Upload
vothu
View
218
Download
0
Embed Size (px)
Citation preview
𝑃𝑃 = (∑𝑑𝑑𝑖𝑖 ∕ 𝑑𝑑𝑃𝑃𝑖𝑖
𝑛𝑛
𝑖𝑖=1
)
−1
Andreas Hogg
DEVELOPMENT AND CHARACTERISATION OF ULTRATHIN LAYER PACKAGING FOR IMPLANTABLE MEDICAL DEVICES
Graduate School for Cellular and Biomedical Sciences
University of Bern
Development and Characterisation of Ultrathin Layer Packaging for Implantable Medical Devices
PhD Thesis submitted by
Andreas Hogg
Thesis advisor
PD Dr. Jürgen Burger Institute for Surgical Technology and Biomechanics
Faculty of Medicine of the University of Bern
Co‐advisor
Prof. Dr. Dr. Rolf Vogel ARTORG Center for Biomedical Engineering Research
Faculty of Medicine of the University of Bern
ii
iii
Accepted by the Faculty of Medicine, the Faculty of Science, and the Vetsuisse
Faculty of the University of Bern at the request of the Graduate School for
Cellular and Biomedical Sciences
Bern, Dean of the Faculty of Medicine
Bern, Dean of the Faculty of Science
Bern, Dean of the Vetsuisse Faculty Bern
iv
v
It is only a question of matching.
‐ Herbert Keppner
vi
vii
Keywords
Packaging, biomedical implants, implantable devices, electronic devices, long‐
term, corrosion protection, biocompatibility, conformity.
Barrier layer, protective multilayer, thin film, coating, deposition, CVD, PECVD.
Tightness, diffusion, helium diffusion, permeation, water permeation, water
vapour permeation, calcium, calcium mirror test, calcium corrosion.
Parylene, silicon dioxide, SiO2, SiOx, HMDS, HMDSO, organosilicon, silane.
Abstract
State of the art implantable medical devices are packaged by conventional glass
or metal housings which are reliable but limited in the miniaturization
potential, as well as cost‐intensive. Hermetic and biocompatible thin‐film
packaging is a key technology for further miniaturization of smart micro‐
medical implants (SMMS). The development of innovative thin‐film and ultra‐
tight packaging would permit substantial miniaturisation of implantable sensor
and actuator systems. This results in less‐invasive surgical interventions and
ultimately better patient quality of life. It is hypothesised that multilayered thin
films will provide superior hermeticity for packaging of implantable devices,
combined with high mechanical stability. Up to now, there is little data of the
permeability at physiological conditions through thin‐film multilayer laminates.
The motivation of this thesis is the development and analysis of new
generation biocompatible packaging for implantable medical devices based on
thin‐film technologies. In particular, the packaging based on multilayered
organic / inorganic hybrid structures is investigated both theoretically and
experimentally.
viii
For this reason, a novel automated single‐chamber deposition reactor for the
deposition of organic and inorganic thin‐film materials has been developed
during this research work. Different deposition technologies have been
investigated and the most promising plasma enhanced chemical vapour
deposition process has been integrated into the reactor system. Through this
innovative technology platform, the materials can be deposited directly on the
electronic circuits of the device. This proposed method of direct encapsulation
for implantable devices, by alternating different coatings, combines low water
permeability, chemical and mechanical stability of the packaging with small
dimensions, and low‐cost manufacturing.
On the one hand, inorganic materials like ceramic or metallic thin films could
provide sufficient hermetic tightness due to their high intrinsic atomic
packaging density. However, they tend to form pinholes and cracks on non‐
uniform and complex substrates. On the other hand, polymeric thin films like
poly‐para‐xylylene have proven to be pinhole‐free at a micrometer scale but
have a lower tightness and mechanical stability. The combination of
ceramic/metal and polymer layer properties creates new composite material
which can solve these issues.
The second part of the thesis focuses on analyses of the deposited multilayer
structures. In order to characterize the tightness of the multilayer packaging, a
specialized helium gas permeability test system has been developed according
industrial standards. In parallel, the water vapour permeation has been
evaluated by the use of highly sensitive calcium mirror tests. These tightness
investigations showed significantly improved barrier properties compared to
standard polymer layers, in terms of helium and water vapour permeation.
Regarding the material and interface composition of the barrier structures,
Fourier transform infrared spectrometry and X‐ray photoelectron spectroscopy
ix
has been applied. In order to gain insight into the morphology of the deposited
multilayers, a combination of focused ion beam and scanning electron
microscopy was used. In addition, the topological aspects of multilayers and
detection of local thickness deficiencies have been investigated by application
of a developed algorithm for defining ideal layer conformity. The use of atomic
force microscopy revealed the surface structures of the deposited layers.
In order to understand gas diffusion through multi‐layered thin film structures,
theoretical permeation models were conceived which were later evaluated by
finite element simulations. The simulated data of the different theoretical
models was finally compared to the experimental diffusion measurements.
Acknowledgments
First of all, I would like to express my deepest gratitude to my two enthusiastic
and passionate supervisors PD Dr. Jürgen Burger and Prof. Dr. Herbert Keppner
for their support, suggestions, and the long discussion during my thesis.
I would like to heartily thank my co‐advisor Prof. Dr. Dr. Rolf Vogel, my external
co‐referee Dr. Andreas Blatter, and my mentor Prof. Dr. Marco Caversaccio for
the discussions and their important thesis inputs.
I would give a special thanks to Dr. Yanik Tardy, the head of the Codman R&D
department in le Locle, for the promotion and the support of this research
work during the years.
I would express my gratitude to our director of the Institute of Applied Micro
Technologies (IMA, He‐Arc) Prof. Dr. Oksana Banakh, who provided the
required infrastructure and facilities and humour during this thesis.
x
This work would not have been possible without the support of my fantastic
team and friends, Thierry Aellen for the excellent discussion, imagination,
motivation and the long nights during the patent phase, Stefanie Uhl for their
superb contribution, inputs and the hundreds of deposition, Benjamin Graf for
the marvellous development work of the deposition reactor and programming,
François Feuvrier for the outstanding engineering work specially for the calcium
test, Fabrice Bisoffi for the imaginative troubleshooting and the multiple pump
repairs and Christiane Broggini for the great administrative work.
My special thanks are extended to the whole collaborators of our institute, Dr.
Alexandra Kaempfer‐Homsy, Catherine Csefalvay, Claudio Prieur, Constantin
Raymond, Daniel Morel, David Grange, Edith Laux, Edouard Guibert, Prof. Dr.
Dr.Harry J. Whitlow, Jean‐Michel Kissling, Jean‐Paul Sandoz, Jérôme Borboën,
Jérôme Charmet, Joël Matthey, Julien Brossard, Laure Jeandupeux, Loïc
Piervittori, Lucien Steinmann, Mario Dellea, Olivier Gloriod, Dr. Patrick
Jeanneret, Pierre‐Alain Montandon, Dr. Pierre‐Albert Steinmann, Dr. Pierre‐
Antoine Gay, Sébastien Brun, Sophie Farine Brunner, Tony Journot, and
Stephan Ramseyer.
In addition, I would take the opportunity to thank the Codman R&D group,
especially Toralf Bork, Martin Pfleiderer, Rocco Crivelli, Didier Balli, Philippe
Margairaz, Danilo Roth, Fabien Luginbuhl, Thierry Pipoz and Karine Bruchon for
the industrial impact of this work.
I would thank Dr. Gilbert Schiltges for the help of the build‐up of the simulation
model and the profound theoretical discussion about the diffusion equations.
I would like to thank for the important effort, provided by Dr. Olaf Kahle form
the Frauenhofer institute PYCO in Berlin, for the collaboration concerning the
WVTR investigations.
xi
I would thank Prof. Dr. Luc Stoppini from the University of Applied Science
(HEPIA) in Geneva, for the cytotoxicity investigations by confocal microscopy.
I would like to express my gratitude to the internship students Vincent
Michaux, Clemens Meyer and Yann Girardet who added an important
contribution to the thesis.
I would specially thanks Dr. Joanna Bitterli for the spectacular FIB/SEM and
AFM images of the multilayer structures.
I would thank the team from Comelec, Wilfred Zutter, Dr. Jean‐François
Laithier, Jacques Bogar, Hicham Damsir for their helpful contribution.
I am heartily thankful for the proofreading’s of Thomas White, Dr. Andreas
Stahel, Dr. Gilbert Schiltges, Dr. Alexandra Kaempfer‐Homsy and Prof. Dr. Dr.
Harry J. Whitlow.
In addition, I would specially thank Serge‐André Maire for the design of the
book cover and Agnès Dervaux Duquenne for the help of the book publication.
I would like to thank the Swiss innovation promotion agency (CTI) for financing
this work by the grants 10836.1 PFLS‐LS and 13207.1 PFLS‐LS.
Finally, I would express my deepest gratitude to my family. Especially to my
Mother and my Father for letting me choose my way and for their precious
support during all these years, and to my sister and my aunt for their helpful
suggestions and motivation. In particular, I would thank my lovely life partner
for her enormous support and encouragement along the thesis.
xii
xiii
Nomenclatures
List of symbols
Permeability coefficent mol
m s Pa
Diffusion coefficient 2m
s
Solubility coefficient 3
mol
m Pa
Solubility coefficient of the amorphous phase 3
mol
m Pa
Volume fraction of the amorphous phase
Defect surface fraction
Particle flux 2
mol
m s
Internal molar concentration 3
mol
m
External molar concentration 3
mol
m
Gas pressure mbar
Measured helium pressure mbar
Helium pressure load mbar
Membrane thickness m
Membrane surface 2m
Time s
Mean particle velocity m
s
xiv
Boltzmann constant m
s
Temperature K
Mass kg
Knudsen number
λ Mean free path m
Number density 3
1
m
Atom diameter nm
Diameter of pore size m
Total defect surface 2m
Porosity [ ]
Tortuosity [ ]
Occupation density 2m
s
Adsorption energy eV
Activation energy eV
Universal gas constant J
mol K
Light intensity 2
W
m
Absorption coefficient 1
m
Calcium volume 3m
Calcium mass g
Water mass g
Specific density 3
g
cm
xv
Molar mass of calcium g
mol
Molar mass of water g
mol
Water transmission rate 2
g
m day
Pressure / flow‐rate conversion factor
List of Acronyms and Abbreviations
AFM Atomic Force Microscopy
ALD Atomic Layer Deposition
Al2O3 Aluminium Oxide (Alumina)
AlxOy Aluminium Oxides
ASTM American Society for Testing and Materials
CCT Critical Cracking Thickness
CSF Cerebrospinal Fluid
CVD Chemical Vapour Deposition
EIS Electrochemical Impedance Spectroscopy
ENA Electrochemical Noise Analysis
FDA Food and Drug Administration
FIB Focused Ion Beam
HMDS Hexamethyldisilazane
HMDSO Hexamethyldisiloxane
ISO International Organization for Standardization
LPCVD Low Pressure Chemical Vapour Deposition
MEMS Microelectromechanical Systems
MIL‐STD Military Standard
MPECVD Microwave Plasma Enhanced Chemical Vapour Deposition
xvi
NaCl Sodium Chloride
OLED Organic Light Emitting Diode
OTM Optical Transmission Measurement
OTR Oxygen Transmission Rate
PAN Polyacrylonitrile
PC Polycarbonate
PCB Printed Circuit Board
PECVD Plasma Enhanced Chemical Vapour Deposition
PEN Polyethylene Naphthalate
PES Polyethersulfon
PET Polyethylene Terephthalate
PS Polystyrene
PVC Polyvinyl Chloride
PVD Physical Vapour Deposition
RF Radio Frequency
RH Relative Humidity
SCCM Standard Cubic Centimeter per Minute
SEM Scanning Electron Microscopy
SEV Secondary Electron Multiplier
SI International System of Units
Si3N4 Silicon Nitride
SixNy Silicon Nitrides
SiO2 Silicon Dioxide
SiOx Silicon Oxides
SMMS Smart Micro‐Medical Implants
UNCD Ultrananocrystalline Diamond
WTR Water Transmission Rate
WVTR Water Vapour Transmission Rate
xvii
Contents
KEYWORDS vii
ABSTRACT vii
ACKNOWLEDGMENTS ix
NOMENCLATURES xiii
CONTENS xvii
1 INTRODUCTION 1
1.1 Medical context 1
1.2 State of the art of long‐term active implants packaging 3
1.2.1 Bulk material packaging 5
1.2.2 Thin‐film packaging 7
1.3 Motivation 13
2 CONCEPT 15
3 THEORY 25
3.1 Permeation 25
3.1.1 Polymer permeation 31
3.1.2 SiOx permeation 33
3.1.3 Combination of the materials 40
3.1.4 Multilayer permeation models 42
3.1.5 Water permeation 46
3.2 Layer uniformity 53
4 MATERIALSANDMETHODS 57
4.1 Helium permeation test system 57
4.2 Water permeation test system 61
4.3 Deposition reactor 64
5 RESULTS 67
5.1 Stead‐state helium permeation 67
5.1.1 Bilayer measurements 70
5.1.2 Multilayer systems 73
5.1.3 Transient helium permeation 74
5.2 Simulation 75
5.2.1 Single layer model 75
5.2.2 Ideal laminate model 80
xviii
5.2.3 Percolative path model 86
5.3 Water permeation 88
5.4 Surface and uniformity analysis 92
5.5 In vitro cytotoxicity and biostability analysis 96
6 CONCLUSIONANDOUTLOOK 99
6.1 Conclusion 99
6.2 Future developments 103
6.3 Outlook 106
7 REFERENCES 107
8 LISTOFPUBLICATIONS 115
8.1 Ultra‐thin layer packaging for implantable electronic devices 117
8.2 Protective multilayer packaging for long‐term implantable medical devices 131
8.3 Ultrathin Multilayer Packaging 139
8.4 Plasma Enhanced Polymer Ultrathin Multilayer Packaging 149
8.5 Packaging with Active Protection Layer 159
8.6 Three Dimensional Packaging for Medical Implants 169
9 APPENDIX 187
9.1 Uniformity algorithm (MATLAB) 187
9.2 Test leaks inspection certificates 188
9.3 Laminar flow facility certificate 190
9.4 Other publications by the author 191
10 CURRICULUMVITAE 193
11 DECLARATIONOFORIGINALITY 197
1 Introduction 1
1 Introduction
1.1 Medicalcontext The term “medical devices” has been defined by the European Union as: “any
instrument, apparatus, appliance, software, material or other article, whether
used alone or in combination, together with any accessories, including the
software intended by its manufacturer to be used specifically for diagnostic
and/or therapeutic purposes and necessary for its proper application, intended
by the manufacturer to be used for human beings” [1].
With regard to this citation, different approaches for medical device
classification can be considered. In particular, form a medical point of view
(device application for diagnostic or therapeutic purposes), for an engineering
aspect (instruments and apparatus), for patient safety (body tissue/device
interface interactions) or for legal and ethical reasons (standards and legal
codes) can be taken as classification indicators. In general, the external and
internal application of the devices can be distinguished as shown in Figure 1.
External devices are used outside of the human body for monitoring and
diagnosis purposes e.g. heart rate monitors, glucose meters, X‐ray computed
tomography or magnetic resonance imaging scanners.
In contrast, internal devices are implanted into the human body and thus, are
also termed implants. Figure 1 divides these devices into active and passive
implants. The most common implants are passive which means that they
function without a power supply. For example hip and knee prostheses,
orthopaedic screws and plates, coronary stents and artificial heart valves are in
this category. There is a growing interest in the development of active medical
implants using sensors and actuators in order to monitor and stimulate human
2 1 Introduction
body functions. The most commonly known active implants are, for example,
cardiac pacemakers and defibrillators, cochlear implants, retina implants and
implantable drug pumps. In order to provide these sensing and actuating
functions, the implants are composed of sensitive and complex electronic
components (e.g. MEMS sensors/actuators, microcontrollers, batteries,
antennas). To guarantee an extended lifetime of over ten years for active
medical implants, different kinds of packaging or encapsulation have been
applied. Nowadays, the size of the implant is often limited by the packaging
method. The progress in medical packaging technologies will allow for the
miniaturization of these implants and thereby lead to novel, less‐invasive
surgery methods which translates to better patient quality of life. The next
chapter is focused on the different packaging solutions of smart implantable
medical devices. In particular, novel encapsulation methods based on thin‐film
technologies in order to reduce the implant size will be discussed.
Figure 1. Classification of medical devices into external, internal, and the sub‐categories passive, and active devices.
1 Introduction 3
1.2 Stateoftheartoflong‐termactiveimplantspackaging To ensure the functionality of biomedical implants, the packaging has to
guarantee integrity during the whole lifetime of the device. For this reason,
different criteria have to be taken into consideration.
First of all, the biocompatibility of the materials used has to be assured because
of the direct tissue contact. Several materials such as the majority of noble
metals, titanium, medical‐grade stainless steel, ceramics like aluminium or
silicon oxide, biocompatible glasses and polymers (e.g. polyethylene,
polypropylene, polytetrafluoroethylene) have been successfully used [2].
Additional biocompatible tests have to be performed if the raw material, the
material structure or composition, the fabrication process, or other parameters
differ from the certified material. Different biocompatibility test scenarios
related to the implant purpose are suggested by the International Organization
for Standardization (ISO 10993‐1) or the United States Food and Drug
Administration (FDA General Program Memorandum #G95‐1) [3].
The second important criterion is the hermeticity of the packaging. The
packaging has to protect the human body against toxic materials being
released from the device electronics. On the other hand, the ingress of the
aggressive body fluids also has to be prevented by the packaging in order to
protect the device components. Well known indicators to define the tightness
of the packaging are the oxygen and the water vapour transmission rate (OTR
and WVTR). In order to measure these rates, different methods have been
introduced, for example, by the American Society for Testing and Materials
(ASTM). Standards such as ASTM F1927, ASTM D1434, ASTM D3985 and ASTM
F2622 for OTR measurement and ASTM E96, ASTM E398 and ASTM F1249 for
WVTR can be applied. A more sensitive WVTR measurement is the calcium
mirror test method, for which presently no standard exists [4]. Other methods
4 1 Introduction
to measure the hermeticity of the entire packaging are defined in the military
standards MIL‐STD‐883 (method 1014), MIL‐STD‐750 (method 1071), and MIL‐
PRF‐38534G (appendix C). These standards are mostly based on mass
spectrometric leak rate measurements of a helium tracer gas in relation to the
internal package volume size. Further details about these test methods are
given in sections 4.1 and 8.1.
It has been observed that all packaging materials leak after a certain time [5].
This permeation depends on the material structure, the atomic density and
matter imperfections. For example, polymeric materials show high WVTR from
10 to 1 x 10‐3 g mm m‐2day‐1 due to their low mass density caused by the
interstices between the polymeric chains. In contrast, metals, ceramics and
glasses show far lower WVTRs of around 1 x 10‐6 to 1 x 10‐10 g mm m‐2day‐1.
Such permeation rates are sufficiently low to protect an implant efficiently
against water in the long term.
Other important criteria are the chemical and mechanical stability during the
life span of the device. The harsh human body conditions can attack the
packaging which leads to a degradation of its barrier function. For this reason,
an indirect biocompatibility study can be carried out in order to investigate the
leachable products and thereby the degradation of the encapsulation according
to ISO 10993 [3]. Other possibilities to investigate the chemical resistivity of the
packaging are based on electrochemical corrosion tests such as corrosion
potentials, electrochemical impedance spectroscopy (EIS), electrochemical
noise analysis (ENA) and potentiodynamic polarization curves. By the use of
these methods, the corrosion resistance can be investigated [6]. To guarantee
the mechanical integrity of the device during the fabrication process, storage,
transport, surgical intervention and implantation, multiple tests have to be
applied as proposed by the ASTM and ISO, for example. Mechanical properties
1 Introduction 5
like flexural, compressive, shear and tensile strengths must be ensured to avoid
damage of the device by environmental forces [7].
Depending on the design requirements of the medical device [8], its packaging
has to be adapted. As an example, active implantable devices often
communicate via radio frequency (RF) to exchange data between an external
reader unit and the implanted device by telemetry [9]. In order to transmit
signals by RF telemetry, the package material has to be suitable for the
communication frequency. Materials, exhibiting low absorption in the specific
electromagnetic band are therefore preferred. The integration of passive
telemetry allows for bidirectional data communication and in addition, energy
transmission with a single RF antenna [10‐12]. The energy is thereby
transmitted by an external electromagnetic RF field through inductive coupling
to the implant’s receiving antenna and the data communication is based on an
absorption modulation [13]. Using such telemetric technology, standard
batteries for internal power supply, which have low battery size / energy ratios,
can be avoided. Hence, the implant can be further miniaturised and the life
time will be increased due to the external power supply.
For the appropriate choice of the packaging material, the mentioned
considerations have to be taken into account.
1.2.1 Bulkmaterialpackaging
One of the first active implantable devices was a ventricular pacemaker which
was implanted in 1958 by Elmqvist and Senning [14]. These early devices were
totally encapsulated in epoxies and functioned properly over one year [15, 16].
Due to the fact that all polymers show high WVTRs and at the same time,
exhibit deterioration of its chemical structure compared to other denser
6 1 Introduction
materials, the packaging strategy has evolved. In the late sixties, the
investigations began with the encapsulation of pacemaker batteries and
electronics using a titanium or stainless steel housing [17, 18]. The longevity of
the devices could thereby be considerably improved. Nowadays, laser‐beam
welded metal housings are state of the art, not only for pacemakers, but also
for other active implantable devices such as programmable infusion pumps
[19]. The electrical connection to pacing electrodes or antennas outside of the
metallic encapsulation for example, can be realised by glass or ceramic
feedthroughs [20]. For this technology, the glass or ceramic materials are
melted or sintered around the electrical metal contacts. This process is well
known and has a high hermetic helium gas tightness [21].
Other encapsulation solutions that are based on by the use of glass housings
have been proposed. This packaging has the advantage of a low absorption of
the electromagnetic RF fields for telemetric systems. WVTR permeation
through glass is typically two orders of magnitude higher than metals but even
a thin‐film glass packaging in the μm‐range provides sufficient hermetic sealing
for long‐term applications [5, 22]. For the fabrication of this hermetic
packaging, the active implant can be inserted, for example, into a capillary glass
tube and encapsulated by welding its ends to glass beads by a CO2 laser beam
[23]. Another possibility that has been investigated is low‐temperature brazing
methods in order to prevent electronic damage. The electronic circuit is
encapsulated between a borosilicate glass substrate and a glass capsule on
which a 2 μm gold film was deposited. A low temperature eutectic brazing ring
(Sn77.2In20Ag2.8, melting point 187 °C) was inserted between the substrate
and the capsule and soldered at a temperature of 195 °C [24].
1 Introduction 7
Nowadays, some polymeric materials, such as silicone rubbers, epoxies,
parylenes, polyurethane, polyetheretherketone or polyimide, have been
investigated for use in implant packaging [25‐30]. An advantage of a thick bulk
polymer encapsulation of the implantable device is its high RF transparency.
The main drawback of polymers is the relative high WVTR compared to glass,
ceramic or metal materials as discussed above. The encapsulation properties of
the polymer strongly depends on the atomic package density, molecular
orientation, crystallinity, chain stiffness and its shrinkage [31].
In contrast to glass, ceramic and metal encapsulations, polymer encapsulation
provides limited sealing properties for long‐term implantation [32].
1.2.2 Thin‐filmpackaging
In addition to encapsulations using bulk materials, different thin‐film packaging
approaches have been investigated for hermetic sealing of medical implants.
The main advantage of a thin‐film packaging is the size reduction of the devices
compared to the conventional methods discussed in the previous section.
Inorganic thin films composed of metals, ceramics or glasses can theoretically
provide sufficient hermetic tightness due to their high intrinsic molecular
density. Nevertheless, during the deposition, they tend to build columnar
microstructures or create defects due to internal stress formation [33, 34]. In
contrast, organic thin films based on poly‐para‐xylylenes, for example Parylene‐
C, can be deposited pinhole‐free down to a thickness of around 50 nm and in
addition, it exhibits stressless deposition properties [35]. Parylene‐C layers are
widely used as an encapsulation material for biomedical applications. Due to its
high conformity, thermal stability (melting point at 290 °C) and defect‐free
characteristics, Parylene has been used for packaging of neural prostheses like
8 1 Introduction
cochlear implants and microelectrodes, for cardiac pace makers, and deep
brain stimulators [26, 36, 37]. The Parylene molecule was synthesized for the
first time by Szwarc in 1947 [38]. The first efficient deposition method was
developed by Gorham in 1966, which was based on a downstream, low
pressure chemical vapour deposition (LPCVD) process [39]. Further details of
the Parylene deposition process are explained in section 4.3. It has to be
mentioned that the WVTR of Parylene‐C of 80 g μm m‐2 day‐1 at 38 °C and 90%
relative humidity (RH) is one of the lowest values for polymeric materials [40].
Other possible polymeric thin‐film barrier coatings for biotechnological
purposes are well described in literature [41]. However, looking at the
molecular structure of polymeric thin films, the low atomic density favours the
permeation of water molecules and thereby provokes swelling, degradation
and corrosion effects of the encapsulated device. Thus, by the use of polymeric
thin‐film layers, the long‐term tightness cannot be ensured.
In order to increase the atomic package density and thereby the hermetic
tightness, several inorganic thin‐film coatings have been investigated. These
films are mostly deposited by plasma enhanced chemical vapour depositions
(PECVD). The PECVD process has the advantages of layer densification during
deposition due to the plasma ion bombardment, good layer conformity and
low process temperatures compared to conventional physical‐ and chemical
vapour depositions (PVD, CVD). Using this technology, for example silicon
oxides, silicon nitrides or aluminium oxides can be deposited as barrier
materials. These barrier coatings have been mostly investigated for the
encapsulation of organic light emitting diodes (OLEDs) or flexible solar cells in
order to reduce the moisture and oxygen permeation [42, 43]. The OLED
components are thereby deposited on polymeric substrates and encapsulated
by transparent thin‐film coatings for the fabrication of flexible devices. For this
1 Introduction 9
purpose, silicon nitride (SiNx) layers have been studied by Wolf et al. [44]. SiNx
layers with a thickness of 25‐400 nm were therefore deposited on polyethylene
terephthalate (PET) and polyethylene naphthalate (PEN) substrates by an
inductively‐coupled PECVD at 70‐120 °C. The WVTR measurements showed
values of 2 x 10‐3 g m‐2 day‐1 (25 °C, 30‐50% RH) for 400 nm thick SiNx layers.
Silicon oxides (SiOx) coatings have also been evaluated and showed the
potential to significantly decrease moisture and oxygen permeation by several
orders of magnitude [45, 46]. Wuu et al. investigated SiOx deposited by PECVD
at temperatures between 80 and 170 °C barriers on flexible polyethersulfon
(PES, 200 μm) substrates [47]. Single SiOx layers with a thickness of 100 nm
exhibited a WVTR of 0.31 g m‐2 day‐1 (25 C, 100% RH). By a double‐sided SiOx
deposition (2 x 100 nm) on the PES substrates, the WVTR could be decreased to
a minimum of 0.1 g m‐2 day‐1. Da Silva Sobrinho et al. found nearly the same
WVTR of 0.2 g m‐2 day‐1 (33 °C, 100% RH) for PECVD SiOx coatings with a
thickness of around 100 nm on 13 μm thick PET substrates [48]. In addition,
they showed that the oxygen transmission rate (OTR) and the WVTR of SiOx and
SiNx are equal for layer thicknesses higher than 100 nm.
Another promising deposition technology is atomic layer deposition (ALD) that
allows the growth of nanoscopic and dense inorganic monolayers (e.g. SiO2,
ZrO2, Al2O3, TiO2, TiN, Ti, Cu, Ni) [49]. For the deposition, usually two gaseous
precursors, which are sequentially introduced into the reactor chamber, are
used. After two precursor cycles, a conformal and pinhole‐free atomic
monolayer is deposited on the substrate and the process can be repeated layer
by layer until the desired coating thickness is achieved. Several studies have
been devoted to the barrier characterisation of ALD films for protection of
10 1 Introduction
organic light‐emitting diodes. For example, ALD aluminium oxide (Al2O3) layers
have been investigated with regard to the WVTR permeation barrier
properties. The investigation reports extremely low WVTR of around 0.06 g m2
day‐1 for a 30 nm thick Al2O3 layer [50]. However, the application of ALD layers
on implantable electronic circuits is time consuming and critical due to the
increased process temperatures needed.
Another approach for hermetic thin‐film encapsulation has been developed by
the deposition of ultrananocrystalline diamond (UNCD) coatings for the
protection of retinal implants [51]. The films were fabricated by microwave
plasma‐enhanced chemical vapour deposition (MPECVD) using an argon‐
methane‐hydrogen precursor mixture. Deposition temperatures between 400
and 800 °C were investigated. The cross‐sections of the films showed dense
and pinhole‐free layers in the observed range. Rabbit eye in vivo tests showed
no acute surface damage of the coated Si samples, but leakage current
measurements using cyclic voltammetry, revealed a need for optimisation.
Further investigation of multilayered structures (UNCD coating deposited on an
additional ceramic layer) are expected to achieve lower leakage currents
compatible for long‐term retinal devices.
As the previous studies showed, the decrease of OTR and WVTR is limited using
single layer structures. In contrast, multilayered thin‐film structures open a
route to improved barrier properties. The first thin‐film multilayer barrier
structures were reported by Shaw et al. in 1994 [52]. Their approach was to
coat an oriented polypropylene substrate with an alternation of organic
acrylate and aluminium thin films. OTR measurements showed a decrease of
oxygen permeation from 1’000 to 0.1 cm3 m‐2 day‐1 (STP) using these barrier
structures. Affinito et al. improved the barrier properties by the use of
1 Introduction 11
acrylate/alumina(Al2O3)/acrylate stacks [53]. The Al2O3 was deposited by
reactive sputtering (PVD) and the polymeric films by flash evaporation of a
poly(ethylene glycol) diacrylate monomer. A barrier stack with thicknesses of
1 μm acrylate/ 25 nm Al2O3 / 240 nm acrylate on 50 μm thick polyester
substrates showed a WVTR lower than 0.0155 g m‐2 day‐1. Further work of this
group showed a defect decoupling effect of the polymeric layer for
successional aluminium oxide layers [54]. Low WVTRs of 8 x 10‐5 g m‐2 day‐1
(25 °C, 40% RH) were achieved by five dyads (5 x AlOx/acrylate bilayers) on PET
substrates.
A silicon based multilayer structure (SiOx and SiOxCyHz) was proposed by Colite
et al. [55]. This structure was composed of polymeric SiOxCyHz surface
smoothing and adhesion layers, and inorganic SiOx‐like barrier layers, deposited
by initiated CVD and PECVD, respectively. The clean layer interfaces due to
their single‐chamber process improved the interlayer adhesion. The lowest
WVTR value of 0.01 g m‐2 day‐1 (25 C, 98% RH) could be achieved by a 300 nm
barrier layer composed of three SiOx/SiOxCyHz dyads deposited on a 125 μm
PEN substrate.
Another silicon based graded organic/inorganic multilayer structure was
proposed by Yan et al. [56]. The advantage of graded interfaces is an improved
adhesion between the adjacent interlayers. The barrier was composed of
inorganic SiOxNy and organic SiOxCy layers. For their fabrication, the precursor
gas composition was changed gradually during the PECVD process. A WVTR of
2 x 10‐5 g m2 day‐1 (23 C, 50% RH) for barrier layers deposited on polycarbonate
substrates was measured.
12 1 Introduction
Chiang et al. reported silicon nitride (SiNX)/Parylene based barrier coatings [57].
For cyclic bending tests, two Parylene/SiNX dyads, with thicknesses of 600 nm
Parylene and 100 nm SiNX, were deposited on top of 175 μm thick
polycarbonate films (PC). Cyclic bending tests showed that the additional
Parylene layer on top of the SiNX could suppress crack formation. After the
3000‐times cycling test, the OTR and WVTR remained constant at 0.1 cm3 m‐2
day‐1 and 0.01 g m2 day‐1, respectively. Similar OTR results of
PC/Parylene/SiNx/Parylene barriers were observed by Schaepkens et al. [58].
Another Parylene‐C based multilayer structure was proposed by Chang et al.
for the encapsulation of retinal implants [59]. Their structure was composed
first of a 5 μm thick Parylene‐C layer, followed by a titanium / gold metal
coating with the thicknesses of 0.02 and 0.3 μm, respectively, and finally a
second 5 μm thick Parylene‐C layer. The Parylene‐C layers were deposited with
the conventional Gorham process and the metal layer by electron beam
physical vapour deposition (EBPVD) on top of an electrical amplifier chip. The
coated devices were immersed into saline solution (NaCl concentration of
9 g / l) and heated at 97 °C for accelerated aging tests. The signal loss (failure)
of the amplifier was chosen as indicator for the efficiency of the barrier
protection. The mean time to failure of these discrete components was 37
days. This corresponds, by use of an Arrhenius relationship, to a lifetime of 6.7
years at 37 °C. Attention has to be taken to their choice of the mean time
failure indicator. For long‐term implants, a restricted drift indicator will be
more appropriate as a device failure criterion. In that case, the proposed
barrier structure will not be sufficient for a long‐term application.
1 Introduction 13
1.3 Motivation
Current packaging for long‐term smart medical implants are composed of bulk
polymer, glass, or metal encapsulations. These technologies drastically limit
further miniaturisation of the devices. There is a need of reliable, cost‐effective
batch‐manufacturing packaging processes at wafer‐level, to protect the
electronic, mechanical and microelectromechanical systems (MEMS) of such a
device against the exposure of body fluids. Conversely, the body tissue has to
be protected from possible reaction products between the implant and the
body liquids. In addition, a continuous size reduction is essential for the
development of less‐invasive smart micro‐medical implants (SMMS), which can
be inserted into the human body by catheter‐based surgery. Such devices can
be easily placed via blood vessels, such as the femoral artery or other body
orifices. Due to the reduced size, these next generation SMMS will allow for
monitoring and stimulation of a wide range of body functions. Recent
outstanding advances in medical research show, for example, the possibility of
optic nerve (retina), acoustic nerve (cochlea) or deep brain stimulation. These
new discoveries will benefit from the development of novel SMMS implants for
specific disease treatments. For a further miniaturisation of such smart
implants, a superior biocompatible thin‐film packaging concept has to be
developed.
At the moment, no thin‐film packaging coating exists for the encapsulation of
long‐term smart medical implants. Most proposed thin‐film barriers have
insufficient tightness, i.e., a high WVTR. In contrast, coatings with high barrier
performance have been developed for application in flexible OLED displays or
organic solar cells. These concepts cannot be directly applied to medical
implants due to the different environmental conditions. For example, the
lifetime of OLEDs has to be ensured at standard ambient condition, whereas
14 1 Introduction
long‐term implants are exposed to the aggressive saline body fluids at
increased temperature (≈ 37 °C) during many years. In addition, the
biocompatibility of the chosen packaging materials and the combination
thereof, is crucial. In order to overcome these scientific challenges, a novel
packaging strategy is proposed in this PhD thesis.
This means that multiple scientific issues have to be resolved. First of all, a
novel thin‐film deposition system has to be developed to deposit uniform
multilayer barriers over the SMMS. In order to deposit a multilayer barrier with
different layer materials, various deposition technologies can be used. A single
chamber reactor system, where the different deposition methods are
integrated, is preferable due to the advantage of uncontaminated and well‐
defined interface properties between the alternate layers of a stack. In
addition, the appropriate process and its parameters for conformal deposition
at near‐ambient temperature was to be found, which avoids the degradation of
the device during its exposure to the deposition process. The deposited
material has to be analysed afterwards, in order to optimise the deposition
parameters to give the highest packaging performance together with
biocompatibility. Due to the lack of hermeticity test methods for ultra‐tight
thin‐film barriers, conventional tests have to be modified or novel methods to
be developed. Finally, the experimentally obtained hermeticity values are
compared to theoretical calculations based on physical transport models,
combined with finite element methods.
2 Concept 15
2 Concept
The development of implantable devices requires the management of several
complex challenges: medical implants must comply with strict hermeticity and
biocompatibility standards; they have to be mechanically stable and must be as
small as possible. Furthermore, external communication with the implant is
preferred. The development of new high‐performance medical devices that are
smaller and less expensive, is often constrained by the selection of the
packaging material.
The main objective of this work is the development of a novel thin film concept
that can be applied to numerous implantable electronic devices, where size
reduction is the key priority. In order to illustrate the potential of new hermetic
thin‐film concepts on SMMS, the recently developed Codman flow transducer
[60] has been chosen to serve as a model for implant size reduction through
the application of multilayer barriers compared to conventional glass
packaging.
The Codman company develops and markets a wide range of diagnostic and
therapeutic products for the treatment of central nervous system disorders,
with a focus on intractable pain management and paediatric and adult
hydrocephalus. Hydrocephalus is one of the most common congenital disorders
of the central nervous system. The treatment of hydrocephalus requires the
implantation of a catheter shunt system for cerebrospinal fluid (CSF) drainage.
Little is known about the effective CSF flow rates and patency of these catheter
systems which are key factors for the successful treatment. For an optimised
patient diagnostic and monitoring of the correct valve function, a flow sensor
has been developed that can detect the CSF flow through the shunt system. For
encapsulation of the flow sensor electronics, a conventional glass packaging
has been chosen. Nevertheless, this bulk packaging increases the total size of
16 2 Concept
the implant and creates unused cavities, and thus limits the miniaturization
potential. Figure 2 shows two flow sensor packaging concepts, with the
conventional glass base plate and glass capsule housing on the right side, and
the miniaturized multilayer packaged flow sensor without glass capsule on the
left side.
Figure 2. Design study of an implantable flow sensor for the treatment of hydrocephalus using the conventional glass packaging technology (right) compared to the new multilayer packaging (left).
By positioning the electronic components into the base plate and the
application of the multilayer thin‐film packaging directly on top of the sensor
and the printed circuit board (PCB), the glass capsule and hence, the glass
soldering process can be avoided. Hence, a size reduction of around 70% of the
total volume can be achieved. This reduction represents a highly significant
improvement for the development of new SMMS implants.
In order to fabricate these thin multilayer structures, the polymer poly(chloro‐
para‐xylylene), also known as Parylene‐C, has been chosen as basic material.
The deposition of Parylene is highly conformal and produces films that are
2 Concept 17
considered to be pin‐hole free (from about 50 nm on Si wafer). This, in
combination with its biocompatibility and its excellent mechanical and
chemical properties, makes Parylene a material of choice for packaging
purposes. Furthermore, due to its high thermal stability (a melting point of 290
°C), Parylene‐coated devices can be sterilized (up to 135 °C) using conventional
techniques prior to operation. Nowadays, Parylene is commonly used as
biocompatible package on stents and other devices. However, long‐term
implantable device compatibility data are lacking. The reduced scratch‐
resistance of Parylene‐packaged devices always implies a certain risk of
damage. Thus, the loss of biocompatibility due to harsh handling limits the use
of Parylene for clinical use. In general, all polymers have reduced mechanical
scratch resistivity compared to ceramics or metals. Moreover, their molecular
structure and their reduced density favours diffusion through the polymeric
chains and swelling effects from liquid exposure. Thin‐films made of ceramics
or metals have a higher molecular package density and thereby tightness.
Nevertheless, they tend to create defects such as pinholes and cracks and the
coating can be less conformal, depending on the deposition process. These
drawbacks result from directionality and internal stress introduced by plasma‐
assisted non‐equilibrium processes.
Based on these considerations, this work proposes novel packaging concepts
comprised of Parylene‐based multi‐stack laminates in combination with
hermetic non‐Parylene barrier layers. The concept takes advantage of the fact
that the combination of the two layer materials favours the desired properties
of each individual layer, while reducing their respective drawbacks. The typical
morphology of such a multilayer stack is represented schematically in Figure 3.
18 2 Concept
Figure 3. Schematic model of direct diffusion of a bulk polymeric single layer (left) and a multilayer structure
(right). The multilayer structure shows a highly increased effective diffusion pathway compared to a single
polymeric layer (red) due to the hermetic inorganic barrier layers (yellow). The percolative pathway arises due
to the defects in the inorganic layers.
The illustration on the left side shows the direct diffusion across the bulk
polymeric layer due to the reduced density. The integration of high density
inorganic layers increases the effective diffusion path‐length through the
coating compared to the bulk polymeric layer. Thus, a stack of polymer and
ceramic layers has the potential to combine the advantages of both materials
in such a way that the percolative pathway of permeants is increased and
directly linked to the crack/pinhole density of the ceramic layer. Moreover, the
interfaces themselves can dominate the physical/chemical permeation
properties of the entire stack. This means that, with regard to the number of
interfaces, increasing the number of thinner layers, rather than having fewer
thicker layers, can enhance the overall barrier tightness. This can be explained
by the material transition zone of the two materials and by capillary effects in
the case of poor adhesion between the layers. In order to increase the
adhesion of the interlayers, the total stress of the barrier stack has to be
matched. Well‐defined deposition conditions, interface treatments, layer
adhesion, and material compositions of the individual layers therefore have to
be taken into account during the deposition process.
2 Concept 19
For the realisation of the above presented working principle, different material
combinations can be proposed. The choice of the stack partner and its
deposition process are important questions to be answered. Finally, three
different concepts have been evaluated as the most promising approaches. The
following figure shows stack concepts based on Parylene with inorganic, high‐
density, and liquid interlayers.
Figure 4. Concept of different multilayers barriers based on Parylene layers interspaced with different high‐hermetic interlayers.
Figure 4 shows the conformal coating strategies over a substrate with
rectangular feature sizes such as electronic components. The first concept is
based on inorganic layers such as silicon oxides, silicon nitrides, titanium oxides
or aluminium oxides. These ceramic materials have the advantage of high
molecular density and thereby, high barrier properties. For biocompatibility,
molecular density, conformity, and deposition at low temperature, SiOx has
been identified as an optimal inorganic interlayer material. To achieve low
deposition temperatures, a PECVD process with the precursor monomer
hexamethyldisiloxane (HMDSO) was chosen. Plasma‐enhanced processes have
a twofold advantage, which is desired to attain the concept objective. First,
they have the potential to deposit thin‐film layers at reduced temperatures
20 2 Concept
(compatible with Parylene layers). Second, an increased molecular density due
to ion bombardment can be obtained. Using a low‐power (50‐100 W) high‐
frequency (13.56 MHz) capacitively coupled plasma, the monomer is
dissociated and deposited on the electrode. For the PECVD process, additional
oxygen gas (O2) is injected into the reactor to form dense SiOx layers. The
increase of the HMDSO/O2 precursor ratio allows tailoring of the purely
inorganic coatings (SiO2) to lower densities, giving more polymer‐like properties
(SiOxCy). A detailed description of the fabrication process of SiOx thin‐film
coatings [61] and of polymeric / inorganic multilayer structures is presented in
sections 8.1 and 8.3.
The second multilayer concept is based on standard Parylene and plasma‐
dissociated high‐density Parylene‐like layers. To obtain these barrier layers, the
Parylene monomers are dissociated by plasma initiation to form amorphous
carbon‐like layers, which are much denser than Parylene itself. The monomer
gas flow rate is determined by the amount of vaporized monomer, which
depends on the temperature of the vaporisation chamber. The monomer gas is
injected as precursor to the 13.56 MHz driven PECVD process. In order to
optimise the fragmentation for layers with higher density, the RF frequency can
be matched to specific resonance frequencies that favour the cracking of the
Parylene molecules. Using this PECVD process, the degree of monomer
fragmentation can also be modified by varying the ratio of RF power and
monomer gas flow. Hence, the layer composition can be tailored, from
polymer‐like layers to diamond‐like carbon coatings, using a higher RF power.
In order to further increase the density of such a layer, the RF electrode can be
biased, which enhances the effect of ion bombardment. Additional
densification can be achieved by copolymerisation using monomer precursors
like C2H4, CH4, or H2.
2 Concept 21
Figure 5 shows an optical microscope image of a deposited high‐density plasma
Parylene‐C multilayer structure with 8 alternating layer dyads (pairs). Further
details of the fabrication process of polymeric / high‐density polymeric
multilayer structures are presented in section 8.4.
Figure 5. Optical microscope image of a multilayer composed of conventional Parylene‐C and high‐density plasma Parylene‐C layers.
As for the SiOx coatings, a compromise has to be found between denser
diamond‐like carbon layers and their associated stress formation. Thicker
barrier layer increase the tightness of the total stack but also increase the
stress. In order to reduce the stress between the high‐density and the adjacent
Parylene layers, their thicknesses have to be matched. The elastic Parylene
layers help to relieve the stress which is introduced during PECVD depositions.
Each Parylene layer will form a relaxation zone as shown in Figure 6. A
limitation of the Parylene layer thickness is set by the requirement that the
effective percolative pathway is diminished. Above this thickness, the total
barrier tightness decreases. A trade‐off has to be found between barrier
tightness and stress formation.
22 2 Concept
Figure 6. Illustration of a barrier stack comprising of barrier layer (d) and polymeric Parylene layers (D). The
introduced stress by the PECVD deposition can be absorbed by the elastic Parylene layers.
The integration of these high‐density multilayers to implantable devices with
complex surface structures requires a pinhole‐free conformal overgrowth of all
implant surface patterns. In order to achieve this, a homogenization of the
profile can be done by means of an intermediate smoothing layer as shown in
Figure 7. For this purpose, a smoothing micro filler (SMF), such as silicone or
epoxy, can be applied on the implant so that the total conformity of thin
packaging barriers can be improved. The addition of this SMF will smooth,
homogenize, remove defaults and cover/round sharp contours/edges of a
medical device structure before the barrier stack is deposited. Furthermore, its
roughness is reduced in a way that the subsequent deposition of the multilayer
seals the implant hermetically at molecular level. Due to the elastic properties
of the SMF, a further stress reduction can be achieved by its application. The
SMF allows for multilayer packaging of any implantable device in a way that
independent of its three dimensional structure, a conformal coating can be
achieved. The surface energy of the SMF has to be as low as possible for this
2 Concept 23
reason. Detailed explanations of the working principle and fabrication methods
of the SMF are presented in section 8.6.
Figure 7 Principle of the packaging concept using a smoothing micro filler (yellow) and a high‐density multilayer barrier (red/orange).
The last concept represents a completely different package strategy by
alternating Parylene and thin liquid layers. The application of very thin liquid
stack‐partners has the potential to suppress stress and defect formation. In
addition, it can be expected that liquids with a thickness greater than a few
nanometres are pinhole‐free. The interfaces of the solid to liquid phase
transition will thereby reduce permeation through the barrier stack. In
addition, the liquid can act as a getter material which absorbs specific atoms
like oxygen or water molecules. In order to improve the water tightness of such
a stack, the liquids have to be hydrophobic. For better adhesion properties, the
liquids should have an improved wetting on Parylene layers and a low viscosity.
Moreover, the Parylene surface can be treated or activated in order to increase
the liquid wettability.
An advantage of the Parylene process is that the layers can be deposited stress‐
free and conformally on low vapour pressure liquids ( < 7 Pa ) without shape
deformation [35]. Hence, liquids like biocompatible silicone oils or polyethylene
glycol are preferred. The liquids can be deposited by condensation, dip coating
or spraying in situ in the Parylene deposition chamber. Another advantage of
24 2 Concept
this stack concept is that the last liquid layers of a stack can be used as
functional drug‐eluting coating. Active pharmaceutical agents such as
antibacterial, anti‐inflammatory or growth simulating drugs can therefore be
added to the liquid. For a controlled release of the drugs, the porosity of the
adjacent Parylene layers can be tailored by varying the total deposition
pressure during their deposition. Further descriptions concerning the
polymer / liquid layer concept and its fabrication process are shown in section
8.5 and [62].
Finally, the work of this thesis was mostly focused on polymeric / inorganic
multilayer barriers. The following chapter will give a theoretical insight into the
permeation mechanism through polymeric, inorganic and combined multilayer
barriers.
3 Theory 25
3 Theory
In this chapter, the permeation through polymers and ceramic layers will be
discussed. Permeation is an intrinsic material property and dependent on the
structure of the materials, which results in different diffusion models. The
theory and models presented, are focused on Parylene and silicone oxide
multilayer materials.
3.1 Permeation
For the presented multilayer structures, the permeation is defined as a
penetration of the solid material by permeants. These permeants are often in
liquid or gaseous phase, such as water or oxygen, respectively. The permeation
process can be explained by the general diffusion theory and is composed of
four different stages as shown in Figure 8.
Figure 8. Illustration of the permeation process composed of the four different stages: adsorption, absorption, diffusion and desorption. The particles are indicated in blue and the solid membrane in orange.
For homogenous materials, the entire permeation process at steady state is
defined by the permeability coefficient . is detemined as a product of the
diffusion coefficent and the solubility coefficent as shown in equation ( 1 ).
26 3 Theory
The material‐specific coefficent quantifies the diffusion process and the
absorption process.
∙ ( 1 )
The permeability specifies the gas volume per time which penetrates the
material under a defined pressure. Different units of the permeability
coefficient are used depending on literature. For the sake of simplicity, only the
international system (SI) of units was introduced in this thesis. As consequence,
[ ] is specified in mol m‐1 s‐1 Pa‐1.
In general, the diffusion process in solids is much slower, as opposed to the
sorption processes. Thus, the diffusion is the primary time determining stage of
the permeation. The diffusion coefficient at constant pressure and temperature
can be described by the first Fick law as follows:
⋅ , ( 2 )
where is the particle flux or the amount of diffusing substance per time and
per area, and the internal molar concentration gradient at constant
pressure and temperature. The diffusion for one dimension can be described
by:
∙dd
( 3 )
3 Theory 27
which shows that the particle flux is proportional to negative molar
concentration gradient. Figure 9 depicts the negative diffusion gradient for two
constant boundary concentrations and for a material with the thickness .
Figure 9. Illustration of the linerar concentration gradient with respect to the constant boundary conditions on either side of the slab of material.
In order to determine the concentration of gases, it is more practical to
measure the differential pressure across the membrane as compared to
concentration measurements. Therefore, the solubility coefficient is used and
this can be described by Henry’s law as follows:
∙ ( 4 )
The law gowerns the solubility or absorption of specific gas molecules which
are dissolved into the material. For a constant temperature, the internal
concentration of absorbed molecules in the material is proportional to the
applied gas pressure outside of the material surface.
28 3 Theory
The ideal gas law describes the relation between the pressure, temperature
and concentration of a gas as follows:
∙ ∙ ∙ ( 5 )
By use of the ideal gas law, the external concentration at a given applied
pressure can be calculated. Using Henry’s law ( 4 ) and the ideal gas law ( 5 ),
the following relation can be found between the external and internal
concentration:
∙ ∙ ∙ ( 6 )
For the calculation of no further assumption has to be made. For the sake
of simplicity, the conversion factor can be used to express the relationship
between and as follows [45]:
( 7 )
By applying Fick’s first law and Henry’s law to the steady‐state concentration
distribution of a gas, a phase transition can be found if < 1. This will lead to a
step in the distribution which depends on as shown in Figure 10.
3 Theory 29
Figure 10. Illustration of gas concentration distribution on a gas-solid interface.
By integration of Fick’s first law ( 3 ) and in combination with Henry’s law ( 4 )
and the permeability equation ( 1 ), assuming that the sorption isotherm is
proportional to temperature and pressure and moreover, the diffusion
coefficient is constant, the relation between the steady‐state particle flux of a
thin‐film with a thickness d and the differential gas pressure ∆ can be
described as:
⋅ ∙dd
∙∆
( 8 )
Equation ( 8 ) can be used in case of steady‐state permeation. If the particle
flux is multiplied by the time t, the total mass, permeating through the
membrane, can be calculated. The total number of particles which have
permeated the substrate with a surface over the time can then be
expressed by:
∙ ∙ ( 9 )
30 3 Theory
For transient permeation through homogenous material, Fick’s second law can
be applied. It predicts the time‐dependent particle flux, where the
concentration is both, position‐ and time‐dependent. Fick’s second law for one
dimension can be described by:
d ,
ddd
∙d ,
d ( 10 )
where is the diffusion coefficient and c the position and time‐dependent
concentration of soluted particles in the material. If the internal concentration
of the second Fick law is replaced by equation ( 4 ), equation ( 10 ) can be
written as follows (assuming to be constant over and to be constant with
respect to time and location ):
dd
∙dd
( 11 )
where is the external applied gas pressure load on the material. The equation
( 11 ) shows that the transient behaviour of only depends on , whereas the
steady‐state value of depends on the product ∙ (which is equal to the
permeability coefficient ). Figure 11 shows an example of the time‐dependent
particle flux through a solid material if a Heaviside function pressure step
(green) is applied. By comparing the experimental and numerical values for the
steady‐state flux and for the flux decay, both parameters and can be
determined (parameter estimation by inverse fitting). This method will be used
in paragraph 5.1.3.
3 Theory 31
Figure 11. Illustration of the time-dependent response of the particle flux (blue) through a solid material by application of a Heaviside pressure step.
3.1.1 Polymerpermeation
Due to the nature of polymers, their material density is much lower compared
to ceramics or metals. Polymers build long molecular chains; therefore the
package density is reduced by the free volume interstices between the
entangled chains. The permeation of polymers is enhanced compared to dense
inorganic materials, because particles can easily penetrate through these free
volumes [63]. For example, the free space between the chains of
polyacrylonitrile (PAN) are in the range of 0.1 to 1 nm [64]. The general
polymer permeation is widely discussed in literature [65‐67]. Standard
amorphous polymers like thermoplastics (e.g. polystyrene), have fewer chain
interconnections and wide free volumes as shown on the left side in Figure 12.
Thermosetting plastics such as epoxy and silicone, have the same basic
structure as thermoplastics but the polymeric chains show a higher degree of
cross‐linking. For this reason, the free volume is reduced. The highest
molecular density, and therefore the lowest diffusion is achieved by crystalline
structured polymers, e.g. the semi‐crystalline Parylene [68].
32 3 Theory
Figure 12. Illustration of chain arrangements of different plastic types. The chains of amorphous thermoplastics are attracted by van der Waals forces. The chains of the cross‐linked thermosets are connected by strong covalent bonds. The crystalline thermoplastics have more van‐der‐Waals bonds and less free volume as compared to the two others.
The ratio between the amorphous and crystalline phase in a polymer is directly
proportional to the gas solubility of the material [69], notably the helium
solubility [70]. The permeation impact of the crystalline polymer fraction of
Parylene has been shown in literature [68]. The helium solubility of a semi‐
crystalline polymer can thereby be described as follows:
∙ ( 12 )
where is the solubility of the amorphous phase and the volume fraction
of the amorphous phase. For polymers and other solid materials, the solubility‐
induced matter transport proceeds much faster than the transport by diffusion.
Hence, at steady‐state measurements, the diffusion determines the
permeation process [71].
3 Theory 33
3.1.2 SiOxpermeation
In order to decrease the permeability of the polymeric thin films, additional
high density inorganic layers can be integrated to form tight composite barrier
laminates. For this reason, PECVD deposited silicon dioxide was chosen as a
material. Due to the nature of PECVD deposition, the SiOx thin film cannot be
considered as a perfect defect‐free layer. Multiple investigations have shown
that SiOx layers have defects (pinholes and cracks) mostly in the nanoscopic size
range [72‐74]. Figure 13 gives a closer look at the different defect scenarios
which may occur. Depending on the defect size, different diffusion mechanisms
can occur. The defects are thereby simplified as cylindrical tubes.
Figure 13. Illustration of different defect types related to the pore size diameter, which defines the diffusion mechanism at standard temperature and pressure.
Due to the defect inclusion in SiOx thin films, the entire permeation is not
determined by the solid lattice diffusion but rather by defect induced diffusion
[45]. To get a closer look at the diffusion mechanisms, the kinetic gas theory
has to be taken into account. Applying this theory, gas particles are assumed to
be hard spheres which are moving around the gas atmosphere with Maxwell–
34 3 Theory
Boltzmann distributed velocities. These particles collide elastically with other
particles and the surrounding pore walls. From the Maxwell–Boltzmann
distribution, the mean velocity of these gas particles can be described by:
8 ∙∙∙
( 13 )
where is the Boltzmann constant, the absolute temperature and the
mass of the individual gas particle. The law shows that the mean velocity of the
molecule is pressure independent. Using the mean velocity of the particles, the
number of collisions per time can be calculated. Hence, an average mean free
path distance , at which a particle is moving before a collision with another
particle occurs, is estimated by the kinetic gases theory as follows:
1
√2 ∙ ∙ ∙ ( 14 )
where is the number density and is the atom or molecule diameter. The
number density can be determined by the use of the ideal gas law which is
pressure and temperature dependent. Table 1 gives a summary of gas particles
diameters.
Table 1: Summary of gas particle diameters .
Particle Diameter [nm]
He 0.22
Ne 0.28
Ar 0.32
H2O 0.32
O2 0.36
N2 0.37
3 Theory 35
A common rule to define the type of the diffusion mechanism is based on the
estimation of the Knudsen number . This number describes the different gas
dynamics regions. express the ratio between the mean free path and the
pore size that the particles have to penetrate.
( 15 )
Figure 14 shows the relation between the defect diameter and the Knudsen
number at the specific helium conditions (25 mbar partial helium pressure at
298.15 K) which are used in the experimental part 4.1 of this thesis.
Figure 14. Relation between the pore size diameter and the Knudsen number for a partial helium pressure of 25 mbar at 298.15 K, calculated using equations ( 14 ) and ( 15 ).
For < 0.01, the particles have more interaction with other gas particles as
compared to the surrounding pore walls. This scenario corresponds to laminar
flow characteristics. For that case, pinholes larger than 1 mm will be
penetrated by helium particles by free diffusion as shown in Figure 13. The free
gas diffusion can be described by a combination of the Chapman‐Enskog theory
and the Sutherland constant [75]. In these conditions one gas is diffusing into
0.001
0.01
0.1
1
10
100
1000
10000
0.001 0.01 0.1 1 10 100 1000
Pore Size Diameter [μm]
Knudsen Number [‐]
Laminar Flow
Knudsen Transition
Molecular Flow
36 3 Theory
another gas medium and particle collisions with the pore walls are insignificant.
For self‐diffusion of one gas into a vacuum, the free diffusion coefficient can by
simplified. In that case, the free diffusion coefficient can be calculated by
the multiplication of one third of the mean velocity and the mean free path
as shown in equation ( 16 ).
13∙ ∙ ( 16 )
For a partial helium pressure of 25mbar at 298.15 K, the free diffusion
coefficient is of around3 10 m /s.
For pore sizes smaller than 10 m ( > 1), the atoms interact more often with
the pore walls than with adjacent particles. The pressure‐independent Knudsen
diffusion becomes dominant. For this reason, the mean free path of equation (
16 ) is replaced by the pore size diameter . Hence, the diffusion coefficient
for Knudsen diffusion can be approximated by:
13∙ ∙ ( 17 )
Using of equation ( 17 ), the diffusion coefficients for pinholes with pore
sizes of 100 to 1 nm are in the range of 4 10 to4 10 m /s,
respectively, as shown in Figure 15.
If the Knudsen number is between 0.01 and 1, the two diffusion coefficients
and play a role for the total diffusion. This area is called Knudsen transition
regime [76]. Figure 15 shows and and the combined diffusion
3 Theory 37
coefficient related to . The figure demonstrates the interaction
between the two diffusion processes contributing to at a partial helium
pressure of 25mbar at 298.15 K.
Figure 15. Relation between , , and in the Knudsen transition regime with respect to the pore size
diameter for a partial helium pressure of 25mbar at 298.15 K, calculated using the equations ( 16 ) and ( 17 ).
In micro‐porous media, the total Knudsen coefficient depends on the porosity
of the layer [77, 78]. The porosity fraction is defined as the ratio between the
void volume and the total volume. For the corresponding SiOx layers, the
porosity is very low. In the literature, surface defect densities of around 80
pinholes / mm2 with a mean pore size of 1.2 μm were found [79]. This
corresponds to a porosity of9 10 . For multilayered structures, the two
additional parameters constrictivity and tortuosity have to be taken into
account. The constrictivity means that the cross‐section of the total diffusion
path changes. In this work, the pores are assumed to be cylindrical with a
constant diameter and unconnected from each other. For this reason the
constrictivity was not applied. The tortuosity describes the elongation of the
devious effective diffusion path through the porous structure compared to
1.E‐07
1.E‐06
1.E‐05
1.E‐04
1.E‐03
1.E‐02
1.E‐01
1.E+00
1.E+01
1.E‐09 1.E‐07 1.E‐05 1.E‐03
Série1
Série2
Série3
Pore Size Diameter [μm]
Diffusion coefficient [m
2/s]
38 3 Theory
straight‐path diffusion. The tortuosity depends on the mean angle between
the concentration gradient direction and the pore direction [80]. It can be
calculated as follows:
1
cos ( 18 )
If it is assumed that all pores have the same geometric form, direction, and a
homogenous contribution, the diffusion coefficient of porous media can be
estimated from the Knudsen diffusion coefficient of one single pore by:
∙ ( 19 )
For higher than 1000, corresponding to pore sizes smaller than 10 nm,
particles are adsorbed on the pore walls. In the case of surface diffusion, it is
assumed that only a monolayer of particles is adsorbed [71]. A part of these
particles diffuse onto the surface to areas with a lower occupation density. This
requires that the particle energy is higher than the activation energy. The
surface diffusion coefficient can be calculated as follows:
∙ ∙ ( 20 )
where is the occupancy density of particles on the surface, the
adsorption energy, and the activation energy which defines the energy that
a particle needs to move across adsorption points on the surface.
3 Theory 39
In the case of surface diffusion, has to be higher than . Otherwise the
particle would be absorbed into the bulk material. Surface diffusion coefficients
are typically in the range of 1 10 to1 10 m /s.
Finally, permeants will also diffuse through the solid material. This effect can be
explained by vacancy‐atom diffusion and lattice defect diffusion. The lattice
diffusion through the bulk material can be described generally by a
temperature‐dependent Arrhenius‐type law as follows:
∙ ∙ ( 21 )
where is a pre‐exponential factor or the diffusion coefficient at infinite
temperature. defines the energy which a particle needs to migrate from one
site to another in the solid lattice of the material. Activation energies for
helium diffusion through silicon oxide layers on porous substrates are in the
range of 5 kcal/mol [81]. Due to the negative coefficient in the exponential
function, will increase with higher temperatures. For dense material with
amorphous or crystalline structures, is much lower compared to the surface
diffusion coefficent . Depending on the material and its structure, is in the
range between 1 10 and1 10 m /s.
For example, for a Pyrex glass (80% SiO2, Corning 7744) a of 7.74
10 m /s at 300 K was measured [82].
In the case of the amorphous SiO2, a tetrahedral local‐ordered network was
proposed by Zachariasen [83] as shown in Figure 16. The SiO2 forms thereby
circular configurations where particles have the possibility to penetrate these
interstices and can diffuse through the solid silicon oxide.
40 3 Theory
Figure 16. The illustration shows the glassy structure model of silicone dioxide of Zachariasen. The tetrahedral SiO2 formation builds circular arrangements where particles can penetrate.
3.1.3 Combinationofthematerials
As mentioned in the previous chapter, SiOx is the material of choice due to its
tightness and biocompatibility. These ceramic layers are tight at molecular‐
level; however the layer growth tends to incorporate defects due to stress
formation. In terms of these defects, the layer thickness is limited to a critical
cracking thickness (CCT). If the layer thickness exceeds the CCT, a greater
number of defects are created due to the high internal stress in the layer.
A laminar stack of polymer and ceramic layers has the potential to combine the
advantages of both materials while reducing their disadvantages.
If Parylene‐C is used as the polymeric interlayer, the polymer will act as a stress
relieving layer due to its high elasticity. Moreover, the polymeric deposition
method capability to overgrow and coat defects and pinholes helps to decrease
the total permeation of the multilayer. Parylene is often deposited by a CVD
3 Theory 41
process which means that the dimer [2.2] paracyclophane (or halogenated
derivates thereof) is vaporised to a gaseous monomer phase, which
polymerises afterwards onto the substrate to from poly‐p‐xylylene (Parylene)
[39]. Theoretically, the size of the monomer molecule govern the smallest
defects which can still be filled by the Parylene deposition. Therefore, the
molecule size has been investigated by the use of the chemistry development
software ChemSketch. The software takes into account the covalent and the
van der Waals radii as well as bonding forces. For a Parylene‐C monomer, a
molecule size of 0.66 to 0.76 nm was calculated, as shown in Figure 17.
Figure 17. Calculated three‐dimensional structure of a Parylene‐C monomer. The carbon atoms are indicated in grey, hydrogen atoms in red and the chlorine atom in green. The van der Waals radius, which indicated the theoretical hard sphere of the atom, is visualised by the spherical points around the nucleus.
From the calculated Parylene‐C molecule size, it can be hypothesized that
defects down to 1 nm can potentially be filled by this CVD process. The
molecular polarisation, due to the chlorine atom, implies that the substrate
surface energy can influence the above mentioned defect coating performance.
≈ 0.76 nm
≈ 0.66nm
42 3 Theory
Finally, the number of laminate alternations (dyads) plays an important role in
the total barrier performance. As soon as particles have permeated the
Parylene layer, the diffusion is strongly attenuated by the adjacent high‐density
SiOx interface. Most of the particles have to diffuse through the SiOx defects.
The defects of all SiOx interlayers thus constitute a percolative path through the
entire multilayer, which is much longer compared to the barrier thickness. For
this reason, the particle transport is dominated by the alternation of interfaces
rather than by the total multilayer coating thickness.
3.1.4 Multilayerpermeationmodels
For multilayer laminate structures, the different material permeations have to
be taken into account. If the materials are perfectly homogeneous, the total
permeation of the multilayer can be assumed to be proportional to the inverse
of the serial addition of each layer thickness divided by its permeation. For this
case the homogenous laminate model can be applied [45].
For a n‐layer system, the total layer permeability coefficient can be
calculated by:
( 22 )
where is the number of layers, the individual layer thickness, the total
layer thickness, and the individual permeability coefficient. If it is assumed
that for a bimaterial system (e.g. polymeric/ceramic stack) the thickness of
each material is constant, the equation ( 22 ) can be simplified as follows:
3 Theory 43
1∙ ( 23 )
where is the number of dyads, and are, respectively, the thickness
and permeability coefficient of a single polymer layer, and and are the
thickness and permeability coefficient of a single ceramic layer as illustrated in
Figure 18.
Figure 18. Illustration of a homogenous laminated multilayer model composed of a polymeric material (red) and a ceramic barrier material (yellow). The permeation is represented by blue arrows.
As already discussed in section 3.1.2, the ceramic layers can contain perforating
defects. If the permeation of the polymer is much higher than the
permeation of the ceramic , the total surface of the defects determine the
overall permeation, as long as the defect density remains relatively low. For
this case, the defect surface fraction can be calculated by the total defects
surface divided by the total surface .
( 24 )
Dyad
44 3 Theory
Related to section 3.1.1, the polymer used has the possibility to seal defects
down to 1nm. Therefore, it can be assumed that the defects are filled by the
polymeric material as illustrated in Figure 19.
Figure 19. Illustration of defect permeation model where defects are filled by the polymeric material.
The total permeation can be calculated by addition of the permeation through
the defect‐free part of the stack and the permeation through the defects.
Figure 20 shows a multilayer stack, where the defects are aligned in a row.
Figure 20. Illustration of defect permeation model where the defects are aligned. The defect permeation is represented by dark blue arrows and the defect‐free permeation by the pale blue arrow.
3 Theory 45
For a row aligned multilayer stack, the following equation can be applied:
1 ∙1∙ ( 25 )
In real ceramic layers, the defects are not aligned but rather distributed
arbitrarily and their position and the size are random as illustrated in Figure 21.
The defects create a percolative path, which is much longer than the thickness
of the total multilayer barrier. The effective diffusion path length
( /cos ) between two adjacent ceramic layers is thereby considerably
increased. Thus, the repetition of stacks increases the effective diffusion path
for permeants through the multilayer and thereby, drastically reduces the
particle permeation. For this case the second term of the formula ( 25 ) has to
be adapted to:
1∙ 1
∙ ( 26 )
Figure 21. Illustration of percolative pathway model where the effective diffusion length by stacked barrier dyads. The increased defect permeation pathway is represented by dark blue arrows and the defect‐free permeation by the pale blue arrow.
46 3 Theory
3.1.5 Waterpermeation
To protect biomedical implants from body fluids, the water permeation is an
important consideration, since the implant must resists permeation of aqueous
solutions. Compared to inert gas diffusion, the permeation of liquids is not only
dependent on the physical properties of the material (porosity, density) but
also on physiochemical effects due to electro osmotic pressure, capillarity and
swelling. These effects are caused by the polar characteristics of the water
molecule and the ion content of body fluids. The layer interfaces will thereby
play and dominant role for the total barrier tightness. A common indicator of
water tightness is the water vapour transmission rate (WVTR). It describes the
mass of water which passes through a thin film with the surface area per day.
Different studies investigated the water vapour permeation of barrier films [84‐
86]. For highly tight barrier films, standard measurement methods cannot be
applied due to the lack of ultra‐low WVTR detectors [87]. For this reason, the
so‐called calcium mirror test is used [88]. Using this method, the highly reactive
elemental calcium (Ca) is corroded by the permeated water and forms calcium
hydroxide (Ca(OH)2) and hydrogen. The total chemical corrosion can be
described by the following stoichiometric reaction:
Ca 2H O → Ca OH H ( 27 )
of:
Ca H O → CaO H ( 28 )
CaO H O → Ca OH ( 29 )
3 Theory 47
In order to measure this calcium degradation, two different calcium methods
are proposed in literature [89, 90]. The first method is based on a resistivity
measurement of a metallic calcium thin film. A test substrate is therefore
coated with a 250 nm thick elemental calcium film. On this highly corrosive
calcium layer, the thin‐film barrier is deposited. The test substrate is
subsequently exposed to water vapour. The ingress of water will provoke
formation of non‐conductive Ca(OH)2, which increases the measured resistivity
of the Ca layer.
In the second method, a test substrate is coated with a thin elemental Ca film
of around 100 nm and protected by the barrier structure. During exposure to
water vapour, one side of the metallic calcium mirror is illuminated by a
homogenous backlight and observed by a grey‐scale camera on the opposite
side. Due to the water exposure, the calcium mirror is corroded to transparent
calcium hydroxide. This effect increases the global optical transmission of the
test samples. From a grey scale image analysis, the remaining calcium thickness
can be determined. Thus, the ingress of water can be calculated from the
stoichiometric reaction ( 27 ). Using the optical calcium test, two different
degradation effects can be observed [91]. For a barrier layer, which has a
constant water permeation rate over the entire surface, a homogenous
corrosion of the Ca layer can be observed as shown on the left side of the
Figure 22. If the barrier structure has deficiencies due to pinholes or cracks,
local corrosion occurs, which forms perforating cavities as illustrated on the
right side of the figure.
48 3 Theory
Figure 22. Illustration of corrosion effects. On the left side, a constant decrease of the calcium thickness due to
homogenous corrosion will increase the optical transmission of the calcium layer. On the right side, the defect
corrosion will also increase the optical transmission by gap growth.
For homogenous corrosion, the thickness of the elemental calcium layer is
degraded uniformly. The relationship between the thickness and the
transmission can be explained by the Lambert‐Beer law as follows:
( 30 )
where and are, respectively, the intensities of the source and transmitted
light, is the absorption coefficient and the thickness of the Ca layer. In the
case of defect corrosion, standard geometric optics can be used for defects
down to the micrometer range. The intensity will be proportional to the defect
surface as follows:
∙ ( 31 )
where is the total defect surface and is the entire surface exposed to
water vapour. For defects below the micrometer range, Huygens‐Fresnel
3 Theory 49
diffraction effects have to be taken into account but are insignificant for the
presented Ca transmission measurement. For experimental measurement of Ca
degradation, a combination of both mechanisms can be expected as shown in
Figure 23.
Figure 23. The figure shows the corrosion progression of a barrier layer over time. Homogenous and defect
corrosion can be observed. Image reproduced with permission from [91].
If the optical transmission is measured and the corrosion mechanisms are
known, the thickness of the Ca layer can be calculated using a weighted mean
of the results from both methods [91]. Figure 24 shows the relation between
the optical transmission and Ca thickness of each effect and the combination of
these two border cases, if the reaction ( 27 ) is not oversaturated.
Figure 24. Illustration of the relationship between optical transmission and calcium thickness, depending on the
different permeation mechanisms.
As explained above, the mass of permeated water can be calculated, if the
mass of corroded calcium is known. Using the optical transmission
50 3 Theory
measurements, the local corroded calcium thickness of the layer can be
determined by a grey‐scale analysis. As depicted in Figure 25, the volume of
corroded calcium can be calculated by integration of over the entire
surface.
d d ( 32 )
Figure 25. Illustration of corroded calcium volume (shaded area) and unaffected calcium (grey).
Hence, the mass of corroded calcium is calculated by multiplication of the
specific calcium density of 1.55 g cm‐3 and by:
∙ ( 33 )
The water vapour transmission rate is defined by the mass of permeated water
per surface area and the exposure time as follows:
⋅ ( 34 )
By use of the stoichiometric relation ( 27 ) and integration of equation ( 33 )
into ( 34 ), the WVTR can be calculated by:
3 Theory 51
∙ ∙
⋅ ( 35 )
where (=2) is the stoichiometric coefficient of equation ( 27 ), and
and are the molar mass of water (18 g mol‐1) and calcium (40.1 g mol‐1),
respectively.
If the optical calcium transmission test is applied on stacked multilayer
structures ( > 2 SiOx interlayer), no line‐ or crack‐shaped corrosion pattern can
be observed. Longitudinal patterns are only generated by cracks of single
barrier layers. The observed highly localized round corrosion pattern leads to
the conclusion that the defect diffusion determine the permeation process.
This circular pattern can be explained by lateral diffusion into the polymeric
layer and defect diffusion in the ceramic layer as shown in Figure 26.
Figure 26. Schematic crack permeation through multilayered barrier structure that consists of four Parylene layers and three high‐density and tight SiO2 interlayers. The first, second, and third water propagation waves are depicted in the colours blue, red, and green, respectively. Due to lateral diffusion, only round corrosion patterns are formed.
52 3 Theory
The permeant diffuses homogenously through the first polymeric layer.
Assuming that a longitudinal crack pattern exists in the first ceramic layer, the
water will be transported through it and gives rise to a lateral diffusion front in
the second Parylene layer that is indicated by the blue waves. As soon as it
reaches the closest spot of the crack in the underlying second ceramic layer,
the water will leak into the third polymeric layer underneath and creates at the
first time a concentric circle‐shaped diffusion front (red). The third ceramic
layer re‐enforces the tendency of circular water propagation (green) that gives
rise to a random point distributed corrosion pattern of the calcium on the
bottom of the multilayer stack.
3 Theory 53
3.2 Layeruniformity
For high hermetic ceramic barrier layers, the permeation strongly depends on
the conformity of the thin‐film laminate. In order to achieve an optimal
hermeticity of the polymeric/ceramic multilayers, a constant coverage of
subsequent layers is required. Local deficiencies can lead to failure of the entire
packaging. Therefore a method was developed to verify the ceramic barrier
film uniformity and thereby exclude local deficiencies.
State of the art methods, which describe layer uniformity, are based on
polished two‐dimensional substrates, such as standard silicon wafers [92‐95].
These methods measure the thickness distribution on the flat substrate and
allow for the calculation of the average thickness deviation [96].
Due to the lack of applicable methods to define the three‐dimensional
uniformity, an algorithm was developed to quantify the conformity of the
deposited multilayer stacks.
Figure 27 illustrates the working principle of this algorithm. The green curves
represent the lower and upper boundaries of a thin film cross‐section with a
non‐uniform film thickness that is deposited over an arbitrary curved substrate.
By use of the algorithm, the orthogonal line is constructed at any point on the
lower boundary of the thin film and at its endpoint is set at the length
corresponding to the layer thickness. The upper boundary of an ideal uniform
packaging layer without variation of thickness is represented by the upper
dashed red curve. The difference between the points of the ideal envelope
(red) and the upper green boundary indicates the non‐conformity.
54 3 Theory
Figure 27. Algorithm to define the ideal conformal layer in comparison to the experimental results. Assuming l is the uniform thickness of the layer and the slope at x0 of the lower curve f(x0) is f’(x0), the slope of the normal n(x0) is n’(x0).
By the use of the coordinates and from the lower border line of a thin
film, the transformed coordinates ( , ) of the envelope of an ideal layer
can be calculated as follows:
1
tan ( 36 )
where is the slope of the normal of the lower boundary and is the
angle between the x‐axis and the normal .
To define and for an ideal layer thickness , the following
trigonometric relations were used:
sintan
1 tan ( 37 )
f(x)
g(u)
x
u
x0
u0
f(x0)
g(u0)
x0-
l
f’(x0)
n’(x0)
3 Theory 55
cos1
1 tan ( 38 )
By insertion of the equation ( 36 ) in ( 37 ), the coordinate transformation
→ of the ideal layer envelope can be calculated as follows:
∙ tan
1 tan ′ ∙ 1 ′ ( 39 )
The other coordinate transformation → of the ideal layer envelope can be
calculated by use of the equations ( 36 ) and ( 38 ) as follows:
1 tan 1 ′ ( 40 )
The sign of the functions ( 39 ) and ( 40 ) depends on the angle . Assuming
that the curvature is continuous and differentiable, the developed
algorithm is limited by the ideal layer thickness . This means that can only
have values up to the smallest curvature radius of .
The application of these equations is shown in the experimental section 5.4
and 8.2.
56 3 Theory
4 Materials and Methods 57
4 MaterialsandMethods
One of the most important properties of the thin‐film barrier layer is its
tightness. The layer has to prevent atomic and molecular penetration to
protect the implantable device against corrosion and, conversely, the human
body from possible toxic components of the implanted electronic circuits. This
layer tightness can be characterised by several methods which are mostly
based on concentration differences between the two sides of the layer. This
chapter focuses on the developed of helium and water permeation
measurement setups. These two measurements take the atomic diffusion and
the molecular water diffusion into account. The last section 4.3 deals with the
development of the deposition reactor and the fabrication procedure for the
multilayer structures.
4.1 Heliumpermeationtestsystem
The helium permeation is based on the atomic diffusion of helium through the
material and quantified according to the US military packaging standard MIL‐
STD‐883 [97]. Helium is chosen due to its small atomic diameter, which will
penetrate the material more easily than larger atoms or molecules. Another
advantage is the preclusion of interactions between the permeant and the
material itself due to its inert gas property. In addition, the helium content of
ambient air is very low at around 5 ppm [98] and thus, contributes only a low
background to the measurement system in the event of diffusion through
polymeric gasket rings.
For the measurement, the barrier layers were deposited onto a self‐supporting
polyvinylchloride (PVC) membrane. These membranes are introduced into a
high vacuum system to measure the gas permeation as shown in Figure 28.
58 4 Materials and Methods
Therefore, one side of the membrane is loaded with a defined gas pressure
(e.g. 20mbar). The pressure is controlled by a precise needle valve (micro flow
leak valve, Alcatel) and a bypass rotary vane pump. This pressure is measured
by a gas‐type‐independent capacitive transducer with a ceramic diaphragm
(VSK 3000, Vacuubrand). On the other side of the membrane, an ultra‐high
vacuum (UHV) is generated by means of a primary pump and a turbomolecular
pump (Turbovac 50, Leybold). A supplementary ion getter pump is connected
to the UHV section to further diminish residual gases. The ion getter pump
(TiTa 10S Ion Pump, Gamma Vacuum) is composed of titanium / tantalum
plates which gives an enhanced pumping rate for helium. The UHV pressure is
measured by a Bayard‐Alpert ion gauge (Atmion, Vacom). Due to the helium
concentration difference, helium atoms penetrate the membrane and (a
fraction) will be measured by a quadrupole mass spectrometer (LC‐D300M,
DYCOR®), equipped with an additional secondary electron multiplier (SEV). The
SEV allows for an increased minimal detectable helium partial pressure down
to 5 × 10‐14 mbar. Using this test setup, the permeation of different noble gases
(e.g. He, Ar, Ne) or other gases such as O2, H2O or CO2 can be measured.
Figure 28. Illustration of the diffusion system. The self‐supporting test membrane (red) is between the helium pressure and a UHV section. Helium atoms are depicted by blue points.
4 Materials and Methods 59
The value, measured by the mass spectrometer, corresponds to the partial
pressure of specific atoms or molecules depending on their mass‐to‐charge
ratio. To characterise the permeation, the mass flow of helium gas has to be
calibrated by means of the partial pressure and the pump speed of the vacuum
system. Different helium test leaks (TL7 and TL8, Inficon) with a predefined
flow‐rate of 7.3 × 10‐7 and 3.5 × 10‐8 mbar l‐1 s‐1 were therefore connected to
the UHV section. Hence, the measured partial helium pressures were
correlated to the calibrated helium leak rates. To calculate absolute
permeation rates, the measured helium pressure background level of around
1 × 10‐11 mbar was subtracted from the measured value.
Figure 29. Picture of the developed diffusion test setup.
60 4 Materials and Methods
For the self‐supporting membranes, a PVC substrate was chosen due to its high
helium permeation coefficient. To ensure homogenous and reproducible
helium diffusion, pre‐stamped membranes with a thickness of 75 µm were
used (Nitto 224, Permapack). The membranes are placed on top of a thin steel
ring in order to avoid deformation by fixation forces. In addition, the steel ring
guarantees a well‐defined surface aperture for the permeation measurement.
Finally, the barrier layers are deposited onto the self‐supporting membrane
structure as shown in Figure 30.
Figure 30. Self‐supporting membrane structure composed of a PVC membrane and a thin steel ring. The membranes are placed onto a Viton O‐ring and introduced into the diffusion test setup.
To characterise the maximum membrane load of the elastic regime, the
hysteretic behaviour, as a result of the applied pressure load, was measured.
For this purpose, different helium pressure loads (cycled) between 0 and
400 mbar were applied across to the coated membranes. After each load, the
pressure is decreased to the preceding pressure load in order to compare the
partial pressure difference of the two measurements. Figure 31 shows the
plastic behaviour of a PVC membrane, coated with a multilayer stack of five
Parylene‐C layers (1 µm) alternated with four SiOx interlayers (240 nm). The
plastic behaviour of the membrane could be detected at a threshold pressure
load difference of 100 mbar, corresponding to a calculated strength of about
Membrane
Steel‐ring
O‐ring
Support
4 Materials and Methods 61
2.6 N. To guarantee the mechanical integrity of the multilayer stacks, the
helium permeation was investigated at a helium pressure load of 20 mbar.
Figure 31. Hysteresis measurement which indicates maximal pressure load to prevent plastic membrane deformation. The helium pressure load (red) and the measured partial helium pressure (blue) are indicated.
4.2 Waterpermeationtestsystem
In order to measure the hermeticity of a package, the water vapour
transmission rate (WVTR) is often used as a common indicator. It specifies the
permeated mass of water vapour through the encapsulation in g m‐2 day‐1. The
WVTR can be detected by mass spectrometry or by humidity indicators. To
measure low water permeation rates (< 1 x 10‐3 g m‐2 day‐1), elemental calcium
indicators have been established due to their highly corrosive reaction with
water as low as 1 x 10‐6 g m‐2 day‐1. Therefore, multilayer barrier stacks have
been compared to conventional Parylene‐C layers according to the ASTM F
1249 standard. The WVTR investigations have been carried out in collaboration
with Frauenhofer institute PYCO in Germany.
For this reason, conventional Parylene‐C and multilayer structures were
deposited on top of 50 x 50 mm2 and 36 μm thick polyethylene terephthalate
(PET, Mylar™) carrier substrates. After deposition, the films were dried in
vacuum at 60°C for 48 h. For the backside encapsulation of the calcium mirrors,
62 4 Materials and Methods
glass slides of the same size and a thickness of 700 μm were cut and dried on a
hot plate at 250 °C for 2 h. After the drying process, thin calcium films (Ca,
distilled, dendritic, Aldrich) of around 110 nm thickness (±10%) and size of
21 x 30 mm2 were deposited by thermal evaporation onto the barrier films. The
backsides of the calcium mirrors were encapsulated using the dried glass slides
and an UV‐cured adhesive as shown in Figure 32. The adhesive is applied by a
dispenser over the whole encapsulation area. Finally, the adhesive was cured
under a UV‐light source.
Figure 32. Illustration of an encapsulated calcium mirror in order to investigate WVTR of the barrier films. The mirror has been deposited directly on the barrier film and afterwards encapsulated by an UV‐adhesive and a glass substrate on the backside.
In order to guarantee the least possible corrosion of the calcium film before the
measurements, the calcium layer was deposited and manipulated under a
nitrogen (N2) glove box facility at the PYCO institute as shown in Figure 33.
Figure 33. Nitrogen glove box facility used at the Frauenhofer institute PYCO in order to guarantee least possible corrosion of the test samples prior to measurements. Image reproduced with permission from the Frauenhofer institute PYCO.
4 Materials and Methods 63
To investigate WVTR, the encapsulated calcium mirrors were placed into a
climate chamber at 38 °C and 90% relative humidity (RH) and at ambient
atmosphere (25 °C and 35% RH). The samples were analysed over a period of 2
months. During the first two weeks, the measurements were performed daily,
and then weekly. For the period of measurements, samples were taken out of
the climate chamber for about 1h.
The WVTR analyses have been performed using two different test setups. The
optical transmission setup (OTM) is composed of a laser diode above the
calcium mirror test sample and a photo diode on the rear side. This detector
will measure the transmission of the corroded calcium mirrors (Figure 34a)).
The laser screened the measurement area for each sample at 70 points in order
to obtain a WVTR value that corresponds to the total mirror surface. From the
measurement of the calcium light transmission, the WVTR can be calculated.
The second test setup (MicroPerm) is equipped with a CCD camera on one side
and a homogenous backlight on the other side (Figure 34 b)). The setup allows
for the characterisation of the defect formation as described in section 3.1.5.
Figure 34. Different test setups at the PYCO institute in order to investigate the WVTR. The optical transmission setup (OTR) is shown on the left and the CCD equipped MicroPerm on the right side. Image reproduced with permission from the Frauenhofer institute PYCO.
b)a)
64 4 Materials and Methods
4.3 Depositionreactor
For the fabrication of ultra‐tight multilayer barrier stacks, a novel single
chamber deposition system has been developed. A high conformal deposition
process, in addition to the standard Parylene method, were chosen and
integrated into a single deposition chamber. The LPCVD Parylene process was
therefore combined with a PECVD SiOx process to guarantee highly conformal
coatings of multilayer barriers, as shown in Figure 35. A laminar flow facility
(clean room class 10) was constructed around the reactor chamber in order to
decrease particle contamination and to guarantee optimal conditions for part
handling.
Figure 35. Developed single‐chamber deposition reactor.
The Parylene was deposited by the conventional Gorham process [39], where a
solid Parylene precursor, e.g. a stable cyclic dimer (di‐p‐xylene) or halogenated
derivatives, is vaporized at a temperature of around 150 °C. This vaporized
precursor then passes through a pyrolysis oven which decomposes the dimer
4 Materials and Methods 65
into reactive monomers at a temperature of around 650 °C. In the deposition
chamber, the reactive monomers polymerise to deposit in a stress‐free way
onto the substrate at room temperature. Typical Parylene layer thicknesses
vary between 100 nm and 10 m.
For the second multilayer material, SiOx was chosen due to its excellent barrier
properties at low deposition temperatures [46, 79]. For the SiOx deposition, the
organosilane HMDSO was used as precursor. The low‐power plasma‐coating
process, carried out typically at 50 W for 5 min, allows a deposition
temperature below 50 °C. Thus, damage or unsoldering of the implantable
electronic circuits components can be excluded. Figure 36 shows the developed
fully‐automated user interface of the multilayer deposition reactor. The reactor
valves are automatically switched depending on the multilayer structure and
the process parameters. All important process parameters are recorded in
order to verify the process cycles after deposition. More detailed process cycles
are described in sections 8.1 and 8.2.
Figure 36. User interface to control and monitor deposition process parameters. The reactor valves are switched automatically depending on the pre‐programmed multilayer structure. All important process parameters are monitored and recorded.
66 4 Materials and Methods
5 Results 67
5 Results
This chapter presents to the results which were obtained during the thesis
work. In the first section, detailed helium permeation measurements and their
corresponding simulation models, water permeation measurements, and
surface / uniformity analyses are presented. More detailed experimental
results are presented in two additional publications listed in sections 8.1 and
8.2. During the thesis, four world‐wide patents have been granted. These
patents are based on the three multilayer concepts presented in chapter 2 and
listed in sections 8.3 to 8.6.
5.1 Stead‐stateheliumpermeation
As described in chapter 4.1, the helium gas permeation of Parylene‐C, SiOx, and
multilayer laminate barriers has been measured using the diffusion test system
developed.
In order to measure a particle flux by a mass spectrometer, the flow‐rate
related to the measured partial helium gas pressure has to be calibrated.
For this purpose, two standardised test leaks have been connected to the test
system. The calibrated leaks TL8 and TL7 have a constant helium flow‐rate of
3.5 10 and 7.3 10 mbar l‐1 s‐1, respectively. These flow ranges have
been chosen according to the measured permeation rate of the barrier layers.
Figure 37 shows the relation between measured partial helium pressure and
corresponding flow‐rate.
68 5 Results
Figure 37. Relationship between the helium test leak flow‐rate and measured partial helium pressure.
Figure 37 shows the linear relationship between the helium test leak flow‐rates
and the measured partial pressures. As a result of these measurements, a
pressure / flow‐rate conversion factor = 310.64 l s‐1 could be determined. By
use of and equation ( 8 ), the permeability coefficient can be calculated as
follows:
∙∆
∙∙ ∙
∙ ( 41 )
First measurements have been focused on the determination of for
commercially‐available polyethylene terephthalate (PET, Mylar®,
DuPontTeijinFilms) in order to verify the performance of the measurement
system. Figure 38 shows the obtained particle flux measurements related to
the PET thicknesses of 12, 19, 23, 36, and 75 μm and their calculated ,
respectively. The measurements revealed a mean = 4.1 x 10‐16 mol m‐1 s‐1
Pa‐1 (STP), which is in agreement with literature values [45].
1.00E‐10
6.00E‐10
1.10E‐09
1.60E‐09
2.10E‐09
2.60E‐09
3.10E‐09
1.00E‐081.10E‐072.10E‐073.10E‐074.10E‐075.10E‐076.10E‐077.10E‐078.10E‐07
Helium Test Leak Flow Rate [mbar l s‐1]
Measured Partial Helium Pressure
[mbar]
3.0 x 10‐9
2.5 x 10‐9
2.0 x 10‐9
1.5 x 10‐9
1.0 x 10‐9
5.0 x 10‐10
1.0 x 10‐10
1 x 10‐8 2 x 10‐7 4 x 10‐7 6 x 10‐7 8 x 10‐7
5 Results 69
Figure 38. Relation between the measured helium particle flux of PET membranes and their thicknesses. The secondary axis shows the corresponding permeation coefficients.
Figure 39 shows the proportional relationship between the measured partial
helium pressure and the reciprocal of , as predicted by the first Fick law ( 3 ).
Figure 39. Relation between measured partial helium pressure and the reciprocal membrane thickness.
In order to produce self‐supporting substrates, prefabricated PVC membranes
were chosen due to their high helium permeability. The helium permeability of
uncoated membranes was first verified by the test system. The PVC carriers
3.00E‐16
4.00E‐16
5.00E‐16
1E‐09
1.1E‐08
2.1E‐08
3.1E‐08
4.1E‐08
5.1E‐08
6.1E‐08
7.1E‐08
8.1E‐08
0 20 40 60 80
PET Thickness [μm]
Particleflux [m
ol m
‐2s‐1]
Permeationcoefficent
[mol m
‐1s‐1Pa‐
1]
8 x 10‐8
7 x 10‐8
6 x 10‐8
5 x 10‐8
4 x 10‐8
3 x 10‐8
2 x 10‐8
1 x 10‐8
1 x 10‐9
5 x 10‐16
4.5 x 10‐16
4 x 10‐16
3.5 x 10‐16
3 x 10‐16
1E‐10
3E‐10
5E‐10
7E‐10
9E‐10
1.1E‐09
1.3E‐09
1.5E‐09
1.7E‐09
0.01 0.03 0.05 0.07 0.09
1/d [ μm‐1 ]
PartialHelium Pressure
[mbar]
1.7 x 10‐9
1.5 x 10‐9
1.3 x 10‐9
1.1 x 10‐9
9 x 10‐10
7 x 10‐10
5 x 10‐10
3 x 10‐10
1 x 10‐10
70 5 Results
showed a reproducible permeability coefficient of = 3 x 10‐15 mol m‐1 s‐1
Pa‐1 at STP for a PVC thickness = 75 μm, a predefined membrane surface by
the steel ring aperture of = 2.6 cm2, and a helium pressure load = 20 mbar.
The measured value of is comparable to literature data [99].
5.1.1 Bilayermeasurements
For first bilayer barrier experiments, standard Parylene‐C layers deposited on
the PVC carriers were used, in order to define the helium permeability
coefficient of Parylene. For this purpose, Parylene‐C layers with different
thicknesses of 1.9, 4.8, 9 and 15.3 μm have been deposited on the 75 μm PVC
carrier membranes and subsequently measured. By use of equation ( 22 ), the
permeation coefficient 2 of the additional deposited Parylene layer can be
calculated as follows:
( 42 )
where is the thickness of the deposited layer, and and the total
thickness and the total permeation coefficient of the membrane, respectively.
The measured helium particle flux and the corresponding permeation
coefficient of Parylene‐C are shown in Figure 40.
5 Results 71
Figure 40. Relation between the measured helium particle fluxes of Parylene‐C coated PVC membranes and their thicknesses. The secondary axis shows the corresponding permeation coefficients.
As described in section 3.1.2, the SiOx layers determine the high gas tightness
of the multilayer barrier. The SiOx composition and thereby the material
properties, such as the tightness, depends strongly on the deposition
parameters [100]. For this reason, a systematic study was performed to obtain
an insight into the relation between the precursor gas composition and the
helium permeation of the deposited layer. Hence, the oxygen/HMDSO gas ratio
for the PECVD deposition was varied and correlated to the He permeation of
the corresponding SiOx layers. Figure 41 shows permeation measurements for
the gas ratios 7.5, 10, 12.5, 15, and 17.5 with SiOx thicknesses between 70 and
700 nm.
3E‐16
5E‐16
7E‐16
4.00E‐08
5.00E‐08
6.00E‐08
7.00E‐08
8.00E‐08
9.00E‐08
1.00E‐07
1.10E‐07
0 5 10 15
Thickness [μm]
Particleflux [m
ol m
‐2s‐1]
Permeationcoefficent
[mol m
‐1s‐1Pa‐
1]
1.1 x 10‐7
1 x 10‐7
9 x 10‐8
8 x 10‐8
7 x 10‐8
6 x 10‐8
5 x 10‐8
4 x 10‐8
8 x 10‐16
7 x 10‐16
6 x 10‐16
5 x 10‐16
4 x 10‐16
3 x 10‐16
72 5 Results
Figure 41. Relationship between the measured helium pressures of SiOx coated PVC membranes and their thicknesses depending on the precursor oxygen/HMDSO gas ratio. The CCT of each ratio is shown by the red dashed line.
It could be shown that an increase of the oxygen content leads to a higher
tightness of the deposited layer due the formation of stoichiometrically denser
SiO2. Figure 41 reveals the CCT (red dashed line) of each gas composition at
which the permeation of deposited layer starts to re‐increase. The CCT is
decreased for higher oxygen contents due to the stress formation during the
deposition. For the different SiOx layers, a permeation coefficient
between 6 x 10‐17 and 5 x 10‐18 mol m‐1 s‐1 Pa‐1 could be calculated, which is in
the same range as reported in literature [45, 81, 82]. For ratios higher than
17.5, a decay of tightness could be observed. It can be concluded that
depending on the application (rigid or flexible) of the multilayer barrier, the
optimal SiOx composition has to be chosen.
8.0E‐10
1.0E‐09
1.2E‐09
1.4E‐09
1.6E‐09
1.8E‐09
2.0E‐09
2.2E‐09
0 200 400 600 800
Ratio 7.5
Ratio 10
Ratio 12.5
Ratio 15
Ratio 17.5
SiOx Layer Thickness [nm]
Partial Helium Pressure [mbar]
CCT
5 Results 73
5.1.2 Multilayersystems
The results of the SiOx precursor study were implemented into the multilayer
fabrication process and an oxygen/HMDSO gas ratio of 12.5 was used for the
barrier stack deposition. This ratio has been chosen as an optimal trade‐off
between tightness and formation of low‐stress multilayer structures. Laminates
with 1, 2, and 4 SiOx interlayers were fabricated and their helium permeations
were measured. Figure 42 shows the corresponding results of the permeation
measurements which were compared to single SiOx and Parylene‐C layers.
Figure 42. Comparison between SiOx, Parylene‐C and multilayer barriers deposited on PVC carrier membranes. The multilayer structure of 5.9 μm (4 SiOx interlayers) shows the same tightness compared to a Parylene‐C layer of 40 μm. The SiOx layer achieves the tightness maximum at around 500 nm.
Figure 42 shows the improvement in degree of tightness using multilayer
barriers compared to single Parylene‐C and SiOx layers. It could be proved that
a barrier stack with four SiOx interlayers fulfils the acceptance criterion of the
long‐term packaging standard MIL‐STD‐883 [97]. Further details concerning the
MIL‐STD‐883 standard are described in detail in section 8.1.
1.0E‐10
6.0E‐10
1.1E‐09
1.6E‐09
2.1E‐09
0 5 10 15 20 25 30
SiOx Layer
Parylene‐C
Multilayer
Layer Thickness [μm]
Partial Helium Pressure [mbar]
74 5 Results
5.1.3 Transientheliumpermeation
The above described steady‐state experiments allowed determination of the
permeability coefficient . To obtain independently the two coefficients and
( ∙ ), the dynamic (transient) behaviour of helium permeation,
described by the second Fick law ( 10 ), can be used. As a consequence,
dynamic tests on PET, PVC and Parylene‐C single‐layer membranes were
performed. For these tests, a helium pressure difference of 20 mbar was
applied across the membrane and the steady‐state permeation was measured.
Afterwards, the helium was pumped out by the rapid opening of a fast‐acting
valve. The valve has been installed directly between the helium load container
and the pump system, whereby its opening gave a fast decay of the pressure
load. Finally, the pressure decay was measured and used in the time‐
dependent diffusion simulations to determine the coefficients and . The
first dynamic tests have been performed on 23 μm thick PET membranes. The
corresponding graph is shown in Figure 43.
Figure 43. Transient behaviour of a 23 μm thick PET membranes (blue) by application of a fast helium pressure load decay (red).
0
5
10
15
20
25
0
1E‐08
2E‐08
3E‐08
4E‐08
5E‐08
6E‐08
0 2 4 6 8 10
Measured He Pressure
He Pressure Load
Time [s]
Measured Partial Helium Pressure [mbar]
Helium Pressure Load
[mbar]
6 x 10‐8
5 x 10‐8
4 x 10‐8
3 x 10‐8
2 x 10‐8
1 x 10‐8
0
5 Results 75
In order to synchronize the measurements of the mass spectrometer and the
pressure gauge, the system clock of the computer has been used for both data
acquisitions. Noise reduction could be achieved by a repetitive measurement of
5 times followed by an averaging. The value, when the decay of the partial
helium pressure has reached 50% of the initial pressure, was chosen as
parameter, taken for the estimation of and by inverse fitting. The
parameter estimation by inverse fitting will be described in detail in the section
5.2.1. This time‐dependent measurement has been applied in addition to PVC
and Parylene‐C membranes.
5.2 Simulation
The goal of this simulation study was the better understanding of the nano‐
and microscopic diffusion of gases through multilayered thin‐film structures.
The conception of different theoretical models, concerning the permeation for
helium gas, was therefore necessary. A finite element simulation was
subsequently carried out, using the COMSOL software application. Finally, the
simulations of the different theoretical models were compared to the
experimental diffusion measurements.
5.2.1 Singlelayermodel
The first simulation model focused on self‐supporting single PVC membranes.
For this purpose, the membrane geometry was designed using COMSOL. A
radius of 200 μm and a thickness of 75 μm were chosen for the membrane’s
dimensions as an optimal trade‐off between a good representation of the real
membrane and a fast computation of the model.
76 5 Results
The following initial boundary conditions were chosen to simulate realistic
conditions using the model:
⦁ The concentration at the upper side of the membrane.
⦁ The concentration at the lower side of the membrane = 0 mol m‐3.
⦁ The initial concentration across membrane = 0 mol m‐3.
⦁ No flux passes through the membrane side‐walls.
Figure 44 shows the simulation of the steady‐state concentration gradient of a
PVC membrane, if a helium pressure load difference of 20 mbar is applied
across the membrane.
Figure 44. Simulation of the concentration gradient of a 75 μm thick PVC membrane by application of a helium gas pressure of 20 mbar.
5 Results 77
Finally the simulated particle flux at steady state has been compared to the
experimental values for the verification of the model. The simulated values
were in good agreement with the experimental measurements. The same
simulation has also been carried out for single PET and Parylene‐C membranes.
Simulated transient behaviour
From the experimental steady‐state permeation measurements, the stationary
value of can be determined experimentally, which yields ∙ , if
membrane thickness is known, as described by equation ( 2 ). By minimisation
of the error between the experimentally‐determined decay (section 5.1.3) and
the simulated decay, the value of the diffusion coefficient , and thereby ,
can be estimated by inverse fitting.
In the following section the procedure to extract the parameters and is
described, once the time‐dependent experiments of section 5.1.3 have been
performed.
1. Using the experimental pressure load data, the time of pressure decay is
determined and marked as time .
2. From the mass spectrometer partial helium pressure data, the following
parameters are determined:
a. The steady‐state flux .
b. The time at which the flux has decreased to 50% of . This
yields the time ∆ ( ) that has been elapsed from
beginning of pressure load decay to arrive at /2.
78 5 Results
3. ∙ ∙ can be determined from the known thickness of the
membrane and from the experimental data.
4. In the COMSOL simulation, the following variables are introduced:
a. The external concentration , which is given via ideal gas law by
the external helium pressure load.
b. The internal concentration ∙
c. The diffusion coefficient / , where is a mathematical
substitution variable.
d. This yields a system with two unknowns, and . It is
guaranteed by the choice of the unknowns that by changing , the
product ∙ remains constant, so that the steady‐state flux
only depends on .
5. As a result, Fick’s law at steady state can be written as: ∙
6. This leaves only one unknown , which is now determined by matching
the experimental ∆ to the numerically calculated ∆ .
a. The experimental measured pressure load is integrated into the
simulation as input boundary, as shown in Figure 45.
Figure 45. Normalized pressure function used in simulation.
5 Results 79
b. A parameter study on has been performed, as shown in Figure 46.
Figure 46. Parametric study of the simulated flux for different values of s [‐].
c. The study yields the value of , which best fits the dynamic decay at
50 % of the experimental measurement.
Figure 47. Enlarged section of the parametric study. The value of 50% of for the measured time ∆ is highlighted with the dashed red line. The line indicates the corresponding value of .
80 5 Results
As does not influence , the steady‐state value of stays constant. From this
parametric study, and are known. The diffusion coefficient then can be
calculated by / and by either / or / ∙ .
By use of the mentioned fitting method, the following and values for PVC,
PET and Parylene‐C has been investigated and summarized in Table 2.
Table 2: Summary of the solubility and diffusion coefficients of PVC, PET, and Parylene‐C.
PVC PET Parylene‐C
[‐] 0.008 0.014 0.016
[mol m‐3 Pa‐1] 3.2 x 10‐6 5.7 x 10‐6 6.5 x 10‐6
[m2 s‐1] 9.2 x 10‐10 7.2 x 10‐11 4.9 x 10‐11
This is in good agreement with literature values, were a PET helium solubility
coefficient of 3.5 x 10‐6 mol m‐3 Pa‐1 was found [70]. The table shows the
increased helium transparency of PVC compared to PET and Parylene.
5.2.2 Ideallaminatemodel
The laminate model describes the ideal scenario, where the permeation of a
multilayer structure can be calculated by superposition of homogeneous single
layer permeations, as described in equation ( 22 ). Due to the fact that SiOx
layers cannot be produced as self‐supporting membranes, a bilayer SiOx / PVC
was simulated. Using this model, the permeation coefficient of the SiOx layer
could be determined, assuming that the size distribution of defects and their
position are homogenous. This will lead to an ‘overall’ diffusion coefficient,
which includes the effect of defects.
5 Results 81
Experimental measured of a PVC (75 um) and SiOx (170 nm) bilayer revealed a
helium particle flux of 8.23 x 10‐8 mol m‐2 s‐1. Applying the ideal laminate
theory, a of 3.26 x 10‐17 mol m‐1 s‐1 Pa‐1 can be calculated. Then, assuming
a solubility of 6.1 x 10‐6 mol m‐3 Pa‐1 [45], gives a coefficient of
5.3 x 10‐12 m2 s‐1.
By introducing the values of and the corresponding internal
concentration of the top layer surface into Comsol, the related total flux
can be simulated. Figure 48 shows a simulation of a bilayer model with a
membrane radius r = 50 μm and a total membrane thickness of 75.17 μm,
composed of the SiOx (170 nm) coated PVC membrane (75 μm). A total flux of
8.22 x 10‐8 mol m‐2 s‐1 was obtained by simulation. This is in a good agreement
with the experimental measurements.
Figure 48. Simulation of SiOx coated PVC membrane where the concentration gradient over the total
membrane is illustrated.
Due to the lower diffusion coefficient of SiOx compared to that of PVC, the
highest concentration decay is associated with the SiOx coating on top of the
membrane.
82 5 Results
The SiOx simulations were subsequently integrated into the model in order to
build up the multilayered structure. Using the ideal laminate theory implies
that no pinholes are modelled in the SiOx layer. For each layer, a ‘global’
diffusion coefficient was used, which includes the homogenous defect
distribution. For this reason, the value of = 5.34 x 10‐12 m2 s‐1 of the
preliminary bilayer study was taken. For the PVC membrane and the Parylene‐C
layer, the values of = 9.2 x 10‐10 m2 s‐1 and = 4.9 x 10‐11 m2 s‐1 have
been used for the simulations. In order to decrease the calculation time, a
reduced surface of 5 x 5 μm was simulated. The individual layer thicknesses of
Parylene‐C and SiOx were set to 170 nm and 1 μm, respectively. The helium
concentration (20 mbar, STP) was applied to the top surface of the multilayer
structure. The following figures show the concentration decay over multilayer
structure with 1, 2, and 4 SiOx interlayers at steady state.
Figure 49. Enlarged cross‐section of a single multilayer coating (2 x Parylene‐C and 1 x SiOx layers) on top of the
PVC carrier membrane. The simulation shows the concentration gradient at steady state.
5 Results 83
Figure 49 illustrates the barrier function of SiOx interlayer due to the higher
concentration decay compared to the polymeric layers. For this scenario, a
total particle flux of 4.38 x 10‐8 mol m‐2 s‐1 could be simulated. The same
simulation parameters have been applied to a multilayer with two SiOx
interlayers as show in Figure 50.
Figure 50. Enlarged cross‐section of a single multilayer coating (3 x Parylene‐C and 2 x SiOx layers) on top of the
PVC carrier membrane. The simulation shows the concentration gradient at steady state.
Compared to the single multilayer, a higher concentration decay over the total
barrier could be observed. For the dual multilayer, a = of 3.7 x 10‐8 mol m‐2 s‐1
was simulated.
84 5 Results
Finally, a multilayer with four SiOx interlayers was simulated, as shown in Figure
51. This stack shows the highest concentration decay over the total barrier at
steady state. In addition, the simulation revealed the inferior influence of the
Parylene‐C layer. Using this model, a = 2.5 x 10‐8 mol m‐2 s‐1 was calculated.
Figure 51. Enlarged cross‐section of a single multilayer coating (5 x Parylene‐C and 4 x SiOx layers) on top of the
PVC carrier membrane. The simulation shows the concentration gradient at steady state.
In order to validate the model concept, the time‐dependent behaviour has
been simulated and compared to the experimental measurements for a
corresponding multilayer stack. Figure 52 a) shows the experimental measured
time‐dependent behaviour of the multilayer with four SiOx interlayer. A
dynamic decay at 50 % of the experimental measurement was achieved after
5.27 s. A comparable lag time could be revealed by the proposed simulation
model, as shown in Figure 52 b).
5 Results 85
Figure 52. Comparison of time‐dependent behaviour between experimental measured multilayer (a) and the
corresponding simulation model (b).
By the use of the homogenous laminate theory, the simulations predicts higher
permeation than the experimental values of the corresponding multilayer
barriers. This is consistent with literature, where refined models have to be
applied for multilayer structures composed of polymeric and inorganic layers
[54, 101]. The simulation confirmed that the diffusion through the proposed
multilayered structures cannot be simplified by a succession of homogeneous
materials, but corresponds rather to a percolative pathway model. This can be
0
5
10
15
20
25
1.00E‐08
2.10E‐07
4.10E‐07
6.10E‐07
8.10E‐07
1.01E‐06
1.21E‐06
1.41E‐06
1.61E‐06
0 50 100 150
Measured Partial He Pressure
He Pressure Load
Helium Pressure
Load
[mbar]
Measured Partial Helium Pressure
[mbar]
Time [s]a)
b)
86 5 Results
explained by the influence of material interfaces and the additional defect
diffusion. Figure 53 shows the discrepancy between the experimentally
measured multilayer and the simulated homogenous multilayer model.
Figure 53. Comparison between measured helium permeation (green curve) and simulated permeation (purple
curve) of multilayer barriers.
5.2.3 Percolativepathmodel
In order to refine the simulation to match to the experimental measurements,
the additional diffusion mechanism of the percolative pathways should be
integrated into the simulations. The Knudsen diffusion through micro‐ and
nanoscopic SiOx barriers defects up to 10 μm should be considered (mentioned
in section 3.1.2), whereas the diffusion coefficient of the defect‐free SiOx
surface area has to be decreased. In this way, the proposed homogenous finite
element model should be expanded with this additional diffusion mechanism.
As described in the experimental part of the thesis, no signifcant defects
1.0E‐10
6.0E‐10
1.1E‐09
1.6E‐09
2.1E‐09
0 5 10 15 20 25 30
Experimental SiOx Measurements
Experimental Parylene‐C Measurements
Experimental Multilayer Measurements
Multilayer Simulation
Layer Thickness [μm]
Partial Helium Pressure [mbar]
5 Results 87
greater than 0.25 µm, could be observed by AFM, SEM, and plasma oxygen
etching investigations. Comparing the helium permeation values for bulk quartz
crystal (SiO2) [102] and the experimentally measured values of the deposited
SiOx thin‐films, the latter was found to have an inferior hermeticity by a factor
of 100. Hence, it is hypothesised that this decrease is mostly caused by the
presence of nanoscopic defects. In order to define the size and the distribution
of these defects, an indirect single‐defect observation could be used. To do so,
a study of the corrosion propagation would allow for further investigations of
the initial defect size, using the calcium mirror test and a high‐resolution CCD
camera. If the defect density per area and the mean defect pinhole size of the
SiOx layer is known, a simplified approach could be made by simulation of
diffusion through a single SiOx pinhole defect, as shown in Figure 54. Hence, the
calculation time of the simulation can thereby be kept as low as possible. The
integration of the entire pinhole size and distribution parameters into the
equations of the FEM model would give a further insight into the multilayer
diffusion mechanisms.
Figure 54. Enlarged cross‐section of the SiOx coating on top of the PVC membrane with the integration of a
single pinhole defect.
SiOx Pinhole
88 5 Results
5.3 Waterpermeation
The degradation of the calcium deposited below the multilayer barrier were
investigated by the formation of transparent calcium hydroxide due to
moisture penetration through the barrier layer (section 3.1.5). For this reason,
29 calcium mirrors were deposited on Parylene‐C and multilayer structures.
The mirrors were analysed by the MicroPerm and the OTM test setups (section
4.2). Additional calcium tests are described in sections 8.1 and 8.2.
By use of the MicroPerm setup, the defect corrosion of the encapsulated
calcium mirrors was studied. Therefore, Parylene‐C layers of 1 μm, single SiOx
multilayer with a thickness of 170 μm SiOx and 2 x 1 μm Parylene‐C, double SiOx
multilayer with a thickness of 2 x 170 μm SiOx and 3 x 1 μm Parylene‐C, and
quadruple SiOx multilayer with a thickness 4 x 170 μm SiOx and 5 x 1 μm
Parylene‐C were investigated. The CCD images after a 130 h climate chamber
exposure at 38 °C and 90% RH are shown in Figure 55.
Figure 55. CCD MicroPerm images of calcium mirrors deposited on 1 μm thick Parylene‐C (a), single (b), double
(c), and quadruple multilayers (d) after a 130 h exposure to 38 °C and 90% RH. The pictures reveal higher water
vapour tightness after each additional SiOx layer due to the increased grey‐scale.
5 Results 89
Figure 55 shows the different corrosion behaviour of the barrier structures. The
Parylene‐C coated calcium layers are completely corroded after an exposure
less than a few hours. The single multilayers revealed an increased crack and
pinhole density. The double multilayers showed fewer cracks and slower
corrosive progress of the calcium mirrors. For the quadruple multilayers, only
minor defects are visible. These defects can be explained by deposition errors,
dust particle contamination after the cleaning process, or handling damage.
The measurements of the optical transmission by the OTM were performed at
70 points for each sample as shown in Figure 56. Only the points in the middle
were used to calculate an averaged WVTR in order to avoid boundary effects.
Figure 56. CCD‐picture of encapsulated calcium mirror with a surface area of 21 x 30 mm2. The blue grid shows
the 70 areas where the optical transmission has been measured by use of the OTM setup. Each point can be
selected to calculate the average WVTR.
The thicknesses of the calcium mirrors were calculated from the optical
transmission related to the two different corrosion behaviours, as explained in
section 3.1.5. For uniform thickness reduction, equation ( 30 ) was applied with
90 5 Results
a calcium absorption coefficient = 0.04 [103]. Concerning the defect
corrosion, equation ( 31 ) was used, where the intensity is proportional to the
defect surface. For the calculation of the WVTR, a weighted average of 2:1 from
the uniform and the defect corrosion, respectively, was chosen. By use of this
linearization, the measured transmission curves show mostly the expected
uniform degradation of the calcium thickness over time, if no saturation occurs.
The barrier layer lag time, that is the delay where the calcium is not affected
after the water vapour exposure (corresponding to a complete tightness), has
been excluded for these WVTR calculations.
The results of the WVTR determination in g m‐2 day‐1 at ambient conditions
(25 °C and 35% RH) are summarized in Figure 57. This measurement was taken
from the samples after an exposure of 1500 h.
Figure 57. Summary of measured WVTR at 25 °C and 35% RH of pure PET carrier substrates (36 μm), coated
with Parylene‐C (1 μm), and multilayer structures (1, 2, and 4 SiOx interlayers).
The PET and Parylene‐C coated PET substrates show high WVTR, as expected.
For the barriers structures, WVTRs of 5 x 10‐3 g m‐2 day‐1 for single multilayers,
1.5 x 10‐3 g m‐2 day‐1 for double multilayers, and 6 x 10‐4 g m‐2 day‐1 for
1.E‐04
1.E‐03
1.E‐02
1.E‐01
1.E+00
1.E+01
0 1 2 3 4 5 6
PET
Parylene
ML1
ML2
ML4
Barrier Layer Thickness [μm]
WVTR
[ g m
‐2day
‐1]
5 Results 91
quadruple multilayers could be quantified. The measured WVTRs show a decay
that is approximately proportional to the number of SiOx layers of the barrier
structure. This behaviour expresses the contribution from additional diffusion
processes of the multilayer interfaces and the roles of the different dynamic
effects associated to the percolative pathway model. The influence of, for
example nano‐capillarity and electro‐osmotic pressure, has not been
considered.
In order to measure the layer tightness performance related to human body
conditions, the coated calcium mirrors were exposed to 38 °C and 90% RH. The
measurements were acquired once a day for 120 h. During the data acquisition,
the samples were taken out of the climate chamber for about 1 h each day.
Figure 58 shows a summary of the multilayer structure compared to a
calculated single Parylene‐C layer with a WVTR of 80 g μm m‐2 day‐1 [40].
Figure 58. Summary of measured WVTR at 38 °C and 90% RH of PET carrier substrates (36 μm), coated with
multilayer structures (1, 2, and 4 SiOx interlayers) compared to standard Parylene‐C layers.
Compared to the first measurements at 25 °C and 35% RH, the PET and
Parylene‐C coated substrates could not be measured due to the fast calcium
1.E‐03
1.E‐02
1.E‐01
1.E+00
1.E+01
1.E+02
1.E+03
0 2 4 6 8
Multilayer
Barrier Layer Thickness [μm]
WVTR
[ g m
‐2 day
‐1]
92 5 Results
mirror degradation. In contrast, a slight degradation of the multilayer tightness
could be observed due to the increased humidity and temperature. The highest
tightness was measured for the quadruple multilayer structure with a WVTR of
5 x 10‐3 g m‐2 day‐1. This ultra‐low WVTR shows the exceptional performance of
the barrier structure which is three decades of magnitude lower than standard
Parylene‐C (commonly used for packaging purposes of medical implants). The
measurements seem to indicate an Arrhenius behaviour with a small activation
energy characteristic. In order to quantify the specific diffusion mechanisms at
a higher temperature than the measurement conditions, only the first values
(exposure of 120 h) have to be taken into consideration to avoid the
perturbation of the experimental acquisitions. A more detailed study is in
preparation, where the calcium mirrors can be measured inside the climate
chamber for longer durations at 38 °C and 90% RH.
5.4 Surfaceanduniformityanalysis
In order to analyse the surface of the multilayer structures, scanning electron
microscopy (SEM) and atomic force microscopy (AFM) imaging was performed.
Using these methods, surface roughness, potential pinholes, and cracks can be
observed. Interface and material composition was analysed by Fourier
transform infrared spectroscopy (FTIR) and X‐ray photoelectron spectroscopy
(XPS) and are described in detail in section 8.1.
For the surface analysis, polished silicon wafers were coated with conventional
Parylene‐C layers of 1 μm thickness and multilayer structures (4 x Parylene‐C
(1μm) and 3 x SiOx interlayers (250nm)). The surface analysis was performed by
an atomic force microscope (3100 Dimension, Digital Instruments) in tapping
mode (150 kHz). A scan range of 10 x 10 µm was chosen for the analysis of the
Parylene‐C and SiOx layers as shown in the following figure.
5 Results 93
Figure 59. AFM images with a 10 x 10 µm scan area of a standard 1 µm Parylene‐C layer (a) and a multilayer
composed of four 1 µm Parylene‐C layers and three 250 nm SiOx interlayers (b). For the 3‐D images, the base of
the vertical axis was shifted arbitrary.
Figure 59 reveals a root mean square roughness of 6 nm for the Parylene‐C
layer and 32 nm for the multilayer stack. It can be assumed that the total
multilayer roughness is increased due to roughness accumulation of the
different deposition steps. In addition, the different surface plasma treatments
for surface activation and cleaning, influence the roughness. However, no
signifcant defects (> 0.25 µm) were observed within the surfaces studied.
For structural analysis of the multilayer barrier, the deposited coatings were
ablated by a focused ion beam (FIB) in order to obtain a clean cross‐section
prior to the analysis by a high‐resolution field emission scanning microscope.
The same multilayer structures as for the AFM experiments have been used. A
94 5 Results
thin platinum coating was deposited onto the insulating barrier layer in order
to avoid charging effects and enhance the contrast. The Figure 60 shows the
micrographs of the SEM/FIB analysis.
Figure 60. SEM images of FIB ablated section of the multilayer barrier structure (a), surface analysis of the
multilayer stack (b) and enlarged multilayer cross‐section (c).
The local ablation by FIB, in order to analyse the multilayer cross‐section, is
shown in Figure 60 a). Clean edge trimming can be achieved by appropriate FIB.
The surface of the multilayer structure is visualised in Figure 60 b). Equal to the
AFM measurements, no significant defects could be observed (> 0.25 µm).
Figure 60 c) depict the enlarged multilayer cross‐section from Figure 60 a). This
5 Results 95
cross‐section image confimed a Parylene‐C thickness to be around 1 μm and a
SiOx thickness of around 240 nm.
The SiOx layer conformity was investigated by application of the algorithm,
developed in section 3.2. By the use of the image treatment software ImageJTM,
the lower and upper boundary of the presented SiOx interlayer were extracted
(Figure 60 c)). The extracted SiOx boundaries (black curves) from the
micrograph acquisition and the grey scale image treatment noise (grey dashed
zone) are shown in Figure 61. The red curve shows an ideal conformal layer
that was calculated by application of the equations ( 39 ) and ( 40 ) on the
lower SiOx boundary with an ideal layer thickness = 239 nm. This red envelope
follows the upper boundary of the SiOx thin film within the errors of the
measurement and the image treatment noise. Hence, the deposited SiOx layer
can be considered conformal for the applied multilayer stack on the observed
surface. This leads to the conclusion that the hermetic multilayer packaging will
not be influenced by local thickness variations.
Figure 61. Application of the conformity algorithm on the SiOx border lines (black curves) of the multilayer thin film of Figure 60 c). The perfect conformal layer thickness = 239 nm (red curve) has been constructed by application of the algorithm on the lower black border line. The grey dashed zones show the noise errors of image acquisition and treatment. The analysed SiOx layer can be considered as conformal because the calculated red curve (ideal) is within the acquisition and image treatment error of the upper SiOx border line.
Length [nm]
Amplitude [nm]
96 5 Results
5.5 Invitrocytotoxicityandbiostabilityanalysis
The in vitro cytotoxicity and biostability of the materials, intended for use as
conformal barrier coatings for medical implants, were tested by a LIVE/DEAD
cell study. This study was performed in collaboration with the University of
Applied Science in Geneva (Hepia). In order to investigate the in vitro
cytotoxicity of the materials used for the barrier structures, layers of 2.2 μm
thick Parylene‐C, 150 nm thick SiOx (O2/HMDSO of ratio 10), and 2.9 μm thick
multilayer structure (2 x Parylene‐C and 1 x 150 nm SiOx) were deposited on a
polystyrene (PS) cell culture cluster and sterilized with ethylene oxide prior to
tests. Because of their high sensitivity regarding toxic substances, a human
neural progenitor cell line (ReNcell) was seeded on the different materials.
After an exposure time of 24h and 72h under climatic conditions according to
the ISO standard 10993‐5 (cytotoxicity), the living and dead cells were observed
and compared to a control cluster. The vitality and the mortality of the cells can
be determined by confocal microscopy, using fluorescent markers. Vital Calcein
and propidium iodide staining were used for live and dead cell detection,
respectively. The following figures show the fluorescent microscope images of
the control cluster, the Parylene‐C layer, and the multilayer structure.
5 Results 97
Control after 24h Control after 48h Control after 72h
Figure 62. Control ReNCell cultures on plastic microplate as control cluster (PS).
For the analysed Parylene‐C, SiOx and multilayer coatings, no increased number
of dead cells could be observed compared to the ReNCell control cultures, as
shown in Figure 63 and 64. It can be infered that the developed materials, used
for the multilayer barrier coating, showed no cytotoxic reaction under the
applied test conditions. Additional tests have to be performed in order to proof
the entire biocompatibility (e.g. sensitization, irritation, genotoxicity) of the
multilayer barriers.
Vital Calcein staining
Propidium Iodide staining
Dead Cell
Dead Cell
Dead Cell
Vital Calcein staining
Propidium Iodide staining
98 5 Results
Control after 24h Control after 72h
Figure 63. ReNCell cultures on single Parylene‐C layer.
Control after 24h Control after 72h
Figure 64. ReNCell cultures on multilayer barrier (2 x 1 μm Parylene‐C and 2 x 150 nm SiOx).
Vital Calcein staining
Parylene C treated
Control PS Parylene C treated
Propidium Iodide stainingC
BA
A+B+C
Vital Calcein staining
Vital Calcein staining
A B
C A+B+CPropidium Iodide staining
Vital Calcein staining
6 Conclusion and Outlook 99
6 ConclusionandOutlook
6.1 Conclusion
State of the art implantable medical device packaging are based on
conventional glass or metal bulk encapsulations. They are reliable for long‐term
stability but limit the potential for further miniaturization of smart micro‐
medical implants. The latter aspect is important to facilitate the placement of
implants in areas of the body where little space is available e.g. in small arteries
or veins, in bones, or in organs. Hence, hermetic and biocompatible thin‐film
packaging is a growing key technology which enables the use of these
innovative SMMS devices over a more diverse range of disciplines.
The scope of this thesis focused on synthesis and analysis research directed
towards novel thin‐film packaging concepts that are required for the next‐
generation implants.
The research contribution was based on the fabrication of defined hermetic
multilayer structures. This objective could be attained by the development of a
novel single‐chamber deposition reactor with a combined CVD‐PECVD process,
allowing for in situ manufacturing of the entire multilayer. These processes
have been chosen due to their high step coverage deposition for three‐
dimensional structures at ambient temperature. The single‐chamber reactor
aspect is of utmost importance, because, as is well known, multilayer laminates
depend strongly on their interface properties, in that the number of interfaces
predominate the physical/chemical permeation properties of the entire stack.
Air exposure, associated with the transfer from one coating system to another,
implies that properties and reproducibility of the package can be lost
completely. The in situ deposition then allows for precise interface conditioning
in terms of adhesion layer as well as shear‐stress relief. By process control and
100 6 Conclusion and Outlook
automation, deposition conditions and interface treatments could be defined,
and thereby the material composition and the layer thicknesses could be
managed. A specialised LabviewTM‐based software application has therefore
been developed. In addition, the single‐chamber system allowed for
uncontaminated, well‐defined, and reproducible layer interfaces and thus
yielded improved adhesion and hermeticity properties.
The main analytical part conducted within this thesis was focused on the
characterisation of the innovative hermetic barrier structures and their
validation for long‐term implant purposes. In order to investigate the barrier
performance, different measurement concepts were applied. The first method
was based on pure physical gas diffusion. Thus, interactions between the
permeated atoms and the barrier material could be excluded. Helium gas was
therefore used as permeation indicator due to its small atomic diameter and
inert chemical behaviour. Due to the lack of applicable methods for the
measurement of ultra‐tight thin‐film barriers, the bulk packaging test standard
MIL‐STD‐883 has been adapted in order to measure the intrinsic hermeticity of
deposited thin films on permeable carrier substrates. A systematic study of the
precursor gas composition of the coating process, related to the deposited SiOx
thickness, revealed an optimal trade‐off between hermeticity and stress
formation at an O2/HMDSO gas ratio of 10 and a total thickness of 170 nm for
the static use of the multilayer barriers. In other words, the SiOx layer, and
thereby the multilayer structure, can be tailored depending on the application
(rigid or flexible) and the intended lifetime of the medical implant. By
integration of these SiOx coatings into a barrier stack, a quadruple multilayer (4
x 170 nm SiOx and 5 x 1 μm Parylene‐C) was able to fulfil the long‐term
packaging acceptance criterion of the MIL‐standard.
Compared to gas diffusion, liquid water vapour exposure evokes additional
interaction between the barrier materials and the permeants, such as swelling,
6 Conclusion and Outlook 101
capillarity, and electro‐osmotic effects. Moreover, transport mechanisms of
permeants through the interface are responsible for a number of phenomena
encountered in biological, chemical and physical processes. In particular, the
role of ions in sorption and interface interactions are prevalent. These
dominate the diffusion in the multilayer structure. In order to evaluate the
polar water vapour permeation under human body conditions, a calcium mirror
test was used. Using this test, low WVTR’s of 5 x 10‐3 g m‐2 day‐1 at body
conditions (38 °C and 90% RH of the ASTM F1249) and 6 x 10‐4 g m‐2 day‐1 at
room temperature (25 °C and 35% RH) were measured for the same quadruple
multilayer structure, as mentioned above. In fact, the WVTR is assumed to be
caused by the condensation of the water vapour on the coating surface which
results to a water transmission with a reduce drain force compare to a total
water immersion. Due to this fact, an adapted calcium mirror test verified the
high barrier performance by direct liquid contact of saline solution droplets
(NaCl concentration of 1 g/l, 9 g/l and 20 g/l).
Different analysis were performed in order to investigate the material surface,
homogeneity, composition, interfaces and thickness. First, the surface
morphology was examined using an AFM. A root mean square roughness of
6 nm for the Parylene‐C layer and 32 nm for the multilayer stack could be
measured. The increased roughness of the multilayer can be attributed to the
accumulation of the roughness associated with additional deposition steps and
plasma treatments. The combined dual beam analysis (FIB/SEM) showed the
same surface roughness as the AFM. However, by the use of AFM and SEM, no
signifcant defects (> 0.25 µm) could be detected within the observed surfaces.
In addition, the material homogeneity and the individual layer thicknesses of
the multilayer stack could be verified. By the use of the dual beam method,
gradual interfaces could be identified. A finer‐scale study of these layer
102 6 Conclusion and Outlook
interfaces by XPS deep etching disclosed their material composition. The
atomic SiOx concentration of 58% oxygen, 38% silicon, and 4% carbon gradually
changed up to 20 nm, where the composition of Parylene‐C becomes uniform.
The high oxygen content of the SiOx layer could be verified by additional FTIR
analysis. The FTIR fingerprints of the single SiOx and Parylene‐C layers were
superimposed for the multilayer stack. This result indicated the integrity of
each individual layer by the single chamber deposition. The coating conformity,
using the developed single‐chamber deposition process, could be verified by
the application of a specifically developed algorithm. This analysis showed no
local SiOx thickness deficiencies for the deposited multilayer structures.
Finally, a first in vitro cytotoxicity study showed good tolerance of Parylene‐C,
SiOx and multilayer coatings for ReNCell cultures.
The final analysis part of the current work was focused on the proposition of a
theoretical model of the helium gas diffusion. Different models have been built
up for a better understanding of nano‐ and microscopic diffusion mechanisms,
using the finite element software tool Comsol multiphysics®. The Comsol
simulations have been compared to the results obtained by experimental
measurements. The simulation confirmed that the diffusion through the
proposed multilayered structures cannot be simplified by a succession of
homogeneous materials, but corresponds rather to a percolative pathway
model that takes into account the material defects of the barrier layers (in our
case a Knudsen diffusion through defects in the SiOx layer) where the material
interfaces play an important role.
This thesis showed, by preparation (synthesis) and measurements (analysis)
that through the use of the novel thin‐film technology made it possible to
replace the conventional bulk packaging concepts. The packaging volume can
6 Conclusion and Outlook 103
thereby be reduced to a minimum. This study will allow for the development of
next generation SMMS medical implants. Further, new hermeticity test
methods for medical device thin‐film packaging have been developed.
6.2 Futuredevelopments
Looking at the multilayer laminate development, a further increase of the
security margin of thin‐film structures is possible. The stress formation of the
deposited layers is a crucial factor for the barrier performance. Increased stress
can provoke defects such as cracks and pinholes. Hence, by further
investigations related to the deposition speed and the precursor
concentrations, the stress formation in SiOx can be diminished and even
compensated. Moreover, a SiOx stress reduction will improve the adhesion to
the substrate and also the adjacent interlayers. Preliminary studies exhibit an
interesting indicator for the evaluation of low‐stress layers by wrinkle
formation of SiOx layers on soft polymeric surfaces. Advanced wrinkle analysis
could lead to stress‐free barrier layers and thereby increased hermeticity.
Additional investigations on the influence of the total deposition pressure while
maintaining constant precursor ratios could improve the SiOx layer conformity.
In order to meet the requirement of industrial partners with high volume
production needs, an upscaling of the existing SiOx deposition process will be
necessary.
Finally, in order to proof the entire biocompatibility of the proposed multilayer
barriers for implantable applications, additional biocompatibility tests such as
sensitization, irritation, carcinogenicity, and different toxicity evaluations have
to be performed.
104 6 Conclusion and Outlook
As described in the thesis, the multilayer development was destined for rigid
applications. Due to the increased demand of flexible implants or electronic
circuits, the proposed barrier structure could be tailored to reduce defect
formation during deflection. The flexibility of the SiOx material can be changed
by the variation of its carbon content. A decrease of the O2/HMDSO ratio
results in the incorporation of non‐dissociated carbonic precursors, which
forms flexible polymeric SiOxCy layers. The loss of hermeticity compared to the
denser SiOx layer therefore has to be compensated by additional SiOxCy
barriers, up to an optimal trade‐off between flexibility and hermeticity.
Another scientific challenge will be the exact prediction of the influence of SiOx
and Parylene‐C thicknesses on the total barrier performance regarding the
WTR. Continuing investigation between the difference of WVTR and WTR, by
variation of the layer thicknesses and the number of interfaces, could reveal
further important indicators.
In this context, the proposed finite element model with multiple diffusion
mechanisms should be expanded. As mentioned in the experimental part of the
thesis, no signifcant defects greater than 0.25 µm, could be observed. By
comparison between helium permeation values of SiO2 bulk crystal (quartz)
and the experimentally measured values of the deposited SiOx thin films
(PECVD), a decrease of the hermeticity by a factor of 100 was deduced. Hence,
it is hypothesised that this decrease is mostly caused by the presence of
nanoscopic defects. In order to define the size and the distribution of these
defects, an indirect defect observation by the study of their corrosion
propagation could reveal these parameters. The calcium mirror test can
therefore be modified using a high‐resolution CCD camera for single defect
observation. The integration of these parameters into the FEM equations of the
6 Conclusion and Outlook 105
proposed percolative pathway model would give a further insight into the
actual diffusion mechanisms.
Other barrier materials could potentially meet the stringent requirements for
long‐term packaging. One prospective material would be silicon nitride which is
well known as good barrier material but there is still an actively discussed
controversy about its biocompatibility. The integration of silicon nitrides (SixNy)
into the Parylene‐C multilayer depends on the application of the implant. Thus,
specific biocompatibility tests have to be conducted.
The emerging ALD technology is also a promising candidate for the integration
into the multilayer laminates. It is conceivable to replace the SiOx PECVD
process by a low temperature PEALD. Alternatively, an additional ALD step
could be integrated on top of the SiOx layer for additional defect reduction.
Another possibility would be a variation of the existing process. For an
industrial application, the proposed plasma Parylene multilayer concept is
interesting because only one precursor gas (chloro‐p‐cyclophane) is needed.
The additional HMDSO plasma step can thereby be omitted which leads to a
simplified process with higher throughput. For these reasons, this concept is
worthy of further investigations.
The exceptional properties of graphene, such as its mechanical resistivity,
flexibility and diffusion barrier properties, could make it a future material of
promise for packaging solutions. Moreover, its high electrical conductivity, and
the possibility of integrating graphene‐based transistors, enables the
integration of electric circuits into the barrier using the same material.
106 6 Conclusion and Outlook
6.3 Outlook
Due to the promising outcome of this thesis, some different novel perspectives
have been established. First of all, applications of multilayer barriers on real
medical implants are foreseen. Multiple feasibility studies have been initiated
to investigate the protection performance of the barriers on implantable
electronic devices. For example, the deposition on contact lenses with
integrated sensor systems for intraocular pressure monitoring will be carried
out. Different studies for the protection of implantable pump electronics,
hearing aid devices, and intracranial flow and pressure sensors are in process.
In addition, the work of this thesis opens the possibility for the miniaturization
of existing implants or the novel development of innovative SMMS. Long‐term
prospection will involve the development of flexible electronic implants for
catheter‐based interventions. Furthermore, these barrier structures could be
applied to non‐medically‐related electronic or sensor applications at ambient
conditions where long‐term reliability is the key factor.
7 References 107
7 References
[1] DIRECTIVE 2007/47/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL, Official Journal of the Europe, L 247 (2007), pp. 21‐55. [2] B.D. Ratner, A.S. Hoffman, F.J. Schoen, J.E. Lemons, Biomaterials Science: An Introduction to Materials in Medicine, 2nd ed., Elsevier Academic Press, Amsterdam, 2004. [3] International Organisation for Standardization, ISO 10993 series of standards, Biological evaluation of medical devices. [4] G. Nisato, M. Kuilder, P. Bouten, L. Moro, O. Philips, N. Rutherford, P‐88: Thin Film Encapsulation for OLEDs: Evaluation of Multi‐layer Barriers using the Ca Test, SID Symposium Digest of Technical Papers, 34 (2003) 550‐553. [5] K.M. Striny, Assembly Techniques and Packaging of VLSI Devices, in: Sze, S. M. (ed), VLSI technology, 2d ed., McGraw‐Hill, New York etc., (1988), pp. 566‐611. [6] F. Mansfeld, The use of electrochemical techniques for the investigation and monitoring of microbiologically influenced corrosion and its inhibition – a review, Materials and Corrosion, 54 (2003) 489‐502. [7] G. Loeb, R. Peck, J. Singh, Y.‐H. Kim, S. Deshpande, L. Baker, J.T. Bryant, Mechanical loading of rigid intramuscular implants, Biomedical microdevices, 9 (2007) 901‐910. [8] Design Control Guidance For Medical Device Manufacturers, U.S. Food and Drug Administration, FDA 21 CFR 820.30 and Sub‐clause 4.4 of ISO 9001, (1997), pp. 1‐53. [9] F. Graichen, R. Arnold, A. Rohlmann, G. Bergmann, Implantable 9‐Channel Telemetry System for In Vivo Load Measurements With Orthopedic Implants, Biomedical Engineering, IEEE Transactions on, 54 (2007) 253‐261. [10] P.A. Neukomm, H. Kündig, Passive wireless actuator control and sensor signal transmission, Sensors and Actuators A: Physical, 21 (1990), pp. 258‐262. [11] P.R. Troyk, I.E. Brown, W.H. Moore, G.E. Loeb, Development of BIONTM technology for functional electrical stimulation: bidirectional telemetry, Engineering in Medicinemn and Biology Society, Proceedings of the 23rd Annual International Conference of the IEEE, 2 (2001), pp. 1317‐1320 [12] X. Sun, Y. Zheng, X. Peng, X. Li, H. Zhang, Parylene‐based 3D high performance folded multilayer inductors for wireless power transmission in implanted applications, Sensors and Actuators A: Physical, 208 (2014), pp. 141‐151.
108 7 References
[13] S. Chatzandroulis, D. Tsoukalas, P.A. Neukomm, A miniature pressure system with a capacitive sensor and a passive telemetry link for use in implantable applications, J Microelectromech S, 9 (2000), pp,.18‐23. [14] R. Elmqvist, A. Senning, An implantable pacemaker for the heart, Proceedings of the Second International Conference on Medical Electronics, Smyth CN, ed., Tliffe & Sons, (1959) pp. 253–254. [15] W. Greatbatch, W.M. Chardack, A transistorized implantable pacemaker for the long term correction of complete atrioventricular block, Proc NEREM, 1:8, (1959) p. 643. [16] D.A. Nathan, S. Center, C.‐y. Wu, W. Keller, An implantable synchronous pacemaker for the long term correction of complete heart block, The American journal of cardiology, 11 (1963), pp. 362‐367. [17] J. Buffet, Technological progress in pacemaker design: hermetic sealing, Medical progress through technology, 3 (1975), pp. 133‐142. [18] T.F. Hursen, S.A. Kolenik, Nuclear Energy Sources, Ann Ny Acad Sci, 167 (1969), pp. 661‐673. [19] R. Venugopalan, A. Ginggen, T. Bork, W. Anderson, E. Buffen, In Vivo Study of Flow‐Rate Accuracy of the MedStream Programmable Infusion System, Neuromodulation, 14 (2011), pp. 235‐240. [20] M. Forde, P. Ridgely, Implantable Cardiac Pacemakers, in: J. D. Bronzino (ed), Medical Devices and Systems, The Biomedical Engineering Handbook, 3rd ed., CRC Press, Taylor and Francis Group, Boca Raton, FL, (2006). [21] A. Murari, H. Albrecht, A. Barzon, S. Curiotto, L. Lotto, An upgraded brazing technique to manufacture ceramic‐metal joints for UHV applications, Vacuum, 68 (2002) 321‐328. [22] J.H.‐C. Chang, Y. Liu, Y.‐C. Tai, in: Micro Electro Mechanical Systems (MEMS), 2014 IEEE 27th International Conference on, 2014, pp. 1127‐1130. [23] J. Singh, R.A. Peck, G.E. Loeb, Development of BIONTM technology for functional electrical stimulation: hermetic packaging, Engineering in Medicine and Biology Society, Proceedings of the 23rd Annual International Conference of the IEEE, 2 (2001), pp. 1313‐1316. [24] A. Ginggen, Y. Tardy, R. Crivelli, T. Bork, P. Renaud, A Telemetric Pressure Sensor System for Biomedical Applications, IEEE Transactions on Biomedical Engineering, 55 (2008) 1374‐1381. [25] A. Christou, J.R. Griffith, B.R. Wilkins, Reliability of hybrid encapsulation based on fluorinated polymeric materials, Electron Devices, IEEE Transactions on, 26 (1979), pp. 77‐83. [26] C. Hassler, R.P. von Metzen, P. Ruther, T. Stieglitz, Characterization of parylene C as an encapsulation material for implanted neural prostheses, Journal of Biomedical Materials Research Part B: Applied Biomaterials, 93B (2010), pp. 266‐275.
7 References 109
[27] H. Miyake, T. Ohta, Y. Kajimoto, M. Matsukawa, A New Ventriculoperitoneal Shunt with a Telemetric Intracranial Pressure Sensor: Clinical Experience in 94 Patients with Hydrocephalus, Neurosurgery, 40 (1997), pp. 931‐935. [28] R.A. Weiner, M. Korenkov, E. Matzig, S. Weiner, W.K. Karcz, T. Junginger, Early Results with a New Telemetrically Adjustable Gastric Banding, Obes Surg, 17 (2007), pp. 717‐721. [29] D.F. Lovely, M.B. Olive, R.N. Scott, Epoxy moulding system for the encapsulation of microelectronic devices suitable for implantation, Med. Biol. Eng. Comput., 24 (1986) 206‐208. [30] L.P. Bucky, H.P. Ehrlich, S. Sohoni, J.W.J. May, The Capsule Quality of Saline‐Filled Smooth Silicone, Textured Silicone, and Polyurethane Implants in Rabbits: A Long‐Term Study, Plast Reconstr Surg, 93 (1994) 1123‐1131. [31] J.H. Briston, L.L. Katan, Plastics Films, 2nd ed., Longman Scientific & Technical / The Plastics and Rubber Institute, New York, (1983). [32] P.E.K. Donaldson, The Encapsulation of Microelectronic Devices for Long‐Term Surgical Implantation, IEEE Transactions on Biomedical Engineering BME‐23 (1976), pp. 281‐285. [33] A.G. Dirks, H.J. Leamy, Columnar microstructure in vapor‐deposited thin films, Thin Solid Films, 47 (1977), pp. 219‐233. [34] L.B. Freund, S. Suresh, Thin film materials stress, defect formation and surface evolution, Cambridge University Press, Cambridge, (2006). [35] J. Charmet, O. Banakh, E. Laux, B. Graf, F. Dias, A. Dunand, H. Keppner, G. Gorodyska, M. Textor, W. Noell, N.F. de Rooij, A. Neels, M. Dadras, A. Dommann, H. Knapp, C. Borter, M. Benkhaira, Solid on liquid deposition, Thin Solid Films, 518 (2010), pp 5061‐5065. [36] E.M. Schmidt, J.S. McIntosh, M.J. Bak, Long‐term implants of Parylene‐C coated microelectrodes, Med. Biol. Eng. Comput., 26 (1988), pp. 96‐101. [37] J.P. Seymour, Y.M. Elkasabi, H.‐Y. Chen, J. Lahann, D.R. Kipke, The insulation performance of reactive parylene films in implantable electronic devices, Biomaterials, 30 (2009), pp. 6158‐6167. [38] M. Szwarc, Some remarks on the CH2[graphic omitted]CH2 molecule, Discussions of the Faraday Society, 2 (1947), pp. 46‐49. [39] W.F. Gorham, A New, General Synthetic Method for the Preparation of Linear Poly‐p‐xylylenes, Journal of Polymer Science Part A‐1: Polymer Chemistry, 4 (1966), pp 3027‐3039. [40] L.K. Massey, Permeability Properties of Plastics and Elastomers: A Guide to Packaging and Barrier Materials, 2nd ed., Plastics Design Library / William Andrew Pub., Norwich, NY, USA, (2003), p.132. [41] G. Ozaydin‐Ince, A.M. Coclite, K.K. Gleason, CVD of polymeric thin films: applications in sensors, biotechnology, microelectronics/organic electronics,
110 7 References
microfluidics, MEMS, composites and membranes, Rep Prog Phys, 75 (2012) 016501. [42] G. Dennler, C. Lungenschmied, H. Neugebauer, N.S. Sariciftci, M. Latrèche, G. Czeremuszkin, M.R. Wertheimer, A new encapsulation solution for flexible organic solar cells, Thin Solid Films, 511–512 (2006), pp. 349‐353. [43] J.S. Lewis, M.S. Weaver, Thin‐film permeation‐barrier technology for flexible organic light‐emitting devices, Ieee J Sel Top Quant, 10 (2004), pp. 45‐57. [44] R. Wolf, K. Wandel, C. Boeffel, Moisture Barrier Films Deposited on PET by ICPECVD of SiN(x), Plasma Processes and Polymers, 4 (2007) 185‐189. [45] A.P. Roberts, B.M. Henry, A.P. Sutton, C.R.M. Grovenor, G.A.D. Briggs, T. Miyamoto, M. Kano, Y. Tsukahara, M. Yanaka, Gas permeation in silicon‐oxide/polymer (SiOx/PET) barrier films: role of the oxide lattice, nano‐defects and macro‐defects, J Membrane Sci, 208 (2002) 75‐88. [46] A.G. Erlat, R.J. Spontak, R.P. Clarke, T.C. Robinson, P.D. Haaland, Y. Tropsha, N.G. Harvey, E.A. Vogler, SiOx Gas Barrier Coatings on Polymer Substrates: Morphology and Gas Transport Considerations, The Journal of Physical Chemistry B, 103 (1999) 6047‐6055. [47] D.S. Wuu, W.C. Lo, C.C. Chiang, H.B. Lin, L.S. Chang, R.H. Horng, C.L. Huang, Y.J. Gao, Plasma‐deposited silicon oxide barrier films on polyethersulfone substrates: temperature and thickness effects, Surface and Coatings Technology, 197 (2005), pp. 253‐259. [48] A.S. da Silva Sobrinho, M. Latrèche, G. Czeremuszkin, J.E. Klemberg‐Sapieha, M.R. Wertheimer, Transparent barrier coatings on polyethylene terephthalate by single‐ and dual‐frequency plasma‐enhanced chemical vapor deposition, Journal of Vacuum Science & Technology A, 16 (1998) 3190‐3198. [49] M. Leskelä, M. Ritala, Atomic layer deposition (ALD): from precursors to thin film structures, Thin Solid Films, 409 (2002) 138‐146. [50] S.‐H.K. Park, J. Oh, C.‐S. Hwang, J.‐I. Lee, Y.S. Yang, H.Y. Chu Ultrathin Film Encapsulation of an OLED by ALD, Electrochemical and Solid‐State Letters, 8 (2005) H21‐H23. [51] O. Auciello, B. Shi, Science and Technology of Bio‐Inert Thin Films as Hermetic‐Encapsulating Coatings for Implantable Biomedical Devices: Application to Implantable Microchip in the Eye for the Artificial Retina, in: D. Zhou, E. Greenbaum (Eds.) Implantable Neural Prostheses 2, Springer Science+Business Media, New York, (2010), pp. 63‐84. [52] D.G. Shaw, M. Langlois, Use of vapor deposited acrylate coatings to improve the barrier properties of metallized film, Annual Technical Conference Proceedings, Society of vacuum coaters, (1994), p. 240.
7 References 111
[53] J.D. Affinito, M.E. Gross, C.A. Coronado, G.L. Graff, I.N. Greenwell, P.M. Martin, A new method for fabricating transparent barrier layers, Thin Solid Films, 290–291 (1996), pp. 63‐67. [54] G.L. Graff, R.E. Williford, P.E. Burrows, Mechanisms of vapor permeation through multilayer barrier films: Lag time versus equilibrium permeation, J Appl Phys, 96 (2004) 1840‐1849. [55] A.M. Coclite, G. Ozaydin‐Ince, F. Palumbo, A. Milella, K.K. Gleason, Single‐Chamber Deposition of Multilayer Barriers by Plasma Enhanced and Initiated Chemical Vapor Deposition of Organosilicones, Plasma Processes and Polymers, 7 (2010), pp. 561‐570. [56] M. Yan, T.W. Kim, A.G. Erlat, M. Pellow, D.F. Foust, J. Liu, M. Schaepkens, C.M. Heller, P.A. McConnelee, T.P. Feist, A.R. Duggal, A Transparent, High Barrier, and High Heat Substrate for Organic Electronics, Proceedings of the IEEE , 93 (2005), pp. 1468‐1477. [57] C.C. Chiang, D.S. Wuu, H.B. Lin, Y.P. Chen, T.N. Chen, Y.C. Lin, C.C. Wu, W.C. Chen, T.H. Jaw, R.H. Horng, Deposition and permeation properties of SiNX/parylene multilayers on polymeric substrates, Surface and Coatings Technology, 200 (2006) 5843‐5848. [58] M. Schaepkens, T.W. Kim, A. Gün Erlat, M. Yan, K.W. Flanagan, C.M. Heller, P.A. McConnelee, Ultrahigh barrier coating deposition on polycarbonate substrates, Journal of Vacuum Science & Technology A, 22 (2004) 1716‐1722. [59] J.H.C. Chang, L. Yang, K. Dongyang, T. Yu‐Chong, Reliable packaging for parylene‐based flexible retinal implant in Digest Tech. Papers Transducers‘13 Conference (2013), pp. 2612‐2615. [60] T. Bork, A. Hogg, M. Lempen, D. Muller, D. Joss, T. Bardyn, P. Buchler, H. Keppner, S. Braun, Y. Tardy, J. Burger, Development and in‐vitro characterization of an implantable flow sensing transducer for hydrocephalus, Biomedical microdevices, 12 (2010), pp. 607‐618. [61] D. Magni, C. Deschenaux, C. Hollenstein, A. Creatore, P. Fayet, Oxygen diluted hexamethyldisiloxane plasmas investigated by means of in situ infrared absorption spectroscopy and mass spectrometry, Journal of Physics D: Applied Physics, 34 (2001), pp. 87‐94. [62] H. Keppner, M. Benkhaira, Method for producing a plastic membrane device and the thus obtained device, Patent WO2006063955, (2006), pp. 1‐25. [63] B.U. Duncan, J.; Roberts S., in, Review of Measurement and Modelling of Permeation and Diffusion in Polymers, NPL REPORT DEPC MPR 012, 2005, pp. 68. [64] V. Ravindrachary, H.R. Sreepad, A. Chandrashekara, C. Ranganathaiah, S. Gopal, Thermally induced microstructural changes in undoped and doped
112 7 References
polyacrylonitrile: A positron‐annihilation study, Physical Review B, 46 (1992) pp. 11471‐11478. [65] R.M. Barrer, Diffusion in and through solids, Univ. Press, Cambridge, 1941. [66] M. Lomax, Permeation of gases and vapours through polymer films and thin sheet, Barking, 1980. [67] A.J.c. Malkin, Experimental methods of polymer physics (measurement of mechanical properties, viscosity, and diffusion), Prentice‐Hall, Englewood Cliffs, N.J. , (1983). [68] J. Charmet, J. Bitterli, O. Sereda, M. Liley, P. Renaud, H. Keppner, Optimizing Parylene C Adhesion for MEMS Processes: Potassium Hydroxide Wet Etching, J Microelectromech S, PP (2013) pp. 1‐10. [69] G. Gorrasi, M. Tortora, V. Vittoria, E. Pollet, B. Lepoittevin, M. Alexandre, P. Dubois, Vapor barrier properties of polycaprolactone montmorillonite nanocomposites: effect of clay dispersion, Polymer, 44 (2003) 2271‐2279. [70] A.S. Michaels, J.A. Barrie, W.R. Vieth, Solution of Gases in Polyethylene Terephthalate, J Appl Phys, 34 (1963) 1‐&. [71] K. Kessler, in, O2‐ und H2O‐Permeation durch SiOx beschichtete PET‐Folien, 1994, pp. 117 S. [72] A.G. Erlat, R.J. Spontak, R.P. Clarke, T.C. Robinson, P.D. Haaland, Y. Tropsha, N.G. Harvey, E.A. Vogler, SiOx gas barrier coatings on polymer substrates: Morphology and gas transport considerations, J Phys Chem B, 103 (1999), pp. 6047‐6055. [73] A.S.D. Sobrinho, G. Czeremuszkin, M. Latreche, G. Dennler, M.R. Wertheimer, A study of defects in ultra‐thin transparent coatings on polymers, Surf Coat Tech, 116 (1999), pp. 1204‐1210. [74] A.S.D. Sobrinho, G. Czeremuszkin, M. Latreche, M.R. Wertheimer, Defect‐permeation correlation for ultrathin transparent barrier coatings on polymers, Journal of Vacuum Science & Technology a‐Vacuum Surfaces and Films, 18 (2000) 149‐157. [75] T.G. Cowling, D. Burnett, S. Chapman, The mathematical theory of non‐uniform gases an account of the kinetic theory of viscosity, thermal conduction and diffusion in gases, 3rd ed., Cambridge University Press, Cambridge, 1990. [76] V. Michalis, A. Kalarakis, E. Skouras, V. Burganos, Rarefaction effects on gas viscosity in the Knudsen transition regime, Microfluid Nanofluid, 9 (2010) 847‐853. [77] P. Grathwohl, Diffusion in natural porous media contaminant transport, sorption/desorption and dissolution kinetics, Kluwer Academic Publishers, Boston etc., 1998. [78] J.L. Vázquez, The porous medium equation; Mathematical theory, Clarendon Press, Oxford, 2007.
7 References 113
[79] A.S. da Silva Sobrinho, G. Czeremuszkin, M. Latrèche, G. Dennler, M.R. Wertheimer, A study of defects in ultra‐thin transparent coatings on polymers, Surface and Coatings Technology, 116–119 (1999) 1204‐1210. [80] J. van Brakel, P.M. Heertjes, Analysis of diffusion in macroporous media in terms of a porosity, a tortuosity and a constrictivity factor, International Journal of Heat and Mass Transfer, 17 (1974) 1093‐1103. [81] W.G. Perkins, D.R. Begeal, Diffusion and Permeation of He,Ne,Ar,Kr and D2 through Silicon Oxide Thin Films, J Chem Phys, 54 (1971) 1683‐&. [82] W.A. Rogers, R.S. Buritz, D. Alpert, Diffusion Coefficient, Solubility, and Permeability for Helium in Glass, J Appl Phys, 25 (1954) 868‐875. [83] W.H. Zachariasen, THE ATOMIC ARRANGEMENT IN GLASS, J Am Chem Soc, 54 (1932) pp. 3841‐3851. [84] F. de los Santos, G. Franzese, Relations between the diffusion anomaly and cooperative rearranging regions in a hydrophobically nanoconfined water monolayer, Phys Rev E, 85 (2012) 010602. [85] J. Greener, K.C. Ng, K.M. Vaeth, T.M. Smith, Moisture permeability through multilayered barrier films as applied to flexible OLED display, J Appl Polym Sci, 106 (2007) 3534‐3542. [86] Y.G. Tropsha, N.G. Harvey, Activated rate theory treatment of oxygen and water transport through silicon oxide poly(ethylene terephthalate) composite barrier structures, J Phys Chem B, 101 (1997) 2259‐2266. [87] M. Stevens, S. Tuomela, D. Mayer, Water Vapor Permeation Testing of Ultra‐Barriers: Limitations of Current Methods and Advancements Resulting in Increased Sensitivity, Society of Vacuum Coaters, 48th Annual Technical Conference Proceedings, (2005) p. 189. [88] M.O. Reese, A.A. Dameron, M.D. Kempe, Quantitative calcium resistivity based method for accurate and scalable water vapor transmission rate measurement, Rev Sci Instrum, 82 (2011). [89] G. Nisato, P.C.P. Bouten, P.J. Slikkerveer, W.D. Bennett, G.L. Graff, N. Rutherford, L. Wiese, in: Proc. Int. Display Workshop / Asia Display, Nagoya, Japan, (2001), pp. 1435‐1438. [90] R. Paetzold, A. Winnacker, D. Henseler, V. Cesari, K. Heuser, Permeation rate measurements by electrical analysis of calcium corrosion, Rev Sci Instrum, 74 (2003) 5147‐5150. [91] O. Kahle, M. Bauer, The Ca‐test revisited, in: Michel B. and Lang K.‐D. (ed), Smart Systems Integration and Reliability, Goldenbogen, Dresden, (2010). [92] L. Sansonnens, J. Bondkowski, S. Mousel, J.P.M. Schmitt, V. Cassagne, Development of a numerical simulation tool to study uniformity of large area PECVD film processing, Thin Solid Films, 427 (2003) 21‐26.
114 7 References
[93] V. Mohammadi, W.B. de Boer, L.K. Nanver, An analytical kinetic model for chemical‐vapor deposition of pureB layers from diborane, J Appl Phys, 112 (2012). [94] S. Daniels, D.C. Cameron, Examination of thin film uniformity at the bottom of a hole structure using a 3D sputter simulation package, J Phys Iii, 6 (1996) 1213‐1218. [95] C.T. Wang, S.C. Yen, Theoretical‐Analysis of Film Uniformity in Spinning Processes, Chem Eng Sci, 50 (1995) 989‐999. [96] J.H. Wang, J.D. Shao, K. Yi, Z.X. Fan, Layer uniformity of glancing angle deposition, Vacuum, 78 (2005) 107‐111. [97] MIL‐STD‐883H, Department of Defense, Test Method Standard Microcircuits, Method No. 1014.13, Seal (26 February 2010). [98] E. Gluckauf, F.A. Paneth, The Helium Content of Atmospheric Air, Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences, 185 (1946) 89‐98. [99] L.K. Massey, Permeability Properties of Plastics and Elastomers: A Guide to Packaging and Barrier Materials, 2nd ed., Plastics Design Library / William Andrew Pub., Norwich, NY, USA, (2003). [100] S. Sahli, S. Rebiai, P. Raynaud, Y. Segui, A. Zenasni, S. Mouissat, Deposition of SiO2‐Like Films by HMDSN/O2 Plasmas at Low Pressure in a MMP‐DECR Reactor, Plasmas Polym, 7 (2002), pp. 327‐340. [101] R.I. Hickson, S.I. Barry, G.N. Mercer, Critical times in multilayer diffusion. Part 1: Exact solutions, International Journal of Heat and Mass Transfer, 52 (2009), pp. 5776‐5783. [102] H. Greenhouse, Hermeticity of electronics packages, Noyes Publishing, Park Ridge, (1999). [103] J. Marfaing, R. Rivoira, Optical absorption and plasma resonance in very thin films of calcium, Physical Review B, 15 (1977) 745‐749.
8 List of publications 115
8 Listofpublications
Papers
8.1 Ultra‐thinlayerpackagingforimplantableelectronicdevices
8.2 Protectivemultilayerpackagingforlong‐termimplantablemedicaldevices
Patents
8.3 UltrathinMultilayerPackaging
8.4 PlasmaEnhancedPolymerUltrathinMultilayerPackaging
8.5 PackagingwithActiveProtectionLayer
8.6 ThreeDimensionalPackagingforMedicalImplants
116 8 List of publications
8 List of publications 117
Paper I
Ultra‐thin layer packaging for implantable electronic devices
DOI:10.1088/0960‐1317/23/7/075001
A. Hogg, T. Aellen, S. Uhl, B. Graf, H. Keppner, Y. Tardy and J. Burger,
J. Micromech. Microeng., 23, (2013), 075001, pp. 1‐12.
Copyright IOP Publishing (2014)
Reproduced with permission of the publisher.
118 8 List of publications
Ultra-thin layer packaging for implantable electronic devices
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
2013 J. Micromech. Microeng. 23 075001
(http://iopscience.iop.org/0960-1317/23/7/075001)
Download details:
IP Address: 157.26.65.15
The article was downloaded on 21/05/2013 at 15:41
Please note that terms and conditions apply.
View the table of contents for this issue, or go to the journal homepage for more
Home Search Collections Journals About Contact us My IOPscience
IOP PUBLISHING JOURNAL OF MICROMECHANICS AND MICROENGINEERING
J. Micromech. Microeng. 23 (2013) 075001 (12pp) doi:10.1088/0960-1317/23/7/075001
Ultra-thin layer packaging forimplantable electronic devicesA Hogg1,2, T Aellen2, S Uhl2, B Graf2, H Keppner2, Y Tardy3
and J Burger3,4
1 Graduate School for Cellular and Biomedical Sciences, University of Bern, Switzerland2 Haute Ecole Arc Ingenierie (HES-SO), La Chaux-de-Fonds, Switzerland3 Medos International Sarl, Le Locle, Switzerland4 ARTORG Center for Biomedical Engineering Research, University of Bern, Bern, Switzerland
E-mail: [email protected]
Received 13 December 2012, in final form 19 April 2013Published 21 May 2013Online at stacks.iop.org/JMM/23/075001
AbstractState of the art packaging for long-term implantable electronic devices generally uses reliablemetal and glass housings; however, these are limited in the miniaturization potential and costreduction. This paper focuses on the development of biocompatible hermetic thin-filmpackaging based on poly-para-xylylene (Parylene-C) and silicon oxide (SiOx) multilayers forsmart implantable microelectromechanical systems (MEMS) devices. For the fabrication, acombined Parylene/SiOx single-chamber deposition system was developed. Topologicalaspects of multilayers were characterized by atomic force microscopy and scanning electronmicroscopy. Material compositions and layer interfaces were analyzed by Fourier transforminfrared spectrometry and x-ray photoelectron spectroscopy. To evaluate the multilayercorrosion protection, water vapor permeation was investigated using a calcium mirror test. Thecalcium mirror test shows very low water permeation rates of 2 × 10−3 g m−2 day−1 (23 ◦C,45% RH) for a 4.7 μm multilayer, which is equivalent to a 1.9 mm pure Parylene-C coating.According to the packaging standard MIL-STD-883, the helium gas tightness was investigated.These helium permeation measurements predict that a multilayer of 10 μm achieves thehermeticity acceptance criterion required for long-term implantable medical devices.
(Some figures may appear in colour only in the online journal)
1. Introduction
Packaging which is compatible with miniaturization is a keytechnology for future smart implantable medical devices thatare significantly smaller than existing devices and enablesless invasive surgical interventions. Standard metal or glasshousings increase the implant size and are expensive. Thin-film packaging has the advantage of allowing for reliableand cost-efficient packaging processes of wafer-level batchmanufacturing.
The packaging has to provide a mechanically andchemically stable barrier for two reasons: to protect the bodytissue from potentially toxic material such as the implantelectronics, and to protect the implant itself from moistureand reactive ion penetration. Furthermore, any foreign bodyresponse to the packaging/tissue interface represents a risk
for the patient after implantation. For this reason, thebiocompatibility of the material has to be ensured accordingto ISO 10993 [1].
In the case of active medical implants, the packagingpermits the communication with the implanted device bymeans of antennas which connect the telemetric interfaceto an external reading unit. In radio frequency (RF) passivetelemetry, inductive coupling systems are used to provide theimplant with energy and in return, to transmit digital data fromthe implant [2].
Biocompatible encapsulation of implantable medicaldevices can be achieved with packaging materials that includepolymers, glasses, metals, and ceramics.
For metallic packaging, biocompatible titanium cans orplates that are laser welded to a supporting base plate areoften used. Electromagnetic signals can be transmitted to the
0960-1317/13/075001+12$33.00 1 © 2013 IOP Publishing Ltd Printed in the UK & the USA
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
implant by active telemetry, using limited carrier frequencieslocated in the kHz range. This enables communication byRF telemetry antennas placed within the hermetically sealedmetallic enclosure. Faster communication is achieved bycarrier frequency in the MHz range using passive telemetry.Hermetic feedthroughs, providing conductive leads to anantenna outside the metallic enclosure, are therefore integrated[3]. Furthermore, these ceramic or glass feedthroughs are alsoused for implantable electrodes in pacemakers, defibrillators[4, 5], multichannel neuromuscular stimulators [6], andcochlear implants [7].
Compared to metallic packing, biocompatible glasses(e.g. borosilicate glass) have the advantage of being transparentto RF and optical communication. In contrast, glasses aretypically two orders of magnitude more permeable to moisturethan metals. However, even a glass packaging with a thicknessin the micrometer range can provide a long-term hermetic sealfor months or years [8].
Ginggen et al have shown that a durable long-term glasspackaging solution is competitive and resistant to human fluidsfor an implantable pressure sensor that monitors intracranialpressure [9].
For the most part, packaging made of biocompatible glassrequires low-temperature hermetic bonding between the glassenclosure and the substrate of the implant. This is essential inorder to avoid damage to the implant electronics. Hermeticglass packaging of implantable electronic devices can beachieved, for example, by soldering a borosilicate glass capto a borosilicate base plate with a low temperature brazingprocess of around 185 ◦C. However, the metallic brazing ring,used to encapsulate the implant, modifies the inductivity of theimplant antenna by mutual inductive coupling. It represents ashort-circuited metallic conductor which absorbs a part of theRF energy and thus reduces communication distances betweenthe biomedical implant and an external control unit [10].
Polymer encapsulation generally depends on chainstiffness, molecular orientation, free volume, cohesive energydensity, moisture sensitivity, temperature and the crystallinityof the materials used [11].
Bulk polymer encapsulation of the electronics withinimplantable devices has the advantage of being fullytransparent to RF communication. It has been used in remoteadjustable gastric band systems [12, 13]. However, comparedto metal, glass and ceramic packaging, polymer encapsulationprovides a diffusion barrier with limited hermetic properties[14, 15]. They have water vapor permeation rates that areseveral orders of magnitude higher compared to metal- orglass housings [8] and therefore cannot be used for long-termimplantable electronic devices.
The polymer poly-para-xylylene, also known as Parylene,has been widely used as a polymeric thin-film encapsulationmaterial for biomedical applications (e.g. stents and implantedneural prostheses [6]).
Parylene is deposited in a one-step process using adownstream low pressure chemical vapor deposition (LPCVD)process known as the Gorham process [16]. The depositionof Parylene is highly conformal and produces films that areconsidered to be pinhole-free (from a thickness of about 50 nm
on Si wafers). Due to its excellent step-coverage and lack ofstress, Parylene is an ideal candidate for the fabrication ofperfectly smooth surfaces [17].
Combining biocompatibility with excellent mechanicaland chemical properties, Parylene is a material of choice forpackaging applications. Furthermore, due to its high thermalstability (melting point at 290 ◦C), Parylene coated devices canbe sterilized (up to 135 ◦C) using conventional techniques priorto operation. However, Parylene has a low mechanical scratchresistance. Furthermore, looking at its molecular structure, thereduced density favors permeation and swelling effects whenexposed to liquids and a loss of tightness cannot be excludedby principle.
Compared to organic Parylene thin films, inorganic thinfilms made of ceramics have a higher molecular packagedensity and tightness but tend to create cracks and pinholes.The latter drawbacks are associated with internal stress due tothe deposition processes.
It has been shown that the hermeticity of flexible electro-optical devices like organic light emitting diodes (OLED)on polymer-based substrates can be significantly enhancedby inorganic moisture barrier coatings with thicknesses inthe range of 25–200 nm. The deposition of inorganic barrierlayers (e.g. SiNx or SiOx) by plasma-enhanced chemical vapordeposition (PECVD) on such substrates can decrease the entirepermeation by several orders of magnitude to an asymptoticminimum [18].
Chiang et al have characterized the degradation of a thinfilm of silicon nitride (SiNX) deposited onto a polycarbonate(PC) substrate after cyclic bending. These investigationsshowed that the barrier performance of the structure wassignificantly improved by depositing a smoothing Parylenethin film on the top of the SiNx layer. The degradation measuredby means of oxygen permeation after a cyclic bending testfor a Parylene/SiNx/PC structure was improved by morethan one order of magnitude as compared to the SiNx/PCstructure [19].
Based on these considerations this paper will focus onan increased hermetic thin-film packaging for biomedicalimplants by the development of multi-stacks composed ofParylene layers in combination with tight inorganic layers [20].The typical morphology is schematically sketched in figure 1.This concept favors the desired properties of each individuallayer, while reducing their disadvantages.
In order to reduce the macroscopic stress of the multilayerand to optimize its adhesion to the surface of the biomedicalimplant, the thicknesses of the organic and inorganic layershave to be matched: thick Parylene layers help to releasestress, but reduce the total molecular density of the stack fora given total thickness and number of stacks. Thick high-density, inorganic layers increase the total barrier hermeticity.However, they also increase the total stress, and reduce theconformity and adhesion behavior of the multilayer on thesurface of the biomedical implant, shown in figure 2.
The working principle of organic–inorganic multilayers,especially Parylene-SiOx multilayers, is based on a percolativepath for molecules penetrating the protecting barrier coating.The effective diffusion path-length in the Parylene layer
2
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
Figure 1. Illustration of a conformal multilayer stack deposited onan electronic circuit. The multilayer is composed of four Parylenelayers (red) combined with three high-density SiOx layers (yellow).Parylene layers are pinhole-free, highly conformal, but lacking inmolecular density. The high-density SiOx layers in the stack have alow permeability. However, they tend to create pinholes. Thecombination of both in a stack has the potential to overcome thedrawbacks of both layers.
Figure 2. Illustration of a multilayer structure composed ofinorganic layer d which tend to increase the total stress σ of thestack. The thick organic layers D can release this stress by arelaxation zone RZ. The circles of the figure correspond topolymeric unit cells.
between adjacent pinholes in the ceramic layer is therebystrongly increased. As soon as a permeate overcomes the firstParylene layer, its further permeation is strongly reduced toovercome the subsequent inorganic layer having an increaseddensity, expect an adjacent pinhole is present. The pinholemight be caused by stress or micro- and nanoscale structuraldistortions of the inorganic layer. If the pinhole distribution inthe inorganic layers of the multilayer is random, the risk ofpinhole alignment across the inorganic layers is statisticallyvery low. Thus, the repetition of many stacks increases theeffective diffusion path for molecules through the multilayerand drastically reduces the probability of molecular transport.Based on this percolative path principle, alternate organic–inorganic multilayers have the potential to come close to thevision of ideal barriers for next generation biomedical implantpackaging.
2. Materials and methods
2.1. Fabrication of multilayers
For hermetic packaging composed of high-density stackpartners in addition to conventional Parylene, a process mustbe chosen that allows for a maximum of conformity for thedeposition of these layers, as depicted in figure 1. Exclusivelyplasma-enhanced processes have the potential to deposituniform layers at reduced temperatures which are compatibleto Parylene layers.
For that purpose, a standard LPCVD reactor has beenmodified for PECVD deposition of layers different fromParylene as shown in figure 3. Further advantages of theprocess used are ambient temperature deposition, formationof a uniform coating, and a better controllability of the processparameters. To improve the reproducibility of depositedmultilayers and monitor process parameters, a LabVIEWsoftware was developed to control and monitor the gas flowand to display and store important process parameters (e.g.deposition time, reactor temperature and plasma intensity).
Parylene deposition is achieved by the conventionalGorham process. A detailed process description is givenin the literature [16]. During Parylene deposition, theprecursor vapor pressure inside the deposition chamber istypically 7 × 10−2 mbar and the substrate temperatureis kept at room temperature. For multilayer stacks theprecursor dichloro[2.2]paracyclophane dimer, also known asParylene-C, was used (Galxyl C, Galentis Srl). The precursorwas vaporized at a temperature of around 130 ◦C. The latterwas varied during deposition to keep a constant chamberpressure. A Parylene layer of 1 μm thickness was deposited in12 min.
Before the deposition of a first Parylene layer formultilayer stacks, a pre-cleaning of the substrate wasperformed by applying an argon plasma at 50 W for 1 min(5 sccm Ar flow rate). These process parameters were selectedto obtain conformal and pinhole-free Parylene layers [21, 22].
To enhance the overall tightness of the barrier coating,the inorganic material SiOx was chosen as an interlayer dueto its high molecular density at low deposition temperatures[23, 24]. SiOx thin-film layers were deposited by PECVD usingorganosilicon monomer precursors like hexamethyldisiloxane(HMDSO) [25]. For the PECVD of inorganic thin films, anadditional plasma generator was used. A capacitively coupledhigh-frequency plasma at 13.56 MHz is utilized to exciteor dissociate the entering process gas created by organicand inorganic precursors. The process gases are injected intothe deposition chamber through automated valves which arealternatively opened or closed. For the in situ plasma process,a controlled plasma is formed adjacent to the substrates byRF energy applied to the sample holder via the connected RFgenerator The plasma dissociate the entering process gasesinto ions and radicals that will be deposited onto the substratesurface.
The oxygen content of the SiOx layer can be adjustedby variation of the HMDSO/O2 ratio. The HMSDO/O2 ratioallows a smooth adjustment of the barrier layer property frompurely inorganic, ceramic properties with a maximum density
3
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
Figure 3. Schematic of the multilayer deposition system: the Parylene monomer flow through the reactor chamber or the bypass (blue) andthe HMDSO monomer flow for SiOx deposition (red) are shown.
at x = 2 (pure SiO2 composition) to layers having morepolymer-like properties due to the presence of organic groupsand therefore reduced density for x <2 [26, 27]. The advantageof a purely ceramic property of SiO2 is its hermeticity due tothe high density. However, ceramic layers increase internalstress in the multilayer which might create cracks withinthe deposited layer and thus reducing hermeticity. The goalof the PECVD process optimization for SiOx deposition isto find the right balance between density/hermeticity andstiffness/internal stress of the SiOx multilayer.
In our experiments, the SiOx barrier layers were depositedat a residual gas pressure of 3 × 10−2 mbar and a plasma RFpower of 50 W. Using a HMDSO/O2 ratio of 4:3 (8 sccmO2 and 6 sccm HDMSO flow rates), the plasma depositionwas started as soon as the total chamber pressure has beenstabilized at 14 × 10−2 mbar. A layer of 240 nm thicknesswas deposited in 5 min.
As Parylene tends to adhere poorly to materials withsmooth or non-porous surfaces, an argon plasma at 50 W for35 s (5 sccm flow rate) was used to activate the surface of SiOx
prior to the deposition of a subsequent Parylene layer.
2.2. Gas diffusion setup
Usually, the hermeticity of thin-film protection layers ischaracterized by the permeation rates of oxygen and watervapor across the barrier layer according to the ASTMF1249 standard. However, most commercial systems are time-consuming and not sensitive enough to measure the very lowpermeation rates of high performance encapsulating thin filmsand multilayers [28].
Therefore, a novel test method has been developed,based on the characterization of multilayers deposited on self-supporting polyvinylchloride (PVC) membranes. The coated
membranes are introduced in a vacuum system with adjustablegas pressure on one side of the membrane. A well-definedpartial pressure of gas (e.g. He, Ne, Ar, O2) is maintained inthe test volume using an injection valve and a bypass pump asdepicted in figure 4.
To guarantee homogeneous substrate permeation ofhelium, a pre-stamped PVC membrane of 75 μm waschosen (Nitto 224, Permapack). The PVC material wasselected due to the high gas permeation coefficient and thereproducible diffusion flux. The permeability of the membranewas quantified by use of a mass spectrometer on the other sideof the membrane. For He permeation, a background noise ofthe test setup of 1 × 10−11 mbar was measured.
The pressure control of the permeation test system permitsnot only static but also dynamic pressure loads. By measuringthe hysteresis of the permeation rate as a function of partialinert gas pressure applied to the membrane (pressure load),potential defects induced in the membrane can be identified.
Figure 5 shows a measurement where the He load pressurehas been varied between 0, 50, 100, 200 and 400 mbar ona PVC membrane coated with five 1 μm thick Parylene-Clayers and four 240 nm thick SiOx interlayers. No hysteresisof the He-permeation of the membrane could be observeduntil a pressure load of 100 mbar, which indicates that upto this pressure the mechanical integrity of the membraneis not affected. Therefore, a pressure of 25 mbar has beenapplied to all multilayers and thin films for He-gas permeationmeasurements to avoid inelastic membrane deformation.
3. Experimental results
3.1. Atomic force microscopy
In order to investigate the surface roughness of the Parylene-and multilayer thin-film structure and to visualize potential
4
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
Figure 4. Schematic of the setup to measure the gas permeation through self-supporting membranes. The multilayer, which is depositedonto a highly permeable test membrane (red), is introduced into the test chamber using an O-ring support.
Figure 5. He permeation cycling of a multilayer composed of a PVC membrane coated with five 1 μm thick Parylene-C layers and four240 nm thick SiOx interlayers. The He pressure that has been applied to one side of the multilayer is indicated inside the graph. The He partialpressure measured on the other side of the multilayer with a mass spectrometer shows no hysteretic behavior for pressures below 100 mbar.
holes and cracks, atomic force microscopy (AFM) wasperformed.
For this purpose, a single 1 μm thick Parylene-C layer anda multilayer composed of four 1 μm thick Parylene-C layersand three 250 nm thick SiOx interlayers were deposited on apolished silicon wafer.
The AFM analysis was performed by a 3100 dimensionmicroscope (Digital Instruments, Santa Barbara, CA) usingsilicon tips with a spring constant of 5 N m−1 and a tip radius<10 nm (Tap 150-AL-G from Budget Sensors, USA). Thetapping mode at a resonance frequency of 150 kHz was used.The AFM measurements were performed over a scan area of10 × 10 μm. Before measurements, the samples were purgedwith nitrogen.
For the single Parylene-C layer, a 6 nm RRMS (roughness,root mean square) was measured. The multilayer showedenhanced feature sizes due to the accumulated roughnessof each deposition step, which increased the RRMS to32 nm. The plasma cleaning/surface activation and stressformation by SiOx interlayer deposition were supposed toplay an important role for the increased roughness. Dueto multiple measurements, the observed area is assumed tobe representative for the entire barrier coating. No defects(>0.25 μm) could be visualized within the observed surface.
3.2. Scanning electron microscopy
Material homogeneity, interfaces and microstructural filmthickness characterization were investigated by a dual beamstation (Zeiss NVision 40 CrossBeam) with a focused ionbeam (FIB) and a high-resolution field emission scanningelectron microscope (SEM). A gallium ion beam was usedto mill a rectangular hole into the deposited multilayer. Thesidewalls of the milled area were analyzed using scanningelectron microscopy to obtain a cross-section of the layer.
Figure 6 shows the multilayer cross-section using the dual-beam technology. The layer thicknesses of around 1 μm forParylene and 240 nm for SiOx could be verified. The grayscale comparison in the FIB/SEM cross-section was used toreveal the SiOx layer embedded in Parylene. The thickness ofthe interfaces is found to be in good correlation with the XPSdeep etching measurements (section 3.4). Moreover, figure 6exhibits the high uniformity of the deposited SiOx layer.
By extraction of the upper interface curvature usingimage analysis tools, a root mean square roughness (RRMS)of 29 nm was measured and illustrated in figure 7. ThisRRMS corresponds well to AFM measurements of anothersection of the same multilayer surface. Hence, it is assumedthat the RRMS is representative for the whole sample. Theroughness was calculated using a software application suppliedby the microscope manufacturer (Leica).
5
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
Figure 6. SEM image of the conformal multilayer structure composed of four 1 μm Parylene-C layers and three 240 nm SiOx interlayers.Only the upper multilayer structure is shown.
Figure 7. Image analysis by extracting upper SiOx/Parylene interface using Leica microscope software. The analysis shows RRMS of 29 nmcorresponding to ISO 4287.
3.3. Fourier transform infrared spectrometry
In order to analyze the qualitative chemical composition ofdeposited single and multilayer structures, Fourier transforminfrared spectroscopy (FTIR) was used. This vibrationalspectroscopy can be applied to reveal the presence of specificfunctional groups. Chemical characterization of layers onpotassium bromide (KBr) disks was performed by FTIR witha Scimitar FTS 2000 spectrometer equipped with a DTGSdetector in transmission mode. The spectra were acquired from4000 cm−1 to 400 cm−1 with a resolution of 4 cm−1 and arepetition of 100 scans.
Figure 8 shows FTIR spectra of SiOx layers of 50 nm,200 nm and 250 nm, a self-supporting Parylene-C membraneof 1 μm and a SiOx (250 nm) coated Parylene layer (1 μm).For better visibility, the spectra are shifted in absorbance.
Despite lower absorbance peak intensities due to a thinnerlayer, the spectrum for the 50 nm thick layer exhibits thesame characteristics as the spectra of the 250 nm thick layer.The three typical peaks for SiO2 formation are visible at1050 cm−1, 800 cm−1 and a third around 440 cm−1 which areattributed to a high oxygen concentration of the deposited layer[26]. These peaks are associated with Si-O-Si asymmetricalstretching, rocking and bending vibration modes, respectively.Nevertheless, incorporation of silanol groups could be detectedby the broad band around 3200–3600 cm−1 assigned to Si-OHstretching. For the 250 nm layer, the peak at 880 cm−1 indicatesan additional silanol inclusion by Si-OH bending vibrations[29]. With respect to an incorporation of carbon components,minor peaks around 2930–2850 cm−1 and 1270 cm−1, formed
by CHx and Si-CH3 stretching band vibrations, were detected.These groups are integrated into the layer due to partialHMDSO dissociation, along with reactions of formed Sidangling bonds with ionized molecules (e.g. OH, CHx,hydrogen radicals or other oligomers). The appearance ofthe Si-OH, CH and Si-CH3 peaks are related to the lowoxygen/HMDSO partial pressure ratio [26, 30] in order toreduce the stress formations of deposited SiOx layers. Thepeaks around 2330 cm−1 are attributed to CO2 air absorptionand are not relevant.
The violet curve in figure 8 shows a typical FTIR spectrumof Parylene-C film based on the standard Gorham process[31]. The spectrum was measured with a self-supportingParylene membrane. For this reason, typical artifacts due toFabry–Perot interferences were measured. IR bands at 3019—2860 cm−1 are assigned to aromatic and aliphatic carbon–hydrogen band stretching vibrations. Peaks at 1608, 1558,1493 and 1402 cm−1 are formed by top-distributed benzenering stretching modes (C-C). The peak at 688 cm−1 shows thetypical C-Cl bond of Parylene-C.
The FTIR spectrum from a Parylene-C/SiOx/Parylene-Cmultilayer (2.25 μm) is depicted by the black curve in figure 8.The peaks show that the characteristics of the individual 1 μmsingle Parylene-C and the 250 nm SiOx layers are ideallysuperimposed for the multilayer coating. This result indicatesthat the different deposition processes in a single chamber areunaffected by each other for the bulk material and attests theintegrity of each individual layer. The FTIR analysis confirmsan oxygen content of about 70% in the multilayer SiOx thinfilms [26].
6
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
Figure 8. FTIR transmission specta of SiOx (50, 200 and 250 nm), Parylene-C thin film of 1 μm as well as a multilayer composed of twolayers of Parylene (1 μm) with an intermediate SiOx thin film (250 nm). All thin films were deposited on a KBr substrate except for the 1 μmParylene-C thin film which was a self-supporting membrane. The base of the vertical axis was shifted arbitrarily for better comparison.
3.4. X-ray photoelectron spectroscopy
X-ray photoelectron spectroscopy (XPS) technique wasused for quantitative chemical composition analysis ofthe multilayer structure and its interfaces. The spectrawere acquired with a Kratos Analytical Axis HSispectrometer, equipped with a quasi-monochromatic Al-Kax-ray (1486.6 eV) source, operated at 15 kV and 10 mA.The x-ray penetration depth was less than 10 nm and theelectron acquisition angle was fixed at 0◦. The spectra arerecorded in the 0–1500 eV kinetic energy range with anenergy resolution dE of about 0.68 eV on insulators. Electricalcharging compensation of insulating areas was performed by aconstant flux of low-energy electrons onto the sample. For theXPS interface analysis, an etching was performed by an argonion beam sputtering. The ion beam etching was characterizedby a filament emission current of 10 mA at a pressure of5 × 10−8 mbar and calibrated by a constant discharge currentof 250 nA onto a metallic reference sample. An averagesputtering rate of 1 nm per etch step was measured by confocalchromatic microscopy in phase-shifting interferometry mode.
For the experiment, a Parylene-C layer of 1 μm wasdeposited on a polished silicon wafer and coated with a SiOx
layer of 30 and 70 nm thickness.Atomic concentrations of the elements are obtained by
the integrated peak areas of the XPS spectra after a Shirleybackground subtraction. The contents of silicon (2p), oxygen(1 s) and carbon (1 s) of the 70 nm SiOx coating were analyzedrelated to the etching depth (figure 9(a)). The measurementshows a constant concentration of silicon, oxygen, and carbonof the SiOx layer. The discrepancy of the initial values can beattributed to atmospheric contamination of the surface. Thelarge presence of oxygen and silicon elements in the layerapproaching 58% and 38%, respectively, compared to thelow content of carbon (4%), is in agreement with the FTIRmeasurements and literature [30, 32]. Investigation regarding
chemical bonds of deposited 70 nm SiOx layers could beachieved due to binding energy shifts of core levels. The C-Cbinding at 285 eV was taken as reference for the analysis. Thetetragonal binding configuration of the silicon was analyzedand depicted in figure 9(b). The deconvolution by peak fitting(70:30 ratio of the Gauss-Lorentz function) into chemicalstates of the Si 2p peak was in agreement with the literature[32, 33]. The presence of quartz Si(-O-)4, assigned to the peakposition at 104.3 eV, is the most abundant fraction at 69%.The other chemical states of Si(-O-)3(-C-) and Si(-O-)2(-C-)2,with peaks centered at 102.7 and 101.7 eV, respectively, havea concentration of 27% and 4% [32].
A similar analysis was performed for a 30 nm SiOx coatingon top of a 1 μm Parylene film (figure 10(a)). After the firstetching steps until 8 nm, a constant concentration of oxygen(59%), silicon (38%), and carbon (2%) was associated withthe homogeneous SiOx bulk material. The concentrations areclose to the ones obtained for the 70 nm SiOx layer, showingthat the material composition is reproducible and does notdepend on the layer thickness. This was also confirmed bythe measurement of a non-depicted 300 nm SiOx coating.The start of the interface region from the homogeneous SiOx
material to the Parylene-C material was observed after 8 nmdue to the decrease of the oxygen and silicon concentrationsand the increase of carbon and chlorine concentrations. Thefact that the interface for the 30 nm SiOx sample appearsalready after 8 nm can be attributed to deposition parameterswhich influence the layer roughness and subsequently, thedepth information of the XPS measurement. The XPS analysisrevealed a Parylene/SiOx interface with a total thickness ofabout 20 nm. This is in agreement with the SEM cross-sectioninvestigations (figure 6). Pure Parylene-C material was formedafter the 20 nm interface where the chlorine and carbon ratioremains constant at 1:8 [34]. The residual silicon and oxygenelements at 35 nm can be explained due to preferential etchingof Parylene and material redeposition induced by the XPS
7
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
(a) (b)
Figure 9. The XPS deep etching analysis shows chemical concentrations of carbon, oxygen and silicon in a 70 nm SiOx coating on top of aParylene-C layer (a) and silicon chemical bonds analysis due to binding energy shifts of core levels (b).
(a) (b)
Figure 10. The XPS deep etching analysis shows chemical concentrations of carbon, oxygen, silicon and chlorine of a 30 nm SiOx coatingon top of a Parylene-C layer (a) and the chlorine analysis, showing the spin-orbital splitting (b).
deep etching process. Moreover, no structural defects in theSiOx layers could be observed due to the lack of chlorineatoms at the first 8 nm. The chlorine configuration of depositedParylene-C was analyzed after an etching depth of 30 nm. Thedeconvolution of the Cl 2p peak into chemical states is depictedin figure 10(b) and shows the spin-orbital splitting into theCl p3/2 peak at 200.4 eV and the Cl p1/2 peak at 202.0 eVwhich is in good agreement with other XPS investigations ofParylene-C [34, 35].
3.5. He-gas permeation
In all permeation measurements of single Parylene layers andalternating Parlyene-C/SiOx multilayers, a He-gas pressure of25 mbar was used where it was shown that no defects havebeen induced by the load gas pressure applied to the layers.
In order to convert the partial helium pressure measuredby the mass spectrometer into a He leak rate, a calibratedhelium leak with a constant flow rate of 5 × 10−8 mbar∗l
swas connected to the mass spectrometer and a corresponding
partial He pressure of 1.89 × 10−10 mbar has been measured.The constant He-background noise measured by the massspectrometer was 1 × 10−11 mbar.
The helium permeation of Parylene single layers andSiOx/Parylene-C multilayers, measured by this gas diffusiontest, are illustrated in figure 11. Each multilayer is composedof alternating Parylene layers of about 1 μm and inorganicSiOx layers of about 240 nm thickness (see also figure 6). Ncorresponds to the number of inorganic layers of the stack.Multilayer composed of N = 1, 2 and 4 SiOx interlayers areshown by the red curve. The blue and black curves illustratethe permeated He gas through single Parylene-C and SiOx
membranes respectively, measured by the mass spectrometer.Multilayers show a significantly enhanced He-gas
tightness compared to a single Parylene-C layer with the samethickness. The helium gas permeability of a multilayer stackcomposed of four inorganic layers (N = 4) with an overallthickness of 6 μm is equivalent to a single Parylene layer witha thickness of 24 μm.
8
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
Figure 11. He-permeation results of SiOx and Parylene thin films compared to multilayers composed of up to 5 Parylene (1 μm) and 4 SiOx
layers (240 nm). The multilayer of 6 μm thickness shows the same hermeticity for He-gas as a 24 μm Parylene thin film. A multilayer of10 μm thickness would be able to fulfil the MIL-STD-883 standard acceptance criterion of the equivalent air leak rate L = 1 ×10−6 atm cm2 s−1.
3.6. Water vapor permeation
For encapsulation of biomedical implants and thereforecorrosion protection, stringent requirements for the barriermaterial are demanded. A known indicator of the barriertightness is the measurement of the water vapor transmissionrate (WVTR) and the oxygen transmission rate (OTR)due to their corrosive behavior. The most sensitive testfor WVTR, the so-called MOCON test, is based on apulse-modulated infrared detection combined with absolutecoulometry (modified ASTM F1249) and is limited to aWVTR of 5 × 10−4 g m−2 day−1 [36]. Furthermore, usingthe MOCON test, measurements close to the detection limitare time consuming and take more than 1 week. To quantifylower WVTR for tight barrier coatings (e.g. for devices suchas flexible organic light emitting diodes or thin-film solarcells [37, 38]) the calcium mirror test is preferred [39, 40].A comparison between the two test methods is given in theliterature [18].
Using the Ca test, effective transmission rates down to10−6 g m−2 day−1 can be detected [28]. Thus, the calciumis degraded to transparent calcium hydroxide by oxygenand water permeation, allowing for the visualization of localdefects in the barrier coating. However, it was proven that waterpermeation is mainly responsible for calcium degradation(more than 95%) at 38 ◦C and 90% RH [41, 42]. Furthermore,it was shown that calcium totally hydrolyzes at ambientatmosphere and room temperature to form calcium hydroxideCa(OH)2 [42]. The transparent calcium products do not appearto act as a significant passivation and retarding layer [43].
To quantify the WVTR of tight multilayer barrierstructures that are presented in this paper, the calcium testmethod was applied. The set of chemical reactions involvedis represented in equations (1)–(3). The oxygen reaction (4)does not influence the measurement of the WVTR within theexperimental test conditions of 23 ◦C and 45% RH.
Water reaction
Ca + 2H20 → Ca(OH)2 + H2 (1)
composed of
Ca + H20 → CaO + H2 (2)
CaO + H20 → Ca(OH)2 (3)
Oxygen reaction
2Ca + O2 → 2CaO (4)
The multilayers and the Parylene thin films were depositedonto a 200 nm thick calcium layer (Ca, 99%, granular, Sigma-Aldrich) which was deposited onto transparent glass substratesusing a thermal evaporation system (Univex 300, Leybold).The thickness was verified using an in situ oscillating quartzcrystal sensor (XTC, Inficon). After deposition, the testsamples were exposed to controlled ambient atmosphere atroom temperature (23 ◦C and 45% RH).
4. Discussion
Using the calcium mirror test, the WVTR through a multilayerbarrier coating composed of four 1 μm thick Parylene-C layersand three 240 nm thick SiOx interlayers was compared toa standard Parylene-C coating of comparable thickness of4.7 μm. The calcium degradation was investigated by theformation of transparent calcium hydroxide due to atmosphericmoisture penetration through the barrier layer shown infigure 12(a). The total optical transmission of three calciumtest samples was analyzed by granulometry after 336 h ofexposure to the ambient atmosphere at room temperature(23 ◦C and 45% RH). The observed calcium oxide/hydroxidetransformation of 10.92% shown in figure 12(b) allows forthe calculation of the amount of converted calcium by H2Opermeation. The stoichiometric relationship of equation (1)
9
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
(a) (b)
Figure 12. (a) Calcium mirror test of a multilayer structure composed of four 1 μm thick Parylene-C layers and three 270 nm thick SiOx
interlayers after 336 h of exposure to ambient atmosphere conditions. (b) Image analysis of corroded calcium using two color images. Thedegraded, non-metallic calcium appears black against pure metallic calcium in white.
leads to a calcium/water equivalent of 2 which is used tocalculate the WVTR using the following formula:
WVTR =mCa · n · M[H2O]
M[Ca]
A · t(5)
where mCa is the amount of converted calcium, n = 2 is thestoichiometric coefficient of the reaction (1), M[H2O] andM[Ca] are the molar masses of water (18 g mol−1) and calcium(40.1 g mol−1), A is the total exposed surface to water vaporand t is the exposure time.
Using equation (5), a WVTR (23 ◦C, 45% RH) of2 × 10−3 g m−2 day−1 for a 4.7 μm thick multilayerbarrier and a total exposed test surface of 4 cm2 could bedetermined. The high barrier performance is consistent withthe literature where for example a WVTR (25 ◦C, 98% RH)of 1 × 10−2 g m−2 day−1 was achieved by a three times SiOx-like/SiOxCyHz barrier [30].
In case of the multilayer, a corrosion pattern of highlylocalized circles rather than lines or scratches could beobserved on the Ca surface. This can be explained by thepermeation of water through micro- and nanoscopic defects(pinholes and cracks) as opposed to a homogeneous, isotropicbulk permeation. These types of defects are well known, andare created by the nature of SiOx plasma deposition processes[30].
The WVTR of a standard Parylene-C film, normalizedwith respect to film thickness, is 83 g ·μm m−2 day−1 [44].Considering that the WVTR permeability is proportional to thelayer thickness and that the deposited thin films are homoge-neous, a Parylene-C thin film of about 1.9 mm thickness wouldprovide an equivalent 1 × 10−2 g ·μm m−2 day−1 WVTR, asthe developed multilayer barrier.
The US Military Standard MIL-STD-883 describes anon-destructive fine leak test that is routinely used in themanufacturing of implantable biomedical devices to verifythe hermeticity of laser welds at the encapsulation ofmicroelectronic and semiconductor components [45].
According to this standard, the device which has to betested is placed in a gas-tight chamber. In the case of the‘fixed’ test method, this chamber is pressurized with a tracergas under well-specified conditions (e.g. He gas at a pressurepE of 5 bar during t1 of 2 h) so that the gas can penetrate intothe device due to leaks of the seal.
After the so-called bombing pressure exposure time t1, thepressure is relieved and the device is transferred to a vacuumchamber where the tracer gas is extracted during a defineddwell time t2, and the leak rate R1 is measured by means of
a leak detector. Under these conditions, the MIL-STD-883specifies reject limits for the obtained leak rate R1.
Using the standard for the ‘flexible’ test method, the valuesfor t1 and t2 can be chosen so that the measured leak rate R1 isgreater than the detection limit of the leak detector used.
The standard specifies an equivalent air leak rate L as anacceptance criterion that depends on the device package cavityvolume V.
Based on the measured leak rate R1, the equivalentstandard air leak rate L can be calculated using the followingformula:
R1 = L × PE
PA×
√MA
M×
⎛⎜⎝1−e
−(
L×t1V×PA
×√
MAM
)⎞⎟⎠
× e−
(L×t2
V×PA×
√MAM
)(6)
R1 = The measured leak rate in atm cc s−1 HeL = The equivalent standard leak rate in atm cc/s airPE = The pressure of exposure in atmospheres absolutePA = The atmospheric pressure in atmospheres absoluteMA = The molecular weight of air in gramsM = The molecular weight of tracer gas (He) in gramst1 = The time of exposure to PE in secondst2 = The dwell time in secondsV = The internal device package volume in cubic
centimetersIn the gas diffusion setup described above, the volume V
as used in MIL-STD-883 corresponds to the spherical volumeV = 0.45 cm2 that is surrounded by the multilayer membranewith a surface area A = 2.83 cm2.
In this setup, the multilayer is exposed to a constant Hepressure of Pv = 25 mbar on one side of the multilayer and thevacuum of the mass spectrometer on the opposite side of themultilayer.
These conditions correspond to the ‘flexible’ test methodof MIL-STD-883 with an infinite bombing pressure exposuretime t1 = ∞ after which the exposure pressure is identical tothe pressure inside the internal volume V of the device packagecavity (pv = pE ).
The test condition of a constant He pressure pv in the gasdiffusion setup corresponds to a measurement of the leak rateafter a dwell time t2 = 0 s.
For an internal cavity volume V > 0.4 cm3, the MIL-STD-883 specifies a maximal equivalent air leak rate L =1 × 10−6 atm cm2 s−1.
10
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
As shown in figure 11, a multilayer on the 75 μm PVCsubstrate—composed of eight 1 μm thick Parylene layers andseven 240 nm thick SiOx layers, thus having a total thicknessof about 10 μm—passes the acceptance criteria of the MIL-STD-883.
5. Conclusions
The challenge of developing and analyzing the potential ofthin-film-based packaging for smart implants is addressed inthis paper. Thin-film packaging is able to coat all features of adevice with a minimum increase in overall volume.
The thin-film approach is based on the application ofstacked multilayers whereby one layer consists of a polymerand the other of an inorganic compound. The stack is depositedby a novel developed in situ deposition process, wherebyperiodic switching from CVD to PECVD in one chambercould be carried out without removing the sample betweenthe processing steps.
As a first thin-film material the polymer Parylene-Cwas chosen because of its well-known conformal growthproperties. However, as for all polymers, tightness at molecularlevel cannot be assumed a priori. As a second layer, SiOx
was chosen due to its particularly high tightness at molecularlevel. SiOx, as for most of the inorganic layers, is much denser.However, high internal stress leads to cracks and pinholes. Theoverall concept of our approach favors the desired propertiesof each individual layer, while reducing their disadvantages.
It was shown that the IR fingerprint of the Parylene andSiOx layers in the stack is the superimposition of the spectrafrom each individual layer. Finer-scale studies using XPSshowed that there is nevertheless a 20 nm gradual transitionzone between the layers that could be observed until the finalcomposition becomes uniform.
Continuous stacking of layers enhances the roughness ofthe overall multilayer. Based on the layer uniformity, it can beconcluded that the observed roughness will not lead to failureof the package due to local deficiency of thickness.
To characterize the final tightness of the packaging,multilayers on He-transparent substrates were tested accordingto the MIL-STD-883 standard: here a 6 μm thick multilayershowed the same tightness as a single 25 μm thick Parylenefilm. For comparison, a 10 μm thick multilayer would alreadybe able to fulfill the acceptance criterion for long-term medicalimplants.
As is well known, the liquid transport mechanismsin polymers containing ions are physically very differentfrom the transport mechanisms of inert gases, such ashelium. Effects due to swelling, capillarity and electro-osmotic pressure also have to be tested (to date, no standardexists for medical devices). In our experiments, the calciummirror test gave an insight into this aspect: for a multilayercomposed of four Parylene-C layers (1 μm) and three SiOx
(240 nm) layers, a very low water vapor permeation rate of2 × 10−3 g m−2 day−1 at 23 ◦C and 45% RH was observed.This low permeation rate is comparable to a pure Parylene-Clayer of 1.9 mm thickness. The calcium test method shows
the effectiveness of multilayers as a water vapor barrier whichexceeds favorably its helium gas tightness.
In summary, a single-chamber process for stackedmultilayers was developed which allows for the depositionof thin-film barrier coatings for medical implants that havethe potential for long-term protection of the devices againstmoisture permeation.
Acknowledgments
This research was supported by the Swiss innovationpromotion agency (grant 10836.1 PFLS-LS). The authorswould like to thank Sebastien Brun for the XPS analysis,Joanna Bitterli for SEM and AFM measurements, and TonyJournot for his support in the calcium mirror test analysis.
References
[1] International Organisation for Standardization, ISO 10993Series of standards Biological Evaluation of MedicalDevices
[2] Graichen F, Arnold R, Rohlmann A and Bergmann G 2007IEEE Trans. Biomed. Eng. 54 253–61
[3] Buchler P, Simon A, Burger J, Ginggen A, Crivelli R,Tardy Y, Luechinger R and Olsen S 2007 IEEE Trans.Biomed. Eng. 54 726–33
[4] Forde M and Ridgely P 2006 Implantable Cardiac PacemakersMedical Devices and Systems (The Biomedical EngineeringHandbook) 3rd edn ed J D Bronzino (Boca Raton, FL: CRCPress, Taylor and Francis Group)
[5] Duffin E G 2006 Implantable defibrillators Medical Devicesand Systems (The Biomedical Engineering Handbook)3rd edn, ed J D Bronzino (Boca Raton, FL: CRC Press,Taylor and Francis Group)
[6] Hassler C, von Metzen R P, Ruther P and Stieglitz T 2010J. Biomed. Mater. Res. Part B: Appl. Biomater. B 93 266–74
[7] McDermott H 1989 IEEE Trans. Biomed. Eng. 36 789–97[8] Striny K M 1988 Assembly Techniques and Packaging of VLSI
Devices 2nd edn, ed S M Sze (New York: McGraw-Hill)VLSI Technology
[9] Ginggen A, Tardy Y, Crivelli R, Bork T and Renaud P 2008IEEE Trans. Biomed. Eng. 55 1374–81
[10] Bork T et al 2010 Biomed. Microdevices 12 607–18[11] Briston J H and Katan L L 1983 Plastics Films 2nd edn (New
York: Longman Scientific & Technical/The Plastics andRubber Institute)
[12] Weiner R A, Korenkov M, Matzig E, Weiner S, Karcz W Kand Junginger T 2007 Obes. Surg. 17 717–21
[13] Miyake H, Ohta T, Kajimoto Y and Matsukawa M 1997Neurosurgery 40 931–5
[14] Jiang G and Zhou D D 2010 Technology advances andchallenges in hermetic packaging for implantable medicaldevices Implantable Neural Prostheses 2, Techniques andEngineering Approaches 1st edn, ed D D Zhou andE Greenbaum (New York NY: Springer Science+BusinessMedia) pp 27–61
[15] Donaldson P E K 1976 IEEE Trans. Biomed. Eng.BME-23 281–5
[16] Gorham W F 1966 J. Polym. Sci. A-1 4 3027–39[17] Charmet J et al 2010 Thin Solid Films 518 5061–5[18] Wolf R, Wandel K and Boeffel C 2007 Plasma Process.
Polym. 4 185–9[19] Chiang C C, Wuu D S, Lin H B, Chen Y P, Chen T N,
Lin Y C, Wu C C, Chen W C, Jaw T H and Horng R H2006 Surf. Coat. Technol. 200 5843–8
11
J. Micromech. Microeng. 23 (2013) 075001 A Hogg et al
[20] Hogg A, Keppner H, Burger J and Aellen T 2011 PatentWO/2011/018709 1–12
[21] Loeb W E 1971 SPE J. 27 46–51[22] Zhang X et al 1993 Proc. SPIE Submicrometer Metallization:
Challenges, Opportunities, and Limitations 1805 30–41[23] da Silva Sobrinho A S, Czeremuszkin G, Latreche M,
Dennler G and Wertheimer M R 1999 Surf. Coat. Technol.116–119 1204–10
[24] Erlat A G, Spontak R J, Clarke R P, Robinson T C,Haaland P D, Tropsha Y, Harvey N G and Vogler E A 1999J. Phys. Chem. B 103 6047–55
[25] Choi J-K, Kim D H, Lee J and Yoo J-B 2000 Surf. Coat.Technol. 131 136–40
[26] Sahli S, Rebiai S, Raynaud P, Segui Y, Zenasni Aand Mouissat S 2002 Plasmas Polym. 7 327–40
[27] Erlat A G, Wang B C, Spontak R J, Tropsha Y, Mar K D,Montgomery D B and Vogler E A 2000 J. Mater. Res.15 704–17
[28] Nisato G, Kuilder M, Bouten P C P, Moro L, Philips Oand Rutherford N 2003 SID Symp. Digest of TechnicalPapers 34 550–3
[29] Dominguez C, Rodriguez J A, Riera M, Llobera A and Diaz B2003 J. Appl. Phys. 93 5125–30
[30] Coclite A M, Ozaydin-Ince G, Palumbo F, Milella Aand Gleason K K 2010 Plasma Process. Polym. 7 561–70
[31] Licari J J 2003 Coating materials for electronic applicationspolymers, processes, reliability, testing (Norwich, NY:Noyes/William Andrew) pp 163–4
[32] Gruniger A, Bieder A, Sonnenfeld A, von Rohr P R,Muller U and Hauert R 2006 Surf. Coat. Technol.200 4564–71
[33] Alexander M R, Short R D, Jones F R, Michaeli Wand Blomfield C J 1999 Appl. Surf. Sci. 137 179–83
[34] Santucci V, Maury F and Senocq F 2010 Thin Solid Films518 1675–81
[35] Senkevich J J, Yang G R and Lu T M 2003 Colloids Surf. A216 167–73
[36] Stevens M, Tuomela S and Mayer D 2005 Water vaporpermeation testing of ultra-barriers: limitations of currentmethods and advancements resulting in increasedsensitivity, society of vacuum coaters 48th AnnualTechnical Conf. Proc. p 189
[37] Hack M, Forrest S R, Thompson M, Jackson T, Praino Rand Huffman D 2003 United States Air Force ResearchLaboratory, DARPA, Report number 20031202080 pp 6–10
[38] Dennler G, Lungenschmied C, Neugebauer H, Sariciftci N S,Latreche M, Czeremuszkin G and Wertheimer M R 2006Thin Solid Films 511–512 349–53
[39] Nisato G, Bouten P C P, Slikkerveer P J, Bennett W D,Graff G L, Rutherford N and Wiese L Proc. Int. DisplayWorkshop/Asia Display, Nagoya, Japan, 2001 pp 1435–8
[40] Lewis J S and Weaver M S 2004 IEEE J. Sel. Top. Quantum10 45–57
[41] Paetzold R, Winnacker A, Henseler D, Cesari V and Heuser K2003 Rev. Sci. Instrum. 74 5147–50
[42] Cros S, Firon M, Lenfant S, Trouslard P and Beck L 2006Nucl. Instrum. Methods B 251 257–60
[43] Nisato G and Bouten P C P 2002 IST Public ReportIST-2001–34215 pp 1–19
[44] Massey L K 2003 Permeability Properties of Plastics andElastomers: A Guide to Packaging and Barrier Materials2nd edn (Norwich, NY, USA: Plastics DesignLibrary/William Andrew) p 132
[45] MIL-STD-883H Department of Defense, Test MethodStandard Microcircuits, Method No. 1014.13, Seal (26February 2010)
12
8 List of publications 119
Paper II
Protective multilayer packaging for long‐term implantable medical devices,
DOI: 10.1016/j.surfcoat.2014.02.070
Andreas Hogg, Stefanie Uhl, François Feuvrier, Yann Girardet, Benjamin Graf,
Thierry Aellen, Herbert Keppner, Yanik Tardy, and Jürgen Burger,
Surf. Coat. Tech., in press, corrected proof, (2014), pp. 1‐6.
Copyright Elsevier (2014)
Reproduced with permission of the publisher.
131
120 8 List of publications
Surface & Coatings Technology xxx (2014) xxx–xxx
SCT-19263; No of Pages 6
Contents lists available at ScienceDirect
Surface & Coatings Technology
j ourna l homepage: www.e lsev ie r .com/ locate /sur fcoat
Protective multilayer packaging for long-term implantablemedical devices
Andreas Hogg a,c,⁎, Stefanie Uhl c, François Feuvrier c, Yann Girardet c, Benjamin Graf c, Thierry Aellen c,Herbert Keppner c, Yanik Tardy b, Jürgen Burger b,d
a Graduate School for Cellular Biomedical Sciences, University of Bern, Switzerlandb Medos International Sàrl, Le Locle, Switzerlandc Haute Ecole Arc Ingénierie (HES-SO), La Chaux-de-Fonds, Switzerlandd ARTORG Center for Biomedical Engineering Research, University of Bern, Bern, Switzerland
⁎ Corresponding author at: Haute Ecole Arc IngénieriChaux-de-Fonds, Switzerland. Tel.: +41 32 930 25 35; fax
E-mail address: [email protected] (A. Hogg).
http://dx.doi.org/10.1016/j.surfcoat.2014.02.0700257-8972/© 2014 The Authors. Published by Elsevier B.V
Please cite this article as: A. Hogg, et al., Surf
a b s t r a c t
a r t i c l e i n f oAvailable online xxxx
Keywords:Water permeationMultilayer coatingsBiomedical implantsParyleneSilicon oxideConformity
State of the art packaging for implantable devices usesmetal or glass housings that are reliable but limited from aminiaturisation viewpoint as well as cost-intensive. We suggest a hermetic and biocompatible thin film packag-ing based on alternating organic/inorganic coatings for further miniaturisation of smart implantable MEMS de-vices that can be applied for long-term implantation. The combination of high intrinsic molecular densitysilicon oxide (SiOx) and pinhole-free and stress releasing poly-para-xylylene (parylene-C) thin films creates anew composite material, which is optimal for hermetic and biocompatible packaging. A novel single-chamberthin film deposition process was developed for the fabrication of SiOx/parylene thin film multilayer structures,using a modified chemical vapour deposition (CVD) process. According to permeation and conformity aspects,the inorganic layer is the crucial layer of the coating. Permeationmeasurements the highly ceramic SiOx materialrevealed a low helium gas permeation and a non-critical cracking thickness up to 300 nm. Themorphology of themultilayer structure was analysed by scanning electronmicroscopy; an algorithm for defining ideal layer confor-mity was established and no local thickness deficiencies of deposited SiOx layers could be observed. To evaluatethe corrosion protection, an adapted calciummirror test based on water droplet permeation was developed, andthe water permeation of conventional parylene-C layers (4.5 μm)was compared to multilayer stacks composedof 3 SiOx interlayers (4.7 μm).In this paper, it could be shown that by tailoring the thickness ratio between the involved layers, the percolativepathway and thereby, the permeation for directwater exposure could be considerably reduced compared to con-ventional parylene-C single layers with the same thickness.
© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/3.0/).
1. Introduction
Conventional biomedical long-term implants use metal and glasspackaging to protect the human body from potentially non-biocompatible materials and on the other side, to protect the devicefrom body fluids which can lead to its failure. For many biomedicalimplants, titanium is the material of choice due to its high biocom-patibility and its mechanical properties [1]. Cardiac pacemakers forexample, are usually packaged into welded titanium jackets. If im-plants have to communicate with an external reading unit, glass ma-terials like Pyrex or Borofloat are used because of their RFtransparency [2]. In addition, these materials have really low watervapour transmission rates (WVTRs). For metals and glasses, WVTRs
e, Eplatures-Grise 17, 2300 La-: +41 32 930 29 30.
. This is an open access article under
. Coat. Technol. (2014), http:/
of around 1 × 10−10 and 1 × 10−6 g mm m−2 day−1 respectively,were measured compared to polymeric materials which are in therange between 10 and 1 × 10−3 g mm m−2 day−1 [3]. Conventionalencapsulation methods using metal and glass jackets generate largeunused cavities between the enclosed MEMS device and the housingwalls as shown in Fig. 1a. This condition limits the miniaturisationpotential of the conventional encapsulation methods. In order to re-duce these unused cavities, a conformal multilayer barrier structurein the μm-range as shown in Fig. 1b was developed. This barrier layeris able to coat the entire implant by a hermetic thin film [4].
In order to perform a conformal overgrowth on complex three-dimensional structures, plasma enhanced and standard chemical va-pour deposition (PECVD, CVD resp.) processes were chosen to guar-antee an uniform coating. For the creation of close-to-ideal tight thinfilm barriers, polymeric and ceramic materials were combined toform conformal multilayer barrier structures. The well-known poly-mer parylene-C was chosen due to its nearly stress-free deposition
the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).
/dx.doi.org/10.1016/j.surfcoat.2014.02.070
Fig. 1. Illustration of a conventional package using surrounding glass ormetal jackets (a). Aconsiderable decrease in packaging volume can be achieved by the novel thin-film layerencapsulation method for biomedical devices (b).
Fig. 2. A stack of polymeric pin-hole-free layers which exhibits a high permeation at mo-lecular level combined with low-permeation but pin-hole affected layers leads to an in-creased effective percolative diffusion pathway.
2 A. Hogg et al. / Surface & Coatings Technology xxx (2014) xxx–xxx
property, pin-hole-free growth, excellent mechanical and chemicalproperties and high step-coverage [5]. However, the gap betweenthe polymeric chains of parylene-C still allows a certain diffusion ofatoms andmolecules through the layer. In order to reduce this disad-vantage, ceramic silicon oxide (SiOx) layers are incorporated into thebarrier film. Silicon oxide thin films are tight at the molecular level.However, they tend to increase internal stress formation of themultilayers which induces defect sites. If a silicon oxide layer isdeposited on top of an elastic parylene-C layer, this subjacent poly-meric material will act as a stress releasing layer. In addition, astack of polymer and ceramic layers has the potential of combiningthe advantages of both materials in that the percolative pathway ofpermeates is increased as shown in Fig. 2.
The tight ceramic layer will allow diffusion only through its pinholesand cracks. Hence, the percolative pathway is significantly elongatedcompared to a homogenous diffusion that permeates the layer in a di-rect way.
Please cite this article as: A. Hogg, et al., Surf. Coat. Technol. (2014), http:/
2. Experimental
2.1. Material fabrication
In order to fabricate a multilayer coating composed of parylene-Cand SiOx thin films, a combination of a low-pressure chemical vapourdeposition (LPCVD) reactor with a plasma enhanced chemical vapourdeposition (PECVD) process was developed. An automation processwas implemented, to guarantee the reproducibility as well as to controlthe key parameters of the different alternated depositions. Due to thehigh stability of process parameters by means of the automated deposi-tion process, important properties of the protective barrier stack such asa conformal and uniform coating at low deposition temperature(b50 °C) could be achieved.
The deposition of parylene C was made by the use of a modifiedparylene coater (Comelec SA). Based on a conventional Gorham pro-cess [6], the parylene-C precursor is vaporised at a temperature ofaround 130 °C, decomposed to a monomer in the pyrolysis chamberat 650 °C and polymerised on the sample, placed into the depositionchamber at a total pressure of around 7 × 10−2 mbar. These processparameters were selected to obtain highly conformal and pinhole-free parylene layers [7]. The precursor dichloro[2.2]paracyclophanedimer (Galxyl C, Galentis Srl) also known as parylene-C was used[8].
To reduce the permeation of parylene based barrier coatings, aninorganic material consisting of SiOx was chosen as the interlayerdue to its high molecular density at low deposition temperaturesand its reliable deposition technology [9,10]. The SiOx thin-filmlayers are obtained by the dissociation of a hexamethyldisiloxane(HMDSO) precursor and additional oxygen gas molecules using anin-situ capacitively-coupled high-frequency plasma at 13.56 MHz.The SiOx material, resulting from the binding of decomposed ionsand radicals onto the substrate surface, is strongly dependent onthe concentration of injected oxygen and HMDSO precursors intothe deposition chamber. Moreover, the material composition de-pends on the residual gas pressure and the substrate temperature.In particular, the oxygen content of the SiOx layer can be adjustedby variation of the O2/HMDSO ratio. The O2/HMDSO ratio allows asmooth adjustment of the barrier layer composition from purely in-organic, ceramic properties (SiO2) with a maximum density, tolayers having more polymer-like properties (SiOx). This can be ex-plained by the presence of organic groups mostly coming from theHMDSO precursor that are incorporated into the tetrahedral SiO2
structure creating a reduced material density (for x b 2). The advan-tage of a purely ceramic SiO2 composition is its hermeticity due to thehighmolecular density. However, these ceramic layers increase internalstress in the multilayer which might create cracks and pinholes withinthe deposited layer and thus, reduce hermeticity. The goal of thePECVD process optimization for SiOx deposition is to reach the right bal-ance between density/hermeticity and stiffness/internal stress of theSiOx multilayer [11]. In this study, the barrier permeability in functionof the layer thicknesses for a constant O2/HMDSO ratio of 10 was inves-tigated. This ratio showed in a prior investigation the optimal trade-offbetween stress formation and permeation to stack up the multilayerbarriers [4].
To combine the parylene-C and the SiOx layers in a multilayer, theprocess gases are injected into a single deposition chamber through au-tomated valves which are alternatively opened or closed. In this study,1 μm thick parylene-C layers were used, which correspond to a deposi-tion time of 12 min. For the SiOx barrier interlayers, the thicknesses ofthe deposited thin films were approximately 240 nm, obtained by a5 min deposition time, an O2/HMDSO flow rate that corresponds to atotal chamber pressure of 14 × 10−2 mbar (3 × 10−2 mbar residualgas pressure), and a RF plasma power of 50W. As a result, the chemicalstates present in the layer composition are dominated by the tetragonalbinding configuration (≈70%)which can be considered as close to ideal
/dx.doi.org/10.1016/j.surfcoat.2014.02.070
Fig. 3. Experimental set-up for the characterization of He-gas permeation containing a gassupply (He), vacuum pumps, pressure sensors, a mass-spectrometer with an electronmultiplier, and a sample holder.
3A. Hogg et al. / Surface & Coatings Technology xxx (2014) xxx–xxx
ceramic properties [4]. In order to increase the adhesion of parylene onthe substrate, an argon plasma was applied before the deposition(5 sccm Ar flow rate, 1 min, RF power 50 W). Moreover, as parylenetends to adhere poorly to materials with smooth or non-porous sur-faces, an argon plasma at 50 W for 35 s (5 sccm flow rate) was usedto activate the surface of the SiOx layers prior to the deposition of a sub-sequent parylene layer.
2.2. Characterisation
2.2.1. MorphologyFor the analysis of the qualitative layer thickness, the homogeneity,
and the micro-structural morphology, a dual beam station (ZeissNVision 40 CrossBeam) including a scanning electron microscope(SEM)with a combined focused ion beam (FIB) was used. A rectangularsection was ablated into the deposited multilayer by a Gallium FIB. Theside walls of the milled area were studied using the integrated SEM.
2.2.2. SiOx material compositionThe quantitative chemical composition analysis of SiOx layers was
performed by Fourier transform infrared spectroscopy (FTIR) with aScimilar FTS 2000 spectrometer equipped with a deuterated triglycinesulphate (DTGS) detector in transmission mode. Characterization oflayers was investigated on potassium bromide (KBr) discs. The spectrawere acquired from 4000 cm−1 to 400 cm−1 with a resolution of4 cm−1 and a repetition of 100 scans. Themeasurement of thematerialcompositionwas analysed by the detection of vibrationalmodes relatedto the presence of its specific functional groups.
2.2.3. Layer hermeticity measurement at helium gas exposureThe characterisation of the SiOx barrier hermeticity was investigated
by the use of a helium (He) leak test measurement [4]. The test methodis based on coated membranes composed of a deposited SiOx layer ontop of a self-supporting 75 μm polyvinylchloride (PVC) film carrier(Nitto 224, Permapack). The coated membranes were inserted into avacuum system and an adjustable helium gas pressure was applied onone side of the membrane. As soon as a He pressure gradient isestablished across the membrane, the partial pressure in the test cellis increasing and quantified by the use of amass spectrometer equippedwith an additional electron multiplier on the other side of the mem-brane as depicted in Fig. 3. A He background noise of the test setup of1 × 10−11 mbar was measured.
2.2.4. Layer hermeticity measurement at liquid exposureThe measurement of the liquid water permeation is one of the most
important indicators to characterise the barrier layer tightness com-pared to standard gas WVTR measurements [12]. By a direct liquid
Fig. 4. FIB cut-out (a) and cross-section SEM microg
Please cite this article as: A. Hogg, et al., Surf. Coat. Technol. (2014), http:/
contact of water, additional transport mechanisms such as capillarity,swelling and electro-osmotic pressure have to be taken into account.For this reason, an adapted calcium mirror test was applied in order toanalyse the liquid water transmission. The principle of this test methodis the high reactivity of calcium with water. If water will be in contactwith calcium, it reacts instantly and forms calciumhydroxide by the fol-lowing chemical equation:
Caþ 2H2O→Ca OHð Þ2 þ H2: ð1Þ
The water will degrade the metallic calcium into transparent calci-um hydroxide which can be analysed visually.
For this measurement, a 200 nm thick calcium layer was depositedonto a glass substrate carrier. Three calcium test sampleswere protectedby a multilayer (4.7 μm) composed of four 1 μm thick parylene-C layersand three 240nmthick SiOx interlayers. As reference, three calciummir-rors were coated with a pure 4.5 μm thick parylene-C layer.
Using these protected calciummirrors, 3 water droplets with a sodi-um chloride solution of 1 g/l, 9 g/l and 20 g/l and 1 droplet of purewaterwere placed on top of each sample as shown in Fig. 7a. The corrosionprogression of the test samples was recorded by a photo camera and apicture of the actual state was saved every 30 s. Due to evaporation ofthe liquid, the water droplets were renewed every two hours.
3. Results and discussion
3.1. Multilayer morphology
The focused ion beam ablation of a multilayer stack is shown inFig. 4a. The cross-section area was afterwards analysed by a scanningelectron microscope as depicted in Fig. 4b.
The investigation reveals conformal layer thicknesses of around1 μm for the parylene-C layers and 240 nm for the SiOx layers. The
raphs (b) of a parylene-C/SiOx multilayer stack.
/dx.doi.org/10.1016/j.surfcoat.2014.02.070
4 A. Hogg et al. / Surface & Coatings Technology xxx (2014) xxx–xxx
homogenous greyscale of the layer bulk is an indicator of an uniformmaterial composition.
3.2. SiOx layer composition
Homogeneity and chemical characteristics of SiOx material based onthe presented PECVD process with a O2/HMDSO flux ratio of 10 wereanalysed by FTIR spectroscopy.
Fig. 5 shows the three typical peaks at 1050 cm−1, 800 cm−1, and440 cm−1 which are assigned to Si\O\Si asymmetrical stretching,rocking and bending vibration modes, respectively. This characteristicis attributed to a high oxygen concentration during the deposition andcan be assigned to an oxygen content of about 70% in the multilayerSiOx thin film [13] close to pure SiO2 compounds. The non-pure ceramicmaterial is confirmed by the broad band at around 3200–3600 cm−1
assigned to Si\OH stretching attributed to silanol groups. The presenceof these chemical groups is interpreted as inherent partial HMDSO dis-sociation, alongwith reactions of formed Si dangling bondswith ionisedmolecules. The peaks at around 2330 cm−1 are attributed to CO2 air ab-sorption and are not relevant.
3.3. Helium hermeticity of SiOx layers
Optimization of the multilayer packaging strongly depends on theproperties of the much denser SiOx inorganic layer. Regarding thehermeticity, the SiOx is the material which has the higher density com-pared to polymers, but tends to create pinholes andmicroscopic defectsduring the PECVD due to the internal stress formation. The layer thick-ness can be increased until a critical cracking thickness (CCT) isachieved. From this thickness onward, the diffusion barrier is less effi-cient. Looking at the percolative pathway model in Fig. 2, the latter ef-fect leads to a reduction of the total percolation length. Moreover, theinternal stress increases the risk of failure due to mechanical impactsby handling or external liquid environments (e.g. immersion intowater). In terms of adhesion, the internal stress of the SiOx thin filmcan lead to the delamination of the barrier stack. Concerning the confor-mity, the SiOx inorganic layer is less conformal due to the directionalityof the PECVD process. Therefore, a deposition pressure has to be chosenin order to minimise the plasma sheath length that spaces out theionised atoms or molecules from the surface topology. Thus, for a con-formal layer formation, the right balance between the deposition pres-sure and an efficient ionic bombardment for higher film density has tobe found relative to the implantmorphology. On the other hand, the in-ternal stress of the inorganic layer is increased by the local geometricdevice complexity such as edges with negative slope or peak effects.
In order to have a possible adaptation of the inorganic layer depend-ing on applications, the helium permeation through membranes com-posed of a SiOx film on a PVC substrate as a function of its layerthickness was investigated as shown in Fig. 6.
400 900 1400 1900 2400 2900 3400 3900
Abs
orba
nce
[a.u
]
Wavelength [cm-1]
Si-O-Si
Fig. 5. Typical FTIR spectrum of SiOx layers obtained by PECVD using an O2/HMDSO flowrate of 10 with the three typical peaks at 1050 cm−1, 800 cm−1, and 440 cm−1. Thepeaks are associated to Si\O\Si asymmetrical stretching, rocking and bending vibrationmodes respectively.
Please cite this article as: A. Hogg, et al., Surf. Coat. Technol. (2014), http:/
TheHe-gas permeationwasmeasured using the set-up illustrated inFig. 3. For anO2/HMDSO ratio of 10, an on-going island growth at the be-ginning of the PECVD could be observed. After thefirst 10 nm, the nucle-ation phase ends in a continuous SiOx layer growth where the barrierhermeticity of the SiOx layer decreases continuously up to 300 nm.The decrease of helium permeation shows that the defect surface ofSiOx layer decreases as a function of its thickness. Thus, for a multilayerstack the total percolation pathway will be increased up to the CCT of300 nm.
3.4. Water permeation of multilayer barriers
Fig. 7b shows the degradation progress of parylene-C coated samplesafter 5 h,where the diameter of corroded calciumcorrelates to the dropletdiameter. After 8 h, an entire corrosion of the parylene-C coated test mir-rors could be observed, in contrast to the multilayer coated calcium mir-rors, whichwere still unaffected (Fig. 7c). In addition, no visible influenceof the sodium chloride concentration on water permeation could be ob-served. The water droplet test showed the lowwater permeation of mul-tilayer barriers compared to conventional parylene-C layers. Additionalwater vapour transmission tests revealed a WVTR for these 4.7 μm thickmultilayer stacks (3 SiOx interlayers) of 1 × 10−2 g μmm−2 day−1 com-pared to theWVTRof 83 g μmm−2 day−1 of conventional parylene-C sin-gle-layer barriers [4,14].
3.5. Multilayer conformity
In order to achieve an optimal hermeticity by multilayer packaging,the question of conformity of the tightness-crucial SiOx layer arises. It iswell known that parylene layers exhibit a close to perfect conformity. Asit was shown in Fig. 6, the SiOx permeation is related to its thickness.Therefore the total barrier hermeticity is strongly depending on theSiOx thickness. For this reason, an algorithm for defining ideal layer con-formity and detection of local thickness deficiencies in comparison tothe experimental results was developed.
In the schematic sketch shown in Fig. 8, the green curves representthe lower and upper boundaries of an arbitrary thin film cross-sectionthat is deposited over a curved (rough) substrate. At any point of thelower boundary of the thin film, the perpendicular is constructed andits endpoint at the thickness l is determined. The envelope (upperdashed red curve) represents the upper boundary of an ideal conformalpackaging layerwithout deficiency in thickness. For those points, wherethe calculated envelope is above the upper green line, the thin film hasdeficiencies in thickness; hence it is non-conformal.
Fig. 6. Helium permeation of SiOx films on PVC substrates carried out by the permeationtest set-up. For an O2/HMDSO flow ratio of 10, a continuous decrease of SiOx permeationrelated to its thickness could be observed.
/dx.doi.org/10.1016/j.surfcoat.2014.02.070
Fig. 7. Calciummirror test: a) three encapsulated samples by pure 4.5 μm thick parylene-C(left side) are compared to 4.7 μm thickmultilayer stack (3 SiOx interlayers)-packaged Camirrors (right side). b) After 5 h ofwater exposure, the parylene-C packaged layers exhibitcorrosion of the size of thewater droplet. c) After 8 h the parylene-C layers are completelycorroded, whereas the multilayer packaged layers were still unaffected.
5A. Hogg et al. / Surface & Coatings Technology xxx (2014) xxx–xxx
The coordinates u0 and g(uo) of the envelope of an ideal layer can becalculated from the coordinates x0 and f(xo) of the lower boundary of athin film as follows:
x→u and f xð Þ→g uð Þu0 ¼ x0 þ
l
�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þ tan2 αð Þ
q ¼ x0−l
�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þ f 0 x0ð Þ� �−2
q ð2Þ
Fig. 8. Schematic illustration of the algorithm application on a non-conformal layer (greencurves) for defining its ideal conformal layer (dashed red curve) for packaging purposes.The ideal conform curve with the distance l to the lower green curve can be achieved bya coordination transformation from f(x0) and x0 to g(u0) and u0 which allows for the cal-culation of the conformity deviation. (For interpretation of the references to colour in thisfigure legend, the reader is referred to the web version of this article.)
Please cite this article as: A. Hogg, et al., Surf. Coat. Technol. (2014), http:/
g u0ð Þ ¼ f x0ð Þ þ l � tan αð Þ�
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þ tan2 αð Þ
q ¼ f x0ð Þ− l
f 0 x0ð Þ � ð�ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þ f 0 x0ð Þð Þ−2
q
ð3Þ
where l is conformal thickness of the thin film, α is the angle betweenthe normal n(xo) and the x-axis, and f′ (xo) is the derivative of f(xo).
Themorphology-analysis of the SiOx layers (showed in Fig. 4b) usingthe described algorithm, is represented by Fig. 9. Therefore, the lowerand upper interface curves were extracted from Fig. 4b by the use of agrey scale analysis with the ImageJ™ software tool. The illustrationshows the averaged and smoothed upper and lower borderlines(blue) of the 240 nm thin SiOx film between two parylene layers of0.92 μm and 1.10 μm thicknesses. These curves are considered to be ob-tained experimentally. The overlaid red curve was constructed by theuse of the conformity algorithm (l = 240 nm) applied on the lowerblue curve. Thus, the normal distance between the red and the lowerblue curve is constantly 240 nm. The constructed red curve was com-pared to the measured upper blue curve and it may be concluded thatthe SiOx layer grows conformally within the acquisition noise comingfrom the SEM and the image treatment (grey dashed zone). Hence,the layer can be considered to be conformal and therefore, will notlead to failure of the package due to a local deficiency of thickness.
In summary, it can be concluded that for a pre-set level surfacecurvature of limited slopes of 12°, conformal coating is adequatelyachieved by using a PECVD composed of HMDSO and oxygen at a pres-sure of 14 × 10−2 mbar and compares well with an ideal constant thick-ness coating.. However, if vertical slope featuresmust be packaged, specialcare has to be taken into account as shown in [15]. It is well known thatfor higher pressures in PECVD processes, more collisions of the radicalsin the discharge gap reduce the directionality of the deposition. Thus,the conformity will be improved by an increase of the total pressure.
4. Conclusion
In this paper, investigations of multilayered stacks for biomedicalpackaging were presented. These layers were deposited in a single-chamber process, combining CVD and PECVD technologies. Due to thehigh stability of process parameters by means of the automated deposi-tion system, a conformal and uniform coating at ambient temperaturefor smart medical implants could be achieved.
Fig. 9. Plot of the smoothed lower and upper border lines of the SiOx thin film of Fig. 4b(blue curves) including acquisition noise of the SEM and image treatment (grey dashedzone). The red curve was obtained by the algorithm applied on the lower curve. The redcurve is within the acquisition noise of the upper border of the SiOx thin film, hence thelayer can be considered as conformal. (For interpretation of the references to colour inthis figure legend, the reader is referred to the web version of this article.)
/dx.doi.org/10.1016/j.surfcoat.2014.02.070
6 A. Hogg et al. / Surface & Coatings Technology xxx (2014) xxx–xxx
In this study, it could be further shown that by variation of the SiOx
layer thickness, looking at its helium permeation, the barrier stack canbe customised depending on the application.
The application of multilayers on medical devices (rough surfaces,vertical features, negative slopes) must be carefully evaluated forthe different cases. It was shown that by using a total pressure of14 × 10−2 mbar (O2 + HMDSO), a highly conformal layer depositionover at least 12° inclined features could be achieved.
The permeation study was extended from the standard water va-pour transmission test [4] to direct water exposure. The adapted calci-um mirror test revealed that using parylene-C layers containing threeSiOx interlayers leads to a considerably lower liquid water permeationas compared to pure parylene layers having the same thicknesses.
The insight on the effects of percolative pathway-increase that wasgained on a three SiOx/parylene-study can be extended to a higher de-gree of stacking with the potential to further reduce permeation.
In summary, a highly hermetic and conformal packaging for medicaldevices was developed by the use of a novel single-chamber thin-filmdeposition process.
Conflict of interest
Wewish to confirm that there are no known conflicts of interest as-sociatedwith this publication and there has beenno significantfinancialsupport for this work that could have influenced its outcome.
Acknowledgements
This research was supported by the Swiss Innovation PromotionAgency (grant 10836.1 PFLS-LS).
Please cite this article as: A. Hogg, et al., Surf. Coat. Technol. (2014), http:/
References
[1] D.F. Williams, in: R. Kossowsky, N. Kossovsky (Eds.), Materials Sciences and ImplantOrthopedic Surgery, 116, Springer, Netherlands, 1986, pp. 107–115.
[2] T. Bork, A. Hogg, M. Lempen, D. Muller, D. Joss, T. Bardyn, P. Buchler, H. Keppner,S. Braun, Y. Tardy, J. Burger, Biomed. Microdevices 12 (2010) 607–618.
[3] K.M. Striny, in: S.M. Sze (Ed.), VLSI Technology, 2nd ed., McGraw-Hill, New York,1988, pp. 566–611.
[4] A. Hogg, T. Aellen, S. Uhl, B. Graf, H. Keppner, Y. Tardy, J. Burger, J. Micromech.Microeng. 23 (2013) 075001.
[5] J. Charmet, O. Banakh, E. Laux, B. Graf, F. Dias, A. Dunand, H. Keppner, G. Gorodyska,M. Textor, W. Noell, N.F. de Rooij, A. Neels, M. Dadras, A. Dommann, H. Knapp, C.Borter, M. Benkhaira, Thin Solid Films 518 (2010) 5061–5065.
[6] W.F. Gorham, J. Polym. Sci. A1 4 (1966) 3027–3039.[7] J.B. Fortin, T.-M. Lu, Chemical Vapor Deposition Polymerization theGrowth and Prop-
erties of Parylene Thin Films, Kluwer Academic Publishers, Boston, Massachusetts,2004.
[8] M. Szwarc, Discuss. Faraday Soc. 2 (1947) 46–49.[9] A.G. Erlat, R.J. Spontak, R.P. Clarke, T.C. Robinson, P.D. Haaland, Y. Tropsha, N.G.
Harvey, E.A. Vogler, J. Phys. Chem. B 103 (1999) 6047–6055.[10] A.S.D. Sobrinho, G. Czeremuszkin, M. Latreche, G. Dennler, M.R. Wertheimer, Surf.
Coat. Technol. 116 (1999) 1204–1210.[11] A.M. Coclite, F. De Luca, K.K. Gleason, J. Vac. Sci. Technol. A 30 (2012)
061502–061509 (061502).[12] G. Nisato, P.C.P. Bouten, P.J. Slikkerveer, W.D. Bennett, G.L. Graff, N. Rutherford, L.
Wiese, Proc. Int. DisplayWorkshop/Asia Display,Nagoya, Japan, 2001, pp. 1435–1438.[13] S. Sahli, S. Rebiai, P. Raynaud, Y. Segui, A. Zenasni, S. Mouissat, Plasmas Polym. 7
(2002) 327–340.[14] L.K. Massey, Permeability Properties of Plastics and Elastomers: A Guide to Packag-
ing and Barrier Materials, 2nd ed. Plastics Design Library/William Andrew Pub.,Norwich, NY, USA, 2003. 132.
[15] L.Q. Xia, M. Chang, in: R. Doering (Ed.), Handbook of Semiconductor ManufacturingTechnology, 2nd ed., CRC Taylor & Francis, Boca Raton, 2008.
/dx.doi.org/10.1016/j.surfcoat.2014.02.070
8 List of publications 121
Patent I
Ultrathin Multi‐Layer Packaging
WO2011018709 (A8)
Inventors:
Andreas Hogg, Herbert Keppner, Juergen Burger, Thierry Aellen
Also published as:
EP2464421 (A2), US8313819 (B2), CA2770611 (A1), AU2010283553 (A1)
139
8 List of publications 133
8 List of publications 123
Patent II
Plasma Enhanced Polymer Ultra‐Thin Multi‐Layer Packaging
WO2011018707 (A1)
Inventors:
Andreas Hogg, Herbert Keppner, Juergen Burger, Thierry Aellen
Also published as:
EP2464280 (A1), US8313811 (B2), CA2770610 (A1), AU2010283551 (A1)
149
124 8 List of publications
8 List of publications 125
Patent III
Packaging with Active Protection Layer
WO2011018705 (A1)
Inventors:
Andreas Hogg, Herbert Keppner, Jerome Charmet, Juergen Burger, Thierry
Aellen
Also published as:
EP2464281 (A1), US8361591 (B2), CA2770606 (A1), AU2010283549 (A1)
159
126 8 List of publications
8 List of publications 127
Patent IV
Three dimensional packaging for medical implants
US2013330498 (A1)
Inventors:
Andreas Hogg, Yanik Tardy, Thierry Aellen, Herbert Keppner, Juergen Burger
Also published as:
EP2687256 (A1), JP2013255792 (A), CA2817328 (A1), AU2013206095 (A1)
169
128 8 List of publications
9 Appendix 129
9 Appendix
9.1 Uniformityalgorithm(MATLAB)
187
130 9 Appendix
9.2 Testleaksinspectioncertificates
188
9 Appendix 131
189
132 9 Appendix
9.3 Laminarflowfacilitycertificate
190
9 Appendix 133
9.4 Otherpublicationsbytheauthor
1. Development and in‐vitro characterization of an implantable flow
sensing transducer for hydrocephalus
Toralf Bork, Andreas Hogg, Markus Lempen, Daniel Müller, Damien Joss,
Thibaut Bardyn, Philippe Büchler, Herbert Keppner, Stephan Braun, Yanik
Tardy, Jürgen Burger, Biomedical microdevices, Volume 12, Issue 4,
(2010), pp. 607‐618.
2. In‐vitro characterization of an implantable thermal flow sensor for
hydrocephalus
J. Burger, T. Bork, A. Hogg, M. Lempen, D. Mueller, D. Joss, T. Bardyn, P.
Buechler, H . Keppner, Y. Tardy, World Congress 2009 Medical Physics
and Biomedical Engineering, IFMBE Proceedings, Volume 25, Issue 5,
(2010), pp 265‐268.
3. Multiphysics Finite Element Analysis of a CSF Flow Sensor
T. Bardyn, J. Burger, T. Bork, A. Hogg, H. Keppner, P. Büchler, Proceedings
of the 8th International Symposium on Computer Methods in
Biomechanics and Biomedical Engineering, (2008), pp. 1‐6
4. Implantable Thermal Flow Sensor for Neurosurgical Applications with
Asymmetric Design for High Flow Ranges
A. Hogg, M. Lempen, F. Page, F. Zumkehr, J. Burger, H. Keppner, P.
Buechler, T. Bardyn, T. Bork, Y.Tardy, VDI Fortschritt‐Bericht, Biotechnik /
Medizintechnik, VDE, Reihe 17, Nr. 267, (2007), pp 63‐64.
191
134 9 Appendix
10 Curriculum vitae 135
10 Curriculumvitae
Name: Andreas Hogg
Nationality: Swiss
Education and Qualifications:
08 / 2009 – 08 / 2014 PhD Thesis: “Development and Characterisation of
Ultrathin Layer Packaging for Implantable Medical
Devices”, Institute for Surgical Technology &
Biomechanics ISTB, University Bern, Switzerland and
Institute of Applied Microtechnologies IMA,
University of Applied Sciences He‐Arc, Neuchâtel,
Switzerland
08 / 2008 ‐ 02 / 2009 Master Thesis: “Hermetic packaging of implantable
sensors”, Codman Neuro Sciences Sàrl, a Johnson &
Johnson company, Le Locle, Switzerland
03 / 2007 ‐ 07 / 2008 Master of Science in Biomedical Engineering,
specialization in microsensors & actuators, University
of Bern, Medical Faculty, Switzerland
10 / 2006 ‐ 01 / 2007 Diploma Thesis: “Conception and realization of a
power control system for surgical navigated milling”,
MIMED, Technical University Munich, Germany
193
136 10 Curriculum vitae
10 / 2003 ‐ 07 / 2006 Diploma in Microtechnical Engineering,
University of Applied Sciences Bern, Switzerland
09 / 2002 ‐ 06 / 2003 Professional Baccalaureate, Berufsmittelschule Visp,
Switzerland
09 / 1998 ‐ 07 / 2002 Apprenticeship as Electronic Technician,
specialization in telecommunication and instrument
technology, Swisscom AG, Brig, Switzerland
Professional Experience
01 / 2014 – present Group leader Medical Technology at the Institute of
Applied Microtechnologies IMA, University of
Applied Sciences He‐Arc, Neuchâtel, Switzerland
08 / 2009 – present Development engineer at the Center of Smart
Implantable Systems (CSIS), Codman Neuro Sciences
Sàrl, a Johnson & Johnson company, Le Locle,
Switzerland
12 / 2011 – 12 / 2013 Project leader at the Institute of Applied
Microtechnologies IMA, University of Applied
Sciences He‐Arc, Neuchâtel, Switzerland
08 / 2009 – 11 / 2011 Research assistant at the Institute of Applied
Microtechnologies IMA, University of Applied
Sciences He‐Arc, Neuchâtel, Switzerland
194
10 Curriculum vitae 137
03 / 2009 – 07 / 2009 Development engineer (R&D center) in the
framework of a CTI project at Codman Neuro
Sciences Sàrl, a Johnson & Johnson company,
Le Locle, Switzerland
02 / 2007 ‐ 02 / 2009 Research assistant at the University of Applied
Sciences Bern, Department of Micro Technology and
Medical Technology. Biel, Switzerland
07 / 2005 ‐ 09 / 2005 Internship at the Technical University Munich,
Department of Micro Technology and Medical Device
Technology MIMED, Germany
06 / 2000 ‐ 10 / 2001 Internship at Swisscom Broadcasting and 3Vision
Light systems GmbH, Switzerland
Language skills
German Native language
English Fluent in speaking and writing
French Fluent in speaking and writing
195
138 10 Curriculum vitae
11 Declaration of Originality 139
11 DeclarationofOriginality
197
ISBN 978-2-940387-14-4 (printed)
ISBN 978-2-940387-15-1 (online)