International Journal of Impact Engineering 135 (2020) 103406

Contents lists available at ScienceDirect

International Journal of Impact Engineering

journal homepage: www.elsevier.com/locate/ijimpeng

Voitenko experiments with novel diagnostics detect velocities of 89 km/s T

a,⁎ b a a a
Douglas G. Tasker , Young K. Bae , Carl Johnson , Kevin Rainey , Christopher Campbell ,
Dave Oschwalda, Cornelius Reeda
a
Los Alamos National Laboratory, Los Alamos., NM 87545, USA
b
Y.K. Bae Corporation, Tustin, CA 92780-7709, USA

A R T I C LE I N FO A B S T R A C T

Keywords: Using a Voitenko accelerator, a series of experiments were performed with the goal of attaining shock velocities
Voitenko accelerator in gases approaching 90 km/s. The basic apparatus comprised a hemispherical bowl filled with a gas at atmo-
Gases spheric pressure; a metal piston across its diameter; and a small bore evacuated shock tube at its apex. The
Hypervelocity evacuated shock tube was separated from the gas bowl by a thin diaphragm. A combination of a plane wave
PDV
explosive lens and a high explosive pad accelerated the piston to a velocity of the order of 4 km/s and subse-
Microwave Doppler icrowave Doppler
quently compressed the gas in the bowl. The thin diaphragm at the other end of the bowl then ruptured and the
velocimetry
Metastable inner shell molecular state high pressure (shock compressed) gas escaped into the shock tube. A high speed digital framing camera recorded
MIMS the light emission from the shock front, photon Doppler velocimetry monitored the piston and, for the first time,
a 94-GHz microwave interferometer was used to monitor the position of the front. These diagnostics showed that
the shock velocity was 89 km/s in helium and there were interesting features, previously unreported, in the
shock propagation detected by the microwave interferometer.

1. Introduction achieved with a measured velocity of ~89 km/s as described below.

The original Voitenko experiments date back to 1964 [1–3] and
various designs have been reported [4]. A miniature version of these
experiments was reported previously by the authors [5]. Voitenko ob-
served velocities of 91 km/s with a hydrogen gas fill and ~80 km/s with
a helium fill [3]. In the miniature version the maximum shock velocity
attained in the gas was 41 km/s compared to the expected velocity of
80 km/s. It was subsequently understood that this velocity was limited
by the scale of the experiment and the consequent thickness of the
diaphragm.
Typically, the basic apparatus comprises a hemispherical bowl filled
with a gas at atmospheric pressure; a metal piston across its diameter;
and a small bore evacuated shock tube at its apex, Fig. 1. The evacuated
shock tube is separated from the gas bowl by a thin diaphragm. A
combination of a plane wave explosive lens and a high explosive pad
accelerates the piston to a velocity of the order of 4 km/s and subse-
quently compresses the gas in the bowl. The thin diaphragm at the
other end of the bowl then ruptures and the high pressure (shock
compressed) gas escapes into the shock tube.
The purpose of the experiments was to demonstrate that gas shock
velocities approaching 90 km/s could be attained with a Voitenko ap-
paratus; to this end three experiments were performed. This goal was
1.1. MIMS and particle acceleration
Methods of accelerating particles to high velocity and their impacts
with solid targets have long been of interest to various communities
including shock physics, high energy density physics and space travel.
Recently a new avenue of interest is the impact of nanoparticles
(1–10 nm in diameter) at velocities approaching 100 km/s. These par-
ticles undergo shock compression to stresses of ~10 TPa in times of the
order of 10 fs to 100 fs. Bae demonstrated experimentally that as these
excited particles relax they emit soft X-rays with high efficiency [6,7].
Bae proposed that under such conditions, because electron-ion
thermalization takes more than one picosecond while the compression
occurs in femtoseconds, the initial compression energy is mainly me-
chanical, i.e., the compression is “cold.” Under these unique conditions
the nanoparticles are compressed into the “Metastable inner shell mo-
lecular state” (MIMS), which is bound by inner shell electrons with
binding energies exceeding 100 eV. It is the decay of this MIMS state
that is accompanied by X-ray emission.
In Bae's experiments, low density fluxes of nanoparticles were ac-
celerated electrically. To study the physics of hypervelocity short
duration impact with larger fluxes, we propose to accelerate

⁎
Corresponding author.
E-mail address: tasker@lanl.gov (D.G. Tasker).

https://doi.org/10.1016/j.ijimpeng.2019.103406
Received 25 June 2019; Received in revised form 3 September 2019; Accepted 3 October 2019
Available online 04 October 2019
0734-743X/ Published by Elsevier Ltd.
D.G. Tasker, et al.
Fig. 1. Schematic of the Voitenko apparatus reported in this paper.
PWL = =plane wave lens; HE = =high explosive PBX-9501; MP = =metal
piston; CP = =cap plate; D = diaphragm; ST = =the evacuated shock-tube;
PDV = =photon Doppler velocimetry probes (2); MI = =microwave inter-
ferometer. The confinement was transparent PMMA. PMMA attenuators, not
shown, could be added between the metal plate and the explosive to reduce
spall, see Section 3.1. The distance between the diaphragm and the left of the
cap plate is 190.5 mm.
nanoparticles using a combination of explosive and electrical techni-
ques. A miniature explosively-driven shock tube, based on the Voitenko
compressor design [3], has been designed to produce shock speeds in
light gases in excess of 80 km/s. High precision photographic and mi-
crowave techniques are employed to examine the shock progression in
detail with nanosecond resolution. The electrical acceleration of na-
noparticles has yet to be employed in the Voitenko study. Here we only
report the generation of the high velocity gas flows.
2. Experiments
2.1. Previous work
In the miniature system originally performed by the authors [5] the
diameter of the shock tube was close to 1 mm. It is now understood that
the maximum observed velocity was likely limited by the thickness of
the diaphragm that separated the shock tube from the explosive gases.
The bore of the miniature shock tube was chosen to enable the SWIFT
[8] high speed photographic system to be used. Unfortunately, the
design was so tiny that it was extremely difficult to insert and align the
diaphragm across the tube. Moreover, the optimum thickness of the
diaphragm was near impossible to achieve, see the thickness rule
below. In this miniature Voitenko assembly, the maximum gas shock
velocity achieved was 41 km/s.
2.2. The diaphragm thickness rule
Voitenko showed that to attain the maximum gas shock velocity the
mass of the diaphragm should not exceed approximately 5% of the mass
of the gas in the hemispherical bowl. Above that thickness there is
steady decline in velocity, e.g., for a diaphragm with a mass that is 20%
that of the gas the velocity is decreased by 40% [2,3].
In the case of the miniature Voitenko, the diameter of the bowl was
6.35 mm and that of the shock tube 1 mm. At those small sizes the
enclosed mass of air at STP was ~500 µg and the maximum 5% dia-
phragm (mica) mass translated to thicknesses of <10.7 µm for air; si-
milarly it was <1.5 µm for helium and <780 nm for hydrogen. It was
extremely difficult to cleave the diaphragm material (mica) thinner
than 25 µm, yet nanometer thicknesses were required. Consequently,
the mica diaphragms used in the miniature apparatus with helium were
an order of magnitude too thick and the observed gas wave speeds were
low.
2
International Journal of Impact Engineering 135 (2020) 103406
2.3. Large scale design and diaphragm
To alleviate the problems of manufacturing diaphragms for the
miniature tubes, the shock tube bore was increased in diameter to
12.7 mm (from the previous 1-mm bore), which allowed the use of both
a 100-mm diameter gas bowl diameter and microwave velocimetry, see
below. With this diameter and the consequent increased mass of gas,
the maximum mica diaphragm thicknesses became: 35 µm in air; 5 µm
in helium; and 2.5 µm in hydrogen. Moreover, in this range of thick-
nesses, polyimide or polyester films could be used. Both polyimide and
polyester have lower densities than mica (hence larger maximum
thicknesses) and are readily available in a range of thicknesses, i.e., no
cleaving is required. For polyimide those maximum thicknesses become
73 µm, 11 µm, and 5.3 µm respectively. Fortuitously, 12-µm thick
polyimide film (without adhesive) was available, which was considered
close enough to the 5%-maximum thickness to be useful. (Note: unlike
mica, polyimide can stretch under shock loading so it may not rupture
as readily as mica, see the Conclusions.)
In the experiments reported here, the shock tube was evacuated to a
pressure of typically 500 mTorr. It was necessary to ensure that the 12-
µm thick polyimide diaphragm could withstand the differential pres-
sure of one atmosphere of gas pressure on the upstream side and va-
cuum on the downstream side. Consequently, all diaphragms were
tested statically in the apparatus.
Once the explosive is detonated, the working gas trapped between
the explosively-driven piston and the diaphragm is elevated to pres-
sures of the order of tens of GPa by multiple reflections and the sound
speed is estimated to be a few tens of km/s [1]. For a classic shock tube
the escape velocity of the gas, after the diaphragm ruptures, is primarily
determined by the sound speed of the gas c0 at its elevated pressure. From
fluid mechanics the escape velocity is given by [9]
2c0
Uescape = −
γ−1 (1)
where ϒ is the ratio of specific heats of the gas (~1.3 for hydrogen).
Clearly the maximum shock speed in the shock tube is primarily de-
termined by the sound speed of the gas at the elevated pressure. So, for
the highest escape velocity, hydrogen is preferred, followed by helium
and so on. From a practical standpoint, hydrogen is highly flammable
and so helium was chosen, albeit with a penalty in the maximum at-
tainable velocity.
2.4. Optical diagnostics
The SWIFT optical diagnostic [10] used in the previous studies at
the Los Alamos National Laboratory was designed for detonator-scale
experiments and could not accommodate explosive masses of more than
a few grams. The new 12.7-mm diameter apparatus required an ex-
plosive mass of 1.3 kg and was therefore too large for SWIFT; the ex-
periments were performed outdoors. However, the microwave diag-
nostic has no explosive mass limitation and allowed us to record
(successfully) both the velocity of the piston motion and the gas shock
with one diagnostic.
Two Photon Doppler Velocimetry (PDV) probes [11] were used to
directly record the velocity of the metal piston and determine piston
integrity. This was done to ensure uniformity of the drive and the ab-
sence of plate spall. In addition, a Shimadzu model HPV-X high speed
digital framing camera, operating at 5 × 106 frames per second (110 ns
exposure, 200 ns interframe time), was used to observe the luminous
shock front as it travelled along the shock tube. The shock front was
intensely bright, so to prevent saturation of the camera images the lens
was stopped down to F32 and the light was further attenuated with a
stack of neutral density filters equivalent to ND3.9.
D.G. Tasker, et al.
2.5. Millimeter-wave interferometer
Microwave interferometry has been used at the Los Alamos National
Laboratory for decades to study the detonation characteristics of ex-
plosives. These early systems were continuous wave (CW), typically
operated at 8 GHz, and were confined to coaxial cables or coplanar
waveguides, see McCall [12]. Circa 2010, using off-the-shelf compo-
nents and techniques similar to those described by Farrands, Glancy
[13,14] and many others, a new 94-GHz free-space interferometer was
developed to study projectiles, fragment fields, and the detonation ve-
locity within high explosives.
The 94 GHz interferometer, Fig. 2, was used for the measurement of
internal shock and detonation fronts (including shock-to-detonation
and deflagration-to-detonation transitions) within explosives, flyer
impacts on explosives, projectiles, and fragment fields. The system uses
a microwave source and a six-times multiplier to generate the 94-GHz
carrier. A circulator allows the use of a single antenna for the trans-
mission and receipt of data. The interferometer utilizes a mixer to
generate the in-phase (I) and quadrature (Q) signals which are recorded
by a digitizer. Conventional short-time Fourier transform analysis is
used to convert the spectral data to velocities, see for example [15].
The interferometer beam was projected into the shock tube through
the cap plate (Fig. 1) towards the piston. This was done to observe not
only the high velocity gas in the shock tube but also the acceleration of
the piston through the diaphragm, and the diaphragm rupture.
3. Results
3.1. PDV signatures of piston velocities and controll of spall
The first large scale experiment was performed without a plastic
attenuator between the explosive and the metal piston. The PDV velo-
city spectrogram showed evidence of spall, i.e., the piston was over-
driven, and the ultimate piston velocity was just 2250 m/s. That is, the
shock loading of the plate was sufficient to cause it to separate into
layers and perhaps fragments, the front layer (nearest the PDV) was
consequently decoupled from further explosive loading and the velocity
was low. The microwave interferometer detected a weakly reflecting,
short-lived jet (1 µs) travelling near 70 km/s.
The spall effects are functions of the duration and magnitude of the
explosive shock loading and the thickness of the piston. The simplest
modification for the next experiment was to reduce the magnitude and
increase the risetime of the shock by adding a PMMA attenuator. In the
second experiment a 5.66-mm PMMA attenuator plate was added be-
tween the explosive drive and the piston. Neither the explosive mass
nor the piston thickness were changed from those of the first experi-
ment. The attenuator was added to reduce the magnitude of the shock
and increase its risetime. The PDV velocity spectrogram, Fig. 3(a), de-
monstrated that this did improve the piston acceleration by reducing
spall. The resulting shape of the PDV velocity signature was a “stair-
case” that is typical of plate acceleration without spall, see for example
Ferguson [16]. However, notice the light blue cloud of velocity data
surrounding the main spectrogram in Fig. 3(a). This is symptomatic of a
cloud of ejecta from the surface of the piston which indicates that it
may still have been overdriven. In the third experiment the PMMA
3
International Journal of Impact Engineering 135 (2020) 103406
Fig. 2. Millimeter-wave interferometer. Key: MS - micro-
wave signal generator, Agilent E8247C; 6x M – six-fold
Active Multiplier Chain, Millitech AMC-10-RFHB0; C –
20 dB Coupler, Millitech CL3-10-R2000; M – I/Q Mixer,
Millitech MIQ-10-RLNA0; Cir – circulator, Quinstar QJY-
94023 W; A – 500 mW Power Amplifier, Quinstar QPN-
94022730-P2; LSA – level-set attenuator, Quinstar QAL-
W00000; Ant – Conical Horn Antenna, Sage Millimeter
SAC-2309–10-S2; Q and I – quadrature and in-phase
output signals.
attenuator thickness was further increased to 8.76 mm. The PDV data in
Fig. 3(b) show an additional gain in piston velocity to 4200 m/s and less
evidence of particle ejecta. The microwave interferometer detected
strong signals for both the second and third experiments, see below.
3.1.1. Observations of experiment 2
The microwave interferometry data of the second experiment are
shown in Fig. 4. The initial velocity of the jet in the shock tube was
50 km/s, near the time of 45 µs and it decelerated to a velocity of
38 km/s just before the end of the tube, at a time of 50 µs; at that time
the wave appeared to accelerate to 71 km/s with a spread of ± 3 km/s.
Note also another possible peak on the spectrogram approaching
100 km/s near 50 µs.
The camera data showed comparable velocities of the luminous
front (~45 km/s on average) and then a bright flash at the end of the
tube ahead of the luminous front, with a dark zone between the flash
and the luminous front; see the inset of Fig. 4. The bright flash appeared
coincident with the second 71-km/s wave observed by the microwave
diagnostics.
3.1.2. Observations of experiment 3
In the fastest experiment, experiment 3, the camera video shows a
bright luminous front travelled down the tube at near 35 km/s and then
a bright flash appeared ahead of the luminous front at the bottom of the
shock tube, Fig. 6, with a dark zone between the flash and the luminous
front. This behavior was similar to that of experiment 2.
The microwave data in the same figure exhibited a complex wave
behavior. The jet travel started with a relatively steady velocity of
35 km/s for the majority of the travel in the tube. Then, just before the
end of the tube, the velocity stepped up to several simultaneous states
in a staircase formation with velocities ranging from 35 to 55 km/s.
Finally, the wave apparently accelerated at the end of the tube to ve-
locities spread between 70 and 90 km/s. Note again another weaker
peak on the spectrogram near 100 km/s at 43.1 µs.
4. Discussion
4.1. Interpretation of imagery
In the video data from experiment 3, Fig. 6, the cool luminous front
can be seen travelling well ahead of the brightly heated zone. This
means the temperature of the jet front was probably below the ioni-
zation energy (24.6 eV) of helium; consequently it was neither visible
nor reflective. The real shock front is probably not visible in the video
data (possibly because of the large optical attenuation applied to the
lens) and could have been further advanced than the visible front. This
is interpreted to mean that the camera and microwave data near the
end of the tube are indeed real because the jet front (approaching
90 km/s) became detectable due to heating by the backward flowing
gas.
In other words, there was a fast, near invisible portion of the jet flow
travelling at near 90 km/s followed by a slower moving tail which was
hotter and brighter. The fast portion became visible as it ploughed into
the reflected gas flow at the end of the tube. In short, the wave did not
accelerate towards the end of the tube, it merely becomes more visible.
D.G. Tasker, et al. International Journal of Impact Engineering 135 (2020) 103406

Fig. 3. PDV piston spectrograms, velocities in m/s vs. time in µs. (a) With 5.66 mm thick PMMA attenuator, showing a typical staircase velocity rise in the absence of
spall. The initial velocity was 2700 m/s at 23 µs and the final velocity 3350 m/s. Note the light blue “cloud” of velocity data above and below the main spectrogram
which is evidence of particle ejecta. (b) With 8.76-mm thick PMMA attenuator. The initial velocity was 2900 m/s at 21 µs and the maximum velocity before impact at
the end was 4200 m/s. [On all spectrograms in this paper the color indicates the signal strength as in the vertical bar: blue at the bottom is low; red at the top is
high.]. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 4. Experiment 2: Microwave spectrogram (colors) with velocities in km/s vs time in µs; raw voltage data (MWave) overlaid in black (mV), right–hand scale. The
main signature decays to ~38 km/s near the time of 50 µs. F: velocity peak of between 68 and 74 km/s, likely corresponding to the bright flash at the end of the tube.
Top left inset: image from one frame of video camera showing the bright flash before the arrival of the main luminous front at the end of the tube.

(Unfortunately, the time of shock arrival at the cap plate was not re-
corded.)
4.1.1. Diaphragm rupture
It is likely that rupture of the thin diaphragm (described in
Section 1) was not instantaneous. We know from studies of shock
loaded polyimide [17] (i.e., the diaphragm material) that it can stretch
like a balloon before failure. It is therefore possible that the rupture
started as a small hole or crack that propagated across the width of the
diaphragm at a finite velocity. Consequently, the gas would initially
escape as a thin jet, rather than an instantaneous release across the full
diameter. Such a thin jet may be the origin of the invisible wave that
travelled ahead of the bright and reflective main front.
From this standpoint, polyimide may not be the best choice for a

4
D.G. Tasker, et al.
Fig. 5. Experiment 3: Microwave spectrogram (colors) with velocities in km/s vs time in µs; raw voltage data (MWave) overlaid in black (mV), right–hand scale.
There was a near-steady velocity of 35 km/s which continued until 42.5 µs when it jumped to 51 km/s. At 43.1 µs there was another jump to velocities of between 75
and 89 km/s (F) corresponding to the bright flash at the end of the tube. The camera data are shown in Fig. 6.
diaphragm. Note that mica is brittle by comparison. From the same
studies of shock loading polyimide [17] we know that Parylene C
fractures rather than stretching and can be vapor deposited [18], so it
may be a better choice for this application.
4.1.2. Detection of diaphragm rupture
It was difficult to detect any details of the piston motion and dia-
phragm rupture from the microwave data because the velocities were
too low to accurately discriminate them from the baseline. The data in
Fig. 5 show possible evidence of diaphragm rupture or piston move-
ment in the vicinity of 4 km/s at 37.7 µs, but the data are too uncertain
to be useful.
Fig. 6. Experiment 3: Five frames from the video, image to image time 1.6 µs
from left to right, showing the progress of a luminous front travelling down the
190.5-mm length of the shock tube. The hemisphere (not visible) was situated
at the top. Exposure time 110 ns per frame. (The video data of experiment 2
were similar but slower.) C: cool luminous front; F: the brilliant flash at the end
of the tube ahead of the bright luminous zone (B). The dots show the top and
bottom of the shock tube. The leftmost image corresponds to the time of
~37.8 µs in Fig. 5.
5
International Journal of Impact Engineering 135 (2020) 103406
5. Conclusions
The combination of the photographic and microwave data provides
strong evidence of the existence of wave speeds of 90 km/s or more in
these Voitenko experiments. There are clearly many features of the
experiments that are not fully understood and merit further study.
Moreover, the explosive drive system should be optimized to get the
highest plate velocity without spall.
Despite the good results, polyimide may not be the best choice of
diaphragm because it stretches easily and probably does not rupture
instantaneously; Parylene C may be a better choice. (Mica also does not
stretch easily. It would be a good choice if a viable technique for
cleaving it to the required micron-scale thicknesses could be developed,
or if it could be purchased in those thicknesses.)
The microwave interferometry data provided a rich suite of in-
formation; it is clearly a powerful diagnostic for such experiments.
Additional diagnostics to record shock arrivals time and correlate them
with the microwave signatures would be helpful in future experiments.
Acknowledgments
This collaborative study by the Los Alamos National Laboratory and
the Y.K. Bae Corporation was supported by a contract from US Defense
Threat Reduction Agency (DTRA). We are especially grateful to Drs. J.
Davies and W. Wilson of DTRA for their generous support. LA-UR-19-
20670 Version 3.
Supplementary materials
Supplementary material associated with this article can be found, in
the online version, at doi:10.1016/j.ijimpeng.2019.103406.
D.G. Tasker, et al.
References
[1] Voitenko AE, Kirko V. Efficiency of a high-explosive plasma compressor. Combust
Explos Shock Waves 1975;11(6):813–5.
[2] Voitenko AE. Acceleration of a gas compressed under conditions of acuteangled
geometry. J Appl Mech Tech Phys 1966;7(4):79–82.
[3] Voitenko AE. Generation of high speed gas jets. Dokl Akad Nauk SSSR
1964;158(6):1278–80.
[4] https://en.wikipedia.org/wiki/Voitenko_compressor.
[5] Johnson CE, Murphy MJ, Tasker DG, Lieber MA. Experiments on a miniature hy-
pervelocity shock tube. Proc. APS Shock Conference, American Physical Society,
Shock Compression of Condensed Matter, Topical Group. 2013.
[6] Bae Y. Metastable innershell molecular state (MIMS). Phys Lett A
2008;372(29):4865–9.
[7] Bae YK. Table-top generation and spectroscopic study of ~10 TPa high-energy
density materials with C + 60 hypervelocity (v ~ 100 km/s) impact. Proc. Shock
Compression of Condensed Matter - 2013. 2013.
[8] Murphy MJ, Clarke SA. Ultra-High-Speed imaging for explosive-driven shocks in
transparent media. In: Chalivendra V, Song B, Casem D, editors. Dynamic Behavior
of Materials, volume 1. New York: Springer; 2013. p. 425–32.
[9] Harlow FH, Amsden AA. Fluid dynamics, a LASL monograph. Book Fluid Dynamics,
A LASL Monograph. Los Alamos Scientific Laboratory, University of California;
1971.
[10] Murphy MJ, Johnson CE. Preliminary investigations of HE performance
International Journal of Impact Engineering 135 (2020) 103406
characterization using SWIFT. Proc. APS Shock Conference, American Physical
Society, Shock Compression of Condensed Matter, Topical Group. 2013.
[11] Strand OT, Goosman DR, Martinez C, Whitworth TL, Kuhlow WW. Compact system
for high-speed velocimetry using heterodyne techniques. Rev. Sci. Inst.
2006;77(8):083108.
[12] McCall, G.H., Bongianni, W.L., and Miranda, G.A.: ‘Microwave interferometer for
shock wave, detonation, and material motion measurements’, Rev Sci Instrum1985,
56, (8), pp. 1612–8.
[13] Farrands JL, Cawsey GF. Microwave observation of detonation. Nature
1956;177(4497):34–5.
[14] Glancy, B.C., Sandusky, H.W., and Krall, A.D.: ‘Microwave interferometry of shock
waves. II. Reacting porous media’, J Appl Phys1993, 74, (10), pp. 6328–34.
[15] Kittell, D.E., J.O. Jr., M, and Son, S.F.: ‘Using time-frequency analysis to determine
time-resolved detonation velocity with microwave interferometry’, Rev Sci
Instum2015, 86, (4), pp. 044705.
[16] Ferguson JW, Taylor P. Application of heterodyne velocimetry and pyrometry as
diagnostics for explosive characterisation. J Phys Conf Ser 2014;500(14):142014.
[17] Sanchez NJ. Dynamic initiator imaging at the advanced photon source: under-
standing the early stages of initiator function and subsequent explosive interactions.
Proc. 20th Biennial Conference of the APS Topical Group on Shock Compression of
Condensed Matter. 2017.
[18] Gaynor JF, Jay Senkevich J, Desu SB. A new method for fabricating high perfor-
mance polymeric thin films by chemical vapor polymerization. J Mater Res
1996;11(7):1842–50.