integrated microsystems for continuous glucose monitoring, interstitial fluid...
TRANSCRIPT
Integrated microsystems for continuous glucose monitoring, interstitial fluid sampling
and digital microfluidics
FEDERICO RIBET
Doctoral Thesis
Stockholm, Sweden, 2020
ii
Front cover illustrations:
One of the solutions proposed in this thesis, aimed at measuring glucose
levels in the skin using a minimally invasive and painless device, for
diabetes care.
- (Left) Envisioned use of the continuous glucose monitoring system.
- (Center) Cross-section of the developed system inserted in the skin.
- (Right) Miniaturized glucose sensor assembled into the lumen of a
hollow microneedle.
KTH Royal Institute of Technology
School of Electrical Engineering and Computer Science
Division of Micro and Nanosystems
TRITA-EECS-AVL-2020:7
ISBN 978-91-7873-415-3
Akademisk avhandling som med tillstånd av Kungliga Tekniska högskolan
framlägges till offentlig granskning för avläggande av teknologie doktorsexamen
i elektroteknik och datavetenskap fredagen den 14:e februari klockan 10:00 i Sal
F3, Lindstedtsvägen 64, Stockholm.
Thesis for the degree of Doctor of Philosophy in Electrical Engineering and
Computer Science at KTH Royal Institute of Technology, Stockholm, Sweden.
© Federico, January 2020
Tryck: Universitetsservice US AB, 2020
iii
Abstract
Interdisciplinary research between medicine and microsystem engineering creates new
possibilities to improve the quality of life of patients or to further enhance the
performance of already existing devices. In particular, microsystems show great
potential for the realization of biosensors and sampling devices to monitor bioanalytes
with minimal patient discomfort. Microneedles offer a minimally invasive and painless
solution to penetrate the epidermis and provide access to dermal interstitial fluid (ISF),
to monitor various substances without the need for more invasive and painful extraction
of blood. Diabetes, for example, requires continuous monitoring of the glucose levels in
the body (CGM) to avoid complications. Although glucose is traditionally measured in
finger-prick blood, CGM, which is performed in ISF, has been proven to be beneficial in
the management of the disease. However, current commercial solutions are still
relatively large and invasive. In this work, an electrochemical glucose sensor 50 times
smaller than competing commercial devices was combined with a hollow silicon
microneedle and shown to be able to measure glucose levels in the dermis in vivo. A
scalable manufacturing method for the assembly of the two separately fabricated
components and their electrical interconnection was also demonstrated. At the same
time, a single data point may be sufficient in other situations, such as when only the
presence of a certain biomarker or drug needs to be assessed. Although continuous
monitoring is not required in these cases, the patient would still benefit by avoiding
blood extraction. However, there are no simple devices currently available to reliably
sample and store ISF. A painless microneedle-based sampling device designed to extract
1 µL of ISF from the dermis was realized. The sampled liquid is metered and stored in a
paper matrix embedded in a microfluidic chip. The sample could then be analyzed using
state-of-the-art tools, such as mass spectrometry.
On the other hand, device miniaturization also creates issues for sensor performance.
In certain types of electrochemical gas sensors, such as nitric oxide sensors used for
asthma monitoring, the reduced size results in a shorter device lifetime. These sensors
typically operate with a liquid electrolyte, subject to evaporation, and their long-term
stability tends to be proportional to the electrode size. To address this issue, a gas
diffusion and evaporation controlling platform to be integrated with this type of sensors
was proposed. Such a platform opens or seals the sensing compartment on demand,
potentially enabling sensor recalibration and evaporation reduction when the sensor is
not in use. The device is based on electrowetting-on-dielectric actuation of low-vapor-
pressure ionic liquid microdroplets on partially perforated membranes. The platform was
then modified to create a zero-insertion loss and broad-band-operation laser shutter.
Keywords: Microelectromechanical systems (MEMS), biosensors, biomedical devices,
continuous glucose monitoring (CGM), glucose sensors, microneedles, interstitial fluid
sampling, minimally-invasive technologies, electrochemical sensors, heterogeneous
integration, wire bonding, magnetic assembly, electrowetting-on-dielectric (EWOD),
digital microfluidics, ionic liquids, gas sensors, optical switches, laser shutters.
Federico Ribet, [email protected]
Division of Micro and Nanosystems
School of Electrical Engineering and Computer Science
KTH Royal Institute of Technology, SE 100 44 Stockholm, Sweden
iv
Sammanfattning
Tvärvetenskaplig forskning mellan medicin och mikrosystemteknik skapar nya
förutsättningar att förbättra livskvaliteten för patienter eller ytterligare
förbättra prestandan hos existerande medicintekniska hjälpmedel. I synnerhet
uppvisar mikrosystem stor potential för att förverkliga biosensorer och
provtagningsanordningar för att mäta bioanalyter med minimalt besvär för
patienten. Mikronålar erbjuder en minimalinvasiv och smärtfri lösning för att
penetrera överhuden och ge tillgång till dermal interstitiell vätska (ISF) för att
mäta olika ämnen, utan behov av mer invasiv och smärtsam extraktion av blod.
Vid diabetes exempelvis, krävs kontinuerlig övervakning av glukosnivån i
kroppen (CGM) för att undvika komplikationer. Även om glukos traditionellt
mäts i kapillärblod från fingret, har CGM, som utförs i ISF, visat sig vara
fördelaktigt i bahandlingen av sjukdomen. Men nuvarande kommersiella
lösningar för CGM är fortfarande relativt stora och invasiva. I detta arbete
kombinerades en elektrokemisk glukossensor, 50 gånger mindre än
konkurrerande kommersiella sensorer, med en ihålig mikronål och visade sig
kunna mäta glukosnivåer i hudlagret in vivo. En skalbar tillverkningsmetod för
montering av de två separat tillverkade komponenterna och deras elektriska
sammankoppling visades också.
I andra provtagningssituationer kan en enda datapunkt vara tillräcklig, till
exempel när endast närvaron av en viss biomarkör eller läkemedel behöver
bekräftas. Även om det inte krävs kontinuerlig övervakning i dessa fall skulle det
underlätta för patienten att slippa blodprovtagning. För närvarande finns det
emellertid inga enkla enheter tillgängliga för att pålitligt inhämta och lagra ISF.
I det här arbetet, har en smärtfri mikronålbaserad provtagningsanordning
utformad för att extrahera 1 µL ISF från dermis realiserats. Den extraherade
vätskan mäts och lagras i en pappersmatris inbäddad i ett mikrofluidikchip.
Provet kan sedan analyseras med hjälp av moderna analysverktyg, såsom
masspektrometri.
Å andra sidan skapar miniatyrisering ibland också problem för sensorers
prestanda. I vissa typer av elektrokemiska gassensorer, såsom kväveoxidsensorer
som används för astmaövervakning, medför den reducerade storleken en kortare
livsläng. Dessa sensorer innehållar ofta elektrolyt som kan avdunsta, och deras
långsiktiga stabilitet tenderar att vara proportionell mot elektrodstorleken. För
att hantera detta, utvecklades en diffusions- och avdunstningsbegränsande enhet
som kan integreras med denna typ av sensorer. En sådan enhet öppnar eller
försluter sensorn på begäran, vilket möjliggör kalibrering och förhindrar
avdunstning när sensorn inte används. Enheten är baserad på s.k.
elektrovätning på dielektrika och möjliggör manövrering av mikrodroppar av
jonvätska med lågt ångtryck på delvis perforerade membran. Enheten
modifierades sedan för att skapa en optisk slutare utan införingsförlust.
Federico Ribet, [email protected]
Avdelningen för Mikro- och Nanosystem, Skolan för Elektroteknik och Datavetenskap,
Kungliga Tekniska Högskolan, 100 44 Stockholm, Sverige
v
Contents
List of Publications ............................................................................................ vii
List of Abbreviations ........................................................................................... ix
Thesis Overview .................................................................................................... xi
1. Introduction and medical motivations ........................................................ 1
1.1 Minimally-invasive monitoring .................................................................. 1
Needs and currently used biological matrices ................................ 1 1.1.1
Physiology of the skin ...................................................................... 2 1.1.2
Interstitial fluid ................................................................................ 3 1.1.3
Laboratory versus point-of-care testing .......................................... 4 1.1.4
1.2 Diabetes mellitus ........................................................................................ 4
1.3 Asthma and breath monitoring .................................................................. 7
2. Technological background ............................................................................. 9
2.1 Miniaturization, MEMS and integration ................................................... 9
Heterogeneous integration ............................................................ 11 2.1.1
2.2 Microfluidics .............................................................................................. 11
Capillary-driven microfluidics ....................................................... 12 2.2.1
Electrowetting ................................................................................ 13 2.2.2
Digital microfluidics ....................................................................... 15 2.2.3
2.3 Microneedles ............................................................................................. 16
2.4 Electrochemical sensing ........................................................................... 18
Reference and pseudo-reference electrodes ................................... 19 2.4.1
Electrochemical sensing and functionalization methods ............. 20 2.4.2
Enzymatic biosensors ..................................................................... 20 2.4.3
Problems in miniaturized gas sensors .......................................... 21 2.4.4
3. Thesis objectives ............................................................................................. 23
4. Continuous glucose monitoring systems .................................................. 25
4.1 Sensing technologies ................................................................................. 25
Electrochemical glucose sensing .................................................... 25 4.1.1
Other sensing principles ................................................................ 27 4.1.2
4.2 Commercial state-of-the-art of CGMS ..................................................... 28
4.3 Microneedle-based CGM solutions ........................................................... 29
vi
4.4 A combined miniaturized solution ........................................................... 31
4.5 Miniaturization of a glucose sensor ......................................................... 32
Sensor design and functionalization ............................................. 32 4.5.1
Sensor performance in vitro ........................................................... 35 4.5.2
4.6 A real-time minimally-invasive CGMS .................................................... 37
Microneedle and CGMS design and realization ........................... 37 4.6.1
Performance in vitro and in vivo ................................................... 38 4.6.2
4.7 What’s next? .............................................................................................. 40
5. Microsensor fabrication and integration .................................................. 41
5.1 Microsensor fabrication ............................................................................ 41
5.2 Microchip assembly and contacting techniques ...................................... 42
5.3 Controlled magnetic assembly ................................................................. 44
5.4 Edge wire bonding .................................................................................... 46
5.5 Microchip and substrate fabrication for integration studies .................. 48
5.6 Results ....................................................................................................... 49
Magnetic assembly process characterization ................................ 51 5.6.1
Contact quality and reliability ...................................................... 52 5.6.2
6. Interstitial fluid sampling ............................................................................ 55
6.1 The use of ISF as a specimen in the clinic ............................................... 55
6.2 State-of-the-art of minimally-invasive sampling ..................................... 56
Blood sampling ............................................................................... 56 6.2.1
ISF sampling devices ..................................................................... 57 6.2.2
6.3 A device for volume-metered ISF sampling ............................................. 58
Design and implementation ........................................................... 59 6.3.1
Characterization ............................................................................. 60 6.3.2
6.4 Perspectives............................................................................................... 61
7. Gas diffusion and evaporation control for gas sensors ......................... 63
7.1 Digital microfluidic platforms .................................................................. 63
Ionic Liquids as movable microdroplets ........................................ 64 7.1.1
7.2 A diffusion and evaporation controller for gas sensors ........................... 65
Performance ................................................................................... 67 7.2.1
8. EWOD-based optical shutters ...................................................................... 69
8.1 Liquid-based optical shutters ................................................................... 69
8.2 Presented solution .................................................................................... 71
8.3 Characterization ....................................................................................... 73
9. Summary and Outlook .................................................................................. 75
Perspectives ........................................................................................................ 77
Acknowledgments................................................................................................ 79
Bibliography ......................................................................................................... 81
Paper Reprints ..................................................................................................... 97
vii
List of Publications
This thesis is based on the following papers published in peer-reviewed
international journals and a manuscript:
I. "Ultra-miniaturization of a planar amperometric sensor targeting
continuous intradermal glucose monitoring," F. Ribet, G. Stemme, and N.
Roxhed, Biosensors and Bioelectronics, vol. 90, pp. 577-583, 2017.
II. "Real-time intradermal continuous glucose monitoring using a minimally
invasive microneedle-based system,” F. Ribet, G. Stemme, and N. Roxhed,
Biomedical microdevices, vol. 20 (4), pp. 101, 2018.
III. "Vertical integration of microchips by magnetic assembly and edge wire
bonding," F. Ribet*, X. Wang*, M. Laakso, S. Pagliano, F. Niklaus, N.
Roxhed, and G. Stemme, in print, Microsystems and Nanoengineering, 2020.
IV. "Minimally invasive and volume-metered extraction of interstitial fluid:
bloodless point-of-care sampling for bioanalyte detection," F. Ribet, M.
Dobielewski, M. Böttcher, O. Beck, G. Stemme, and N. Roxhed, manuscript,
Jan. 2020.
V. "Gas diffusion and evaporation control using EWOD actuation of ionic liquid
microdroplets for gas sensing applications," F. Ribet, L. De Pietro, N.
Roxhed, and G. Stemme, Sensors and Actuators B: Chemical, vol. 267, pp.
647-654, 2018.
VI. “Zero-insertion-loss optical shutter based on electrowetting-on-dielectric
actuation of opaque ionic liquid microdroplets” F. Ribet, E. De Luca, F.
Ottonello-Briano, M. Swillo, N. Roxhed, and G. Stemme, Applied Physics
Letters, vol. 115 (7), pp. 073502, 2019.
viii
The contribution of the author to each publication, major (X≥80%) (•••),
substantial (80%>X>33%) (••), minor (X≤33%) (•):
Design Fabrication Experiments Analysis Writing
I. ••• ••• ••• ••• •••
II. ••• ••• ••• ••• •••
III. ••• •• •• •• ••
IV. ••• ••• •• ••• •••
V. ••• ••• ••• ••• •••
VI. ••• ••• •• ••• •••
This work has also been presented at the following peer-reviewed
international conferences (not included as paper reprints):
VII. “Microneedle-based system for minimally invasive continuous monitoring
of glucose in the dermal interstitial fluid,” F. Ribet, G. Stemme, and N.
Roxhed, in Proceedings of IEEE 31st International Conference on Micro
Electro Mechanical Systems (MEMS), pp. 408-411, 2018.
VIII. “Ionic liquid microdroplet manipulation by electrowetting-on-dielectric for
on/off diffusion control,” F. Ribet, L. De Pietro, N. Roxhed, and G. Stemme,
in Proceedings of IEEE 31st International Conference on Micro Electro
Mechanical Systems (MEMS), pp. 1181-1184, 2018.
IX. "Microneedle and sensing microprobe integrated system for minimally
invasive continuous glucose monitoring,” F. Ribet, G. Stemme, and N.
Roxhed, 5th International conference on Microneedles, pp. 33, 2018. Oral
presentation.
X. “Zero-loss optical switch based on ionic liquid microdroplet EWOD
actuation,” F. Ribet, E. De Luca, F. Ottonello-Briano, M. Swillo, N.
Roxhed, and G. Stemme, in Proceedings of IEEE Transducers &
Eurosensors XXXIII, 2019.
Other peer-reviewed papers with minor contributions by the author, not
included in this thesis:
XI. “Human Cell Encapsulation in Gel Microbeads with Cosynthesized
Concentric Nanoporous Solid Shells,” X. C. Zhou, T. Haraldsson, S. Nania,
F. Ribet, G. Palano, R. Heuchel, M. Löhr, and W. van der Wijngaart,
Advanced Functional Materials, vol. 28 (21), pp. 1707129, 2018.
ix
List of Abbreviations
2.5D Two-Dimensional projections in the third direction
3D Three-Dimensional
BMIM-BF4 1-Butyl-3-methylimidazolium tetrafluoroborate
BMIM-PF6 1-Butyl-3-methylimidazolium hexafluorophosphate
BSA Bovine Serum Albumin
CE Counter Electrode
CGM Continuous Glucose Monitoring
CGMS Continuous Glucose Monitoring System
CV Cyclic Voltammetry
DBS Dried Blood Spots
DM Diabetes Mellitus
DMF Digital MicroFluidics
DRIE Deep Reactive Ion Etching (or Bosch Process)
EC Electrochemistry
EMIM-BF4 1-Ethyl-3-methylimidazolium tetrafluoroborate
EWOD ElectroWetting-On-Dielectric
FAB Free Air Ball
FEM Finite Element Modeling
FET Field-Effect Transistor
GA GlutarAldehyde
GOx Glucose Oxidase
IC Integrated Circuit
IL Ionic Liquid
IoT Internet of Things
IR InfraRed
ISF InterStitial Fluid
ISFET Ion-Sensitive Field-Effect Transistor
ISO International Organization for Standardization
ITO Indium Tin Oxide
LC-MS/MS Liquid Chromatography – Tandem Mass Spectrometry
LED Light Emitting Diode
LFA Lateral Flow Assay
MEMS MicroElectroMechanical Systems
MN MicroNeedles
MOSFET Metal-Oxide-Silicon Field-Effect Transistor
MS Mass Spectrometry
x
NHE Normal Hydrogen Electrode
NO Nitric Oxide
OCP Open Circuit Potential
OCT Optical Coherence Tomography
PBS Phosphate Buffered Saline
PoC Point of Care
PTFE PolyTetraFluoroEthylene
Q-RE Quasi Reference Electrode (or pseudo-reference electrode)
RE Reference Electrode
Redox Reduction and Oxidation (reaction)
RIE Reactive Ion Etching
SEM Scanning Electron Microscope
SOI Silicon On Insulator
WE Working Electrode
xi
Thesis Overview
The presented work explores areas in which the miniaturization, design, and
integration of microelectromechanical systems (MEMS) and microfluidic
technologies can improve or provide novel technical solutions. In the past decades,
the development and spread of micromanufacturing technologies offered new
possibilities for the realization of devices for an extremely broad range of
applications, and have been explored and developed especially in conjunction
with consumer electronics. However, for medical applications, this potential is
still far from being deeply explored. In this work, various microsystems and
related manufacturing technologies were designed, realized and characterized.
Such devices and technologies aimed at specific medical applications, including
diabetes monitoring, minimally invasive bodily fluid sampling and asthma
monitoring. One of the main focuses of the thesis is on minimally invasive
monitoring of bioanalytes in interstitial fluid in the skin, using microneedle-
based systems. At the same time, some of these technologies could also be used
for different applications (e.g. heterogeneous integration of electronic components)
or modified to serve different purposes (e.g. realization of laser shutters).
Structure of the thesis
The thesis is composed of nine chapters and is meant to be read in combination
with the reprints of the articles (I to VI) appended to the thesis. All technical
details are reported in these articles, while the main body of the thesis provides
an overview of the involved fields and the obtained results, aimed to put them
into a broader perspective. The first three chapters are introductory and provide
the essential background to understand the theory behind the developed devices
or methods, and the targeted medical and technological needs.
- Chapter 1 provides information about the motivation, the relevance and
the medical needs targeted by the work performed in the thesis, including
point-of-care bodily fluid sampling, diabetes care, and asthma monitoring.
- Chapter 2 provides an overview of the technological background related, in
particular, to electrochemical sensors, microneedles and more generally
microelectromechanical systems (MEMS). MEMS technologies are the
common denominator for the development of most presented devices. This
chapter touches upon some key aspects of the work such as
microfabrication, system integration, and microfluidics.
xii
- Chapter 3 summarizes the objectives of the thesis, tackling the needs
highlighted in the first two chapters.
After this general introduction, chapters 4 to 8 focus on the research and
development tracks towards a specific device or technology.
- Chapter 4: The design and realization of a miniaturized continuous glucose
monitoring system.
- Chapter 5: Methods for microsensor fabrication and integration.
- Chapter 6: A painless and volume-metered ISF sampling strategy.
- Chapter 7: Gas diffusion control for electrochemical gas sensing.
- Chapter 8: The realization of a liquid-based optical shutter.
Finally, Chapter 9 provides a summary of the achieved results and some
perspectives for future developments.
1
Chapter 1
Introduction and medical motivations This chapter introduces some key aspects to contextualize the work performed in
this thesis. The aim is both to provide the essential medical background to
understand the human physiology and the needs involved, and to point out the
relevance and potential impact of this work, including the development of
minimally invasive techniques for bioanalyte monitoring in skin and breath.
1.1 Minimally-invasive monitoring
To access information on the health of a patient beyond the assessment of the
symptoms, analyses of bodily samples needs to be performed. Assessment of the
concentration or presence of specific analytes in the body is an essential aspect of
medical practice for patient monitoring and disease diagnostics. Such an
assessment enables doctors and nurses to get information, either qualitative or
quantitative, on the status of a patient and take medical decisions based upon
that. Depending on the disease or the condition to be assessed, different bodily
matrices are to be used, including tissues (e.g. via biopsy) and different fluids
(both liquids and gases). In general, it is straightforward to say that an ideal
specimen to measure a certain analyte would be the one that satisfies the
following three characteristics: (i) it can provide sufficiently precise information
to lead to the correct medical conclusions, and it can be obtained with (ii) the
minimum patient discomfort and (iii) with minimum direct and indirect costs for
the healthcare system. Minimally invasive monitoring technologies try to
minimize the damage, the pain and the discomfort for the patient in accessing
the necessary sample from the body. For example, this might consist in the
miniaturization of the used components or in enabling access to alternative
matrices, easier to access but still able to provide the same information of the
standard ones.
Needs and currently used biological matrices 1.1.1
The standard practice is currently to use blood as a measurement matrix for a
large variety of analytes, e.g. glucose, cholesterol, medical drugs, etc. However,
blood extraction entails invasive and painful methods, whether done via finger
prick or intravenously. Less invasive sampling techniques would be beneficial,
especially for point-of-care (PoC) testing, where operations need to be performed
2
by patients or untrained personnel, as later discussed. Other non-invasive
sampling strategies involve measurements in urine and hairs, while minimally-
invasive techniques use saliva, tears, and sweat [1], [2]. Some of these options
have recently been tested for continuous bioanalyte monitoring, but the targeted
analyte is not always present in these matrices at a sufficient concentration to be
detected. Additionally, the correlation between their concentration in these
matrices and the systemic concentration in blood (gold standard) is currently
unclear or known to vary with bodily secretion rates, leading to unreliable
measurements. [1], [3]
Being able to monitor substances with the same confidence as nowadays done
in blood, but without the same pain and discomfort for the patients, would
represent a dramatic improvement towards a more patient-centered healthcare.
An emerging matrix for minimally-invasive monitoring is interstitial fluid (ISF),
having a good correlation with blood values for several analytes (e.g. glucose) and
being potentially minimally-invasively accessible from within the skin. In this
thesis, two devices that enable minimally invasive monitoring of analytes in ISF
are presented, in chapters 4 and 6, respectively.
Physiology of the skin 1.1.2
The skin is the largest organ in the human body. It acts as a sensory and
regulatory element, and as a selective permeable barrier between the
environment and the other organs and tissues. A cross-section of the skin is
illustrated in Fig. 1. There are two main layers: epidermis and dermis. Below the
dermis, a third layer of tissue is present: the hypodermis (or subcutaneous tissue).
Fig. 1. Illustration of the cross-section of the skin, showing the different layers and their internal
composition (from[4]).
3
The outermost layer, the epidermis, serves as a waterproof barrier enclosing the
body and acts as a protection against infections. The middle layer, the dermis,
protects the body from external stress and strain, and hosts thermo- and
mechanoreceptors. Instead, the subcutaneous tissue mainly consists of connective
and fat tissue. Its main purposes are to attach the skin to muscles and bones, and
to connect nerves and blood vessels to the skin. The thickness of the different
layers strongly varies across different body locations, with the overall skin
thickness ranging from 0.05 mm on the eyelids to more than 1.5 mm on the feet
soles. Considering the forearm, the typical location used e.g. for blood sampling,
the average skin thickness is about 1 mm [5].
Interstitial fluid 1.1.3
ISF is a liquid present in all parts of the body between tissue cells, outside the
cells themselves and the blood vessels. It accounts for about one-fourth of the
water content in the human body. The composition of the ISF is rather similar to
that of blood plasma, but with lower protein content. In addition to water, it
contains salts, glucose, small proteins, lipids, amino acids, fatty acids, hormones,
neurotransmitters, and other small biomolecules [6]. This similarity is due to the
continuous molecular exchange between the blood in capillary vessels and the
interstitial space occurring by diffusion, hydrostatic pressure, osmotic pressure,
active transporters, and transcytosis across the capillary endothelium. However,
active transporters and tissue specificities are responsible for the small ISF
composition differences among different organs. Among other functions, the ISF
is responsible for shuttling water, nutrients, and waste products between the
cells and the bloodstream by diffusion.
Because of this similarity with blood plasma, ISF is gaining a growing interest
as a monitoring matrix. In particular, as mentioned, skin ISF becomes very
interesting for monitoring applications because of its accessibility. Moreover, the
concentration of several bioanalytes is closely correlated to the one in blood,
which is typically used as a medical standard. Substances such as glucose, for
example, are present in the same concentration as in blood, with just a short
physiological delay, quantified in the order of 4 to 12 minutes, due to the
diffusion time from the blood capillaries [7], [8]. Additionally, by targeting skin
ISF, monitoring is not only potentially achieved in a minimally-invasive fashion,
but continuous monitoring with transdermal or intradermal devices can also be
performed, providing a complete temporal picture of the evolution of the targeted
analyte concentration.
Among the different skin compartments, the dermis proportionally contains
the largest amount and the most homogenously distributed ISF. In the
hypodermis, inhomogeneity in size of the adipocytes and scarcer capillary
distribution detrimentally affect the local ISF analyte concentrations, [9] whilst
in the epidermis blood capillary are not present at all. As a consequence, the
dermis is the ideal measurement location within the human skin.
4
Laboratory versus point-of-care testing 1.1.4
Laboratory testing consists in sampling the specimen from the patient and then
performing the necessary analyses using medical laboratory equipment present
elsewhere. This is the traditional approach and it has pros and cons. Its main
advantage is that specialized laboratories can run reliable measurements using
state-of-the-art tools and highly specialized personnel. The quality and efficiency
of the measurement is thus difficult to beat. On the other hand, the main
disadvantage is the time delay. Sample transport and analysis might require
several hours or days before the outcome is obtained by the patient.
A different approach is PoC testing, which consists in performing the required
specimen analysis directly at the location of the patient in need. This can be at
home, at the bedside, or at the doctor’s office. Some PoC tests are nowadays
extremely widespread. Some common examples are lateral flow tests such as
pregnancy tests, colored test strips used e.g. for urine analysis, and handheld
strip readers such as the ones used for glucose monitoring (see section 1.2 and
chapter 4) [10]. However, there are some possible drawbacks or complications
also in the use of PoC solutions. First of all, it is difficult to guarantee the same
measurement reliability and quality of laboratory testing. PoC devices have to be
simple enough to be cost-effective (both for sampling and measurement) and be
used by less trained personnel or self-performed, while still providing medically
valid information. For certain tests, which involve analytes with very low
concentrations or for simultaneous detection of multiple substances, it is not
always possible to gather the same amount of information with portable and
disposable devices. Moreover, a fast answer is not always needed and, thus, the
cost and complications are not balanced by the usefulness. Therefore, in brief,
PoC is preferable when a quick assessment is required or when technology allows
obtaining the required information at the same or lower cost and effort with
respect to laboratory testing, and with negligible side effects. An example of a
condition requiring constant PoC monitoring, out of the clinic, is diabetes.
1.2 Diabetes mellitus
Diabetes mellitus is a metabolic disorder related to the regulation and control of
glycemia, i.e. the concentration of blood glucose. Glucose is a sugar and is one of
the main energy sources for the cells. In a healthy individual, the regulation of
glucose levels is fundamental and is mainly performed by insulin, a hormone
produced by beta cells in the pancreas. In normal individuals, the metabolism of
carbohydrates and sugars is regulated by insulin, which is in turn upregulated or
downregulated according to the glucose levels detected by the pancreatic beta
cells. Thanks to this action, glucose concentration oscillations in the body are
controlled and limited in amplitude. However, in diabetic individuals, a faulty
response to this mechanism occurs, leading to unstable and dangerous
oscillations of the bodily glucose levels. This disorder can be generated, most
commonly, by two mechanisms. In type-1 (T1DM or juvenile diabetes) the
pancreas is unable to produce insulin, while in type-2 (T2DM or senile diabetes)
5
the body is unable to properly use the produced insulin, in a process also called
insulin resistance.
According to the International Diabetes Federation, diabetes is currently
affecting 400 million people (5% prevalence), 90% of which have T2DM, and
causes 5 million deaths worldwide yearly. And these numbers are rising, now at
an alarming pace of about +2.5% yearly [11]. This situation is not only strongly
affecting a large number of individuals, but has also generated enormous
economic pressure on the healthcare systems: approximately 10% of global
healthcare direct expenses are currently associated to diabetes. Among these
costs, hospitalization accounted for the largest portion. Moreover, also indirect
welfare costs such as social benefits, absenteeism, and early retirement have to
be added to these numbers, providing a clear picture of the medical and societal
importance of improved diabetes monitoring and treatments [12]. As of today,
diabetes is not curable, and is responsible for several complications, both in the
short and long term. Possible complications include cardiovascular disorders,
neuropathy, nephropathy, retinal damages and improper healing of lower limb
wounds leading to amputations. The combination of diabetes with several chronic
comorbidities, i.e. co-occurring medical conditions, is unfortunately very common,
especially with cardiovascular diseases and hypertension [13], [14].
To prevent or limit complications, patients need to monitor their glucose levels
to get the correct treatment, e.g. drugs or insulin shots in the case of T1DM. This
is a clear example in which PoC testing is necessary since it can offer a platform
for continuous monitoring of the patient’s condition at home. The conventional
method for self-monitoring involves fingertip pricking with a lancet to access a
drop of capillary blood (Fig. 2a), which is then placed in contact with a blood
glucose meter [15]. This concept, while nowadays reliable and considered a
standard, entails a series of problems. Firstly, blood sampling is only performed
at discrete time points (typically 4 - 5 times a day at most) and thus provides a
very limited picture of the glycemic evolution over time, as depicted in Fig. 2b. By
only having a few data points, there is a large risk of missing important events
and it is impossible to see trends, i.e. whether the concentration is falling or
rising at sampling time. Damages are caused by large oscillations even more than
by constant hyperglycemia, and it is therefore important to carefully
continuously track glucose levels over time, not to miss peaks, but also to
minimize oscillations and act promptly. Secondly, regular pricking exposes to a
higher risk of infections and, thirdly, the pain and discomfort may often result in
reduced patient compliance.
6
Fig. 2. (a) Capillary blood extraction by finger pricking using a lancet. (b) Example of evolution of
the bodily glucose levels over a day, highlighting the lack of information provided by a limited
number of data points.
Therefore, continuous glucose monitoring (CGM) is essential to improve
treatment quality [16]. CGM provides both real-time tracking and a clearer
picture of the glycemia evolution in association with different behaviors of the
patient. Since it requires constant access to a bodily matrix, CGM is not
performed in blood, but in ISF, due to accessibility reasons. In the skin, ISF is
potentially easily accessible with minimally-invasive methods. Therefore,
continuous glucose monitoring systems (CGMS) provide a combination of better
monitoring quality with reduced invasiveness and discomfort for the patients.
As further discussed later (chapter 4.2), despite some great technical
improvements in the last five years, current CGMS are not yet ideal in terms of
design, size, invasiveness, and cost. While the efficacy and importance of CGM
have been proven, widespread adoption of CGMS is still limited because of their
invasiveness and relatively high cost, and because in most countries they are not
yet listed in governments’ health guidelines and, thus, are not sponsored by
national healthcare systems [17]. The Holy Grail of diabetes treatment would be
the realization of a minimally-invasive closed-loop CGMS, i.e. a system that can
sense glycemia, process the information and automatically delivery the exact
amount of insulin or drugs, without any need for the patient to actively perform
any action, and without creating pain and significant discomfort to the user.
While the single components and the data processing capability are potentially
available (even if they would still benefit from several technical and physiology-
related improvements), the sensing reliability is the main issue. It is extremely
challenging to create transdermal or implantable sensors that can last for a long
time (weeks) and guarantee ~100% reliability. Reliability is fundamental since
wrong dosage of insulin, especially in hypoglycemic conditions, can result in
severe complications and even death.
These aspects highlight the relevance of improved diabetes monitoring from
the patient and societal perspectives, as well as the main problematics that the
field, from an engineering perspective, has to face to enable widespread
7
distribution: reduce the cost, minimize the invasiveness to increase patient
compliance, and improve reliability. In this thesis, chapters 4 and 5 present a
system designed to address these requirements, to monitor glucose in real-time
and with minimal patient discomfort.
1.3 Asthma and breath monitoring
Monitoring specific analytes to acquire information about a certain disease is a
common way of assessing the health status of a patient. In the case of diabetes,
glucose concentration is tracked, being the molecule directly causing
complications. In other diseases, however, there is not a responsible molecule
that can be directly monitored. In these cases, different biomarkers that are
related to the status and evolution of the disease are measured instead. Their
presence and concentration allow extrapolating information about the disease
under investigation. An example of such a disease is asthma. Asthma is a chronic
inflammatory condition that results in the swelling and narrowing of the airways,
and that in different forms affects more than 5% of the population. Nitric oxide
(NO) is a gas used as a marker for asthma. Since it is produced by the lungs in
case of inflammation, the NO concentration relates to the degree of inflammation
present in asthmatic patients. Therefore, measurement of the NO concentration
in breath is performed to gather information about the patient condition. In this
thesis, a technical solution specifically designed for potentially improving gas
sensing system, such as the ones used today for asthma monitoring, is presented.
The technological challenges and needs related to such sensing systems are
described in the next chapter (section 2.4.4).
8
9
Chapter 2
Technological background
This chapter provides an overview of the technologies involved in the realization
of the devices and the methods presented in the thesis. The aim is to provide the
minimum theoretical and technological background to understand the physics
and chemistry behind the operation of the realized devices and their
manufacturing methods, as well as to point out some limitations in the state-of-
the-art of such technologies for the targeted applications. The key technologies
and components used for the realization of all demonstrated devices and methods
are: (i) MEMS and microsystem integration, (ii) capillary-driven and digital
microfluidics, (iii) microneedles, and (iv) electrochemical sensors.
2.1 Miniaturization, MEMS and integration
The rise of consumer electronics devices initially started in the 1920s with radio
broadcasting, followed by the spread of TVs in the 1950s. Meanwhile, in 1947, the
first type of transistor was invented, followed by the planar MOSFET, in 1959.
These last two inventions can be easily considered among the most
revolutionizing ones in human history. Their direct impact on society and
consumer electronics has been tremendous, and MOSFET had soon become the
most manufactured device in history. This is due to two main factors. The first
one is that these devices can be made using silicon as the main bulk material.
And that luckily silicon is relatively cheap and easy to get since ¼ in weight of
the Earth’s crust is made of it. The other factor is technological, and related to
the keyword “planar”, which had multiple important consequences. In particular,
the planar process on silicon wafers eventually enabled parallelization and
miniaturization. Planarity allowed the use of parallel fabrication techniques,
such as photolithography, implantation and thin-film deposition, key technologies
to realize semiconductor ICs and microchips on a large scale. Due to the
enormous demand, large investments in fabrication infrastructure could be made.
The ultimate consequence of these factors is as easy as fundamental: it soon
became tremendously cheap and standard to produce ICs. Additionally, the
smaller, the cheaper, so that since then, the transistor size continuously shrunk
(today, down to 7 nm) and the total number of MOSFETs per IC roughly doubled
every two years (Moore’s law) [18]. However, the consequences of the
10
development of silicon microfabrication technologies did not affect only the
possibility to mass-produce MOSFETs. The development and spread of silicon
micromanufacturing technologies initially developed for and explored especially
in conjunction with consumer electronics, also offered at the same time new
possibilities for the realization of devices for an extremely broad range of
applications.
The second wave of devices enabled by these silicon manufacturing
technologies happened in the 1990s with microelectromechanical systems
(MEMS), such as inkjet nozzles, accelerometers, pressure sensors, microphones,
and gyroscopes. Commercially, the widespread of MEMS devices was (and still is)
driven especially by the automotive and, later, by the smartphone industries.
Using silicon as the structural material, [19] the same microfabrication
technology and infrastructure could be used and adapted where necessary,
especially for etching and creating 2.5D or suspended structures.
Miniaturization capabilities provided by microfabrication technologies
resulted in three different and interesting effects. The first is the size reduction
per se, enabling portability and the possibility to integrate multiple functions in
less space. The second is the possibility to have more devices per area during
production, i.e. to drastically reduce the cost per device. The third is the rise of
totally new phenomena at a small scale, due to the different physical and
chemical properties of materials and structures characterized e.g. by a high
surface-to-volume ratio. Some of these properties can be used to build devices
that would not even work in their macroscopic counterparts. Several branches of
research evolved from the MEMS field, including optical and radiofrequency
applications, microfluidics and bioMEMS. In some cases, since fabrication
methods evolved from silicon-based technologies, the only thing remaining in
common between these areas is the feature sizes. However, a common
denominator is that sensors and actuators of different types are becoming
smaller and cheaper, allowing for the third big wave: the Internet of Things (IoT).
IoT is a system of several connected devices, able to continuously monitor us and
our surroundings, without the need for human interaction, collecting and
transmitting information on every aspect, including environmental monitoring,
smart-home appliances, transportation, power management, and healthcare. The
advancement of microsystem technologies is a crucial component to achieve a
smarter and more interconnected world.
For medical applications, the potential of miniaturization and MEMS-derived
technologies is still largely unexplored. Interdisciplinary work between medicine
and engineering is nowadays creating possibilities to improve or even
revolutionize some biomedical technologies, taking advantage of the mature
manufacturing infrastructure developed for the consumer electronics industry,
and by the presence of new different phenomena to be exploited at the micro and
nanoscale. Chapters 4 to 7 report some examples in which miniaturization and
MEMS technologies can lead to the creation of devices providing direct benefit to
patients or improving medical technologies currently in use.
11
Heterogeneous integration 2.1.1
Micromachining and MEMS technologies nowadays provide reliable, scalable and
low cost-per-device fabrication possibilities. However, they still have one
fundamental limitation: they do not enable real 3D geometries. At the same time,
many emerging applications require out-of-plane features and structures with
complex 3D geometries to fulfill their function. Examples of such devices are
inductors, antennas, optical attenuators, flow sensors, and biosensing systems
[20]–[23]. A way to create such systems is to combine and integrate different
planar structures, manufactured using standard microfabrication methods into
3D structures. Therefore, the controlled assembly and electrical interconnection
of heterogeneous microcomponents into out-of-plane functional devices also
became a need, and a challenge, in the MEMS field. A method to enable assembly
and contact of microchips to create 3D structures is presented in chapter 5.
2.2 Microfluidics
As mentioned, one of the “spin-offs” of the MEMS-field is microfluidics. Also
microfluidics initially took advantage of the mature silicon manufacturing
technologies to create miniaturized devices and exploit new phenomena, before
shifting into other fabrication methods, including molding, lamination and 3D
printing [24]. Liquid handling at the microscale entails a series of interesting
features, which have been exploited, especially for biomedical applications, to
create devices commonly called “lab-on-chip”. The idea is that by integrating
multiple fluidic functions on a chip, the operation can be controlled, automatized
and be made user-independent. This can simplify the tests or studies that are
currently made in biomedical laboratories, or replace them with devices that can
be used for point-of-care applications and that, potentially, can be low cost and
thus disposable because of the small size, the materials used, and the simplicity
of their fabrication and assembly.
Microfluidics studies the behavior of fluids in the sub-microliter volume range.
A fluid, when considering small droplets or its flow in a confined channel with
sizes in the sub-mm range, gives rise to physical phenomena that are different
from the ones observed in the macroscopic world. For example, in channels at
this scale, laminar flow is prevalent with respect to turbulent flow, as described
by the Reynolds number (Re = ρνL/μ, where ρ = density, ν = flow velocity, L =
characteristic length and μ = viscosity), a ratio between inertial and viscous force
components. The smaller this number, the less inertia affects the fluid behavior.
Because of this, for example, the fluid flow profile can be determined by knowing
the initial boundary conditions and liquid mixing happens by diffusion only [25].
Another implication of the small sizes is that surface properties and short-
distance forces (e.g. Van der Waal forces) play a major role in microfluidic devices,
while gravitational forces can often be neglected. As a consequence, cohesion
forces within a liquid at a liquid-air interface prevail over adhesion forces and
result in the minimization of the surface area, because molecules at the interface
are pulled inwards. This effect creates a surface tension (γ). The formation of
droplets and the floating of insects on water are, for example, due to this
12
phenomenon. In most practical situations, this behavior is studied at a solid-
liquid-gas interface, such as a liquid droplet on a surface (or a liquid within a
channel). In this situation, a contact angle can be defined as the angle at the
three-phase intersection between the liquid and the substrate. Since microfluidic
devices typically operate in air, the only variables are the liquid’s and substrate’s
properties. This angle provides important information on the wetting behavior of
a certain liquid-solid system: if the angle is lower than 90°, the liquid
spontaneously wets the surface (or the channel). If water is used to measure this
angle, such a surface is called hydrophilic. If the angle with water is instead
larger than 90°, the surface is called hydrophobic. The implications of the contact
angle for capillary-driven and digital microfluidic devices are described in the
next two sections.
Capillary-driven microfluidics 2.2.1
When the contact angle is lower than 90°, spontaneous substrate wetting occurs.
As a consequence, when the liquid is confined, it spontaneously flows inside a
closed channel, without the need for an externally-induced pressure. The
tendency of a liquid to spontaneously rise within a channel from a reservoir is
called capillarity. Capillary action is a function of the surface tension, the contact
angle and the radius of the channel: the larger is the surface tension γ and the
smaller are the contact angle θ and the radius r, the stronger this effect is. In
particular, a hydrophilic tube placed horizontally always fills with water. A
vertically placed tube will instead fill until gravitational forces are strong enough
to stop the process, according to the following equation:
ℎ = 2 𝛾 𝑐𝑜𝑠𝜃
𝜌 𝑔 𝑟 , (1)
where ρ is the density and g is the gravitational acceleration. This behavior is
used for the realization of passive microfluidic devices, where the liquid flow is
simply driven by surface chemistry and not by external pumps.
Capillary flow can be induced not only in channels but also in porous
hydrophilic media. Since external pumps are not necessary, such devices are very
suitable for PoC applications. An example of such devices is lateral flow assays
(LFA). LFAs are based on capillary-induced liquid flow through a matrix,
typically made of cellulose [10], [26]. Pregnancy tests are the most widespread
LFS on the market for PoC testing. In pregnancy test strips, a urine sample is
driven through a paper matrix, until the liquid crossing two lines along the flow
direction. These lines are functionalized to provide the readout by color change.
One line is meant as a control, while the second changes color only if a hormone
reacts with the immunoassay [27]. On the other hand, while very simple and
cost-effective, an obvious drawback of purely capillary-driven is that the flow is
continuous and passively controlled only by the design, not allowing for complex
or on-demand operations. Nevertheless, smart designs can still provide
conditional or timed valves, and other functions [28], [29]. In this thesis, the
devices presented in chapters 4 and 6 make use of capillary-driven microfluidics,
both in channels and in porous media, to operate.
13
Electrowetting 2.2.2
Under normal ambient circumstances, the wetting behavior is controlled only by
surface and liquid properties. However, by applying an external electrical field,
the apparent contact angle of a liquid containing charged ions (e.g. tap water) can
be modified. This effect is called electrowetting (EW). The electric field is
typically applied through metallic electrodes, but if a metal is directly positioned
in contact with a liquid and a sufficient voltage is applied, electrochemical (EC)
reactions would start to occur. For example, electrolysis starts to occur already at
1.2 V and other reactions might damage the electrodes at even lower voltages.
This makes the practical voltage window for EW very limited and consequently
makes EW not practically useful to manipulate liquids in devices. To solve this
problem, the metal surfaces can be passivated with an insulator to prevent such
unwanted reactions. This principle is called electrowetting-on-dielectric (EWOD)
and is illustrated in Fig. 3a-b.
Fig. 3. (a) Under normal conditions, a liquid droplet has a set contact angle θ0 on a surface. (b) If
a voltage is applied across the droplet, the internal charges redistribute and result in an apparent
change in contact angle.
Under these conditions, the electrodes are preserved from electrochemical
interactions with the liquid and the change in contact angle is thus reversible. If
the quality of the dielectric materials is ideal, the applicable voltage window is
tremendously extended and limited only by the breakdown of the dielectric, in
the order of 600 V/μm for Si3N4, for example. In the non-ideal case, the situation
slightly complicates and will be discussed in chapter 7.1. Therefore, EWOD can
be used to dynamically modify the wettability of a surface by electrical means.
The theoretical change in contact angle can be calculated combining Lipmann’s
and Young’s equations:
cos(𝜃𝑣)– cos (𝜃0) = 𝜀𝑟 𝜀02 𝛾𝐿𝐺 d
𝑉2 , (2)
where θv and θ0 are the contact angle after and before applying the voltage V,
respectively, d is the dielectric thickness and γLG is the liquid-gas interfacial
energy [30].
What is of particular interest for microfluidic devices is that motion of the
droplet can be induced by asymmetrically applying an electric field to a liquid
droplet using EWOD. The concept is illustrated in Fig. 4a-c.
14
Fig. 4. (a) Initial condition at t0 of a droplet placed between two hydrophobic plates. The top plate
is connected to ground, while the bottom electrodes are initially at 0 V. (b) At the time t’, a voltage
is applied to an electrode, inducing contact angle variation in the neighboring side of the droplet.
The droplet is thus attracted in that direction and starts moving. (c) Motion stops when the droplet
is centered on top of the biased electrode, at t’’. When the voltage is released, the droplet recovers the
initial shape, as in sub-figure (a), while remaining in its new position.
In steady-state (Vapp = 0 V), the droplet stays in its current position (Fig. 4a).
When a voltage is applied between electrodes on its side, the droplet is attracted
by the electrical field (Fig. 4b) and if the resulting force is sufficient (i.e. the field
is strong enough) the droplet moves onto the actuated electrode, minimizing the
system energy (Fig. 4c). The effect of EWOD actuation is maximized when the
initial contact angle is larger than 90° so that a transition between hydrophobic
and hydrophilic behavior is induced when the field is applied.
Coplanar actuation
The most common EWOD actuation configuration for the motion of droplets in
digital microfluidics (DMF) devices is the parallel-plate configuration, as
illustrated in Fig. 4a-c. In this case, electrodes are present both in the substrate
and in a cover plate placed on top of the droplet, parallel to the substrate.
Typically, the substrate has multiple separately addressable electrodes, actuated
to move the droplet, whilst the cover plate features only a single large ground
electrode.
Another possibility to induce motion is to have all active electrodes, negative
and positive, in the same plane embedded in the same substrate. This
configuration is called single-side coplanar [30]. The operation principle of
coplanar actuation is illustrated in Fig. 5.
15
Fig. 5. Coplanar actuation working principle. The field is generated only by a couple of electrodes
embedded in the same substrate (transparency is used to show the electrodes through the dielectric
and hydrophobic top coatings). Droplet motion is induced when a neighboring couple of electrodes
is actuated (e.g. the red electrode couple in the drawing) and stops when the droplet is centered on
these electrodes. Actuation can be achieved, regardless of the presence of a cover plate. Color code:
blue = liquid droplet, dark green = substrate, light and transparent green = dielectric layers, yellow
= disconnected electrodes, orange = actuated electrode couple.
The theoretical contact angle variation in this configuration can be calculated
according to equation 3 [30]:
cos(𝜃𝑣)– cos (𝜃0) = 𝜀𝑟 𝜀02 𝛾𝐿𝐺 d
(𝐴𝑎𝐴𝑡(
𝐴𝑔
𝐴𝑎 + 𝐴𝑔)
2
+ 𝐴𝑔
𝐴𝑡(
𝐴𝑎𝐴𝑔 + 𝐴𝑎
)
2
)𝑉2 , (3)
where Aa, Ag, At are the active electrode, ground electrode, and total electrode
areas, respectively. From this equation, we can deduce that the actuation is
maximum when Aa = Ag = ½ At, and that the actuation, in this case, is half as
efficient as in the case illustrated in Fig. 4 and one fourth as efficient as in the
case illustrated in Fig. 3. This is quite intuitive since the active electrode area is
half overall. Indeed, provided a certain area, the gap area in between the
electrode should be minimized to maximize At. The main advantage of this
configuration is that the cover plate can be any kind of passive hydrophobic
surface, focusing the microfabrication only on the bottom substrate, and the cover
plate can even be omitted if desired since it would act only as a protection from
the environment. This actuation scheme is used to create the devices reported in
chapters 7 and 8.
Digital microfluidics 2.2.3
Commonly, lab-on-chips are characterized by continuous flow in confined
channels. However, through EWOD actuation, single liquid microdroplets can
also be individually controlled. Such handling technology is called digital
microfluidics. EWOD manipulation of microdroplets in digital microfluidics has
been widely used as a low-volume and reconfigurable alternative to conventional
continuous-flow microfluidics based on microchannels [31]–[35]. Since single
microdroplets can be electrically manipulated through electrodes embedded in a
16
substrate, DMF platforms enable versatile and programmable/reprogrammable
operations to be performed, unlike channel-based fluidic chips which are limited
to their specific design and operation scheme. In addition to transport, DMF
allows also relatively complex operations such as microdroplet creation, mixing
and splitting [34], [36]. Moreover, they allow the usage of a very small reagent
quantity, in the range of nL down to pL, potentially enabling cost-effective assays
even with expensive reagents. While biomedical applications are one of the main
fields of application, including immunoassays, and analysis and processing of
proteins and DNA [34], [35], other types of devices have been created. Some
examples of alternative applications include electrical and RF switches [37], [38],
particle collection on filter meshes [39], and optical devices [40], [41].
However, there is an obvious drawback when only using water or solvent-
based microdroplets, which is the fast droplet evaporation and thus limited
device lifetime. To limit this effect, where practically possible, the droplets are
confined in sealed environment or surrounded by immiscible liquids such as oil
for water [32]. However, many applications do not allow the use of these solutions
and alternative strategies need to be adopted to create durable devices that can
operate in “open-air” environments. To avoid this problem, the use of non-volatile
liquids, such as liquid metals and ionic liquids, has been proposed [37], [38], [42]–
[44]. Among them, the use of liquid metals is less desirable. Mercury is
undesirable because of its toxicity, and gallium alloys such as Galinstan are not
durable because of fast surface oxidation [38], [45]. In this thesis, two durable
digital microfluidic devices based on ionic liquid microdroplet manipulation, used
for two different applications, are presented in chapters 7 and 8, respectively.
2.3 Microneedles
Microneedles (MN) are a type of microfluidic platform that can exploit capillarity
during operation and that constitute an essential building block for some of the
devices presented in this thesis. MNs are tiny needles with characteristic
dimensions in the range of hundreds of µm. A first direct implication of their
small size is quite intuitive: unlike the long hypodermic needles connected to
syringes we are all familiar with, used e.g. for venous blood sampling and vaccine
delivery, their insertion is considerably less painful for the patients. In fact,
below a certain length and width, the chances to hit a nerve termination during
insertion are so low that their insertion and removal is likely to be painless [46].
Also after removal typically neither pain nor soreness is perceived. A second
interesting implication of the shorter penetration length is that they allow
accessing a different location: the skin. The average depth of the dermis on the
forearm is approximately 1 mm, so that microneedles operate in a different
region with respect to standard needles that are inserted subcutaneously or
intramuscularly. A third implication is that, because of their reduced
invasiveness, the use of MNs has the potential to reduce a series of common side-
effects and risk factors, such as local trauma, inflammation, and infections,
because of the reduced skin damage compared to their larger counterparts.
Finally, a fourth implication of the size reduction is the potential reduction of
discomfort, anxiety and thus increase of patient compliance: needle phobia is
17
considered to be a significant problem in clinical practice, with more than half of
adults showing aversion and stress related to needle insertion, and up to 70% in
case of pediatric patients [47].
Due to the listed reasons, microneedles are gaining growing attention in both
cosmetics (e.g. microneedling, using Dermaroller® and Dermapen®) [48], [49]
and medicine. Focusing on medical applications, they have been used for drug
delivery and sensing [46]. As of today, the most explored use case for
microneedles is drug delivery [50]–[52].
Microneedles types can be divided into solid, dissolving/swellable and hollow
[53], as illustrated in Fig. 6a-c, respectively. Solid MNs are filled pins or pyramid-
like structures, which are used to prick the skin or coated with functional
materials to perform either sensing or sampling. Dissolving MNs are a sub-type
of solid MNs which are chemically designed either to be degraded and absorbed
by the skin after insertion, for drug delivery applications, or to absorb liquid and
compounds, for sampling applications. They are typically made of hydrogels or
other polymers. Hollow MNs differ from solid MNs by the presence of a lumen
along their length connecting the tip area to the opposite end, as in standard
hypodermic needles (Fig. 6d). Hollow MNs can be considered microfluidic chips
since their lumen is typically smaller than 200 μm in diameter. As a consequence,
if their lumen walls are hydrophilic, spontaneous filling occurs if their opening is
placed in contact with a liquid reservoir. Hollow MN-based devices using these
features are reported in chapters 4, 5 and 6.
Fig. 6. (a-c) Cross-section illustration of solid (a), dissolvable (b) and hollow (c) MNs, respectively,
inserted in the skin. (d) 3D illustration and cross-section of an example of a hollow microneedle.
Microneedles are typically made of one of these four material categories: metals,
silicon, glass, and polymers (various plastics, hydrogels, SU-8, etc.) [54].
Depending on the application and the chosen MN material, different fabrication
methods are involved [54]. Metallic MNs can be obtained by shortening or cutting
standard hypodermic needles [55]–[57]. Silicon MNs are processed using
standard micromachining tools, such as dry (RIE and DRIE) and wet etching [52],
[58]–[61]. Glass MNs are often obtained by pulling glass capillary tubes and are
currently not viable for large-scale production [62], [63]. Polymeric needles can be
manufactured using a series of techniques, including 3D printing [64]–[66] and
replica or injection molding [67]–[70].
18
2.4 Electrochemical sensing
EC sensors are another key component of two of the realized devices, presented
in chapters 4 and 7, respectively. As of today, electrochemical sensors are the
largest group of chemical sensors. In many cases, they have reached commercial
maturity over a decade ago and are well established in clinical and analytical
laboratory practice, for example for measuring pH, oxygen, glucose, cholesterol,
and lactate [71], [72].
Electrochemistry involves the transfer of charges from electrodes to a liquid or
solid material through a redox chemical reaction. By controlling the chemical
reactions at the electrodes, it is possible to use the charge transfer process as a
sensing mechanism for one of the reactants. Electrochemical sensing cells are
formed by a minimum of two electrodes, but typically three. In such a cell, a
working electrode (WE, or sensing electrode) and a counter electrode (CE, or
auxiliary electrode) are always present and immersed in an electrolyte, which can
be a solid or a liquid, to create a closed electrical circuit. The electrochemical
reactions can be spontaneous, i.e. the reaction and charge transfer is driven only
by the electro-catalytic surface properties of the WE in combination to one or
more specific chemical components in the surrounding electrolyte, as in galvanic
cells (i.e. batteries). Most often, however, the desired redox reaction needs
additional energy to take place (electrolytic cell). This is provided by shifting the
electrical potential of the electrode, with respect to the electrolyte, to the level
required to induce the non-spontaneous redox reaction. To do this, a specific
voltage difference (or bias) needs to be externally applied. However, the potential
in the electrolyte is unknown and cannot be directly measured. Only the
difference in potential between two immersed electrodes can be measured. The
potential of an electrode in a solution, called half-cell potential E, is described by
the Nernst equation [73]:
𝐸 = 𝐸0 +
𝑅𝑇
𝑛𝐹 ln
[𝑜𝑥]
[𝑟𝑒𝑑], (4)
where R is the gas constant, T is the temperature, F is the Faraday constant, n is
the number of electrons involved, [ox] and [red] are the concentrations of the
oxidized and reduced species, respectively, and E0 is the standard potential. E0 is
a relative value, chosen in relation to another arbitrary electrode reaction, which
is typically a normal hydrogen electrode (NHE) [73], [74]. By knowing the
potential of both half-cell reactions, the potential difference between the two
electrodes can be obtained. On the other hand, in most of the sensing applications,
the conditions within the solution change over time and not all processes are
constant or controllable at the WE and CE. For this reason, a third electrode
(reference electrode) is commonly present in the electrochemical cell to provide a
known potential reference to the system. A standard electrochemical cell is
illustrated in Fig. 7a and, schematically, in Fig. 7c.
19
Fig. 7. (a) Illustration of a standard electrochemical three-electrode cell. (b) Illustration of a
common macroscopic Ag/AgCl reference electrode, in which a chlorinated silver wire is kept in a
saturated KCl solution. (c) Schematic representation of the circuit. (d) Illustration of a cell in
which a Q-RE is used, as demanded by miniaturized sensors: the Q-RE is immersed directly in the
same electrolyte as WE and CE.
Reference and pseudo-reference electrodes 2.4.1
A reference electrode (RE) is an electrode that has a well-characterized and
stable redox reaction within a certain electrolyte, also during operation, since
only small current flows through it so that its stability is not compromised. The
RE’s role is to provide a stable potential reference to guarantee a constant and
specific bias to the WE. Standard REs are formed by an electrode immersed into
an electrolyte in a glass vial that is connected to the main electrolytic solution
through a porous junction, as illustrated in Fig. 7b. This allows the RE to be
electrically connected to the cell but still operate in a controlled solution that
remains constant and, thus, it remains at a fixed known potential value, as
described in Eqn. (4). Standard REs are, for example, NHE, Ag/AgCl and the
Saturated Calomel Electrode (SCE) [73].
While their stability is excellent, in miniaturized systems it is usually
impossible to use such an approach, i.e. the reference electrode cannot be
integrated into the system together with a separate vial and electrolyte. In this
situation, a pseudo-RE (also called quasi-RE, Q-RE) is used instead (Fig. 7d). The
Q-RE is an electrode directly immersed in the measurement solution, serving the
same function in the cell as a normal RE. Q-REs only work if the measurement
solution allows the desired reaction to happen and if this reaction is not
dynamically influenced by other factors, thereby providing stable reference
overtime. In other words, if e.g. the Q-RE is sensitive to the concentration of a
substance and to the pH, it behaves as desired only if both such concentration
and pH are sufficiently constant over time in the measurement solution. This
20
creates challenges for the stability of the sensing microsystems, additional to the
ones already caused by electrode miniaturization itself, since the RE stability is
proportional to its size, in a first approximation [71].
Electrochemical sensing and functionalization methods 2.4.2
Once a reference potential value is set and can be considered constant over time
during operation, external circuitry is used to monitor the potential and the
current flowing through the electrochemical cell. This action is performed by a
potentiostat, a tool able to simultaneously measure the current or voltage
conditions of the cell and, if necessary, to apply a fixed or dynamic voltage bias to
the WE against the RE (or Q-RE). To detect the presence of substances in an
electrolytic solution or to characterize and functionalize the electrodes
themselves, different strategies and measurement principles can be used. In this
work, amperometry, open circuit potential, and cyclic voltammetry have been
used. Amperometry consists in measuring the generated current when a fixed
bias is applied. Open circuit potential (OCP) monitors the voltage between two
electrodes without any applied bias. Cyclic voltammetry (CV) is based on the
measurement of the current while a dynamic bias, sweeping back and forth
between two voltage values, is applied. Other electrochemical measurement
principles not used in this work include e.g. pulse voltammetry, chrono-
potentiometry, and coulometry [74].
Enzymatic biosensors 2.4.3
The electrochemical sensors used in this work fall under the class of enzymatic
biosensors. Biosensors, by definition, use a biological recognition element to
achieve selective sensing of a specific chemical. In this case, such an element is
an enzyme. The WE is functionalized with such enzyme, which has the role of
catalyzing the decomposition of the target molecule into an electroactive molecule.
This secondary molecule is then electrochemically oxidized or reduced on the
surface of the WE itself when a sufficient potential is applied. In this way, an
electrical current is generated and it can be measured by amperometry. Unless
the reaction is limited or influenced by other factors, the resulting current is
proportional to the initial concentration of the target molecule. Moreover, to
obtain a readout that is proportional only to the target molecule, such a
secondary molecule should also be non-endogenous or, at least, have constant
local concentration over time.
Five fundamental parameters characterize this type of amperometric sensors:
sensitivity, selectivity, linearity, stability and response time.
- The sensitivity is the ratio between the sensor response (change in current
or voltage) for a certain change in the concentration of the target analyte.
The larger the sensitivity, the larger the readout signal is.
- Selectivity is the property of only sensing the targeted analyte. Depending
on the electrolytic environment, other endogenous molecules might be
electroactive and undergo oxidation/reduction at the WE at the applied
potential. Such molecules are called interfering species and, if their
concentration is too high or dynamic, their contribution to the readout
21
current might hide the signal variations caused by the targeted molecule.
For a correct readout, it is important to prevent the reaction of such
interfering species at the WE site.
- Linearity is the property of having proportionality between input and
readout. If the enzymatic reaction is dependent on more reactants than the
target molecule alone, an excess of all other reactants must be guaranteed
at all times, to avoid early saturation of the reaction.
- The stability is the ability to maintain the response constant over time for
the same analyte concentration. In EC sensors, this is largely dependent
on the RE.
- The response time is the time required by the sensor to reach a stable
output value, given a change in the targeted analyte.
Problems in miniaturized gas sensors 2.4.4
Electrochemical sensors can also be used to monitor gases for medical
applications. Nitric oxide (NO) sensors are an example that falls under this
category. NO is used as a marker for asthma and it can be monitored in the
breath. Therefore, there is a commercial interest to integrate such sensors into
portable handheld meters and, of course, at an accessible cost. Fig. 8a shows the
dimension of an exemplifying commercial device, whose portability is limited by
the necessity to buffer the exhaled breath, to compensate for the long response
time of the sensor [75], [76]. Previous work showed that miniaturized sensors can
potentially improve response time and reduce the cost and complexity of the
system, potentially allowing these detectors to really become portable [77]. Fig.
8b illustrates the cross-section of an electrochemical gas sensor of this type: the
sensing compartment is filled with a liquid electrolyte and is open to the
environment to let the gas molecules diffuse in and out.
Fig. 8 (a) Illustration of a commercial handheld asthma breathe analyzer. The size is large
because of technical limitations in the sensing system used. (b) Illustration of the cross-section of a
miniaturized electrochemical gas sensor.[77] The sensing chamber filled with the liquid electrolyte
is open to the environment to allow gas diffusion into the sensing area.
22
However, a major problem reported for this type of miniaturized sensors is the
short lifetime, mainly due to the evaporation of the liquid electrolyte from the
open area. Another problem of miniaturized electrochemical sensors is that the
stability typically relates to the Q-RE size and thus small sensors are prone to
suffer from significant response drift over time. If the sensor is continuously
exposed to the environment, recalibration cannot be performed. In fact, the
insulation of the sensing chamber from the dynamic environment has to be
ensured for the duration of the recalibration procedure. In this thesis, a solution
to enhance the lifetime of electrochemical gas sensors is presented in chapter 7.
23
Chapter 3
Thesis objectives
To tackle some specific aspects of the problems presented in the first two
chapters, on both the technological and the medical side, this thesis work aimed
to:
1. Demonstrate that it is feasible to miniaturize a glucose sensor to be
inserted within the dermis while preserving sufficient performance. The
required size reduction is one order of magnitude compared to the smallest
state-of-the-art glucose sensor, i.e. from 0.35 mm2 to 0.04 mm2. (Paper I)
2. Demonstrate that the combination of such a miniaturized sensing probe
with a hollow microneedle can be used in humans and can provide real-
time data on the glycemic levels in vivo. (Paper II)
3. Demonstrate that such a system can potentially be manufactured on a
large scale by developing a method to assemble and electrically contact the
fragile sensing microprobes into microneedles. (Paper III)
4. Demonstrate that a known volume of ISF can be sampled and stored from
the human skin using a simple and compact microfluidic device. (Paper IV)
5. Demonstrate that quantitative data on the concentration of bodily analytes
can be obtained from a metered amount of sampled ISF stored in such a
device. (Paper IV)
6. Create a device to be integrated with electrochemical gas sensors able to
extend their lifetime, by enabling recalibration and limiting electrolyte
evaporation when not in use. (Paper V)
7. Demonstrate that the use of ILs, combined with EWOD actuation, enables
new “open-air” digital microfluidic devices for diverse applications.
(Papers V and VI)
8. Demonstrate the versatility of such a microfluidic platform, realizing an
optical shutter with improved performance compared to state-of-the-art
liquid-based optical shutters, and in particular featuring zero insertion
loss in the open state and broad-band operation. (Paper VI)
24
25
Chapter 4
Continuous glucose monitoring systems
This chapter presents the development of a miniaturized real-time CGMS. It
starts with a brief overview of the CGMS landscape. This overview includes the
working principles behind glucose sensing and already marketed CGM devices,
as well as other alternative technologies with a potential for marketability when
combined with microneedles. Then, the rest of the chapter focuses on the design
and the characterization of an ultra-miniaturized glucose sensor and, finally, on
the combination of such microsensor with a microneedle chip to perform CGM in
dermal ISF in vivo, minimally invasively and in real time.
4.1 Sensing technologies
Even if competitive alternatives are emerging, [78], [79] electrochemical (EC)
sensing is the most common strategy for continuous glucose monitoring, as well
as for standard detection in blood [80]–[82].
Electrochemical glucose sensing 4.1.1
Glucose sensing can be performed electrochemically using amperometry [81], [83].
Recognition of glucose is provided by an enzyme, glucose oxidase (GOx), which
transforms glucose into hydrogen peroxide (H2O2) according to the following
reaction:
𝐺𝑙𝑢𝑐𝑜𝑠𝑒 + 𝑂2 𝐺𝑂𝑥→ 𝐺𝑙𝑢𝑐𝑜𝑛𝑖𝑐 𝑎𝑐𝑖𝑑 + 𝐻2𝑂2. (5)
The produced H2O2 is then anodically oxidized at the surface of the WE, when
this is biased at +0.6 V:
𝐻2𝑂2 +0.6 𝑉→ 𝑂2 + 2𝐻
+ + 2𝑒−. (6)
A potentiostat maintains the WE bias constant and measures the generated
current. The circuit is closed by the CE reaction, balancing the current as follows:
2𝐻+ + 1
2𝑂2 + 2𝑒
− → 𝐻2𝑂. (7)
In this thesis, if not specified otherwise, EC potentials use Ag/AgCl as the reference value.
26
In the ideal case, the generated current is proportional to the initial
concentration of glucose. However, there are two important aspects to consider
while designing a glucose sensor of this type, especially for in-vivo applications
[84].
In the reaction reported in Eqn. (5), H2O2 production is dependent not only on
glucose but also on oxygen. This means that an excess of oxygen with respect to
glucose must be ensured throughout the measurement for correct operation. If
this condition is not satisfied, the response to glucose saturates because the
reaction becomes oxygen-limited and linearity is lost.
Moreover, in the reaction in Eqn. (6), a bias of +0.6 V is required for the
reaction to happen. At this potential, other endogenous bodily substances such as
uric acid and ascorbic acid are electroactive. Therefore, if they can reach the
electrode, they would undesirably be oxidized and would thus contribute to the
generated current causing a false glucose reading. Additionally, some other
exogenous compounds such as paracetamol, a commonly used medical drug, are
known to alter glucose readings.
To guarantee stability and allow miniaturization, a Q-RE needs to be
integrated on the same sensor as the WE and CE. This represents a key
challenge for this type of miniaturized biosensors. A few different Q-RE materials
have been proposed, with Ag/AgCl and IrOx being the best candidates for in-vivo
operation [83], [85]–[87]. Most commercial devices use Ag/AgCl as Q-RE, because
of its well-known behavior and reliability in macroscopic situations [88]. However,
extreme miniaturization of Ag/AgCl electrodes is problematic because of the
limited lifetime of microfabricated chlorinated silver electrodes. The reaction
involves the dissolution of the chlorinated layer, which consist of few hundreds
nanometers in thin-film electrodes, and when the process is completed the
potential drastically changes [89]. Additionally, the AgCl solute is toxic and
creates concerns in terms of inflammatory response. Specific coatings have been
proposed to reduce and slow down this process [90], [91]. IrOx, as Q-RE, is
instead limited by the strong dependence of its potential on pH (between -64 and
-68 mV/pH) [83], [92]. However, in vivo, skin ISF is buffered and regulated at pH
7.1 ± 0.3, making this issue negligible [93]. Moreover, IrOx shows good
biocompatibility, mechanical stability, and limited long-term potential drift.
Compared to Ag/AgCl, it also shows lower sensitivity to electroactive species
present in ISF. Finally, it can be deposited using standard microfabrication
methods in sub-100-nm thin-film electrodes, followed by a one-step
electrochemical activation. Alternatively, IrOx can also be directly deposited both
by using cleanroom technologies and by electro- or electroless deposition [86], [94],
[95].
The sensing method described above, based on reactions (5) and (6), is defined
as first-generation glucose sensing [96]. To facilitate the reaction, mediators such
as ferrocene can be used [97], [98]. This principle is called second-generation. In
this case, the mediator is reduced during the GOx reaction and then oxidized at
the electrode surface. Moreover, if a direct electron transfer pathway between the
enzyme and the electrode is created, a third-generation glucose sensor is realized
[99]. These advanced sensing mechanisms are less susceptible to interfering
27
species, since they can operate at a lower voltage, as well as to a condition of
oxygen deficiency.
State of the art
Historically and commercially, electrochemical enzymatic sensing is the most
common method to sense glucose. The first EC enzymatic glucose sensor dates
back in 1962 and was introduced by Clark and Lyons. Standard handheld glucose
meters use sensing strips containing an electrochemical sensor to measure the
glucose concentration in capillary blood [80], [84]. Also most CGM systems use
the same principle to measure glucose in ISF [88], [99]. The size of these
transdermal strips exceeds 5 x 0.6 mm2. Smaller EC sensors have been realized,
but always exceeding 3 mm in length and > 0.4 mm2 overall footprint [83], [89],
[100]. IrOx was chosen as the Q-RE material in some of these miniaturized
sensors [83], [86], [89].
More recently, several studies reported glucose sensing using non-enzymatic
EC sensors [101]. In this case, glucose is directly oxidized at the electrode surface,
without using any enzyme. This can be achieved by using specific electrode
materials, chemistries, and nano-structuring [96]. The main limitations,
especially for in-vivo use, are related to the difficulty to achieve competitive
selectivity, non-toxicity, fast kinetics, surface anti-fouling and operation at ISF
pH.
Other sensing principles 4.1.2
Alternative strategies for glucose sensing have been proposed, including the use
of functionalized field-effect transistors, impedance spectroscopy, and optical
techniques [78].
The ISFET principle is based on a MOSFET structure, in which the gate is
coated with an ion-sensitive membrane. When the pH varies, the FET
conductance is modified [102]. The use of ISFETs for glucose sensing was
proposed long ago, e.g. using GOx as a recognition element to generate pH
variations to be detected by the FET structure [103], [104]. However, progress in
miniaturization and stabilization to create valid commercial products has been
limited.
Impedance spectroscopy is an interesting strategy because it potentially
allows non-invasive monitoring of glucose from outside the skin [105]. However,
the measured impedance variations are due to electrolyte changes in the tissue,
which can be related to hypo- or hyperglycemia, and not directly due to glucose
concentration variations. Consequently, the selectivity is not ideal, since other
competing factors might influence this type of readout.
Finally, other strategies are based on optical measurement techniques,
suitable for both non-invasive monitoring and implantable devices, including
optical coherence tomography (OCT), IR spectroscopy, Raman spectroscopy, and
fluorescence intensity reading [106]. Among them, at the moment, the use of
fluorescence is the most promising alternative. Often, a fluorescent label is bound
to GOx and uses energy transfer from the reaction with glucose, or it exploits
either local oxygen consumption or hydrogen peroxide production to generate the
28
optical signal. Alternatively, other chemistries can be used to interact with
glucose and generate a variable fluorescent signal [107]. The excitation can be
performed with an LED and the readout is typically performed using a
photodiode.
4.2 Commercial state-of-the-art of CGMS
To perform CGM, there are commercially available solutions on the market.
These solutions are based on two main sensing methods: electrochemical
transcutaneous sensors and implantable fluorescence-based optical sensors. In
both cases, sensors are inserted (totally or partially, as illustrated in Fig. 9a-b,
respectively) through the skin to measure glucose concentration in ISF over time,
and have an electronic circuit for readout and data transmission placed outside
the skin, on an adhesive patch. Such a circuit communicates wirelessly, via e.g.
Bluetooth, with an external reader that can be standalone, a smartphone or a
smartwatch (Fig. 9c).
Fig. 9. (a) Cross-section illustration of a transcutaneous sensor, as in most commercial
electrochemical sensors. (b) Cross-section illustration of a fully-implanted sensor, where no hard-
wire is present with the external electronics. In both (a) and (b) the power supply and the basic
readout are performed by the electronics integrated on the adhesive patch outside of the skin. These
drawings are not necessarily in scale. (c) The electronic patch communicates wirelessly with an
external reader providing a user interface, data processing, and long-term data storage.
Commercially, the first and currently most used solution is electrochemical
sensing. Large biomedical companies as Abbott, Medtronic, and Dexcom have
developed electrochemical transcutaneous sensors, as illustrated in Fig. 9a.
These products are based on flexible sensing probes protruding perpendicularly
from an adhesive patch. The adhesive patch has a double function: (i) it keeps the
29
sensor in position and protect it after its insertion, and (ii) it hosts the electronics
to supply power, perform the readout, and communicate with the external reader.
The sensing probes are pushed through the skin using either a syringe-like or a
spring-loaded inserter. In both cases, needles longer than 6 mm and wider than
600 μm are used for the procedure [108]–[110].
The latter principle, fluorescence, was instead more recently commercially
implemented by Senseonics™, with Eversense® [111]. What is also varying in the
practical commercial implementation between these two sensing strategies is the
insertion and operative location. In this case, the sensor is implanted under the
skin. In terms of user experience, the claimed advantages are a longer duration
(90 days against 14 days at most for the former products), reduction of the
patients’ worries about losing or damaging a sensor or the patch tape, and the
separation between the readout and the sensor itself, as illustrated in Fig. 9b,
meaning that the electronic patch can also be removed and recharged. The
drawbacks are, at the same time, also related to the sensor implantation:
patients cannot self-perform the insertion in this case and need to visit a clinic
every three months and undergo a small surgical procedure. This is a significant
complication for wide-spread adoption, both in terms of cost and convenience. The
size of such a sensor is also significant and is far from being considered as
minimally invasive: 18.3 mm by 3.5 mm.
Another previously proposed strategy for measuring glucose, non-invasively,
was to use reverse iontophoresis to extract glucose molecules from the skin. It is
based on the creation of a low electrical current through the skin, which drives
migration of charged ions (e.g. Na+ and Cl-) which in turn creates an electro-
osmotic flow transporting glucose molecules towards the skin surface [112], [113].
This principle was brought to the market with GlucoWatch® (Cygnus Inc.).
However, there were two fundamental problems in using this technique: the
extracted glucose was in the range of 0.1% of that in blood, posing challenges in
the sensor sensitivity requirements, and the creation of skin irritation, redness,
and itching, which led to its discontinuation. [114]
4.3 Microneedle-based CGM solutions
A promising solution to improve CGMS is the use of microneedle-based platforms.
As mentioned in chapter 2.3, microneedles offer a series of advantages with
respect to larger implanted or transcutaneous technologies. Microneedles are
gaining attention as they offer virtually painless insertion and reduce patient
discomfort, potentially resulting in better compliance and satisfaction.
Additionally, there are also specific advantages when using them for CGM:
because of the size, they allow access to ISF in the dermis and their insertion
results in reduced damage to the skin. Commercial alternatives currently
measure glucose in the subcutaneous tissue, which is not the ideal location for
reliable and real-time glycemia monitoring, due to the local heterogeneity and
limited distribution of capillaries [9]. ISF is more abundant in the dermis and the
glucose concentration dynamics there is closer to that in capillary blood [115],
[116]. Moreover, microneedle insertion results in reduced tissue trauma during
insertion and post-insertion inflammation, [46] which are responsible for an
30
increased foreign body reaction, adversely affecting long-term monitoring [117],
[118]. Finally, the use of microneedle platforms has been demonstrated for
minimally-invasive delivery of drugs, including insulin, [63], [119]–[121]
potentially enabling the two functions to be integrated into the same system.
However, currently proposed solutions suffer from some technical and design
issues. Two main implementation strategies have been shown for CGM, using
hollow and solid MNs, respectively. Hollow MNs systems had issues in
performing cost-effective and real-time monitoring. Most of these systems were
based on the extraction of ISF via the MNs lumens, to a sensing compartment
placed on the MNs’ backside. This approach required complex active extraction
mechanisms based on vacuum suction, microfluidic pumps, and valves, or device
pre-filling with liquid [116], [122]–[129]. Moreover, if diffusion is required
through the entire length of the MN to reach the sensing compartment,
significant measurement delays are introduced [46], [130], [131]. For example, to
travel 1 mm in ISF a glucose molecule would take in the order of 20 to 25
minutes. This delay, combined with the physiological delay blood-to-ISF (~4-12
minutes), impedes real-time monitoring and thus precludes the use of such
CGMS to make medical decisions, such as insulin dispensing. Finally, if the MN
design is not optimized, there is a significant risk of clogging during insertion, for
“volcano”-like MNs, or unreliable insertion, for non-sharp MNs [132].
Coated solid microneedle arrays are another viable alternative, which does
not suffer from those issues related to fluid extraction, delays, and clogging [67],
[133], [134]. The main drawback related to these systems is related to the non-
planarity of the sensing area. As a consequence, the electrode microfabrication
and functionalization procedures are more complex. The choice of functional
materials for the coating is limited by technical factors, including the necessity of
electrodepositing the layers, to achieve conformal electrode functionalization, and
the need to guarantee robustness upon intradermal insertion, which is not
straightforward for the typically used polymeric and enzymatic functional layers.
Finally, the sensing electrodes cover a large surface and this might result in
larger costs with respect to miniaturized planar sensors.
Some of the state-of-the-art designs are illustrated in Fig. 10a-d, to portrait
features and issues associated with the different implementations. A possible
solution to combine the advantages of the introduced technologies, minimizing
some of their drawbacks, is presented in the next section.
31
Fig. 10. Simplified illustrations of examples of CGMS implementations based on MNs. (a) A pre-
filled sensing compartment communicates with the dermal space through the MNs’ lumens.
Glucose molecules need to travel purely by diffusion to reach the electrodes [130]. (b) Also in this
case, the sensor is on the backside of the MNs and the system requires a relatively complex
microfluidic scheme to operate [128]. (c) Solid microneedle array, where the WE is coated with an
enzymatic layer [135]. (d) Combined solid-hollow MN solution, where the MNs are coated similarly
to sub-figure c, and then assembled in a hollow MN array to facilitate insertion [134].
4.4 A combined miniaturized solution
The aim is to perform minimally invasive, reliable and minimally delayed CGM
in the dermal interstitial fluid, with an easy to fabricate system and a simple
plug-and-measure approach. The proposed solution takes inspiration from the
respective advantages of the previously presented sensors: commercial devices
measuring in-situ and microneedle-based devices measuring from the dermis in a
minimally invasive manner.
Commercial CGMS operate inside the body by placing the sensor in direct
contact with ISF. Moreover, after insertion, no actuator is needed for the sensor
to operate, preserving a certain overall simplicity. Finally, sensors and patches
are typically fabricated separately, without complicating the cross-compatibility
of the manufacturing process. On the other hand, it would be preferable to
minimize the overall size of the transdermal sensor, which is also forcing to
measure glucose in the subcutis, as a consequence.
Microneedle-based solutions instead offer minimally invasive monitoring and
painless insertion, with minimal tissue trauma and inflammation. Moreover,
they offer access to the ISF in the dermis, which is a better measurement location
than the subcutis. On the other hand, it would be preferable to avoid active
and/or continuous ISF extraction, as well as the time delay due to molecular
diffusion, in the case of hollow microneedles, and the complex functional coating
of 3D structures, in the case of solid microneedles.
32
In this thesis, a CGMS in which the sensing is performed in situ from within
the lumen of a microneedle is proposed. A single ultra-miniaturized
electrochemical sensor is inserted in a hollow microneedle. When the device is
inserted into the skin, the microneedle lumen fills with ISF by capillary action
through the opening, without the need for external actuation mechanisms.
Molecular exchange is fast since it occurs on a minimal distance because the
sensing electrodes are directly facing the side-opening of the microneedle. The
entire concept is illustrated in Fig. 11a-c.
Fig. 11. Overview of the envisioned concept. (a) Eventually, the device is designed to be used as a
patch to be placed on the arm. (b) A microneedle is inserted intra-dermally, with a miniaturized
sensor fitting in its lumen. (c) The microneedle protects the sensor during insertion and operation.
The electrodes on the microsensor face the side opening of the microneedle.
Indeed, performance, adaptability to large-scale manufacturing (see chapter 5)
and sensor cost have to be in line with currently available commercial solutions,
while being less invasive for this concept to be of real potential impact.
4.5 Miniaturization of a glucose sensor
The first challenge to face in the realization of such envisioned CGMS is the
following: is it possible to miniaturize a glucose sensor enough to fit within the
lumen of a microneedle and thus in the dermal region, while preserving
performances in line with ISO guidelines? Indeed, the miniaturization of this
type of sensors intrinsically entails challenges not only in the manufacturing
process but also in signal amplitude, stability, and choice of active materials.
Sensor design and functionalization 4.5.1
As a glucose sensing mechanism, enzymatic electrochemical sensing has been
identified as the most promising strategy for miniaturization, integration
simplicity and cost containment. Optical sensors are currently impossible to
miniaturize and integrate to the same extent and, as mentioned earlier, the other
strategies still show significant problems or drawbacks.
33
The targeted sensor active area, i.e. the maximum area to fit on a microprobe
that in turn fits inside a microneedle lumen, is approximately 75 x 700 μm2. To
provide a visual idea of the targeted dimensions, the sensor must fit on the tip of
a human hair. The comparison in size between the realized microsensor and
commercial sensors is illustrated in Fig. 12a-b. The achieved reduction in active
surface area of the sensor is more than 50 fold, while the overall volumetric size
reduction is of more than 200 fold compared to commercial solutions. Also
compared to the smallest glucose sensors reported in the literature,
miniaturization of one order of magnitude or more was required [83], [89], [100].
Fig. 12. (a) Illustration of skin cross-section comparing, in scale, the realized sensor to the latest
implantable and transcutaneous commercial sensors inserted therein. (b) SEM micrograph
showing a commercial transcutaneous sensor and the realized sensor side-to-side.
On this small microprobe surface, all the three electrodes (WE, CE, Q-RE)
forming an electrochemical cell are present. The realized overall shape of the
sensor chip resembles a “T” (hereafter called “T”-shape) being composed by said
microprobe featuring the sensing electrodes and a wider top part featuring
electrical contacts (Fig. 13b).
WE and CE are made of platinum. For the Q-RE, due to the reasons listed in
section 4.1, IrOx was chosen. In the fully characterized design (Fig. 13a), each
electrode has a surface area of 0.012 mm2 (170×70 μm2), with an inter-electrode
distance of 10 μm, to maximize the active surface and minimize ohmic losses.
Initial studies were performed on large flat silicon chips (Fig. 13a), while at a
later stage the substrate was machined in the form of a 65-μm-thick microprobe
(Fig. 13b). Another electrode configuration was tested, featuring long parallel
electrodes, as the one shown in Fig. 13b-c; in this case, the electrodes are shown
on a silicon microprobe. This second design might be preferable in terms of in-
vivo sensing because it is exposed to the same conditions once in the microneedle,
i.e. all electrodes have the same distance from the opening towards the skin.
However, as shown in Fig. 13c-d, this second electrode configuration at this scale
does not geometrically allow for separate functional coatings by, for example,
normal drop-casting or liquid dispensing methods.
34
Fig. 13. (a) SEM micrograph of a planar three-electrode sensor before functionalization. The
footprint area is only 530 x 70 μm2 (0.037 mm2). (b) SEM micrograph of a sensor on a microprobe,
with a size fitting within a microneedle lumen. (c) Picture of a functionalized microsensor with
“parallel” electrodes. (d) Picture of a microsensor with “serial” electrodes. In this case, separate
functionalization of the different electrodes can be achieved with a precision dispenser.
The electrode functionalization is a fundamental part of the process. To
operate, the sensor needs to be functionalized. There are a few important
functions, such as glucose recognition, selectivity, and linearity that need to be
guaranteed by the sensor. These functions are provided by three different coating
layers, as illustrated in Fig. 14. First of all, the sensor is coated with GOx (in a
crosslinked matrix of bovine serum albumin and glutaraldehyde) to enable the
reaction in Eqn. (5) and generate H2O2. Then, to guarantee excess of oxygen and
extend the linearity range, a polyurethane layer is used as diffusion barrier for
glucose. Finally, a Nafion perm-selective coating rejects common interfering
species such as uric acid and ascorbic acid because of charge exclusion.
Fig. 14. Illustration of the cross-section of the platinum WE functionalized to achieve
measurement sensitivity, selectivity and linearity. The combination of the three coating layers (GOx,
polyurethane, and Nafion) provides the desired functionality to the sensor.
35
Sensor performance in vitro 4.5.2
The microsensor was initially tested in vitro, under relevant conditions
mimicking important aspects of in-vivo operation in ISF. Measurements were
performed in a stirred phosphate buffered saline (PBS) solution (0.01 M, pH 7.4).
The four main parameters to be evaluated to demonstrate the reliability of a
biosensor are: stability, sensitivity, selectivity, and linearity.
To study the stability, the open circuit potential of the integrated IrOx Q-RE
was measured over four days, versus a macroscopic commercial Ag/AgCl RE. The
result is shown in Fig. 15a. The potential remained very stable, within ± 10 mV,
which is satisfactory and in line with larger sensors in previous works [89], [136].
Small transitory peaks are visible once a day and are due to PBS re-fill
operations to compensate for the liquid evaporation over time.
Once achieved satisfactory stability, the sensing response of the three-
electrode microsensor was characterized. A cell bias of +0.5 V was applied with a
commercial potentiostat. Aliquots of interfering species (uric acid and ascorbic
acid) and glucose (at increasing concentration in steps of +50 mg/dL) were added,
while amperometrically measuring the response over time. Fig. 15b shows the
overall measured current over eight hours, with a glucose concentration of up to
500 mg/dL, which represents the full range of interest. Standard blood glucose
meters typically guarantee operation in the range of 20-500 mg/dL. Fig. 15d
shows the same amperometric data, but as a function of the glucose concentration,
highlighting the linearity range, up to 200 mg/dL. The linearity range can be
extended, by sacrificing the sensitivity, and vice versa.
Finally, Fig. 15c shows the results of a separate measurement, in which the
current generated by glucose at standard physiological levels in healthy
individuals (100 mg/dL), and by interfering species at both high physiological and
toxic concentrations is compared, proving selectivity. ISO guidelines require ± 20%
resolution (above 100 mg/dL glucose concentration), because such an error is
unlikely to lead to a wrong medical decision. The cross-sensitivity, at therapeutic
levels, introduced an error < 2.5%, while at high toxic concentrations, the
maximum error was 8.0%, proving selectivity even under extreme testing
conditions. Another source of errors is noise, which can be averaged out and
would otherwise affect by ± 5% the readout. Repeatability and reproducibility are
the remaining sources of errors if regular recalibration is not performed.
Therefore, the realized device provides a proof-of-concept that shows the
possibility to miniaturize a glucose sensor to the extent required to fit within the
dermal layer of the skin (footprint < 0.04 mm2), while preserving functionalities
in line with ISO guidelines.
36
Fig. 15. (a) OCP measurement of the integrated IrOx Q-RE vs. a standard Ag/AgCL RE. (b)
Chrono-amperometry of the microsensor response to glucose and interfering species. (c)
Contribution of the different electroactive substances to the overall current, showing glucose
selectivity even at a toxic concentration of interfering species. The Nafion layer rejects charged
species such as uric acid (UA) and ascorbic acid (AA). (d) Generated current vs. glucose
concentration over the entire range of interest. The polyurethane coating enhances linearity at the
expense of sensitivity.
37
4.6 A real-time minimally-invasive CGMS
The second query towards the realization of such MN-based CGMS is whether
the realized sensor can be combined with a microneedle, to support insertion and
protect the sensing microprobe, and perform minimally-invasive real-time
glucose monitoring in the skin in vivo. Here, the focus is on sensing properties
and skin interfacing. The manufacturing and integration aspects are analyzed
and presented separately, in chapter 5.
Microneedle and CGMS design and realization 4.6.1
The CGMS is composed of two independently fabricated components: the sensing
microprobe (described in section 4.5) and a microneedle chip. The latter is based
on a 2 x 4 mm substrate with an out-of-plane hollow silicon microneedle (Fig. 16a
and Fig. 17a). The microneedle features a complex shape, shown in Fig. 16b, to
provide two main important functions. The first is the sharp tip, with a radius
below 100 nm, [119] which allows for reliable insertion [131]. The second is the
side opening. By having the opening on the sidewall instead of at the tip, the risk
of clogging is drastically reduced and better protection to the microsensor
inserted in the lumen is provided. Finally, because of its size, the microneedle
results in painless intradermal insertion. A comparison between a commercial
CGMS insertion needle and this microneedle is shown in Fig. 16c. The
microfabrication, design, and development of the microneedle shape are not part
of this thesis work, and have been described and studied in depth elsewhere [58].
Fig. 16. (a) CGMS on a fingertip. The microneedle is highlighted and barely visible. (b) Enlarged
SEM micrograph showing the hollow silicon microneedle shape and the side opening. (c) SEM
micrograph comparing the microneedle to a commercial CGMS insertion needle used to insert
Abbott Freestyle Libre’s sensor, as an example.
In initial tests, each microsensor was laboriously manually inserted in the lumen
of the microneedle and fixated to the microneedle chip base by epoxy glue. Wires
were attached with conductive epoxy glue to relatively large (500 x 500 μm2)
contacts on the sensor chip. The microneedle chip is then attached to a holder,
used to perform insertion and to fixate the interconnections to the external
electronics. The result of an assembled CGMS is shown in Fig. 17a. Fig. 17b
38
shows the three-electrode sensing microprobe inside the lumen of the
microneedle, with the opening enlarged for visualization purposes.
Fig. 17. (a) Picture of a manually assembled and contacted device. (b) Zoomed-in SEM
micrograph showing the three sensing electrodes inside the microneedle lumen. For visualization
purposes only, the opening was enlarged by laser ablation.
Performance in vitro and in vivo 4.6.2
The complete CGMS has also been characterized, first in vitro and then in vivo.
The MN lumen, whose walls are made of SiO2, is hydrophilic and thus fills
spontaneously when its opening is placed in contact with a liquid reservoir.
Moreover, the in-vitro studies confirmed the previously shown results.
Additionally, another parameter of particular interest is, in this case, the
response time of the system. To be able to measure in real-time the response time
of the system (sensor plus microneedle housing) needs to be faster than the
physiological variation rates in the human body. Fig. 18a shows the chrono-
amperometric result for +50 mg/dL glucose concentration steps, up to 250 mg/dL.
The response time (90% rise) at each step was 315 s on average, only 15 s more
than for the sensor alone (Fig. 15b). Translated in a variation rate, this means
9.5 mg/dL/min which is much larger than typical maximum physiological glucose
variation rates of ~3 mg/dL/min [137]. Therefore, the CGMS can track extreme
glycemic variation rates in real-time.
To verify in-vivo functionality, the device was applied to the forearm of a
healthy volunteer, following ethical standards (IRB approval: EPN Stockholm,
Dnr. 2015/867–31/1). The device insertion is a fundamental part of the procedure
to guarantee correct operation, especially in this situation where the microneedle
has to be worn for hours or days. Because of their length, microneedles do not
penetrate the skin as easily as hypodermic needles do, even with the advanced
geometry shown in Fig. 16b. This is due to the mechanical and elastic properties
of the skin, and its ability to deform under the applied pressure [138]. Insertion
speed plays a major role in piercing the skin and achieving deep and reliable
microneedle penetration [139]. For this purpose, a spring-loaded inserted was
39
used, to generate an insertion speed of approximately 3 m/s. Successful insertion
was verified by measuring the resistance drop upon wetting between two
electrodes. If skin penetration was achieved, ISF filled the hydrophilic MN lumen.
Penetration, unlike with manual application, proved to be extremely consistent
using this method. To stabilize the microneedle chip on the skin once inserted, a
thin layer of medical-grade adhesive tape was placed under the chip base, to
fixate it to the skin and limit unwanted motion or ejection of the microneedle
from the skin, in case of motion of the arm by the volunteer.
Then, to test the in-vivo sensing functionality, a so-called glucose tolerance
test was performed: starting from fasting conditions, the person orally intakes a
relatively large dose of sugar (35 g). The glycemic curve is recorded over time
until it falls back to normally low values. In healthy individuals, this occurs
within a couple of hours. The values recorded with the CGMS were compared to
capillary blood values measured with a commercial blood glucose meter. A two-
point calibration was used. The results are shown in Fig. 18b. The CGM curve
follows the reference glycemic curve, with a small delay which is in line with the
physiologically expected value of approximately 10 minutes between blood and
ISF values.
Fig. 18. (a). Chrono-amperometry of a CGMS response in vitro at increasing glucose
concentrations. Inset: same current data plotted versus glucose concentration. (b) In-vivo
characterization, in a volunteer’s forearm. Results of a standard glucose tolerance test, where the
green continuous line represents the CGMS reading from the skin ISF, averaged every minute to
cancel noise, and the blue dots are the capillary blood values.
In conclusion, we realized an extremely miniaturized system able to perform in-
situ CGM in the dermis, with minimal delay with respect to glycemia oscillations,
relying on passive lumen wetting with ISF by capillary action, over a few hours.
Therefore, this device potentially enables minimally invasive, simple, and real-
time CGM for patients affected by diabetes.
40
4.7 What’s next?
The studies performed demonstrated proof-of-concept operation of the proposed
microneedle-based sensing system. Once demonstrated the concept, the next step
is to build a complete system and further improve the sensing performance in
parallel. Indeed, what is missing to create a complete CGMS is the electronic
readout and wireless communication systems, which eventually need to be
integrated. These electronic components are in principle readily available with
minor modification requirements and, if desired, could be optimized for maximum
patient comfort and minimal cost of the disposable parts. Data analysis,
processing, and disease modeling are also essential aspects for collecting reliable
information.
Nevertheless, there is a series of necessary studies to demonstrate long-term
operation in vivo. As mentioned before, the performed in-vivo studies lasted only
for the approximately two hours needed to perform a glucose tolerance test. This
limitation came from the experimental setup: the device had to remain attached
to the spring-loaded inserter during the whole duration of the experiment. This is
because a transferable patch with wireless electronics was not developed at this
stage. As a result, the device could not move conformally with the volunteer’s arm,
completely limiting movement freedom throughout the entire test. Loss of
electrical contact overtime was, in fact, the main failure mode. This might be due
to different reasons, including the ejection of the needle from the skin, formation
of bubbles, ISF evaporation or foreign body reaction hindering ISF entrance in
the lumen of the microneedle. After realizing a transferable adhesive patch, more
studies on the interface between the microneedle opening and the skin would be
essential. Only then, the long-term operation could be assessed.
Furthermore, something relevant to be explored in this regard is a change in
the size and/or the number of microneedle openings. The microneedle geometry
was earlier developed primarily for drug delivery, and not for sensing
applications [58], [119]. Increasing the area through which the lumen
communicates with the dermal space or even create pathways through it could
further enhance both liquid filling and molecular exchange by diffusion.
Moreover, some improvements regarding the sensing technology, potentially
using also more complex chemistries (second or third generation) could be helpful
to further improve performance and, possibly, remove the need for initial
calibration.
Finally, the last research step towards commercialization would be to perform
studies on a larger and more diverse population (gender, age, etc.) and skin types
for final validation in the form of clinical trials. These trials would need to
include operation under everyday-life conditions, such as during physical
activities.
41
Chapter 5
Microsensor fabrication and integration
A miniaturized CGMS has a potential impact only if a scalable manufacturing
and integration strategy can be developed. This represents the third query: can
the system be manufactured, assembled and electrically packaged in a manner
that potentially enables large-scale production, despite the size reduction and the
consequent fragility of the components?
In the first place, to preserve simplicity, the CGMS was conceived so that the
different components (sensor and microneedle) can separately be fabricated using
standard silicon micromanufacturing methods. Then, what was missing is a
method to handle and vertically assemble sensing probes within the microneedle
lumen, and to then create electrical interconnections between the sensor
electrodes and the packaging. This need comes with strict requirements: the
silicon microchips are tiny, fragile and sensitive to solvents and temperatures
above 40°C, due to the functionalization with an enzyme.
In this chapter, the microfabrication process of the microsensors alone is
presented and, then, after reviewing the state-of-the-art for heterogeneous
microchip assembly and electrical interconnection, a novel integration method to
potentially enable large-scale manufacturing of the CMGS is presented. This
method was generalized and characterized for different microchip sizes, and is
also applicable to the exact geometry of the microsensor-microneedle CGMS, as
reported at the end of this chapter.
5.1 Microsensor fabrication
The microsensors were fabricated starting from an oxidized silicon substrate,
using standard photolithography and microfabrication technologies as illustrated
in Fig. 19a-d. Platinum was deposited and patterned to form the sensing
electrodes, interconnections and contact pads. The Q-RE area was covered with
an iridium layer. Then a silicon nitride insulating layer was deposited and
patterned to passivate the interconnections. To obtain an IrOx layer on the Q-RE,
cyclic voltammetry was used, cycling the electrode potential between oxygen and
hydrogen evolution extremes in a PBS solution to create a thin oxide layer on the
surface [83], [140]. Then, the silicon substrate was etched from the backside to
thin down the silicon area in correspondence of the sensor area, enough to fit in
the microneedle lumen (< 70 μm), and then from the frontside to define the
42
perimeter of the chip. When this process was completed, the electrodes could be
functionalized with GOx, polyurethane and Nafion, to fulfill the functional
requirements described in the previous chapter. These layers were dispensed
from liquid solutions via drop-casting. Details regarding fabrication and
processing methods are reported in the appended papers I and II. To achieve
better coating uniformity and control, inkjet printing [141], [142] was tested as
well. Inkjet allows printing of pL-sized droplets, with few microns placement
accuracy, unlike drop-casting. On the other hand, liquid solutions can be
correctly jetted only if their viscosity and surface tension are within specific
ranges, and if they do not clog the channels due to excessive volatility of the
solvents. As a consequence, of the three solutions only the GOx-BSA layer could
be modified and successfully printed. However, the enzyme performance
significantly deteriorated due to the printing process, making it a non-viable
approach for miniaturized sensors.
Fig. 19. Microfabrication process steps. (a) Platinum electrode deposition and patterning. (b)
Iridium deposition on the Q-RE area and passivation of the interconnections. (c) Oxidation of the
iridium electrode and silicon substrate machining to form the microsensor shape (d)
Functionalization of the electrodes by drop-casting.
This process provided functional sensors ready to be combined with the
microneedle chips. However, as mentioned, up to this stage, the microsensors
needed to be singularly assembled and electrical contacted, in a manual and
time-consuming manner. To enable large-scale manufacturing, a method to (i)
assemble and (ii) contact the microsensors had to be created.
5.2 Microchip assembly and contacting techniques
Standard silicon micromachining techniques, while established and cost-efficient,
do not allow the creation of complex 3D geometries. Therefore, the creation of 3D
MEMS structures requires the integration of multiple separately fabricated
components and to achieve large-scale manufacturing is a challenge. In
particular, the heterogeneous integration of microsystems can be divided into two
43
sub-problems: (i) handling, placement and orientation of the microchips, and (ii)
creation of reliable electrical interconnections between the microchips and the
external packaging.
Regarding the first sub-problem, various methods have been proposed to
assemble microcomponents, including contact-based and contactless approaches.
Contact-based methods can consist in pick-and-place strategies [143]–[147] and
transfer printing [148], [149]. However, these methods suffer from scalability
issues or difficulty to vertically orient the chips and are by nature less compatible
with fragile devices. Contactless methods for in-situ out-of-plane actuation of
planar structures attached to a substrate include magnetically induced bending
[20]–[22], [150]–[152], internal or thermal stress [153]–[156], or surface-tension-
driven assembly [157]–[159]. These techniques are not versatile and do not allow
integration of planar microchips into a separate receiving substrate. Besides,
they require exotic materials or high temperatures, incompatible with biosensing
applications. Other stochastic contactless approaches to place discrete
components on a surface or into a template of holes include controlled vibration
stimuli [160] and magnetic manipulation [161]–[165]. Magnetic assembly is of
particular interest because enables contactless manipulation even of fragile
components without exposure to harsh environments. However, these methods do
not provide the possibility to perform a controlled vertical assembly. Previously,
only vertical magnetic-field-assisted assembly of natively-ferromagnetic nickel
wires used as through-silicon vias was demonstrated [166]–[168]. In all these
applications, orientation was not fully controlled.
The limitation in the orientation also limited the possible alternatives for the
second sub-problem: the electrical interconnection. Previously proposed solutions
for contacting vertical chips were limited to simple devices as resistors or
capacitors, with two interchangeable ends [169], [170]. Moreover, they involved
incompatible harsh processing such as high temperature, plasma etching,
physical polishing and contamination with polymers and solvents. To realize
large-scale and versatile electrical contacts, it would be preferable to use a well-
established and high-throughput industrial technique. One such technique is
wire bonding. Despite being a serial process, wire bonding is commonly used for
mass productions and can perform up to 200 bonds per minute. Not only it is fast,
but it provides also versatility, thanks to the pattern recognition feature
normally present in standard commercial tools. However, wire bonding typically
requires planar metallic pads to allow attachment. Therefore, to contact
heterogeneously vertically assembled microchips, an unconventional process had
to be developed. Other unconventional applications using wire bonding were
previously proposed [171] and direct bonding to planar silicon holes was
previously demonstrated [172], but none included contacting of heterogeneously
assembled microchips.
Therefore, no available method allows performing assembly and contacting
procedures compatible with the requirements (in terms of orientation, shape, size,
microfabrication, fragility, and sensibility) of the described glucose biosensors. In
this work, a heterogeneous 3D integration approach was designed and developed
to potentially enable its large-scale manufacturing. A combination of shape-
44
matching and magnetic-field-assisted manipulation was used to perform and
control the assembly process. Then, an unconventional wire bonding approach
(named here “edge wire bonding”) was developed to create electrical connections
between the vertically placed microchips and the packaging. The entire
integration process is performed at room temperature and without involving
harsh processing or exotic materials.
5.3 Controlled magnetic assembly
The first step was to perform directional vertical assembly of the microchips into
matching receiving holes. In the biosensor case, this would consist in fitting the
microprobe into the lumen of a microneedle. In this work, different widths of the
microchips (and of the receiving holes) were tested as demonstrators, to show the
general applicability of this method. To perform the assembly, the microchips
had to be collected, moved and correctly oriented to eventually fall inside the
receiving holes into the receiving substrate. To achieve this, a manipulation
technique based on the interaction between the microchips, the receiving
substrate, and an external magnetic field was developed. This technique is
inspired by previous magnetic-field-assisted assembly methods [165], [168], [173],
[174] and modified to achieve the desired result. The magnetic responsiveness is
provided by a ferromagnetic thin-film coating (nickel) on the microchip backside.
Then, the vertical orientation of the assembled microchips is controlled by a
combination of shape-matching and a striped patterning of the ferromagnetic
coating on the microchip backside. The process is illustrated in Fig. 20a-b.
Fig. 20. (a) Magnetic assembly principle. Microchips, coated with a ferromagnetic layer of Ni,
react to the presence of a magnetic field by vertically lifting on the substrate. If the magnet is
dynamically moved beneath the receiving substrate, the microchips move to follow the field. (b) By
using a striped pattern of the Ni layer, a preferential lifting orientation is provided along the
striping direction. Due to shape matching, provided by the “T”-shape, microchips cannot be
assembled upside down.
45
When a magnet is present and moved underneath the receiving substrate, the
microchips react to the dynamic magnetic field and can be manipulated without
direct contact.
Nickel thin films are magnetically anisotropic and, as a consequence, the
nickel-coated microchips are subject to a magnetic torque when an external
magnetic field is present [175]. Since their easy axis (i.e. the direction of
energetically favorable magnetization orientation) is parallel to the nickel film
plane, when a magnetic field is applied perpendicularly to the receiving substrate
the microchips tend to lift vertically [176]. The magnetic torque and lateral force
generated are proportional to the Ni thickness and the magnetic field strength
and have to overcome gravitational and friction forces to enable microchip lifting
and motion. This could be guaranteed with a 3-μm-thick Ni layer. However, to
achieve lifting in a specific direction, the easy axis has to be limited along such a
predetermined direction. Patterning the nickel layer into narrow stripes forces
the easy axis to be parallel to the striping direction, thus resulting in preferential
lifting, as shown in Fig. 21a-b [177]–[179]. This approach allowed achieving
vertical lifting independently on the microchip dimensions and aspect ratio, as
shown in Fig. 21c-d.
Fig. 21. (a, b) Effect of the striped pattern. With a plain Ni layer, lifting directionality is not
guaranteed. Moreover, by switching the polarity of the magnetic field, it is possible to rotate the
microchips upside down. (c, d) The striping allows, independently on the aspect ratio, to lift the
microchip vertically along the pre-determined orientation.
46
Importantly, by switching the magnetic field polarity, the microchips tend to flip
upside down. This feature is particularly useful to assemble all microchips during
the process and not only the ones that are initially correctly oriented. In addition
to the striping, the microchips are provided with a wider top part (“T”-shape), as
shown in Fig. 20b. This top part is wider than the receiving holes, preventing the
microchips from being assembled upside down. The combination of shape-
matching and nickel patterning enables an efficient assembly process.
5.4 Edge wire bonding
The second step, after having the microchips in place, is to create electrical
interconnections with either the receiving substrate or an external package. This
is performed using wire bonding, taking advantage of the very mature and
widespread infrastructure, while developing an unconventional process to serve
the contacting needs in this heterogeneous integration application.
The process is illustrated in Fig. 22a-f and is based on a standard ball-stitch
bonding loop using gold wires. The microchips feature frontside electrodes and
trenches on their topside in correspondence of each electrode (Fig. 22b). Each
trench is meant to host a ball bond connection, while the other end of the
interconnection is attached to a contact pad on the receiving substrate. The
trenches have a trapezoid shape to enhance the mechanical robustness of the ball
bond. The gold free air ball (FAB) used for the ball bond is slightly wider than the
trench. When the FAB is pressed on a trench (Fig. 22c-e), the plastic deformation
of gold is induced. The deformation results in two important effects: (i) the ball
bond wedges inside the tapered trench sidewalls, providing mechanical fixation,
and (ii) the ball bond deforms around the edge and contacts the frontside
electrodes, due to a certain offset toward the front side of the microchips (Fig. 22f).
47
Fig. 22. (a) Illustration of an array of assembled chips, to be wire bonded. (b) Each microchip
features three frontside electrodes, interconnected by thin-film resistors for characterization
purposes. In correspondence of each electrode, a trapezoid-like trench is present on the top side for
mechanical fixation of the ball bond. (c) Zoomed-in view of the bonding concept. The deformation
of the ball bond around the edge results in an electrical connection to the frontside electrodes. (d)
Backside view of an interconnected microchip. (e) Bonding process on an assembled microchip. The
FAB, slightly wider than the receiving trench, is squeezed into it. (f) To create ball deformation
around the microchip edge to contact the frontside electrode, the bonding process is performed on
the trench with a certain offset towards its front side.
48
5.5 Microchip and substrate fabrication for integration
studies
The microchips and the receiving substrate for the studies on magnetic assembly
and wire bonding were fabricated as illustrated in Fig. 23a-b, and as explained in
detail in the appended article III. In brief, the microchips shape was defined by
DRIE on the front side, and then Au electrodes were deposited and patterned. On
the backside, the microchips were thinned down to 65 μm, and a Ni layer was
deposited. The microchips remained in position due to the presence of small Si
tabs between the wafer and the microchip perimeter. Ni patterning and
microchip release were performed using femtosecond laser ablation. The
receiving wafer was processed with DRIE to create through holes (receiving holes)
and then Au bond pads were deposited and patterned.
Fig. 23. (a) Microfabrication scheme for the realization of the microchips. (b) Microfabrication
scheme for the realization of the receiving substrate.
49
5.6 Results
Fig. 24 shows an array of integrated microchips. Fig. 25a-c show a close view of
the results of the wire bonding procedure from different angles.
Fig. 24. SEM micrograph of a portion of an array of assembled and contacted microchips. Inset: A
closer view of a microchip within the array.
Fig. 25. SEM micrographs of the microchip top part after wire bonding. (a) Frontside view of the
ball bond. (b) Backside view. (c) Partial side view, highlighting the area in which the gold-to-gold
electrical connection is formed.
50
The method has been tested also for microchips having dimensions equal to
the sensing microprobes described in the previous chapter. Fig. 26b shows that
the wire bonding method can be performed even at this small scale, despite the
fragility of the silicon microprobes.
Fig. 26. (a) Illustration of a “T”-shaped microchip, featuring a microprobe with the same
dimensions of the three-electrode microsensors described in the previous chapter. (b) SEM
micrograph of the microprobe assembled in the lumen of a microneedle, on which wire-bonded
interconnections were created. (c). Assembly process. When the magnet is not present, the
microprobes lie down on the substrate. (d) When the magnetic field is present, the microprobes lift
vertically and start following the magnetic field until they fall into a receiving hole. A gentle
vibration of the substrate facilitated smoother motion at this scale.
51
The silicon trenches are narrower than the ones in Fig. 25a-c (50 μm versus 60
μm) and distributed only on a 280-μm-wide “T”, as reported in Fig. 26a. Fig. 26c-d
are extracted from a video (supplementary video 4 of appended article III) in
which microprobes are assembled in receiving holes with dimensions comparable
to the microneedle lumen (90 μm diameter here). As shown in the video, the
process is quite efficient when probe-shaped microchips are assembled in circular
holes. Substrate vibration significantly facilitated smoother motion and increased
assembly efficiency with microprobes. No vibrations were applied with larger
chips (330-μm-wide up to 1.5-mm-wide rectangular chips in matching rectangular
holes) in the tests reported below.
Then, the efficiency of the assembly process, and the quality and reliability of
the electrical contact were studied.
Magnetic assembly process characterization 5.6.1
Magnetic assembly can be seen as a stochastic process. To analyze its behavior
and efficiency, a statistical analysis was performed on small in-line arrays and
modeled to extrapolate the behavior in case of larger arrays. These were
performed using an excess of microchips compared to the number of receiving
holes to maintain a constant “successful assembly event” probability throughout
the process. The permanent magnet was moved back and forth on the array until
the entire array was completely assembled. This procedure was repeated 50
times per each experiment. The mean number of sweep to fill a hole, after the
first assembly event within the array, tends to converge to a constant value (Fig.
27a). These values can be used to calculate the probability density to fill an array
(Fig. 27c), which after integration provides estimates of the total expected
number of magnet sweeps to fill an array (Fig. 27d). This behavior can also be
used to model and make a prognosis for larger arrays (dashed lines in Fig. 27c-d).
For example, to reach ~100% completion probability, 60 magnet sweeps are
needed for a six-hole array and 180 for a thousand-hole array. As mentioned
earlier, the efficiency here is strongly limited by the rotational freedom imposed
by the rectangular holes (e.g. as shown in Fig. 27b). Ordered oriented results
come in a trade-off with assembly speed. For example, for the rectangular holes,
the mean number of sweep needed for the assembly is around 20, whilst this
number is drastically reduced, below 2, for microprobes in circular holes.
Depending on the final application, circular or rectangular receiving holes might
be needed. Importantly, the contacting proposed technique can be used
independently on the orientation of the microchips within the receiving holes, due
to the pattern recognition feature available in standard wire bonding tools. A
series of factors influences the assembly process and efficiency. Apart from the
rotational freedom in the form of receiving hole shape (round versus rectangular)
and gap size, the total number of chips, the presence of vibrations, and the shape
itself of the microchips affected the assembly. Further experiments and details
about the used theoretical models are reported in the appended article III and its
supplementary information.
52
Fig. 27. (a) Mean number of magnet sweeps to fill one hole in an array, as the assembly progresses,
for different array sizes, from a single hole to an array of six in-line holes. (b) Illustration of the
rotational freedom for rectangular chips in holes with different shapes and different gap sizes. (c)
Probability density distribution of filling an array for different array sizes. The probability
densities are modeled as hypo-exponential distributions. (d) Probability distributions of filling an
array for different array sizes.
Contact quality and reliability 5.6.2
The quality of the electric contacts was evaluated by measuring the resistance
through the thin-film resistors (see Fig. 22b) before and after bonding. Before
bonding, the resistance was measured by directly probing on the frontside
electrodes, while after wire bonding the resistance was measured from the bond
pads on the receiving substrate. The measured resistance increase can thus be
used to quantify the worst-case contact resistance, including the gold wire
resistance (~0.04 Ω) and the stitch bond resistance. The average difference was
0.19 Ω. Therefore, the contact resistance was concluded to be negligible,
demonstrating excellent electrical conductance. Operation with continuous
electrical currents up to 20 mA was demonstrated. The absence of unwanted
short circuits was also verified.
The robustness of the wire bonds was assessed using a pull tester. The edge
ball bond proved to be the most robust part of the wire bond and to be more than
seven times stronger than standard requirements [180]. Before the failure of the
ball bond, both detachment of the stitch bond (minimum applied force needed:
~90 mN, Fig. 28b) and rupture of the wire connection itself (minimum applied
force needed: ~170 mN) always occurred before ball bond failure.
53
Fig. 28. (a) Measured resistance values before (left) and after (right) wire bonding. The contact
resistance is negligible. (b) Results of a pull test. Failure never occurred at the ball bond site.
Therefore, a heterogeneous method for vertical assembly and electrical
connection of microsensors was designed and implemented. The method was
tested and verified for different microchip sizes and receiving substrate
geometries. The process did not involve high temperature or harsh chemicals and
is thus compatible with enzymatic sensors and other sensitive and fragile devices.
54
55
Chapter 6
Interstitial fluid sampling
This chapter focuses on ISF microsampling devices as minimally invasive
alternatives to blood sampling. First, the current status of the clinical use of ISF
as a measurement matrix is introduced. Then, the state-of-the-art of blood and
ISF microsampling techniques, respectively, is reported. After this overview of
the field, the design, development, and characterization of a MN-based device for
volume-metered ISF microsampling are presented.
6.1 The use of ISF as a specimen in the clinic
The possibility to use ISF as a monitoring matrix depends on the specific
analyte of interest and its ability to cross the capillary vessels’ endothelium to
enter the interstitial space. As anticipated in section 1.1.3, molecular size cutoff,
transcytosis and active transport are mostly responsible for the differences in
composition between ISF and blood. ISF is nowadays used as a measurement
matrix for CGM, as described in chapter 4. Except for glucose, ISF is mostly used
for pharmacokinetic studies for therapeutic drug monitoring and development
[181], [182]. The use of ISF as a measurement matrix is currently limited by two
strongly interdependent factors. The first is the limited number of extensive
studies on the correlation of the concentrations and dynamics of compounds in
blood (or plasma) and ISF. The second is the lack of commercial sampling devices
providing easy access to ISF. For most applications, the main benefits of
substituting blood with ISF would be for patients, since ISF sampling can
potentially be sampled minimally invasively. Realistically, unless a sampling
method could provide easy access to ISF to perform necessary large-scale studies,
there would not be sufficient driving force from the medical community to
migrate from blood standards to another matrix.
Another situation of interest for using ISF instead of blood is when a simple
yes/no answer to the presence of one or more specific analytes (e.g. alcohol,
medical drugs, and drugs of abuse) would be sufficient. In this case, unlike for
quantitative analysis, precise and specific blood standard values would not be
necessary. Thus, as long as these substances are verified to be present in ISF and
at a detectable concentration, there might be no need to perfectly know the ISF-
vs-blood dynamics or to achieve high measurement precision through large
sample volumes.
56
6.2 State-of-the-art of minimally-invasive sampling
Minimally-invasive biofluid sampling techniques are used to minimize the pain,
discomfort and physical damage to the patient, while still extracting the same
liquids used for clinical assessment, whether it is blood or ISF.
Blood sampling 6.2.1
To collect blood, there is no way around inflicting a certain degree of damage to
the patient skin to break some capillary vessels and cause bleeding. While most
standard measurements are still performed in large blood volume, requiring
venipuncture, several applications would not necessarily require that. Finger
pricking became a standard in some applications, such as glucose monitoring. In
general, at least from a technological perspective, the research direction is to
create sampling devices and methods to minimize invasiveness and enable PoC
solutions, minimizing the need for trained staff and easing logistics.
Perspectives on DBS and mass spectrometry
Dried blood spots (DBS) consist in the storage of small volumes of capillary blood
into a cellulose-based substrate. This technique was proposed in the 1960s and
has been widely used for pediatric purposes, for example [183]. One of the main
advantages of DBS is the fact that, once dried, the blood is adsorbed by the
porous paper matrix and can be stored there for days or even up to months in
desiccated ambient conditions. This simplifies tremendously the logistics
normally required by blood or plasma transportation, which involves constant
refrigeration and is thus costly. Additionally, biohazards are reduced by handling
DBS instead of blood in its liquid form [184].
On top of handling and logistics, also the sample collection is facilitated for
DBS. Venous blood sampling provides few milliliters of blood, which need to be
extracted via venipuncture. Instead, DBS only require smaller amounts of
capillary blood, in the range of a few microliters, which are extracted via finger
pricks. Therefore, DBS sampling can be easily performed at the PoC by the
patients themselves, unlike venous blood extraction that requires trained medical
staff (phlebotomists). This is of particular interest also for newborns and ill
patients, for example.
On the other hand, there are some issues related to DBS collection and
analysis. A major issue reported for DBS collection systems is, for example, the
difficulty to obtain a homogeneous spot with a known blood volume for
quantitative analysis. Recently some solutions based on paper saturation or
microfluidic platforms have been proposed to collect known amounts of blood
independently on, for example, hematocrit levels [184]–[186]. Moreover, the
smaller amount of sample collected results in measurement challenges, especially
in terms of sensitivity required by the analysis tools to detect the targeted
analytes. A sensitive and precise characterization method that allows analyte
detection from small sample volumes is thus necessary. In some cases,
immunoassays have directly been integrated in a format similar to LFA (see
section 2.2.1), with limitations in terms of sensitivity, selectivity, and multi-
analyte detection possibilities [183], [187]. Alternatively, mass spectrometry (MS)
57
has been proposed for DBS characterization. MS is often used in clinical and
analytical chemistry to measure small and diluted substances. More recently,
tandem mass spectrometry (MS/MS) allowed detection of multiple analytes
during the same measurement with higher selectivity, providing a clear
advantage compared to LFA-like immunoassays and other techniques. In
particular, tandem mass spectroscopy combined with liquid chromatography (LC-
MS/MS) is currently used in pharmacological studies for drug development. LC is
used for physical separation, while MS/MS relies on a double ionization process
to differentiate compounds according to their mass-to-charge ratio, providing
high sensitivity and specificity. For these reasons, LC-MS/MS is nowadays
collecting growing interest also for DBS characterization.
Blood microsampling devices
DBS takes a step towards PoC applications, providing a solution for sample
storage and easing the logistics. However, the sampling procedure itself,
performed using finger pricks, is still painful. An additional step can be taken to
minimize the invasiveness of the sampling procedure as well. For blood
extraction, some solutions based on microneedles have been proposed [188]–[190].
Two of these works were based on a single hydrophilic stainless steel microneedle
pricking the skin [188], [189]. To collect blood, either a reservoir or a microfluidic
chip (including an on-chip assay) was attached to the microneedle backside. The
filling was helped by under-pressure inside the chip or by wicking fibers
enhancing capillary action. In both works, animal studies on either rabbits or
rats were performed. In the third work, a microneedle array was used to create
damage to the skin, while the collection was based on on-chip-stored vacuum
suction [190]. The blood was collected in a reservoir until the filling was
confirmed by blood visualization through a “window” at the end of the reservoir
itself.
ISF sampling devices 6.2.2
To sample ISF instead of blood, there are two major aspects to be taken into
account. Firstly, ISF tends to be retained by the skin and, unlike blood, does not
flow out spontaneously because no net hydrostatically-induced pressure is
present. Therefore, an external force must be applied to collect ISF from the skin,
e.g. mechanical pressure, vacuum suction, capillary-induced flow, etc. Moreover,
ISF sampling rates are limited by the ISF diffusion within the dermis, rather
than by the device itself [191]. A consequence of these properties is that ISF
collection is typically slower than what it is achievable with blood extraction.
Secondly, if too much damage is inflicted on the skin during pricking, bleeding is
induced, defeating the initial purpose. Therefore, ISF sampling can and must be
performed in a minimally invasive manner.
The first method to sample ISF was based on suction blisters [192]. Such
blisters are formed by applying vacuum for 1-2 hours to a small area of the skin,
resulting in the detachment of the epidermis from the dermis and in the
accumulation in that space of a liquid called blister suction fluid, which could
then be sampled by puncturing the site and collecting the liquid into a vial [193]–
58
[195]. Suction blister fluid is reported to be similar or equivalent to ISF, while
some studies reported specific composition differences [1], [57], [59], [194], [196]–
[198]. Another option previously proposed was to use microdialysis [60], [199].
However, these methods are relatively invasive, time-consuming and quite
patient-unfriendly [2].
ISF extraction can alternatively be performed in a minimally invasive manner
using microneedles [191], [200]. Three main strategies have been proposed, using
hollow MNs, solid “poke-and-blot” MNs, and swellable MNs, respectively. Hollow
MNs are inserted intradermally and ISF flow is induced into their lumens, which
are typically made hydrophilic to facilitate sampling, by vacuum suction [116],
[201] or by applying mechanical pressure on the surrounding skin area [57], [59],
[131], [202]. Another proposed strategy was the use of solid MNs to repeatedly
prick the epidermis and then collect ISF by simply blotting with an absorbing
matrix the pricked area [203], [204] or by applying vacuum suction to collect the
fluid in a reservoir [191]. Swellable MNs are made of hydrogels or other porous
polymeric materials instead. Once inserted in the skin, they can absorb ISF by
swelling or filling their internal pores, incorporating the target molecules to be
studied [70], [205]–[210]. Other devices have been proposed for the analysis of
small ISF volumes in combination with microneedles, but the sampling in vivo
has not been reported [211], [212].
Some of the main limitations of these interesting methods are related to their
complexity, long sampling times or incompatibility with current analytical
standards in the clinic, or the risk of contamination by sweat. Moreover, many of
these studies were validated only on animal models or ex-vivo skin. An additional
value for medical acceptance of the use of ISF in the clinics could be provided by a
simple device that guarantees sampling of a certain volume-metered amount of
ISF, stored in a stable form, and compatible with characterization techniques
currently used in medical laboratories for the same final medical assessment.
6.3 A device for volume-metered ISF sampling
An important technological requirement to enable future clinical use of ISF as a
monitoring matrix is to provide a simple platform that is designed to
simultaneously satisfy requirements in both ISF sampling and subsequent ISF
characterization. In other words, a device should be thought to be widely usable
(i.e. cost-effective) and to provide a straightforward manner to perform the
desired measurements on a known sample volume once the sample has been
collected. To move towards this goal, a possible implementation could involve the
characterization chain used in DBS after collection, i.e. collect and store the
sample in a cellulose matrix and then use LC-MS/MS for analysis. As mentioned,
such a method enables simple logistics and measurement in small (few µL)
sample volumes. Additionally, collecting a volume-metered amount would enable
quantitative analysis, as previously proposed [213], [214]. This thinking,
combined with a microsampling device able to extract ISF painlessly from human
skin in vivo, could result in a step forward towards the use of ISF as monitoring
matrix for a broader number of medical applications in the clinics and for PoC
sampling.
59
Design and implementation 6.3.1
With the abovementioned requirements in mind, a volume-metering ISF
microsampling device was designed and realized. The device was designed to
sample ISF from within the skin of the human forearm and to store the liquid in
a paper matrix. Sampling relies on a single stainless steel microneedle whose
backside is connected to a simple microfluidic chip enclosing the paper matrix.
The paper can then be removed and used for off-line analysis. The microneedles
were obtained by ablating standard 32 G hypodermic needles. The microfluidic
chip was based on simple laminated plastic and adhesive layers. Additional
details on the fabrication procedure and the material choice are reported in the
appended manuscript IV.
The system is composed of two main parts: (i) the sampling device described
above, and (ii) an inserter featuring a pressuring ring. The latter is responsible
for the insertion of the MN in the skin and for the generation of the mechanical
overpressure needed to overcome the physiological ISF retention by the skin to
guide the ISF into the MN lumen. Fig. 29a illustrates one of the possible
implementations of the described system, with an armband-like solution applied
to the forearm, featuring a combined applicator and pressurizing ring. Fig. 29b
illustrates the cross-section of the sampling device inserted into the skin.
Fig. 29. Illustration of the system features. (a) The system applied to the forearm. The green
armband-like structure includes an integrated inserter and pressurizing ring. (b) Cross-section of
the device inserted in the dermis. The ring applies a mechanical pressure inducing filling of the
microneedle lumen, connected, in turn, to the paper matrix. (c) Top view illustration of the paper
matrix, with a filling with ISF over time, and transporting the water-soluble blue dye along the
flow direction. Once the liquid flow reaches the visualization window at the end of the path,
sampling is stopped.
60
Volume-metering is provided by the geometry of the paper and by a colored
dye present in the paper itself. During sampling, the liquid progresses within the
paper matrix transporting the soluble colored dye, as shown in the top view
illustration in Fig. 29c. Once the liquid reaches the visualization window, the
device is removed from the skin. The realized device is simple, compact and
potentially cost-effective.
Fig. 30a-c shows an assembled disposable device and compares it with a
commercial lancet commonly used for finger pricking.
Fig. 30. Pictures of the assembled disposable device. (a) Sampling device on a fingertip. (b) Side-
to-side comparison between a traditional lancet for finger pricking and the sampling device. (c)
Closer view of the stainless steel microneedle.
Characterization 6.3.2
The microfluidic chip is designed to collect ~1 µL of ISF. The reliability was
characterized in vitro. In most cases, sampling was completed within 5 minutes,
with a variation between 3 and 8 minutes (n = 15). In several tests, the
pressuring ring was manually held in position during sampling. The MN
insertion was reported as painless by all volunteers. An example of a device used
for the sampling procedure in vivo is shown in Fig. 31a-d.
61
Fig. 31. Sampling procedure performed in vivo, on a human forearm, overtime. The timeframes
are taken at (a) 20 s, (b) 80 s, (c) 130 s, and (d) 190 s, respectively, after starting the sampling
process. The entire paper matrix is here visible through a transparent plastic foil for visualization
purposes.
Samples can then be analyzed using LC-MS/MS, with a similar protocol to the
one previously developed for DBS analysis [214]. Early results are reported in the
appended manuscript IV, and demonstrate the possibility to detect caffeine in 1
μL samples of artificial ISF, spiked with physiologically relevant caffeine
concentrations.
6.4 Perspectives
The development of the ISF sampling system is still in a relatively early phase.
Additional studies and design improvements are carried on as this thesis is being
written. There are several studies and optimization that have to be performed to
demonstrate or to achieve the final goal in terms of performance and usage. As a
first step, the characterization of the volume metering feature and the sampling
time in different individuals and with different operators should be performed to
prove the metering reliability and the viability of the system usage in everyday
life conditions. Moreover, the removal of the paper matrix is currently done
manually after completion of the sampling procedure, but it would ideally be
automated at a later stage to allow drying without operator intervention.
Ongoing LC-MS/MS measurements will provide an idea of the clinical validity of
the concept as well as indicate exact requirements in terms of required sampled
volumes. Depending on the substance to be measured and its concentration range
in ISF, the required amount of sample might even be reduced below 1 µL, leading
to a significant sampling time reduction. Design optimizations could be
investigated to minimize the necessary overpressure applied to the skin, to
further reduce patient discomfort. Moreover, the pressuring procedure was still
62
performed manually in several tests. Different designs of the combined
pressuring inserter are currently under evaluation, in order to simultaneously
maximize patient comfort, sampling speed and independence on the operator.
Finally, while materials are inexpensive, the assembly procedure needs to be
automated to enable large-scale manufacturing.
63
Chapter 7
Gas diffusion and evaporation control for gas
sensors
In this chapter, a technical solution aiming at the extension of the lifetime of
electrochemical gas sensors, e.g. for breath monitoring for asthma, is presented.
This type of sensors suffers from a limited lifetime due to the evaporation of the
liquid electrolyte and the response drift, especially when miniaturized. To tackle
these issues, a gas diffusion and evaporation controlling system based on
electrowetting-on-dielectric (EWOD) actuation of ionic liquid microdroplets is
proposed. This device is based on ionic liquid microdroplets, which are electrically
moved on a perforated membrane covering the sensing compartment to expose or
seal the sensor itself. Moreover, the device is compatible with direct integration
with miniaturized gas sensors. While developed for this specific application, the
realized digital microfluidic platform can also be adapted to serve different
purposes, such as to control the venting of channels in microfluidics, to perform
particle collection, or to realize an optical shutter as described in chapter 8.
7.1 Digital microfluidic platforms
Devices based on DMF platforms have been developed mainly as a low-volume
and reconfigurable alternative to traditional channel-based microfluidics to
perform bioassays and serve other biomedical applications [35]. Another field of
application for EWOD that is collecting growing interest is the realization of
optofluidic devices [215].
While interesting because of the opportunity it provides for purely electrical
control of liquid motion, continuous operation with DMF platforms has some
critical requirements to be addressed. In biological applications, for example,
surfaces are prone to biofouling that overtime hinders correct droplet motion.
When biomolecules or other precipitating substances are not present in the
moving droplets, as in the work presented in this thesis, failure modes are
related to the electrodes and coatings used for actuation. Failure in DMF
platforms is often induced by an excessively large applied voltage, damaging the
dielectric layer. The hydrophobic-dielectric layer failure occurs in gradual steps
at increasing voltage, depending on the quality of each layer and not necessarily
in the following order. Typically, first, the hydrophobic coating slowly degrades
64
due to ion injection [216]. Then, if and where pin-holes are present, the liquid
penetrates and reacts, forming visible gas bubbles and damaging the metal
electrodes due to electrolysis. Eventually, dielectric breakdown occurs when the
dielectric capacity is overcome. Therefore, the quality and uniformity of the
dielectric and hydrophobic layers, combined with the material choice, are
essential. The thinner the dielectric layer (and the larger the dielectric constant),
the lower the actuation voltage can be. However, at the same time, the thinner
the layer the more difficult it is to obtain a uniform and compact dielectric layer
to achieve safe and robust operation. Another common related drawback for
several reported EWOD platforms is the large actuation voltage required, which
is often larger than 100 V and up to 300 V [217].
Moreover, a major problem of digital microfluidic platforms to create
functional devices is their limited durability due to the possible evaporation of
the microdroplets overtime when operating in “open-air”. Sealing of the system is
not a viable approach in all devices, including the application targeted in this
chapter, i.e. the control of gas diffusion to and from a compartment. Because of
the need for a free path for the gas molecules to go through the device (from the
environment to the sensing compartment of a gas sensor) when in the open state,
the microdroplets have to be exposed to the external environment as well. For
this reason, ionic liquids were identified as an ideal microdroplet material,
substituting the water-based microdroplets commonly used, to enable “open-air”
EWOD applications.
Ionic Liquids as movable microdroplets 7.1.1
The use of ionic liquids (IL) provides the possibility to overcome these droplet
evaporation issues, because of their extremely low vapor pressure that results in
a negligible evaporation rate compared to water [218]. To provide a practical idea,
the shelf life of a 250 nL IL droplet (half-sphere of ~ 1 mm diameter), calculated
as the time for losing 20% of its volume, corresponds to more than 100 years at
room temperature, instead of a few hours such as in the case of water [219], [220].
Ionic liquids are synthetic organic salts featuring a low melting point, often below
room temperature. The formation of a stable crystal lattice is prevented by the
poor coordination of their ions. This is in turn due to the bulkiness and
asymmetry of the ionic molecules that compose them, which soften the long-range
Coulomb interactions [218]. Nonetheless, the charge interaction and molecular
size are enough to prevent relevant evaporation even at high temperatures. Their
ionic composition is very variable so that thousands of combinations to create
different ILs are possible. Different ILs can present very different properties: in
addition to viscosity and surface tension properties, they can also be either
hydrophilic or hydrophobic, and hence miscible or immiscible with water.
Their main applications are currently as “green” but powerful solvents and
electrolytes. While ILs can be synthetically tailor-designed, in this thesis only
commercially-available liquids were used. In addition to the shelf lifetime
provided by the negligible evaporation rate, ionic liquids are good candidates for
this application because of their viscosity and diffusivity properties. The
diffusivity of gases in IL is approximatively five orders of magnitude smaller than
65
in air, providing a good barrier to be used for gas sealing [221]. However, for
digital microfluidic applications, only investigations of EWOD-induced spreading
and motion have previously been reported [222]–[229]. The exploitation of ionic
liquids to build functional digital microfluidic devices is still in its infancy [42]–
[44], [230], [231]. A particular challenge in the use of ILs for EWOD actuation is
related to their low surface tensions, often resulting in a relatively small initial
contact angle. This characteristic results in increased wetting and larger contact
angle hysteresis. These drawbacks impede or complicate the motion of
microdroplets, especially on irregular surfaces like perforated membranes, as
necessary in the presented application. ILs with a static contact angle θ0 < 90°,
even on very hydrophobic surfaces such as PTFE, are thus not suitable for the
targeted use, both because of pinning on the holes present on the actuation
substrate and of the large voltage requirements to induce motion under such
conditions. Therefore, the correct choice of the movable liquid is crucial to achieve
performance and robustness. In this work, EMIM-BF4, BMIM-BF4, and BMIM-
PF6 were used. They feature a θ0 of 105°, 96°, and 91°, respectively, and large
contact angle (CA) variations upon voltage application (ΔCA) of 26°, 27° and 32°,
respectively, when using the actuation electrode designs and materials described
in the appended article V.
From a practical perspective, the correct operation and long lifetime of a
device would not be limited by the evaporation of the IL microdroplets. Operation
and lifetime are instead strongly related to the quality of the deposited dielectric
and hydrophobic layers, in combination with the actuation parameters and the
choice of the IL for the desired application [216], [232]. Design, choice of the
electrode materials and fabrication processes are important aspects in this regard.
The discussion about the material choices, the deposition techniques and
thicknesses used are described in detail in the appended article V, and the
resulting combination is schematically illustrated in the next section, in Fig. 33a.
7.2 A diffusion and evaporation controller for gas sensors
As introduced in chapter 2.4.4, in liquid electrolyte-based electrochemical gas
sensors, the lifetime and correct long-term operation suffer from (i) electrolyte
evaporation and (ii) the lack of a sealing mechanism between the sensing
chamber and the environment to potentially allow performing recalibration,
needed for long-term drift compensation [233], [234].
In state-of-the-art sensors, the electrolyte evaporation is either limited by
minimization of the area through which the sensor communicates with the
environment, e.g. by using porous hydrophobic filters [234], or by substituting the
liquid electrolyte with a solid/gel electrolyte [234], [235]. Both solutions can be
problematic in certain situations: by minimizing the open area also the response
time and sensitivity worsen, while the use of solid electrolyte alternatives is not
compatible with all gases to be measured. Regarding long-term drifts, the effect
can be limited by improving the quality of the RE, which is however extremely
challenging especially in miniaturized devices, where only Q-RE can be used. An
example of such a situation, a miniaturized liquid-electrolyte-based
electrochemical NO sensor for asthma monitoring, was earlier reported [77].
66
The integration of an on-demand, ON/OFF, gas diffusion- and evaporation-
controlling mechanism could both improve the stability and extend the sensor
lifetime, enabling prolonged use of miniaturized sensors of this type. Intuitively,
such a task could be performed mechanically, for example by an electrically
controlled shutter. However, at this (sub-mm) scale, the realization of a sealing
mechanical shutter entails complications especially in term of stiction in humid
environments, such as the top of a liquid reservoir. For this reason, ionic liquid
movable microdroplets were chosen instead, to cover and uncover the holes of a
partially perforated membrane separating the sensing chamber from the
environment. The concept is illustrated in Fig. 32a-b.
Fig. 32. Illustration of the working principle. (a) In the open state, the perforated portion of the
membrane is free, allowing gas exchange between the sensing chamber and the environment. (b) In
the closed state, an IL microdroplet is moved onto the perforated area of the membrane and covers
the openings, blocking gas exchange and evaporation.
The controlled motion of the IL microdroplets is electrically performed by EWOD
actuation. Additionally, this type of actuation also features low power
consumption.
An illustration of the cross-section of the device is reported in Fig. 33a. In
particular, the actuation is performed using co-planar EWOD (c.f. section 2.2.2):
two couples of electrodes are embedded in a partially perforated membrane to
enable back-and-forth movement of an ionic liquid microdroplet between the open
(solid part of the membrane) and the closed (perforated part of the membrane)
positions on demand. Fig. 33b shows a top-view illustration of the device and its
actuation scheme. The choice of coplanar actuation is motivated by the possibility
of eventually having only the bottom substrate to be microfabricated, thus
reducing the final cost of the device, and reducing the complexity of its assembly
and electrical connection.
67
Fig. 33. (a) Cross-section illustration of the microfabricated device. (b) Top-view illustration of the
device controlled by EWOD actuation.
Performance 7.2.1
The performances of the device were characterized in terms of actuation voltage,
actuation speed, gas diffusion control (using CO2 as test gas), and water
evaporation reduction. Three ionic liquids were tested: EMIM-BF4, BMIM-BF4,
and BMIM-PF6. The first two are miscible with water and, thus, could be used for
gas diffusion control, but not for evaporation limitation. Here, only the
characterization of BMIM-PF6 as actuated microdroplet using the design
illustrated in Fig. 33 is reported, while the complete characterization, with other
ILs, different materials, and electrode designs, is reported in the appended article
V and in [236].
EWOD Actuation
Actuation could be performed both with DC and AC applied voltage, but AC has
proven more reliable, as also previously reported [216], [237], [238]. Toggling of
the microdroplet position was achieved with an applied AC peak voltage as low as
22.5 V. At this voltage, the toggling time was approximately 2 s. To reduce the
actuation time down to 100 ms, the applied voltage was increased to 37.5 V.
These values, if compared to other EWOD platforms often requiring around 100
V or more, are very competitive [239].
Gas diffusion and evaporation control
The gas diffusion controlling performance was characterized using the setup
illustrated in Fig. 34a. A commercial CO2 sensor was used to quantify the time
difference between open and closed state to reach a certain concentration level
within the sensing chamber. The theoretically expected performance, using
diffusivity values from the literature, [221], [240], [241] was simulated using a
finite element modeling software, as shown in Fig. 34b. The simulation result is a
difference of more than three orders of magnitude diffusion time between the two
situations, both in the initial transient and around equilibrium during operation.
The behavior was later also experimentally validated. Fig. 34c shows the
difference between open and closed state: 214 s versus 29.4 hours to reach 20% of
the external CO2 concentration, which corresponds to three orders of magnitude
difference.
68
The control of evaporation was characterized in terms of the ratio of water
loss, from a chamber sealed with the device, between the open and the closed
state. One order of magnitude reduction of evaporation was achieved, as shown in
Fig. 34d. According to diffusivity values, the ratio should be larger. Diffusivity of
water and small gas molecules, such as CO2, is expected to be similar [242]. A
possible explanation for this behavior could be related to density-driven or
Marangoni induced convective transport of condensed water domains formed at
the bottom of the ionic liquid droplet bottom [243].
Fig. 34. (a) Illustration of the experimental setup for gas diffusion characterization. (b) FEM
simulation of the diffusion of CO2 through the device in the closed and open position, i.e. with or
without an ionic liquid microdroplet sealing the openings of a sensing chamber. To reach 90% of
the ambient concentration, the time necessary changes from 0.6 s to 10.7 hours in the two
configurations. (c) Experimental study of the time necessary to reach 20% of the external CO2
concentration, in closed and open positions, respectively. (d) Weight loss over time by water
evaporation when the device is in an open or closed position, respectively.
Therefore, a microfabricated system based on EWOD-actuation of IL
microdroplets was realized. The system was proven to reduce electrolyte
evaporation rates and slow down gas diffusion from a compartment, potentially
allowing, upon integration, for prolonged use of miniaturized electrochemical gas
sensors.
69
Chapter 8
EWOD-based optical shutters
The realized digital microfluidic platform based on ionic liquid microdroplet
actuation, introduced in the previous chapter, can be modified to serve different
applications. In this chapter, taking advantage of the same enabling features of
ionic liquid microdroplets in combination with “open-air” EWOD actuation, a
zero-insertion-loss broad-band optical shutter design is presented and
characterized.
8.1 Liquid-based optical shutters
Optical shutters are a common building block of optical setups and integrated
devices. To implement this and similar functions, such as irises [40], [244] and
lenses, [245], [246] manipulation of liquids has been proposed. These devices
control the shape or position of liquids to achieve their functions. This is
performed by EWOD actuation, using different designs depending on the
targeted application. Compared to mechanical alternatives, [247]–[250] EWOD
simultaneously offers simple and low-power actuation, polarization independence
and avoids the passage of scattered light. In most cases, liquid-based shutters
consist of two immiscible liquids, such as water and oil, one of which is mixed
with an opaque dye. By controlling the respective position, light can be
transmitted through a pseudo-transparent path (open state) or be absorbed by
the dyed liquid (closed state). Fig. 35a-d illustrate some previously proposed
implementations of this type [41], [251]–[253]. Other similar implementations
have also been proposed [254], [255]. Alternatively, the control of the shape can
be used as a lens to focus or defocus the light into an opening (Fig. 35e, [253]) or
to exploit diffraction and total internal reflections to control light passage [256],
[257]. All these solutions suffer from a common limitation: the light, even in the
open state, has to pass through several attenuating and refractive layers, such as
glass, ITO, Teflon, and water or oil. This inevitably results in insertion losses in
the open state. Moreover, they can operate only at wavelengths for which all
these materials are transparent, meaning that they are not universal in terms of
wavelength transmission. Additionally, most of them share another drawback:
they are not bi-stable, i.e. a voltage needs to be continuously applied to maintain
the shutter either open or closed.
70
Fig. 35. Different implementations of EWOD-based optical shutters. None includes a direct optical
through-path. (a) Dyed oil and water interface motion to create a pseudo-transparent optical path
[251]. (b) Motion of a water droplet surrounded by dyed oil to allow/block light transmission [41].
(c) Shape modification of dyed water in oil around a transparent glass post [252]. (d) Lateral
spreading of a dyed water droplet surrounded by oil to block light [253]. (e) Focusing or defocusing
the incoming light into an opening, by changing the shape of a lens formed by the interface between
two liquids [258].
71
8.2 Presented solution
To realize a broad-band zero-insertion-loss shutter, an optical through-path
needs to be present, so that in the open state no physical barrier is present to
attenuate or scatter the light. While the previous solutions could not allow a
complete through-path, because of the volatility of the used liquids, the use of
ionic liquids enables “open-air” operation. The design and operation principle are
illustrated in Fig. 36a-b. In the closed state, an opaque microdroplet blocks light
transmission through the device. In the open state, the droplet is moved away
and light can freely be transmitted through the device, with no physical barrier.
Fig. 36. Illustration of the operation principle. (a) In the open state, the laser beam can freely be
transmitted through the device. No physical barrier is present. (b) In the closed state, an opaque
microdroplet covers the through-path and the laser beam is thus blocked.
The microfabrication process followed the same procedure described in the
previous chapter (and in the appended articles V) while using a slightly different
design of the photomasks. The actuation scheme is also the same, with coplanar
electrodes. The cross-section is illustrated in Fig. 37a, while Fig. 37b shows a
partial top view of the device before and after cover plate placement. A basic
visual demonstrator is reported in Fig. 37c.
The main difference in this case, in addition to the presence of an optical
through-path in the final assembled device, consists in the need to modify the
ionic liquids. Ionic liquids as EMIM-BF4, BMIM-BF4 and BMIM-PF6 are fairly
transparent. Therefore, to make them opaque, a colored dye was mixed in. Some
important factors to be considered in the choice of the dye are (i) its miscibility
with ILs and (ii) its influence on the surface tension of the mixture, (iii) its optical
properties, whose requirements may vary depending on the targeted application.
Poor miscibility results in dye precipitation and thus fouling of the actuation
electrodes, while excessive reduction of the surface tension may hinder EWOD
actuation and create pinning effects. For example, water-immiscible ILs such as
BMIM-PF6 could not be homogeneously mixed with the selected dye. Instead, the
mixtures with EMIM-BF4, BMIM-BF4 resulted homogeneous and only minimally
affected the surface tension, so that the contact angle remained above 90°, as
reported in Fig. 38a-b.
72
Fig. 37. (a) Cross-section illustration of the device. (b) Partial top view of the device, before (left)
and after (right) placement of the cover plate. (c) Simple visual demonstrator of operation with a
laser pointer.
Fig. 38. Contact angle dependence on the concentration of the dye in EMIM-BF4 (a) and BMIM-
BF4 (b), respectively. Top row: natural contact angle on a PTFE surface. Middle and bottom rows:
contact angle after dye mixing at 0.1 wt% and 10 wt% concentration, respectively.
73
8.3 Characterization
In terms of actuation, the parameters are similar to the ones described in the
previous chapter, with an actuation voltage of 25 V and a corresponding
actuation time of less than 300 ms. The optical performance was characterized in
terms of maximum attenuation in the closed state, at different dye concentrations,
with both ILs (Fig. 39b), using the setup illustrated in Fig. 39a. In this case, a
laser with a wavelength of 532 nm was used, passing through a 250 nL
microdroplet, which is the same size shown in Fig. 37c and used in the previous
EWOD actuation experiments. A maximum attenuation of 78 dB was achieved.
The table in Fig. 39c summarizes the performance compared to state-of-the-art
alternatives. The presented device features zero insertion loss, a larger maximum
allowed laser power, low actuation voltage, and attenuation in line with the best
state-of-the-art alternatives. The actuation speed can be further improved by
increasing the actuation voltage, as earlier described.
Fig. 39. (a) Optical setup used for the characterization in the visible spectrum. (b) Dependence of
the closed-state attenuation on the dye concentration, using setup (a), with EMIM-BF4 (blue) and
BMIM-BF4 (red). Maximum attenuation = 78 dB. (c) Table comparing the performance of our
device to state-of-the-art alternatives. The attenuation and insertion loss refer to the visible
74
spectrum. [a]=[40], [b]=[41], [c]=[252]. (d) Characterization in the mid-IR. Zero insertion loss is
achieved, while under the same conditions state-of-the-art alternatives suffer from large losses.
The device was also tested in the mid-infrared (mid-IR) range (λ = 4.2 μm). The
results are illustrated in Fig. 39d. At such wavelength, competitors suffer from
large losses in the open state, because glass significantly absorbs it. The realized
shutter correctly operates also in the mid-IR spectrum, instead. By changing the
dye, enhanced performance in terms of attenuation for different wavelengths
could potentially be obtained, provided that the alternative dye is miscible and
does not significantly affect the liquid surface tension.
Therefore, the presented platform provides a compact solution for the
implementation of liquid-based optical shutters, with improved performances in
terms of actuation voltage, insertion losses, and operation bandwidth.
Additionally, the shutter also features zero static power consumption due to its
bi-stable switching principle.
75
Chapter 9
Summary and Outlook
This thesis work aimed at the development of microsystem technologies for
applications mostly focused on health monitoring. The presented technologies
have been developed to the extent of demonstrating proof-of-concept. Referring to
the objectives listed in chapter 3:
Objective #1. A glucose sensor with a footprint of only 580 x 70 µm2 was
realized. This represents 50-fold miniaturization compared to commercial
state-of-the-art, while preserving functionality in line with ISO standards
in terms of sensitivity, selectivity, measurement range, and response time.
Thus, the reduced size potentially enables in-situ intradermal monitoring.
Objective #2. The microsensor was combined with a side-opened hollow
silicon microneedle and tested in vivo. After painless insertion in the
human forearm, glucose levels could be tracked in ISF with a minimal
delay compared to glycemia dynamics (~10 minutes, as physiologically
expected). Passive access and in-situ analysis of the dermal interstitial
fluid was achieved, enabling a simple plug-and-measure approach.
Objective #3. To demonstrate the CGMS potential also in terms of large-
scale manufacturing, despite the fragility and small size of the components,
a method based on magnetic-field-assisted assembly for microchip
handling and edge wire bonding was created. The combination of nickel
patterning and shape matching enabled control over the assembly
direction, while the developed wire bonding technique enabled the creation
of reliable electrical connections between the vertically-placed microchips
and the external packaging.
Objective #4. A device to sample a known volume (~1 µL) of ISF has been
created. The disposable chip is small and potentially inexpensive, and the
system does not require complex external actuators, such as vacuum
pumps. The precision in terms of volume extracted was experimentally
characterized, but the device operation needs to be fully proven in vivo,
also with different sampling conditions and skin types. The minimally-
invasive device potentially enables painless and bloodless sampling of ISF.
76
Objective #5. The characterization of the ISF extracted and stored in paper
is currently ongoing, using LC-MS/MS. The protocol is under developed
and needs to be verified on ISF samples from human volunteers.
Objective #6. A microfabricated system enabling control of evaporation and
gas diffusion through a perforated membrane was realized. The device was
proven to reduce the evaporation rate of water from a compartment by one
order of magnitude and to slow down gas diffusion from the environment
to the sensor by more than three orders of magnitude. The integration of
such a device with miniaturized electrochemical gas sensors can thus
potentially prolong their lifetime.
Objective #7. While initially developed for a specific application, the
created platform can be modified and used for a wide range of applications,
due to the enabling possibility of realizing DMF platforms operating in
“open-air”. The device can be employed to serve similar diffusion-control
needs in other kinds of devices or, for example, potentially be used to
create particle collectors, optical shutters or implement more classical lab-
on-a-chip applications. The developed EWOD-based technology also offers
relatively low voltage (25 V) and low power operation. Moreover, the
created design provides bi-stable operation and, thus, zero static power
consumption.
Objective #8. A laser shutter based on IL microdroplet manipulation was
realized. The device features zero insertion loss in the open state and
broad-band operation, unlike other currently available liquid-based
implementations. Moreover, actuation voltage, toggling speed, and
maximum attenuation are in line or better, if compared to EWOD-based
alternatives.
The applications in objectives 1 to 5 are aligned with the United Nations
Sustainable Development Goal #3: good health and well-being [259]. In particular,
the improvement and widespread adoption of CGMSs can partly help in
“reducing by one-third premature mortality from non-communicable diseases
through prevention and treatment and promote well-being”, by monitoring
patients affected by diabetes. ISF sampling devices can indirectly help in
“strengthening the prevention of substance abuse” and “halve the number of
global deaths and injuries from road traffic accidents” by e.g. simplified roadside
testing and/or eased PoC monitoring.
In conclusion, a series of technological solutions to tackle problems in
healthcare products have been developed. Among them, two have a major
potential impact on end-users: (i) the miniaturization and widespread adoption of
CGMS and (ii) the reduction of invasiveness and discomfort for biofluid sampling.
The advantages could be significant not only in terms of reduction of patient
discomfort and potentially improved life quality, but also in terms of costs for
healthcare systems, due to complications and comorbidities of diabetes and due to
the facilities and personnel required to perform blood sampling. No pain, same
gain™.
77
Perspectives
To a smaller or larger extent, all of the presented technologies would require
further development and, in some cases, also an assessment of long-term
functionality under relevant conditions to fully prove satisfaction of the objectives
listed in chapter 3. To analyze the perspectives of this work, the developed
technologies will be divided into three areas: CGM (objectives I to III, in papers I
to III), ISF sampling (objectives IV and V, in manuscript IV) and digital
microfluidics (objectives VI to VIII, in papers V and VI).
For CGM, the steps required to bring the developed technology to the market
have been detailed in section 4.7 and can be summarized as follows:
The sensor chemistry could be further improved, by creating a third-
generation sensor or by enhancing the homogeneity of the coatings, to
minimize or remove the need for initial calibration and to ensure prolonged
usage in vivo.
Long term reliability of the interface microneedle-ISF has to be proven and
is currently one of the most critical aspects of the presented CGM
technology. To better assess this, the creation of a transferable adhesive
patch with integrated electronics is required. If necessary, microneedle
shape might need to be modified to optimize such interface. Operation in
different skin types and age/gender populations has to be assessed, as well
as correct functioning in normal living conditions, such as during physical
activities. In these regards, there is currently a lack of studies in the
scientific literature, as well as a lack of standardization in terms of
insertion and operation with microneedle-based devices.
The complete microfabrication process, combining sensing and assembly-
contacting functionalities, needs to be tested.
For ISF sampling, the work is still in progress while this thesis is being written.
The future perspectives can be summarized as follows:
The system design can be further optimized both to minimize the
mechanical overpressure required to extract ISF and to minimize the
sampling time, to improve patient comfort.
Device operation has been verified on a limited population. As in the case
of CGMS, more studies on diverse skin types and different populations are
required to guarantee correct functioning independently on the patient.
The characterization using mass spectrometry on a larger number of
samples, on real ISF samples, and from multiple individuals will provide
additional information, both in terms of precision of the volume-metering
feature and in terms of the minimum amount of ISF sample needed to
reliably detect the targeted compounds.
Being that ISF is currently not widely accepted as a clinical standard,
more studies in terms of ISF characterization would provide additional
78
information regarding the applicability to specific use cases (e.g. drug-of-
abuse monitoring). If all the previous aspects are verified, the system is
potentially simple and cost-effective enough to be marketed, but a major
barrier could still be represented by the reluctance of the medical
community in shifting from blood standards.
In terms of ionic liquid manipulation using EWOD to build digital microfluidic
platforms:
Verification of long-term operation of the proposed devices after
integration with the other components in the system, i.e. onto an
electrochemical gas sensor or into an optical setup, is required. The
efficacy of a recalibration process to compensate for Q-RE drifts,
potentially enabled by the presence of this device, still has to be proven.
The fabrication process can be further optimized to enhance conformity of
the holes’ sidewall coverage to limit droplet pinning, and in the quality of
the material deposition processes, thus increasing robustness and further
reducing actuation time and voltage.
Find additional applications. The realized devices successfully addressed
issues of current technologies, but the conceived platform might be used to
tackle different needs in other fields.
In conclusion, at the current stage, CGM and ISF sampling have significant
market potential. By performing the proposed additional studies, the real
potential for their marketability could be assessed and lead to future
developments outside of the academic environment.
79
Acknowledgments
First of all, I would like to express my gratitude to Ass. Prof. Niclas Roxhed and
Prof. Göran Stemme for giving me the possibility to join the division of Micro
and Nanosystems at KTH to pursue my doctoral studies. Thank you also for your
trustful and knowledgeable supervision. There is another thing that I would like
to acknowledge to Prof. Stemme: having created a great environment called MST.
I believe that it is rare to find such a pleasant, open and exciting place where to
work and do research. It has been a real pleasure to spend five years here. I
would also like to thank Prof. Frank Niklaus for the precious feedback and
guidance in the microsystem integration project, and Prof. Wouter van der
Wijngaart for the invested time as the advance reviewer of this thesis.
In addition to my supervisors, the results reported in this thesis have been
obtained with the fundamental collaboration of several persons: Miku Laakso,
Xiaojing Wang, Mikolaj Dobielewski, Simone Pagliano, Sara Cavallaro,
Luca De Pietro, Elisa Orso, Eleonora De Luca, Floria Ottonello Briano,
and Marcin Swillo. Thank you for your great collaboration, brilliant
contribution, and generous efforts. Thank you, Cecilia Aronsson and Mikael
Bergqvist for your support, positive mood, and amazing technical skills that
boarder on art. And thank you, James Campion, Alessandro Enrico and
Eleonora De Luca for the fundamental help in proofreading various
manuscripts. I would also like to acknowledge all the professors and teachers I
had over the years for transmitting me something that in a way or another
helped my personal and professional development.
Of course, none of these studies would have been possible without the funding
received from Vinnova, the European Research Council, the Swedish
Foundation for Strategic Research and the Foundation Olle Engkvist
Byggmästare. Thank you for this possibility and for sustaining scientific
development. We also acknowledge our collaborators Debiotech SA and
Brighter AB.
I would then like to immensely thank everyone that passed by MST for
being always so collaborative and for creating a wonderful environment, sharing
both working and fika times, after-work beers, the Friday lunches out, the
beautiful mustaches, the sill lunches and all the other fantastic celebrations, the
kickoff parties (ehm, I meant meetings…), and so on in the last 5 years. I learned
so much from you all, and truly enjoyed the atmosphere you all contributed to
create.
80
Thank you, Sweden for your hospitality. We had some incomprehension
regarding weather, healthcare and some minor issues, but I’m truly appreciating
(most of) my time here. And thank you, EU for making most of the European
continent a welcoming home for our generation.
An honorable mention then goes to my second family in Stockholm: my
teammates (and wags) from the glorious AC Azzurri 1971. We shared some
beautiful successes as well as some very painful moments (figuratively and
literally). Playing and hanging out with you made me feel a bit more at home.
Forza! And thank you to all my old friends, for not forgetting about me even after
so many years apart and always welcoming me back. Thank you, Alex,
Alessandro, Andrea, Giorgio, Luca, Marco, Matteo, Stefano, Umberto,
Valerio and co. And to all the nice people I met over the years, even if ways have
departed.
Finally, but most importantly, a heartfelt thanks to my family, that supported
me towards this achievement with immense affection and unconditional love. To
Marisa, Fulvio, Daniela, Carlo, Dario, and Mara. To Elsa, who I miss and to
whom I cannot express how grateful I am. To Elvio, Ida, and Guido, with whom
I cannot share this moment. To my father, Erminio, for always supporting and
believing in me, challenging me to do better. To my mother, Erica, for her love
and for making me feel brilliant her way, by endlessly repeating, and I quote,
“why didn’t I make a dumber son that couldn’t go anywhere else than staying at
home?!”. To my li’l old sister, Elisa, for the childhood fights and the successive
harmony… as well as for never distracting me from my job with unimportant
communications, such as having gotten married one day. Apart from jokes, grazie
di cuore. Thanks for always being there for me, even if I’m not always there with
you, knowing that I’m not good at it. I’m trying to do something good, and who
knows if I will eventually manage one day. And finally, thank you, Dr. Eng.
Eleonora De Luca, for sharing this journey with me. I mean only the
professional one, indeed. I’ll go for a hand-shake… A tap tap... Ok, maybe a hug
Grazie di esserci.
Federico Ribet,
Stockholm, Sweden
January 17th, 2020
81
Bibliography
[1] S. Pichini, I. Altieri, P. Zuccaro, and R. Pacifici, “Drug monitoring in nonconventional
biological fluids and matrices,” Clin. Pharmacokinet., vol. 30, no. 3, pp. 211–228, 1996.
[2] S. Capiau, J. W. Alffenaar, and C. P. Stove, “Alternative Sampling Strategies for
Therapeutic Drug Monitoring,” in Clinical Challenges in Therapeutic Drug Monitoring:,
Elsevier Inc., 2016, pp. 279–336.
[3] S. R. Corrie, J. W. Coffey, J. Islam, K. A. Markey, and M. A. F. Kendall, “Blood, sweat, and
tears: developing clinically relevant protein biosensors for integrated body fluid analysis,”
Analyst, vol. 140, no. 13, pp. 4350–4364, 2015.
[4] “Skin cross-section.” [Online]. Available:
commons.wikimedia.org/wiki/File:501_Structure_of_the_skin.jpg#filelinks.
[5] K. Hwang, H. Kim, and D. J. Kim, “Thickness of skin and subcutaneous tissue of the free
flap donor sites: A histologic study.,” Microsurgery, vol. 36, no. 1, pp. 54–8, Jan. 2016.
[6] N. Fogh-Andersen, B. M. Altura, B. T. Altura, and O. Siggaard-Andersen, “Composition of
interstitial fluid,” Clin. Chem., vol. 41, no. 10, pp. 1522–1525, 1995.
[7] B. Aussedat, M. Dupire-Angel, R. Gifford, J. C. Klein, G. S. Wilson, and G. Reach,
“Interstitial glucose concentration and glycemia : implications for continuous subcutaneous
glucose monitoring Interstitial glucose concentration and glycemia : implications for
continuous subcutaneous glucose monitoring,” Am. J. Physiol. - Endocrinol. Metab., vol.
278, pp. 716–728, 2000.
[8] M. S. Boyne, D. M. Silver, J. Kaplan, and C. D. Saudek, “Timing of Changes in Interstitial
and Venous Blood Glucose Measured with a Continuous Subcutaneous Glucose Sensor,”
Diabetes, vol. 52, no. 11, pp. 2790–2794, 2003.
[9] E. Cengiz and W. V Tamborlane, “A tale of two compartments: interstitial versus blood
glucose monitoring.,” Diabetes Technol. Ther., vol. 11 Suppl 1, no. SUPPL.1, pp. S11-6, Jun.
2009.
[10] A. K. Yetisen, M. S. Akram, and C. R. Lowe, “Paper-based microfluidic point-of-care
diagnostic devices,” Lab Chip, vol. 13, no. 12, pp. 2210–2251, 2013.
[11] L. Guariguata, D. R. Whiting, I. Hambleton, J. Beagley, U. Linnenkamp, and J. E. Shaw,
“Global estimates of diabetes prevalence for 2013 and projections for 2035,” Diabetes Res.
Clin. Pract., vol. 103, no. 2, pp. 137–149, 2014.
[12] A. Marcellusi et al., “Economic aspects in the management of diabetes in Italy,” BMJ Open
Diabetes Res. Care, vol. 4, no. 1, pp. 1–11, 2016.
[13] B. M. Leon, “Diabetes and cardiovascular disease: Epidemiology, biological mechanisms,
treatment recommendations and future research,” World J. Diabetes, vol. 6, no. 13, p. 1246,
2015.
[14] B. K. Bailes, “Diabetes mellitus and its chronic complications.,” AORN J., vol. 76, no. 2, pp.
82
266–276, 278–282; quiz 283–286, 2002.
[15] J. Wang, “Electrochemical glucose biosensors,” Chem. Rev., vol. 108, pp. 57–69, 2008.
[16] D. Nathan et al., “The effect of intensive treatment of diabetes on the development and
progression of long-term complications in insulin-dependent diabetes mellitus,” N. Engl. J.
Med., vol. 329, no. 14, pp. 977–986, 1993.
[17] F. Cosentino et al., “2019 ESC Guidelines on diabetes, pre-diabetes, and cardiovascular
diseases developed in collaboration with the EASD,” Eur. Heart J., pp. 1–69, 2019.
[18] T. Simonite, “Moore’s Law Is Dead. Now What?,” MIT Technology Review, 2016. [Online].
Available: https://www.technologyreview.com/s/601441/moores-law-is-dead-now-what/.
[19] K. E. Petersen, “Silicon as a mechanical material,” in PROCEEDINGS OF THE IEEE,
1982, vol. 70, no. 5, pp. 58–95.
[20] J. Zou, C. Liu, D. R. Trainor, J. Chen, J. E. Schutt-Ainé, and P. L. Chapman,
“Development of three-dimensional inductors using plastic deformation magnetic assembly
(PDMA),” IEEE Trans. Microw. Theory Tech., vol. 51, no. 4 I, pp. 1067–1075, 2003.
[21] Yong-Kyu Yoon, Bo Pan, J. Papapolymerou, M. Tentzeris, and M. G. Allen, “Surface-
micromachined millimeter-wave antennas,” in The 13th International Conference on Solid-
State Sensors, Actuators and Microsystems (TRANSDUCERS), 2005, pp. 1986–1989.
[22] Y. W. Yi and C. Liu, “Magnetic actuation of hinged microstructures,” J.
Microelectromechanical Syst., vol. 8, no. 1, pp. 10–17, 1999.
[23] Chang Liu et al., “Out-of-plane permalloy magnetic actuators for delta-wing control,” in
Proceedings IEEE Micro Electro Mechanical Systems, 2005, pp. 7–12.
[24] B. K. Gale et al., “A review of current methods in microfluidic device fabrication and future
commercialization prospects,” Inventions, vol. 3, no. 3, 2018.
[25] C. Y. Lee, C. L. Chang, Y. N. Wang, and L. M. Fu, “Microfluidic mixing: A review,” Int. J.
Mol. Sci., vol. 12, no. 5, pp. 3263–3287, 2011.
[26] S. K. Mahadeva, K. Walus, and B. Stoeber, “Paper as a platform for sensing applications
and other devices: A review,” ACS Appl. Mater. Interfaces, vol. 7, no. 16, pp. 8345–8362,
2015.
[27] I. A. Darwish, “Immunoassay Methods and their Applications in Pharmaceutical Analysis:
Basic Methodology and Recent Advances.,” Int. J. Biomed. Sci., vol. 2, no. 3, pp. 217–35,
2006.
[28] G. A. Lenk, G. Stemme, and N. Roxhed, “Delay valving in capillary driven devices based on
dissolvable thin films,” 18th Int. Conf. Miniaturized Syst. Chem. Life Sci. MicroTAS 2014,
pp. 216–218, 2014.
[29] M. Zimmermann, P. Hunziker, and E. Delamarche, “Valves for autonomous capillary
systems,” Microfluid. Nanofluidics, vol. 5, no. 3, pp. 395–402, 2008.
[30] U.-C. Yi and C.-J. Kim, “Characterization of electrowetting actuation on addressable
single-side coplanar electrodes,” J. Micromech. Microeng., vol. 16, no. 16, pp. 2053–2059,
2006.
[31] K. Choi, A. H. C. Ng, R. Fobel, and A. R. Wheeler, “Digital Microfluidics,” Annu. Rev. Anal.
Chem., vol. 5, no. 1, pp. 413–440, 2012.
[32] M. G. Pollack, a D. Shenderov, and R. B. Fair, “Electrowetting-based actuation of droplets
for integrated microfluidics,” Lab Chip, vol. 2, no. 2, p. 96, 2002.
[33] S. K. Cho and H. Moon, “Electrowetting on dielectric (EWOD): New tool for bio/micro fluids
handling,” Biochip J., vol. 2, no. 2, 2009.
83
[34] E. Samiei, M. Tabrizian, and M. Hoorfar, “A review of digital microfluidics as portable
platforms for lab-on a-chip applications,” Lab Chip, vol. 16, no. 13, pp. 2376–2396, 2016.
[35] H. H. Shen, S. K. Fan, C. J. Kim, and D. J. Yao, “EWOD microfluidic systems for
biomedical applications,” Microfluid. Nanofluidics, vol. 16, no. 5, pp. 965–987, 2014.
[36] S. K. Cho, H. Moon, and C. J. Kim, “Creating, transporting, cutting, and merging liquid
droplets by electrowetting-based actuation for digital microfluidic circuits,” J.
Microelectromechanical Syst., vol. 12, no. 1, pp. 70–80, 2003.
[37] P. Sen and C.-J. Kim, “A fast liquid-metal droplet switch using EWOD,” 2007 IEEE 20th
Int. Conf. Micro Electro Mech. Syst., no. January, pp. 767–770, 2007.
[38] P. Sen and C. J. Kim, “A liquid-solid direct contact low-loss RF micro switch,” J.
Microelectromechanical Syst., vol. 18, no. 5, pp. 990–997, 2009.
[39] Y. Zhao and S. K. Cho, “DROPLET MANIPULATION ON HIGH-OPENING MICRO
FILTER MESHES,” 2008.
[40] P. Müller, R. Feuerstein, and H. Zappe, “Integrated Optofluidic Iris,” J.
Microelectromechanical Syst., vol. 21, no. 5, pp. 1156–1164, 2012.
[41] P. Muller, A. Kloss, P. Liebetraut, W. Mönch, and H. Zappe, “A fully integrated optofluidic
attenuator,” J. Micromechanics Microengineering, vol. 21, p. 125027, 2011.
[42] P. Dubois et al., “Ionic liquid droplet as e-microreactor.,” Anal. Chem., vol. 78, no. 14, pp.
4909–17, 2006.
[43] X. Hu et al., “Ionic liquid based variable focus lenses,” Soft Matter, vol. 7, no. 13, p. 5941,
2011.
[44] Y. S. Nanayakkara, H. Moon, and D. W. Armstrong, “A tunable ionic liquid based RC filter
using electrowetting: A new concept,” ACS Appl. Mater. Interfaces, vol. 2, no. 7, pp. 1785–
1787, 2010.
[45] T. Liu, P. Sen, and C. J. Kim, “Characterization of nontoxic liquid-metal alloy galinstan for
applications in microdevices,” J. Microelectromechanical Syst., vol. 21, no. 2, pp. 443–450,
2012.
[46] L. Ventrelli, L. Marsilio Strambini, and G. Barillaro, “Microneedles for Transdermal
Biosensing: Current Picture and Future Direction,” Adv. Healthc. Mater., vol. 4, no. 17, pp.
2606–2640, 2015.
[47] H. R. Jeong, H. S. Lee, I. J. Choi, and J. H. Park, “Considerations in the use of
microneedles: pain, convenience, anxiety and safety,” J. Drug Target., vol. 25, no. 1, pp.
29–40, 2017.
[48] M. T. C. Mccrudden, E. Mcalister, A. J. Courtenay, P. González-Vázquez, T. R. Raj Singh,
and R. F. Donnelly, “Microneedle applications in improving skin appearance,” Exp.
Dermatol., vol. 24, no. 8, pp. 561–566, 2015.
[49] S. Bhatnagar, K. Dave, and V. V. K. Venuganti, “Microneedles in the clinic,” J. Control.
Release, vol. 260, no. March, pp. 164–182, 2017.
[50] N. S. Rejinold, J. H. Shin, H. Y. Seok, and Y. C. Kim, “Biomedical applications of
microneedles in therapeutics: Recent advancements and implications in drug delivery,”
Expert Opin. Drug Deliv., vol. 13, no. 1, pp. 109–131, 2016.
[51] Y. C. Kim, J. H. Park, and M. R. Prausnitz, “Microneedles for drug and vaccine delivery,”
Adv. Drug Deliv. Rev., vol. 64, no. 14, pp. 1547–1568, 2012.
[52] B. Stoeber and D. Liepmann, “Arrays of hollow out-of-plane microneedles for drug delivery,”
J. Microelectromechanical Syst., vol. 14, no. 3, pp. 472–479, 2005.
[53] T. N. Tarbox, A. B. Watts, Z. Cui, and R. O. Williams, “An update on
84
coating/manufacturing techniques of microneedles,” Drug Deliv. Transl. Res., vol. 8, no. 6,
pp. 1828–1843, 2018.
[54] A. E. G. Cass and S. Sharma, Microneedle Enzyme Sensor Arrays for Continuous In Vivo
Monitoring, 1st ed., vol. 589. Elsevier Inc., 2017.
[55] C. G. Li, H. A. Joung, H. Noh, M. B. Song, M. G. Kim, and H. Jung, “One-touch-activated
blood multidiagnostic system using a minimally invasive hollow microneedle integrated
with a paper-based sensor,” Lab Chip, vol. 15, no. 16, pp. 3286–3292, 2015.
[56] M. Rajabi et al., “Flexible and stretchable microneedle patches with integrated rigid
stainless steel microneedles for transdermal biointerfacing,” PLoS One, vol. 11, no. 12, pp.
1–13, 2016.
[57] B. Q. Tran et al., “Proteomic Characterization of Dermal Interstitial Fluid Extracted Using
a Novel Microneedle-Assisted Technique,” J. Proteome Res., vol. 17, no. 1, pp. 479–485,
2018.
[58] N. Roxhed, T. C. Gasser, P. Griss, G. A. Holzapfel, and G. Stemme, “Penetration-enhanced
ultrasharp microneedles and prediction on skin interaction for efficient transdermal drug
delivery,” J. Microelectromechanical Syst., vol. 16, no. 6, pp. 1429–1440, 2007.
[59] R. L. Smith et al., “Silicon microneedle array for minimally invasive human health
monitoring,” SPIE, vol. 1049102, no. February 2018, 2019.
[60] J. D. Zahn, D. Trebotich, and D. Liepmann, “Microfabricated microdialysis microneedles
for continuous medical monitoring,” 1st Annu. Int. IEEE-EMBS Spec. Top. Conf.
Microtechnologies Med. Biol. - Proc., pp. 375–380, 2000.
[61] H. Gardeniers, J. W. Berenschot, M. Der Boer, and M. Hefetz, “Silicon micromachined
hollow microneedles for transdermal liquid transfer,” vol. 12, no. February 2002, pp. 855–
862, 2014.
[62] A. M. Römgens, D. L. Bader, J. A. Bouwstra, F. P. T. Baaijens, and C. W. J. Oomens,
“Monitoring the penetration process of single microneedles with varying tip diameters,” J.
Mech. Behav. Biomed. Mater., vol. 40, pp. 397–405, 2014.
[63] E. Larrañeta, R. E. M. Lutton, A. D. Woolfson, and R. F. Donnelly, “Microneedle arrays as
transdermal and intradermal drug delivery systems: Materials science, manufacture and
commercial development,” Mater. Sci. Eng. R, vol. 104, pp. 1–32, 2016.
[64] A. Aksit et al., “In-vitro perforation of the round window membrane via direct 3-D printed
microneedles,” Biomed. Microdevices, vol. 20, no. 2, 2018.
[65] P. R. Miller, X. Xiao, I. Brener, D. B. Burckel, R. Narayan, and R. Polsky, “Microneedle-
based transdermal sensor for on-chip potentiometric determination of K+,” Adv. Healthc.
Mater., vol. 3, no. 6, pp. 876–881, 2014.
[66] S. D. Gittard, A. Ovsianikov, B. N. Chichkov, A. Doraiswamy, and R. J. Narayan, “Two
Photon Polymerization of Microneedles for Transdermal Drug Delivery,” Expert Opin Drug
Deliv, vol. 7, no. 4, pp. 513–533, 2010.
[67] S. Sharma, Z. Huang, M. Rogers, M. Boutelle, and A. E. G. Cass, “Evaluation of a
minimally invasive glucose biosensor for continuous tissue monitoring,” Anal. Bioanal.
Chem., vol. 408, no. 29, pp. 8427–8435, 2016.
[68] K. Takeuchi, N. Takama, B. Kim, K. Sharma, P. Ruther, and O. Paul, “A porous
microneedle array connected to microfluidic system for ISF collection,” 2018 IEEE CPMT
Symp. Japan, ICSJ 2018, pp. 85–88, 2019.
[69] M. Wang, L. Hu, and C. Xu, “Recent advances in the design of polymeric microneedles for
transdermal drug delivery and biosensing,” Lab Chip, vol. 17, no. 8, pp. 1373–1387, 2017.
85
[70] H. Chang et al., “A Swellable Microneedle Patch to Rapidly Extract Skin Interstitial Fluid
for Timely Metabolic Analysis,” Adv. Mater., vol. 29, no. 37, pp. 1–8, 2017.
[71] J. Janata, Principles of chemical sensors, 2nd ed. Springer US, 2009.
[72] Y. Wang, H. Xu, J. Zhang, and G. Li, “Electrochemical sensors for clinic analysis,” Sensors,
vol. 8, no. 4, pp. 2043–2081, 2008.
[73] C. Hamann, A. Hamnet, and W. Vielstich, Electrochemistry, 2nd ed. Wiley-VHC, 2007.
[74] L. Bard, A.Faulkner, Electrochemical Methods, 2nd ed. Wiley, 2001.
[75] H. K. Gatty, MEMS-based electrochemical gas sensors and wafer-level methods. 2015.
[76] “Veryday breath asthma analyzer,” 2019. [Online]. Available:
https://veryday.com/case/aerocrine/.
[77] H. K. Gatty, G. Stemme, and N. Roxhed, “A wafer-level liquid cavity integrated
amperometric gas sensor with ppb-level nitric oxide gas sensitivity,” J. Micromechanics
Microengineering, vol. 25, p. 105013, 2015.
[78] S. K. Vashist, “Non-invasive glucose monitoring technology in diabetes management: A
review,” Anal. Chim. Acta, vol. 750, pp. 16–27, 2012.
[79] R. Gifford, “Continuous glucose monitoring: 40 years, what we’ve learned and what’s next,”
ChemPhysChem, vol. 14, no. 10, pp. 2032–2044, 2013.
[80] J. Wang, “Electrochemical Glucose Biosensors,” Chem. Rev., vol. 108, no. 2, pp. 814–825,
2008.
[81] A. Heller and B. Feldman, “Electrochemical glucose sensors and their applications in
diabetes management,” Chem. Rev., vol. 108, no. 7, pp. 2482–2505, 2008.
[82] A. Heller and B. Feldman, “Electrochemistry in diabetes management,” Acc. Chem. Res.,
vol. 43, no. 7, pp. 963–973, 2010.
[83] C. Li, C. H. Ahn, L. A. Shutter, and R. K. Narayan, “Toward real-time continuous brain
glucose and oxygen monitoring with a smart catheter,” Biosens. Bioelectron., vol. 25, no. 1,
pp. 173–178, 2009.
[84] S. K. Vashist, D. Zheng, K. Al-Rubeaan, J. H. T. Luong, and F. S. Sheu, “Technology
behind commercial devices for blood glucose monitoring in diabetes management: A
review,” Anal. Chim. Acta, vol. 703, no. 2, pp. 124–136, 2011.
[85] M. Waleed Shinwari, D. Zhitomirsky, I. A. Deen, P. R. Selvaganapathy, M. Jamal Deen,
and D. Landheer, “Microfabricated reference electrodes and their biosensing applications,”
Sensors, vol. 10, no. 3, pp. 1679–1715, 2010.
[86] V. M. Tolosa, K. M. Wassum, N. T. Maidment, and H. G. Monbouquette,
“Electrochemically deposited iridium oxide reference electrode integrated with an
electroenzymatic glutamate sensor on a multi-electrode array microprobe,” Biosens.
Bioelectron., vol. 42, no. 1, pp. 256–260, 2013.
[87] A. Weltin, J. Kieninger, B. Enderle, A. K. Gellner, B. Fritsch, and G. A. Urban, “Polymer-
based, flexible glutamate and lactate microsensors for in vivo applications,” Biosens.
Bioelectron., vol. 61, pp. 192–199, 2014.
[88] B. Feldman, R. Brazg, S. Schwartz, and R. Weinstein, “A continuous glucose sensor based
on wired enzyme technology -- results from a 3-day trial in patients with type 1 diabetes.,”
Diabetes Technol. Ther., vol. 5, no. 5, pp. 769–79, Jan. 2003.
[89] H. Yang et al., “An iridium oxide reference electrode for use in microfabricated biosensors
and biochips.,” Lab Chip, vol. 4, pp. 42–46, 2004.
[90] F. Moussy and D. J. Harrison, “Prevention of the Rapid Degradation of Subcutaneously
86
Implanted Ag/AgCl Reference Electrodes Using Polymer Coatings,” Anal. Chem., vol. 66,
no. 5, pp. 674–679, 1994.
[91] H. Suzuki, A. Hiratsuka, S. Sasaki, and I. Karube, “Problems associated with the thin-film
Ag/AgCl reference electrode and a novel structure with improved durability,” Sensors
Actuators B Chem., vol. 46, no. 2, pp. 104–113, 1998.
[92] R. K. Franklin et al., “Iridium oxide reference electrodes for neurochemical sensing with
MEMS microelectrode arrays,” Proc. IEEE Sensors, vol. 2005, pp. 1400–1403, 2005.
[93] Y. Marunaka, “Roles of interstitial fluid pH in diabetes mellitus: Glycolysis and
mitochondrial function.,” World J. Diabetes, vol. 6, no. 1, pp. 125–35, 2015.
[94] K. Yamanaka, “Anodically electrodeposited iridium oxide films (AEIROF’s) from alkaline
solutions for electrochromic display devices,” Jpn. J. Appl. Phys., vol. 28, no. 4, pp. 632–
637, 1989.
[95] S. A. M. Marzouk, S. Ufer, R. P. Buck, T. A. Johnson, L. A. Dunlap, and W. E. Cascio,
“Electrodeposited Iridium Oxide pH Electrode for Measurement of Extracellular
Myocardial Acidosis during Acute Ischemia,” Anal. Chem., vol. 70, no. 23, pp. 5054–5061,
Dec. 1998.
[96] K. E. Toghill and R. G. Compton, “Electrochemical non-enzymatic glucose sensors: A
perspective and an evaluation,” Int. J. Electrochem. Sci., vol. 5, no. 9, pp. 1246–1301, 2010.
[97] W. Schuhmann, “Electron-transfer pathways in amperometric biosensors. Ferrocene-
modified enzymes entrapped in conducting-polymer layers,” Biosens. Bioelectron., vol. 10,
no. 1–2, pp. 181–193, 1995.
[98] A. Chaubey and B. D. Malhotra, “Mediated biosensors,” Biosens. Bioelectron., vol. 17, no.
6–7, pp. 441–456, 2002.
[99] E. Csöregi, D. W. Schmidtke, and a Heller, “Design and optimization of a selective
subcutaneously implantable glucose electrode based on ‘wired’ glucose oxidase.,” Anal.
Chem., vol. 67, no. 7, pp. 1240–1244, 1995.
[100] A. Weltin, B. Enderle, J. Kieninger, and G. A. Urban, “Multiparametric, flexible
microsensor platform for metabolic monitoring in vivo,” IEEE Sens. J., vol. 14, no. 10, pp.
3345–3351, 2014.
[101] S. Park, H. Boo, and T. D. Chung, “Electrochemical non-enzymatic glucose sensors.,” Anal.
Chim. Acta, vol. 556, no. 1, pp. 46–57, Jan. 2006.
[102] M. Kaisti, “Detection principles of biological and chemical FET sensors,” Biosens.
Bioelectron., vol. 98, no. June, pp. 437–448, 2017.
[103] Y. Miyahara, T. Moriizumi, and K. Ichimura, “Integrated enzyme fets for simultaneous
detections of urea and glucose,” Sensors and Actuators, vol. 7, no. 1, pp. 1–10, 1985.
[104] N. Ito et al., “A novel blood glucose monitoring system based on an ISFET biosensor and its
application to a human 75 g oral glucose tolerance test,” Sensors Actuators B. Chem., vol. 1,
no. 1–6, pp. 488–490, 1990.
[105] A. Caduff, E. Hirt, Y. Feldman, Z. Ali, and L. Heinemann, “First human experiments with
a novel non-invasive, non-optical continuous glucose monitoring system,” Biosens.
Bioelectron., vol. 19, no. 3, pp. 209–217, 2003.
[106] N. S. Oliver, C. Toumazou, A. E. G. Cass, and D. G. Johnston, “Glucose sensors: A review
of current and emerging technology,” Diabet. Med., vol. 26, no. 3, pp. 197–210, 2009.
[107] M. Kaiho, J. Sawayama, Y. Morimoto, and S. Takeuchi, “Glucose-Responsive Fluorescent
Hydrogel,” pp. 534–537, 2017.
[108] “Abbott Freestyle Libre,” 2019. [Online]. Available: https://www.freestylelibre.se/.
87
[109] “Medtronic CGM,” 2019. [Online]. Available: https://www.medtronic-diabetes.co.uk/about-
diabetes/continuous-glucose-monitoring.
[110] “Dexcom CGM,” 2019. [Online]. Available: https://www.dexcom.com/en-IN.
[111] “Eversense Diabetes Monitoring,” 2019. [Online]. Available:
https://www.eversensediabetes.com/sensor.
[112] R. T. Kurnik, B. Berner, J. Tamada, and R. O. Potts, “Design and Simulation of a Reverse
Iontophoretic Glucose Monitoring Device,” J. Electrochem. Soc., vol. 145, no. 12, pp. 4119–
4125, 1998.
[113] A. J. Bandodkar and J. Wang, “Non-invasive wearable electrochemical sensors: A review,”
Trends Biotechnol., vol. 32, no. 7, pp. 363–371, 2014.
[114] A. J. Bandodkar, W. Jia, C. Yardimci, X. Wang, J. Ramirez, and J. Wang, “Tattoo-based
noninvasive glucose monitoring: A proof-of-concept study,” Anal. Chem., vol. 87, no. 1, pp.
394–398, 2015.
[115] W. Groenendaal, K. a Schmidt, G. von Basum, N. a W. van Riel, and P. a J. Hilbers,
“Modeling glucose and water dynamics in human skin.,” Diabetes Technol. Ther., vol. 10,
no. 4, pp. 283–93, 2008.
[116] P. M. Wang, M. Cornwell, and M. R. Prausnitz, “Minimally invasive extraction of dermal
interstitial fluid for glucose monitoring using microneedles.,” Diabetes Technol. Ther., vol.
7, no. 1, pp. 131–41, 2005.
[117] A. El-Laboudi, N. S. Oliver, A. Cass, and D. Johnston, “Use of microneedle array devices
for continuous glucose monitoring: a review.,” Diabetes Technol. Ther., vol. 15, no. 1, pp.
101–15, 2013.
[118] Y. Wang, S. Vaddiraju, B. Gu, F. Papadimitrakopoulos, and D. J. Burgess, “Foreign Body
Reaction to Implantable Biosensors: Effects of Tissue Trauma and Implant Size,” J.
Diabetes Sci. Technol., vol. 9, no. 5, pp. 966–977, 2015.
[119] N. Roxhed, B. Samel, L. Nordquist, P. Griss, and G. Stemme, “Painless drug delivery
through microneedle-based transdermal patches featuring active infusion,” IEEE Trans.
Biomed. Eng., vol. 55, no. 3, pp. 1063–1071, 2008.
[120] K. Ita, “Transdermal delivery of drugs with microneedles—potential and challenges,”
Pharmaceutics, vol. 7, no. 3, pp. 90–105, 2015.
[121] K. Van Der Maaden, W. Jiskoot, and J. Bouwstra, “Microneedle technologies for
(trans)dermal drug and vaccine delivery,” J. Control. Release, vol. 161, no. 2, pp. 645–655,
2012.
[122] G. E. Gattiker, K. V. I. S. Kaler, and M. P. Mintchev, “Electronic Mosquito: designing a
semi-invasive Microsystem for blood sampling, analysis and drug delivery applications,”
Microsyst. Technol., vol. 12, no. 1–2, pp. 44–51, 2005.
[123] P. Miller et al., “Towards an Integrated Microneedle Total Analysis Chip for Protein
Detection,” Electroanalysis, vol. 28, no. 6, pp. 1305–1310, 2016.
[124] F. Ricci, D. Moscone, and G. Palleschi, “Ex vivo continuous glucose monitoring with
microdialysis technique: The example of GlucoDay,” IEEE Sens. J., vol. 8, no. 1, pp. 63–70,
2008.
[125] H. W. Vesper, P. M. Wang, E. Archibold, M. R. Prausnitz, and G. L. Myers, “Assessment of
trueness of a glucose monitor using interstitial fluid and whole blood as specimen matrix.,”
Diabetes Technol. Ther., vol. 8, no. 1, pp. 76–80, 2006.
[126] H. Yu, D. Li, R. C. Roberts, K. Xu, and N. C. Tien, “An interstitial fluid transdermal
extraction system for continuous glucose monitoring,” J. Microelectromechanical Syst., vol.
21, no. 4, pp. 917–925, 2012.
88
[127] S. Zimmermann, D. Fienbork, A. W. Flounders, and D. Liepmann, “In-device enzyme
immobilization: Wafer-level fabrication of an integrated glucose sensor,” Sensors Actuators,
B Chem., vol. 99, no. 1, pp. 163–173, 2004.
[128] S. Zimmermann, D. Fienbork, B. Stoeber, A. W. Flounders, and D. Liepmann, “A
microneedle-based glucose monitor: Fabricated on a wafer-level using in-device enzyme
immobilization,” TRANSDUCERS 2003 - 12th Int. Conf. Solid-State Sensors, Actuators
Microsystems, Dig. Tech. Pap., vol. 1, no. May 2014, pp. 99–102, 2003.
[129] S. A. Ranamukhaarachchi, C. Padeste, U. O. Hafeli, B. Stoeber, and V. J. Cadarso, “Design
considerations of a hollow microneedle-optofluidic biosensing platform incorporating
enzyme-linked assays,” J. Micromechanics Microengineering, vol. 28, no. 2, 2018.
[130] A. Jina et al., “Design, development, and evaluation of a novel microneedle array-based
continuous glucose monitor.,” J. Diabetes Sci. Technol., vol. 8, no. 3, pp. 483–7, 2014.
[131] E. V. Mukerjee, S. D. Collins, R. R. Isseroff, and R. L. Smith, “Microneedle array for
transdermal biological fluid extraction and in situ analysis,” Sensors Actuators, A Phys.,
vol. 114, no. 2–3, pp. 267–275, 2004.
[132] B. Chua, S. P. Desai, M. J. Tierney, J. A. Tamada, and A. N. Jina, “Effect of microneedles
shape on skin penetration and minimally invasive continuous glucose monitoring in vivo,”
Sensors Actuators, A Phys., vol. 203, pp. 373–381, 2013.
[133] S. Sharma et al., “A pilot study in humans of microneedle sensor arrays for continuous
glucose monitoring,” Anal. Methods, vol. 10, no. 18, pp. 2088–2095, 2018.
[134] J. R. Windmiller et al., “Bicomponent Microneedle Array Biosensor for Minimally-Invasive
Glutamate Monitoring,” Electroanalysis, vol. 23, no. 10, pp. 2302–2309, 2011.
[135] P. Bollella, S. Sharma, A. E. G. Cass, and R. Antiochia, “Microneedle-based biosensor for
minimally-invasive lactate detection,” Biosens. Bioelectron., 2018.
[136] S. Yao, M. Wang, and M. Madou, “A pH Electrode Based on Melt-Oxidized Iridium Oxide,”
J. Electrochem. Soc., vol. 148, no. 4, p. H29, 2001.
[137] T. C. Dunn, R. C. Eastman, and J. A. Tamada, “Rates of glucose change measured by blood
glucose meter and the gluco watch biographer during day, night, and around mealtimes,”
Diabetes Care, vol. 27, no. 9, pp. 2161–2165, 2004.
[138] M. L. Crichton, B. C. Donose, X. Chen, A. P. Raphael, H. Huang, and M. A. F. Kendall,
“The viscoelastic, hyperelastic and scale dependent behaviour of freshly excised individual
skin layers,” Biomaterials, vol. 32, no. 20, pp. 4670–4681, 2011.
[139] F. J. Verbaan et al., “Improved piercing of microneedle arrays in dermatomed human skin
by an impact insertion method,” J. Control. Release, vol. 128, no. 1, pp. 80–88, 2008.
[140] J. Juodkazyte, B. ??ebeka, I. Valsiunas, and K. Juodkazis, “Iridium anodic oxidation to
Ir(III) and Ir(IV) hydrous oxides,” Electroanalysis, vol. 17, no. 11, pp. 947–952, 2005.
[141] L. Setti, A. Fraleoni-Morgera, B. Ballarin, A. Filippini, D. Frascaro, and C. Piana, “An
amperometric glucose biosensor prototype fabricated by thermal inkjet printing,” Biosens.
Bioelectron., vol. 20, no. 10 SPEC. ISS., pp. 2019–2026, 2005.
[142] Y. H. Yun et al., “A glucose sensor fabricated by piezoelectric inkjet printing of conducting
polymers and bienzymes.,” Anal. Sci., vol. 27, no. 4, p. 375, 2011.
[143] K. Aoki et al., “Microassembly of semiconductor three-dimensional photonic crystals,” Nat.
Mater., vol. 2, no. 2, pp. 117–121, 2003.
[144] N. Dechev, S. Member, W. L. Cleghorn, and J. K. Mills, “Microassembly of 3-D
Microstructures Using a Compliant , Passive Microgripper,” J. Microelectromechanical
Syst., vol. 13, no. 2, pp. 176–189, 2004.
89
[145] K. Mølhave, T. Wich, A. Kortschack, and P. Bøggild, “Pick-and-place nanomanipulation
using microfabricated grippers,” Nanotechnology, vol. 17, no. 10, pp. 2434–2441, 2006.
[146] S. Zimmermann, T. Tiemerding, and S. Fatikow, “Automated Robotic Manipulation of
Individual Colloidal Particles Using Vision-Based Control,” IEEE/ASME Trans.
Mechatronics, vol. 20, no. 5, pp. 2031–2038, Oct. 2015.
[147] M. C. Tarhan, R. Yokokawa, L. Jalabert, D. Collard, and H. Fujita, “Pick-and-Place
Assembly of Single Microtubules,” Small, vol. 13, no. 32, p. 1701136, Aug. 2017.
[148] A. Carlson, A. M. Bowen, Y. Huang, R. G. Nuzzo, and J. A. Rogers, “Transfer printing
techniques for materials assembly and micro/nanodevice fabrication,” Adv. Mater., vol. 24,
no. 39, pp. 5284–5318, 2012.
[149] B. Corbett, R. Loi, W. Zhou, D. Liu, and Z. Ma, “Transfer print techniques for
heterogeneous integration of photonic components,” Prog. Quantum Electron., vol. 52, pp.
1–17, Mar. 2017.
[150] C. Liu et al., “Out-of-plane permalloy magnetic actuators for delta-wing control,” in
Proceedings of IEEE Micro Electro Mechanical Systems, 2005, p. 7.
[151] J. Zou, J. Chen, C. Liu, and J. E. Schutt-Ainé, “Plastic deformation magnetic assembly
(PDMA) of out-of-plane microstructures: Technology and application,” J.
Microelectromechanical Syst., vol. 10, no. 2, pp. 302–309, 2001.
[152] E. Iwase and I. Shimoyama, “Multistep sequential batch assembly of three-dimensional
ferromagnetic microstructures with elastic hinges,” J. Microelectromechanical Syst., vol.
14, no. 6, pp. 1265–1271, 2005.
[153] A. Qualtieri, F. Rizzi, G. Epifani, A. Ernits, M. Kruusmaa, and M. De Vittorio,
“Microelectronic Engineering Parylene-coated bioinspired artificial hair cell for liquid flow
sensing,” Microelectron. Eng., vol. 98, pp. 516–519, 2012.
[154] L. Li, J. Zawadzka, D. Uttamchandani, and S. Member, “Integrated Self-Assembling and
Holding Technique Applied to a 3-D MEMS Variable Optical Attenuator,” J.
Microelectromechanical Syst., vol. 13, no. 1, pp. 83–90, 2004.
[155] J. H. Comtois and V. M. Bright, “Applications for surface-micromachined polysilicon
thermal actuators and arrays,” Sensors Actuators, A Phys., vol. 58, no. 1, pp. 19–25, 1997.
[156] T. Ebefors, E. Kalvesten, C. Vieider, and G. Stemme, “New robust small radius joints
based on thermal shrinkage of polyimide in V-grooves,” J. Micromech. Microeng., vol. 8, pp.
188–194, 1998.
[157] R. R. A. Syms, “Equilibrium of Hinged and Hingeless Structures Rotated Using Surface
Tension Forces,” J. Microelectromechanical Syst., vol. 4, no. 4, pp. 177–184, 1995.
[158] R. R. A. Syms, S. Member, E. M. Yeatman, V. M. Bright, and G. M. Whitesides, “Surface
tension-powered self-assembly of microstructures - The state-of-the-art,” J.
Microelectromechanical Syst., vol. 12, no. 4, pp. 387–417, 2003.
[159] K. L. Scott, T. Hirano, H. Yang, H. Singh, R. T. Howe, and A. M. Niknejad, “High-
Performance Inductors Using Capillary Based Fluidic Self-Assembly,” J.
Microelectromechanical Syst., vol. 13, no. 2, pp. 300–309, 2004.
[160] J. H. Hoo, K. S. Park, and R. Baskaran, “Template-based self-assembly for silicon chips
and 01005 surface-mount and Karl F B ohringer,” J. Micromechanics Microengineering, vol.
24, p. 045018, 2014.
[161] C. G. J. Fonstad, “Magnetically-Assisted Statistical Assembly - a new heterogeneous
integration technique,” MIT Technical Report. 2002.
[162] S. B. Shetye, I. Eskinazi, and D. P. Arnold, “Self-assembly of millimeter-scale components
using integrated micromagnets,” IEEE Trans. Magn., vol. 44, no. 11, pp. 4293–4296, 2008.
90
[163] Q. Ramadan, Y. S. Uk, and K. Vaidyanathan, “Large scale microcomponents assembly
using an external magnetic array,” Appl. Phys. Lett., vol. 90, no. 17, p. 172502, Apr. 2007.
[164] Y. B. Lin et al., “Multi-force-field assisted self-assembly process of thermoelectric device,”
in The 17th International Conference TRANSDUCERS & EUROSENSORS, 2013, pp. 818–
821.
[165] J. H. Hoo, K. S. Park, Ç. Varel, R. Baskaran, and K. F. Böhringer, “Wafer-level high
density integration of surface mount technology components in through-silicon trenches,”
in IEEE 25th International Conference on Micro Electro Mechanical Systems (MEMS),
2012, pp. 373–376.
[166] A. C. Fischer, S. J. Bleiker, T. Haraldsson, N. Roxhed, G. Stemme, and F. Niklaus, “Very
high aspect ratio through-silicon vias (TSVs) fabricated using automated magnetic
assembly of nickel wires,” J. Micromechanics Microengineering, vol. 22, no. 10, p. 105001,
2012.
[167] S. J. Bleiker et al., “High-Aspect-ratio through silicon vias for high-frequency application
fabricated by magnetic assembly of gold-coated nickel wires,” IEEE Trans. Components,
Packag. Manuf. Technol., vol. 5, no. 1, pp. 21–27, 2015.
[168] M. Asiatici, M. J. Laakso, A. C. Fischer, G. Stemme, and F. Niklaus, “Through silicon vias
with invar metal conductor for high-temperature applications,” J. Microelectromechanical
Syst., vol. 26, no. 1, pp. 158–168, 2017.
[169] K. L. Scott, T. Hirano, H. Yang, H. Singh, R. T. Howe, and A. M. Niknejad, “High-
Performance Inductors Using Capillary Based Fluidic Self-Assembly,” J.
Microelectromechanical Syst., vol. 13, no. 2, pp. 300–309, Apr. 2004.
[170] S. H. Tsang, “Automated assembly of hingeless 90◦ out-of- microstructures,” J.
Micromechanics Microengineering, vol. 17, p. 1314, 2007.
[171] A. C. Fischer, J. G. Korvink, N. Roxhed, G. Stemme, U. Wallrabe, and F. Niklaus,
“Unconventional applications of wire bonding create opportunities for microsystem
integration,” J. Micromechanics Microengineering, vol. 23, no. 8, 2013.
[172] A. Hirshberg and D. Elata, “Direct wire-bonding of silicon devices without metal pads,”
Proc. IEEE Int. Conf. Micro Electro Mech. Syst., pp. 757–760, 2009.
[173] a C. Fischer, S. J. Bleiker, T. Haraldsson, N. Roxhed, G. Stemme, and F. Niklaus, “Very
high aspect ratio through-silicon vias (TSVs) fabricated using automated magnetic
assembly of nickel wires,” J. Micromechanics Microengineering, vol. 22, no. 10, p. 105001,
2012.
[174] S. J. Bleiker et al., “High-Aspect-ratio through silicon vias for high-frequency application
fabricated by magnetic assembly of gold-coated nickel wires,” IEEE Trans. Components,
Packag. Manuf. Technol., vol. 5, no. 1, pp. 21–27, 2015.
[175] H. J. Cho and C. H. Ahn, “Magnetic Microactuators—Techniques and Applications,” in
MEMS/NEMS, Boston, MA: Springer US, 2006, pp. 1631–1654.
[176] C. A. Grimes and J. K. Lumpp, “The soft magnetic properties of stripes fabricated using
laser ablation of multilayer thin films,” J. Appl. Phys., vol. 79, no. 8, p. 5497, Aug. 1996.
[177] H. Le Gall et al., “Two-dimensional arrays of magnetic stripes and dots patterned by laser
ablation,” J. Appl. Phys., vol. 81, no. 8, pp. 4727–4729, Apr. 1997.
[178] A. García-Arribas et al., “Tailoring the magnetic anisotropy of thin film permalloy
microstrips by combined shape and induced anisotropies,” Eur. Phys. J. B, vol. 86, no. 4, p.
136, Apr. 2013.
[179] C. Pawashe, S. Floyd, and M. Sitti, “Modeling and Experimental Characterization of an
Untethered Magnetic Micro-Robot,” Int. J. Rob. Res., vol. 28, no. 8, pp. 1077–1094, Aug.
91
2009.
[180] US Department of Defense, “Mil-Std-883G,” Test Method Standard - Microcircuits, no. w/
CHANGE 3. 2018.
[181] T. K. L. Kiang, S. A. Ranamukhaarachchi, and M. H. H. Ensom, “Revolutionizing
therapeutic drug monitoring with the use of interstitial fluid and microneedles technology,”
Pharmaceutics, vol. 9, no. 4, 2017.
[182] T. K. L. Kiang, V. Schmitt, M. H. H. Ensom, B. Chua, and U. O. Häfeli, “Therapeutic Drug
Monitoring in Interstitial Fluid: A Feasibility Study Using a Comprehensive Panel of
Drugs,” J. Pharm. Sci., vol. 101, no. 12, pp. 4642–4652, Dec. 2012.
[183] M. Wagner, D. Tonoli, E. Varesio, and G. Hopfgartner, “THE USE OF MASS
SPECTROMETRY TO ANALYZE DRIED BLOOD SPOTS,” Mass Spectrom. Rev., vol. 35,
pp. 361–438, 2016.
[184] G. Lenk, S. Sandkvist, A. Pohanka, G. Stemme, O. Beck, and N. Roxhed, “A disposable
sampling device to collect volume-measured DBS directly from a fingerprick onto DBS
paper,” Bioanalysis, vol. 7, no. 16, pp. 2085–2094, 2015.
[185] P. Denniff and N. Spooner, “Volumetric absorptive microsampling: A dried sample
collection technique for quantitative bioanalysis,” Anal. Chem., vol. 86, no. 16, pp. 8489–
8495, 2014.
[186] L. A. Leuthold et al., “New microfluidic-based sampling procedure for overcoming the
hematocrit problem associated with dried blood spot analysis,” Anal. Chem., vol. 87, no. 4,
pp. 2068–2071, 2015.
[187] B. Wilcken and V. Wiley, “Newborn screening,” Pathology, vol. 40, no. 2, pp. 104–115, 2008.
[188] T. Li, A. Barnett, K. L. Rogers, and Y. B. Gianchandani, “A blood sampling microsystem
for pharmacokinetic applications: Design, fabrication, and initial results,” Lab Chip, vol. 9,
no. 24, pp. 3495–3503, 2009.
[189] C. G. Li, H. A. Joung, H. Noh, M. B. Song, M. G. Kim, and H. Jung, “One-touch-activated
blood multidiagnostic system using a minimally invasive hollow microneedle integrated
with a paper-based sensor,” Lab Chip, vol. 15, no. 16, pp. 3286–3292, 2015.
[190] T. M. Blicharz et al., “Microneedle-based device for the one-step painless collection of
capillary blood samples,” Nat. Biomed. Eng., vol. 2, no. March, 2018.
[191] P. P. Samant and M. R. Prausnitz, “Mechanisms of sampling interstitial fluid from skin
using a microneedle patch,” in Proc Natl Acad Sci U S A, 2018.
[192] U. Kiistala, “Suction blister device for separation of viable epidermis from dermis.,” J.
Invest. Dermatol., vol. 50, no. 2, pp. 129–137, 1968.
[193] J. Kool et al., “Suction blister fluid as potential body fluid for biomarker proteins,”
Proteomics, vol. 7, no. 20, pp. 3638–3650, 2007.
[194] A. C. Müller et al., “A comparative proteomic study of human skin suction blister fluid
from healthy individuals using immunodepletion and iTRAQ labeling,” J. Proteome Res.,
vol. 11, no. 7, pp. 3715–3727, 2012.
[195] S. Kayashima et al., “Suction effusion fluid from skin and constituent analysis: New
candidate for interstitial fluid,” Am. J. Physiol. - Hear. Circ. Physiol., vol. 263, no. 5 32-5,
1992.
[196] M. Muller, A. Pena, and H. Derendorf, “MINIREVIEW-Issues in Pharmacokinetics and
Pharmacodynamics of Anti-Infective Agents: Distribution in Tissue,” … Agents …, vol. 48,
no. 5, pp. 1441–1453, 2004.
[197] P. Liu, M. Müller, and H. Derendorf, “Rational dosing of antibiotics: The use of plasma
92
concentrations versus tissue concentrations,” Int. J. Antimicrob. Agents, vol. 19, no. 4, pp.
285–290, 2002.
[198] T. K. L. Kiang, U. O. Häfeli, and M. H. H. Ensom, “A comprehensive review on the
pharmacokinetics of antibiotics in interstitial fluid spaces in humans: Implications on
dosing and clinical pharmacokinetic monitoring,” Clin. Pharmacokinet., vol. 53, no. 8, pp.
695–730, 2014.
[199] D. O. Scott, L. R. Sorenson, K. L. Steele, D. A. L. Puckett, and C. E. Lunte, “In Vivo
Microdialysis Sampling for Pharmacokinetic Investigations,” Pharmaceutical Research: An
Official Journal of the American Association of Pharmaceutical Scientists, vol. 8, no. 3. pp.
389–392, 1991.
[200] R. F. Donnelly, K. Mooney, E. Caffarel-Salvador, B. M. Torrisi, E. Eltayib, and J. C.
McElnay, “Microneedle-Mediated Minimally Invasive Patient Monitoring,” Ther. Drug
Monit., vol. 36, no. 1, p. 1, 2013.
[201] H. Yu, D. Li, Y. Ji, X. Zhang, and K. Xu, “An Interstitial Fluid Transdermal Extraction
Chip with Vacuum Generator and Volume Sensor for Continuous Glucose Monitoring
Overview of the Microfluidic Chip,” Key Eng. Mater., pp. 571–575, 2013.
[202] R. M. Taylor, P. R. Miller, P. Ebrahimi, R. Polsky, and J. T. Baca, “Minimally-invasive,
microneedle-array extraction of interstitial fluid for comprehensive biomedical applications:
transcriptomics, proteomics, metabolomics, exosome research, and biomarker
identification,” Lab. Anim., pp. 1–5, 2018.
[203] C. Kolluru, M. Williams, J. Chae, and M. R. Prausnitz, “Recruitment and Collection of
Dermal Interstitial Fluid Using a Microneedle Patch,” Adv. Healthc. Mater., vol. 8, no. 3,
2019.
[204] C. Kolluru et al., “Plasmonic Paper Microneedle Patch for On-Patch Detection of Molecules
in Dermal Interstitial Fluid,” ACS Sensors, vol. 4, no. 6, pp. 1569–1576, 2019.
[205] A. V. Romanyuk, V. N. Zvezdin, P. Samant, M. I. Grenader, M. Zemlyanova, and M. R.
Prausnitz, “Collection of analytes from microneedle patches,” Anal. Chem., vol. 86, no. 21,
pp. 10520–10523, 2014.
[206] E. Caffarel-Salvador et al., “Hydrogel-forming microneedle arrays allow detection of drugs
and glucose in vivo: Potential for use in diagnosis and therapeutic drug monitoring,” PLoS
One, vol. 10, no. 12, pp. 1–21, 2015.
[207] A. Mandal et al., “Cell and fluid sampling microneedle patches for monitoring skin-
resident immunity,” Sci. Transl. Med., vol. 10, no. 467, p. eaar2227, 2018.
[208] J. Chen, M. Wang, Y. Ye, Z. Yang, Z. Ruan, and N. Jin, “Fabrication of sponge-forming
microneedle patch for rapidly sampling interstitial fluid for analysis,” Biomed.
Microdevices, vol. 21, no. 3, pp. 1–10, 2019.
[209] E. M. Cahill et al., “Metallic microneedles with interconnected porosity: A scalable
platform for biosensing and drug delivery,” Acta Biomater., vol. 80, pp. 401–411, 2018.
[210] K. Takeuchi, N. Takama, B. Kim, K. Sharma, O. Paul, and P. Ruther, “Microfluidic chip to
interface porous microneedles for ISF collection,” Biomed. Microdevices, vol. 21, no. 1, 2019.
[211] S. A. Ranamukhaarachchi, C. Padeste, M. Dübner, U. O. Häfeli, B. Stoeber, and V. J.
Cadarso, “Integrated hollow microneedle-optofluidic biosensor for therapeutic drug
monitoring in sub-nanoliter volumes,” Sci. Rep., vol. 6, no. April, pp. 1–10, 2016.
[212] L. M. Strambini et al., “Self-powered microneedle-based biosensors for pain-free high-
accuracy measurement of glycaemia in interstitial fluid,” Biosens. Bioelectron., vol. 66, pp.
162–168, 2015.
[213] G. Lenk, S. Ullah, G. Stemme, O. Beck, and N. Roxhed, “Evaluation of a Volumetric Dried
93
Blood Spot Card Using a Gravimetric Method and a Bioanalytical Method with Capillary
Blood from 44 Volunteers,” Anal. Chem., vol. 91, no. 9, pp. 5558–5565, 2019.
[214] O. Beck, N. Kenan Modén, S. Seferaj, G. Lenk, and A. Helander, “Study of measurement of
the alcohol biomarker phosphatidylethanol (PEth) in dried blood spot (DBS) samples and
application of a volumetric DBS device,” Clin. Chim. Acta, vol. 479, no. January, pp. 38–42,
2018.
[215] K. Mishra, D. van den Ende, and F. Mugele, “Recent developments in optofluidic lens
technology,” Micromachines, vol. 7, no. 6, 2016.
[216] B. Koo and C.-J. Kim, “Evaluation of repeated electrowetting on three different
fluoropolymer top coatings,” J. Micromechanics Microengineering, vol. 23, no. 6, p. 67002,
2013.
[217] W. C. Nelson and C. C. J. Kim, “Droplet Actuation by ( EWOD ): A Review,” J. Adhes. Sci.
Technol., vol. 26, no. August 2012, pp. 1747–1771, 2012.
[218] H. Weingärtner, “Understanding ionic liquids at the molecular level: Facts, problems, and
controversies,” Angew. Chemie - Int. Ed., vol. 47, no. 4, pp. 654–670, 2008.
[219] K. R. J. Lovelock, A. Deyko, P. Licence, and R. G. Jones, “Vaporisation of an ionic liquid
near room temperature,” Phys. Chem. Chem. Phys., vol. 12, no. 31, pp. 8893–8901, 2010.
[220] O. Aschenbrenner, S. Supasitmongkol, M. Taylor, and P. Styring, “Measurement of vapour
pressures of ionic liquids and other low vapour pressure solvents,” Green Chem., vol. 11, no.
8, pp. 1217–1221, 2009.
[221] M. B. Shiflett and A. Yokozeki, “Solubilities and diffusivities of carbon dioxide in ionic
liquids: [bmim][PF6] and [bmim][BF4],” Ind. Eng. Chem. Res., vol. 44, no. 12, pp. 4453–
4464, 2005.
[222] S. Millefiorini, A. H. Tkaczyk, R. Sedev, J. Efthimiadis, and J. Ralston, “Electrowetting of
ionic liquids,” J. Am. Chem. Soc., vol. 128, no. 9, pp. 3098–3101, 2006.
[223] D. Chatterjee, B. Hetayothin, A. R. Wheeler, D. J. King, and R. L. Garrell, “Droplet-based
microfluidics with nonaqueous solvents and solutions.,” Lab Chip, vol. 6, no. 2, pp. 199–
206, 2006.
[224] H. Li, M. Paneru, R. Sedev, and J. Ralston, “Dynamic electrowetting and dewetting of ionic
liquids at a hydrophobic solid-liquid interface,” Langmuir, vol. 29, no. 8, pp. 2631–2639,
2013.
[225] J. Restolho, J. L. Mata, and B. Saramago, “Electrowetting of Ionic Liquids: Contact Angle
Saturation and Irreversibility,” J. Phys. Chem. C, vol. 113, no. 21, pp. 9321–9327, 2009.
[226] Y. S. Nanayakkara et al., “A fundamental study on electrowetting by traditional and
multifunctional ionic liquids: Possible use in electrowetting on dielectric-based microfluidic
applications,” Anal. Chem., vol. 80, no. 20, pp. 7690–7698, 2008.
[227] X. Hu et al., “Electrically Switchable Capillarity of Ionic Liquids,” J. Adhes. Sci. Technol.,
vol. 26, no. 12–17, pp. 2069–2078, 2012.
[228] S. Zhang et al., “Enhanced and reversible contact angle modulation of ionic liquids in oil
and under AC electric field,” ChemPhysChem, vol. 11, no. 11, pp. 2327–2331, 2010.
[229] P. A. L. Wijethunga, Y. S. Nanayakkara, P. Kunchala, D. W. Armstrong, and H. Moon,
“On-chip drop-to-drop liquid microextraction coupled with real-time concentration
monitoring technique,” Anal. Chem., vol. 83, no. 5, pp. 1658–1664, 2011.
[230] A. Eftekhari, Ionic Liquid Devices. Knovel, 2018.
[231] A. Al Amin, A. Jagtiani, A. Vasudev, J. Hu, and J. Zhe, “Soft microgripping using ionic
liquids for high temperature and vacuum applications,” J. Micromechanics
94
Microengineering, vol. 21, no. 12, p. 125025, 2011.
[232] B. Raj, M. Dhindsa, N. R. Smith, R. Laughlin, and J. Heikenfeld, “Ion and liquid
dependent dielectric failure in electrowetting systems,” Langmuir, vol. 25, no. 20, pp.
12387–12392, 2009.
[233] E. Bakker and Y. Qin, “Electrochemical Sensors,” Anal. Electochemistry, pp. 171–202, 2002.
[234] J. Chou, “Electrochemical Sensors,” in Hazardous Gas Monitors, McGraw-Hill & SciTech
Publishing, 1999, pp. 27–35.
[235] F. Opekar and K. Štulík, “Electrochemical sensors with solid polymer electrolytes,” Anal.
Chim. Acta, vol. 385, no. 1–3, pp. 151–162, 1999.
[236] F. Ribet, L. De Pietro, N. Roxhed, and G. Stemme, “IONIC LIQUID MICRODROPLET
MANIPULATION BY ELECTROWETTING-ON-DIELECTRIC FOR ON / OFF
DIFFUSION CONTROL,” IEEE 31st Int. Conf. Micro Electro Mech. Syst., 2018.
[237] Y. S. Nanayakkara, S. Perera, S. Bindiganavale, E. Wanigasekara, H. Moon, and D. W.
Armstrong, “The effect of AC frequency on the electrowetting behavior of ionic liquids,”
Anal. Chem., vol. 82, no. 8, pp. 3146–3154, 2010.
[238] C. Dong et al., “On the droplet velocity and electrode lifetime of digital microfluidics:
voltage actuation techniques and comparison,” Microfluid. Nanofluidics, vol. 18, no. 4, pp.
673–683, 2015.
[239] F. Mugele and J.-C. Baret, “Electrowetting: from basics to applications,” J. Phys. Condens.
Matter, vol. 17, no. 28, pp. R705–R774, 2005.
[240] C. Moya, J. Palomar, M. Gonzalez-Miquel, J. Bedia, and F. Rodriguez, “Diffusion
Coefficients of CO2 in Ionic Liquids Estimated by Gravimetry,” Ind. Eng. Chem. Res., vol.
53, no. 35, pp. 13782–13789, 2014.
[241] S. H. Barghi, M. Adibi, and D. Rashtchian, “An experimental study on permeability,
diffusivity, and selectivity of CO2 and CH4 through [bmim][PF6] ionic liquid supported on
an alumina membrane: Investigation of temperature fluctuations effects,” J. Memb. Sci.,
vol. 362, no. 1–2, pp. 346–352, 2010.
[242] K. Saihara, Y. Yoshimura, S. Ohta, and A. Shimizu, “Properties of Water Confined in Ionic
Liquids,” Sci. Rep., vol. 5, no. 1, p. 10619, 2015.
[243] H. Kim, K. Muller, O. Shardt, S. Afkhami, and H. A. Stone, “Solutal Marangoni flows of
miscible liquids drive transport without surface contamination,” Nat. Phys., no. July, pp.
1–7, 2017.
[244] L. Li, C. Liu, H. Ren, and Q.-H. Wang, “Adaptive liquid iris based on electrowetting.,” Opt.
Lett., vol. 38, no. 13, pp. 2336–8, 2013.
[245] J. Wan, B. Chen, T. Zhang, S. Zhang, and Y. Duan, “Optical switch with a micro liquid
droplet,” in SPIE, 2018, no. 10827.
[246] L. Li, R. Yuan, L. Luo, and Q. Wang, “Optofluidic Zoom System Using Liquid Optical Path
Switchers,” IEEE PHOTONICS Technol. Lett., vol. 30, no. 10, pp. 883–886, 2018.
[247] L. Eldada, “Optical communication components,” Rev. Sci. Instrum., vol. 75, no. 3, pp. 575–
593, 2004.
[248] R. R. A. Syms, H. Zou, J. Stagg, and D. F. Moore, “MEMS variable optical attenuator with
a compound latch,” Microelectron. Eng., vol. 73–74, pp. 423–428, 2004.
[249] R. R. A. Syms, H. Zou, J. Stagg, and H. Veladi, “Sliding-blade MEMS iris and variable
optical attenuator,” J. Micromechanics Microengineering, vol. 14, no. 12, pp. 1700–1710,
2004.
95
[250] C. Marxer, P. Griss, and N. F. De Rooij, “A Variable Optical Attenuator Based on Silicon
Micromechanics,” IEEE Photonics Technol. Lett., vol. 11, no. 2, pp. 233–235, 1999.
[251] C. U. Murade, J. M. Oh, D. Van Den Ende, and F. Mugele, “Electrowetting driven optical
switch and tunable aperture,” Opt. Express, vol. 19, no. 16, pp. 15525–15531, 2011.
[252] C. Liu, L. Li, and Q. H. Wang, “Bidirectional optical switch based on electrowetting,” J.
Appl. Phys., vol. 113, no. 19, pp. 193106–1, 2013.
[253] L. Li, C. Liu, and Q.-H. Wang, “Optical switch based on tunable aperture,” Opt. Lett., vol.
37, no. 16, p. 3306, 2012.
[254] J. L. Jackel, S. Hackwood, and G. Beni, “Electrowetting optical switch,” Appl. Phys. Lett.,
vol. 40, no. 4, pp. 3–6, 1982.
[255] Y. Li, R.-Y. Yuan, L. Li, S. Rong, and Q. H. Wang, “Electrowetting optical switch with
large aperture tuning range,” J. SID, vol. 25, no. 12, pp. 725–730, 2017.
[256] C. Liu, D. Wang, L. Yao, L. Li, and Q. Wang, “Electrowetting-actuated optical switch based
on total internal reflection,” Appl. Opt., vol. 54, no. 10, pp. 2672–2676, 2015.
[257] R. D. Montoya, K. Underwood, S. Terrab, A. M. Watson, V. M. Bright, and J. T. Gopinath,
“Large extinction ratio optical electrowetting shutter,” Opt. Express, vol. 24, no. 9, p. 9660,
2016.
[258] L. Li, C. Liu, H. R. Peng, and Q. H. Wang, “Optical switch based on electrowetting liquid
lens,” J. Appl. Phys., vol. 111, no. 10, pp. 1–5, 2012.
[259] “UN development programme,” 2019. [Online]. Available:
https://www.undp.org/content/undp/en/home/sustainable-development-goals.html.
96
Paper Reprints