Department of Chemical Engineering, National Taiwan University of Science and Technology, 43, Keelung Road, Section 4, Taipei 106, Taiwan
Accepted 14 April 2004 Available online 4 June 2004
Abstract
The influence of surface additives, iodine and indium, on copper chemical vapor deposition (CVD) in its initial growth period is studied. The Cu CVD has been carried out on TaN/SiO2 and Cu(sputtered)/TaN/SiO2 substrates using Cu–
hexafluoroacetylacetonate–vinyltrimethylsilane (Cu(hfac)(vtms)) precursor. Addition of iodine enhances the horizontal growth and reduces the vertical growth of Cu nuclei on the TaN surface. Addition of indium increases the number density of copper nuclei significantly on TaN. Both effects accelerate the initial growth rate and shorten the initial period. The X-ray photoelectron spectra confirm that iodine floats out to the growing copper surface, but indium does not. Since iodine is floating on the copper surface, it is a surfactant and its influences persist throughout the deposition. On the other hand, the influences of extranucleation sites provided by indium cease at the end of initial growth. Hence, Cu CVD on the In-added TaN surface develops a continuous film quickly, but its growth afterwards is similar to that without indium. On the Cu(sputtered)/TaN/SiO2 substrate, the surface is featured with faceted and protruded Cu grains in the initial growth. The iodine adsorption on sputtered copper worsens the initial morphology and leads to a porous and less conductive film. After 1-h deposition at 145 8C, the sheet resistivity of Cu film on iodine-adsorbed TaN/SiO2substrate is 1.8 0.1 mO-cm.
# 2004 Elsevier B.V. All rights reserved.
Keywords: Chemical vapor deposition; Copper; Thin film; Surface morphology; Resistivity
1. Introduction
Chemical vapor deposition (CVD) of copper metal has been extensively studied for its applications in metallization of integrated circuits in the growth kinetics [1–4], surface reaction and selectivity
[5–7], precursor design and decomposition behavior [8–11]. Despite the massive efforts, implementation of copper CVD in the back-end processes is restricted to depositing a seed layer to impart the needed conduc-tivity for later electroplating process. One intrinsic problem in copper CVD is copper nuclei grow in the Volmer–Weber mode on the nitride barrier surfaces.
The Volmer–Weber mode of Cu CVD is particularly apparent at high deposition temperatures. The ten-dency to grow many protruded copper grains leads to pores between grains and a rugged surface [12]. To
Applied Surface Science 236 (2004) 165–174
*Corresponding author. Tel.:þ886 2 2737 6618;
fax:þ886 2 2737 6644.
E-mail address: [email protected] (D.-S. Tsai).
implement the copper CVD, the porosity has to be removed and surface morphology needs to be controlled. Application of a surface additive to control the growth mode is a plausible approach.
Modifying the growth mode through the influence of surface additive is certainly an old concept; the term
‘‘surfactant’’ is much more discussed in physical vapor deposition[13,14]than chemical vapor deposi-tion. Atomic hydrogen has been identified as a sur-factant beneficial to the heteroepitaxial growth of germanium on Si(1 1 1) surface in molecular beam environment where Si surfaces with and without adsorbed hydrogen can be compared [15]. The sur-factant role of hydrogen is rarely mentioned in GeH4/ SiH4CVD, because hydrogen adatoms always exist on Si surface in dehydrogenation reactions [16]. Water vapor enhances the nucleation and the growth rate in copper deposition, if a proper water dosage is applied and Cu–hexafluoroacetylacetonate–vinyltrimethylsi-lane (Cu(hfac)(vtms)) is used as precursor [17–19].
When Cu(hfac)2 is the copper source, addition of water vapor appears to be less beneficial [20].
Adsorbed iodine has been reported to be both catalyst and surfactant in copper CVD using Cu(hfac)(vtms) as precursor. Since iodine is not adsorbed on the TiN surface, it is adsorbed on copper after the first Cu layer has been deposited[21,22]. Hence, the iodine effects on the initial Cu growth are not understood. Iodine is also found to reduce the surface roughness and extend the surface reaction-limited region in Ru metal CVD [23]. Palladium has been sputtered onto the TiN/SiO2 substrate for enhancing the nucleation in copper MOCVD [24]. Indium is a potential surfactant for copper CVD; theoretical calculations indicate that indium assists in the layer-by-layer growth on Cu(1 1 1) and Cu(1 0 0) crystal planes[25–27].
In this paper, we report the effects of iodine and indium on Cu CVD, especially in its initial growth period. The initial growth period is defined as the time span from copper nucleation on the surface to the point when copper nuclei cover the whole surface and form a continuous film. Surface additive iodine is adsorbed as soon as copper deposition occurs on TaN. Indium is sputtered on the TaN/SiO2/Si(1 0 0) substrate before Cu CVD. The evolutions of copper nuclei in CVD on these substrates and the substrate without surface additives in the initial period are recorded and com-pared. The film resistivity and the surface roughness of
continuous films are also measured to understand the influence of surface additives after the initial growth.
2. Experimental details
2.1. Preparation of copper thin films
Copper CVD was carried out in a warm-wall reactor at growth temperature 140–240 8C and reaction cham-ber pressure 0.3 Torr. The reactor was pumped down to 5 106Torr before each deposition run. The liquid copper source Cu(hfac)(vtms) – commercial name CupraSelect – was purchased from Schumacher.
CupraSelect was evaporated in a bubbler which was held at 40 8C, and carried into the reactor by high-purity argon gas at 20 sccm/min. The delivery line between the bubbler and the reactor was maintained at 45 8C. The TaN/SiO2substrate was prepared by reac-tive sputtering a TaN barrier layer (85 nm) on top of SiO2(100 nm)/Si(1 0 0) substrate in N2/Ar atmosphere under a total pressure of 102Torr, using an RF power of 250 W. The CuS/TaN/SiO2substrate was prepared by sputtering another copper layer (5 nm) on top of the TaN/SiO2substrate.
Since iodoethane C2H5I adsorbs on the copper surface dissociatively[28], but not on the TaN surface, the iodoethane adsorption was planned right after Cu(hfac)(vtms) decomposed on the TaN/SiO2 sub-strate. The iodine source iodoethane was stored in an evaporator filled with argon gas at 300 Torr. A mixture of C2H5I vapor and Ar was released into the reactor chamber, while the evaporator pressure decreased from 300 to 150 Torr. The dose of iodoethane was estimated as 7 104mol for one discharge. Then, the precursor inlet valve was switched from the exit pipeline to the reactor and the deposition proceeded. As soon as the copper nuclei formed on the TaN surface, iodoethane would be dissociatively adsorbed. The period of C2H5I presence in the reactor was around 1 min, the flowing argon quickly carried C2H5I out of the chamber. This sub-strate was denoted as IA/TaN/SiO2. An indium layer of thickness 4 nm was sputtered on the TaN/SiO2 sub-strate using RF power of 100 W; the subsub-strate was denoted as InS/TaN/SiO2. Several CuS/TaN/SiO2 sub-strates were dosed with iodoethane in the way
166 T.-Y. Chang et al. / Applied Surface Science 236 (2004) 165–174
described above. These substrates were denoted as IA/ CuS/TaN/SiO2.
2.2. Microstructure characterization and film resistivity
The surface morphology was examined using a field-emission scanning electron microscope FESEM (JSM-6500F JEOL), equipped with an energy disper-sive spectrometer (INCA Energy, Oxford Instru-ments). The number density of nuclei, the average nucleus size and the fractional area being covered by copper are estimated from FESEM images, using software Sigma Scan Pro 5.0. The surface roughness was measured using atomic force microscopy (Nano-scope III, Digital Instrument), and the images were acquired in tapping mode. The surface composition was determined by X-ray photoelectron spectroscopy (XPS) using a Thermo VG Scientific Theta Probe system under a base pressure of 109Torr. The Al Ka line at 1486.68 eV was the X-ray source and the Ag 3d5/2line at 368.26 eV was the calibration refer-ence before measurement. XPS peak positions and integrated intensities were obtained through the curve fitting, using Avantage v1.68 Software of Thermo VG
Scientific. The Cu sheet resistivity was measured by the van der Pauw method, using a 4-point probe. Ten measurements at different locations were taken on each specimen. The sheet resistivity was calculated from the equation, 4.53 xf(V/I), in which xfwas the film thickness, V was the potential difference between two inner probes and I was the current driven through the two outer probes. In discussion of electrical prop-erty evolution in the initial growth period, the sheet resistance V/I, instead of sheet resistivity, was used when the film thickness was ill-defined.
3. Results and discussion 3.1. Surface morphology evolution
The area being covered by copper nuclei, the num-ber density of nuclei, and the average nucleus size on TaN/SiO2 substrates are plotted against the growth temperature in Fig. 1. The growth time is fixed at 3 min. The average nucleus size increases with the growth temperature. On the other hand, both the fractional areas covered by Cu nuclei and the number density of Cu nuclei first increase with the growth
Fig. 1. Percentage of area covered by copper, nucleus number density and copper nucleus size on the TaN/SiO2substrate at various growth temperatures. The growth time is 3 min.
T.-Y. Chang et al. / Applied Surface Science 236 (2004) 165–174 167
temperature and then decrease. The transition in initial growth occurs between 200 and 220 8C. As the sub-strate temperature increases from 200 to 220 8C, the number density and the fractional area covered by copper both plunge, while the copper nucleus size increases significantly in the vertical direction. The transition temperature in surface morphology is con-sistent with reports in the literature. Yang et al.
observed irregular/nonuniform grain growth which led to poorly connected large crystallites and films of high resistivity above 200 8C [17]. Lee et al.
reported a similar observation on the transition in surface morphology, and attributed the formation of globular grains and voids to the growth kinetics switch from the surface-reaction to the mass-transport limited mechanism[12]. Apparently, the surface morphology at a growth temperature higher than 200 8C is shifted towards an inferior starting point that shall lead to a porous Cu film of rugged surface.
SinceFig. 1suggests the copper surface morphol-ogy grown at temperatures less than 200 8C is more suitable for improvement, a low deposition tempera-ture 145 8C is chosen to study the effects of surface additives. Morphological evolutions during the initial copper growth on TaN/SiO2, IA/TaN/SiO2, InS/TaN/
SiO2, CuS/TaN/SiO2, IA/CuS/TaN/SiO2substrates are illustrated in Fig. 2. Fig. 2(a) and (b) are FESEM micrographs of TaN/SiO2 surface after copper is grown for 2 and 6 min, Fig. 2(c) and (d) are those of IA/TaN/SiO2,Fig. 2(e) and (f)are those of InS/TaN/
SiO2, Fig. 2(g) and (h) are those of CuS/TaN/SiO2, Fig. 2(i) and (j)are those of IA/CuS/TaN/SiO2.Fig. 2(a) and (b)indicates that copper nuclei are quite small and sparsely populated on TaN/SiO2 in the beginning.
After 2-min CVD, nucleus sizes are 50.2 2.6 nm, the number density of Cu nuclei is estimated 1.21 109cm2. After 6-min growth, both the average nucleus size and the number density of Cu nuclei appeared increased, 139.3 14.1 nm and 4.84 109cm2. Under the influences of surface iodine, Fig. 2(c) and (d)depicts that Cu nuclei become larger and more populated in the first few minutes. The average Cu nucleus size is 91.7 12.1 nm and the number density 6.65 109cm2after 2-min growth.
After 6 min, the Cu nuclei begin to coalesce; therefore, some Cu grains are very large and the number density drops to 3.14 109cm2. The FESEM image shows that these large Cu grains are formed through
extend-ing and connectextend-ing smaller nuclei in the horizontal direction; therefore, they are rather flat. The average grain size is 169.6 34.3 nm.
The influence of indium is very different (Fig. 2e and f). The early surface morphology of InS/TaN/SiO2 is featured with densely populated small Cu nuclei.
The average Cu nucleus size is 31.2 2.1 nm, and the number density is 69.77 109cm2 after 2-min growth. Since the number density of Cu nuclei is very high, they could start to coalesce before 2 min.
Fig. 2(f)shows that the coalescence action has elimi-nated so many nucleus boundaries after 6 min that Cu nuclei are no longer distinctive units. It can be con-cluded from these micrographs that both iodine and indium have positive effects on the surface morphol-ogy in the initial growth; these effects improve the Cu surface morphology of a continuous film later on TaN surface.
It is of interest to examine how the CVD copper grows on a sputtered copper surface and compare its features with those on the TaN surface.Fig. 2(g) of CuS/TaN/SiO2indicates that CVD copper nucleates in the first 2 min easier and grows faster on the sputtered Cu surface than on the TaN surface. The average nucleus size is 82.6 8.2 nm and the number density of grain 4.13 109cm2.Fig. 2(h)shows that the Cu nuclei grow larger and become Cu grains; these grains develop clear crystal facets after 6 min. The average Cu grain size is 165.0 29.1 nm. The number density of Cu grains is slightly lower, 3.92 109cm2. Influences of adsorbed iodine on the sputtered Cu surface are quite different from those on the TaN surface. FESEM images ofFig. 2(i) and (j)exhibits the largest Cu grains and the most isolated Cu grains in all five specimens. The average Cu grain size on IA/ CuS/TaN/SiO2 is 204.4 17.1 nm after 2 min and 353.9 29.1 nm after 6 min. The number density is 0.64 109cm2 at 2 min and 0.41 109cm2 at 6 min. The adsorbed iodine on IA/CuS/TaN/SiO2 assists in developing large and protruded copper crys-tallites.
The percentages of surface area that are occupied by Cu nuclei in the early 7 min on these five substrates at 145 8C are summarized inFig. 3. The variation of Cu area with time represents the film forming rate. As a whole, the copper film forming rate is higher on the TaN surface than that on the sputtered Cu surface.
With the help of indium, the CVD copper grains on
168 T.-Y. Chang et al. / Applied Surface Science 236 (2004) 165–174
Fig. 2. Surface morphology evolution with growth time. The Cu CVD is carried out at 145 8C and 0.3 Torr. FESEM micrographs of the TaN/SiO2surface after CVD (a) 2 and (b) 6 min; those of the IA/TaN/SiO2surface after (c) 2 and (d) 6 min; those of the InS/TaN/SiO2surface after (e) 2 and (f) 6 min; those of the CuS/TaN/SiO2surface after (g) 2 and (h) 6 min; those of the IA/CuS/TaN/SiO2surface after (i) 2 and (j) 6 min.
T.-Y.Changetal./AppliedSurfaceScience236(2004)165–174
InS/TaN/SiO2occupy the surface easily over 80% in the first 5 min. On the other hand, the film forming is the most difficult on IA/CuS/TaN/SiO2, the copper
surface percentage increases slowly from 19.8 (2 min) to 41.9% (7 min). After 7-min deposition, the percentages of surfaces covered by CVD copper are ranked as follows: 95.0% (InS/TaN/SiO2), 93.4%
(IA/TaN/SiO2), 83.4% (TaN/SiO2), 74.5% (CuS/TaN/
SiO2), 41.9% (IA/CuS/TaN/SiO2). The deposited Cu develops a film most quickly on the InS/TaN/SiO2 surface.
Fig. 4 summarizes the variations of the copper nucleus sizes on the five substrates. The average nucleus size and the standard deviation in nucleus sizes both increase with the deposition time. The copper nuclei on InS/TaN/SiO2 are small and their size is very uniform. They connect each other early because the nuclei number density is huge. Another microstructural feature of Cu nuclei on InS/TaN/SiO2 is that their nucleus boundaries are blurred when they begin to connect after 4 min. In contrast, the copper nucleus boundaries on IA/TaN/SiO2are legible and the nuclei are larger than those on TaN/SiO2and InS/TaN/
SiO2. The average nucleus size jumps from 169.6 (6 min) to 320.3 nm (7 min) as many Cu nuclei merge.
Fig. 2. (Continued ).
2 3 4 5 6 7
Area being covered by Cu nuclei (%)
Growth time (min)
Fig. 3. Percentage of area covered by copper nuclei vs. time at 145 8C.
170 T.-Y. Chang et al. / Applied Surface Science 236 (2004) 165–174
Since several merged nuclei are considered to be one big irregular-shaped nucleus, the standard deviation in nuclei sizes increases drastically, from 34.3 (6 min) to 62.3 nm (7 min). The average size of Cu nuclei on IA/ CuS/TaN/SiO2is the largest among the five substrates.
These nuclei grow in size, but remain divided. Hence, the standard deviations in grain size are similar in the first 7 min.
Variations in the number densities of copper nuclei in the initial growth are illustrated in Fig. 5. The number density on the InS/TaN/SiO2substrate is the highest in the beginning, it quickly drops to a magni-tude similar to those of the other four substrates in 4 min, which is the end of initial growth. On both of IA/TaN/SiO2and TaN/SiO2surfaces, the number den-sity of nuclei first increases, then decreases. This type of variation demonstrates that the initial growth fol-lows the nucleation-and-coalescence process. The number density on IA/TaN/SiO2 is higher than that of TaN/SiO2; it descends earlier too. The number density on CuS/TaN/SiO2and IA/CuS/TaN/SiO2does not change much in the first 7 min. Nucleation of CVD copper on the Cu(sputtered) surfaces appears to be difficult.
Table 1lists the root-mean-square roughness RMS values of the five substrates after 10- and 30-min growth. After 10-min deposition, the nuclei on TaN/
SiO2, IA/TaN/SiO2, InS/TaN/SiO2, CuS/TaN/SiO2 have been sufficiently connected and form a contin-uous film, but those on IA/CuS/TaN/SiO2remain
iso-lated. Both iodine and indium are effective to help deposit a smooth surface. RMSvalues of 10-min films on IA/TaN/SiO2 (11.7 nm) and InS/TaN/SiO2 (12.6 nm) are low, compared with that of TaN/SiO2 (14.4 nm). The surface of 10-min CVD film on CuS/ TaN/SiO2 is rough, RMS¼ 22.5 nm. After 30-min CVD, the copper film of thickness 208 nm on IA/ TaN/SiO2 is still the smoothest surface, RMS¼ 26.6 nm. The RMS value of InS/TaN/SiO2(36.3 nm) is similar to that of CuS/TaN/SiO2(37.0 nm) and IA/ CuS/TaN/SiO2(36.6 nm). The most rugged surface is the copper thin film on TaN/SiO2, RMS¼ 46.9 nm.
3.2. Surface composition
Surface compositions of copper thin films grown at 145 8C for 1 h are listed onTable 2. The surface iodine
2 3 4 5 6 7
Cu nucleus size (nm)
Growth time (min)
Fig. 4. Copper nucleus size vs. time at 145 8C.
2 3 4 5 6 7
Copper nucleus number density ×10 9 (cm-2 )
Growth time (min)
Fig. 5. Copper nucleus number density vs. time at 145 8C.
Table 1
Root-mean-square roughness RMSof the copper thin films on five substrates after 10- and 30-min CVD
Substrate RMS(nm)
10 min 30 min
TaN/SiO2 14.4 46.9
IA/TaN/SiO2 11.7 26.6
InS/TaN/SiO2 12.6 36.3
CuS/TaN/SiO2 22.5 37.0
IA/CuS/TaN/SiO2 99.9a 36.6
aCu grains on IA/CuS/TaN/SiO2 after 10 min remain discon-nected.
T.-Y. Chang et al. / Applied Surface Science 236 (2004) 165–174 171
is detected on both of IA/TaN/SiO2and IA/CuS/TaN/
SiO2 substrates. The presence of iodine is also con-firmed in the element mapping of FESEM. Since iodine is adsorbed only in the early moment and the deposited copper is quite thick, iodine is supposed to be buried in the Cu layer. Substantial amount of surface iodine – 0.5 and 0.8% – suggests that iodine atoms float out during deposition. Thus, iodine is a surfactant in Cu CVD [13]. The iodine presence on TaN surface also appears to reduce oxygen impurity of the TaN surface. The oxygen impurity level of Cu film on IA/TaN/SiO2is not detectable, while that on TaN/
SiO2 and In/TaN/SiO2 substrate is 0.95 and 0.7%;
respectively. Quite different from iodine, there is no
respectively. Quite different from iodine, there is no