Jpn. J. Appl. Phys. Vol. 40 (2001) pp.

2650–2657
Part 1, No. 4B, April 2001

2001
c The Japan Society of Applied Physics

Copper Electrodeposition: Principles and Recent Progress

Jonathan R EID∗
Novellus Systems, Portland Technology Center, Wilsonville, OR 97070, USA
(Received August 31, 2000; revised manuscript received October 20, 2000; accepted for publication October 25, 2000)

The capability of copper electroplating to produce void free filling of sub-micron high aspect ratio features has made it the
process of choice for copper interconnect formation. Several aspects of copper electrodeposition including the basic electro-
chemistry and electrochemical kinetics, mass transport phenomena, potential gradients in solution, electrolyte composition,
and the influence of various organic additives have been studied for over 50 years. Much of this basic understanding can be
applied to development of integrated circuit (IC) copper electroplating processes. Other aspects of copper electroplating are
unique to IC applications. These include the interactions of very thin seed layers with the electroplating process, the basic
“bottom-up” filling mechanism necessary for seam free filling, and the metallurgical properties of sub-micron scale deposits.
The dependence of IC filling processes upon plating bath chemistry and polarization characteristics are discussed in this paper.
KEYWORDS: copper, electrodeposition, semiconductor, plating, electrochemistry

1. Introduction trolytes therefor reduce the tendency for a higher deposition

rate near the edge of a wafer which occurs during deposi-
1.1 Process chemistry
tion on resistive seed layers electrically contacted at the wafer
Copper electroplating baths used in integrated circuit (IC)
edge.5) Low acid electrolytes have been used for application
applications are formulated using a stable electrolyte solu-
such as copper foil electroplating where part geometry is not
tion containing copper sulfate and sulfuric acid. The kinet-
complex and substrate resistance is high. Selection of an acid
ics of Cu electrodeposition from this electrolyte have been
concentration for wafer applications must balance the benefit
studied for over 50 years and are well documented.1) In order
of higher acid on cell geometry related variations with the re-
to control deposit metallurgy and influence the geometric de-
duction of seed layer impact on thickness distribution which
pendence of deposition kinetics on the wafer surface organic
low acid allows. Wetting performance and oxide dissolution
additives are added to these electrolytes. The relationships
behavior must also be considered.
of deposition kinetics and deposit metallurgy to the concen-
Organic additives added to copper electroplating baths fall
trations and nature of organic additives found in commercial
into three categories.6–8) Accelerators are mercapto contain-
plating baths is complex and not fully understood.
ing species which locally accelerate current at a given voltage
Copper sulfate which is added to the plating electrolyte dis-
where they adsorb on the Cu surface during electroplating.
solves and results in free hydrated cupric ion. This ion is re-
Accelerators9) are usually present in the plating bath in the
duced to copper metal during electroplating. Typical cupric
concentration range of 1 to 25 parts per million. Suppressors
ion concentrations in commercial use today range from 10
are polymers such as polyethylene glycol which tend to form
to about 60 grams/liter. High copper concentration are useful
a current suppressing film on the entire wafer surface, espe-
when a rapid deposition process is required to achieve desired
cially in the presence of chloride ion (which can be considered
tool through-put. Lower copper concentrations result in a uni-
as a co-suppressor). Suppressors10, 11) are usually present in
formly increased charge transfer resistance at all points on the
the plating bath at high concentrations (100–2000 ppm) and
wafer surface. These solutions will therefor minimize kinetic
are not readily consumed. This results in a concentration at
variations and improve deposit thickness distributions across
the interface which is not strongly dependent on their rate
the wafer.2) In IC applications, copper sulfate concentration
of mass transfer or diffusion to the surface. Levelers are a
must also be adequate to avoid depletion of cupric ion within
second class of current suppressing molecules. Levelers7, 12)
high aspect ratio features at deposition rates adequate for high
are usually added to the plating bath at a low concentration
speed processing.
so that leveler concentration at the interface is mass transfer
Sulfuric acid is usually added to the plating electrolyte
dependent, especially if the molecule is consumed or incor-
(10–325 g/L) to increase solution conductivity and improve
porated in the deposit. Using levelers, currents at isolated
wetting or oxide dissolution on seed surfaces.1, 3, 4) In gen-
mass transfer locations such as the inside of a via are less sup-
eral, more conductive solutions minimize variation of poten-
pressed, while protruding surfaces or corners to which mass
tial gradients within the plating solution, and thus result in
transfer by diffusion or migration is more efficient are more
more uniform and geometry independent interfacial kinetics.
suppressed.
As a result, these electrolytes are used for applications such as
Under typical plating conditions, current behavior in cop-
very high aspect ratio (20 : 1) printed circuit board through-
per deposition in electroplating baths are described by a sim-
hole plating. As acid concentration is decreased the resistance
plified form of the current-overpotential equation:13)
of the electrolyte increases significantly. Electrodeposition
using higher resistance electrolytes tends to diminish the ef- i = i o {C o (0, t)/C 0 }e(−αn F(E−Eeq )/RT ) , (1)
fect of other resistances (seed layer for example) in the elec-
while at high applied voltages the limiting current (i l ) is de-
troplating system on thickness distribution. Low acid elec-
scribed by the equation:

∗ E-mail
i l = n Fm o C o . (2)
address: jon.reid@novellus.com

2650
 Jpn. J. Appl. Phys. Vol. 40 (2001) Pt. 1, No. 4B J. R EID 2651

With no potential applied to the copper seed on the wafer sur-

face the seed will remain at an equilibrium potential (E eq )
relative to a reference electrode. This potential is defined
by the fundamental reduction/oxidation properties of copper
in the electrolyte and the concentration of cupric ion in the
electrolyte. When an external negative potential (E) is ap-
plied to the copper surface, current will flow and cupric ions
will begin to deposit on the seed. The current flow will be
proportional to the exchange current (i o ), which is a mea-
sure of the equilibrium reduction/oxidation kinetics of the
system, and which determines the sensitivity of current flow
to applied potential. A system with large exchange current
will support a relatively large current at a given applied po-
tential. Current flow is also sensitive the concentration of
cupric ion at the metal interface (C o(0,t) ) relative to the bulk
concentration (C o ) at a given time. When the concentra-
Fig. 1. Polarization behavior of copper electrolyte with catalytic additive
tion of metal ion at the interface decreases as the result of only and with complete additive system. Plot shows possible current den-
a lack of mass transfer there will be a linearly related de- sity differences between wafer areas (feature base vs field) due to differ-
crease in current at a given applied potential. Most impor- ences in additive adsorption.
tant, there is an exponential kinetic relationship between ap-
plied potential and current over the range of practical inter-
est. This relationship includes the constants R(gas constant), drops are also possible.14)
T (temperature), n(electrons/mole), F(Coulombs/mole), and Bottom-up fill acceleration begins when preferential ad-
the factor α which measures the relative sensitivity reduction sorption of accelerating additives takes place near the base
and oxidation processes to potential change. This behavior of IC features. It can easily be shown that voltage drops be-
is a deviation from linear I –E relations described by Ohm’s tween the field and the base of any feature will not impact
law. At very high applied potentials, eq. (2) applies and the current flow.2, 15) It is thus possible to evaluate filling accelera-
current flow becomes proportional to the rate at which metal tion capability of a plating bath by comparing current–voltage
ion is delivered to the metal surface by convection, migration, response curves for a complete additive system (matching ad-
and diffusion as quantified by the mass transfer coefficient sorption at the surface of the wafer) with those obtained in
(m o ) and the concentration of metal ion in solution. the presence of accelerating additives only (matching adsorp-
Based on these equations, the kinetics at a single point on tion at the feature base). Figure 1 shows polarization curves
the wafer surface are described as a function of potential, ad- measured at a rotating disk electrode in a copper plating elec-
ditive adsorption conditions, and interfacial cupric ion con- trolyte containing a complete additive system, and for the
centration. In real wafer deposition systems, the interfacial same electrolyte containing only accelerating additives.8) In
potential, adsorption of additives, and cupric ion concentra- this example, the current at a given potential in the system
tion usually vary slightly across a wafer surface, and between containing accelerator only is about 7× that measured in the
the surface and the base of a high aspect ratio feature. This presence of the complete additive system. This current dif-
kinetic variability leads to within-wafer thickness distribu- ference can be interpreted as the possible bottom-up fill ca-
tion variations on the order of 1–2% sigma in well designed pability of a chemistry which develops adsorption properties
equipment, as well as to the desired phenomena of bottom-up suppressing growth at the wafer surface while accelerating
growth where growth rates are accelerated to at least 10× growth within features.
within small features.
In order to achieve uniformity’s of 1%, mass transfer and 1.2 Seed layer impacts
interfacial potential characteristics at all points on the wafer Properties of damascene features determined by the phys-
surface must be relatively uniform. Based on eq. (1), uniform ical vapor deposition (PVD) seed layer which impact plating
mass transfer characteristics can be expected to lead to uni- process performance include feature shape, continuity or ag-
form cupric ion and additive adsorption characteristics across glomeration of the seed, and degree of seed oxidation. While
the wafer. The degree of mass transfer uniformity required is the plating process can achieve highly accelerated bottom-up
achieved through proper equipment design. filling, a feature shape which is very necked at the opening
Interfacial potential uniformity across the wafer surface is is difficult to fill because it will plate shut leaving a “center
influenced by both the voltage drops in solution between the void” owing to a finite deposition rate on the sidewall.16) An
wafer and the anode, and by the voltage drop across the resis- example is shown in Fig. 2.
tive wafer seed layer. Takahashi2) showed that by increasing When PVD films become very thin within high aspect ra-
the interfacial voltage drop associated with the charge trans- tio features there is a tendency to form discontinuous or ag-
fer reaction, the effect of seed resistance and voltage drops in glomerated Cu layers.17, 18) When this takes place the plating
solution are reduced to negligible levels. Other approaches in- process does not initiate at a uniform rate on all surfaces. Typ-
clude use of highly resistive electrolyte solutions which mini- ically, this results in accelerated growth beginning above the
mize seed layer effects at the expense of increasing sensitivity base of high aspect ratio features and leaves large voids “bot-
of voltage drops in solution to cell geometry. Numerous hard- tom voids” at the base of the via as shown in Fig. 3(a). When
ware solutions to generation of uniform interfacial potential
2652 Jpn. J. Appl. Phys. Vol. 40 (2001) Pt. 1, No. 4B J. R EID

larization behavior of the electrolyte in the presence of in-

dividual additive components, and hysteresis of current flow
between anodic and cathodic potential sweeps.
Electroplating chemistries initially used for IC applications
were designed to provide ductile deposits with a uniform
thickness distribution across large surfaces and within high
aspect ratio PWB holes. These solutions typically include an
accelerator, a suppressor, and one or more levelers. Based
on the criteria used to select chemicals for these solutions, a
slight enhancement of filling in high aspect ratio (AR) fea-
tures might be expected to result from mass transfer behav-
ior of the leveler species. As reported previously,19) these
baths did show some ability to fill moderate aspect ratio fea-
tures. The success of more recently developed additive sys-
tems which do not contain levelers suggest the fill mechanism
by which the PWB era solutions enable bottom-up fill may not
be exclusively related to leveler behavior.8, 18)
Figure 4(a) shows polarization behavior of electrolyte con-
taining cupric ions, sulfuric acid, and chloride. Current is
measured as potential is first swept negatively from the equi-
librium value of +50 mV vs SCE to −280 mV vs SCE, and
Fig. 2. Seed profile of a trench feature showing pinch off near the opening
following PVD seed deposition and fill the result following plating.
then cycled back to the equilibrium potential. Results are
shown for 3 electrode rotation rates corresponding to different
rates of mass transfer. In this simple system, current increases
in an approximately exponential manner until mass transfer
growth accelerated growth begins at the base of the feature
begins to limit overall current flow. At 100 rotations per
full fill is achieved as shown in Fig. 3(b).
minute (RPM), mass transfer limitation play a role in overall
Finally, the degree of oxidation of the seed layer can play
current magnitude over much of the practical current density
a role in both filling and general deposit quality formed dur-
range (10–50 mA/cm2 ), while at 1500 RPM mass transfer ef-
ing copper deposition. While some degree of oxidation may
fects on current flow are minimal until current densities above
enhance wetting in acidic solutions owing to the high solubil-
50 mA/cm2 are reached. It should be noted that the current at
ity of oxide in the acid, excessive oxidation (>∼ 20 A) can
any given potential during the cathodic potential sweep is es-
both entirely consume metallic seed within high aspect ra-
sentially equal to the current at the same potential during the
tio features18) and result in non-uniform wetting on the wafer
anodic potential sweep. This lack of hysteresis between the
surface.
potential cycles indicates rapidly equilibrating surface mor-
phology and growth characteristics.
2. Experimental
Polarization curves of the same electrolyte following addi-
Polarization curves were measured by scanning at a rate of tion of a PWB type additive containing accelerator, leveler,
2 mV/second starting at the equilibrium potential of copper and suppressor species are shown in Fig. 4(b). Two major po-
in the electrolyte. A Pt rotating disk electrode with a surface larization differences following additive addition are noted.
area of 0.166 cm2 was prepared for each cycle by anodic dis- First, currents are sharply reduced at a given voltage. This
solution of copper leaving a metallic Pt surface on which each behavior reflects blockage of growth sites, and/or film forma-
cycle was started. Potentials were controlled using a Prince- tion which reduces mass transfer to the metal interface. Sec-
ton Applied Research 273 potentiostat and measured relative ond, there is a hysteresis behavior in the current flow between
to a saturated calomel electrode (SCE). Polarization and elec- the cathodic and anodic potential sweeps. In this system (at
trodeposition studies were carried out in an electrolyte con- 2 mV/s current sweep rate), the current flow at a given voltage
taining 175 g/L sulfuric acid, 17.5 g/L cupric ion, and 50 mg/L reflects previous growth history; currents at a given voltage
chloride ion except as specified. Electrodeposition for IC fea- during the cathodic sweep are lower than those observed dur-
ture fill evaluation was performed on full 8 inch wafers plated ing the subsequent anodic sweep. This behavior implies that
using Novellus Sabre equipment. once a high current begins at a given location on the wafer
surface it will tend to continue.
3. Results and Discussion Trench fill results obtained by plating in additive free and
PWB additive containing solutions are shown in Figs. 5(a)
3.1 Plating chemistry polarization behavior and associated
and 5(b) respectively. The trench fill results provide a good
filling capability
means to evaluate bottom-up fill acceleration capability of a
In this section, the polarization behavior of a printed wiring
plating chemistry on a well seeded surface. Without additives
board (PWB) era copper plating chemistry is compared to
present (Fig. 5(a)), the copper deposition rate on all surfaces
polarization behavior of an IC plating chemistry with much
within the feature is approximately equal. This leads to con-
greater bottom-up fill capability. Characteristics of the polar-
formal growth and a large center void extending much of the
ization behavior which may relate to filling performance are
length of the 5 : 1 aspect ratio trench. This result is observed
considered. These include overall current suppression, po-
over a current density range of approximately 5–50 mA/cm2 .
 Jpn. J. Appl. Phys. Vol. 40 (2001) Pt. 1, No. 4B J. R EID 2653

Fig. 3. (a.) Cross sectional via images showing (left to right) agglomerated seed coverage, metal profile following partial fill by elec-
troplating, and final fill result following electroplating. (b.) Cross sectional via images showing (left to right) smooth seed coverage
only, metal profile following partial fill by electroplating, and complete fill following electroplating.

Fig. 4. Polarization curves measured at 100, 500, and 1500 RPM in (a.) copper plating electrolyte containing copper sulfate, sulfuric
acid, and chloride ion and (b.) measured following addition of 6 ml/L catalytic additive and 2 ml/L suppressor/leveler PWB additive
to the base electrolyte. All anodic and cathodic scans were taken at 2 mV/s.

When PWB additives are included in the electrolyte com-

plete filling of the 5 : 1 AR trenches is achieved, thus demon-
strating the existence of bottom-up fill. A small center void
is, however, observed following plating of 6 : 1 aspect ratio,
0.17 µm trenches as shown in Fig. 5(b). While fill achieved
with this type of bath has been adequate for 0.18 µm genera-
tion device production,8) elimination of center voids in higher
aspect ratio structures than approximately 5 : 1 requires plat-
ing chemistries designed for IC applications.
A second aspect of fill performance is the ability to nucleate
and grow copper on marginal seed coverage, and thus dimin-
ish the bottom voids of the type shown in Fig. 3(a). This is
Fig. 5. Fill results for a 5 : 1 AR, 0.25 µm trench (a.) following elec-
best tested using via structures since PVD coverage within a
trodeposition in an electrolyte containing copper sulfate, sulfuric acid, and via is approximately 50% of that within a trench of the same
chloride ion with no additives present. Fill for a 6 : 1 AR, 0.17 µm trench aspect ratio. Figure 6 shows fill in a 0.21 × 0.9 µm via plated
(b.) following addition of 6 ml/L catalytic additive and 2 ml/L suppres- in a PWB chemistry. Large bottom voids, but no center voids,
sor/leveler PWB additive to the base electrolyte.
are evident in all vias. This result is believed to reflect a lack
of copper electrodeposition near the via base at the beginning
of the filling process in conjunction with more rapid growth
At current densities above approximately 50 mA/cm2 , deple- along the better seeded sidewalls higher in the via. Finally,
tion of cupric ion within the feature further limits growth growth acceleration begins on the well seeded sidewalls re-
within the feature and void volume increases.
2654 Jpn. J. Appl. Phys. Vol. 40 (2001) Pt. 1, No. 4B J. R EID

Fig. 6. Fill results for 0.21 × 0.9 µm vias following electrodeposition in a

PWB chemistry containing 6 ml/L catalytic additive and 2 ml/L suppres-
sor/leveler in the base electrolyte.

sulting in bottom-up fill beginning somewhat above the via

base and leaving a void at the base.
In Fig. 7(a), polarization curves measured at nomi-
nal working concentrations of a two component (accelera-
tor/suppressor) IC additive are shown. During the initial
cathodic potential sweep, the currents measured using this
chemistry are significantly lower than those measured using
the PWB chemistry. For example, at −250 mV the current
flow in this bath is approximately 8 mA/cm2 compared to
15–20 mA/cm2 in the PWB bath. This polarization behavior
difference means that higher voltages must be applied using
this chemistry in order to achieve a given current flow. Higher
voltages are well known to be more effective in achieving uni-
form nucleation characteristics on variable surfaces. Since
the higher level of current suppression will always apply to
growth on the wafer field, a more polarizing electrolyte can
be considered useful in achieving growth rate differentiation
between the field and surfaces within a feature which may be-
come depolarized. A second difference between the two baths
is the extremely large hysteresis in current flow at a given po-
tential between the cathodic and anodic sweeps observed for
the IC bath. This difference implies that localized high cur-
rents will propagate more efficiently in the IC bath.
Figure 7(b) shows polarization curves of the electrolyte
measured following addition of only the suppressor compo-
Fig. 7. Polarization curves measured at 50, 300, and 1000 RPM in a cop-
nent at nominal concentration. As anticipated, currents in
per plating solution containing copper sulfate, sulfuric acid, and chloride
the presence of the suppressor additive component alone are ion, (a.) with addition catalytic additive and suppressor at nominal concen-
much lower than observed using the complete additive. Like- trations, (b.) with addition a polymer like suppressor additive at nominal
wise (Fig. 7(c)), currents in the presence of the accelerator concentration, and (c.) with addition of catalytic additive at nominal con-
centration, (All anodic and cathodic scans were taken at 2 mV/s).
component alone are much higher at a given potential than
observed using the complete additive. At a given potential up
to approximately −220 mV, the current density in the accel-
erator only case is at least 100× that in the suppressor only started for any reason on the wafer.
case. An ideal additive, in which only suppressor molecules Figure 8 shows trench and via fill results measured using
adsorbed on the field surface of the wafer and only accelera- the two component additive system at concentrations corre-
tor molecules adsorbed at the feature base, would allow this sponding to Fig. 7(a). As can be seen in Fig. 8(a), 0.12 µm
degree of fill acceleration. This degree of acceleration has not trenches with AR’s up to 10 : 1 are fully filled. The fill-
been observed, probably because all additive components in ing capability of the IC chemistry with highly suppressed
existing additive systems adsorb on the field surface of the polarization behavior and significant hysteresis between an-
wafer (see Fig. 1). It should also be noted that electrolytes odic and cathodic potential scans is clearly superior to that
containing only accelerator, or only suppressor, show little observed using the PWB chemistry. In Fig. 8(b), it is seen
or no hysteresis of the polarization curves between anodic that 0.21 × 0.9 µm vias are fully filled. The lack of bottom
and cathodic potential sweeps. These solutions would not voids (which were observed after deposition using the PWB
be expected to propagate localized accelerated growth which chemistry) is believed to result from the more effective or uni-
 Jpn. J. Appl. Phys. Vol. 40 (2001) Pt. 1, No. 4B J. R EID 2655

Fig. 8. Fill results for 0.12 µm, 10 : 1 AR trenches (a.), and 0.21 µm, 4.5 : 1 vias (b.), and multi-size vias (c.) following deposition in a
copper plating solution containing copper sulfate, sulfuric acid, and chloride ion along with catalytic and suppressor IC additives at
nominal concentrations.

form nucleation20) resulting from the more polarized condi-

tions using the IC additives. Figure 8(c) shows that bottom
voids again appear at higher aspect ratio’s using the IC chem-
istry.

3.2 Fill response to plating chemistry

The following outline of points can be made in summariz-
ing observations in working with both PWB era and IC copper
plating bath additive systems in standard sulfuric acid elec-
trolytes.8, 18)
Fig. 9. Fill results for 0.12 µm, 10 : 1 AR trenches (a.), and multi-size vias
1. Strongly accelerated fill has not been achieved using sup- (b.) following deposition in a low acid copper plating solution contain-
pressing polymers alone or with chloride ion. Growth in these ing copper sulfate and chloride ion along with catalytic and suppressor IC
baths appears to be largely conformal. additives at optimized concentrations.
2. Suppressing polymer concentration can be increased to at
least 4× beyond an optimal (plateau) level without impacting
fill. These concentrations can extend upwards of 1000 mg/L,
well beyond a level at which concentration gradients within
features are likely to develop.
3. Bottom-up filling can be achieved, and is generally most
pronounced, in the absence of levelers.
4. The addition of levelers to bottom-up filling chemistries
often results in top center voids.
5. As chloride ion concentration is increased from zero, the
filling acceleration in vias increases to a maximum value then
decreases as chloride concentration is further increased. Con- Fig. 10. Fill results for 0.25 µm, 6 : 1 AR trenches (a.), and multi-size
vias (b.) following deposition in an electrolyte containing cop-
formal fill has been noted at very high chloride levels. per-propanesulfonate, propanesulfonic acid, and chloride ion along with
6. As accelerator concentration is increased, bottom-up fill catalytic and suppressor IC additives at concentrations optimized for this
increases from near zero to a maximum rate and then dimin- electrolyte.
ishes to conformal behavior. This is similar to the trend for
chloride ion (point 5).
7. For a given additive system, too low a copper concentra- Chemicals, Philadelphia Pa.
tion diminishes bottom-up fill capability in high aspect ratio Figure 9 shows via and trench fill results obtained using a
features. low acid (pH 2) sulfate based electrolyte. Trench fill is equiv-
8. Accelerated growth in electrolytes containing a two com- alent to performance measured in standard electrolyte with
ponent (accelerator/suppressor) additive continues over a features as small as 0.12 × 1.2 microns showing complete fill.
damascene feature following bottom-up fill. The addition of Vias smaller than 0.24 × 1.2 microns exhibit bottom voids
a leveler component serves to suppress current on the rapidly while larger features show complete fill. Via fill results show
growing surface, once it protrudes above the field, thereby the same or possibly a greater degree of bottom voids than ob-
leading to a relatively uniform deposit thickness. served using high acid electrolytes despite the lower expected
dissolution rate of copper in this electrolyte. This result sug-
3.3 Fill behavior in alternate electrolytes gests that initial seed properties, rather than dissolution of the
Fill performance has been evaluated in a variety of alter- seed in the plating bath, are most essential to bottom void
nate electrolytes. These include copper sulfate solutions at elimination.
near-neutral pH’s and mercaptopropanesulfonic acid/copper Figure 10 shows trench and via fill using a propane sul-
propanesulfonate based electrolytes supplied by ATOFINA
2656 Jpn. J. Appl. Phys. Vol. 40 (2001) Pt. 1, No. 4B J. R EID

Fig. 11. Fill evolution profiles for various feature sizes and nominal deposit thickness’ as generated in an IC plating bath using optimized
accelerator and suppressor concentrations.

fonic acid/copper propane sulfonate electrolyte. As is seen in After electrodeposition of 80 nm of Cu, the total Cu thick-
Fig. 10, complete trench fill was achieved in features as small ness on the field is approximately 230 nm including the
as 0.25 × 1.5 microns, and strongly accelerated bottom-up fill 150 nm seed layer. At this point in the filling sequence, the
is evident from the protrusions over the trenches. Further op- largest features (0.9 and 1.7 µm) show conformal growth of
timization of process conditions and additive concentrations the 80 nm layer on all surfaces within the trench. The next
in this electrolyte appears likely to yield trench fill results smallest trench size, 0.6 µm, exhibits slight fill acceleration in
equivalent to sulfuric acid based baths. Via fill results were the feature corners as evidenced by the thicker deposit in that
not substantially different than those obtained in copper sul- location. At 0.3 µm, the sidewalls in the upper portion of the
fate based electrolytes with bottom voids being observed in trench show conformal growth, however, a distinct surface of
0.24 × 1.2 micron features and larger vias showing complete copper growing parallel to the base of the trench has formed
fill. The similarity in via bottom void formation observed us- slightly above the base of the trench. This surface is charac-
ing organic and non-organic anion based electrolytes, despite teristic of accelerated bottom-up growth. Once formed, this
the expected wetting differences, suggest wetting limitations surface moves rapidly up the feature yielding complete fill as
of the electrolyte do not contribute strongly to poor electrode- has taken place in the 0.15 µm trenches where the growth sur-
position behavior on discontinuous seed. face is approximately level with the dielectric plane.
No remarkable differences in trench or via fill performance After electrodeposition of 130 nm, slight fill acceleration in
were achieved among the electrolytes tested, although the the corners of 0.9 and 1.7 µm trenches has become evident.
concentrations of organic additive which gave optimum per- The 0.6 µm trenches now show a flat growth surface which
formance varied as the electrolyte is changed. The universal has evolved from the initially accelerated corner growth seen
observation of bottom voids in small vias suggests the need after the 80 nm deposition. At the 0.3 µm trench size, the
exists for improved PVD coverage, or seed repair methods, flat growth surface noted at 80 nm nominal growth has pro-
which do not increase the effective aspect ratio of features ceeded above the field by approximately 100 nm. Note that
prior to the final filling step. This will become increasingly the vertical growth of the bottom-up fill surface in the 0.3 µm
important as thinner seed layers are required for future device trench, between 80 and 130 nm nominal deposit thickness,
generations. is approximately 700 nm. This is a 14× acceleration rela-
tive to the nominal thickness change (50 nm) and suggests an
3.4 Fill evolution characteristics approximately 140 mA/cm2 current density on the localized
Figure 11 shows fill evolution for trench sizes of 0.15, 0.3, bottom-up growth surface.
0.6, 0.9 and 1.7 µm following electrodeposition of 80, 130, After 195 nm deposition, more pronounced corner growth
195 and 240 nm of copper on the wafer field surface at a nom- is noted in the 0.9 and 1.7 µm trenches and fill of the
inal current density of approximately 10 mA/cm2 . 0.6 µm trench is nearly complete. The vertical growth of the
 Jpn. J. Appl. Phys. Vol. 40 (2001) Pt. 1, No. 4B J. R EID 2657

bottom-up fill surface in the 0.6 µm trench, between 130 and adsorbing accelerating additives accumulate within high as-
195 nm nominal deposit thickness, is approximately 550 nm. pect ratio features and displace adsorbed suppressor. This
This is an 8.5× acceleration relative to the nominal growth accumulation initially takes place at the feature base corners
rate. where surface area changes most significantly. Highly polar-
After 240 nm deposition, significant corner growth is noted ized electrolytes which maintain strong current suppression
in the 1.7 µm trenches. The 0.9 µm trenches have under- on the wafer surface and aid in uniform nucleation within fea-
gone the transition from accelerated corner growth to rapid tures are beneficial to high AR feature filling. The mechanism
bottom-up fill to form the characteristic flat growth surface. and limitations of bottom-up filling appear to be consistent in
Acceleration of growth relative to the field in the 0.9 trench different electrolytes.
size is approximately 8× between 195 and 240 nm nominal
deposit growth.
1) F. A. Lowenheim: Modern Electroplating (Wiley and Sons, New York,
Further deposition time results in similar transitions to 1974).
rapid bottom up growth fronts in features as large as approx- 2) K. Takashi: J. Electrochem. Soc. 147 (2000) 1414.
imately 3.0 µm. Exact delay time, or nominal deposit thick- 3) H. Brown and R. Fellows: US patent 2882209, Apr. 14, 1959.
ness, required for the transition to accelerated bottom-up fill 4) O. Kardos, D. Arcelesi and S. Valayil: US patent 3956120, May 11,
1976.
varies with nominal current density for all feature sizes. Very 5) E. Broadbent, E. J. McInerney, L. Gochberg and R. J. Jackson: J. Vac.
low current densities never result in transition to bottom-up Sci. & Technol. B 17 (1999) 2594.
fill, and currents above an optimum value generally increase 6) D. R. Turner and G. R. Johnson: J. Electrochem. Soc. 109 (1962) 798.
the nominal thickness required for transition. 7) L. Mirkova, S. Rashkov and C. Nanev: Surf. Technol. 15 (1982) 181.
8) J. Reid and S. Mayer: Advanced Metallization Conf. 1999 Proc. MRS,
The feature size dependence of initiation of bottom-up fill Warrendale, PA, p. 53.
acceleration and the observation of initially accelerated cor- 9) A. Steponavicius, S. Lichusina, A. Surviliene and A. Sudavicius:
ner growth are consistent with the requirement of a threshold Chemija 3 (1997) 64.
surface area change within the feature prior to fill accelera- 10) M. Hill and G. Rogers: J. Electroanal. Chem. 86 (1978) 179.
11) J. Kelly and A. West: J. Electrochem. Soc. 145 (1998) 3472.
tion. During conformal filling, localized concentration of the 12) J. Kelly, C. Tian and A. West: J. Electrochem. Soc. 146 (1999) 2540.
strongly adsorbing accelerating species takes place. Concen- 13) A. J. Bard and L. R. Faulkner: Electrochemical Methods (Wiley and
tration is initially most pronounced in the base corners of the Sons, New York, 1980).
trenches where surface area change is also initially signifi- 14) H. Mori: US Patent 5443707, Aug. 22, 1995.
15) S. Mayer, R. Contonlini, V. Bhaskaran, R. J. Jackson, J. Reid, J. Martin,
cant. In features between 0.3 and 0.9 µm in width, the total D. Morrissey and R. Schetty: Electrochem. Soc. Proc. 99–31 (2000)
surface area within the trenches has decreased by 40–50% at 174.
the point when rapid growth acceleration begins. 16) J. Reid, V. Bhaskaran, R. Contonlini, E. Patton, R. J. Jackson, E.
Broadbent, T. Walsh, S. Mayer, J. Martin, D. Morrissey, R. Schetty and
S. Menard: Proc. of IITC 1999 (1999) p. 284.
4. Conclusions 17) J. Hartman, H. Yeh, H. Atwater and I. Hashim: Mater. Res. Soc. Proc.
564 (1999) 257.
Based on studies of polarization behavior, fill evolution 18) J. Reid, S. Mayer, E. Broadbent, E. Klawuhn and K. Ashtiani: Solid
characteristics, fill in multiple electrolytes, and fill depen- State Technol. (2000) p. 86.
dence on additive chemistry several conclusion can be drawn 19) P. C. Andricacos, C. Uzoh, J. O. Dukovic, J. Horkans and H. Deligianni:
regarding requirements for bottom-up fill. Fill is most readily IBM J. Res. Dev. 42 (1998) 567.
20) V. Isaev: J. Electroanal. Chem. 453 (1998) 25.
achieved using additives containing a polymer suppressor and
a mercapto accelerator. Bottom-up fill initiates when strongly