T. WAECHTLER, ET AL., J. ELECTROCHEM. SOC., VOL. 156, NO. 6, H453-H459 (2009) – POSTPRINT OF THE ORIGINAL ARTICLE 1
Copper Oxide Films Grown by Atomic Layer
Deposition from Bis(tri-n-butylphosphane)-
copper(I)acetylacetonate on Ta, TaN, Ru, and SiO
2
Thomas Waechtler
∗k
, Steffen Oswald
†
, Nina Roth
‡
, Alexander Jakob
‡
, Heinrich Lang
‡
, Ramona Ecke
∗
,
Stefan E. Schulz
∗§
, Thomas Gessner
∗§∗∗
, Anastasia Moskvinova
¶
, Steffen Schulze
¶
, and Michael Hietschold
¶
∗
Center for Mircotechnologies (ZfM), Chemnitz University of Technology, D-09107 Chemnitz, Germany
†
Leibniz Institute for Solid State and Materials Research (IFW), D-01069 Dresden, Germany
‡
Department of Inorganic Chemistry, Institute of Chemistry,
Chemnitz University of Technology, D-09107 Chemnitz, Germany
§
Fraunhofer Research Institution for Electronic Nano Systems (ENAS), D-09126 Chemnitz, Germany
¶
Solid Surfaces Analysis and Electron Microscopy Group, Institute of Physics,
Chemnitz University of Technology, D-09107 Chemnitz, Germany
k
E-mail: thomas.waechtler@zfm.tu-chemnitz.de
∗∗
Electrochemical Society Active Member.
c
The Electrochemical Society, Inc. 2009. All rights reserved. Except as provided under U.S. copyright law, this work may
not be reproduced, resold, distributed, or modified without the express permission of The Electrochemical Society (ECS).
The archival version of this work was published in Journal of The Electrochemical Society, Vol. 156, No. 6, pp. H453–H459
(2009), ISSN 0013-4651, Digital Object Identifier (DOI): 10.1149/1.3110842 (http://dx.doi.org/10.1149/1.3110842).
Abstract— The thermal atomic layer deposition (ALD)
of copper oxide films from the non-fluorinated yet liq-
uid precursor bis(tri-n-butylphosphane)copper(I)acetylacetonate,
[(
n
Bu
3
P)
2
Cu(acac)], and wet O
2
on Ta, TaN, Ru and SiO
2
substrates at temperatures of < 160
◦
C is reported. Typical
temperature-independent growth was observed at least up to
125
◦
C with a growth-per-cycle of ∼ 0.1 Å for the metallic sub-
strates and an ALD window extending down to 100
◦
C for Ru.
On SiO
2
and TaN the ALD window was observed between 110
and 125
◦
C, with saturated growth shown on TaN still at 135
◦
C.
Precursor self-decomposition in a chemical vapor deposition
mode led to bimodal growth on Ta, resulting in the parallel
formation of continuous films and isolated clusters. This effect
was not observed on TaN up to about 130
◦
C and neither on Ru or
SiO
2
for any processing temperature. The degree of nitridation
of the tantalum nitride underlayers considerably influenced the
film growth. With excellent adhesion of the ALD films on all
substrates studied, the results are a promising basis for Cu
seed layer ALD applicable to electrochemical Cu metallization
in interconnects of ultralarge-scale integrated circuits.
I. INTRODUCTION
W
HILE copper is now widely accepted as conductor
material for interconnect systems of ultralarge-scale
integrated circuitry (ULSI) [1], there is an increasing need
for extremely thin, continuous Cu films that are conformal
in demanding nanoscale geometries. The conventional dama-
scene technology applied for creating the metallization system
of ULSI devices requires Cu seed layers as starting point for
electrochemical Cu filling (ECD) of vias and conductor lines
embedded in patterned SiO
2
or low-k / ultra low-k material [1],
[2]. So far sputtering is the method of choice for depositing
such seed layers, but with reduced line width and increasing
aspect ratio of the interconnect features this method tends to
fail due to its inherent non-conformal deposition characteristic
as schematically shown in Fig. 1.
Gas-phase chemical deposition methods, such as chemical
vapor deposition (CVD) and atomic layer deposition (ALD),
are being discussed as alternative methods in this respect [3]–
[6]. ALD, in particular, is a viable technique to obtain contin-
T. WAECHTLER, ET AL., J. ELECTROCHEM. SOC., VOL. 156, NO. 6, H453-H459 (2009) – POSTPRINT OF THE ORIGINAL ARTICLE 2
Cu
Dielectric cap
SiO
2
or ULK
Barrier
PVD Seed
ECD
Cu
Void
Lower
interconnect
level
Via level
Upper
interconnect
level
Fig. 1. Schematic of Cu ULSI dual damascene metallization: Especially in
high-aspect-ratio features, such as vias, the sputtered PVD seed layer is prone
to nonconformalities (left), which may result in the formation of voids during
the subsequent electrochemical Cu deposition (right).
uous, ultrathin films with very good step coverage in extreme
geometries and on three-dimensional nanoscale objects due to
the self-limiting film growth behavior [7], [8].
As ALD relies on the reaction of a chemisorbed monolayer
of a metal precursor with a co-reactant to form the desired
material, a suitable precursor combination must be found.
In contrast to ALD processes of noble metals, where also
oxidizing agents can be used to obtain metal films [9]–[15],
metallic copper films are only accessible by ALD directly
when the Cu precursor is reduced. This can be accomplished
by plasma-enhanced ALD (PEALD) with atomic hydrogen as
the reducing agent [16]–[18]. However, these approaches dete-
riorate ALD’s ability to conformally coat complex geometries,
because especially hydrogen plasmas tend to deplete down in
nanoscale trenches and shadowed areas. As a consequence,
this leads to reduced step coverage and non-conformal film
growth, compromising a major advantage inherent to ALD.
From this point of view, thermal ALD is generally favored.
Pure thermal approaches to form metallic Cu have been
proposed with [CuCl] and H
2
as the precursors, requiring el-
evated processing temperatures between 360 and 425
◦
C [19]–
[21]. These processes yielded isolated Cu clusters only; due to
the strong agglomeration tendency Cu exhibits especially on
substrates like silica, refractory metals or refractory metal ni-
trides [22]–[27]. Therefore such processes are problematic for
applications where ultra-thin, continuous films are required. In
addition, the use of solid precursors with high melting points
such as metal halides further complicates the situation as the
precursors must be sublimed in situ, making it cumbersome
to guarantee a constant and reproducible precursor flow rate
which is indispensable for reproducible processes applicable
to mass production.
Using metal-organic precursors, in contrast, lower sublima-
tion temperature, or in many cases precursor liquid delivery or
bubbling as well as reduced processing temperatures are pos-
sible. In this respect, Cu(II) β-diketonates have been studied.
With copper(II) hexafluoroacetylacetonate, [Cu(hfac)
2
], ALD
was reported from 230
◦
C [28], [29]. Although fluorinated
precursors exhibit the advantage of greater volatility compared
to their non-fluorinated counterparts, adhesion problems of
the deposited films have been observed as a result from
fluorine-containing residues accumulating at the interface to
the substrate in Cu CVD processes from [(TMVS)Cu(hfac)]
(CupraSelect) (TMVS = trimethylvinylsilane) [30]–[32]. As
an alternative, ALD was studied using the non-fluorinated
copper(II) tetramethylheptanedionate, [Cu(thd)
2
], and direct
reduction to form Cu. However, these processes turned out
to be highly substrate-dependent, as film growth relied on
catalytic underlayers such as Pt or Pd [33], [34].
Apart from Cu(II) precursors, Cu(I) amidinates have re-
cently been reported to react with hydrogen already at tem-
peratures ≤ 200
◦
C to give copper films by ALD [35]–[37].
However, most of these precursors are solids under standard
conditions. Furthermore, smooth and continuous ALD Cu
films are only obtained on metallic nucleation layers such as
Co or Ru, while discontinuous islands grow on materials like
WN
x
[38], [39]. With respect to a practical application, such
requirements would result in rather complex processes.
To circumvent theses issues associated with direct precur-
sor reduction for growing metallic Cu films, alternatives are
being discussed with respect to simple hence cost-effective
processes. CVD and ALD approaches include the formation
of films of copper compounds such as nitrides [40]–[42] or
oxides [28], [29], [43]–[47] which are subsequently reduced.
In this respect, we here investigate the ALD of oxidic copper
films from bis(tri-n-butylphosphane)copper(I)acetylacetonate,
[(
n
Bu
3
P)
2
Cu(acac)] (
n
Bu = n-butyl) for the first time. This
liquid, non-fluorinated Cu(I) β-diketonate precursor is ac-
cessible by a straight-forward synthesis methodology from
inexpensive educts. Hence it also appears attractive for later
large-scale industrial application at reasonable cost. In our
current study, we concentrate on pure thermal ALD processes
to eliminate the drawbacks of PEALD associated with reduced
film conformality in high aspect ratio features. For similar
reasons and also in order to avoid extensive oxidation of
substrate materials as well as the use of costly equipment,
ozone-based oxidation processes are not investigated. Rather,
a mixture of water vapor and oxygen ("wet oxygen") is used
as oxidizing agent during ALD. With the target application
of seed layers for ULSI damascene metallization in mind,
ALD was carried out on tantalum as well as tantalum nitride
diffusion barrier layers. Ruthenium, as a candidate for Cu
nucleation layers [12], [48], [49] and potential component of
advanced diffusion barrier systems [50]–[55], was also used
as substrate. For comparison to the conductive substrates,
depositions were further carried out on SiO
2
.
II. EXPERIMENTAL
A. Precursor Considerations
Bis(tri-n-butylphosphane)copper(I)acetylacetonate,
[(
n
Bu
3
P)
2
Cu(acac)] (Fig. 2) was synthesized under inert
gas atmosphere by published methods [56], [57].
Under standard conditions, the precursor is a pale yellow
liquid and can be stored in inert gas atmosphere for months
without decomposition. Vapor pressure measurements show
that the precursor has a vapor pressure of 0.02 mbar at 98
◦
C
(Fig. 3), being comparable to [Cu(acac)
2
] but considerably
lower than fluorinated substances such as [(TMVS)Cu(hfac)]
or [Cu(hfac)
2
] [58]–[60]. However, for ALD a high vapor pres-
sure is not as critical as for CVD where high deposition rates
are desirable. Since ALD relies on forming a chemisorbed
monolayer of precursor molecules on a substrate, also less-
volatile liquids can be applied to ALD.
T. WAECHTLER, ET AL., J. ELECTROCHEM. SOC., VOL. 156, NO. 6, H453-H459 (2009) – POSTPRINT OF THE ORIGINAL ARTICLE 3
Fig. 2. Structure of [(
n
Bu
3
P)
2
Cu(acac)] (molecular weight: 567.3 g/mol).
0.001
0.01
0.1
1
10
2.0 2.5 3.0 3.5 4.0
1000/T (1/K)
Vapor Pressure (mbar)
(TMVS)Cu(hfac)
Cu(hfac)
Cu(acac)
( Bu P) Cu(acac)
126.9
60.2
12.6
Temperature (°C)
n
3 2
2
2
Fig. 3. Vapor pressure data for [(
n
Bu
3
P)
2
Cu(acac)] compared to other
commonly used Cu CVD/ALD precursors [58]–[60].
In a previous report [57], Cu CVD by thermal disproportion-
ation was shown at 220
◦
C with this precursor. As differential
scanning calorimetric studies (DSC) of the molecule showed
major decomposition peaks only at 237 and 255
◦
C [57], we
chose the precursor as a viable candidate for low-temperature
ALD studies.
B. ALD Experiments
ALD experiments were carried out in a cold-wall, 4 in.
single-wafer vertical flow reactor equipped with a load-lock
chamber [61], [62]. The process control system of the orig-
inal CVD reactor had been modified to enable automated,
cyclic processing [63]. The deposition chamber was evacuated
with a combination of a turbomolecular and a roots pump,
both backed with rotary pumps, to achieve a base pressure
of < 3 × 10
−6
mbar. The ALD processes themselves were
carried out at pressures between 0.8 and 1.5 mbar. During
deposition, the chamber walls were kept at 85
◦
C to avoid
precursor condensation. Each ALD cycle consisted of the steps
summarized in Table I. Due to the large reactor volume of
nearly 13 liters, relatively long pulses were necessary both to
reach saturation of the precursor chemisorption as well as to
guarantee sufficient purging of the chamber.
TABLE I
STEPS OF AN ALD CYCLE.
Step Description Pulse length (s)
1 Precursor exposure 3–5
2 Argon purging 5
3 Oxidation (oxygen + water vapor) 7–11
4 Argon purging 5
The copper precursor was stored at room temperature under
Ar atmosphere in a stainless steel stock bottle. During process-
ing it was evaporated by a Bronkhorst liquid delivery system
(LDS) with Ar as carrier gas (flow rate ∼ 700 sccm). This
approach avoids exposing the precursor stock to heat and thus
leads to a longer shelf life of the substance. After evaporating
between 85 and 100
◦
C at a flow rate of 10 to 20 mg/min and
mixing with carrier gas, the precursor vapor was transported to
the deposition chamber via heated stainless steel tubes. Water
vapor was generated by a bubbler, also using Ar as carrier gas,
at a temperature of 45 to 50
◦
C. With an Ar flow of 200 sccm,
approximately 18 to 20 mg/min H
2
O were evaporated. In
parallel to the water vapor, 90 sccm of oxygen were flown
during the oxidation pulses. The purging steps were realized by
supplying 145 sccm of Ar. The 4 in. wafers that were used as
substrates were heated resistively by a graphite heater during
the ALD. The processing temperature ranged from 100 to
155
◦
C. As starting layers, 40 nm of TaN or combinations
with Ta (Ta/TaN, i. e., 20 nm Ta on top of 20 nm TaN), the
preferred diffusion barrier system for ULSI Cu interconnects,
were sputtered onto the Si prior to the ALD processes. The
sputtering was realized in a Balzers CLC 9000 equipment. The
Ta/TaN stacks were formed in continuous processes by turning
off the N
2
flow when the desired TaN thickness was reached.
Because of this, a continuous transition was obtained from
TaN to Ta, and in some cases the Ta was also unintentionally
nitrided. Si substrates coated with 100 nm Ru on top of a
10 nm Ti adhesion layer, both prepared by evaporation, were
purchased from Advantiv Technologies, Inc., Fremont, CA
(USA) and used as received. For depositions on silica, 4 in. Si
wafers were thermally oxidized in a Centrotherm tube furnace
prior to the ALD processes. All substrates were exposed to air
before the ALD and were not pretreated in situ.
C. Sample Characterization
The surface structure and morphology of the ALD films
was studied by field-emission scanning electron microscopy
(SEM) on a LEO 982 Digital Scanning Microscope as well
as atomic force microscopy (AFM) with a Digital Instru-
ments NanoScope IIIa in tapping mode using standard silicon
tips. Spectroscopic ellipsometry with a SENTECH SE 850
ellipsometer was applied to determine the thickness of the
ALD films. Selected samples were studied in more detail
by cross-sectional transmission electron microscopy (TEM)
with respect to the film structure as well as to verify the
thickness values obtained from ellipsometry. A Philips CM20
TEM equipped with a field-emission gun was used for these
investigations. Additionally, electron energy loss spectroscopy
(EELS) and electron diffraction analyses were carried out
T. WAECHTLER, ET AL., J. ELECTROCHEM. SOC., VOL. 156, NO. 6, H453-H459 (2009) – POSTPRINT OF THE ORIGINAL ARTICLE 4
Fig. 4. (a) Top-view SEM image of a Cu/Cu
x
O ALD film on Ta grown at
135
◦
C. The film was partially etched to expose the Ta surface. The bimodal
growth characteristic has led to cluster formation in parallel to the growth of a
continuous film. (b) Cross-sectional TEM image of a perfect ALD film grown
on TaN at 125
◦
C. No clusters are visible. By spectroscopic ellipsometry, a
film thickness of 3.6 nm was determined.
on this TEM. Chemical analysis of the samples was fur-
ther realized by X-ray photoelectron spectroscopy (XPS)
on a PHI 5600 (Physical Electronics) with Al K
α
X-rays
(1486.7 eV) used for excitation. The adhesion of the films
was examined with the tape test using Tesa 4129 tape with
an adhesion force of 8 N per 25 mm.
III. RESULTS AND DISCUSSION
For the ALD experiments carried out on Ta a bimodal
growth characteristic was experienced, leading to the parallel
formation of continuous films and separated islands. In an
earlier report [64], it was already shown that when using
wet oxygen rather than only O
2
or water vapor during the
ALD much better results can be obtained, which is due to
the enhanced oxidizing activity of the H
2
O/O
2
combination.
Fig. 4 a shows a top-view SEM image of a Ta sample where
the ALD film has been partially etched away. Between the
clusters, a continuous film can be seen. On TaN, in contrast,
continuous films without any larger aggregates were obtained
up to 125
◦
C (Fig. 4 b).
Fig. 5. (a) Cross-sectional TEM image of an ALD film grown on TaN at
135
◦
C. Besides a continuous film of ∼ 5 nm thickness (ellipsometry: 4.9 nm),
clusters of a size between 15 and 20 nm are visible. (b) EELS map for Cu of
the same sample.
For ALD films grown at 135
◦
C, very small clusters in
the range of 15 to 20 nm were detected by cross-sectional
TEM imaging (Fig. 5). With increasing temperature, however,
larger clusters started to appear on TaN, too. Together with
an increase in the growth per cycle (GPC), this points to
self-decomposition of the precursor setting in. In fact, it was
shown theoretically by Machado et al. [65], [66] that especially
Ta is very reactive toward metal-organic precursors, causing
them to decompose and even form fragments of the ligands.
As that study further points out, this effect is much less
pronounced on TaN surfaces. Apparently this results in a
much better controlled ALD growth on TaN than on metallic
Ta in the current experiments. This is also expressed by the
self-saturated growth regime obtained on TaN at 135
◦
C as
reported before [64], although a slight increase in the GPC
due to beginning CVD effects is already experienced at this
temperature. The respective graph is reprinted here in Fig. 6.
The growth characteristic for this process shown in Fig. 7
displays a linear behavior with increasing number of ALD
cycles in the lower range, while for more than about 200
cycles, the data suggest a steeper increase. We believe this to
be due to a change in the surface chemistry once the substrate
gets completely covered with the growing film.
Fig. 8 displays XPS data of the Cu 2p3/2 core level and
the Cu LMM Auger transition for ALD films on TaN. The
samples were fabricated at different processing temperatures
both on stoichiometric and Ta-rich TaN. XPS analyses carried
out approximately 40 days after deposition as well as 100
T. WAECHTLER, ET AL., J. ELECTROCHEM. SOC., VOL. 156, NO. 6, H453-H459 (2009) – POSTPRINT OF THE ORIGINAL ARTICLE 5
115 °C
125 °C
135 °C
145 °C
155 °C
930935940945
0
0.2
0.4
0.6
0.8
1
Binding Energy (eV)
Normalized Intensity
Cu 2p3/2
182022242628303234
0
0.2
0.4
0.6
0.8
1
Binding Energy (eV)
Normalized Intensity
Ta 4f
Cu or
Cu
2
O
metallic Ta
bound Ta
(TaN)
(a)
(c)
560565570575580
0
0.2
0.4
0.6
0.8
1
Binding Energy (eV)
Normalized Intensity
Cu LMM
CuO
Cu(OH)
2
CuO or
Cu(OH)
2
(b)
Cu
Cu
2
O
CuO
Cu(OH)
2
ALD process temperature:
Fig. 8. XPS core level spectra of (a) Cu 2p3/2 and (c) Ta 4f as well as binding energy data obtained for the Cu LMM Auger transition (b) from ALD
samples on TaN. The processes of 400 cycles were carried out with precursor and purging steps of 5 s and 11 s oxidation pulses at 115, 125, 135, 145, and
155
◦
C.
Fig. 6. GPC as a function of precursor pulse length for ALD on TaN at
135
◦
C. Reprinted with permission from Ref. [64].
Fig. 7. Evolution of film thickness for ALD on TaN at 135
◦
C with respect
to the number of cycles, based on ALD runs according to Table I with 5 s
precursor and 11 s oxidation pulses, separated by 5 s purging pulses. (The
dotted lines are drawn to guide the eye.)
days later reveal that there is a general tendency of Cu
2
O
formation, which is in accordance with other reports where
preferably Cu(I) oxide was produced from [Cu(hfac)
2
] [45],
[67] or [Cu(acac)
2
] [46] and H
2
O or H
2
O
2
. This is shown by
the Cu 2p3/2 peak (Fig. 8 a) at about 932.4 eV [68]. However,
because this signal could also be assigned to metallic Cu [68],
[69], one has to take into account the Auger spectra for the
Cu LMM transition (Fig. 8 b) as well. There Cu would be
expected at a binding energy of 567.7 to 567.9 eV [69], [70]
while Cu
2
O should give a signal at 570 eV [68], [69]. [We
note that, in Refs. [68]–[70], the kinetic energy of the Auger
electrons is given, while in the current study we report the
binding energy, i. e., the difference of the X-ray excitation
energy and the kinetic energy of the Auger electrons.] Since no
signal to be assigned to Cu(0) is present in the Auger spectra,
we can conclude the presence of Cu(I) oxide. This is further
supported by electron diffraction studies (Fig. 9). Apart from
substrate signals originating from TaN, the diffraction reflexes
obtained are best fitted to Cu
2
O rather than CuO or Cu.
However, the XPS core level spectra in Fig. 8 a also display
signals at higher binding energies. In this respect, CuO would
be expected at 933.2 eV [68] together with a satellite between
940 and 945 eV [68] and an Auger signal for the Cu LMM
transition at 568.5 eV [69]. Apart from CuO, the Cu 2p satellite
is also typical for Cu(OH)
2
together with a major signal at
935.1 eV [69] and an Auger peak at 570.4 eV (i. e., at a similar
position as for Cu
2
O) [69]. Because the Cu 2p3/2 core level
signal as well as an Auger peak for CuO are not developed,
we can further conclude that Cu(OH)
2
is present in addition to
Cu
2
O. However, it is known from earlier studies by Baklanov
et al. [71] that Cu samples exposed to air display the formation
of a Cu hydroxide surface contamination. As we had also
