Full text of DOI:10.1557/JMR.2002.0350

Characteristics of copper films produced via atomic
layer deposition
Jinshan Huo and Raj Solanki
Department of Electrical and Computer Engineering, OGI School of Science and Engineering at
OHSU, Beaverton, Oregon 97006
James McAndrew
American Air Liquide, Chicago Research Center, Countryside, Illinois 60525
(Received 22 April 2002; accepted 19 June 2002)
Properties of copper films produced using atomic layer deposition (ALD) were
characterized. Composition, morphology, and electrical properties of these films grown
on glass, as well as Ta, TiN, and TaN on silicon wafers were examined. The resistivity
of films thicker than about 60 nm was near bulk value. Films were deposited using a
two-step ALD process in which copper(II)-1,1,1,5,5,5,-hexafluoroacetylacetonate
hydrate and water vapor were introduced in the first step and a reducing agent was
introduced in a subsequent step. Five reducing agents were evaluated, with the best
results obtained using isopropanol or formalin. The optimum deposition temperature
with isopropanol was about 260 °C, whereas it was about 300 °C with formalin. These
films were also investigated as seed layers for electrodeposition of thicker Cu layers
for possible interconnect applications. Excellent fills in high aspect ratio trenches were
demonstrated.
all the surface reactions are saturated, making the growth
I. INTRODUCTION
process self controlled. As a result, ALD is capable of
depositing conformal films whose thickness can be de-
termined accurately by the number of deposition cycles.
In addition, the separate dosing of precursors assures that
no detrimental gas-phase reactions will take place. Also
in practice, it has been observed that the temperatures
needed for depositing high-quality films by ALD are
lower than those required by CVD.
The choice of the precursors is critical for the ALD
process to work. Although several precursors have been
successfully used for Cu deposition using CVD (e.g., see
Ref. 2), these sources will not necessarily be suitable for
the ALD process, since the latter is driven by sequential
surface chemistry. ALD of Cu was first reported by
Tuppo et al., who used metallic zinc to reduce copper (I)
chloride.3 Subsequently, hydrogen was used as the re-
ducing agent;4 however in this case, significantly higher
deposition temperature (approximately 400 °C) was re-
quired. The higher volatility of the organometallic Cu
sources simplifies the deposition process. However, one
should note that Cu I sources are generally not well
suited for ALD since they tend to be thermally less stable
and therefore easily decompose on contact with the sub-
strate. On the other hand, Cu II sources are thermally
more stable and hence better suited for the ALD process.
Martensson and Carlsson have produced self-limited
Cu deposition in the 190 to 260 °C range using Cu
Electrodeposition of copper for fabrication of micro-
electronic device interconnects has now become a rou-
tine process. Prior to electrodeposition or electroplating,
the wafers require a thin layer of Cu, which is called a
seed layer. Currently, modified versions of sputtering are
employed for deposition of these seed layers. As the
device dimensions shrink, one of the challenges faced is
finding a way to lay down a uniform seed layer in high
aspect ratio trenches and vias of demascene structures. It
is important that these layers be highly conformal, con-
tinuous, and smooth. To achieve this, we have been in-
vestigating use of atomic layer deposition (ALD) for
producing thin conformal copper seed layers. Our initial
results showing feasibility of this technique were re-
ported recently.1 Here we present further results, includ-
ing electrodeposition on ALD Cu seed layers.
Atomic layer deposition is a monolayer stepwise
growth process that proceeds by exposing the substrate
surface alternately to each precursor followed by a nitro-
gen pulse to remove the excess species and byproducts of
the reaction. Therefore, unlike chemical vapor deposition
(CVD), in ALD the precursors are introduced to the sub-
strates separately. Hence with this sequential mode, the
duration of each exposure is adjusted so that each surface
reaction goes to completion before starting the next re-
action. Under properly adjusted experimental conditions
2394
J. Mater. Res., Vol. 17, No. 9, Sep 2002
© 2002 Materials Research Society
http://journals.cambridge.org
Downloaded: 29 Dec 2014
IP address: 131.111.164.128

J. Huo et al.: Characteristics of copper films produced via atomic layer deposition
(II)-2,2,6,6-tetramethyl-3,5-heptandionate or Cu(thd)2
vapor pressure of Cu(hfac)2 и xH2O is greater than that of
the anhydrous compound. In the absence of added H2O,
dehydration by the dry carrier gas may occur. Thus it is
important to add H2O to the carrier gas stream to maxi-
mize the amount of copper delivered per pulse.10,11 It has
been shown by Lecohier et al. that an increased reaction
rate occurs even when water vapor is added to the proc-
ess chamber separately from the copper precursor, so
there is an additional benefit beyond vapor pressure en-
hancements.12 Awaya and Arita postulated that, in the
presence of water, the reaction proceeds by dissociation
of water coordinatively bonded to copper, forming oxi-
dized copper and Hhfac, followed by the reduction of
copper oxide by hydrogen.6 Lecohier et al. observed that
both hydrogen and helium carrier gases gave the same
reaction rate in the presence of water vapor on Pt seed
layer substrates, indicating that the reduction by hydro-
gen was not important.13 They proposed that the rate-
determining step was H-transfer from H2O to hfac and
that resulting OH species could be removed by combi-
nation to form H2O2 or by reaction with H2 if present.
The film growth rate observed by Lecohier et al. was
proportional to the water vapor concentration; i.e., they
operated in a regime where water vapor was the limiting
reagent. Cohen et al. clearly showed that Cu(hfac)2 de-
composes upon adsorption on a metal surface to form
Cu(hfac) and adsorbed hfac, with the latter very likely
decomposed.9
Considering the above, we propose a two-step mecha-
nism. In the first ALD step, Cu(hfac)2 и xH2O is deliv-
ered to the surface with an excess of water vapor, where
it is adsorbed and decomposed to Cu(hfac) и yH2O (y ഛ
x) and hfac, followed by reaction to form Hhfac and
oxidized copper. The reaction proceeds until the surface
is saturated with oxidized copper, hfac, and decomposi-
tion products. During the subsequent purge step, there
may be some loss of copper by reformation of Cu(hfac)2
on the surface. Hhfac has been shown to effectively react
with CuO to form Cu(hfac)2 at 200 °C.14
Reduction occurs in the second step, where reducing
gases such as alcohols transfer hydrogen to hfac on the
surface to form volatile Hhfac. They also reduce oxidized
copper to metallic copper and water. We observed that
formalin and isopropanol produce acceptable copper
films even in the absence of hydrogen. Films of about the
same thickness and resistivity are achieved on bare glass,
Si, and Si coated with TaN when H2 is replaced with Ar.
Use of hydrogen carrier gas gave a somewhat “cleaner”
looking film. Previously, in the absence of other reducing
agents, it was observed that selective deposition on Pt
seeded surfaces occurred with He carrier gas, whereas
nonselective deposition occurred with H2 as the carrier
gas.12 In the present case, the other reducing agents em-
ployed gave rise to nonselective deposition, even in the
absence of H2.
as the Cu source reduced with H2.5 However, this process
requires a layer of platinum/palladium for Cu deposition
to occur.
The Cu precursor for our current investigation is cop-
per (II)-1,1,1,5,5,5-hexafluoroacetylacetonate hydrate or
Cu (hfac)2 и xH2O. This source has been examined ex-
tensively for CVD deposition of Cu films, with process
modifications that include plasma-enhanced CVD and
addition of water or ethanol that have lead to significant
improvement in the film properties.2,6–8 To promote ef-
fective reduction of Cu(hfac)2 molecules into high-purity
copper films, the selection of external reducing agents
and film deposition conditions play extremely important
roles. Application of such reducing agents in CVD met-
allization process has been shown by Cohen et al. to
result in both increased density of Cu nuclei and faster
removal of the hfac ligands, thereby allowing for depo-
sition of smoother Cu thin films with excellent step
coverage.9
II. EXPERIMENTAL
In this investigation, we examined five different re-
ducing agents. These are methanol, ethanol, isopropyl
alcohol (IPA), formalin (approximately 37% formalde-
hyde in 10–15% methanol and water), and carbon mon-
oxide. The carrier gas for the reducing agent and
Cu(hfac)2 was H2. In the case of Cu(hfac)2, H2 was first
bubbled through water, which is an important step. Car-
bon monoxide was delivered as a 30% mixture with H2
or N2 (mixtures were supplied by Air Liquide, Morris-
ville, PA). The Cu films were deposited on 50 × 50 mm
substrates that included glass plates and silicon wafers
pre-coated with a 500-nm layer of SiO2, followed by a
barrier layer about 6 nm thick. The three different barrier
layers examined were Ta, TaN, and TiN.
The Cu deposition was performed in a microchemistry
F-120 ALD reactor. A typical ALD cycle consisted of a
1.5-s-long Cu(hfac)2 pulse, then a 1.2-s N2 purge, fol-
lowed by a 1-s reducing agent pulse, and finally another
1.2-s N2 purge. The Cu source was heated to 75 °C, and
the reducing agents were all at room temperature. The
chamber pressure was 5 mtorr. Although deposition of
Cu was observed in all cases around 230 °C, the results
reported here were produced at 300 °C, unless other-
wise noted.
III. PROPOSED MECHANISM
As mentioned previously, the carrier gas for the
Cu(hfac)2 was H2 that was first bubbled through water.
The importance of water has been studied rather exten-
sively for CVD deposition of Cu from Cu(hfac)2. The
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2395
http://journals.cambridge.org
Downloaded: 29 Dec 2014
IP address: 131.111.164.128

J. Huo et al.: Characteristics of copper films produced via atomic layer deposition
Although CO is a strong reducing agent, the quality of
significant improvements in the resistivity of the Cu
films. This is most likely due to incorporation of impu-
rities such as carbon and oxygen, resulting either from
incomplete dissociation of methanol or partial removal of
the hfac ligands from the Cu precursor. XPS analysis
showed about 3 at.% carbon and 4 at.% oxygen.
Results with ethanol were significantly better. The
films were again smooth and appeared brightly copper
colored. Thinner films were mirrorlike; however, thicker
films appeared slightly milky due to surface roughness.
The best Cu films were produced with IPA and formalin
as the reducing agents, with formalin producing slightly
better films. These films appeared brightly copper col-
ored and were much smoother than those produced with
ethanol. The Cu films discussed below were all produced
with the formalin chemistry.
Films with thickness up to about 120 run were mir-
rorlike. However, thicker films tended to be hazy due to
the surface roughness. The adhesion of these films was
checked using the Scotch tape pull test, in which the tape
was examined for traces of Cu under an optical micro-
scope. The copper films on TiN and TaN adhered well
and did not peel off. Adhesion of Cu to Ta and glass was
not as good; segments of the film tended to peel off with
the Scotch tape. We should note that the Ta-coated wa-
fers were exposed to air for an extended period of time
prior to deposition; hence its surface was most likely
oxidized. However, Cu deposited at 320 °C had signifi-
cantly better adhesion on both surfaces.
Cu films deposited with it were not as good as with
ethanol, IPA, and formalin. The deposition rate with al-
cohols and formalin was about 0.22 nm/cycle, whereas
the deposition rate with CO/H2 was up to 50% higher.
The higher deposition rate with CO is most likely due to
a CVD component from residual CO remaining in the
reactor between the pulses. This was evident when
the pressure in the reactor was monitored. The deposition
rate was lower in balance nitrogen than in balance hy-
drogen (30% CO in either case). The Cu films in both
cases were rough and had high resistivity. These films
had carbon and oxygen concentrations of about 12 and
14 at.%, respectively, measured by x-ray photoelectron
spectroscopy (XPS) analysis. The dark appearance of
these films was probably due to the presence of Cu2O.
Therefore, in summary, the following two-step ALD
mechanism is proposed for ALD of Cu. In the first step,
Cu(hfac)2 и xH2O is transported, followed by adsorption
on the surface. Decomposition and then reaction with
water follows, giving rise to oxidized Cu and adsorbed
hfac. In the second ALD step (after the N2 purge), oxi-
dized Cu is reduced to metallic Cu and volatile Hhfac is
formed. The Ellingham diagram in Fig. 1 illustrates the
effectiveness of the various reducing agents. Formalde-
hyde is not only a strong reducing agent but can also
supply H:
HCHO + Cu2O ס 2Cu + HCOOH
HCHO + H2O ס HCOOH + 2[H]
,
(1)
(2)
Grazing incidence x-ray diffraction was used to inves-
tigate the structure of Cu films on all four types of sub-
strates. In all cases, the diffraction spectra showed peaks
corresponding to (111), (200), (220), (311), and (222)
orientations, as shown in Fig. 2. A comparison of the
relative intensities of these peaks with powder diffraction
standards (JCPDS) data indicates a nonpreferential
orientation.
,
where [H] indicates hydrogen taken up by another spe-
cies (e.g., hfac). Similarly, isopropanol supplies [H] on
reduction to acetone:
CH3CH2OCH3 ס CH3COCH3 + 2[H]
.
(3)
On the other hand, CO is a strong reducing agent but
cannot supply [H]. Note also ethanol and methanol can-
not reduce H2O to H2 at the temperatures of interest, but
CO, HCHO, and IPA can. This may indicate that the
reason for high resistivity when using EtOH and MeOH
is residual H2O on the surface, presumably leading to
oxide formation in the subsequent N2 purge step.
IV. RESULTS AND DISCUSSION
A. Film properties
When Cu(hfac)2 and methanol were used as the
sources, the Cu films deposited were smooth and con-
formal. Resistivities of these films tended to be higher;
e.g., resistivity of 200-nm thick films was 5.3 ␮⍀ cm.
These films appeared reddish probably due to presence
of CuO2. Our efforts in varying the deposition param-
eters (temperature and pulse time) did not produce
FIG. 1. Ellingham diagram showing the effectiveness of various re-
ducing agents.
2396
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 29 Dec 2014
IP address: 131.111.164.128

J. Huo et al.: Characteristics of copper films produced via atomic layer deposition
B. Film resistance
C. Surface morphology and film composition
One of the motivations for using copper as an inter-
connect metal is its low resistivity; hence, the goal of
any deposition process is produce Cu films with resis-
tivity as close as possible to its bulk value. Resistivity of
the deposited films as a function of their thickness is
plotted in Fig. 3. The resistivity was determined by first
measuring the sheet resistivity using a four-point probe
and multiplying by the film thickness measured using
cross-sectional micrographs from a scanning electron
microscope. It can be seen that for films of thickness
greater about 75 nm, the resistivity was near bulk value
(1.68 ␮m cm). However as the films became thinner,
the resistivity increased, which was expected due to in-
terface scattering. We estimated the mean free path of
electrons in this film to a first approximation from the
film thickness where the resistivity starts to increase rap-
idly. This estimate gave us a value for the mean free path
to be about 50 nm, which is close to the bulk value.
Theoretical mean free path of electrons in Cu was cal-
culated to be about 42 nm.15
The surface morphology of the ALD films was in-
vestigated using an atomic force microscope (AFM).
An AFM scan of a 80-nm-thick Cu film is shown in
Fig. 4(b). As a comparison, the surface roughness of a
CVD Cu film (∼60 nm thick) deposited independently at
a microelectronics company is shown in Fig. 4(a). The
CVD film was deposited using Cu(hfac)(tmvs) (tmvs ס
trimethylvinylsilane) on a similar substrate. The average
surface roughness of the ALD film was 4.12 nm com-
pared to 8.35 nm for the CVD film. For electrodeposi-
tion, a smooth seed layer is desirable, especially within
vias and trenches.
The Cu films deposited using the formalin and IPA
chemistries had less impurity incorporation then with
methanol and ethanol. XPS depth profiles showed carbon
concentration of 1.5 at.%, with only a trace of O and H
bonds. Our main concern was F incorporation since it can
affect the adhesion of the film to the barrier layer and is
believed to reduce the electromigration resistance. Rela-
tively low F concentration of 0.25 at.% was detected,
mostly near the barrier layer interface.
FIG. 2. Grazing incidence x-ray diffraction of the ALD copper film.
FIG. 4. AFM scans showing the surface roughness of copper films
deposited via (a) CVD and (b) ALD.
FIG. 3. Resistivity of the Cu film as a function of thickness.
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2397
http://journals.cambridge.org
Downloaded: 29 Dec 2014
IP address: 131.111.164.128

J. Huo et al.: Characteristics of copper films produced via atomic layer deposition
cidence x-ray diffraction pattern of this film was not
significantly different from that of the seed layer shown
in Fig. 2.
V. CONCLUSIONS
We examined several properties of atomic layer de-
posited Cu films important for microelectronics applica-
tions. It is demonstrated that this is a viable technique for
deposition of high-purity copper films with bulk resis-
tivity. The deposited films are smooth and conformal,
especially in high aspect ratio structures and hence are
well suited as seed layers for electrodeposition. This in-
vestigation is still in progress to determine the back-end
parameters, such as electromigration resistance. Finally,
we also demonstrated atomic-layer etching of copper
films using a process that is the reverse of the deposition
process discussed above. Details of these results will be
presented in the near future.
REFERENCES
1. R. Solanki and B. Pathangey, Electrochem. Solid-State Lett. 3,
479 (2000).
2. T.T. Kodas and M.J. Hampden-Smith, The Chemistry of Metal
CVD (VCH, New York, 1994), Chap. 4.
3. M. Juppo, M. Ritala, and M. Leskela, J. Vac. Sci. Technol. A15,
2330 (1997).
4. P. Martensson and J.O. Carlsson, Chem. Vap. Dep. 3, 45 (1997).
5. P. Martensson and J.O. Carlsson, J. Electrochem. Soc. 145, 2926
(1998).
6. N. Awaya and Y. Arita, Jpn. J. Appl. Phys. 32, 3915 (1993).
7. B. Lecohier, R.K. Philipoz, B. Calplini, T. Stumm, and
H. Van den Bergh, J. Phys. Suppl. 1, 287 (1991).
8. B. Zeng, E.T. Eisenbraum, J. Liu, and A.E. Kalayeros, Appl.
Phys. Lett., 61, 2175 (1992).
9. S.L. Cohen, M. Liehr, and S. Kasi, Appl. Phys., Lett. 60, 50 (1992).
10. N. Awaya, K. Ohno, and Y. Arita, J. Electrochem. Soc. 142, 3173
(1995).
11. C-M. Chiang, T.M. Miller, and L.H. Dubois, J. Phys. Chem. 97,
11781 (1993).
FIG. 5. Cross-sectional SEM showing (a) the conformality of 20-nm-
thick ALD Cu film used as a seed layer and (b) electrodeposited Cu on
the ALD seed layer.
D. Electrodeposition
As a final test of these seed layers, electrodeposition of
thicker Cu films was investigated. The typical thickness
of the seed layers was about 30 nm, although layers as
thin as 10 nm were also used. A commercial plating so-
lution was used in our electrodeposition cell. Figure 5
shows a scanning electron microscope cross-sectional
view of the electrodeposited Cu in trenches. The depos-
ited Cu completely fills these structures and no voids can
be observed. The electrodeposited film passed the Scotch
tape pull test, indicating good adhesion. The grazing in-
12. B. Lecohier, B. Calpini, J-M. Philippoz, and H. van den Bergh,
J. Appl. Phys. 72, 2022 (1992).
13. B. Lecohier, B. Calpini, J-M. Philippoz, and H. van den Bergh,
Appl. Phys. Lett. 60, 3114 (1992).
14. F. Rousseau, A. Jain, T.T. Kodas, M. Hampden-Smith, J.D. Farr,
and R. Muenchausen, J. Mater. Chem. 2, 893 (1992).
15. N.W. Ashcroft and N.D. Mermin, Solid State Physics (Holt,
Rinehart, and Winston, New York, 1976), p. 52.
2398
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 29 Dec 2014
IP address: 131.111.164.128