Molecular Caulking: A pore sealing CVD polymer for ultralow k dielectrics

Article abstract of DOI:10.1149/1.1751195

Porosity has been introduced in existing low-k interlayer dielectrics to further reduce their dielectric constant. It is desirable to deposit a metallic layer on top of the porous dielectric by chemical vapor deposition (CVD). However this presents the challenge of preventing the precursor from penetrating into the porous dielectric and depositing metal within this insulating layer. In the present paper a low-k CVD polymer capping (Molecular Caulking) is deposited at room temperature onto the porous ultralow k dielectric methyl silsesquioxane. Experiments show that the Molecular Caulking prevents precursor penetration during subsequent metallorganic CVD. In addition, while the Molecular Caulking itself slightly penetrates into the methyl silsesquioxane, it does not appreciably increase surface roughness or film dielectric constant.

Full text of DOI:10.1149/1.1751195

Journal of The Electrochemical Society, 151 ͑7͒ F157-F161 ͑2004͒
F157
0
013-4651/2004/151͑7͒/F157/5/$7.00 © The Electrochemical Society, Inc.
Molecular Caulking
A Pore Sealing CVD Polymer for Ultralow k Dielectrics
Christopher Jezewski,a,z Christopher J. Wiegand, Dexian Ye,
b
b
b
b
c
a
Anupama Mallikarjunan, Deli Liu, Chowming Jin, William A. Lanford,
Gwo-Ching Wang, Jay J. Senkevich, and Toh-Ming Lu
b
b
b
a
Department of Physics, University of Albany, Albany, New York 12222, USA
Department of Physics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute, Troy,
b
New York 12180, USA
c
Texas Instruments, Incorporated, Dallas, Texas 75265, USA
Porosity has been introduced in existing low-k interlayer dielectrics to further reduce their dielectric constant. It is desirable to
deposit a metallic layer on top of the porous dielectric by chemical vapor deposition ͑CVD͒. However this presents the challenge
of preventing the precursor from penetrating into the porous dielectric and depositing metal within this insulating layer. In the
present paper a low-k CVD polymer capping ͑Molecular Caulking™͒ is deposited at room temperature onto the porous ultralow
k dielectric methyl silsesquioxane. Experiments show that the Molecular Caulking prevents precursor penetration during subse-
quent metallorganic CVD. In addition, while the Molecular Caulking itself slightly penetrates into the methyl silsesquioxane, it
does not appreciably increase surface roughness or film dielectric constant.
©
2004 The Electrochemical Society. ͓DOI: 10.1149/1.1751195͔ All rights reserved.
Manuscript submitted June 11, 2003; revised manuscript received January 18, 2004. Available electronically May 20, 2004.
In future gigascale integrated circuits ͑ICS͒ resistive-capacitive
RC͒ delay is an increasingly important issue. Carbon-doped oxides
final cure in an N ambient at 420°C. The resulting films contain
2
1
͑
50% porosity, and the pores are interconnected. Pore size ranges
and aromatic polymers are examples of materials being investigated
from 0.5-4 nm, with an average pore size of 1.5 nm. MSQ has a
2
8
to replace SiO as the interlayer dielectric ͑ILD͒. Both materials
2
nominal stoichiometry of SiO1.5(CH)0.5
.
possess lower dielectric constants and will lower the contribution to
RC delay. In order to reduce the dielectric constant further, it is
generally accepted that the ILD will contain some amount of poros-
ity. The introduction of porosity results in a number of other unde-
sirable properties such as a reduction in mechanical strength and
susceptibility to penetration of chemicals. Most importantly, during
chemical vapor deposition ͑CVD͒ exposure of the porous dielectric
to gaseous precursors that are expected to infiltrate an open pore
II
Copper CVD experiments were done via Cu ͑tmhd) and H in a
2
2
vertical, low pressure, warm-wall reactor. The precursor bubbler was
held at a constant temperature of 127.5 Ϯ 0.6°C and delivered with
15 sccm of argon carrier gas. The substrate was kept at 217
Ϯ 5°C and the chamber walls and precursor transfer lines all held
at 150 Ϯ 5°C. The total pressure of argon, H , and precursor, was
2
approximately 2 Torr. The deposition time was 30 min for all ex-
periments. Bare MSQ and several MC/MSQ films of varying MC
thickness were placed side-by-side on the substrate heater in each
experiment. For further details on the copper CVD process em-
ployed in this study, we refer to an earlier publication.6
film or even a closed pore film if the pore wall thickness in the
nanoscale dimensions3-6 cause degradation of film properties. CVD
will typically have a reactive sticking coefficient much less than one
in order to have good conformal coverage. Indeed, one way to re-
duce this penetration would be to increase the reactive sticking co-
efficient, e.g., by increasing the deposition temperature. However,
this would reduce the conformality of the deposition. Clearly poros-
ity and conformal CVD on high aspect ratio substrates are funda-
mentally at odds.
Cobalt deposition experiments were performed in a vertical, low
pressure, warm-wall reactor. Co (CO) was sublimed at room tem-
2
8
perature. The substrate was kept at 60 Ϯ 2°C and the deposition
time was 2 min. No carrier or purge gas was used, and deposition
pressure was approximately 18 mTorr. Bare MSQ and several MC/
MSQ films of varying MC thickness were placed side-by-side on the
substrate heater in each experiment.
Several methods have been studied recently to solve this prob-
lem. A recent review details some of the currently proposed strate-
9
7
MC thin films were deposited using the Gorham method. The
gies to seal porous dielectrics. A new sealing layer, Molecular
reactor consisted of a sublimation furnace, a pyrolysis furnace, and a
bell-jar-type deposition chamber. Base pressure in the deposition
Caulking, is presented here. MC films are deposited by CVD at
room temperature using a free radical polymerization process. Pre-
liminary results are promising. The approach taken was to measure
the new sealant’s ability to prevent penetration of metal precursors
copper, cobalt͒ during CVD. The depth distribution of deposited
metals were measured by Rutherford backscattering spectrometry
RBS͒. In addition, changes in the dielectric constant as a result of
Ϫ6
chamber was at mid 10 Torr. During growth the deposition cham-
ber pressure was in the low mTorr range and deposition rates were
between 0.09-0.14 Å/s. A detailed description of the reactor and
͑
deposition process has been described elsewhere.1
0,11
Briefly, the
precursor ͓2.2͔-paracyclophane was sublimed at a temperature of
55°C. The sublimed precursor flew into a high-temperature region
650°C͒ of the reactor inlet where it was quantitatively cleaved into
͑
1
͑
MC, were determined by metal insulator semiconductor ͑MIS͒ ca-
pacitance measurements. Deposited film thickness was determined
by both ellipsometry and ion beam backscattering using the 5.75
two p-xylylene monomers by vapor-phase pyrolysis. These reactive
intermediates were then transported to a room temperature deposi-
tion chamber where upon condensation, spontaneous polymerization
took place. Linear chains of poly͑p-xylylene͒ with unterminated end
groups were formed. Bulk poly͑p-xylylene͒ has a k value of ap-
proximately 2.65 perpendicular to the plane of the film.12 In this
work ultrathin films ͑1-5 nm͒ were deposited. Si͑100͒ 50 ⍀ cm
substrates were rinsed in ethanol, followed by deionized water,
blown dry with nitrogen, and then placed side-by-side with the po-
rous MSQ in the deposition reactor. Ultrathin poly͑p-xylylene͒ films
4
12
MeV He elastic nuclear resonance of C. Finally, the effect of MC
deposition on surface topology was measured by atomic force mi-
croscopy ͑AFM͒.
Experimental
The porous methyl silsesquioxane ͑MSQ͒ film was deposited by
spin coating and went through a series of baking stations before the
͑1-5 nm͒ are low molecular weight and more oligomeric than poly-
meric. Molecular weight increases with film thickness. Annealing
thin poly͑p-xylylene͒ films has shown indications of conversion
z
E-mail: jezewc@rpi.edu
Downloaded on 2015-05-15 to IP 193.61.135.34 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

F158
Journal of The Electrochemical Society, 151 ͑7͒ F157-F161 ͑2004͒
from low polymer to high polymer.13 It is advantageous to have the
pore sealant to be a high polymer since it possesses more robust
chemical bonds. In addition, unreacted monomers or larger
oligimers that are volatile at higher temperatures will diffuse out of
the film upon annealing. Therefore after deposition, samples were
annealed in forming gas at 250°C for 30 min.
The deposited film thickness was measured by a variable angle
spectroscopy ellipsometer ͑VASE, J. A. Wollam, Lincoln, NB͒ on
the silicon samples. VASE measurement interpretation is difficult on
MSQ films, so thickness measurements were used from the silicon
wafers and assumed similar growth on the MSQ films. The thickness
of MC was determined by using the Cauchy coefficients of poly͑p-
xylylene͒ (A ϭ 1.6, B ϭ 0.01) or an index of refraction of 1.458
n
n
at 634.1 nm.
Copper growth was characterized by RBS with the 4.0 MV Dy-
namitron accelerator at the Ion Beam Laboratory, Department of
Physics, University at Albany. Measurements were made with 2.0
4
MeV He particles. The RBS-determined areal density was con-
verted into an equivalent thickness by dividing by the bulk atomic
2
2
3
density of copper 8.45 ϫ 10 atom/cm . Spectra were collected
2
with a 20 mm area beam spot, 2-4 ␮C of charge, and with 3 nA of
Figure 1. RBS spectra showing penetration of copper into the interface on
bare MSQ/SiO2 /Si and 1.1 nm thick MC/MSQ/SiO2 /Si. The double arrows
show the thickness of the MSQ film as determined by the width of the silicon
signal in MSQ. The double arrow length also corresponds with the signal
peak of penetrated copper.
4
current. Ion beam backscattering using the 5.75 MeV He elastic
nuclear resonance of 12C was done at the same facility.
Two samples were chosen for extraction of the k value, 540 nm
MSQ/5 nm SiO /Si ͑control sample͒ and 250°C annealed 3.7 nm
2
MC/540 nm MSQ/5 nm SiO /Si. Top aluminum dots of 0.5, 1, and
2
1
.5 mm diam were electron-beam evaporated via a shadow mask. In
order to get good ohmic contact at the back side of the silicon wafer,
00 nm of aluminum was sputter deposited. The capacitance-voltage
C-V) characteristics of the Al/low-␬ ͑stack͒/5 nm SiO /Si struc-
If MC deposition passivates the MSQ/SiO2 interface, there
would be no deposition to signify the penetration of the copper
precursor. A nonselective metal precursor Co (CO ) was chosen to
3
(
2
2
8
tures were measured with a HP 4280A 1 MHz capacitance meter/CV
plotter. At least five measurements were performed for each capaci-
tor size for each sample.
Surface morphology was measured using an atomic force micro-
scope ͑AFM, AutoProbe CP͒ made by Park Scientific Instruments,
TM Microscope. A triangular silicon cantilever with silicon conical
tip ͑Veeco Metrology Group͒ was used in noncontact mode to mea-
sure the surface topography. The tip radius of curvature is Ͻ10 nm
and had a half apex angle of 12°.
provide further evidence that MC/MSQ/SiO /Si physically blocks
2
the pores. Cobalt CVD is nonselective and readily grows on MC.
Deposition was approximately 0.5 nm/min at 60°C, a temperature
where cobalt deposition is surface-reaction controlled and thus ex-
pected to grow at the surface and penetrate the porous MSQ. Figure
3
shows RBS spectra of the three samples, Co/2.7 nm
MC/MSQ/SiO /Si, Co/2.7 nm MC/MSQ/SiO /Si, and Co/bare
2
2
MSQ/SiO /Si. The bare MSQ shows penetration of cobalt sup-
2
ported by the long tail of the cobalt surface peak, while the other
samples have only surface deposition. The width of penetrated co-
Results and Discussion
From previous work, copper was shown to deposit inside the
6
porous dielectric rather than on the surface during CVD. The cop-
per precursor penetrated the interconnected porous MSQ and selec-
tively deposited at the interface between the MSQ film and the sili-
con substrate. This was observed by RBS and scanning electron
microscopy ͑SEM͒. RBS can be used to measure the amount of
penetrated material quantitatively down to monolayer levels. The
SEM only has a sensitivity of several nanometers, so RBS was used
to see small amounts of precursor that penetrated through the MC
sealant. To illustrate this, Fig. 1 shows RBS spectra after copper
deposition on bare MSQ/SiO /Si and a 1.1 nm MC/MSQ/SiO /Si
2
2
sample. The arrows labeled surface copper and surface silicon show
the kinematic energy of backscattering from copper and silicon, re-
spectively, at the surface. The observed peak at about 1400 KeV
shows that copper is deposited at the interface between the MSQ
and the SiO /Si substrate.
2
Figure 2 is a plot of the amount of RBS-determined copper de-
posited by CVD at this MSQ/SiO interface as a function of MC
2
thickness. Data from two separate experiments are shown in Fig. 2
͑
circles and triangles͒. The data in Fig. 2 show that a MC film
thickness of only 1.1 nm resulted in a 96% reduction in copper
penetration during CVD. After 3.5 nm MC, the copper penetration
goes below the 0.05 nm detection limit of RBS. Reactor modifica-
tions, including precursor and purge gas inlet lines, resulted in dif-
ferent growth rates of the two experiments, but the MC liner thick-
ness sufficient to prevent penetration was consistent.
Figure 2. RBS is used to determine the equivalent thickness of copper depo-
sition on/in MC/MSQ/SiO /Si as a function of MC thickness. Two data sets:
2
circles and triangles are from experiments before and after reactor alter-
ations, respectively. Data point size is larger than error bars.
Downloaded on 2015-05-15 to IP 193.61.135.34 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 151 ͑7͒ F157-F161 ͑2004͒
F159
Figure 3. RBS spectra of cobalt on 2.7 nm MC/Si, 2.7 nm
MC/MSQ/SiO /Si, and bare MSQ/SiO /Si. Cobalt penetration is only ob-
2
2
Figure 5. 4He 5.75 MeV ion beam backscattering spectra of
served on the bare MSQ/SiO /Si substrate. The 2.7 nm MC/Si and 2.7 nm
2
MC/MSQ/SiO /Si and bare MSQ/SiO /Si samples. View has been ex-
2
2
MC/MSQ/SiO /Si samples show cobalt deposition only at the surface.
2
panded to show only the carbon region of spectra, and the silicon substrate
contribution has been subtracted from both spectra to emphasize relative
carbon content in each sample.
balt is consistent with the silicon width ͑thickness͒ in MSQ and
shows that cobalt penetrates completely through the MSQ film with-
out MC sealing.
Figure 4 shows a schematic representation of how the MC layer
may be sealing the pores in MSQ. MSQ films used here have a pore
size distribution of 0.5-4 nm. From Fig. 4, one expects that signifi-
cant blocking of pores would require a MC thickness on the order of
the average pore diameter, consistent with our result of 1.1 nm in
Fig. 2.
the complexity of measuring the thickness deposited on
MSQ/SiO /Si. Given that during CVD, precursor gases can diffuse
2
into the porous material, some penetration of the MC layer was
expected. The quantitative measure of deposited film content in/on
the porous MSQ/SiO /Si was determined by adding a stoichio-
2
metric atomic percentage of hydrogen (C8H8) to the measured car-
4
Capacitance measurements showed that the dielectric constant of
bon content as determined by He ion beam backscattering analysis
1
2
a MSQ/SiO /Si film covered with 3.7 nm of MC had essentially the
at 5.75 MeV. C exhibits a strong ͑␣,␣͒ elastic scattering resonance
in the energy region between 5.5-5.8 MeV. In this energy region the
cross section is more than 100 times the Rutherford cross section.
The resonance is sufficiently broad to enable depth profiling over
depths on the order of micrometers. Essentially ion beam analysis at
these energies is similar to RBS at 2 MeV, providing there is accu-
rate absolute cross section data available at the energies used.
2
same dielectric constant as the bare MSQ/SiO /Si film. The effec-
2
tive dielectric constant of the low k stack was calculated from the
slope of the measured accumulation capacitance vs. capacitor area.
The measured dielectric constant of porous MSQ was 2.26, this
increased to 2.30 after an annealed MC coating. The ϳ1.7% in-
crease in dielectric constant after annealing at 250°C was less than
the sample-to-sample variation in capacitor area.
1
4
Davies et al. have measured the non-Rutherford cross sections in
the energy region of 5.5-5.8 MeV.
The MC thickness as determined by ellipsometry on SiO is an
2
Measurements of carbon content were made on bare
indirect method, but was meant for quantitative comparison due to
MSQ/SiO /Si and 3.5 nm MC/MSQ/SiO /Si samples. Amorphous
2
2
carbon ͑H-Square, Ca͒ was used as a standard. Amorphous carbon is
dense and has a mirror smooth surface. No detectable bulk contami-
nation was determined by RBS, and hydrogen depth profiling
showed no bulk contamination of hydrogen at a detection limit of
ϳ0.1%. Given that bare MSQ/SiO /Si and MC/MSQ/SiO /Si are
2
2
approximately the same thickness, the carbon content of
MC/MSQ/SiO /Si was obtained by comparing the carbon contents
2
in both films using the backscattering geometry and cross-sectional
data determined by the amorphous carbon standard. An increase in
1
7
2
carbon content of 1.54 ϫ 10 atom cm was found for the 3.5 nm
MC/540 nm MSQ/SiO /Si film. Adding a stoichiometric quantity of
2
17
hydrogen (C H ) brings the total areal density to 3.08 ϫ 10 atom
cm . The bulk density of MC is 1.02 ϫ 10 atom cm , so an
8
8
2
23
3
equivalent MC thickness of approximately 30 nm was deposited
in/on MSQ/SiO /Si. Figure 5 shows the substrate subtracted ion
2
beam backscattering spectra. Qualitatively it can be seen that the
carbon content of the front two-thirds of the film has a statistically
significant increase in carbon content. To estimate the increase in
dielectric constant, a uniform penetration of the MC layer was as-
sumed. The MSQ films are 540 nm thick and contain 50% porosity,
so an MC equivalent thickness of 30 nm evenly distributed through
Figure 4. Illustration of possible MC deposition behavior in the connected
pore structure of porous MSQ.
Downloaded on 2015-05-15 to IP 193.61.135.34 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

F160
Journal of The Electrochemical Society, 151 ͑7͒ F157-F161 ͑2004͒
Figure 6. AFM measurements: ͑a͒ bare
MSQ/SiO /Si, ͑b͒ MC/MSQ/SiO /Si, ͑c͒
2
2
plot of rms roughness as a function of MC
thickness, and ͑d͒ lateral correlation
length as a function of MC thickness.
the MSQ film would fill 11.1% of the porosity. An upper bound to
the calculation of the composite film dielectric constant can be
found by adding the contribution of components in parallel
whereas the lateral correlation length continues to increase. As the
oligimer grows it tends to align itself parallel to the substrate sur-
face.
1
5
Vaeth et al.18 showed inhibited polymer growth on a copper sur-
face. 335 nm of polymer grew on a silicon substrate before any
growth occurred on copper. Selectivity is a highly desired quality of
the free radical polymerization process used here. For dual dama-
scene structures in the back end of line processes there would be no
growth on the underlying copper layer. Only the dielectric would be
coated, and there would be no series contribution to ohmic contact
from the MC film. This would alleviate the need to etch back the
dielectric liner and therefore reduce the number of processing steps.
Thermal stability of poly͑p-xylylene͒ films deposited with the
ktot ϭ k P ϩ k P ϩ k P ,
͓1͔
1
1
2
2
3
3
where ktot , k , k , and k are the dielectric constants of the total
1
2
3
film, air, MC, and the dense MSQ, respectively. P , P , and P are
1
2
3
the respective fractions of air, MC, and the dense MSQ. An increase
of approximately 4% in the dielectric constant should be expected.
Capacitance-determined dielectric constant increase of 1.5 Ϯ 3.3%
for a 3.7 nm thick MC/MSQ/SiO /Si is consistent with these re-
2
sults.
19
reactor used in these studies has been studied previously. Thermal
Roughness was measured by AFM and quantitative information
about surface morphology was extracted from a height-height cor-
relation function defined elsewhere.16 Figure 6a and b show the
desorption studies ͑TDS͒ revealed the onset of decomposition at
about 400°C. In addition, small thickness changes due to isothermal
annealing in vacuum were detectable after 2 h at 410°C. Future
work will test the thermal stability of poly͑p-xylylene͒ films as used
in this application.
AFM images of bare MSQ/SiO /Si and 5 nm MC/MSQ/SiO /Si
2
2
samples, respectively. At 4 nm, a suitable MC thickness to prevent
penetration, the root mean square ͑rms͒ roughness is 0.62 nm greater
Conclusions
than the bare MSQ/SiO /Si sample. The morphology after deposi-
2
tion shows an initial increase in roughness followed by an apparent
smoothing as deposition proceeds ͑Fig. 6c͒. The lateral correlation
length continues to increase with deposition thickness and increases
A polymer liner as thin as 3.5 nm has been shown to prevent
copper penetration during CVD. The liner has a low dielectric con-
stant ͑2.65͒ and effectively seals the porous dielectric. The amount
of liner penetration resulted in a negligible increase in dielectric
constant as determined by capacitance measurements. AFM results
show a 0.62 nm increase in rms roughness of the sealed MSQ sur-
face. The sealant is a selective CVD process and will not deposit on
the underlying Cu surface. Selective growth in this manner reduces
the number of process steps by possibly eliminating the need to etch
back the liner before further metallization steps in a via or trench
structure.
from an initial 20 nm for the bare MSQ/SiO /Si to 44.6 nm in the 5
2
nm thick MC/SiO /Si coated sample. An increase in the lateral cor-
2
relation length indicates that as the film grows there is some long
range smoothing. The lateral correlation length is roughly the wave-
length of fluctuation of the surface. The lateral correlation length
changes are evident in the AFM images of bare MSQ ͑Fig. 6a͒ and
5
nm MC coated MSQ ͑Fig. 6b͒ by an overall increase in surface
feature size.
During deposition of p-xylylene monomer (-C H -), two chemi-
The free radical polymerization film deposition process used here
contains beneficial fundamental properties, such as, initial reactions
of condensation followed by two to three orders of magnitude
8
8
cal reactions can occur at the surface, initiation, in which new poly-
mer molecules/chains are generated, and propagation, in which ex-
1
7
17
isting polymer molecules are extended to higher molecular weight.
It is, perhaps, this effect which results in the decrease in roughness,
growth rate increase by subsequent propagation. The rapid growth
rate after initiation should aid in quickly sealing off the top layers to
Downloaded on 2015-05-15 to IP 193.61.135.34 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 151 ͑7͒ F157-F161 ͑2004͒
F161
further penetration and thus result in a conformal low dielectric
constant liner. In the future we will present results on alternative
sealants including adhesion and thermal stability.
and Interfaces, American Vacuum Society ͑AVS͒, p. 52 ͑2002͒.
. C. Jezewski, W. Lanford, J. J. Senkevich, D. Ye, and T.-M. Lu, Chem. Vap. Depo-
sition, 9, 305 ͑2003͒.
. K. Maex, M. R. Baklanov, D. Shamiryan, F. Iacopi, S. H. Brongersma, and Z. S.
Yanovitskaya, J. Appl. Phys., 93, 8793 ͑2003͒.
. M. Morgen, E. T. Ryan, J.-H. Zhao, C. Hu, T. Cho, and P. S. Ho, Annu. Rev. Mater.
Sci., 30, 645 ͑2000͒.
6
7
8
9
Acknowledgments
Support of this study is provided by the SRC Center for Ad-
vanced Interconnect Systems and Technologies ͑CAIST͒.
. W. Gorham, J. Polym. Sci., Part A: Polym. Chem., 4, 3027 ͑1966͒.
0. J. Fortin and T.-M. Lu, J. Vac. Sci. Technol. A, 18, 2459 ͑2000͒.
1
1
1
1. J. B. Fortin and T.-M. Lu, Chem. Mater., 14, 1945 ͑2002͒.
Rensselaer Polytechnic Institute assisted in meeting the publication costs
of this article.
2. J. B. Fortin and T.-M. Lu, Chemical Vapor Deposition: The Growth and Properties
of Parylene Thin Films, p. 58, Kluwer Academic Publisher, Norwell, MA ͑2004͒.
3. J. J. Senkevich, G.-R. Yang, and T.-M. Lu, Colloids Surf., A, 216, 167 ͑2003͒.
4. J. A. Davis, F. J. D. Almeida, H. K. Haugen, R. Siegele, J. S. Forster, and T. E.
Jackman, Nucl. Instrum. Methods Phys. Res. B, 85, 28 ͑1994͒.
1
1
References
1
2
. S. P. Murarka, Mater. Sci. Eng., R., 19, 87 ͑1997͒.
. The National Technology Roadmap for Semiconductors, 2001 ed., Semiconductor
Industry Association, Santa Clara, CA ͑2001͒.
15. D. E. Aspnes, Thin Solid Films, 89, 249 ͑1982͒.
16. Y.-P. Zhao, G.-C. Wang, and T.-M. Lu, Experimental Methods in the Physical
Sciences: Characterization of Amorphous and Crystalline Rough Surfaces: Prin-
ciples and Applications, Vol. 37, p. 18, Academic Press, New York ͑2000͒.
17. W. F. Beach, Macromolecules, 11, 72 ͑1978͒.
18. K. M. Vaeth and K. F. Jense, Chem. Mater., 12, 1305 ͑2000͒.
19. J. P. Fortin and T.-M. Lu, Chemical Vapor Deposition Polymerization: The Growth
and Properties of Parylene Thin Films, pp. 70-76, Kluwer Academic Pub., Nor-
well, MA ͑2004͒.
3
4
. M. E. Thomas, D. M. Smith, S. Wallace, and Nancy Iwamoto, in Proceedings of the
International Interconnect Technology Conference (IITC), p. 223 ͑2002͒.
. W. F. A. Bresling, A. Satta, J. Schumacher, T. Abell, V. Sutcliffe, A.-M. Hoyas, G.
Beyer, D. Gravesteijn, and K. Maex, in Proceedings of the International Intercon-
nect Technology Conference (IITC), p. 288 ͑2002͒.
5
. W. F. A. Bresling, A. Satta, J. Schumacher, A. M. G. Beyer, K. Maex, O. Kilpella,
and H. Sprey, in Proceedings of 3rd International Conferences on Microelectronics
Downloaded on 2015-05-15 to IP 193.61.135.34 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).