctuation occurs within δp while gradient in the intermediate region (δ-δp) remains
stationary. The gradient within the stationary region depends on the hydrodynamics of
the system.
The concentration at the electrode then can be written by taking into account the steadystate depletion and modifying Eqn. 8-8.

1  8
C j  Cbulk , j 
n j FD j   2

i j

p  odd
1
e
p2
   2 D j t  
  p2

 
4 2  


1/2
  i j     4 D j t 1    

 nFD    

 
j 

(8-10)
133
Eq. (8-6), (8-7) and (8-10) must be solved simultaneously for each species in the
system. For Cu3Ge electrodeposition, this results in a system of six algebraic equations
that must be solved to calculate Cs,Cu, Cs,Ge, iCu, iGe, iH and E. Hydrogen evolution is
assumed to be not limited by its mass-transfer8. The set of equations is solved by using
the ‘fsolve’ command in MATLAB which utilizes an iterative trust region dogleg
algorithm to calculate the variables, for a given pulse waveform. MATLAB codes for
solving this set of equations are given in section 8-4.
134
8-2. Determination of Transport and Kinetics Parameters
In order to accurately model the effect of mass transfer and kinetics on the pulse
electrodeposition of Cu3Ge, the transport parameters, i.e., diffusion coefficients (D) and
kinetic parameters, i.e., exchange current densities (i0) and Tafel slopes need to be
determined. However, these parameters are not available in literature for a coppergermanium co-deposition process from a tartrate electrolyte. Therefore, these parameters
were determined experimentally.
8-2.1. Determination of the kinetic parameters
Polarization curves for each species were obtained via weight measurements.
Here, Cu-Ge was co-deposited on a copper rotating disk electrode at various current
densities. A rotation speed of 1000 rpm was used. The composition of the deposit was
analyzed by EDS and partial current densities of each species were obtained using
Faraday’s law. Partial current densities (ij) and overpotentials (η) were then plotted to
construct the polarization curves for each species (figure 8-4). Equilibrium reduction
potentials for copper and germanium were taken as their open circuit potentials listed in
chapter six. Partial current densities of H2 were calculated by subtracting the copper and
germanium partial currents from the total applied current, assuming H2 evolution is the
only side reaction that occurs. The Butler-Volmer equation, shown in Eq. 8-11, was
fitted to each polarization curves using OriginTM software by least squares fitting method.
Least square fitting yielded best guess values for the i0 and b (inverse of Tafel slope).
These values are tabulated in table 8-1.
135
i  i0 e
  nF 


 RT 
2
i 0 (A/cm )
 i0 e b 
8  11
b
αc
-4
5.5
0.071
-8
10.7
0.069
-10
15.3
0.393
Copper
2 x 10
Germanium
3 x 10
Hydrogen
4 x 10
Table 8-1: Kinetic parameters for copper, germanium and hydrogen obtained via least
square fitting of the polarization curves in figure 8-1. αc is the cathodic charge transfer
coefficient.
136
3.0x10-2
-2
2.5x10-2
2.0x10
-2
1.5x10
-2
1.0x10
-2
5.0x10
-3
-10
i0 : 3.7 x 10
A/cm
2
Current Density (A/cm2)
Current Density (A/cm2)
3.0x10
Cu
b : 15.3
(a)
2.5x10
b : 10.8
Ge
-2
2.0x10
1.5x10-2
1.0x10-2
(b)
5.0x10-3
0.0
0.0
0.4
i0 : 2.7 x 10-8 A/cm2
-2
0.6
0.8
1.0
1.2
0.4
0.6
Overpotential (V)
0.8
1.0
1.2
Overpotential (V)
-2
Current Density (A/cm2)
3.0x10
i0 : 2 x 10-4 A/cm2
-2
2.5x10
b : 5.4
H2
-2
2.0x10
-2
1.5x10
-2
1.0x10
(c)
-3
5.0x10
0.0
0.4
0.6
0.8
1.0
1.2
Overpotential (V)
Figure 8-4: Polarization curves of (a) hydrogen (b) germanium and (c) copper on a
copper rotating disk electrode from the copper and germanium containing electrolyte at
1000 rpm. Least squares fitted curves are shown as well as the best fit values for the
kinetic parameters.
137
8-2.2: Determination of the transport parameters
Diffusion coefficient of cupric ions in a tartrate complexed electrolyte is listed in
literature10. A diffusion coefficient of 6  10 6 cm2/s was used in the pulse plating model.
Diffusion coefficient of germanium in solution, however, is not available in the literature.
Therefore, this value was obtained experimentally via cyclic voltammetry. Cyclic
voltammograms were performed on a steel rotating disk electrode in the copper-free
alkaline tartrate electrolyte. Potential was scanned from -0.42 V (vs. SCE) to -2.4 V (vs.
SCE) in the cathodic direction and to +0.1 V (vs. SCE) in the anodic direction and
currents at the electrode were recorded (Figure 8-5). Germanium reduction peak is
observed at about -1.2 V (vs. SCE). The discrepancy between the equilibrium reduction
potential (-0.364 V vs. SCE) and the observed one could be attributed to the effects of
complexation between germanium and tartrate. The peak current densities (ipeak) are
related to the scan rates (υ) by the following equation11:
1/ 2
i peak
 F3 
3/ 2 1/2
1/ 2
 0.4463 
 n D CBulk
 RT 
(8-12)
Here F is the Faradays constant. R and T are the universal gas constant and temperature.
D is the diffusion coefficient of germanium. Cyclic voltammograms on the steel RDE for
various scan rates are shown in Figure 8-5. The peak current increases linearly with scan
rates (Figure 8-6) and diffusion coefficient can be calculated from the slope of the linear
curve by using Eqn. 8-12. For germanium, the diffusion coefficient is calculated to be
2.2  10 8 cm2/s. The two order of magnitude difference between the values of diffusion
138
coefficients of Cu and Ge could be explained by the nature of the Ge complex in solution
which might form a bulkier complex than Cu, thereby reducing its mobility.
Figure 8-5: Cyclic voltammograms on a steel rotating disk electrode from the copperfree alkaline tartrate electrolyte. Potential was scanned from -0.42 V (vs. SCE) to -2.4 V
(vs. SCE) in the cathodic direction and to +0.1 V (vs. SCE) in the anodic direction. The
figure above has been magnified to better illustrate the peak currents.
139
Peak current density (A/cm2)
4.0x10-2
DGe= 4 x10-8 cm2/s
3.0x10-2
2.0x10-2
1.0x10-2
0.4
0.6
0.8
1.0
(Scan rate)1/2 [V/s]1/2
Figure 8-6: A plot of peak current densities vs. scan rates, obtained from Figure 8-4.
Diffusion coefficient can be calculated from slope of the linear plot. Slope was obtained
by fitting a straight line using least squares method in origin.
A diffusion layer thickness of 20 µm for copper was assumed. Gas evolution at
the cathode is used as the mode of convection since 75 % of the total current goes toward
hydrogen evolution. There is no known correlation between gas evolution at the
electrode and effective diffusion layer thickness. Under natural convection, diffusion
layer can extend as much as 300 µm into the bulk while for forced convection (such as
140
gas evolution) diffusion layer thickness can fall anywhere in the range from 5 to 100 µm.
Therefore, a value of 20 µm is a reasonable one. Germanium diffusion layer thickness
1/3
D 
was calculated from the copper one by scaling it by a factor of  Ge  . This resulted in
 DCu 
a diffusion layer thickness of 4 µm for germanium.
141
8-3. Effects of mass-transfer and kinetics in Cu3Ge pulse plating
Film composition and faradaic efficiency of the Cu3Ge pulse plating process were
measured and compared to those predicted by the model to validate it. Germanium
composition in the film was calculated using its partial current via Faradays law. Partial
current density was obtained by solving Eq. (8-6), (8-7) and (8-10) for concentration at
the electrode at end of pulses as outlined in section 8-1. Partial current densities and
germanium composition of the film at various total applied current densities are shown in
Figure 8-7. Experimentally obtained germanium concentration in the film at various
applied pulse current densities are also shown in Figure 8-7. The steady-state pulse
model agrees reasonably well with the experimental data. Germanium incorporation in
the deposit is kinetically controlled at current densities less than 200 mA/cm2. Between
200-300 mA/cm2, germanium composition goes through a maximum indicating the
transition to deposition under mass transfer control. Experimentally, this limiting current
peak is observed at slightly higher current densities, between 300-400 mA/cm2.
However, both model and experimental data indicate that germanium reaches its mass
transfer limit prior to copper in galvanostatic pulse plating situations. This is further
illustrated in Figure 8-8. Concentration of copper and germanium at the electrode are
shown at the end of different current density pulses. Concentration of germanium at the
electrode reaches zero at 350 mA/cm2 while copper does not reach its mass transfer limit
until 550 mA/cm2.
142
Ge Content (at. %)
30
25
Experiment
20
15
10
Steady-state Model
5
0
0
100
200
300
400
500
600
Current Density (mA/cm2)
Figure 8-7: Atomic concentration of germanium in the alloy at different pulse current
densities. A pulse ‘on’ time (ton) = 2.2 ms and duty cycle = 0.02 were used both in the
steady-state model and in the experiments. Diffusion boundary layer thicknesses of
copper and germanium were taken as 20 µm and 4 µm, respectively.
143
Dimensionless Concentration
1.0
0.8
0.6
Cu
0.4
Ge
0.2
0.0
100
200
300
400
500
600
Current Density (mA/cm2)
 C 
Figure 8-8: Dimensionless concentration of copper and germanium  s  at different
 Cbulk 
pulse current densities. Simulation conditions were same as in Figure 8-7.
Faradaic efficiency of the plating process was also calculated by subtracting the
hydrogen evolution current from the total current. Figure 8-9 shows faradaic efficiency
of the Cu3Ge pulse plating at different applied current densities. At about 205 mA/cm2,
the faradaic efficiency is calculated to be about 65 %. Under the same experimental
conditions, for 8 minutes of plating, the faradaic efficiency was measured to be between
25-35 %. This discrepancy is possibly due to gas evolution and its coalescence to
bubbles during plating, thus reducing the faradaic efficiency of the process. Hydrogen
gas bubbles are continuously evolved during high current density plating and coalesce on
144
the electrode surface. The effect of bubble coalescence and removal from the surface on
the faradaic efficiency and mass-transfer were not taken into account in the steady-state
pulse plating model. Figure 8-10 illustrates the effect of plating time on faradaic
efficiency. At short plating time (< 5 minutes), current efficiency is almost 50 %. At
longer times, the efficiency is reduced to almost 16 % due to increased gas evolution and
bubble coalescence, masking portion of the electrode surface.
100
90
Efficiency (%)
80
70
60
50
40
30
20
10
0
0
100
200
300
400
500
600
Current Density (mA/cm2)
Figure 8-9: Faradaic efficiency of Cu3Ge pulse plating at different applied current
densities calculated using the steady-state model. Identical parameters as Figure 8-6
were used.
145
100
90
80
Efficiency (%)
70
60
50
40
30
20
10
0
0
5
10
15
20
25
30
Plating Time (minutes)
Figure 8-10: Faradaic efficiency of Cu3Ge pulse plating at various total plating times.
Experimental condition: i=205 mA/cm2, ton=2.2 ms, toff=100 ms.
In summary, a simplified steady-state model for galvanostatic pulse plating of
binary alloy was developed that takes into account the mass-transfer and kinetics of the
plating process. The model was specifically applied to the electrodeposition of Cu3Ge.
The effect of gas evolution on mass-transfer and faradaic efficiency were not considered
as part of the pseudo-steady-state model. Therefore, for long plating times, this model
does not accurately predict the faradaic efficiency of the process. However, for short
plating times and for predicting film composition, this steady-state pulse plating model
provides an effective modeling tool.
146
8-4. Supplementary Data-Matlab (C++) based pulsed alloy deposition modeling code
function model1 = system3(x,itot)
global deltapcu deltapge deltacu deltage kcu kge gamma t i0h2 bh2 Dcu
Dge i0cu bcu i0ge bge
cbcu=0.000025;
ncu=2;
f=96485;
i0cu=2*10^(-4);
Erev1=0.38;
bcu=5.5;
cbge=0.000096;
nge=4;
Erev2=0.45;
i0ge=2.7*10^(-8);
agent)
bge=10.77;
Erev3=0.42;
R=5.14;
% Bulk Cu concentration
% # of electrons transferred
%
%
%
%
%
%
%
exchange current density of Cu
reversible of Cu vs. SCE (OCP)
inverse of tafel slope
Bulk Ge concentration
# of electrons transferred
reversible potential of Ge vs. SCE (OCP)
exchange current density of Ge (w/o complexing
% inverse of tafel slope (w/o complexing agent)
% H2 reversible potential
% resistance of the solution (in ohms)
model1 = [-x(1)+cbcu*(1)-x(7)*deltacu*kcu/(ncu*f*Dcu)(x(7)*gamma*(deltacu-deltapcu)/(ncu*f*Dcu));
x(7)-(i0cu*(x(1)/cbcu)*exp(bcu*(x(3)-Erev1-x(4))));
x(4)-R*(i0cu*(x(1)/cbcu)*exp(bcu*(x(3)-Erev1-x(4))));
-x(2)+cbge*(1)-x(8)*deltage*kge/(nge*f*Dge)(x(8)*gamma*(deltage-deltapge)/(nge*f*Dge));
x(5)-R*(i0ge*(x(2)/cbge)*exp(bge*(x(3)-Erev2-x(5))));
x(8)-(i0ge*(x(2)/cbge)*exp(bge*(x(3)-Erev1-x(5))));
itot-x(7)-x(8)-i0h2*exp(bh2*(x(3)-Erev3-x(6)))
x(6)-R*(i0h2*exp(bh2*(x(3)-Erev3-x(6))))];
end
:
function callfunction3
format long
global deltapcu deltapge deltacu deltage kcu kge gamma t i0h2 bh2 Dcu
Dge
ncu=2;
%
Dcu=0.000006;
%
Dge=4*10^(-8);
%
nge=4;
%
i0h2=3.7*10^(-10);%
bh2=15.3;
%
Erev3=0.42;
%
# of electrons transferred
Cu diffusion coefficient
Ge diffusion coefficient
# of electrons transferred
exchange current density of H2
inverse of tafel slope
H2 reversible potential
147
iinit=0.010;
% simulation start
mwcu=63.55;
% Cu molecular weight
mwge=72.63;
% Ge molecular weight
ifinal=0.600;
% simulation end
hi=0.010;
% step size
toff=0.100;
% rest period
gamma=t/(t+toff); % duty cycle
deltacu=0.0020;
deltage=0.0004;
deltapcu=(4*Dcu*t*(1-gamma)/3.1416)^1/2;
deltapge=(4*Dge*t*(1-gamma)/3.1416)^1/2;
t=0.0022;
kcu=sqrt((4*Dcu*t/(3.1416*(deltacu)^2)));
kge=sqrt((4*Dge*t/(3.1416*(deltage)^2)));
fid=fopen('simulation results 2.2 ms 0-600mA.txt','w+');
guess = [0.000025 0.000096 2.0 0.1 0.05 0.3 0.010 0]; % initial
conditions at t=0
for itot=iinit:hi:ifinal
results=fsolve(@(x)system3(x,itot),guess,optimset('MaxIter',2000,'TolFu
n',1e-9,'MaxFunEvals',8000,'Display','iter'));
hold on
ih2=i0h2*exp(bh2*((results(:,3))-Erev3-results(:,6)));
volmer, no mass-xfer limitation for H2
% Butler-
figure (1)
plot (itot*1000,results(:,1)*1000,'-ks','LineWidth',3);
hold on
plot(itot*1000,results(:,2)*1000,'-bo','LineWidth',3);
xlabel('Current Density (mA/sq.cm)'); ylabel('Surface Concentration
(Mol/L)');
hold on
legend('Cu','Ge');
figure (2)
plot(itot*1000,results(:,7)*1000,'-ks','LineWidth',3);
hold on
plot(itot*1000,results(:,8)*1000,'-bo','LineWidth',3);
hold on
plot(itot*1000,ih2*1000,'-k^','LineWidth',3);
xlabel('Current Density (mA/sq.cm)'); ylabel('Current Density
(mA/sq.cm)');
hold on
legend('iCu','iGe','iH2');
figure (3)
plot(itot*1000,results(:,3)*-1,'-ks','LineWidth',3);
xlabel('Current Density (mA/sq.cm)'); ylabel('Potential (V vs. SCE)');
hold on
148
icu=results(:,7);
ige=results(:,8);
Gecontent=(ige*mwge*100/nge)/((icu*mwcu/ncu)+(ige*mwge/nge));
calculates % of Ge in deposit (Wge/Wtotal)
eff=(1-(ih2/itot))*100;
%
% faradaic efficiency
fprintf(fid,'%f %0.7f %0.7f %f %f %f %f %f\r\n'
,itot*1000,results(:,1)*1000,results(:,2)*1000,icu*1000,ige*1000,ih2*10
00,results(:,3)*-1,Gecontent);
figure(4)
plot(itot*1000,Gecontent,'-bs','LineWidth',3);
xlabel('Current Density (mA/sq.cm)'); ylabel('Ge Content (%)');
axis([iinit*1000 ifinal*1000 0 60]);
hold on
figure(5)
plot(itot*1000,eff,'k^','LineWidth',3);
xlabel('Current Density (mA/sq.cm)'); ylabel('Efficiency(%)');
axis([iinit*1000 ifinal*1000 0 100]);
hold on
end
hold off
fclose(fid);
149
References:
1. T. R. Rosebrugh and W. L. Miller, J. Phys. Chem. 14, 816 (1910).
2. H. Y. Cheh, J. Electrochem. Soc. 118, 551 (1971).
3. N. Ibl, Surf. Technol. 10, 81 (1980).
4. M. Datta and D. Landolt, Surf. Technol. 25, 97 (1985).
5. A. R. Despic and K. I. Popov, J. Appl. Electrochem. 1, 275 (1971).
6. Y. N. Dordi and U. Landau, Abstract 467, p.1175, The Electrochemical Society
Extended Abstracts, Vol. 79-2, Los Angeles, CA, Oct 14-19, 1979.
7.
K-M. Yin and R. E. White, AIChE Journal. 36, 187 (1990).
8. Z. Liu and A. C. West, Electrochim. Acta. 56, 3328 (2011).
9. A. Ruffoni and D. Landolt, Electrochim. Acta. 33, 1281 (1988).
10. C-H. Liao, PhD Dissertation, Case Western Reserve University, Cleveland, OH
(2012).
11. A. J. Bard and L. R. Faulker, Electrochemical Methods: Fundamentals and
Applications, p. 231, 2nd Ed, Wiley (2004).
12. Y. G. Siver, Russ. J. Phys. Chem., 34, 273 (1960).
13. M. Menon and U. Landau, J. Electrochem. Soc., 137, 445 (1990).
150
Chapter 9
Conclusions and Future Work
9-1. Conclusions
A novel and versatile copper plating chemistry has been developed that provides
high nucleation density of Cu (>1011 nuclei/cm2) on a barrier layer (Ru), bottom-up
metallization with SPS and PEI as additives and demonstrate alloy plating capability.
High nucleation rate and nucleation density of Cu on Ru was achieved by plating Cu
from the alkaline Cu-tartrate-complex electrolyte using high current density pulses.
Combination of the complexed chemistry and pulsing facilitates deposition at high
overpotential which promotes homogeneous nucleation and high nucleation rate.
Bottom-up fill was demonstrated from the same alkaline Cu-complex chemistry using
SPS as a suppressor and PEI as an anti-suppressor, different than the conventional
PEG/SPS (suppressor/anti-suppressor) combination used in the acidic medium.
Electrodeposition of Cu-alloy (e.g., Cu3Ge) was also shown to be possible from these
alkaline complexed electrolytes. A square wave pulse current waveform was utilized for
Cu3Ge electrodeposition that enabled plating at high instantaneous current density,
providing Cu and Ge in a 3:1 ratio. The electroplated Cu3Ge films were observed to be
smooth and compact. XRD indicated formation of a monoclinic crystal structure in the
plated film. Resistivity of the plated Cu3Ge films was found to be higher than the PVD
Cu3Ge, which is possibly due to impurity incorporation and small grain size of the
deposit.
151
The aforementioned multi-functional characteristics of the alkaline coppercomplexed electrolytes make them an attractive candidate for damascene integration in
>20 nm interconnects fabrication. However, there are several issues that need to be
addressed prior to the industrial implementation of the chemistry and associated plating
process.
9-2. Future Work
Incorporation of impurities (C, O and N) in Cu films deposited from the
complexed electrolytes causes Cu resistivity to deviate from its bulk value of 1.7 µΩ-cm.
The impurity incorporation is due to the presence of organic ligands in the electrolyte that
contain carbon, oxygen and nitrogen atoms. The increased resistivity, presumably
associated with impurities in the Cu films is undesirable as it increases the ‘RC’ delay in
devices. However, mechanism for incorporation of these impurities from the complexed
electrolytes is not currently known. The effects of various process parameters, i.e.,
current density, pH, ligand concentration in the electrolyte and bath temperature, on
impurity levels must be characterized in order to identify preferred process condition that
minimizes impurity incorporation. Currently, the impurity level in Cu plated from the
complexed electrolytes in the presence of additives is approximately 3 atomic percent.
This value needs to be reduced by at least an order of magnitude before complexed
electrolytes can be considered as viable replacements for the conventional acid
electrolytes in the damascene process. Another adverse effect of plating Cu from the
alkaline complexed electrolytes is hydrogen co-evolution. This is prominent in the case
152
of alloy plating where high current density pulses are employed. Hydrogen evolution and
coalescence of the gas bubbles at the electrode introduce roughness and porosity in the
deposit that subsequently increase its resistivity. This is evident by the presence of
nanometers sized pores in cross-sections of the Cu3Ge film. Inclusion of the gas bubbles
in the deposit and thereby its porosity can be reduced by employing agitation during
plating to enhance the removal of the bubbles at the electrode as they are generated. The
hydrogen evolution reaction can also be eliminated by using non-aqueous solvents i.e.,
organic or molten salt electrolytes. However, this may introduce additional issues
pertaining bath stability and additives interaction in a non-aqueous environment, as well
as cost and complexity.
Another avenue for further future work pertains to the extendibility of the alkaline
process and its theoretical limit. If one were to assume homogeneous nucleation and
favorable energetics between Ru and Cu atoms, minimum critical thickness of the seed
layer that can be obtained is the atomic radius of Cu which is approximately 0.13 nm.
Hence, the alkaline process theoretically could be extended to features in which the seed
layer is ~0.13 nm. Opportunities for further research exist in exploring new chemistries
and optimizing process parameters that yields homogeneous growth of copper on Ru with
high nucleation rate, thereby enabling electrodeposition of thinner Cu seed layers.
It is widely believed that the established dual-damascene process flow which
involves a Cu seed layer deposition by PVD and subsequent metallization by
electroplating copper from an acidic electrolyte will hit its limit at <20 nm features.
Alternative chemistries for copper deposition must be explored to further extend Moore’s
law for future generation devices. Alkaline copper-complex chemistries offer several
153
advantages over the conventional acid electrolytes such as conformal, <10 nm seed layer
deposition, electrodeposition of Cu-alloys relevant for next generation of interconnects.
Bottom-up electrodeposition of submicron features has also been demonstrated.
Although several concerns remain, alkaline Cu-complex chemistries offer an alternate
route for metallizing interconnect features in 22-nm technology nodes and beyond.
Chapter 10
Appendix: Electrodeposition of Cu-Mn
10-1. Introduction
Continued scaling of interconnect dimensions for improved performances in logic
and memory devices is dictated by Moore’s law. In conventional metallization scheme,
these interconnects are fabricated using a dual-damascene process flow that involves
electrodeposition of copper to fill the patterned features1. As interconnects geometries
shrink for future technlogy nodes, the electromigration (EM) lifetime of pure Cu becomes
a significant issue2. The reduced EM lifetime of pure Cu in smaller geometries leads to
poor reliablity for the Cu interconnect lines. A simple method for improving the EM
lifetime of Cu is to alloy it with ‘impurities’, such as Mn2-5. In this scheme, the alloying
element (Mn) is introduced either in the seed layer or the liner. During subsequent
anneals, manganese segregates at the grain boundaries and the interface between the Cu
and the capping layer due to its higher diffusivity relative to Cu. Accumulation of
manganese at the capping interface impedes Cu diffusion increasing its EM lifetime.
Inclusion of 1-2 at. % manganese has been shown to improve the EM lifetime of Cu lines
by four times2. An additional benefit of a Cu-Mn alloy is its ability to react with SiO2 at
154
the dielectric interface forming a manganese silicate barrier that effectively prevents
interdiffusion of copper and silicon6-9. This ‘self-forming’ barrier process presents an
alternative to the conventional chemical vapor (CVD) or atomic layer deposition (ALD)
of the Ti and Ta-based barriers that become challenging at reduced geometries.
Traditionally, Cu-Mn is deposited onto SiO2 by either sputtering6,7 or CVD8,9 which is
followed by a heat treatment to segregate the manganese. However, due to the
incompatibility of sputtering and CVD with the established dual-damascene process flow,
an electrodeposition route for fabricating Cu-Mn is desired. Electrodeposition offers
several advantages over dry processes including high throughput, low cost of ownership
and the ability to provide bottom-up electroplating.
In spite of its apparent advantages, electrodeposition of copper rich, manganese
alloys from aqueous electrolytes for applications in interconnect metallization has not
received much attention. This is mostly due to the signficant challege associated with
reduction of manganese which is highly electronegative (
compared to copper (
ECu 2 / Cu
EMn2 / Mn
= -1.18 V vs. SHE)
= +0.34 V vs. SHE)10. Most literatures have focused on
the possible application of managanese alloys as sacrificial protection layers for steel11-13.
These manganese-rich copper alloys are, however, not applicable in Cu interconnect
metallization due to its increased resistivity. Efforts to deposit copper-rich manganese
alloy on steel yielded spongy and porous films12. Similar results were obtained for
electrodepostion of Cu-Mn on n-type and p-type silicon14. Due to vigorous hydrogen
evolution, the deposited film was non-uniform and porous rendering it inadequate for
interconnect applications.
155
Previously we have shown applicability of copper-complexed electrolytes in
providing high nucleation density of Cu on Ru15, bottom-up fill in features16 and
depositing Cu3Ge17. We describe here electrodeposition of Cu-Mn from a similar
complexed electrolyte using a pulse current waveform where manganese content in the
film ranges from 1-2 atomic percent. Furthermore, we report that, after annealing,
manganese atoms segregate at the film interface, similar to sputtered or CVD Cu-Mn
films.
10-2. Experimental
A three-electrode setup in a separated cell was used for Cu-Mn electrodeposition.
The counter (anode) and working electrodes (cathode) were separated by a fitted glass
plug. Separation of the anodic and the cathodic compartments prevented contamination
of solution due to formation of manganese oxides on the anode11-13. A platinized
titanium mesh was used as the counter electrode and a saturated calomel electrode (SCE)
served as the reference electrode. The electrolyte contained 0.05 M CuSO4.5H2O (Fisher,
Certified ACS), 0.59 M MnSO4.H2O (Sigma-Aldrich), 0.052 M Ethylenediamine
tetraacetic acid (Fisher, Technical Grade) and 1 M (NH4)2SO4 (Fisher, Certified ACS).
The pH of the electrolyte was between 6.4-6.6. The electrolyte temperature during
plating was 25 oC. A square pulse waveform, similar to the one shown in Figure 3, was
used to electroplate Cu-Mn The pulse train consisted of a 167 ms long, 300 mA/cm2
pulse followed by a 5 mA/cm2 pulse with a duration of 5 s. Electrochemical
measurements were performed on a Cu disk electrode with an active area of 0.32 cm2
156
rotating at 300 rpm. A Hitachi S4500 scanning electron microscope (SEM) equipped
with a Noran Energy Dispersive Spectrometer (EDS) was used to analyze the deposit
composition. A VSP Bio-logic potentiostat provided power for all experiments.
To characterize the mobility of manganese in electroplated Cu-Mn films, Cu-Mn
was electrodeposited onto silicon wafer coupons pre-coated with a Cu seed layer and
Ta/TaN diffusion barrier. Agitation was provided by a stir bar to remove gas bubbles
generated at the electrode during plating. Electroplated Cu-Mn films were annealed in a
N2 environment inside a PureLab HE 4GB 2500 glove box at ~400 oC for 1 hour. A PHI
VersaProbe XPS Microprobe was used to obtain the depth profile of the electroplated
copper-manganese film. Cross-section of the plated film was prepared and imaged using
a FEI Helios Nanolab 650 SEM equipped with a Focused Ion Beam. Roughness of the
plated films was characterized using a KLA-Tencor P-6 Stylus Profilometer.
10-3. Results & Discussions
Electrochemical investigation.- Cathodic potentiodynamic behaviors of 1 M
(NH4)2SO4 + 0.59 M MnSO4.H2O and the complete electrolyte are shown in Figure 1. In
the complete electrolyte containing EDTA, selective complexation occurs between Cu
and EDTA due to a stability constant that is five orders of magnitude higher than MnEDTA18. Hence, in the plating solution manganese is either free in solution (as Mn2+) or
weakly complexed with NH3 (stability constant ~ 0.9)19. Onset of H2 evolution in the
‘Mn only’ electrolyte is observed at -0.7 V vs. SHE. Since manganese reduces at an
electrode potential more negative than -1.18 V vs. SHE, it is impossible to separate the
157
contributions of manganese reduction and hydrogen evolution to the cathodic current.
However, it is clear that the manganese reduction occurs somewhere between -1.1 V and
-1.4 V vs. SHE, probably closer to the less negative end of this range since manganese is
very weakly complexed in solution. In the complete electrolyte containing Cu-EDTA,
onset of copper reduction is observed at -0.19 V vs. SHE, significantly more cathodic
than its standard reduction potential (+0.34 V vs. SHE). This shift is typical of
electrodeposition of copper from a complexed electrolyte15. The plateau observed in the
current density range of 18-20 mA/cm2 is attributed to the limiting current of Cu
electrodeposition, which is calculated for our system to be approximately 23 mA/cm2
using the Levich equation20. The challenge associated with Cu-Mn co-deposition is
readily visible in Figure 2. Electrodeposition of manganese, which occurs within the
potential window of -1.1 V to -1.4 V, is always accompanied by vigorous hydrogen
evolution and plating Cu at its mass transfer limit neither of which is desirable for
obtaining a smooth and compact electrodeposit. This is illustrated in Figure 2. The
potential transients, EDS spectra and surface morphologies of the deposits plated at 20
and 200 mA/cm2 are shown. At 20 mA/cm2, close to the limiting current density of
copper, the plated films were microscopically rough. Large (1-3 µm) spherical particles
were observed which is typical of films deposited close to its mass transfer limit.
Futhermore, no Mn was detected in the plated film at 20 mA/cm2, which is reasonable
since the electrode potential was lower than the manganese reduction potential. At 200
mA/cm2, potential more negative (-1.6 V vs. SHE) than the manganese reduction
potential, ~2 at. % manganese was detected in the film. However, the plated films were
visibly powdery and rough due to vigorous hydrogen evolution and copper being plated
158
at its mass transfer limit. The resulting porosity of the plated film led to poor adhesion of
the film to the substrate which failed the qualitative peel test. Clearly, galvanostatic
plating is not the preferred method for obtaining Cu-Mn films as it leads to rough and
poorly adherent deposits that are not suitable for the intended application in
microelectronics fabrication.
Figure 10-1: Linear sweep voltammetry on a Cu electrode from 1 M (NH4)2SO4 + 0.59
M MnSO4.H2O electrolyte and the complete electrolyte (0.05 M CuSO4.5H2O + 0.052 M
EDTA + 0.59 M MnSO4.H2O + 1 M (NH4)2SO4). Scan rate and rotation speed were 50
mV/s and 300 rpm, respectively. A magnified image of the area of interest is shown in
the inset in bottom right corner.
159
Figure 10-2: Co-deposition of copper and manganese on a copper RDE using DC
waveform. Potential transients, EDS spectra and morphologies of the films deposited at
20 and 200 mA/cm2 are shown. Rotation speed was 300 rpm in all cases.
Pulse electrodeposition.- To circumvent the adverse effects of galvanostatic
plating of Cu-Mn, a pulse waveform similar to the one shown in Figure 3 was utilized to
deposit Cu-2 at. % Mn (Figure 4c). Application of such a high-low pulsed current
waveform allows selective deposition of Mn during the high current density pulses (300
mA/cm2) and Cu during the low current density pulses (5 mA/cm2) yielding Cu-Mn in the
desired stoichiometric ratio. Furthermore, evolution of roughness associated with plating
Cu above the limiting current in a DC waveform is minimized in the pulse waveform due
to the instantaneous nature (167 ms) of the high current density pulse. Roughness of CuMn films deposited using DC and pulse waveforms are compared in Figure 5. For
160
comparison, we select a DC current density of 20 mA/cm2 which is close to the Cu mass
transfer limit and a pulse waveform similar to the one shown in Figure 3. Beneficial
aspects of the pulse waveform are clearly evident as it significantly reduces roughness
while enabling co-deposition of Cu and Mn. The root mean squared (rms) roughness of
the pulse plated deposit was 372 nm, substantially lower than that of the DC plated one
(850 nm). We believe this roughness can be further reduced by decreasing the duration
of the high current density pulse and proper flow control. An additional benefit of the
relatively short duration of the high current density pulse in comparison to the low
current density pulse is that it significantly reduces gas evolution and coalescence of gas
bubbles on the electrode promoting compactness of the deposit. Unlike the film
deposited using a 200 mA/cm2 continuous current, Cu-Mn films deposited using the
pulse scheme were compact (Figure 4a). It passed the qualitative peel test indicating
excellent adhesion to the underlying copper substrate. The faradaic efficiency of the
pulse plating process was calculated from the measured thickness to be approximately 30
%. The low faradaic efficiency is attributed to co-evolution of hydrogen which always
accompanies Mn reduction during the high current density pulse. The deposition rate of
the film using the pulse waveform in Figure 3 was calculated to be approximately 8.3
nm/pulse.
161
Figure 10-3: Schematic of the pulse plating waveform used to deposit Cu-Mn in a 49:1
stoichiometric ratio. A 300 mA/cm2 pulse with a 0.167 s ‘on’ time was followed by a 5
mA/cm2 pulse with a 5 s ‘on’ time. A pulse train consisting of 109 pulses was applied for
a total charge density of 8.2 C/cm2.
162
Figure 10-4: (a) Cross-section of the Cu-Mn film electrodeposited on a Cu seeded wafer
using the pulse waveform shown in Figure 1. A protective layer of platinum was
sputtered on top of the deposit prior to milling (b) a top-down micrograph of the plated
Cu-Mn film without the Pt cap (b) EDS spectra of the Cu-Mn film indicating codeposition of Cu and Mn in a 49:1 ratio.
163
Figure 10-5: Surface roughness profiles of electroplated Cu-Mn using DC (black dashed
line) and pulse (red solid line) plating. Both experiments corresponded to a total charge
density of 8.2 Cb/cm2. The rms roughness of the DC plated film was 850 nm, while that
of the pulse plated film was 372 nm.
Segregation of manganese:- In order to enhance the EM lifetime of
interconnects, Mn atoms must demonstrate higher diffusivity than the Cu atoms after heat
treatment and segregate at the interfaces. The segregation of Mn at the interface after
annealing was investigated via XPS. Mn 3p spectral line (at 48 eV) was chosen for
analysis since the Mn 2p lines could not be deconvoluted from the Cu auger lines and the
Mn 3s had a very low intensity peak. XPS depth profile indicates that after annealing at
~400 oC for 1 hour, manganese atoms in the electrodeposited Cu-Mn films migrate to the
film surface (Figure 6). Virtually no manganese was detected on the surface of the asdeposited film whereas after annealing it showed a distinct 3p peak (Figure 6a),
164
corresponding to a 12 at. % of manganese in the alloy. This was also observed
qualitatively by the apparent color change of the deposit from a dark golden before the
heat treatment to a slightly pink color after (Figure 6a). Similar phenomenon has been
reported for sputtered and CVD Cu-Mn. However, in this work, we report for the first
time the diffusion and segregation of Mn in an electrodeposited Cu-Mn film that is highly
desirable for damascene integration.
Another attribute of a Cu-Mn alloy is its self-forming barrier capability.
However, without an adhesion layer, sputtered Cu had poor adhesion to SiO2. Hence,
Cu-Mn could not be deposited onto Cu seeded wafers from a wet chemistry that did not
already have an underlying adhesion layer such as Ti or TaN. A method for enhancing
the adhesion of Cu to SiO2 without using an intermediate layer must be developed in
order to exploit the self-forming barrier capability of electroplated Cu-Mn.
165
Figure 10-6: (a) XPS spectra of Mn 3p at the surface of the electroplated Cu-Mn film
before and after annealing (b) a depth profile of the manganese content in the alloy before
and after annealing that indicates segregation of Mn at the interface.
10-4. Conclusions
In summary, a Cu-Mn alloy containing 2 at. % of manganese has been
electrodeposited from a complexed electrolyte using a pulse waveform. The deposit is
shown to be compact, adherent and relatively smooth. Mn atoms in the electroplated CuMn alloy demonstrate diffusivity, segregating at the interface after heat treatment.
Electrodeposition of Cu-Mn offers an attractive avenue for metallizing future generation
interconnects that require improved reliability due to its ease of integration.
166
References:
1. P.C. Andricacos, C. Uzoh, J.O. Dukovic, J. Horkans and H. Deligianni, IBM J. of
Res. and Dev. 42, 567 (1998).
2. J. P. Gambino, “Improved Reliability of Copper Interconnects Using Alloying”,
Proc. 17th IEEE IPFA, pp. 1-7, 2010.
3. T. Nogami, T. Bloom, A. Simon, B-Y. Kim, C-K. Hu, et al, “High Reliability 32
nm Cu/ULK BEOL Based on PVD CuMn Seed, and its Extendibility”, Proc.
IEDM, pp. 33.5.1-33.5.4, 2010.
4. C. Christiansen, B. Li, M. Angyal, T. Kane, V. McGahay, Y. Y. Wang and S.
Yao, “Electromigration-resistance enhancement with CoWP or CuMn for
advanced Cu interconnects”, Proc. IRPS, pp. 3E.3.1-3E.3.5, 2011.
5. T. K. Indukuri, R. N. Akolkar, J. S. Clarke, A. Genc, F. Gstrein, M. C. Harmes, B.
Miner, F. Xia, D. J. Zierath, S. Balakrishnan, Microelectron. Eng., 92, 49 (2012).
6. J. Koike and M. Wada, Appl. Phys. Lett. 87, 041911 (2005).
7. M. Haneda, J. Iijima and J. Koike, Appl. Phys. Lett. 90, 252107 (2007).
8. K. Neishi, S. Aki, K. Matsumoto, H. Sato, H. Itoh, S. Hosaka and J. Koike, Appl.
Phys. Lett. 93, 032106 (2008).
9. Y. Au, Y. Lin, H. Kim, E. Beh, Y. Liu and R. G. Gordon, J. Electrochem. Soc.
157, D341 (2010).
167
10. A. Brenner, Electrodeposition of Alloys, Vol. 1, p.137, Academic Press, New
York (1963).
11. J. Gong and G. Zanagari, J. Electrochem. Soc. 149, C209 (2002).
12. J. Gong and G. Zangarai, J. Electrochem. Soc. 151, C297 (2004).
13. F. Mangolini, L. Magagnin and P. L. Cavallotti, J. Electrochem. Soc. 153, C623
(2006).
14. A. Radisic, J. L. Hernandez, Z. Tokei, G. P. Beyer and P. M. Vereecken, ECS
Trans. 3, 69 (2007).
15. A. Joi and U. Landau, “An Alkaline Copper Plating Process Providing High
Nucleation Density on Ru and Bottom-up Fill”, 220th ECS Meeting, Abstract
#1944, 9-14 October 2011, Boston, MA.
16. A. Joi, R. Akolkar and U. Landau, J. Electrochem. Soc., 160, D3001 (2013).
17. A. Joi, R. Akolkar and U. Landau, Appl. Phys. Lett. 102, 134107 (2013).
18. A. E. Martell, Critical Stability Constants., Vol. 2, Plelum press, New York
(1974).
19. R. M. Smith, A. E. Martell and R. J. Motekaitis, “NIST Standard Reference
Database 46: NIST Critically Selected Stability Constants of Metal Complexes
Database”, Version 8 (2004).
20. V. G. Levich, Physiochemical Hydrodynamics, pp. 60, Prentice-Hall, New Jersey
(1962).
168
List of Symbols
d = Spacing between parallel crystal planes
s = Tip distances in a four-point probe
Rs = Sheet resistance, Ω/square
dcrit = Critical coalescence thickness for nucleation
N = Nucleation density
n = Number of atoms in a cluster or nuclei
F = Faradays constant, 96485 C/mole
Z = Number of electrons transferred for reduction of a metal
J = Nucleation rate
k = Boltzmann’s constant
pK = Stability constant of metal complexes
E = Voltage recorded at the electrode
E0= Standard reduction potential of Metals
R = Universal Gas Constant
T = Temperature
ton = Pulse ‘on’ time or pulse width
toff = Pulse ‘off’ time or rest period
C = Concentration of species in solution
Cbulk = Bulk concentration of species in solution
Cs= Concentration of species at the surface of the electrode
D = Diffusion coefficient
ipulse = Pulse current density
169
Greek Letters
β = Broadening of the diffraction line measured at half its maximum intensity
λ = Wavelength of the X-ray
θ = Bragg angle
Φ = Surface energy
η = overpotential
φ = Electrostatic potential at the interface of double layer and electrolyte
δ = Nernst diffusion layer, concentration boundary layer
δp = Pulse boundary layer
γ = Duty cycle of a pulse waveform
170