Low-k SiC
xN
yFilms Prepared by Plasma-Enhanced Chemical
Vapor Deposition Using 1,3,5-trimethyl-1,3,
5-trivinylcyclotrisilazane Precursor
Hung-En Tu, Yu-Han Chen, and Jihperng Leu∗,z
Department of Materials Science and Engineering, National Chiao Tung University, Hsinchu 30049, Taiwan
Low-k silicon carbonitride (SiCxNy) films with k of 3.6–4.6 were prepared by radio frequency plasma-enhanced chemical vapor
deposition at 25 to 400◦C under low power density of 0.15 W/cm3, using a single source precursor, 1, 3, trimethyl-1, 3,
5-trivinylcyclotrisilazane (VSZ), and Ar. At lower deposition temperatures (≤ 200◦C), most cyclic VSZ structures were preserved in the SiCxNyfilms, resulting in a lower density (1.60–1.76 g/cm3), a lower dielectric constant (k∼3.6–3.9) and a fairly good elastic
modulus of 22.0–25.0 GPa. When the deposition temperature was raised to 400◦C, the cyclic N-Si-N linkages were reformed to a dense Si-N structure, with the desorption of CHxbonds, resulting in higher density (2.0 g/cm3), a dielectric constant of 4.6, and an
excellent elastic modulus of 65.2 GPa. The leakage current density of SiCxNyfilms was reduced from 1.5×10−6to 4.0×10−8A/cm2
at 1 MV/cm, upon increasing the deposition temperature from 25◦C to 400◦C. The conduction mechanism of the SiCxNyfilms,
except the film deposited at 400◦C and tested under higher electric field, exhibited Schottky emission due to few charged defects by using a cyclic VSZ precursor and a lower plasma power density of 0.15 W/cm3.
© 2012 The Electrochemical Society. [DOI: 10.1149/2.085205jes] All rights reserved.
Manuscript submitted December 7, 2011; revised manuscript received February 14, 2012. Published March 2, 2012.
As the dimensions of integrated circuits are scaled according to Moore’s law, the increase in propagation (RC) delay, crosstalk noise and power dissipation in the backend interconnects become the lim-iting factors in the ultra-large scale integrated devices.1 To reduce
the capacitance in the backend interconnects, low dielectric constant (k < 3.5) material was first introduced as an interlayer dielectric (ILD).2 Meanwhile, silicon nitride was retained as the etch-stop and
diffusion barrier layer in the dual damascene architecture, because of its excellent etch selectivity and barrier effectiveness, although its dielectric constant is relatively high, at 6.5–7.0.3In order to further
reduce the capacitance in the backend interconnects, the semiconduc-tor industry has consistently tried to achieve lower effective dielectric constants, which involves low-k ILD and low-k etch-stop layer with reduced thickness.4, 5
To reduce the k-value of silicon nitride films, silicon car-bonitride (SiCxNy) thin films have been introduced as an etch
stop/barrier layer,6–8 because of their low dielectric constant and
their properties as effective barriers against Cu diffusion and drift.9, 10 In addition to sputtering deposition and laser vapor
de-position methods, silicon carbonitride films have been prepared by plasma-enhanced chemical vapor deposition (PECVD), using multi-precursors such as SiH4+NH3(N2)+CH411, 12 and SiH(CH3)3
+NH3.8, 13, 14 In recent years, single source precursors, such as
hexamethyldisilazane (HMDS), for low-k SiCxNy applications,9, 15
and tris(dimethylamino)silane16 and
1,3-bis(dimethylsilyl)-2,2,4,4-tetramethylcyclo- disilazane,10 for increased mechanical and
tribo-logical performance, have been the subject of much research, because they retain the ready fragments and allow better control of the com-position of films, compared to multi-precursors.
In this paper, a new, cyclic organosilazane, 1, 3, 5-trimethyl-1, 3, 5-trivinyl- cyclotrisilazane (VSZ) with a cyclic Si-N-Si, pendent CH3, and vinyl groups, as illustrated in Figure 1, is used as a
sin-gle source precursor for the deposition of low-k SiCxNyfilms by the
radio frequency (RF) PECVD technique using a low power density (0.15 W/cm3). This approach may allow the retention of most of the
chemical bondings, such as the pendent Si-CH3for low polarazability
and the use of cyclic Si-N-Si in the starting precursor to reduce the dielectric constant, while maintaining fairly good mechanical strength when the vinyl groups in single source precursor are broken to form a cross-linked structure.17This study used Fourier transform infrared
spectroscopy (FTIR) and X-ray photoelectron spectroscopy (XPS)
∗ Electrochemical Society Active Member. zE-mail: [email protected]
to investigate the effect of deposition temperature on the chemical structure of low-k SiCxNy films. In addition, the films’ properties
and electric characteristics, such as density, elastic modulus, dielec-tric constant, dielecdielec-tric breakdown and leakage behavior were also studied.
Experimental
The SiCxNy films were deposited onto (100) silicon wafers by a
RF (13.56 MHz) parallel-plate PECVD system, in which the electrode spacing and the diameter of the electrode were 20 and 150 mm, respec-tively. 1, 3, 5-trimethyl-1, 3, 5- trivinylcyclotrisilazane (C9H21N3Si3,
VSZ) (Gelest, Inc. 95%) was used as the single source precursor. The liquid precursor was vaporized at 60◦C, to prevent condensation, and carried to the reactor using an argon carrier gas at a flow rate of 20 sccm. The showerhead in the upper electrode distributed the pre-cursor and gases uniformly over the substrate. The deposition pressure and RF power were maintained at 90 mTorr and 50 W (power density = 0.15 W/cm3), respectively, without bias. The deposition
tempera-tures were varied from room temperature to 400◦C, to investigate the influence of deposition temperature on the structures and properties of SiCxNyfilms.
The chemical bonding and composition of SiCxNyfilms were
ex-amined using specular reflectance FTIR (MAGNA-IR Technology Protege 460) at a 50◦incident angle. The FTIR spectra were collected in the 500–4000 cm−1range at a resolution of 4 cm−1, with a total of 32 scans.
XPS analyzes were performed, to examine the chemical bonding and compositions, using a PHI Quantera AES 650) with a monochro-mated Al kαX-ray source (hν = 1486.6 eV) with an energy resolution of 0.1 eV. The surface of the specimen was pre-cleaned by bombard-ment with Ar+ions (5 kV), prior to the collection of XPS data. A neutral gun was employed, to eliminate the charge effect during the XPS measurement. The XPS spectrum was calibrated by the binding energy of the Au 4f7/2line at 84.0 eV. Quantification of XPS data was
achieved using peak areas and experimental sensitivity factors. Nanoindentation tests were carried out using a nanoindentor (MTS Nano Indenter XP System) with a Berkovich tip, in continuous mode, to obtain the reduced modulus (Er). The Oliver–Pharr method18was
then used to determine the elastic modulus (E) of various SiCxNy
films. To eliminate the effect of the Si substrate, the indentation depth was maintained at less than 20% of the film’s thickness. The density of the SiCxNyfilm was measured by X-ray reflectivity (XRR) (Bruker
D8 Discover), with a Cu Kαsource (λ = 0.154 nm), using ω-2θ scan mode. The scanning region ranged from 0◦to 2◦. The XRR data was
The dielectric constant (k) and leakage current of the SiCxNyfilms
were characterized by capacitance-voltage (C-V) (HP 4280) measure-ment and current-voltage (I-V) measuremeasure-ment, using metal-insulator-semiconductor (MIS) structure configuration [Al electrode/SiCxNy
film/Si (50 ohm-cm)] at room temperature. To accurately measure the dielectric constant by C–V dot measurement, three circular aluminum dots of nominal diameters 200, 400, and 800 mm were used, to min-imize the geometric effect. Aluminum electrodes with a thickness of 1μm were coated onto the dielectric films by ULVAC EBX-6D ther-mal evaporator through a shadow mask. Measurements of film thick-ness were made using an n&k Analyzer 1280 (n&k Technology, Inc.) at wavelengths ranging from 190 to 900 nm. The I-V characteristics were measured at room temperature, under N2 purge, using a HP
4156B with a diameter of 200μm as the electrode. For C-V and I-V measurement, the thickness of the SiCxNyfilms was between 100 and
130 nm. For nanoindentation measurement, a nominal thickness of 1 μm was used, to avoid the substrate effect.
Results and Discussion
Infrared and XPS analysis of PECVD films.— The chemical
bond-ing types in the PECVD SiCxNy films prepared at various substrate
temperatures were first characterized by specular reflectance FTIR spectroscopy. Figure 2 shows the FTIR spectra of the starting liquid precursor, VSZ and the PECVD SiCxNythin films, deposited at
vari-Figure 2. FTIR spectra of the VSZ liquid precursor and SiCxNy films
de-posited at various deposition temperatures.
2964 2887, 2956 νsCH3,νaCH3 10, 15, 17, 19 2910 νaCH2 17 3055, 1409 νaCH2,δ CH2in vinyl group 15, 17, 19 3370 3370 ν N-H 20
ous substrate temperatures. The characteristic features of the infrared spectrum of VSZ precursor included Si-N, Si-C, CH3, CH2, C=C and
N-H bonding types, as summarized in Table I. The primary absorption bands at 922 and 1048 cm−1corresponded to the asymmetric stretch-ing vibration modes of the Si-N bond of the cyclic Si-N-Si structure. Si-CH3 bonding includes the stretching vibration of the Si-C bond
at 800 cm−1,9, 10, 15 the bending vibration of the Si-CH
3 bonds at
1265 cm−19,10,15 and the asymmetric stretching vibration of the C-H3bond at 2964 cm−1.10, 15, 17Other bonding included the stretching
vibration of the C=C bonds of Si-CH=CH2 at 1600 cm−1,17, 19 the
asymmetric stretching vibration of the CH2bond in the vinyl group at
3055 cm−1,15, 17, 19the bending vibration of the CH
2bond in the vinyl
group at 1409 cm−1,15, 17, 19 and the stretching vibration of the N-H
bond at 3370 cm−120and the N–H bending mode at 1180 cm−1.15, 20
Upon plasma deposition, significant changes were observed in the infrared spectrum of the SiCxNyfilm deposited at room temperature.
The major absorption bands of VSZ and a representative SiCxNy
film, and their attributed sources are summarized in Table I. The absorption bands related to Si-N at 922 and 1038 cm−1remain the strongest bands. The absorption bands for the C=C at 1600 cm−1 and the CH2bonds in the vinyl groups at 3055 cm−1and 1409 cm−1
disappeared. In contrast, the asymmetric CH2 and symmetric CH3
stretching vibrations at 2910 and 2887 cm−1appeared.10, 15, 17, 19Also,
the broad N-H bands at 3370 cm−1 and 1150 cm−1 appeared with enhanced intensities.20, 21There was a minor loss in the Si-CH
3band at
1265 cm−1and the C-H3absorption band at 2956 cm−1, due to chain
scission of Si-CH3 under plasma conditions. These results showed
that most of the cyclic N-Si-N linkages were preserved in the SiCxNy
films after the PECVD process. Some of the cyclic N-Si-N bonds were broken, which resulted in an increase in the formation of Si-N-H and C-NH bonds. The reactive C=C bonds of the vinyl group were broken to form a cross-linked structure, during the plasma deposition process, which is similar to that reported by Lubguban et al. for low-k organosilicate glass using tetravinyltetramethylcyclotetrasiloxane precursor.17The hypothesis is supported by the appearance of CH
2
vibration modes at 2910 (asym. CH2), related to Si-(CH2)n-Si (n= 1
or 2), after the breaking of the vinyl groups. These are indicators of the cross-linked structure. The other reaction path formed a terminal Si-CH2-CH3, which may be further broken up into Si-CH3and Si-H.
As the deposition temperature was raised to≥100◦C, the absorp-tion intensity of Si-N-Si continued to increase, up to 400◦C, with its absorption band shifting from∼1048 cm−1, at a deposition tempera-ture of 25◦C, to the lower wavenumber, 1038 cm−1, for a deposition temperature of 400◦C. In contrast, the C=C bond at 1600 cm−1 and the -CH2 asymmetric stretching associated with the vinyl group at
Figure 3. The relative peak ratios of Si-N, C=C, CHx, and N-H infrared
absorption bands in the SiCxNyfilms as a function of deposition temperature.
the Si-CH3 band at 1265 cm−1 and the N-H bands at 3370 cm−1
and 1150 cm−1(1625 cm−1CN-H) decreased, as deposition temper-ature was increased, presumably due to the formation of volatile CH4
or NH3. A new absorption peak, for Si-H10, 15 or C-N22 at around
2100 cm−1became visible at T≥200◦C and increased, as deposition temperature was increased. The formation of Si-H was a result of the N-Si-N scission and reaction with hydrogen radicals from the plasma processes, while the C=C bonds of the vinyl groups were broken and attached to the fragments of N-H to form C-NH.
The quantitative change in the Si-N, N-H, CH3, CH2,and Si-H
bonds in the SiCxNyfilms as a function of deposition temperature is
shown in Figure 3, based on the area ratio of their respective infrared absorption bands, shown in Figure 2. As deposition temperature was increased, the ratio of the Si-N bonds increased and reached a plateau at T≥ 200◦C, while the ratio of Si-H bonds increased linearly, as deposition temperature increased. In contrast, the ratio of N-H (3300, 1150 cm−1) and CH3decreased approximately linearly, as deposition
temperature was increased, because the labile CH3bonds were more
volatile at higher deposition temperatures. Also, the formation of CH2
in Si-CH=CH2 and CH2 in Si-(CH2)n-Si, Si-CH2-CH3 or Si-CH2
-CNH showed a decreasing trend at deposition temperatures >25◦C. This result showed that the Si-CH=CH2 of the vinyl groups were
broken to react with another vinyl group, forming Si-(CH2)n-Si and
C-N bonds, during the PECVD process. Simultaneously, the absorption intensity of the Si-H bond increased, as deposition temperature was increased.23
In addition to infrared analysis, the compositions of SiCxNyfilms
deposited at various deposition temperatures and their relative ratios to Si were quantified by XPS. These are summarized in Table II. For reference, the ideal elemental compositions of the VSZ precursor were 20% Si, 60% C and 20% N, i.e. C/Si ratio= 3 and N/Si ratio = 1. When depositing SiCxNyfilm at room temperature, the C/Si and
N/Si ratios were 1.8 and 0.46, which were less than those of the VSZ
Table II. Compositions of SiCxNy films deposited at various deposition temperatures.
Composition, at.% (ratio relative to Si)
Si C N Deposition temperature,◦C 25 30.5 (1) 55.5 (1.82) 14.0 (0.46) 100 48.6 (1) 39.5 (0.81) 11.9 (0.25) 200 49.0 (1) 39.2 (0.80) 11.8 (0.24) 300 51.3 (1) 37.0 (0.72) 11.7 (0.23) 400 65.8 (1) 19.3 (0.29) 14.9 (0.23)
Figure 4. The absorption frequency of Si-N-Si as a function of deposition temperature.
precursor, at 3.0 and 1.0, respectively. The reduction in the C/Si and N/Si ratios of PECVD film deposited at room temperature indicated that the cyclic structure of the N-Si bonds and the branches of Si-CH3bonds were broken up to some degree, during plasma processing.
In addition, the ablation of C and N relative to Si probably occurred due to the formation of volatile CH4and NH3by-products during the
plasma deposition process.
The N/Si ratio of the SiCxNyfilms decreased from 0.46 at 25◦C
to 0.25–0.23 for deposition temperature between 100◦C and 400◦C. In contrast, their respective C/Si ratio decreased from 0.86 at 25◦C to 0.7-0.81 at T∼100–300◦C and then there was a more radical decrease to 0.29 at a deposition temperature of 400◦C. This implied that during the deposition of SiCxNy film at 400◦C by PECVD, the desorption
of CHxbonds may occur simultaneously, due to a pyrolysis process,
which resulted in the rearrangement of the films’ structure and the subsequent formation of dense SiCxNyfilms.23In addition to the XPS
results, Figure 4 shows the position of the absorption band of N-Si-N in SiCxNyfilms deposited by PECVD at various temperatures, as
measured by FT-IR analysis. At temperatures≤200◦C, the Si-N ab-sorption band was centered at 1047–1048 cm−1and still displayed the characteristics of cyclic N-Si-N. As the deposition temperature was raised to 300 and then 400◦C, the Si-N absorption band was down-shifted to 1043 and 1038 cm−1, respectively. This redshift implied that the cyclic N-Si-N structures were broken up and reformed to a dense Si-N structure.24
Properties of PECVD films.— Film density is an important
phys-ical property of SiCxNy films and is sensitive to free volume and
cross-linked structure. Figure 5 shows the density of the SiCxNyfilms
as a function of deposition temperature. The densities of the SiCxNy
films increased monotonically, from 1.6, 1.67, 1.76, 1.83 to 2.0 g/cm3,
as deposition temperature increased through 25, 100, 200, 300, to 400◦C. According to IR and XPS results, the intensities of the CHx
bonds and the Si-C/Si-N ratio for the films decreased with increasing deposition temperature. The carbon content of the SiCxNyfilms also
decreased, as deposition temperature increased. The cyclic N-Si-N structures were broken up and reformed into a dense Si-N structure,24
based on the redshift in the Si-N absorption (Fig. 4), as described in the previous section. Therefore, a rise in deposition temperature caused the scission of Si-CH2CH3 or Si-CH3bonds into Si-H bonds, which
reduced the surrounding free volume in the SiCxNyfilms.25The free
radical species created by the scissile vinyl groups in the VSZ struc-ture were rearranged with a backbone strucstruc-ture to form cross-linked, dense SiCxNyfilms, which resulted in an increase in the films’ density
at higher deposition temperatures.
Figure 6 shows the elastic modulus of the SiCxNy films as a
function of deposition temperature. The elastic moduli varied from 21.0 GPa to 65.2 GPa, as the deposition temperature was increased
Figure 5. The density of the SiCxNy films as a function of deposition
temperature.
from 25◦C to 400◦C. As deposition temperature was increased, the elastic modulus of the SiCxNy films increased linearly, due to the
scission of the Si-N-Si linkages and the desorption of CHxbonds to
increase the crosslinked structure, as confirmed by FTIR. The highest values for the elastic modulus and hardness, obtained at 400◦C, were 65.2 GPa and 6.5 GPa, respectively. This result showed that the hard-ness and elastic modulus of SiCxNyfilms at lower deposition
tempera-tures, using the VSZ precursor, were higher than those for the SiCxNy
films (H= ∼1.9 GPa, E = ∼12.2 GPa) using HMDS9 as the single
source precursor. Thus, for SiCxNy films using a single source
pre-cursor, the VSZ precursor yielded a better mechanical strength than HMDS, primarily due to the fact that its vinyl groups were reformed to a cross-linked structure, under PECVD deposition conditions.
Dielectric constant and leakage behavior.— Figure 7 shows the
dielectric constant of the SiCxNy films as a function of deposition
temperature. As deposition temperature was increased, the dielectric constant of the SiCxNyfilms increased from 3.6 to 4.6. The scission
the Si-C bonds, accompanied by a decrease in carbon content, were observed at higher deposition temperatures. This inverse relationship between the concentration of carbon and the dielectric constant is in agreement with the findings of other reports.17, 26 In addition, the
reduced free volume of the SiCxNy films, due to the formation of
bonding such as Si-H and the formation of cross-linked Si-N structures at higher temperatures, contributed to an increase in not only the dielectric constant, but also in mechanical properties such as the elastic modulus.
Figure 6. Elastic modulus of the SiCxNyfilms as a function of deposition
temperature.
Figure 7. The dielectric constant of the SiCxNy films as a function of
deposition temperature.
Figure 8a shows the leakage current density as a function of the electric field for SiCxNy films deposited at various deposition
tem-peratures. The leakage current density showed a decreasing trend from 1.5×10−6to 4.0×10−8A/cm2at 1 MV/cm, as deposition
tem-perature increased. Their breakdown strengths were all >3 MV/cm. A lower leakage current density was obtained at deposition temperatures ≥ 300◦C. This can be attributed to a decrease in pendent CH
3and an
increase in more stable Si-N bonding, as verified by the XPS results, which showed a decreasing Si-C/Si-N ratio as deposition tempera-ture was increased. Moreover, the SiCxNy films became denser at
higher deposition temperatures, which reduced the carrier transport and leakage current in SiCxNyfilms.14, 27
The conduction mechanism for leakage current is presumably at-tributed to the Schottky emission (SE) mechanism, which was induced by thermionic emission across the potential energy barrier at a metal-insulator interface, for SiCxNyfilms deposited at≤300◦C. The current
density of the SE mechanism is expressed by the following Eq. 1:28
J= A∗T2exp β sE1/2− φs kBT [1]
whereβs= (e3/4πε0ε)1/2, e is the electronic charge,ε0is the
dielec-tric constant of free space,ε is the relative dielectric constant, A∗is the Richardson constant, E is the applied electric field,φsis the
con-tact potential barrier, T is the temperature, and kB is the Boltzmann
constant. The SE mechanism was confirmed by the linear correlation between ln(J/T2) and E1/2,29for films deposited at 25–300◦C as shown
in Figs. 8b–8e. It is believed that silicon dangling bonds in SiCxNy
films create states in the energy band-gap and are responsible for hopping conduction, as proposed by Robertson and Powell.30
Figure 9a shows the leakage current versus the average electric field plot for SiCxNyfilm deposited at 400◦C. The result suggested that the
current conduction mechanism changed from the SE mechanism in the low field to the Frenkel-Poole (F-P) emission mechanism in the high field (>2.0 MV/cm). The current density in the F-P emission mechanism is expressed by the following Eq. 2:28
J= J0exp β P FE1/2− φP F kBT [2]
where J0= σ0E is the low-field current density,σ0 is the low-field
conductivity,βP F = (e3/πε0ε)1/2andφP Fis the energy barrier height
of the trap level. The F-P emission mechanism was confirmed by the linear relationship between ln(J/E) versus to E1/229 for the electric
field ranging from 2.0 to 3.0 MV/cm, as shown in Figure 9b. In the F-P emission mechanism, charge transport was dominated by the carriers that were captured and emitted by charged traps with coulomb potentials. These results indicated that charged defects existed in the the low-k SiCxNyfilm deposited at 400◦C.
Figure 8. (a) The leakage current density versus electric field, and Schottky emission mechanism fitting, for the SiCxNyfilms as a function of deposition
temperature: (b) 25◦C, (c) 100◦C, (d) 200◦C, and (e) 300◦C.
In comparison, F-P emission dominates the current conduction in the PECVD as-deposited SiCxNy films using multi-precursors such
as SiH(CH3)3+NH314or Si(CH3)4+NH3.31Recently, Mallikarjunan
et al.32demonstrates that the optimized carbon bonding in PECVD
SiCxNy film (plasma power density: 7.5–20 W/cm3 if an electrode
spacing of 1 mm is assumed) through the use of BASICN series of precursors with a combination of high Si–N and Si–C bond energies, is able to delay the onset of F-P emission. In our study of PECVD SiCxNyfilms using a single precursor (VSZ), all films exhibited SE
mechanism, except for the film deposited at 400◦C, which was at-tributed to F-P emission mechanism under higher electric field. Our study showed that there are few charged defects in our SiCxNyfilms,
presumably due to less damage by using a cyclic precursor and a lower plasma power density of 0.15 W/cm3.
The applicability of SiCxNy in this study for the
copper/low-k baccopper/low-kend interconnects will be also examined and discussed in two critical areas: (1) resist poisoning and (2) etch selectivity. Re-sist poisoning is found to occur when basic contaminants, primar-ily amines in the as-deposited etch-stop films such as SiCxNyusing
trimethylsilane+NH3+He, diffuse into the photoresist and neutralize
the photo-generated acids.33In comparison, our study is based on a
sin-gle source precursor without NH3and operated at a low power density.
We hypothesize that our SiCxNy films will alleviate or eliminate the
resist poisoning. Yet, this requires experimental validation in the pat-terning processes through the fabrication of multi-level copper/low-k test wafers in the future.
For the copper/low-k backend interconnects, a via-first patterning scheme has been widely used in a dual damascene integration
pro-Figure 9. Conduction mechanism fitting of the SiCxNy film deposited at
400◦C: (a) Schottky emission and (b)Frenkel-Poole emission.
cess flow.34, 35In the via etch process, highly selective etching of the
low-k ILD against the underlying etch-stop layer, such as SiCxNyin
our case, is required. This is to ensure sufficient thickness of SiCxNy
film to protect the underlying copper layer from plasma induced dam-age and oxidation due to diffusion of oxygen. The etch selectivity requirement becomes more challenging as the low-k ILD is reduced to k < 2.6, which is achieved by incorporating high carbon content. Recently Kume et al.35reported that a high etch selectivity (∼10) can
be achieved for a low-k ILD with high carbon content (C/Si= 2.7 for k= 2.55) against a SiCxNyetch-stop layer with C/Si= 0.9 using
N2-Ar-CFxetching gas chemistry. In our study, as shown in Table II, a
wide range of C/Si ratio from 0.29 to 1.82 was obtained with decreas-ing deposition temperature from 400◦C to 25◦C. We can infer that high etch selectivity can be achieved for our SiCxNyfilms deposited
at 100–300◦C (C/Si= 0.72–0.81), and even higher for SiCxNyfilm
deposited at 400◦C (C/Si= 0.29) if a low-k ILD with a high carbon content (C/Si ratio= 2.7) in Kume et al.’s study35is assumed. Even
for SiCxNyfilm deposited at room temperature (C/Si ratio= 1.82), a
high selectivity is still possible by designing the etching gas chemistry to enhance their difference in etch rates of ILD and etch-stop layer with specific carbon contents.
Conclusions
Low-k SiCxNyfilms with k values of 3.6–4.6 were developed and
prepared by RF PECVD (power density= 0.15 W/cm3), at 25 to
400◦C, using 1, 3, 5-trimethyl-1, 3, 5- trivinylcyclotrisilazane as a single precursor and Ar as the carrier gas. At lower deposition tem-peratures (≤200◦C), the vinyl groups of the VSZ were broken and reformed to cross-linked Si-(CH2)n-Si linkages, during the plasma
deposition process. However, most of the cyclic VSZ structures were preserved to create free volume in the SiCxNy films, which resulted
in a lower density (1.60–1.76 g/cm3) and a lower dielectric
con-stant (k∼3.6–3.9), with a fairly good elastic modulus of 22.0–25.0 GPa. When the deposition temperature was raised to≥300◦C, the cyclic N-Si-N linkages were broken up and reformed to a dense Si-N
×10 A/cm at 1 MV/cm and a high breakdown strength >3 MV/cm. On the applicability of our SiCxNyfilms in the copper/low-k
back-end interconnects, advantage and improvement are found in two crit-ical areas: (1) resist poisoning and (2) etch selectivity. It is proposed that the use of a single source precursor and low plasma power den-sity can alleviate or eliminate the resist poisoning, which required further experimental validation. Moreover, high etch selectivity can be achieved for our SiCxNy films deposited at 100–300◦C (C/Si
= 0.72–0.81), and even higher for SiCxNy film deposited at 400◦C
(C/Si= 0.29) if a low-k ILD with a high carbon content (C/Si ratio = 2.7) in Kume et al.’s study35is assumed.
Acknowledgments
The authors are grateful for the financial support by National Sci-ence Council of ROC, under Contract Nos.: NSC 99-2221-E-009-177 and NSC 100-3113-E-007-002-.
References 1. M. T. Bohr, Solid State Technol., 39, 105 (1996).
2. P. Gonon, A. Sylvestre, H. Meynen, and L. V. Cotthem, J. Electrochem. Soc., 150, F47 (2003).
3. A. S. Lee, N. Rajagopalan, M. Le, B. H. Kim, and H. M’Saad, J. Electrochem. Soc., 151, F7 (2004).
4. S. H. Rhee, M. D. Radwin, M. F. Ng, J. I. Martin, and D. Erb, Appl. Phys. Lett., 83, 2644 (2003).
5. M. A. Hussein and Jun He, IEEE Trans. Semicond. Manuf., 18, 69 (2005). 6. D. H. Zhang, L. Y. Yang, C. Y. Li, P. W. Lu, and P. D. Foo, Thin Solid Films, 504,
235 (2006).
17. J. Lubguban, T. Rajagopalan, N. Mehta, B. Lahlouh, S. L. Simon, and S. Gangopadhyay, J. Appl. Phys., 92, 1033 (2002).
18. W. C. Oliver and G. M. Pharr, J. Mater. Res., 7, 1564 (1992).
19. A. M. Coclite, G. Ozaydin-Ince, F. Palumbo, A. Milella, and K. K. Gleason, Plasma
Process. Polym., 7, 561 (2010).
20. Z. Chen, K. Prasad, C. Y. Li, S. S. Su, D. Gui, P. W. Lu, X. He, and S. Balakumar,
Thin Solid Films, 462–463, 223 (2004).
21. Y. Taki, T. Kitagawa, and Osamu Takai, Thin Solid Films, 304, 183 (1997). 22. I. Ferreira, E. Fortunato, P. Vilarinho, A. S. Viana, A. R. Ramos, E. Alves, and
R. Martins, J. Non-Cryst. Solids, 352, 1361 (2006).
23. M. Tada, H. Yamamoto, F. Ito, T. Takeuchi, N. Furutake, and Y. Hayashi, J.
Elec-trochem. Soc., 154, D354 (2007).
24. L. Favennec, V. Jousseaume, V. Rouessac, F. Fusalba, J. Durand, and G. Passemard,
Mater. Sci. Semicond. Process, 7, 277 (2004).
25. A. Zenasni, V. Jousseaume, P. Holliger, L. Favennec, O. Gourhant, P. Maury, and G. Gerbaud, J. Appl. Phys., 102, 094107 (2007).
26. C. Charles-Alfred and V. Jousseaume, Surf. Coat. Technol., 201, 9260 (2007). 27. Y. L. Cheng, J. Wu, T. J. Chiu, S. A. Chen, and Y. L. Wang, J. Vac. Sci. Technol. B,
29, 031207 (2011).
28. S. M. Sze and K. K. Ng, Physics of Semiconductor Devices, John Wiley & Sons, New York (2007).
29. C. C. Chiang, I. H. Ko, M. C. Chen, Z. C. Wu, Y. C. Lu, S. M. Jang, and M. S. Liang,
J. Electrochem. Soc., 151, G93 (2004).
30. J. Robertson and M. J. Powell, Appl. Phys. Lett., 44, 415 (1984).
31. K. Kobayashi, H. Yokoyama, and M. Endoh, Appl. Surf. Sci., 254, 6222 (2008).
32. A. Mallikarjunan, A. D. Johnson, L. Matz, R. N. Vrtis, A. Derecskei-Kovacs, X. Jiang, and M. Xiao, Microelectron. Eng., Available online 5 May 2011.
33. H. Nakao, S. Takei, T. Shinjo, and Y. Nakajima, Advances in Resist Technology
and Processing XXIII, edited by Qinghuang Lin Proc. of SPIE Vol. 6153, 61532O,
(2006).
34. T. Kropewnicki, K. Doan, B. Tang, and C. Bj¨orkman, J. Vac. Sci. Technol. A, 19, 1384 (2001).
35. I. Kume, M. Ueki, N. Inoue, J. Kawahara, N. Ikarashi, N. Furutake, S. Saitoh, and Y. Hayashi, Jpn. J. Appl. Phys., 50, 04DB02 (2011).
