Document [original]

Integrated microwave biosensors on SiGe BiCMOS technology: A

“More Than Moore” approach

Vorgelegt von

Master of Science

Subhajit Guha

geb. in Kolkata, India

Von der Fakultät IV

Elektrotechnik und Informatik

der Technischen Universität Berlin

zur Erlangung des akademischen Grades

Doktor der Ingenieurwissenschaften

(Dr.-Ing)

genehmigte Dissertation

Promotionsausschuss

Vorsitzender: Prof. Dr. Klaus Petermann

Berichter: Prof. Dr. Roland Thewes

Berichter: Dr. rer nat. habil Christian Wenger

Berichter: Prof. Dr. Hermann Schumacher

Tag der wissenschaftlichen Aussprache: 18.01.2017

Berlin 2017

iii

I lovingly dedicate this thesis to my parents and my sister, who have always been a

great support through-out the journey

Abstract

There has been an ever increasing demand for the establishment of Point-of-Care testing systems

for rapid detection and diagnosis of diseases and as well for monitoring vital health parameters.

The development in the area of microfluidic systems as well as microsystem technology as a whole

gave birth to new avenue of research called the Lab-on-a-chip devices, which are an essential part

of point-of-care testing systems. Such devices are expected to perform biochemical analysis with

sensitivity and accuracy of the order of the state of the art bioanalytical laboratories and at the

same time use extremely small volume of the samples. This led to the development of biosensors

with extremely high sensitivity and accuracy especially based on optical techniques. However,

optical techniques although produced extremely sensitive sensor systems, suffered from serious

drawbacks like requirement of labeling compounds, bulky test-benches for measurement and many

more. Therefore, the real goal of establishing miniaturized point-of-care diagnostic system for

rapid measurements was very difficult to achieve.

The obvious choice as an alternative to optical platform was to establish electrical sensing schemes

to avoid the requirement of using labeling compounds and markers. The electrical approaches

explored at the initial phase although circumvented the problems of optical techniques, had other

issues which still continued a bulky overall measurement setup (for e.g. use of reference

electrodes).

In this thesis, “all-electrical” sensor systems operating at the GHz frequency range of the

electromagnetic spectrum, and fabricated on standard CMOS/BiCMOS process have been

explored and demonstrated. The focus of the thesis is to demonstrate the capability of integrating

biological sensing on the standard CMOS/BiCMOS process. This approach takes a step ahead of

the established electrical biosensors with a hybrid approach where the front end electronics for

data acquisition and processing is integrated in a hybrid fashion with the biosensor system. The

approach explored in this thesis has the biosensor on the same technology platform where the front

end electronics for read out, data acquisition and processing are fabricated. This kind of an

approach stems out from the “More than Moore” technique of semiconductor technology roadmap

and offers extremely high sensitivity due to close proximity of the sensor to the front end

electronics. The high-frequency approach on the other hand offers other advantages like nullifying

low-frequency dispersion mechanisms and evading unwanted electrochemical effects at the sensor

and electrolyte interface (avoiding the use of reference electrodes). Sensors operating at high

frequency have dimensions of the order of the biomaterials (cells) that are probed. Therefore, the

high-frequency CMOS compatible sensor approach explored in this thesis takes a step forward

towards establishing simple, low cost, miniaturized point-of-care systems.

Several relevant sensor applications are explored in this thesis in order to demonstrate the

feasibility of the established approach. An immunosensor operating at 6 GHz has been established

and the functionality has been demonstrated with the detection of concentration of creatinine

molecules. The sensor system demonstrates the capability of detection of the concentration of

creatinine molecules in the clinically relevant range and with the sensitivity equivalent to the

established optical techniques. Applications like sensing of glucose concentration in a suspension

and cytometric applications like detection of concentration of particles in a solution has been

shown. Finally, a novel approach to make the overall sensor system flexible and with extremely

rapid measurement capability has been demonstrated. In such a system, the sensor output is a DC

signal, therefore, making the sensor system function with DC inputs and DC outputs, thus setting

the platform for ideal miniaturized point-of-care diagnostic systems.

vii

Zusammenfassung

Die Nachfrage nach „Point-of-Care-Testing“ Systemen zur schnellen Erkennung und Diagnose

von Krankheiten und auch für die Überwachung von lebenswichtigen Gesundheitsparametern

steigt an. Neue Entwicklungen im Bereich der Mikrofluidik-Systeme sowie der

Mikrosystemtechnik als Ganzes ermöglichte einen neuen Forschungszweig für sogenannten Lab-

on-a-Chip Bauelemente, die ein wesentlicher Bestandteil der „Point-of-Care-Testing“ Systeme

sind. Von diesen Bauelementen wird erwartet, dass sie die biochemische Analyse mit der

Empfindlichkeit und Genauigkeit von modernen bioanalytischen Laboren mittels der Verwendung

von extrem kleinen Probenvolumen durchführen können. Das führte, basierend auf optischen

Technologien, zu der Entwicklung von Biosensoren mit extrem hoher Empfindlichkeit und

Genauigkeit. Obwohl durch die Nutzung von optischen Techniken extrem empfindliche Sensoren

hergestellt werden können, existieren schwerwiegenden Nachteilen wie das Markieren von

Molekülen und die Größe der Messaufbauten. Damit war das Ziel der Miniaturisierung von Point-

of-Care-Diagnosesystemen für schnelle Messungen schwer zu erreichen.

Eine Alternative zu optischen Testsystemen war es, elektrische Sensorsysteme zu etablieren, um

das Markieren von Molekülen zu vermeiden. Die elektrischen Ansätze, die in der Anfangsphase

untersucht wurden, umgingen die Probleme der optischen Techniken, besaßen immer noch einen

ziemlich sperrigen Gesamtmessaufbau (z.B. durch die Verwendung von Referenzelektroden).

Daher ist das Ziel, hochempfindliche Sensorsysteme für miniaturisierte Point-of-Care

Diagnostiksysteme herzustellen, bei weitem noch nicht erreicht. In diesem Zusammenhang ist ein

neuer Ansatz erforderlich, der hochempfindliche Biosensoren sowie die Miniaturisierung des

gesamten Sensorsystems ermöglicht.

In dieser Arbeit werden "all-electrical" Sensorsysteme, die im GHz-Frequenzbereich des

elektromagnetischen Spektrums arbeiten und mittels eines Standard-CMOS / BiCMOS-Prozesses

hergestellt wurden, entwickelt und untersucht. Der Schwerpunkt dieser Arbeit ist, die Integration

von biologischen Sensoren mittels eines Standard-CMOS / BiCMOS-Prozesses zu demonstrieren.

Dieser Ansatz geht einen Schritt weiter als in bereits etablierte elektrischen Biosensoren, denn

mittels des hybriden Ansatzes wird die Front-End-Elektronik für die Datenerfassung und -

verarbeitung in einem Hybrid-Ansatz in das Biosensor-System integriert. Der Ansatz in dieser

Arbeit ist, den Biosensor in die gleiche Technologie-Plattform zu integrieren, in der auch die

Front-End-Elektronik zum Auslesen, Datenerfassung und Verarbeitung hergestellt wird. Diese Art

des Ansatzes stammt aus der "More than Moore" Philosophie der Halbleiter-Technologie und

bietet eine, aufgrund der Nähe des Sensors zur „Frontend“ Elektronik, extrem hohe

Empfindlichkeit. Der Hochfrequenz-Ansatz auf der anderen Seite bietet weitere Vorteile das

Ausblenden der niederfrequenten Dispersionsmechanismen und das Verhindern von

unerwünschten elektrochemischen Effekten an der Sensor und Elektrolyt-Grenzfläche, denn es

sind keine Referenzelektroden erforderlich. Sensoren, die bei diesen hohen Frequenzen arbeiten,

haben Abmessungen in der Größenordnung der zu untersuchenden Biomaterialien (Zellen). Daher

viii

liefert der in dieser Arbeit untersuchte Hochfrequenz CMOS-kompatible Sensoransatz einen

wichtigen Schritt nach vorne, auf dem Weg zu einfachen, kostengünstigen, miniaturisierten Point-

of-Care-Systemen.

In dieser Arbeit wurden mehrere relevante Sensoranwendungen untersucht, um die Fähigkeit

dieses Ansatzes zu demonstrieren. Ein Immuno-Sensor, der bei 6 GHz arbeitet, wurde entwickelt

und dessen Funktionalität wurde durch die Detektion der Konzentration von Kreatinin-Molekülen

nachgewiesen. Das Sensorsystem demonstriert die Fähigkeit der Konzentrationsmessung von

Kreatinin Molekülen im klinisch relevanten Bereich mit einer Empfindlichkeit der entsprechend

etablierten optischen Techniken. Weitere Anwendungen wie die Bestimmung der

Glukosekonzentration in einer Suspension und zytometrische Anwendungen wie die

Konzentrationsbestimmung von gelösten Partikeln, wurden gezeigt. Ein neuartiger Ansatz, der das

gesamte Sensorsystem flexibler und extrem schnellen Messzyklen ermöglicht, wurde

nachgewiesen. In einem solchen System besteht das Sensorausgangssignal aus einem DC-Signal

und ermöglicht deshalb, das Sensorsystem mit DC-Eingänge und DC-Ausgänge zu nutzen und

bietet somit die ideale Plattform für miniaturisierte „Point-of-care“-Diagnosesysteme.

Acknowledgement

This thesis would not have been possible without the help and support from a lot of people in my

professional and personal life. A PhD thesis is one of the very few cherished documents in one’s

life. I would like to take this opportunity to thank all the people who have supported me in this

segment of my life while pursuing my doctoral studies in IHP Microelectronics.

First and foremost I would like to thank my parents Mrs. Sankari Guha and Mr. Abhijit Guha and

my sister Ms. Rumpa Guha who have been the pillars of support in my life. I deeply acknowledge

their belief in me and the blessings and wishes that they have bestowed on me. The support they

have extended to me to come to Germany and pursue my career aspirations and continue my

doctoral studies is immense. I thank them from the bottom of my heart for all that they have done

for me.

Dr. rer nat. habil. Christian Wenger, the supervisor of my doctoral studies in IHP, has been in a

true sense an excellent support, a wonderful guide and a very good friend during my stay in IHP.

As a supervisor, he has given me complete freedom to carry out my research activities and

supported me in all my research ventures. As a guide he has given me valuable suggestions to

enhance my research standards and establish my research work in an international platform. As a

friend he has been very supportive in my ups and downs during my stay at IHP and has given me

valuable advices in taking very important steps both in my personal and my professional life. I

deeply and sincerely acknowledge the presence of him during my doctoral studies. I would also

like to acknowledge Prof. Dr. Roland Thewes, chair of Sensors and Actuators department, TU

Berlin. The technical suggestions and advices given by him have increased the technical

significance of the thesis by many folds. I would also like to thank him for taking his time out from

his very busy schedule and agreeing to be the supervising professor of my PhD work. I would also

like to extend my acknowledgement to Prof. Dr. Hermann Schumacher, University of Ulm, for

agreeing to co-supervise the PhD thesis. The valuable suggestions given by him aided in improving

the quality of the thesis.

There have been many people in IHP who have been a valuable help from the technical aspect of

the thesis. I would like to thank all of them for their support. Firstly, I would like to thank Dr.

Klaus Schmalz of the dept. of circuit design, IHP. He has been a great support to me while I was

designing the high-frequency sensors. I had some valuable conversations and suggestions from

him. Dr. Chafik Meliani, guided the work with respect to plaque characterization and imaging. I

would like to acknowledge his guidance in the technical aspects. Dr. Frank Herzel, dept. circuit

design, IHP, was involved in designing the sensor system, which is used for single particle sensing.

I would like to thank him for all the valuable technical insights he provided during the design of

the sensor system. Dr. Axel Warsinke, dept. of biotechnology, University of Potsdam, was

involved in the immobilization of protein molecules for the establishment of immunosensors. My

heartfelt thanks to him for the support. Mr. Alexander Wolf, dept. of material research, IHP was a

hiwi student and supported me while I was designing the microfluidic system for integration with

the sensors. Mr. Farabi Ibne Jamal, also working in the same area of high frequency biosensors,

dept. circuit design, IHP, has been a constant support and a wonderful colleague. We have designed

a lot of circuits together and he was always my support during the critical biological measurements.

I would like to thank him for all the work that we have done together.

My acknowledgement would be incomplete if I don’t mention the name of Prof. Dr. Thomas

Schröder and Prof. Dr. Giovanni Cappellini. Dr. Schröder, head of the department of material

research, IHP, has indeed made my stay here quite wonderful. Although he was not a part of the

PhD work, he has given me innumerable suggestions about paper writing skills, selecting

conferences, writing future proposals and many more. I have thoroughly enjoyed my discussions

with him and acknowledge him for all the valuable suggestions. With Dr. Capellini, I have worked

on designing of Ge LASERs on CMOS technology which was not a part of my PhD work. The

working spirit that I have learnt from him is noteworthy. I deeply acknowledge him for that.

Ms. Ankita Arora, although she was not actively involved during my doctoral thesis, she made a

special place for herself while I was finishing the writing of the dissertation. I would like to express

my heartfelt gratitude to Ankita for being extremely supportive and above all for being an

extremely special person making my life ever so beautiful.

During this thesis work there were a lot of friends who were a personal support. Firstly, Karthik,

my closest friend for a long time now, has been a wonderful support all through-out. He has seen

my ups and downs very closely and has always stood by me. I cannot thank him in words for all

that he has done for me. Payel, a very close friend was a strong support for me always. The long

and deep conversations with her kept me going at all stages of my research life, starting way back

in my IISC days in India. Mudita who is my sister and a dear friend, made my staying in Germany

so easy. Her contribution towards my getting settled in Germany and leading an easy life is

immense. I heartily acknowledge her for that. Iria and Pedro, two of my very good friends in IHP,

indeed made my living in Frankfurt (Oder) very interesting and lively. Living in the same city and

working in the same organization, they really knew the professional and the personal aspects of

me and have been very supportive all through-out. I sincerely thank them for that. I would like to

thank all my IHP colleagues and friends whose names I could not mention here for making my

stay in IHP beautiful. Friends in Germany, like Kiran, Aishcharya, Pragya, were a great company

right from the time I started my master studies in Hamburg and continued during my doctoral

studies in IHP as well

Above all, I would thank almighty God for giving me the strength to pursue my career aspirations.

xii

xiii

Table of Contents

1. Introduction

1.1 Motivation………………………………………………………………………. 01

1.2 Optical biosensors…………………………………………………………… 03

1.3 Electrochemical biosensors………………………………………………05

1.4 CMOS/BiCMOS microwave biosensor platform………………. 08

1.5 Need for microfluidic integration……………………………………. 13

2. Design and Integration

2.1 Introduction……………………………………………………………………..16

2.2 Technology and integration………………………………………………17

2.2.1 Technology……………………………………..……………………… 17

2.2.2 Influence of biological material and integration……….19

2.3 Design of Planar microwave sensors…………………………………24

2.3.1 Coplanar transmission lines…………………………..………24

2.3.2 Interdigitated Capacitor…………………………………….....26

2.4 Design of sensor circuit…………………………………………………. …31

3. Dielectric Immunosensor for Creatinine

3.1 Introduction………………………………………………………………………35

3.2 Proposed CMOS compatible immunosensor

approach…………………………………………………………………………..37

3.2.1 Fabrication and operation of the sensor………..…………38

3.2.2 Immobilization of creatinine……………………………..……..42

3.3 Results and discussion………………………………………………………44

3.3.1 Optical measurement of creatinine concentration…..46

3.3.2 Dielectric measurement of creatinine concentration.48

3.4 Conclusion………………………………………………………………………..53

xiv

4. Detection of Analyte Concentration in Suspensions

4.1 Introduction……………………………………………………………………..54

4.2 Sensor parameter for microfluidic integration….………………56

4.2.1 Planarization of silicon chip………………………………………57

4.2.2 Microfluidic integration to silicon chip…………………..…59

4.3 Results and discussion……………………………………………………….60

4.3.1 Calibration of sensors……………………………………………….60

4.3.2 Particle concentration measurement……………..………..66

4.3.3 Fat and calcium characterization in blood……………..…69

4.3.4 Dielectric imaging of biomaterials………………………..….72

4.4 Conclusion………………………………………………………………………..75

5. Towards Particle Counting

5.1 Introduction…………………………………………………………………………77

5.2 System dynamic analysis……………………………………………………..78

5.2.1 Modeling of dynamic capacitance sensor…………………79

5.2.2 Design of sensor circuit…………………………………………….81

5.2.3 Frequency demodulator architecture……………………….83

5.3 Results and discussion…………………………..…………………………….90

5.4 Dual demodulator architecture…………………………………..……….95

5.4.1 Elimination of noise by time-averaging…………………………..…95

5.4.2 Particle concentration and flow-rate…………………………………99

5.5 Conclusion………………………………………………………………………………….100

6. Conclusion and Outlook…………………………………………………………..101

Design and integration…………………………………………………………….102

Dielectric immunosensor ………………………………………………………..103

Detection of analyte concentration…………………………………………104

Towards particle counting……………………………………………………….105

Outlook……………………………………………………………………………………106

Bibliography……………………………………………………………………………………..108

List of publications……………………………………………………………………………119

List of figures……………………………………………………………………………………121

List of tables……………………………………………………………………………………..129

List of Abbreviations…………………………………………………………………………130

Chapter 1 Introduction

| 1

INTRODUCTION

1.1 Motivation

There is an ever-increasing demand for establishment of point-of-care (POC) testing approaches

in the field of medical applications and health care. The advancement in the POC technology

ensures a positive impact on health, wellness and quality-of-life in both developed and

developing world [1]. One of the primary aims of establishing POC devices is to bring down the

overall time required to produce the diagnostic test results [2, 3, 4] along with easy handling of

medical test devices. Developing sensitive and fast biomedical analysis systems is the foundation

for such POC systems. At the same time, keeping the cost of POC systems low is also of prime

significance [3] in order to address a mass-market. The design of cheap POC systems with

extremely sensitive bio-analysis unit calls for the convergence of various research domains

ranging from life science, chemistry, sensor design, circuit design, microfabrication, system

design and more.

Presently specialized personnel in laboratories, utilizing off-the-shelf components and

instrumentations, carry out most clinical analyses, assuring extreme precision and accuracy of

the obtained results. Typical steps for present day medical diagnostics and clinical analysis is

shown in Fig. 1.1

The tests conducted in clinical laboratories require high incubation and analysis time. Therefore,

although extremely precise, the present clinical diagnostic approaches require a substantial

amount of time to produce the results.

________________________________________________________________________________________

Figure 1.1 Steps for typical clinical diagnostics.

Chapter 1 Introduction

| 2

POC systems aim to reduce this time required for the overall clinical diagnostic, for fast

detection of disease and easy treatment of them. Fig. 1.2 shows the steps for a typical POC

testing. Due to limited number of steps involved in POC systems, the total time between the

decision of clinical diagnosis and reporting of final clinical analysis is considerably small.

Alongside fast analysis time scale, in order to be an effective system comparable to the clinical

laboratories, extremely sensitive sensors (biosensors) are to be used for the clinical analysis in

POC systems.

This multi-disciplinary design approach of the POC systems as mentioned above has given rise

to lab-on-chip (LOC) devices, which are the fundamental bio-analysis blocks of POC systems.

LOC devices aim at developing miniaturized sensor platforms, which integrate several laboratory

functions on a single chip or a single system. Considerable amount of research work has been

dedicated to establish such high sensitive sensors also called biosensors or chem-bio sensors. A

biosensor is a device that is used to detect or/and quantify biomolecules based on a biochemical

reaction [5]. The biomolecule can be a specific protein or DNA, biomarkers, pathogenic

organisms, hormones or other medically relevant analytes [6]. To comply with the standard used

in the clinical laboratories, development of innovative analytical devices with enhanced

sensitivity, specificity, precision, speed, usability and miniaturization is needed. State-of-the-art

________________________________________________________________________________________

Figure 1.2 Steps for point of care diagnostics. The time constraints are considerably reduced when

compared to the standard clinical diagnostic approaches.

Chapter 1 Introduction

| 3

biosensor platforms are mainly based on optical detection schemes. The optical sensors are taken

as gold-standard in the field of biomedical diagnostics and they have made their way to

commercially available non-invasive POC diagnostic devices. One such example is the pulse

oximeter based on optical absorption principles [7] measuring proportion of oxygenated

hemoglobin in blood. Commercially available pulse oximeters are hand held devices showing a

classical example of POC diagnostic system. The success of such clinical diagnostic devices led

to the research of optical biosensors in the area of biomarker detection like immunosensors [8],

cytometric applications [6], proteomic analysis [9], infectious disease diagnostics [10], etc. On

the other hand, electrochemical sensors have also been another cornerstone technique for POC

diagnostic systems. The evolution of the electrochemical glucose sensor into a viable cheap

commercial product ever since its inception in 1960’s is an example of the success of

electrochemical techniques. Therefore, electrochemical techniques for biosensors applied now

to biomolecule detection or cytometric applications are also being researched for application in

POC systems [11,12,13]. Hence, it is worthwhile to review some of these optical and

electrochemical techniques and compare them with the microwave (high-frequency) sensing

technique for biological applications, like immunosensors, cytometry, etc., that has been

explored in this thesis.

1.2 Optical biosensors

As mentioned above, the success of commercially available optical POC diagnostic systems has

prompted the research of optical biosensors for disease detection, viral detection and more.

Optical biosensors are based on detection schemes like fluorescence based, [14-28],

chemiluminescence [29-40], surface plasmon resonance schemes, [41-47], absorbance based [48,

49] for DNA detection, protein analysis, immunosensing applications etc.

Figure 1.3 Schematic of optical immunosensors based on fluorescence detection.

________________________________________________________________________________________

Chapter 1 Introduction

| 4

For example, optical immunosensors have already become the gold-standard approach for

clinical diagnostics for determining various biomarkers. Immunosensors provide highly

repeatable measurements for a wide range of biomarkers relating the immunoassay event into an

optical signal. Fig. 1.3 shows the schematic of a typical optical immunosensor based on

fluorescence detection. This is the most common method of detection using an immunoassay.

Antibodies are immobilized on an electrode surface. Specific antigens bind to the immobilized

antibodies forming an antigen-antibody pair. A second antibody is then bound to the antigen in a

sandwich configuration, where the antigen is between two antibodies [50, 51]. An additional

antibody with a fluorophore label is bound to the top antibody. Optical signal emitted from a

LASER source impinges upon the antigen-antibody pair; fluorescence emission from the

fluorophore labels or markers is detected by a photo-detector, for example a photo-multiplier or a

CCD camera [7]. The fluorescence gives a direct understanding of the concentration of captured

biomarkers in the sandwich approach and therefore, allows for quantitative determination of

captured biomarker concentration. A further extension with optical fiber cables is needed to

bring the optical signal to the front-end electronics. The fluorescence marker based sensing

scheme shows excellent sensitivity and high specificity; such a technique is also suitable for a

large variety of biomolecules. The sandwich scheme of antigen-antibody binding and using

specific markers for an antigen-antibody couple increases the specificity of the system.

Labeling a biomolecule can often lead to change of the properties of the biomolecule which in

turn can lead to falsified output from the sensor. The above mentioned sandwich approach

overcomes the issue by not directly labeling the target molecule but using an intermediate step.

The established ELISA (Enzyme linked immunosorbent assay) is a classic example of this

technique. However, the development cost is increased due to the sandwich approach. Also the

use of labeling compounds makes the overall cost of the sensor high. Also additional antibody

molecules are needed for the sandwich approach and further for the binding of the fluorophore

markers [5]. The translation of the optical sensor signal to electrical signal can lead to signal

degradation and also degradation of signal to noise ratio. Although, commercial products like the

pulse oximeters mentioned above have achieved excellent integration scheme and translation of

optical signal to electrical signal, for biomarker detection and single particle analysis this can be

a limiting factor, because of low concentration of test molecules.

Label free optical techniques are developed using approaches like surface plasmon resonance

(SPR). This utilizes the coupling of the optical signal to a thin metallic surface as shown in Fig.

1.4. The antigen-antibody pair is immobilized on the surface of the thin metal surface. The

optical method in this technique detects the localized refractive index variation around the

vicinity of the metal structures. Binding of specific antigen to the antibody changes the refractive

index around the vicinity of the metal structure, resulting in the shift of the resonance peak of the

reflected light. Being label free, this approach reduces the complexity and cost involved in

fluorophore marker based sensors.

Chapter 1 Introduction

| 5

This technique is limited in the detection of smaller concentration of biomolecules [44], because

of localized refractive index change. The amount of target molecules required to create a

detectable change is high. In other cases, a complex circuit is needed to amplify or detect a

minute change of refractive index, which has to additionally overcome the issues of signal

degradation while translating optical signal into electrical signal. Other methods of increasing the

detection limit of SPR is using nanoparticles or nanostructures which also require stringent

fabrication technique. Also in SPR based biosensors, false signal is a problem in complex

solution like blood or urine [43]. Hand-held POC diagnostic devices have been demonstrated

[45], which could overcome the shortcomings of limited detection or selectivity using the

techniques mentioned above.

1.3 Electrochemical biosensors

Electrochemical sensors are based on interaction of chemical species with electrodes resulting in

electrical signals primarily current (amperometric sensors) [52-60], potential (potentiometric

sensors) [61-65] or impedance (impedimetric sensors) [66-70]. Integration of electrochemical

sensors with the front-end electronic circuits for signal read-out is considerably easier when

compared to optical sensors, due to inherent electrical signal output from the sensor. Therefore,

Figure 1.4 Schematic of optical immunosensors based on surface plasmon resonance.

________________________________________________________________________________________

Chapter 1 Introduction

| 6

at the integration level, which involves packaging and assembling, electrochemical sensors are

less complex when compared to their optical counterparts. This has led to inexpensive, easy to

use POC diagnostic systems, like the commercially available blood-glucose monitoring devices.

On the same lines, there has been considerable effort to extend the research of electrochemical

sensors towards detection of concentration of biomarkers, viral detection and more.

Electrochemical sensors are known for being label free, thus nullifying the issues related to

labeling compounds. However, for applications like immunosensor, labeling is still required.

Fig. 1.5 shows a typical electrochemical (amperometric) immunosensor approach using labeling

technique. Amperometric techniques use a three electrode measurement setup, with a working

electrode, counter electrode and a reference electrode. The immunosensors operate by the

detection of current on the working electrode. The current is generated due to the redox reaction

at the surface of the electrode. It should be noted the same sandwich approach for antibody-

antigen pairing as was seen in optical sensor is used in most of the electrochemical

immunosensors as well. In case of electrochemical sensors, redox markers are used instead of

fluorophore markers in optical sensors. The redox markers initiate a cyclic oxidation-reduction

process at the working electrode and produce a current due to the exchange of electrons [59].

Labeled electrochemical amperometric techniques also offer high specificity due to labeling

technique with fairly less complex front-end integration when compared to optical

immunosensors. However, the limitations caused by labeling compounds still persist in this kind

of electrochemical immunosensors. Therefore, the cost of the sensor system is still considerably

higher due to the use of labeling compounds. In single molecule detection, label free

electrochemical sensors are used [71]. Such sensors no longer suffer from the problems of

labeling compounds.

Figure 1.5 Schematic of optical immunosensors based on amperometric detection technique.

________________________________________________________________________________________

Chapter 1 Introduction

| 7

Electrochemical sensors based on impedance measurements and have been demonstrated by

various research groups like, Goh and Ram [72], Krommenhoek et al. [73], Faenza et al. [74].

Commercial products based on electrochemical impedance spectroscopy have been demonstrated

by Micronit microtechnologies [75], Gamry Instruments [76] and more. Usually impedance

measurements are performed at frequency range of 100 KHz to a few MHz as described by

Krommenhoek et al. [73]. In this frequency range (“low-frequency”), biological suspensions

especially of suspended cells show dielectric dispersions based on their properties, for example,

potential across the cell membrane and cell walls [77]. Therefore, designing the sensors in this

frequency range is competent for detection cellular properties governed by low-frequency

dispersion mechanisms (for e.g., in case of cells, membrane capacitance). Electrochemical

sensors based on impedance measurements also require reference electrode for precise

measurements.

It is seen that an important component of any form of electrochemical sensor is the reference

electrode. The use of reference electrode is needed to keep the sample solution at a

thermodynamic equilibrium. The additional reference electrode can often make the overall

integration scheme cumbersome [59]. However, considerable effort has been put into

miniaturization of reference electrodes, but it is still an irreplaceable component of

electrochemical sensor, therefore, increasing the overall development cost and design complexity

of the sensor. Commercial products like i-STAT (Abbott Point of Care, U.S.A) [70] for blood

analysis, have overcome the miniaturization issue with respect to reference electrodes with

additional effort in the overall design. However, the penetration of electrochemical based

immunosensors, pathogen concentration detector products is still very slow even after 10 years

of research activities. The biggest challenge for electrochemical sensor is the establishment of

automated array platform with fast electrical response and signal processing.

All in all, it is seen the technique of sensing applied for biological purposes is application

specific and is therefore, difficult to establish a universal biosensing technique for POC

diagnostic device. While optical sensing technique is feasible for a vast number of applications

(pulse oximeter, ELISA technique as examples), electrochemical sensing is suited for other

applications like blood glucose monitoring systems. Low-frequency impedance sensors are

suited for determining cellular parameters like cell membrane properties. However, in detection

of concentration of biomarkers, pathogens, high-frequency (GHz range) dielectric sensing is

becoming an extremely attractive option. Concentration of molecules in a medium influences the

permittivity of the medium which can be detected with a capacitive sensor. The use of high

frequency aids in miniaturization of the overall system. Compared to labeled optical sensors used

for the same applications, the high-frequency dielectric sensors require no labeling compounds.

It is an all-electrical sensing technique. Sensitivity of the order of single particle detection can be

demonstrated by the established high-frequency capacitive sensors. Therefore, extremely low

concentrations of analyte can be detected with better signal to noise ratio as compared to label

free optical sensors. When compared to electrochemical techniques high-frequency capacitive

Chapter 1 Introduction

| 8

sensors for dielectric detection require no additional reference electrodes for the system.

Therefore, the complexity with regard to the overall development of the sensor system is

reduced. In detection of concentration of analytes in a suspension high-frequency dielectric

detection avoids the low-frequency dispersion mechanisms, therefore, eliminating the chances of

false data. Integration of these high-frequency sensors with CMOS/BiCMOS technology opens

up the possibility of a new market for POC diagnostics involving detection of concentration of

biomarkers, viruses, analytes and more. Such a market if established, the CMOS integrated high-

frequency biosensors can usher in cheap POC diagnostic products utilizing batch fabrication of

the standard CMOS process. In this thesis, high-frequency capacitive sensors operating at

microwave frequencies and fabricated in standard CMOS process, have been explored for

applications like immunosensor, analyte concentration detection, cytometry.

1.4 CMOS/BiCMOS microwave biosensor platform

The growth of semiconductor industry can be tracked back to the famous paper published by the

INTEL co-founder Gordon Moore in 1965, where he stated, that, “the number of components

that could be incorporated per integrated circuit would increase exponentially with time” [78]

This trend that has been observed since 1970, of number of transistors in an integrated circuit

doubling every eighteen months is called the Moore’s law. As a consequence of this trend,

miniaturization of circuits by scaling down of transistors has been the driving force for

technology advancements. At the same time there has been considerable advancement in

microfabrication technology leading to fabrication of structures of the order of few nano-meters

possible.

The trend of miniaturization with increasing performance is often traded off with power

consumption. This leads to the direction of new transistor concepts, new materials, and is

referred to as “More Moore” [79]. Then there is a second trend, which is characterized by

diversification of functionalities on semiconductor based devices. This trend is referred to as

“More Than Moore”. The additional functionalities incorporated on the semiconductor based

devices also aid in miniaturization and scaling, although not at the same rate as the transistor

scaling. Functional diversification includes interaction with the outside world through sensors

and transducers, and can be implemented with the incorporation of passive components, micro-

electro-mechanical systems (MEMS), surface acoustic wave filters and actuators, microfluidic

integration etc. Thus, More Than Moore is a heterogeneous integration technology of digital and

non-digital applications on the same semiconductor process platform leading to wide variety of

application fields. Establishment of biochips or biosensors is one such field and is depicted in the

technology roadmap in Fig. 1.6.

Chapter 1 Introduction

| 9

The primary advantages of monolithically integrated biochips on a CMOS platform are:

- Signal processing capabilities in close proximity to the sensor

- Batch fabrication for large number of device manufacturing

- Miniaturization of the overall system

- “All-electrical” sensor systems requiring no labeling

Additionally, as mentioned above, if the monolithic CMOS POC diagnostic devices become a

reality for commercial products, low cost can be achieved due to batch fabrication technique.

The concept of CMOS based biochips is shown in Fig. 1.7. The first and second generation

biosensors used functionalized bio-receptor coupled to a transducer. Further microelectronic

circuits were then integrated with the transducer to extract the signal output. The new approach

of complete CMOS biosensor chip involves the elimination of bioreceptors and capability of

using the metal layers of the CMOS process for immobilization or detection of biomolecules.

This is shown in Fig. 1.7 where the biomolecules are shown to be captured on the passivation

layer of the back-end-of-line (BEOL) stack of the CMOS process or the exposed metal layer of

the CMOS process. Complete CMOS biosensors or biochips have then been explored by

researchers for various applications like DNA characterization [80-82], detection of biomarkers

[83], cytometric application [84-86]. DNA sensor developed by Stagni et al. [81] shows an

additional post-processing step of depositing gold electrodes on top of the CMOS chip for DNA

immobilization.

________________________________________________________________________________________

Figure 1.6 Technology road-map showing More Than Moore concept [72].

Chapter 1 Introduction

| 10

On the other hand, CMOS image sensors developed by Hassibi et al. [83] shows the use of

exposed top metal layer of a five metal CMOS process for the capture of biomolecules. The

sensing scheme in the mentioned CMOS sensor is primarily based on the interface capacitance of

the electrodes and the fluid containing the biomolecules.

The advances in RF engineering have further led research groups to explore high frequency

characterization of biological suspensions and biomaterials. As early as 1998, Stuchly et al. [87]

demonstrated biosensors based on waveguide structures. Since then considerable amount of

research work have been devoted to the establishment of microwave based biosensors. Sensors

based on whispering-gallery mode resonator [88], coaxial resonator [89], coplanar lines [90,91],

capacitors [92] have been demonstrated for the distinction of cells and proteins, DNAs,

biomarkers for tumors and cancer etc. Ferrier et al. [93] have also shown interferometric

microwave sensors for detection of single cells.

Fig. 1.8 shows one such approach used by Grenier et al. [91,92]. Coplanar multi-fingered

capacitor is used as a passive sensor structure in order to detect different concentration of cells.

________________________________________________________________________________________

Figure 1.7 CMOS single chip biosensor approach.

________________________________________________________________________________________

Figure 1.8 Passive microwave sensors for detection of concentration of cells [91].

Chapter 1 Introduction

| 11

The capacitive sensor is fabricated on a microwave substrate and a fluidic system is integrated on

top of the sensor. The advantages of microwave sensing towards miniaturization can be

understood from the sensor setup. The dimension of the sensor in the order of micrometers

matches the dimensions of the cells being investigated. Therefore, along with the

miniaturization, the sensitivity is also improved due to the physical dimension. The measurement

setup in such a sensor system involves measuring of the scattering parameters of the passive

capacitor. As mentioned above, the ionic background and the electrode electrolyte interface do

not play any role in microwave sensing. The high frequency of the electrical signal penetrates

through the interface or double layer capacitance, nullifying its influence.

Another often used configuration for microwave detection technique is based on interferometry.

Sensitivity of the order of single cells has been demonstrated by Yang et al. [94]. Such a

configuration of the sensor is shown in Fig. 1.9.

________________________________________________________________________________________

Figure 1.9 Interferometric approach based on passive structures for biosensors [88].

Chapter 1 Introduction

| 12

The interferometric approach offers very high sensitivity due to a differential measurement

technique. The sensor architecture which involves coplanar transmission lines are fabricated on a

microwave substrate and similar scattering parameter measurement is employed as discussed in

case of the capacitive sensor. This is shown in Fig. 1.9. However, all the microwave sensor

configurations are passive structures.

Therefore, the next step is the monolithic integration of microwave sensor architecture on

CMOS/BiCMOS platform. In this thesis work, sensor concepts based on CMOS integrated

microwave sensors are presented. Efforts are being made by few contemporary research groups

around the world to establish CMOS microwave sensors but majorly focused on cytometric

applications like counting of cells [95], or particles based on magnetic makers [96]. This work is

focused on establishing a unique platform for significant biosensor applications like

immunosensors, analyte concentration detection and cytometric applications as well.

A sensing scheme is introduced in the thesis work which is applied to all the sensing applications

explored in this work. The sensing principle is based on the variation of capacitance (here an

interdigitated capacitor) embedded in a CMOS oscillator, causing a shift in the resonant

frequency of the oscillator. The change of capacitance is caused by the variation of fringing

electric fields of the interdigitated capacitor (IDC), due to the change of permittivity on top of it.

The sensor capacitor is fabricated on one of the two top metal layers in a BiCMOS process with

seven metal layers or five metal layers. The sensor topology is shown Fig. 1.10.

For the application of immunosensor detection of creatinine molecule is demonstrated. For the

proposed dielectric immunosensor the permittivity variation is brought about by the different

amounts of anti-creatinine antibodies binding to the creatinine molecules immobilized on the

________________________________________________________________________________________

Figure 1.10 Sensor architecture used in this thesis. Capacitive sensor is embedded in a CMOS oscillator

circuit.

Chapter 1 Introduction

| 13

IDC. In this work, a method to immobilize creatinine molecules on the surface of silicon nitride

(Si3N4) layers has been demonstrated. Si3N4 is the standard passivation layer for CMOS

technologies; therefore, the capability of immobilizing creatinine molecules on its surface helps

to evade any additional post processing steps for future label free sensors used in creatinine

detection.

For the cytometric applications, microfluidic system was integrated with the sensor system for

accurate handling of the probe solution. Polymer based microfluidic system has been used in this

work. A chemical mechanical polishing step is conducted on the fabricated to chip to adhere to

the planarity issues for integration of microfluidic system with the silicon chip. On chip signal

processing capabilities for improved sensitivity and POC system applications is shown in this

work.

1.5 Need for microfluidic integration

Advancement in the field of microfluidic technology has played a key role in the development of

LOC devices. The first advancement of microfluidics towards LOC applications was

demonstrated by Manz et al. [97]. Since then considerable improvement has been made in the

microfluidic technology for precise handling of biological suspensions. Microfluidic platforms

provide a well-defined volume of fluid samples and provide easy handling of fluid samples.

Fluid sample of extremely small volumes can be handled using present microfluidic systems.

The active geometry of the microfluidic systems also matches closely the dimensions of

biological samples (cells, bio-molecules etc.). The main limitation of integration of microfluidic

system with electrical sensors is the overall size of the sensor system, which is often defined by

the size of the microfluidic system.

In this aspect two approaches are being often explored. The first approach is based on silicon

microfluidics. Silicon microfluidics is gaining interest because of easier integration with the

CMOS process technology. Kaynak et al. [98, 99] have explored the integration of a microfluidic

platform on BiCMOS platform with the backside etching of the silicon wafer after the fabrication

of the BiCMOS sensor circuit. Such a technique is extremely advantageous for building

monolithic sensor system and also standard microfluidic components such as microfluidic

switches, valves, mixers, micro-pumps can be implemented on the silicon substrate using

standard CMOS or BiCMOS process steps. However, thinning or polishing of the backside of the

wafer in order to have microfluidic channels of desired dimension is a major challenge in such

systems. Standard 8’’ Silicon wafers have a thickness of 750 µm. Microfluidic channels have

thicknesses of the order 50 µm to 200 µm depending upon the applications. Therefore, in order

to fabricate microfluidic channels of that order based on backside etching of silicon wafer, the

wafer has to be polished or thinned to the order of few 100 µm. This requires complex process

steps. The mechanical stability of the silicon wafer during and after the polishing step is also of

Chapter 1 Introduction

| 14

primary concern, as the thin wafer is prone to bowing and breaking. Therefore, often a less

complex microfluidic integration based on polymer is used for faster system design and sensor

prototype establishment.

The concept of polymer based microfluidic technique was established and shown by Whiteside

et al. [100, 101], for integration of microfluidic systems with sensors based on glass substrate. In

this thesis, polymer based microfluidic technique is used to handle liquid samples and analytes.

In order to integrate the polymer microfluidic systems on BiCMOS silicon chip additional

polishing step of the chip is conducted in order to obtain the topographical planarity required for

the integration process. Polydimethysiloxane (PDMS) is used as the polymer in such

microfluidic approaches and there is a strong affinity between PDMS and glass, resulting in

strong bonding between the glass substrate and PDMS microfluidic system. Therefore, in order

to establish comparable bonding strength between the silicon BiCMOS chip and PDMS

microfluidic system, the process recipe for preparation of the PDMS microfluidic system and the

bonding steps are determined based on the material properties of PDMS and the passivation layer

of the BiCMOS process.

Organization of thesis

The thesis is organized in the following manner. The second chapter deals with the theoretical

aspect of integration of biosensors on a BiCMOS platform. The challenges from the design

aspect and the possible solutions are discussed in this chapter. The choice of the sensor topology

and its functioning is shown.

The subsequent three chapters deal with significant applications of the established biosensor

technology. The first among them is the immunosensor application discussed in chapter 3.

Immunosensors is one of the main applications discussed in the thesis. In this chapter the

application of the sensor operating at 6 GHz as immunosensor for detection of creatinine

concentration is shown. The operation of the sensor system is further compared with the

established optical sensing techniques in terms of sensitivity and dynamic range.

Chapter 4 deals with the sensors operating in the frequency range of 6 GHz to 12 GHz for

primarily detection of concentration of a solute in a suspension. In this aspect glucose sensors are

discussed. Further cytometric application like detection of concentration of particles in a fluid

system has been addressed. A major focus in this chapter is the integration of microfluidic

system with the BiCMOS sensor chip. Possible techniques based on polymer based microfluidic

systems or based on non-conducting wall around the sensor have been explored.

Chapter 5 shows the establishment of a total sensor system, which is capable of particle counting

in a flow-based fluidic system. The main focus of this chapter is the establishment of extremely

sensitive sensor system in order to detect single particles. The other aspect of the sensor system

Chapter 1 Introduction

| 15

is to have DC read-out and the same has been investigated. Further theoretical investigations are

done for the elimination of noise form such biosensors.

Conclusion of the thesis based on the application of the sensors in various biological avenues is

presented in chapter 6. Each application has been concluded depicting the ability of the sensor

architecture to replace the established complex biosensors. An outlook of the work showing the

future perspectives of the thesis is presented in the end.

Chapter 2 Design and Integration

DESIGN and INTEGRATION

2.1 Introduction

In line with the “More Than Moore” diversification of semiconductor technology, integration of

the biosensors on standard CMOS technology platform has the potential to bring in a new market

for POC diagnostic applications [79]. This has been discussed in chapter 1. Bringing together

“biology” and CMOS technology on a single chip is a significant step in terms of integration of

technologies. The integration of biosensors on CMOS technology ensures high yield and

reliability of the overall system; primarily due to the robustness of the established CMOS process

technology [102]. However, it should be emphasized here that the integration of “biology” and

the process steps required for the integration of the biosensor system should not cause any

degradation of the performance of the CMOS active circuit components [103]. CMOS

technology platforms are primarily optimized with the goal of increasing the performance and

yield of the CMOS devices like transistors. However, interaction of biology with the CMOS

active devices can play a significant role affecting their performance [103]. For example, one

such situation is the issue of integration of microfluidic platform with the CMOS/BiCMOS

sensor chip. The fluid in the microfluidic channel influences the passive components and the

active devices in the circuit. Therefore, the design of the sensor chip and the integration of

microfluidics have to be such that the fluid channel has minimum or no interaction with the other

circuit components other than the sensor structure. This consideration has been demonstrated in

this chapter in designing the sensor layout, pad-frame of the chip and the footprint for the

polymer microfluidic system used for sensing of the analytes.

CMOS sensors operating in the frequency range of 5 GHz to 15 GHz are explored in this thesis.

Therefore, influence of “biological” integration on the high-frequency performance of the circuit

components has to be taken into account. One such example is the influence of dielectric loss

due to the sensing material on the overall sensor and the circuit performance. The design of the

sensor circuit has to carefully take into account the dielectric losses that can be incurred during

the sensor operation. This design consideration has been taken into account and demonstrated in

this chapter.

The sensor architecture employed in this thesis is based on a capacitive sensor embedded in a

CMOS oscillator. The oscillator design involves use of inductors as an integral part of the sensor

system. The inductors coupled with the sensor capacitor constitute a resonator, and the

oscillation of the resonator is driven by a pair of transistors. A change in permittivity is detected

by the sensor capacitor which translates into the change in oscillation frequency of the resonator.

The change in permittivity can also influence the inductors by changing its parasitic capacitance.

Chapter 2 Design and Integration

Therefore, the apparent inductance can change or in other words there can be a detuning of the

inductor. The design of the sensor chip layout and the positioning of the “biological” integration

should take into account of this influence.

The sensor used in this thesis is an interdigitated capacitor (IDC). The sensing mechanism is

based on the effect of dielectric materials on the fringing electric fields between the fingers of the

IDC. The analysis and design of the IDC is based on the coplanar transmission line and is

addressed in this chapter.

It is also significant to understand the technology platforms used for the designing of the overall

biosensors. In this aspect it is necessary to understand the back-end-of line (BEOL) metal stack

of the BiCMOS process used. Therefore, this chapter deals with the technology platforms

followed by the design and integration challenges of the biosensor.

2.2 Technology and integration

2.2.1 Technology

The sensors are fabricated in standard BiCMOS technology of IHP namely SG25H1 (0.25 µm

BiCMOS process) and SG13S (0.13 µm BiCMOS process). The technology platforms offer high

performance heterojunction bipolar transistors capable of operating up to 300 GHz. Therefore,

the technology platform is highly suitable for designing high-frequency circuits. Integrated

biosensors on the BiCMOS technology platform bring diversification of the BiCMOS platforms,

which are primarily used for radio-frequency (RF) to microwave communication systems [104].

The performance parameters of the two technologies are given in Table 2.1 and Table 2.2.

Table 2.1: performance parameters of SG25H1

Parameter

High performance

HBT fmax

220 GHz

HBT ft

180 GHz

Maximum HBT breakdown

voltage (BV_CBO)

4.8 V

MIM Capaitor

1.15 ff/µm2

Varactor (Cmax/Cmin)

Chapter 2 Design and Integration

Table 2.2: performance parameter of SG13S

Parameter

High performance

HBT fmax

300 GHz

HBT ft

250 GHz

Maximum HBT breakdown

voltage(BV_CBO)

MIM Capaitor

1.65 ff/µm2

Varactor (Cmax/Cmin)

The transistors constitute the front-end-of-line (FEOL) of the silicon process, with the passive

structures and interconnect fabricated with the metallization layers on the BEOL. Fig. 2.1 shows

the cross-sectional schematic of the BEOL stack of the SG13S and SG25H1 process of IHP.

There are five metallization layers in the BEOL stack for SG25H1 process, while the SG13S

process has seven metallization layers. The top two metallization layers (TM1 and TM2) of both

the processes have higher thickness compared to the other metallization layers. These metal

layers are used to fabricate high quality factor RF passive components like inductors,

transmission lines, etc. The lower metal layers are used for interconnects within the circuits. As

mentioned above, integrated biosensors on such standard BiCMOS/CMOS platform is

considered as major technology advancement and the sensors are termed as the next generation

biosensors. The schematic shown in Fig. 2.2, shows the approach in which the biological

integration is performed on top of the BEOL stack of the CMOS/BiCMOS process. The

_______________________________________________________________________________________

Figure 2.1 Cross section schematic of BiCMOS Back-end-of-line stacks of IHP. a) SG25H1 BiCMOS

process with five metal layers. b) SG13S BiCMOS process with seven metal layers.

a.)

b.)

Chapter 2 Design and Integration

biological integration includes immobilization of proteins or other biomolecules on top of the

passivation layer of the BiCMOS stack or bonding of microfluidic systems carrying biological

suspensions on top of the BiCMOS stack.

One of the metallization layers in BEOL stack is used as the sensor/transducer and the electronic

circuitry for sensor read-out, data acquisition and signal processing is constituted with the

BiCMOS active devices (transistors) and passive components (MIM capacitors, resistors)

integrated in the same stack as mentioned before. Therefore, the next generation biosensors are

essentially single chip solution. The BEOL of the BiCMOS processes of IHP are passivated with

a 400 nm thick layer of silicon nitride (Si3N4) and has a permittivity of 6.7. The intermediate

dielectric between the metallization layers is silicon dioxide (SiO2) with permittivity of 4.2.

Therefore, the biological integration on top of the BEOL processes is separated from the

transistors (at the FEOL of the BiCMOS process), by the above dielectric layers and hence, as

mentioned previously, the biological interaction does not influence the operation of the

transistors.

2.2.2 Influence of biological material and Integration

Planar interdigitated capacitors (IDC) fabricated on the topmost metal layer of the BiCMOS

stack (TM1: top metal 1/TM2: top metal 2) are investigated as sensors in this thesis. In the

immunosensor application explored in this thesis, the sensor is fabricated on the TM1 metal layer

due to surface chemistry requirement for immobilization of biomolecules. While, for the

applications of cytometry and single particle detection, the sensor is fabricated in TM2.

Fabricating the sensor structure comes more from the aspect of biology; the biological

compounds, for e.g., protein molecules in case of the immunosensors are immobilized on top of

________________________________________________________________________________________

Figure 2.2 New sensor approach: Integrated CMOS sensor scheme.

Chapter 2 Design and Integration

the passivation layer of the topmost metal layer. In case of microfluidic interaction where fluids

are analyzed, the microfluidic system is also heterogeneously integrated on top of the BEOL

stack. Therefore, fabricating the sensor structure on the top metal layers ensures proximity of the

sensor to the biological integration. This is significant for the overall sensitivity of the sensor

system.

As mentioned in the previous section, the sensor architecture employed in this thesis is in the

configuration of a resonator where the sensor capacitor is coupled with a pair of inductors. The

schematic of the resonator is shown in Fig. 2.3 (a). The sensor capacitor detects the permittivity

change in the bio-sample. The permittivity change can be brought about by concentration of

particles in a suspension or concentration of immobilized biomolecules on the sensor, etc. This

permittivity change causes a shift in the oscillation frequency of the resonant tank, and is read

out by the CMOS oscillator circuit.

The quality factor of the resonant tank is determined by the quality factors of the individual

components. The fundamental definition of quality factor (Q) is defined as [106],

𝑄=2𝜋𝑓 𝑒𝑛𝑒𝑟𝑔𝑦 𝑠𝑡𝑜𝑟𝑒𝑑

𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑝𝑜𝑤𝑒𝑟 𝑑𝑖𝑠𝑠𝑖𝑝𝑎𝑡𝑒𝑑 (2.1)

where f is the resonance frequency of the LC tank. The frequency shift of the resonant tank is

caused by the real part of the permittivity of the material. However, permittivity is a complex

quantity with two components,

________________________________________________________________________________________

Figure 2.3 a) LC resonant tank circuit with the capacitor acting as the sensor b) quality factor of the LC

tank circuit with and without loss.

a.)

b.)

Chapter 2 Design and Integration

𝜀= 𝜀′+ 𝑗𝜀′′ (2.2)

where 𝜀′ is the real part of the permittivity and 𝜀′′ is the imaginary component of the permittivity

[107]. The imaginary component of the permittivity determines the losses or degradation of

electric field. Therefore, along with sensing of resonance frequency shift there is a variation of

the output electric field power based on the complex permittivity of the material shown in Fig.

2.3 (b). The imaginary part of the permittivity, therefore, influences the quality factor of the

resonator by affecting the loss and in turn the average power dissipation.

The biological suspension carrying the cells, proteins, and other bio-materials are often

extremely lossy [108,109]. With respect to microwave passive structures, the lossy materials

reduce the quality factor of the structure. This is seen in Fig. 2.3 (b), where the magnitude of the

impedance of the tank is seen to reduce with the increasing losses. This is due to increase in

average power dissipated due to the increase of losses.

At this juncture it is important to understand the influence of quality factor on the overall sensor

system. With the reduction in the quality factor, the oscillations of the resonant tank are damped

as shown in Fig. 2.4.

The damped oscillations already originate from the series resistance associated with the inductors

due to finite conductivity of the metallization layers used for the inductor design. The additional

material loss reducing the quality factor of the sensor system further enhances the loss and

results in further damping of the oscillations. From the CMOS active circuit design (oscillator

here) aspect, used to read out this oscillating frequency, there should be compensation

mechanism for the losses. The active circuit of Fig. 2.5 is such a compensation mechanism. The

CMOS oscillator topology employs the strategy of negative resistance to compensate for the

______________________________________________________________________________________

Figure 2.4 Damped oscillation due to loss. The losses are incurred due to the series resistance

accompanying the inductors as well as the imaginary part of the permittivity of the biomaterial.

Chapter 2 Design and Integration

losses [110]. However, here it should be mentioned that the negative resistance compensation

should be comparatively higher compared to the oscillators designed for communication circuits.

In this case the dielectric loss on the sensor should also be taken into account. One can run into

the problem of overcompensating the losses or in other words having a very high compensation

for low loss materials. In that case, the output of the oscillator will not remain sinusoidal due to

saturation of the oscillator. However, in this thesis the main objective is to determine the

variation in the oscillation frequency of the oscillator, therefore, the loss of sinusoidal output of it

is not of high significance. Water was considered to be the dielectric material with highest losses

and air was considered to be with minimum dielectric loss. The oscillator design was performed

to accommodate this range of losses.

More on this is dealt within the subsequent section where the circuit design aspect in the thesis

has been addressed. Therefore, the lossy biomaterial placed on top of the sensor structure plays a

significant role in determining its quality factor and in turn in the performance of the overall

sensor oscillator.

Now it is necessary to understand how the theoretical aspects affect the integration of the sensor

system. As mentioned above the capacitive sensor is generally fabricated on the top metal layers

for high sensitivity. However, if the inductors are placed very close to the sensor, the

biomaterials might actively influence the performance of the inductors. The real part of the

permittivity influences the parasitic capacitance of the inductor, therefore, detuning the value of

the apparent inductance of the inductors. This effect modifies the oscillating frequency of the

oscillator. Therefore, an unwanted parameter gets added to the sensor system, where the shift in

the oscillating frequency has to be mapped to the change in capacitance of the sensor capacitor

and the partial change in the inductance of the inductors, depending on the area of coverage of

the inductor by the biomaterials.

______________________________________________________________________________________

Figure 2.5 Negative resistance compensation done by the active circuit in order to sustain the oscillation

of the resonance tank.

Chapter 2 Design and Integration

Qualitatively this might seem useful, as an additional parameter might increase the sensitivity of

the sensor system, however, quantitatively this is an extremely tricky situation, where, extracting

the true material parameters (concentration, type) from the oscillation frequency shift become

difficult. The imaginary part of the permittivity on the other hand will influence the quality factor

of the inductor. The quality factor of the inductors is degraded by the imaginary part of the

permittivity. Hence, the overall degradation of the quality factor of the resonator is considerably

high due to the degradation of quality factor of both the sensor capacitor and now additionally

the inductors. These effects call for placement of the inductors far from the sensor capacitor.

This is shown in Fig. 2.6 where the inductors are placed far from the sensor. This increases the

area budget of the overall chip.

It is also seen from the sensor chip layout, that the pad-frame for electrical connections is placed

on one side of the chip. And the placement of the pads is made at a distance of 1 mm from the

sensor area. This is done for effective microfluidic integration without leakage. The microfluidic

foot-print is made in a way such that the bonding area of the microfluidic system with the chip is

substantially large to avoid any kind of fluid leakage from the microfluidic channel.

1 mm

_______________________________________________________________________________________

Figure 2.6 Typical layout of a sensor system. The distance of the inductors from the sensors is 500 µm.

This evades the interaction of biomaterials with the inductors.

Chapter 2 Design and Integration

2.3 Design of planar microwave sensors

2.3.1 Coplanar Transmission lines

The model of coplanar transmission lines is used for analyzing the behavior of the interdigitated

capacitor. Therefore, it is significant to delve into the analytical model of the coplanar

transmission lines. The Fig. 2.7 shows a typical coplanar transmission line placed on a substrate

and covered with a dielectric. Theoretical studies of coplanar transmission lines can be done

using a full wave analysis or a quasi-static analysis [111]. The first theoretical analysis of

coplanar structures was done by Wen et al. using conformal mapping technique. Since then,

conformal mapping technique has been a useful approach in order to derive the parameters for

coplanar transmission line.

Conformal mapping technique offers the advantage of converting open geometries into closed

geometries [112], thereby, enabling the derivation of design equations. The Schwarz-Christoffel

transformation method of conformal mapping is often used for the planar transmission lines like

the coplanar transmission lines. The Schwarz-Christoffel transformation is used to map the the x-

axis into a polygon and the upper half of z plane (that is y>0) is mapped as the interior of the

polygon. This technique is extremely useful for mapping of the capacitance of the transmission

line arising due to the fringing fields.

The partial capacitances of the coplanar transmission line shown in Fig. 2.7 can be written as

follows:

C0: This is the capacitance of the structure without any dielectric that is with air as the

surrounding.

C1: This is the capacitance of the structure with only dielectric 𝜀2 of height h1

Figure 2.7 Coplanar transmission line sandwiched between two dielectrics.

Gnd

Chapter 2 Design and Integration

C2: This is the capacitance of the structure with only dielectric 𝜀1 of height h2

The total capacitance of the coplanar transmission line structure is given as:

𝐶=𝐶0+ 𝐶1+ 𝐶2 (2.3)

The individual configurations of the partial capacitances is shown in Fig. 2.8

As described above, the partial capacitances can be obtained by the conformal mapping

techniques [111,112]. The partial capacitance with no dielectric is given as,

𝐶0=4𝜀0𝐾′(𝑘)

𝐾(𝑘) (2.4)

where, K is the first kind of complete elliptical integral and K’(k) = K(k’). The geometry of the

coplanar structure defines the parameters k and k’ and is given as,

𝑘= 𝑐

𝑏 √𝑏2−𝑎2

𝑐2−𝑎2 (2.5)

𝑘′= √1 − 𝑘2 (2.6)

The partial capacitances due to the dielectric layers can be calculated using similar conformal

mapping techniques, however, the function for integral is dependent on the geometry of the

coplanar transmission line and the height of the dielectric. The partial capacitance C1 is given as

Figure 2.8 Partial capacitances of the coplanar transmission line showing the contribution of the

individual dielectric layers

Chapter 2 Design and Integration

𝐶1=2𝜀0(𝜀2−1)𝐾′(𝑘1)

𝐾(𝑘1) (2.7)

The same laws of complete elliptical integral applies. The function k1 is give as

𝑘1=sinh (𝜋𝑐

2ℎ1)

sinh ( 𝜋𝑏

2ℎ1) √𝑠𝑖𝑛ℎ2(𝜋𝑏

2ℎ1)−𝑠𝑖𝑛ℎ2(𝜋𝑎

2ℎ1)

𝑠𝑖𝑛ℎ2(𝜋𝑐

2ℎ1)−𝑠𝑖𝑛ℎ2(𝜋𝑎

2ℎ1) (2.8)

On the same lines, the partial capacitance C2 is given as

𝐶2=2𝜀0(𝜀1−1)𝐾′(𝑘2)

𝐾(𝑘2) (2.9)

The k2 can be derived in the same way as k1 was derived.

2.3.2 Interdigitated Capacitor

The typical structure of an interdigitated capacitor (IDC) is shown in Fig. 2.9. The IDC relies on

strip-to-strip capacitance of parallel conducting fingers. The fringing electric fields between the

fingers of the IDC penetrates into the material placed on top of it and is varied based on the

permittivity of the material.

The significance of choosing IDC as the sensor element can be summarized as follows:

- at the operating sensor frequency, the dimensions of the IDC are of the order of

micrometers, matching the dimension of biomaterials, for e.g., cells, fluid volume etc.

- feasible design of high quality factor IDC structures with high conductive top metal

layers of the BEOL stack

______________________________________________________________________________________

Figure 2.9 Geometry of the interdigitated capacitor showing the spacing and width of the fingers.

Chapter 2 Design and Integration

- purely capacitive structure enables design of simpler read out techniques, for e.g.,

CMOS oscillators

- Penetration depth and capacitance density can be controlled based on the geometry of

the IDC

The analysis of the IDC can be extended from the analysis done for the coplanar transmission

lines in the previous section using the same conformal mapping technique. Conformal mapping

technique can be used in the quasi-static condition. For quasi-static approximation the distance

between the fingers of the IDC on the same electrode (also called the spatial wavelength of the

IDC) has to be smaller than the operating wavelength of the IDC. In this thesis, the operating

wavelength ranges from 60 mm to 20 mm (operating frequency: 5 GHz to 15 GHz). The spatial

wavelength of the IDC being in the range of µm, enables the use of conformal mapping

technique for establishing the governing equations of capacitance of the IDC. Analysis of IDCs

has been done for decades now [112-114]. The conformal mapping, transforms the planar IDC

structure into an orthogonal plane where the planar capacitance is translated to a parallel plate

capacitor. However, in order to understand and correlate the results from the conformal mapping

approach it is significant to understand the model of the IDC on the BICMOS platform used in

this thesis.

Considering, the sensor is fabricated on the top metal layer as is the case in most of the thesis

work, (for the case of immunosensor development, the IDC is in TM1 and the analytical

expression is presented in the concerned chapter) the IDC configuration for sensor operation is

shown in Fig. 2.10.

_______________________________________________________________________________________

Figure 2.10 Cross-sectional schematic of the IDC fabricated on TM2 metallization layer of BEOL stack. The

fringing fields penetrate into the biomaterial placed on top of the passivation layer.

Chapter 2 Design and Integration

From Fig. 2.10 it is seen that the fringing electric fields penetrates into the BEOL SiO2 dielectric

at the bottom. On the top the fringing electric fields penetrate into the Si3N4 passivation layer and

the material placed on top of it. Therefore, the IDC can be analyzed using conformal mapping

technique by considering all the dielectric layers with specified heights. The model used for the

analysis of the IDC is shown in Fig. 2.11

As seen from Fig. 2.11, the cross section of the IDC has two sets of electrodes with voltages V

and –V respectively. The equipotential planes V=0, are the perpendicular planes half-way

between the electrodes. The expression for the capacitance between the electrodes and the

ground potential (equipotential plane) with a dielectric layer of height h can be established by

mapping the planar surface to a parallel plate capacitor using the Schwarz-Christoffel transform

technique. The overall capacitance of the IDC can be evaluated as follows

C0: Capacitance with no dielectric layer, that is, air with infinite thickness

C1: Capacitance with material under test of height h1

C2: Capacitance with Si3N4 passivation, of height h2

Figure 2.11 Model of IDC for evaluation of the capacitance fabricated on TM2 of the BiCMOS stack. The IDC

electrodes have Si3N4 and MUT on top and SiO2 at the bottom. The equipotential surfaces are marked

with vertical dotted lines. The two sets of electrodes are at the potential V and –V.

_______________________________________________________________________________________

Chapter 2 Design and Integration

C3: Capacitance with SiO2 below, of height h3

The total capacitance is the sum of the partial capacitances multiplied by the length L of the

fingers. Additionally, the parallel plate capacitance (Cpp) due to the thickness of the fingers needs

to be taken into account.

𝐶𝐼𝐷𝐶=𝐿 (𝐶0+ 𝐶1+ 𝐶2+𝐶3)+𝐶𝑝𝑝 (2.10)

The partial capacitances can be derived in the same way as was explained in the previous section

for coplanar transmission lines. For number of fingers N >3, the total capacitance is obtained by

multiplying CIDC with a factor (N-3).

The total capacitance can be approximated to,

𝐶𝐼𝐷𝐶_𝑇𝑜𝑡𝑎𝑙=2𝜀0(𝑁−3)𝐿(𝐾′(𝑘)

𝐾(𝑘)+ (𝜀𝑀𝑈𝑇−1)𝐾′(𝑘1)

𝐾(𝑘1)+(𝜀𝑆𝑖3𝑁4−𝜀𝑀𝑈𝑇)𝐾′(𝑘2)

𝐾(𝑘2)+(𝜀𝑆𝑖𝑂2−

1)𝐾′(𝑘3)

𝐾(𝑘3))+(𝑁−3)𝐶𝑝𝑝 (2.11)

The term K is the complete elliptical integral function and k is the geometry and height of

dielectric layer dependent term, as shown in coplanar transmission line derivations. The

influence of the dielectric permittivity of the material under test (MUT) is seen in eq. 2.11. The

Figure 2.12 The partial capacitance components of the IDC for individual dielectric layers.

_______________________________________________________________________________________

Chapter 2 Design and Integration

variation of this permittivity affects the fringing electric field between the fingers of the IDC and

changes its capacitance.

Along with the geometry of the IDC determining the capacitance, another very significant

parameter of the sensor structure designed for high-frequency sensing applications is the self-

resonating frequency (SRF) of the structure. SRF of a structure is defined as the frequency at

which the capacitive contribution of the structure is nullified by its self-inductive contribution

and the structure is purely resistive. This phenomenon is analogous to the general resonant LC

tank circuit operation; however, in the case of self-resonance the capacitive and the inductive

contributions of the same structure are taken into account. The self-resonating phenomenon is

shown in Fig. 2.13 (a).

The resonance peak in the capacitance vs. frequency curve defines the frequency at which the

IDC’s self-inductance nullifies the capacitive contribution. Beyond the self-resonance

phenomenon, as seen in Fig. 4.2 (a), the capacitance is negative, or in other words the structure is

inductive. The SRF is dependent on the size of the sensor and also on the permittivity ambient of

the IDC. The self resonating frequency of the IDC sensor structure decreases with the increase in

its area. However, for a given geometry of the IDC on a specific substrate, the SRF is only

dependent on the permittivity of the material placed on top of it (MUT). With the increase in the

permittivity of the MUT the capacitance of the IDC increases and the SRF reduces as shown in

Fig. 2.13(a). It is significant to design the operating frequency of the sensor system considerably

lower than the SRF of the IDC. This is essentially because in such a condition the electric field in

the IDC is rotation free and the structure is essentially capacitive. The simulated SRF of typical

IDC structures used in this work is beyond 150 GHz and the simulation of one such IDC

structure is shown in Fig. 2.13 (a). In this chapter three sensor systems operating in the frequency

range of 6 GHz – 15 GHz are demonstrated. Therefore, the operating frequency range of the

sensor systems is sufficiently below the SRF of the corresponding sensor IDCs. A typical

Figure 2.13 (a) Self resonance phenomenon of a typical IDC structure. The self-resonance frequency

(SRF) is around 150 GHz (b) Variation of the capacitance of the IDC with respect to permittivity at

the operating frequency range of 6 GHz – 12 GHz.

______________________________________________________________________________________

a.)

b.)

Chapter 2 Design and Integration

variation of the capacitance of the IDC with respect to the increase in permittivity of the MUT is

shown in Fig. 2.13 (b). The direct proportionality of the IDC capacitance to the permittivity of

the MUT can be seen in eq. 2.11, where the capacitance of the IDC was analyzed.

2.4 Design of sensor circuit

The sensor circuit is based on a standard cross-coupled oscillator topology where the sensor IDC

is analogous to the varactor (variable capacitor) used in such a circuit. In a standard voltage

controlled oscillator, the capacitance of the variable capacitor is tuned (varied) using a tuning

voltage, resulting in the resonance frequency scaling of the oscillator. In this thesis, the

resonance frequency of the oscillator is controlled by the variation of the permittivity on top of

the IDC. The circuit topology can be seen in Fig. 2.14. An LC resonant tank structure is

constituted by connecting the multi-fingered IDC in parallel to a couple of series connected

inductors as shown in Fig 2.14. A pair of cross-coupled n-MOS transistors drives the oscillation

of the LC resonant tank. This configuration is a typical negative transconductance oscillator.

The model of the negative transconductance oscillator is shown in Fig. 2.15(a). The model

represents the oscillator as two interconnected one-port networks: an LC tank which acts as the

frequency selective resonator and an active network represented by the cross-coupled oscillator.

In the steady state condition, loss conductance (G’) generated in the tank circuit should be

compensated by the negative transconductance generated by the active circuit, which enables

Figure 2.14 The sensor circuit with IDC sensor embedded in a cross coupled CMOS oscillator.

_______________________________________________________________________________________

Chapter 2 Design and Integration

sustaining of the oscillations. The inductor being made from a metal of finite conductivity

generates the losses due to the resistance of the metal.

The standard approach to represent an oscillator is given by a simple linear feedback system

[110], with an overall transfer function given as,

𝑌(𝑠)

𝑋(𝑠)=𝐻(𝑠)

1−𝐻(𝑠) (2.12)

where, X(s) and Y(s) are the frequency domain representations of the input x(t) and output y(t)

respectively. H(s) is the frequency transformation of the impulse response of the system h(t). The

commonly used Barkhausen’s criteria for sustained oscillation at a frequency  is written as,

1. Loop gain, |H(j)| must be equal to unity

2. Total phase shift around the loop, must be equal to 0° or 180°

The transconductance (gm) of the transistor compensates the losses of the tank circuit. This is the

negative transconductance used in this kind of oscillator design. From the energy point of view,

the active circuit replenishes the energy dissipated periodically in G’, thus, enabling a sustained

Figure 2.15 (a) Model for negative transconductance cross coupled oscillator without considering

dielectric losses. (b) Model for negative transconductance cross coupled oscillator with considering

dielectric losses.

_______________________________________________________________________________________

Chapter 2 Design and Integration

oscillation of the cross-coupled oscillator. Therefore, the transconductance of the cross coupled

pair should be equal to or more than the negative value of G’, (gm ≥ -G’).

In this thesis, the IDC sensor is used as the variable capacitor. The permittivity of the biomaterial

placed on the top of the sensor is written as

𝜀𝑀𝑈𝑇= 𝜀𝑀𝑈𝑇

′+𝑗𝜀𝑀𝑈𝑇

′′

As described in the previous section the real part of the permittivity influences the resonant

frequency of the oscillator. The imaginary part of the permittivity accounts for the dielectric

losses. This dielectric loss is in addition to the resonant tank losses mentioned above. This is

modelled as the additional loss conductance in parallel to the tank loss conductance. This is

shown in Fig. 2.14(b). Therefore, the negative transconductance required to compensate for the

overall losses is higher than the negative transconductance required to compensate the tank

losses. This requires designing of the cross coupled transistor pair with wider channel widths.

The transconductance of a transistor is directly proportional to the width of the transistor [110].

The transconductance of the cross coupled pair should compensate for the maximum losses that

can be incurred during the sensing process. In this thesis, biological samples are primarily

measured in water, which acts as the aqueous suspension for measurement. The cross coupled

pair of nMOS transistors is designed in such a way, that the transconductance of it compensates

the losses for water. As addition of biological materials in water reduces the loss factor, the

maximum loss for which the negative transconductance should be designed is water.

However, from the standpoint of the design of the oscillator, the high transconductance can lead

to loss of sinusoidal nature of the output, for low loss materials. In that case there is an

overcompensation due to the high transconductance of the cross coupled pair. This is however,

not significant for the thesis work, as the resonant frequency of the oscillator is not affected by

this loss of sinusoidal nature. In this work, the resonant frequency of the oscillator and its

variation for different materials is of importance. The cross coupled nMOS pair brings in

additional parasitic capacitance. The parasitic capacitance is determined by the width of the

transistors and the biasing of the transistors. The additional parasitic capacitances affect the

sensitivity of the sensor as the capacitance is in parallel to the sensor capacitor. The nMOS

transistors at the following buffer stage also add parasitic capacitance. A buffer stage is used

following the oscillator, in order to isolate the sensor circuit from additional circuitry following

the sensor for advanced operations (see chapter 5). The resonant frequency of the oscillator is

given as,

𝑓= 1

2𝜋√2𝐿𝐶𝑇𝑜𝑡𝑎𝑙 (2.13)

The total capacitance CTotal, is the sum of the capacitances due to the IDC and the parasitic

capacitances due to the transistors.

Chapter 2 Design and Integration

The same oscillator circuit topology with the embedded IDC sensor is used all through-out the

thesis work. In the subsequent chapters the use of the sensor and the oscillator circuit in total

have been explained for various applications like immunosensors, cytometry, etc.

Chapter 3 Dielectric Immunosensor

DIELECTRIC IMMUNOSENSOR

In this chapter* a CMOS high frequency direct immunosensor operating at 6 GHz (C-band) is

discussed. The functioning of the sensor is shown by the label free determination of creatinine.

The sensor is fabricated in standard 0.13 µm SiGe:C BiCMOS process. The ability to immobilize

creatinine molecules on Si3N4 passivation layer of the standard BiCMOS/CMOS process evades

any further need of cumbersome post processing of the fabricated sensor chip. The sensor is

based on capacitive detection of the amount of non-creatinine bound antibodies binding to an

immobilized creatinine layer on the passivated sensor. The chip bound antibody amount in turn

corresponds indirectly to the creatinine concentration used in the incubation phase. The

determination of creatinine in the concentration range of 0.88 – 880 µM has been successfully

demonstrated. A sensitivity of 35 MHz/10-fold increase in creatinine concentration (during

incubation) at the center frequency of 6 GHz has been manifested by the immunosensor. The

results have been compared with a typical optical measurement technique and the sensitivity is of

the order of established optical indication technique. The C-band immunosensor chip comprising

an area of 0.3 mm2 reduces the sensing area considerably, therefore, requiring sample volume as

low as 2 µl. The small analyte sample volume and label free approach also reduce the

experimental costs in addition to the low fabrication costs offered by standard fabrication

technique of CMOS/BiCMOS process.

3.1 Introduction

ELISA (Enzyme Linked Immunosorbent Assays) has been established as the most standard

technique in medical or clinical diagnostic processes over the last decade. The most common

technique applied in the ELISA approach is based on chromatographic assays and is primarily

used in point-of-care-testing (POCT) [115-118]. However, often such ELISA based techniques

give qualitative or semi-quantitative analysis, therefore, limiting the applications to diagnose

diseases depicting high changes in analyte concentrations. Hence, an ever increasing need for

sensors with qualitative analysis is pressing. Immunosensors are such special “biosensors” based

on selective antibody-antigen binding and providing concentration dependent or quantitative

information. The fundamental scheme of an immunosensor is shown in Fig. 3.1, depicting the

highly interdisciplinary approach of designing such specialized immunosensor. In the first step,

antibodies or antigens are bound on a functionalized transducer. Selective pairing of a particular

antigen-antibody pair in aqueous solution governed by a specific chemical reaction is sensed

using one of the following generalized techniques: optical, chemical, amperometric, dielectric

etc. The sensed signal is detected using a parameter change for the

 Parts of the chapter has been published as “Label free sensing of creatinine using a 6 GHz CMOS near-field dielectric

immunosensor”, Analyst, 2015, 140, 3019-3027 DOI: 10.1039/c4an02194k

Chapter 3 Dielectric Immunosensor

particular sensing scheme used: for e.g., fluorescence marker for optical technique, cyclic

voltammetry signals for amperometric technique, capacitance shift for dielectric measurements,

etc.

The final step involves the conversion or amplification of the detected signal into a reasonable

electrical signal for further signal processing and read-out or display. The cost and complexity of

such a sensor system is often governed by one of the four steps described above. Optical sensing

schemes are being investigated presently to establish immunosensors for POCT. The first

approach includes ELISA-like immunosensor scheme for an autonomous LOC device containing

all of the required reagents (e.g., buffers, enzyme or fluorophore antibody conjugates, etc.),

separation units, pumps, channels and sensors [119, 120]. The second approach also incorporates

labeling compounds (e.g., enzymes, fluorophores, redox mediators) but, in a one-step assay

format [121]. In case of optical technique, the optical markers and the amount of antibodies used

in the sensor system make the application expensive. The above class of immunosensors can be

also classified as indirect immunosensors where the labeling compound aids in the sensing

Figure 3.1 Schematic of a basic immunosensor. The first step includes immobilization of

antibodies/antigens on a transducer surface followed by the binding of corresponding

antigens/antibodies. A sensing scheme is employed to detect the antigen-antibody pair. A signal

processing front-end circuit converts the detected signal to electrical signal.

_______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

scheme. Transduction techniques like surface plasmon resonance (SPR) is also being

investigated recently. Such a technique evades labeling compounds as in the above techniques,

and gives a direct indication of antigen-antibody binding, also referred to as direct

immunosensor. The complexity and cost of such a system arise from the measurement test-bench

and the necessary front end circuit [122]. In parallel to the on-going research for new

immunosensor techniques, there already exists established commercial immunosensors, for e.g.,

BIAcore. Such established state-of-the-art immunosensors require incubation steps as well as

manual pipetting steps [123, 124], therefore, increasing the overall development cost and

complexity.

As a significant application of the developed high-frequency sensors for biomedical applications

shown in this thesis work, establishing a miniaturized direct immunosensor has been a prime

focus. Theoretical studies of microwave interaction with biological materials like biomolecules,

cells, tissues, have been studied in detail over decades and there is considerable volume of

related literature [125, 126]. In this part of the work we explore the design of a single chip

immunosensor for the detection of creatinine molecules. The sensor is made to operate at the

frequency range of 6 GHz. The choice of frequency is made 6 GHz, because the dielectric

permittivity of the aqueous solution (water) used in this work is around 70. The dielectric

permittivity of the given volume of antigens and antibodies in this frequency range is of the order

of 2 to 3. This provides a considerable permittivity contrast for high sensitive immunosensor

design.

3.2 Proposed CMOS compatible immunosensor approach

This section of the thesis demonstrates the development of fully integrated CMOS compatible

immunosensor platform based on high-frequency (6 GHz) dielectric sensing; the sensor platform

has been used to detect creatinine molecules. The primary advantage of using a high-frequency

technique is the miniaturization of the overall sensor system leading to the use of extremely

small sample volumes of the antigen and antibodies. As mentioned previously the “all-electrical”

sensing approach nullifies the need of using labeling markers for detection and also the high

frequency dielectric detection is independent of reference electrodes used for measurements. In

order to develop real miniaturized LOC device, the above features of a high-frequency

immunosensor based on dielectric detection is highly lucrative. The sensing approach is based on

the capacitive detection of dielectric permittivity change. The capacitive sensor is embedded in a

CMOS oscillator, where the sensor acts as a variable capacitor (varactor) and the capacitance

change is translated to the resonant frequency shift of the oscillator.

The functionality of the sensor is determined by detection of the concentration of creatinine

molecules in a competitive immunoassay technique. Creatinine is one of the most often

determined parameters in clinical diagnostic. This is primarily because the concentration of

creatinine in serum and urinary excretion is less influenced by dietary changes such as a high

intake of a creatinine-free diet, unlike urea or nitrogen residues. Creatinine is the index for renal

Chapter 3 Dielectric Immunosensor

glomerular infection; the concentration range in the plasma of a healthy adult person is around

25-150 µM or 2-17 µg mL-1 [127]. However, the concentration is dependent on the age, sex as

well as the demography. This concentration range changes radically for patients with serious

kidney disorders and general debilitation, in which the creatinine concentration increases in the

serum/plasma and decreases in the urine. Standard chemical or optical techniques can be

successfully used to determine creatinine concentration and are considered the gold standard,

especially the Jaffe method [128,129] but with high cost and large complexities. In contrast, the

proposed CMOS high-frequency immunosensor offers a flexible, easy to handle miniaturized

solution with higher detection range of measurement and comparable sensitivity. The target of

this work is to demonstrate the capability of the established CMOS dielectric immunosensor to

detect and screen the increase in creatinine concentration in serum.

3.2.1 Fabrication and operation of the sensor

A multi-fingered planar interdigitated capacitor (IDC) is used as the prototype sensor for

capacitive detection of concentration of creatinine molecules. The IDC along with embedding

oscillator for read-out, has been fabricated in the standard 0.13 µm SiGe:C BiCMOS technology.

The BEOL stack of the process with seven metal layers is shown in Fig. 3.2(a) as was described

in chapter 2. The five lower metal layers are thinner (400 nm) compared to the top metal layers

marked as TM1 and TM2 in Fig. 3.2 (a). The thickness of TM1 and TM2 are 1.5 µm and 2 µm

respectively. The passivation layer of silicon nitride (Si3N4), of thickness 350 nm, on top of TM2

isolates the electrical circuit from the external environment. The planar IDC is fabricated on

TM1 metal layer of the BEOL stack. The reason for fabricating the IDC on the TM1 metal layer

stems from the need of immobilization of creatinine molecules. The surface chemistry developed

for the immobilization technique is based on the dielectric stack comprising of a layer of Si3N4

and SiO2. As seen from Fig. 3.2, the TM1 metal layer has a layer of SiO2 and Si3N4 on top of it.

The IDC is further coupled with a pair of inductors fabricated on TM2 metal layer to form a

resonating tank. The sensor circuit, that is the cross coupled oscillator has been explained in

chapter 2.

Chapter 3 Dielectric Immunosensor

Fig. 3.2 (b) shows the geometry of the planar IDC sensor used in the immunosensor design. The

IDC has five fingers (electrodes) with finger length (L) of 100 µm. The designed IDC has equal

electrode width (w) and inter-electrode spacing (s) of 20 µm. The wider fingers and the gap is

required for high penetration depth of the fringing fields. This is needed for the fringing fields to

extend the SiO2 and Si3N4 layers in order to sense the immobilized molecules on top of the Si3N4

layer. The thickness of the TM1 metal layer being 2 µm. The parallel plate capacitance between

adjacent fingers with silicon dioxide (SiO2) as the dielectric between the fingers, is shown as Cox

in the semi-infinite model of the IDC shown in Fig. 3.2 (c). The fringing electric fields between

adjacent electrodes penetrating into the BEOL oxide layer in the bottom gives rise to the

capacitance due to the substrate also shown as CSiO2_Bottom in Fig. 3.2 (c). The fringing electric

fields between the adjacent fingers on top penetrates into the top oxide layer (SiO2), followed by

the layer of Si3N4 passivation and the dielectric environment above that defined as the material

under test (MUT). This fringing field gives rise to the capacitive contribution shown as CSiO2_top,

CSi3N4 and CMUT in Fig. 3.2(c). Therefore, the total capacitance of a unit cell of the IDC

(comprising two adjacent electrodes) per unit length is given as

Figure 3.2 IDC sensor on BiCMOS back-end-of-line (a) Schematic of BiCMOS back-end-of-line stack

with seven metallization layers. The top two metallization layers are thick and are less resistive. (b)

Geometrical schematic of the IDC showing the length, spacing and width of the fingers. (c) Schematic

of the immobilized creatinine on the sensor surface.

_______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

𝐶𝐼𝐷𝐶 = 𝐶𝑜𝑥 +𝐶𝑆𝑖𝑂2_𝐵𝑜𝑡𝑡𝑜𝑚 +𝐶𝑆𝑖𝑂2_𝑡𝑜𝑝 +𝐶𝑆𝑖3𝑁4 +𝐶𝑀𝑈𝑇 (3.1)

The geometry of the IDC structure determines the penetration depth of the fringing electric field.

The penetration depth (Pd) of the fringing fields from IDC is defined by the following equation

[114].

𝑃𝑑𝐼𝐷𝐶 =𝑤+𝑠

2𝜋 (3.2)

PdIDC is the penetration depth of the fringing fields from the IDC with finger width w and

adjacent finger spacing of s. From the fabricated geometry of the IDC the penetration depth is

calculated to be 8 µm. Therefore, the fringing fields in the bottom accounting for Csub

contribution in eq. (3.1) penetrate only into the oxide layer of the BEOL and do not extend to

the silicon substrate. On the top, the SiO2 above TM1 is 5 µm thick followed by the passivation

layer of Si3N4 350 nm thick. Hence the fringing electric fields penetrate through the oxide and

the passivation into the biomolecules immobilized on top of the passivated IDC surface. The

capacitance of the IDC fabricated in TM1 can be derived following the same principles as was

done in chapter 2. However, the correction term arising due to the SiO2 layer on top of the TM1

has to be included. The equation obtained in eq. 2.11 is now replaced by the additional term

included considering the influence of SiO2 and is given as,

𝐶𝐼𝐷𝐶_𝑇𝑜𝑡𝑎𝑙 = 2𝜀0(𝑁 −3)𝐿(𝐾′(𝑘)

𝐾(𝑘)+ (𝜀𝑀𝑈𝑇 −1)𝐾′(𝑘1)

𝐾(𝑘1)+(𝜀𝑆𝑖3𝑁4 −𝜀𝑀𝑈𝑇)𝐾′(𝑘2)

𝐾(𝑘2)+(𝜀𝑆𝑖𝑂2 −

𝜀𝑆𝑖3𝑁4)𝐾′(𝑘3)

𝐾(𝑘3)+(𝜀𝑆𝑖𝑂2 −1)𝐾′(𝑘4)

𝐾(𝑘4))+(𝑁 −3)𝐶𝑜𝑥 (3.3)

This equation can be understood from the model of the IDC shown in Fig. 3.2 (c), depicting the

various dielectric layers and the sensing layer. The immediate influence of fabrication of the

sensor structure in the TM1 metal layer is the loss of sensitivity due to the influence of the oxide

layer. The SEM image of the fabricated chip shown in Fig. 3.3 depicts the bond pads and the

inductors on TM2 and a focused ion beam (FIB) cut was performed to demonstrate the IDC on

TM1.

Figure 3.3 Scanning electron microscopy (SEM) image of the sensor chip showing the inductors on

topmost metal layer. A focused ion beam (FIB) cutting is performed to expose the IDC sensor surface.

______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

The creatinine molecules and the binding antibodies are of the order of 10 - 30 nanometers, and

therefore, can be conveniently sensed by the IDC when on top of the passivation layer because of

the penetration depth of the fringing fields.

The sensing scheme shown in Fig. 3.4 shows the condition of permittivity variation on top of the

IDC. Following the fabrication of the sensor chip, creatinine molecules were immobilized on the

passivated surface of the IDC as shown as the first step of the sensing scheme in Fig. 3.4. A

competitive immunoassay like approach is used to detect the creatinine concentration. In such an

indirect binding technique, initially, anti-creatinine molecules are incubated in different

concentrations of creatinine, shown in the incubation phase in the Fig. 3.4. Four concentration

levels of creatinine (0.88 µM to 880 µM) were chosen for incubation. This was done to

demonstrate the wide detection range of the sensor. The resultant incubated solution is pipetted

on top of the sensor IDC with previously immobilized creatinine molecules on top of it. Using a

higher concentration of creatinine for incubation of the same amount of anti-creatinine antibodies

would result in larger fraction of the antibodies binding to the incubating creatinine molecules.

Therefore, a smaller fraction of the antibodies binds to the immobilized creatinine molecules on

the IDC surface. An opposite effect is obvious when the concentration of the incubating

creatinine is less.

Figure 3.4 Sensor operation of the dielectric immunosensor. Creatinine molecules had been

immobilized on the passivated surface of the sensor. Anti-creatinine antibodies were incubated in four

different concentrations of creatinine molecules (pre-treatment phase). The four different antibody

solutions are allowed to bind to the immobilized creatinine molecules. Antibody samples incubated

with higher concentration of creatinine have less free antibodies left to bind to immobilized creatinine

molecules.

______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

This indirect binding technique of the previously incubated anti-creatinine antibodies has been

utilized for the detection or sensing of the concentration of creatinine molecules used for

incubation. The assay principle used here to bind the non-complexed free antibodies to a surface

immobilized antigen is well established in clinical diagnostics. Many commercially available

ELISA test kits for measurement of low molecular weight analytes are based on such principle.

Following the phase of pipetting of the incubated anti-creatinine antibody solution, the chips

were washed with water and dried. In order to test the binding of the antibodies to the chip

immobilized creatinine molecules, anti-mouse-antibody peroxide conjugate was added to the

chip. Binding of this conjugate to the anti-creatinine antibodies ensured that the anti-creatinine

antibodies were bound by the chip immobilized creatinine molecules and not released by any

process of denaturation occurring due to the drying phase as explained above.

An additional droplet of water has been added during the electrical measurement of the sensor

chips and is shown in the sensing scheme as the water molecules. The use of additional droplet

of water droplet causes intrinsic sensitivity amplification due to the stark permittivity contrast of

water and the antibodies. When the anti-creatinine antibodies are incubated in higher

concentration of creatinine, smaller fraction of the free antibodies bound to the chip immobilized

creatinine. When lower amount of antibodies binds to the chip immobilized creatinine molecules,

the molecules are surrounded by more amounts of water molecules. Therefore, the variation of

the amount of non-creatinine bound antibodies binding to the immobilized creatinine varies the

amount of water molecules surrounding the immobilized layer. Due to considerable contrast of

permittivities between water and the antibodies this variation causes a sharp change in the Cmut

contribution of the total IDC capacitance. If no water droplet is used in the electrical

experiments, the immobilized creatinine molecules will be surrounded by air. The permittivity of

the antibodies is of the very close to the permittivity of air when compared to their permittivity

with respect to water. Therefore, from the sensing aspect not considerable permittivity change is

seen if the antibodies bind to the immobilized creatinine or the immobilized creatinine molecules

are surrounded by air. Hence, addition of water molecules ensures intrinsic sensitivity

enhancement. It should be noted as well that immunosensors operate with real world samples in

aqueous solution (serum/plasma) where multiple antibodies or antigens are suspended in the

solution. Therefore, use of water molecule for sensitivity enhancement has been done without

losing the generality of the immunosensor application. The capacitance change of the IDC can be

translated to the concentration of creatinine used in the incubation phase. Thus, an indirect

immunoassay based approach has been used in conjunction with an “all-electrical” sensing

scheme.

3.2.2 Immobilization of creatinine

The immobilization surface chemistry of creatinine on top of the Si3N4 surface of the IDC was

established on silicon test-structures passivated with Si3N4. The surface chemistry was

established in co-operation with the Biotechnology group of University of Potsdam. Various

Chapter 3 Dielectric Immunosensor

immobilization techniques based on creatinine butyric acid and a creatinine-bovine serum

albumin conjugate (crea-BSA) were experimented on 1 cm2 test structures or test chips. The

structure of the creatinine molecule is shown in Fig. 3.5 (a). The preparation of creatinine butyric

acid and a creatinine-bovine serum albumin conjugate (crea-BSA) has been dealt in detail by

Benkert et al. [130]. The chemical structure of crea-BSA is shown in Fig. 3.5 (b).

Same procedure was followed for the synthesis of the above compounds. The best results for the

immobilization of the creatinine molecules are obtained with the adsorption of crea-BSA to the

Si3N4 surface. Crea-BSA (10 mgml-1in standard phosphate buffered saline, PBS) was diluted in

the ratio in the ratio 1:10 with aqua bidset. 2 µl of this solution was pipetted on the Si3N4 surface

of the test structures and was further incubated in a humid chamber for one hour. After six

washing steps of the experimental with PBS and three washing steps with aqua bidset, the chip

surfaces, the chip surfaces were blocked with 2.9 ml 3% BSA in PBS (BSA/PBS) for one hour.

In order to establish a control experimental setup to determine the accurate immobilization of

creatinine molecules, control chips were modified with 2 µl of a BSA solution in the same way

as the experimental chips. For the detection of the immobilized creatinine, 0.1 ml of 3 µg ml-1

anti-creatinine antibody in BSA-PBS solution was added and incubated for one hour with 2.5 ml

peroxidase-conjugated goat anti-mouse IgG (H+L) which was obtained from Dianova (Germany)

and diluted in the ratio 1: 5000 in BSA/PBS. After six washing steps with PBS the chips were

incubated with 2.5 ml of a peroxide standard substrate solution (3,3’,5,5’- tetramethylbenzidine,

H2O2 dissolved in 0.1 M acetate buffer pH 5) for one hour and the developed blue colour was

compared visually. Since only in the case of chips with crea-BSA an antibody binding was

observed, it can be concluded that the binding of the crea-BSA to the Si3N4 surface of the test

Figure 3.5 (a) Chemical structure of creatinine (b) Chemical structure of crea-BSA molecule.

_______________________________________________________________________________________

a.)

b.)

Chapter 3 Dielectric Immunosensor

chip is very strong. Therefore, the same principle was used to immobilize the creatinine

molecules on the Si3N4 surface of the sensor IDC.

A 3 mg sample of creatinine butyric acid was activated by a modified carbodiimde method with

EDC amd sulfo-NHS (molecular ratio 1:3:1) in 0.05 M phosphate buffer (pH 5) for 30 minutes

and coupled to 1 mg of carrier protein (KLH) in 0.1 M carbonate buffer at pH 8.5 for three hours.

After gel filtration with PBS equilibrated PD-10 columns (Pharmacia Biotech, Sweden), the

coupling efficiency was evaluated by determination of KLH-coupled creatinine via the Jaffe

method in relation to the protein concentration. In addition, the decrease in free amino groups of

KLH was controlled by 2,4,6- trinitrobenzenesulfonic acid. Monoclonal antibodies were

obtained by immunization of mice with synthesized creatinine-KLH conjugate and the

hybridoma technique with mouse myeloma cell line SP2.

3.2.3 Sensor circuit design

The oscillator circuit with the embedded IDC sensor described in chapter 2 is the sensor circuit.

In this work of immunosensor, the exclusive variation of capacitance in the sensor circuit is

caused due to the variation of the IDC capacitance brought about by the binding of different

concentrations of anti-creatinine antibodies to the immobilized creatinine molecules on the IDC.

For the developed prototype immunosensor operating at 6 GHz, 1 nH inductors were used. The

capacitance of the IDC is simulated using ADS Momentum and is obtained to be 100 fF. The

inductors are fabricated on the TM2 (topmost metal layer) of the BEOL stack, and have a

simulated quality factor of 15 at 6 GHz aiding the overall quality factor of the LC resonance

tank. The “all-electrical” immunosensor requires a 3.3 V external DC power supply for operation

and as measurement equipment an X band spectrum analyzer from Rhode and Schwarz is used.

3.3 Results and discussion

The capacitance variation of the sensor IDC due to binding of anti-creatinine antibodies to the

immobilized creatinine molecules in an aqueous environment was simulated in COMSOL 4.2a.

The effect of the permittivity of the aqueous solution on the sensitivity of the IDC was simulated

to establish the idea of intrinsic sensitivity enhancement as was proposed in the previous section.

The immobilized creatinine molecules were modeled as circular structures with effective

diameter of 10 nm and permittivity of 1 gm/dl 3-4 [131,132]. The antibodies were modeled as

cylindrical pillars of height 30 nm and effective diameter 10 nm and a relative permittivity of 2

for same sample volume. The size of the antibodies is considerably larger when compared to the

creatinine molecules.

The inclusion of the aqueous environment replicates the electrical experimental scenario where a

droplet of water is added during the measurements. Four simulation environments were

established for four different concentrations of creatinine molecules used in the incubation of the

anti-creatinine antibodies. The simulations then indicate that for an aqueous solution

Chapter 3 Dielectric Immunosensor

environment with sufficiently high permittivity compared to the antibodies and creatinine

molecules, the IDC shows considerable increase in its capacitance with increase in concentration

of creatinine molecules used during the incubation phase. The maximum increase of capacitance

is seen when the aqueous solution has the permittivity of water and the least variation when the

aqueous environment is modeled with the permittivity of air. The results are shown in Fig. 3.6.

This is in line with the proposed theory although the theory appears counter-intuitive initially.

The combined effect of the permittivity of the aqueous environment along with the antibodies

and creatinine molecules has to be taken into account. A higher concentration of creatinine

molecules used for the incubation of anti-creatinine antibodies translates to the fact, more

number of anti-creatinine antibodies bind to the creatinine during the incubation phase (see Fig.

3.4). Thus, the number of free antibodies in the incubated solution reduces with the increase in

the concentration of creatinine used during the incubation phase. On pipetting the samples on the

sensor, different concentrations of antibodies bind to the immobilized creatinine molecules. The

same has been simulated. When few antibodies bind to the immobilized creatinine, it indicates

higher concentration of creatinine molecules used for incubation. In case of fewer binding

antibodies, the immobilized creatinine molecules are surrounded by the molecules of the aqueous

environment. When, the permittivity of the aqueous solution is high, this translates to a higher

capacitance of the IDC when lower number of antibodies bind to the immobilized creatinine. The

same is observed in the simulation. In case of water as the aqueous medium, with permittivity of

70 at 6 GHz, the variation in capacitance of the based on the binding antibodies is quite high

Figure 3.6. Typical variation of IDC capacitance as a function of concentration of creatinine molecules

used in the incubation of antibodies. The capacitance increases with increase in creatinine molecule

concentration used during incubation. The capacitance variation is strong when the surrounding

medium for experiment is water while with air the variation is negligible.

______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

(shown in the black curve of Fig. 3.6). However, when the permittivity of the aqueous medium is

replaced by the permittivity of air that is 1, the variation in capacitance is negligible. This can be

accredited to the permittivity contrast. With air, the permittivity was similar to the permittivity of

the antibodies. Therefore, not much variation in capacitance was seen in case of fewer antibodies

binding and more air molecules around the immobilized creatinine molecules and vice versa.

While in the case of the water molecules, whose permittivity was considerably high when

compared to the antibodies, a high variation of capacitance is seen with the binding of the

antibodies. Therefore, it can be stated that, higher concentration of creatinine molecules used for

incubation of the antibodies results in higher capacitance of the IDC while the experiments are

conducted in a relatively high permittivity aqueous environment. From the CMOS oscillator

sensor circuit outlook, this change of capacitance translates to a decreasing resonant frequency

with increasing concentration of creatinine molecules used for the incubation phase.

3.3.1 Optical measurement of creatinine concentration

Prior to the conduction of the “all-electrical” measurements using the sensor circuits, optical

measurements in a standard ELISA like approach is conducted on test chips, primarily for two

reasons: ELISA-like assay procedures and the optical measurements are established reliable

measurement technique and can be used as an independent standard method to compare the

results with the “all-electrical” method; secondly, in order to find the best immobilization and

binding condition which should be later applied to the sensor chips.

Initially several methods in order to immobilize Si3N4 passivated silicon test chips were

compared regarding the anti-creatinine antibodies binding. Hence, creatinine butyric acid and

crea-BSA were covalently or non-covalently immobilized onto different modified test chips. 2 µl

of the immobilization solution was pipetted to the Si3N4 surface, which was the same amount

needed to cover the IDC sensor area. One such test chip is shown in Fig. 3.7.

Figure 3.7. Chip photograph of Si3N4/Si test chip for optical measurement. Chip size is 1 cm x 1 cm.

______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

The immobilized creatinine was then detected with the mono-clonal anti-creatinine antibody

B90-AH5 and a peroxidase-conjugated anti-mouse IgG. The peroxidase activity was detected via

a standard color reaction using 3,3’,5,5’-tetra-methylbenzidine and hydrogen peroxide. Most of

the chips generated no color. Only chips with adsorptive immobilized crea-BSA generated the

typical blue color, whereas, the chips with immobilized BSA generated no color.

Figure 3.8. Indirect competitive assay principle for optical creatinine determination with creatinine-

modified Si3N4 test chips

Figure 3.9. Optical measurement of creatinine concentration. The response slope of the optical

measurement in the range 0.88 to 88 µM shows the dynamic range of the standard measurement

technique.

______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

This result shows that the anti-creatinine antibody was specifically bound to the immobilized

crea-BSA. For creatinine determination an indirect competitive immunoassay principle as shown

in Fig. 3.8 was applied.

The crea-BSA modified Si

test chips were incubated with different creatinine

concentration (0-8.8 mM) with a defined and optimized antibody concentration (0.1 µg

-1

). As mentioned above, the optical measurements were conducted to compare the

standard technique with our proposed “all-electrical” approach. The detection range of the

established optical technique although covers the concentration ranges of creatinine

which is of clinical relevance, but is saturated beyond 88 µM of creatinine concentration.

On comparing the detection range of both measurement techniques, the electrical

approach shows an order of magnitude increase in detection range, as is demonstrated in

the subsequent section.

3.3.2 Dielectric measurement of creatinine concentration

Characterization of sensor

The sensor chip showing the IDC sensor in conjunction with oscillator circuit is shown in

Fig. 3.10.

470 µm

Sensor Area

Figure 3.10. Chip photograph of dielectric sensor. The sensor area (IDC) is marked in red.

______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

The chip was characterised initially to analyse the electrical performance and was

followed by a calibration step with the measurement of glucose solution (see chapter 4 for

glucose solution measurement). The sensor chip draws a current of 27 mA from 3.3 V DC

power supply. The resonance frequency of the sensor oscillator is 6.01 GHz in air that is

with no material placed on top of the sensor. The overall chip area is 0.3 mm

. The

miniaturized sensor area as mentioned above reduces the volume of the probe sample

used in the analysis of the creatinine.

The performance of the sensor oscillator showing resonance frequency shift for varying

IDC capacitance was characterized with glucose solution measurements. Fig. 3.12 shows

the measurement of various concentrations of glucose solutions using the sensor

oscillator. Different concentrations of glucose solutions are pipetted on the sensing area

and the resulting frequency is measured and compared with simulations.

The resultant permittivity of the glucose solution for different concentration of glucose is

calculated analytically using the mixture rules (see chapter 4). It is observed that the resonance

frequency of the sensor oscillator up-shifts with increasing concentration of glucose in the

homogeneous solution of glucose in water. This is attributed to the decrease in resultant

permittivity of the solution with the increasing concentration of glucose as glucose has lower

permittivity as compared to water. With lower glucose concentration the permittivity of the

resultant solution is close to water and the same can be seen in Fig. 3.11. The resonance

Figure 3.11. Calibration of sensor circuit with glucose solution. The red curve shows the simulation and

the black triangles are the measurement results. The resonant frequency up-shifts with increasing

glucose concentration.

______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

frequency of the sensor oscillator is with pure water on top of the IDC is 5.81 GHz. This

measured behavior of the sensor is in close agreement with the theoretical and simulated

behavior of the sensor. Thus, the calibration technique well establishes the operation of the

sensor. The sensor is now extended to the direct measurement of the creatinine.

Creatinine concentration measurement

Several chips from the same wafer need to be characterized in order to determine the

reproducibility and the yield of the fabricated direct immunosensor. Multiple chips on the same

wafer were characterized to estimate the on wafer process variation. The variation observed in

the resonance frequency of the sensor oscillator is not more than 3 MHz. This depicts the high

yield and reproducibility of the sensor chip. The creatinine molecules were immobilized on the

sensor area only, using the surface chemistry on top of the sensor area. The surface chemistry

was used only to modify the sensor area, leaving the inductors. Therefore, the creatinine

molecules did not detune the inductors.

Eight chips with the same resonance frequency of 6.01 GHz were chosen and treated later in

order to immobilize creatinine molecules on the IDC sensor surface. This is followed by the

pipetting step of 2 µl of the pre-treated anti-creatinine antibody samples on the Si3N4 based

surface of the IDC sensor area marked in red in Fig. 3.10. It was mentioned above for the

purpose of electrical measurement with strong permittivity contrast and a better sensor response

a drop of water (1 µl) was added on the sensor area. Fig. 3.12 shows the resonance frequency

shift of two chips treated with two different incubated antibody samples.

Figure 3.12. (a) Resonant frequency peak for chip treated with antibodies incubated with 0.88 µM

creatinine. (b) Resonant frequency peak for chip treated with antibodies incubated with 88 µM

creatinine.

______________________________________________________________________________________

a.)

b.)

Chapter 3 Dielectric Immunosensor

The measurement results of the chips with four different samples of antibody are shown

in Fig. 3.13. The higher the concentration of creatinine in the pre-incubation the lower is

the concentration of antibody binding to the immobilized creatinine molecules on the

IDC. The black curve shows the measurements done with an additional droplet of water

of approximate volume of 1 µl carefully pipetted on the sensing area.

It is seen, with lower concentration of creatinine used in pre-treatment (higher amount of

antibodies binding to the immobilized creatinine), higher is the resonant frequency. This

is in line with the proposed theory and simulation results shown above. As seen from the

measurements, for each step variation of the concentration of the creatinine, the resonant

frequency varies by approximately 35 MHz. With highest creatinine concentration during

the incubation phase, least amount of antibodies binds to the immobilized creatinine

molecules on the IDC. Therefore, the resulting resonant frequency tends towards the

frequency of the oscillator with pure water on top of it 5.73 GHz). The red curve shows

the same experiments done on the same chips with air as the surrounding medium.

As seen from the results there is negligible variation of the resonant frequency of the

oscillator although there is a tendency of frequency downshift. The measurement results

agree closely with our proposed model and simulation. The capacitance variation of the

IDC is strongly dependent on the permittivity contrast between the antibodies and the

Figure 3.13. Measured variation of resonant frequency as a function of creatinine concentration used

in the incubation of antibodies. The black curve shows the resonant frequency for four samples with

increasing concentration of creatinine used in incubation while the experiment was done in aqueous

(water) environment. The resonant frequency downshifts with increasing creatinine concentration. The

red curve shows the same experiment done with air as the surrounding medium.

_______________________________________________________________________________________

Chapter 3 Dielectric Immunosensor

surrounding medium and hence, suited for biosensor, immunosensor applications as most

often a buffer solution is used for such measurements.

It can also be noted from the above measurements that in aqueous solution, the sensor has a

detection range higher than the optical measurement technique. It was shown in Fig.3.10 that the

measurement response with optical technique saturates beyond 88 µM of creatinine

concentration used in the incubation phase. However, in the proposed sensor, the curve although

seems to have a saturating effect, but have considerable sensitivity from 88 µM to 880 µM.

Therefore, the sensor is sensitive beyond the clinically relevant range of 0.88 µM – 88 µM as

well. The increase in the sensitivity range, makes the sensor useful for other clinical diagnostics

as well. The variation of frequency in this range is 25 MHz and is considerably higher than the

process variation of 3 MHz, therefore, showing an order of magnitude higher detection range

compared to the established optical technique. This can be attributed to the contrast of

permittivity between water molecules and the anti-creatinine antibodies. When comparing the

sensitivity of the two approaches, the percentage change of frequency per 10-fold increase in

concentration of creatinine with respect to the total frequency shift over the entire detection

range (~42%) is comparable to the percentage change in E450nm intensity (~ 40%).

In order to determine the error bar in the measurement and also to determine the

reproducibility of the sensor system from the measurement perspective, several sets of

chips with same samples of pre-incubated antibodies were measured at the same time in

aqueous solution. The maximum normalized standard deviation in the resonant frequency

for similar measurement condition is 0.223. The frequency response of two sets of chips

showing maximum measurement variations was plotted. The resonant frequency contrast

for the two sets of chips was observed for all the four antibody samples incubated with

different amounts of creatinine. The maximum drift in the resonant frequency of two

chips with same sample of antibodies was observed to be 4 MHz, shown in Fig.3.14. This

drift in the resonant frequency is approximately a tenth of the measured sensitivity of the

sensor. The observed variation being considerably less than the sensitivity shows high

reproducibility of the immunosensor system.

Chapter 3 Dielectric Immunosensor

3.4 Conclusion

The results show that creatinine can be measured with the developed CMOS high

frequency dielectric immunosensor in the clinically relevant concentration range. The

electrical measurement results show close agreement with an optical standard

measurement technique. The measurement capability in the order of nanomolar

concentration level shows that such a high frequency sensor can be successfully used for

relevant measurements in clinical diagnostics. The measured frequency shift of 35 MHz

in the clinically relevant regime of creatinine concentration of 0.88 µM to 88 µM is much

higher than the effect of process variation. The effect of process variation was measured

and was shown negligible in comparison to the sensitivity of the sensor. This was shown

in the error bar measurement conducted with two sets of chips. Therefore, it can be

deduced that the demonstrated CMOS high frequency sensor has considerable stability for

clinical measurements. The miniaturized sensor design and the exclusion of any labelling

compounds will reduce the costs in comparison to other antibody-based creatinine assays

enormously. Additionally, the capability of immobilization of creatinine molecules on

standard passivation layer of CMOS process (Si

) evades the need of any post

processing techniques of the silicon chip required for immobilization of creatinine

molecules. This result is very significant for future CMOS immunosensors for creatinine

and can be adapted to other antigen antibody couples

Since the already published creatinine enzyme immunoassays and indirect

immunosensors can specifically measure creatinine in real human serum samples it can be

assumed that the developed immunosensor which uses the same antibody is also able to

measure real samples Since nanomolar analyte concentrations can be determined it can be

stated that in future the developed technology can be adapted for other clinically relevant

analytes as well.

Figure 3.14. Error bar measurement for two sets of chips. The maximum frequency drift does between

two chips does not exceed 4 MHz.

_______________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

DETECTION of ANALYTE

CONCENTRATION

This section* of the thesis presents a high-frequency (X-band) CMOS dielectric sensors applied

to various biological applications; primarily to detect concentration of suspended particles in a

solution, to detect concentration of glucose in homogeneous glucose solutions, dielectric imaging

of biomaterials and detection of fat and calcium in blood. Three sets of sensor chips in the

frequency range of 6 GHz to 12 GHz are demonstrated in this chapter. The sensor chips are

fabricated in 0.25 µm or 0.13 µm SiGe:C BiCMOS technology of IHP. Two approaches for fluid

handling are shown in this chapter. The first approach includes using of a typical polymer based

microfluidic system and the second approach includes creation of an insulating wall around the

sensor area of the chip. The post-processing steps required for the microfluidic integration is

explained in this chapter. In that context, the influence of silicon nitride and silicon dioxide

passivation layers on the sensor chip is analyzed. The dielectric sensitivity of the chips is

characterized and calibrated using different organic fluids (alcohols); sensitivity of the sensor

chips were found to be a strong function of the passivation layers on the top of the sensors. The

sensors are further applied to detect fat and calcium present in blood samples, as a first step to

develop minimally invasive technique for plaque characterization in arteries. Further, a prototype

of a typical sensor array system is demonstrated for imaging of biomaterials and can have

potential applications in analyzing cancerous tissues form healthy ones.

4.1 Introduction

The key features that a state-of-the-art biosensor should possess are the capability of in situ and

label free detection of molecules and biological cells within extremely small volumes of the

probe sample with a reduced measurement time. Chapter 3 dealt with one such special biosensor

(immunosensor), the functioning of which required extremely small sample volumes of antigens

and antibodies. On similar lines, there is an increasing demand for developing miniaturized fast

non-invasive systems for detection of molecular concentrations, concentration of cells and

particles in biological suspension [133, 134] which will use significantly less amount of probe

samples as compared to the established sensing systems. Applications like detection of oral

squamous cell carcinoma [135], determining the sickle red blood cells in serum [136],

discrimination of leukemia cells (HL-60) [137], require rapid label free sensing techniques.

*Parts of this chapter have been published as “Integrated high-frequency sensors in catheters for minimally invasive

plaque characterization”, European Microelectronics and Packaging Conference and Exhibition, September 2015,

Friedrichshafen, Germany

“12 GHz CMOS MEMS lab-on-chip system for detection of concentration of suspended particles in bio-suspensions”,

Biodevices, January 2015, Lisbon, Portugal

“An 8 GHz CMOS near field bio-sensor array for imaging spatial permittivity distribution”, IEEE-MTT-S International

Microwave Symposium, May 2014, Tampa, USA DOI: 10.1109/MWSYM.2014.6848459

Chapter 4 Detection of Analyte Concentration

As shown in Fig. 4.1 a typical LOC device should have microfluidic sensing system with

integrated autonomous control and detection circuits on the same measurement and sensing

platform. The medical industry has unprecedented possibilities to extract advantages out of the

developing LOC technology, as it rightly corresponds to the perfect size to reach biomolecules

and cells properties.

The development of such LOC device is analogous to the miniaturization of the age old

computing systems of 1960s to the hand held smart phones with higher computing powers. The

dream of establishing LOC system is to miniaturize the functionalities of the state-of-the-art

biotechnology laboratories to a hand held autonomous compact device.

In this chapter, we report a complete CMOS/microfluidic system for dielectric detection of

suspended particles in biological suspensions in the frequency range of 6 GHz to 12 GHz.

Hybrid integration of the microfluidic system to the CMOS chip is performed as a post process

step after the chip fabrication. Simultaneous electrical and optical measurements of suspended

particles in a solution depict close correlation of both the measurement. The X-band sensor

described in this work aids in avoiding low frequency dispersion mechanisms described

previously. As mentioned previously low-frequency dispersion mechanisms are useful to

determine other properties of the particles but can be problematic for detection of concentration

of particles.

Figure 4.1 Typical schematic of a lab on a chip system showing the microfluidic and detection

system. The same platform houses the microcontroller and the detection circuits (IHP internal).

______________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

4.2 Sensor parameters for microfluidic integration

As mentioned before, the sensor architecture (permittivity controlled oscillator) is the same for

all the application and also the sensor module, IDC. The sensing principle is based on the

variation of fringing field capacitance between the fingers of the IDC, due to the change of

permittivity on top of it. The operation of such a sensor based on a unit cell model has been

explained in chapter 2. In this section of this chapter a more detailed analysis of the sensor for

the microfluidic integration is taken into account: which is the effect of passivation layer on the

sensitivity of the sensor.

The sensor systems are fabricated in the standard BiCMOS process lines of IHP (0.13 µm for 6

GHz sensor system and 0.25 µm for 8 GHz and 12 GHz sensor system). As was explained in

chapter 2 the choice of metal layer stems from the biological integration need. One major

implication of the choice of the metal layer for designing the IDC, along with the influence on its

quality factor is the influence of passivation layer. Passivation layer in micro-fabrication or

CMOS technology is an insulation layer deposited on top of the metal layers (electrodes) to

protect the metal layers from external environment. The passivation layer for a standard

BiCMOS process is 400 nm of Si3N4. The influence of passivation layer is a significant

parameter to be analyzed due to the two different approaches considered here for characterizing

bio-suspensions. The first approach involves using a conventional polymer based microfluidic

system suitable for analysis of fluids in a flow assisted fluid system. Chemical mechanical

polishing is employed in order to planarize the Si3N4 surface for better bonding of the polymer

microfluidic system to the silicon chip. Such a planarization results in the reduced thickness of

the passivation layer. The second approach includes fabricating a non-conductive wall around

the sensor structure for biological suspension. Such an approach is suitable for analysis of fluids

in static condition. In this approach, the thickness of the passivation layer is not reduced. Fig. 4.3

shows the variation of the capacitance of the IDC with increasing permittivity of MUT for varied

thickness of the passivation layer. The variation of the capacitance with increasing permittivity

given by the slopes of the curves in Fig. 4.2 is seen to reduce with the increasing thickness of the

passivation layer. The slope of the curve defines the sensitivity of IDC alone. Such a result is

understandable, as the strength of the fringing electric field tends to decay exponentially along

the z direction. This has been discussed in detail in the analysis of the penetration depth of

electric fields of IDC, in chapter 3. If the IDC is designed on the TM1 metallization layer of the

BiCMOS stack, the passivation layer has a thickness of approximately 5 µm. This is the

combined height of SiO2 on TM1 and Si3N4 on TM2.

Chapter 4 Detection of Analyte Concentration

This is the case for the sensor system operating at 6 GHz where the IDC is designed on the TM1

metallization layer. For this sensor system we use the non-conductive wall approach for fluid

characterization in order to analyze the limits of the sensitivity of the sensor system in terms of

passivation layer thickness. The effect of passivation is of wide interest for near field bio

sensing, as often the electrodes are passivated to prevent them from coming in direct contact with

the bio materials. It is shown with simulations that the sensor is sensitive up to a passivation

thickness of 5 µm and thus can be efficiently used in near field bio sensing applications. The

same has been shown with the measurements using the 5 GHz sensor system.

4.2.1 Planarization of silicon chip

As mentioned above, in order to analyze biological suspensions using the CMOS sensor chip,

two approaches were considered. In the first approach, where a polymer based microfluidic

system is used to handle the bio-suspension, planarization of the silicon chip is required for

precise binding between the polymer microfluidic channels and the silicon chip. This requires

post processing of the chip after fabrication. Chemical mechanical polishing (CMP) step is

employed in order to planarize the Si3N4 passivation surface [138].

The chip fabrication involves the production process of the silicon wafer consisting of the sensor

circuit in a standard BiCMOS process (0.25 µm for 12 GHz sensor chip). As shown in Fig. 4.3

Figure 4.2 Variation of capacitance of the IDC with respect to permittivity of MUT for different

thicknesses of passivation layer on top of the IDC. The variation of capacitance reduces with increasing

thickness of passivation layer.

_________________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

(a) the BEOL of the 0.25 µm BiCMOS process has five metal layers aluminum/tungsten

metallization with silicon dioxide as the interlayer dielectric.

As was mentioned in the passivation analysis in the previous section, the back-end-of-line stack

is topped with Si3N4 passivation surface. The thickness of the TM2 layer is 3 µm as is shown in

Fig. 4.3 (a). Therefore, the topography of the finished sensor chip is at least 3 µm high. There are

stringent requirements for the planarity of the surface of the (Polydimethylsiloxane) PDMS and

the silicon chip for bonding purposes. Therefore, the fabricated sensor chip is not suitable for the

bonding process with an irregularity in the topography of the order of µm. The processing of the

chip is modified for the planarization of the chip surface. The gaps between the TM2 structures

are filled with silicon dioxide and planarized using the CMP technique. High density plasma

(HDP) chemical vapor deposition (CVD) was used to deposit this silicon dioxide. HDP is used

because of its good gap filling properties. The CMP process was stopped several hundred

nanometers above the TM2 layer. Then the oxide was etched back using reactive ion etching

without a resist mask until the TM2 surface was exposed. Fig. 4.3 (b) shows the BEOL stack

after the CMP process. The top Si3N4 layer is completely planarized and TM2 metallization layer

is exposed to the external environment. The gap between the TM2 metal structures is filled with

SiO2 as mentioned above. A scanning electron microscopy image of a planarized chip is shown

in Fig. 4.3 (c).

Figure 4.3 Planarization of the BiCMOS stack for microfluidic integration. (a) Schematic of the back-

end-of-line stack. (b) Schematic of the stack followed by planarization (c) SEM image of a typical

planarized chip.

a.)

Polishing

________________________________________________________________________________________

CMP

b.)

c.)

Chapter 4 Detection of Analyte Concentration

4.2.2 Microfluidic integration to silicon Chip

PDMS microfluidic channels of width 500 µm and height 50 µm were fabricated using SU8

master mold using soft lithography technique. This technique of polymer based microfluidic

system was first introduced by the group of Whitesides of the department of Chemistry in

Harvard University [139]. The same procedure is followed here. The PDMS channel was further

bonded to the CMOS chip using oxygen plasma bonding technique. Fig. 4.4 shows the schematic

view of the CMOS microfluidic system.

The master mold was fabricated from SU8 photoresist patterned on a 4 inch silicon wafer. SU8 is

most commonly used for such fabrication techniques because of the capability of manufacturing

high aspect ratio structures with it. PDMS was prepared using Sylgard 184 silicone elastomer

base (Monomer) and its curing agent (hardener). The monomer and the hardener were mixed in

the ratio 10:1. Other ratios of monomer to hardener were also tried for different values of

elasticity of the PDMS structures. However, the above combination of monomer and hardener

was chosen as it gave the best bonding strength. After thorough mixing, the solution was poured

on the master mould and cured at a temperature of 70°C for ninety minutes. Room temperature

curing is also possible but takes a longer time of approximately a day. The obtained PDMS

structure was carefully peeled off from the mold and stored in a salinized chamber.

Oxygen plasma bonding of the PDMS microfluidic channel to the CMOS chip was performed in

the Reactive Ion Etching (RIE) chamber. Plasma pressure of 16 Pa was used for a time of 30

seconds with an RF power of 65 Watt.

Figure 4.4 Schematic of the microfluidic integration with the CMOS sensor chip.

______________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

Using higher RF power reduces the bonding strength as the PDMS surface which is changed

from hydrophobic to hydrophilic due to the plasma action, to enable the bonding process, is

transformed back to hydrophobic with higher RF power. Careful alignment of the channel on top

of the sensor was the limiting factor of the bonding time. The bonding time was kept within one

minute in order to keep the PDMS in the activated state. Fig. 4.5 shows the process steps of the

PDMS/CMOS hybrid microfluidic system.

4.3 Results and discussion

4.3.1 Calibration of sensors

Organic alcohols and glucose solutions have been used to calibrate the sensors. In the frequency

range of 6 GHz to 15 GHz the dispersion of water can be used to calibrate the sensor. In this

frequency range a small change in the concentration of glucose in the solution can change the

permittivity considerably of the solution. This is due to the dispersion slope of water. The

Figure 4.5 Fabrication and bonding of PDMS microfluidic channel with the CMOS sensor chip. The

PDMS microfluidic system is fabricated using soft lithography approach. Oxygen plasma bonding is

used to bond the PDMS microfluidic channel to the silicon chip.

______________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

frequency dependent dielectric constant of pure water is characterized by a single Debye

relaxation mechanism with the relaxation frequency approximately at 17 GHz shown in eq. 4.2.

The variation of the dielectric constant of water with respect to frequency is shown in Fig. 4.6

[140]. The static dielectric constant of water as seen from Fig. 4.6 is around 78 and the

permittivity at infinite frequency is 4.

The dielectric permittivity of water due to the single Debye mechanism is seen to reduce with

frequency. The dispersion mechanism is given by the equation,

𝜀𝑓= 𝜀∞+𝜀𝑠−𝜀∞

1+( 𝑓

𝑓𝑐)2 (4.2)

Ɛf is the permittivity at the operating frequency f. Ɛs and Ɛ∞ are the static and infinite frequency

permittivity respectively. The characteristic frequency of the relaxation mechanism is given by

fc. At 6 GHz the permittivity is 70 and at 12 GHz the permittivity is 60. The permittivity of water

at the above frequencies is considerably higher when compared to glucose. With the increase of

water content in the glucose solution, the overall permittivity of the solution increases.

Therefore, the concentration of glucose in a suspension can be characterized based on the

variation of the average dielectric concentration of the suspension depending on the glucose

concentration.

As mentioned in the previous section as one of the approaches for fluid handling, a non-

conductive dielectric wall was fabricated around the sensor to analyze the glucose suspensions.

Fig. 4.7 shows a typical sensor chip mounted on board with dielectric wall for fluid handling.

Figure 4.6 Permittivity of water with respect to frequency. The static permittivity of water is 78

while the infinite frequency permittivity is 4. The characteristic frequency of the Debye relaxation

process is 18 GHz [140].

______________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

The sensor chip shown in Fig. 4.8(a) has four sensors with a switched architecture. At a given

time only one sensor is operated by turning on the switch for the respective sensor. The glucose

solution is pipetted on top of the enclosed sensors as shown in Fig. 4.8. The sensor circuit is

similar to the one described in chapter 2.

The 6 GHz glucose sensor has the IDC designed on the TM1 metallization layer of the BEOL

stack. As mentioned in the previous section, this design was adopted to characterize the

feasibility and the detection limit of the sensor system in terms of its sensitivity based on the

thickness of the passivation layer. A systematic way to characterize the sensor was to first

analyze the dielectric permittivity of solutions of known permittivity. The calibration of the

sensor is done with organic fluids of known permittivity. The resonance frequency of the sensor

oscillator is 6.02 GHz with no material placed on top of the sensor, however, with dielectric wall

built around the sensor. The liquids used for calibration of the sensor were PMMA (Ɛ = 2.67),

PDMS (Ɛ = 2.63), Ethanol (Ɛ = 3.2), methanol (Ɛ~12). Fluids with such permittivity values were

chosen in order to determine the resolution of the sensor.

Fig. 4.8 shows the variation of the resonant frequency of the oscillator during the calibration

phase. It is noted that the resonant frequency downshifts with the increasing permittivity of the

fluids and the same is expected from the sensor performance. This is due to the increase in the

Figure 4.7 Sensor with non-conductive wall around the senor for fluid handling. (a) Typical sensor

chip (b) Sensor mounted on board with non-conductive wall.

_________________________________________________________________________________________

a.)

b.)

Chapter 4 Detection of Analyte Concentration

capacitance of the IDC. It is also noted that the PDMS and PMMA having similar permittivity

values at 6 GHz, varying only in the second decimal place and therefore, show identical

frequency shifts.

The sensor shows a sensitivity of 20 MHz/permittivity in the permittivity range of 1 to 20, as

extracted from the measurements. For error bar measurements, several chips are measured and a

variation of 2 MHz to 3 MHz in the normalized (with air on top) resonance frequency is

observed. Therefore, combining the sensitivity and the error bar measurements it can be

concluded that the sensor system shows a detection resolution of 0.5 in the permittivity values.

The same was seen in the PMMA and the PDMS measurement as mentioned above. The sensor

was then used to characterize the concentration glucose in a homogenous solution. Fig. 4.9

shows the variation of the resonance frequency of the sensor oscillator with the concentration of

water in glucose solution. For a 90% saturated glucose solution (10% water), the permittivity as

obtained from the literature is approximately 10. Water, which has a considerably higher

permittivity as shown in the previous section, increases the permittivity of the solution when

added to the saturated glucose. The same trend is noted in the measurement, as seen in Fig. 4.9.

As the water content in glucose solution was increased the resonance frequency of the sensor

was reduced or in other words, the resonance frequency of the oscillator has a direct

proportionality to the glucose concentration as increasing glucose concentration reduces the

permittivity of the solution. The sensors show a sensitivity of 15 MHz downshift per 10%

increase in concentration of water in the glucose solution. In Fig 4.9, the 0% concentration of

water indicates super saturated glucose solution. All the other measurements are normalized to

Figure 4.8 Variation of the resonance frequency of the oscillator with materials of different

permittivities. The oscillating frequency downshifts with increasing permittivity.

________________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

this super saturated glucose solution. The increase in concentration of water is done by

incrementally adding fixed volume of water in the saturated glucose solution.

From the above measurements, it is shown that the IDC although fabricated on TM1 metal layer

of the BEOL stack, with 5 µm of passivation of combined Si3N4 and SiO2 dielectrics, is sensitive

to the change of permittivity on top of it. The fabrication of the sensor in TM1 metal layer was

needed for immunosensor application as was mentioned in chapter 3. The results of glucose

solution calibration of the 5 GHz sensor are presented here to show the difference in the

sensitivity for the two metal layers of the sensor, TM1, TM2. The sensitivity and the resolution

can be dramatically enhanced with the removal of the passivation layer and shifting the IDC on

the TM2 metal layer, thus, bringing the sensor close to the analyte. The same has been done

using the 12 GHz sensor.

The 12 GHz sensor is used in conjunction with a microfluidic system for the characterization of

the biological suspensions. As was mentioned before, the passivation layer is planarized in order

to obtain a sufficiently precise bonding between the chip and the microfluidic system. The IDC is

designed in the TM2 metallization layer and hence, has close exposure to the analyte. The

CMOS chips were characterized electrically prior to microfluidic experiments. The current

drawn by the chip was 12 mA at an operating voltage of 3 V. The oscillating frequency was

measured to be 12.32 GHz with an output power of -5 dBm. Further characterization of the chip

was performed after plasma bonding of the PDMS microfluidic channel with the chip. The DC

operating values of the chip remained unaltered, while the oscillating frequency was measured to

Figure 4.9 Variation of oscillating frequency of the sensor with increasing concentration of water in

glucose solution.

______________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

be 12.20 GHz. The 100 MHz shift of the oscillating frequency was accounted for the influence of

the PDMS on the inductor coils used in the design of the oscillator. This resonant frequency

served as the reference for further measurements, as the microfluidic channel was empty.

The variation of oscillating frequency of the dielectric sensor with materials of different

permittivities was characterized by using organic fluids in the microfluidic system. A downshift

of oscillating frequency was observed with increasing permittivity of the organic fluids, in this

case alcohols. Fig. 4.10 shows the variation of the resonant frequency for different alcohols. At

12 GHz isopropanol and ethanol have almost the same permittivity (Ɛ= 3.8~4.2), as shown by

Belrhiti et al [140] and can be seen in the frequency output plot to be close to each other. When

compared to the 5 GHz sensor, already a better selectivity is noticed. It is also noted that

although the static permittivity of methanol is higher than the static permittivity of acetone, at 12

GHz, the permittivity of methanol is less than the permittivity of acetone described by Kung et al

[141] and the corresponding shift of resonant frequency shows the same.

Sensitivity of 100 MHz/permittivity was observed with the measurements performed with the

organic alcohols. This sensitivity is considerably higher when compared to the 5 GHz sensor

oscillator. In order to estimate the measurement reproducibility microfluidic channels were

bonded to five different sensor chips from the same wafer. Maximum frequency variation of 2

MHz was observed for same measurements and was negligible compared to the sensitivity of the

sensor. The detection limit of the sensor can also be estimated with the measurement of

isopropanol and ethanol. The alcohols have a permittivity difference of 0.7 at 12 GHz and still

Figure 4.10 Variation of the oscillating frequency for different organic alcohols. The resonance

frequency downshifts with increasing permittivity and has a sensitivity of 100 MHz/permittivity.

______________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

show a considerable frequency shift as shown in Fig. 4.10. The CMOS/microfluidic system was

then used to study the effect of water in a homogeneous glucose solution. The variation of

resonant frequency with different concentration of water depicts the variation of permittivity of

the glucose solution with water content. Fig 4.11 shows the downshift of resonant frequency of

the oscillator with increasing water content. Saturated glucose solution has a permittivity of 8 at

12 GHz given by Meriakri et al [142]. The corresponding oscillating frequency is measured to be

11.52 GHz. This is close to the value measured for methanol (Ɛ= 9.2) during calibration. The

obtained results can be extended to determine permittivity of the glucose solution with different

concentration of water. Every 10% increase in the water content shows a frequency down-shift

of 250 MHz, which indicates a permittivity increase of approximately 2.5. The increase in water

concentration is obtained by incremental increase of water volume in the super saturated glucose

solution

Comparing the sensitivity of the two sensors, it is intuitive that the 12 GHz sensor has a higher

sensitivity when compared to the sensor operating at 6 GHz due to the closer proximity of the

sensor to the analyte. This is also observed while determination of the change in permittivity of

the glucose solution with varying concentration of glucose. The glucose solution when measured

using the 12 GHz sensor, shows an increase in permittivity of 2.5/10% increase in water

concentration. This is significantly higher when compared to the measurements performed using

the 6 GHz sensor oscillator. The actual change of the permittivity in the solution due to

concentration of glucose was hidden by the permittivities of SiO2 and Si3N4.

Figure 4.11 Variation of oscillating frequency of the sensor with increasing concentration of water in

glucose solution.

_________________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

4.3.2 Particle concentration measurement

The 12 GHz sensor in conjunction with microfluidic system was utilized to determine the

concentration of suspended microbeads in acetone. The influence of micro-beads or particles in a

solution can be understood by the hindrance of molecular motion given by the Stokes-Einstein

Debye equation [143]. The presence of the particles in an aqueous solution (for e.g., acetone in

this case) influences the Debye relaxation process observed in aqueous solutions as described

previously for water, thus impacting the overall permittivity of the solution. This can be utilized

as a sensing parameter in order to determine the concentration of particles in a solution. Such a

technique is lucrative for in situ, label-free molecular detection in extremely small sample

volumes as well as cytometric applications. A typical biological cell suspension shows the

dispersion mechanisms as shown in Fig. 4.12.

The α and β dispersion mechanisms are low-frequency phenomena. As mentioned previously,

the low-frequency dispersion mechanisms can be utilized to understand other parameters of the

particles or biological molecules.

At high frequency regime (GHz range) γ dispersion dominates and as described above, can be

utilized to determine the concentration of particles in a solution. From the mathematical relation

it can be understood that the characteristic Debye relaxation time increases or the characteristic

frequency reduces with increasing concentration of particles in a suspension [143].

Figure 4.12 Dielectric dispersion curve for biological cell suspension.

______________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

𝜏 = 𝜋𝜂𝑟𝑙2

𝑘𝐵𝑇 (4.4)

r is the radius of the particle, l is the hopping length,  is the viscosity of the particle kB is the

Boltzmann’s constant and T is the temperature. Viscosity is directly proportional to the number

of particles.

The effective permittivity of the solution with particles can be approximated with Maxwell-

Garnett equation [144] or Bruggeman’s approach [145]. Using the Maxwell-Garnett equation the

effective permittivity of the solution of acetone and microbeads in this work can be expressed as,

𝜀𝑒𝑓𝑓 = 𝜀𝑎(1 + 3𝜎(𝜀𝑚𝑏−𝜀𝑎)

(𝜀𝑚𝑏+2𝜀𝑎)) (4.4)

Ɛeff is the effective permittivity of the solution and the permittivity of acetone and the microbeads

are given by Ɛa and Ɛmb respectively. σ is the volume fraction of the microbeads given by

𝜎 = 𝑁𝑉/𝑉𝑡𝑜𝑡 (4.5)

N is the number of microbeads with volume V and Vtot is the total volume of the solution. In our

measurement system we used micro-beads of diameter 10 µm in different concentrations in a

fixed volume of acetone 2 ml.

The beads were thoroughly mixed in order to prepare a homogeneous solution. The variation of

the resonance frequency of the sensor oscillator with respect to the concentration of microbeads

in acetone is shown in Fig. 4.13. Equation (3) shows that effective permittivity of the solution

Figure 4.13 Variation of oscillating frequency of the sensor with varying concentration of

microbeads in acetone.

________________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

has a direct proportionality with concentration of particles in the solution. However, the second

term of the equation (Ɛmb – Ɛa) renders a negative term when the permittivity of acetone is higher

than the permittivity of the particles. Therefore, the overall term has a value less than 1 and the

value reduces with increasing concentration. Thus, the overall permittivity of the solution

decreases with increasing concentration of particles. Therefore, as observed in Fig. 4.14, the

resonance frequency of the oscillator increases with increasing concentration of particles due to

reduced capacitance. The diameter of the beads are 10 µm with average permittivity of 2.

Frequency up-shit of 125 MHz/10 µl increase in bead content in acetone is measured.

4.3.3 Fat and Calcium characterization in blood

Atherosclerosis is the intravascular condition where the artery walls are hardened due to

deposition of calcified fat on its walls [146, 147]. Such hardening of artery walls leads to

extreme medical conditions and is often rated as one of the primary reasons for death due to

heart failure, stroke, etc [148]. Compact CMOS compatible sensors can serve as an interesting

alternative to the state of the art detection techniques of calcified fat which include intra-

vascular-ultrasound-imaging (IVUS), optical-coherence-tomography (OCT) of arteries etc [148,

149]. Although, IVUS is an established technique using high frequency acoustic signals, the

image quality and the axial resolution is still a concern. On the other hand, OCT systems

overcome the shortcomings of IVUS, at the expense of high packaging cost. In this regard, high-

frequency or microwave sensors compatible to CMOS technology are a possible solution. CMOS

high-frequency sensors offer compact low-cost miniaturized solution for efficient imaging of

intravascular modalities with axial resolution in the order of µm. However, development of such

sensors for applications in the area of medical diagnostics like plaque characterization is still on

the horizon.

In order to establish such high-frequency sensors for in-situ applications, initial ex-vivo

validations of such sensors with same materials (blood, fat and calcium) are mandatory, in order

to establish the feasibility of such sensors in the real environment. In this part of the chapter, we

present a CMOS sensor operating at 12.6 GHz, used to discriminate pure blood samples form

blood samples infested with fat and calcium in the liquid phase.

The sensor chip draws a current of 12.5 mA from a 2.5 V power supply. The normalized

resonant frequency of the oscillator with no material on top of the IDC is 12.6 GHz with output

power of -5 dBm. The measurements were conducted in two phases. A calibration step is

performed with organic alcohols to estimate the sensitivity and the selectivity of the sensor. The

second phase of the measurements is performed with the blood samples. The functionality of the

sensor is characterized using organic alcohols. The resonant frequency of the oscillator scales

down with increasing permittivity of the organic alcohols as shown in Fig. 4.14. Selectivity of

the order of 0.5 in absolute permittivity value is demonstrated with detection of isopropanol and

Chapter 4 Detection of Analyte Concentration

ethanol (Ɛ= 3.5 and 4 respectively). However, with the measured error limit of 3 MHz and

sensitivity of 100 MHz/permittivity, the limit of selectivity is extended to the order of 0.1 in

absolute permittivity value. The calibration step was followed by the measurement of binary

mixtures of fat and calcium in blood. A similar control experiment was performed using water as

the suspending medium for different fat and calcium concentrations.

Pure pig-blood samples were procured and binary mixtures were prepared with varied

concentrations of liquid fat and calcium. The suspending medium was blood.

Figure 4.14 Calibration of the sensor system. Resonant frequency downshifts with increasing

permittivity of alcohol.

Figure 4.15 Variation of the resonant frequency of the oscillator with varying fraction of fat and

calcium in blood. The resonant frequency scales up with increasing concentration.

_____________________________________________________________________________________________

___________________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

Fig. 4.15 shows the resonant frequency scaling of the oscillator with respect to the fat and

calcium concentration in the mixture samples. The resonant frequency of the oscillator scales up

with increasing concentration of fat and calcium in the mixture. Permittivity of the suspending

medium, blood Ɛ = 46), is higher (when compared to calcium (Ɛ = 9) and fat (Ɛ = 4.5) [109]. As

far as the specificity of the sensor is concerned, it should be kept in mind that the sensor will be

used to screen calcified fat. Therefore, detection of calcified fat from blood is the main focus of

the sensor.

Therefore, increasing concentration of fat and calcium lowers the overall permittivity of the

mixture, lowering the IDC capacitance resulting in up scaling of the oscillating frequency. The

permittivities extracted from the above measurements along with the aid of the calibration step

fit precisely with the analytically obtained values from binary mixture laws governed by

Lichtenecker equations below.

∈𝑟𝑒𝑠= exp[𝑣1𝑙𝑛 ∈1+ 𝑣2𝑙𝑛 ∈2] (4.6)

Where v1 and v2 are the volume fraction of the two materials in the mixture with permittivity

values

1 and

2 respectively and

res is the effective permittivity of the mixture. Fig. 4.16 shows

the fit of the extracted permittivity values from the measurements with the analytical equation.

The error in the extracted permittivity values is 0.4 % when compared to the theoretically

calculated value.

Figure 4.16 Permittivity of binary mixture: fat and calcium in blood. The extracted permittivity values are

fit to the analytically calculated permittivity values.

_________________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

A similar control experiment was established with binary mixtures of calcium and fat prepared in

water in order to show the detectability of the sensor is dependent on the permittivity contrast of

the suspending medium and calcium and fat. The frequency response of the oscillator is shown

in Fig. 4.17. As expected, the resonant frequency increases with increasing concentration of fat

and calcium in the mixture. It is noted that the discrimination window is much higher when the

suspending medium is water. This can be attributed to the higher permittivity contrast between

water, and calcium and fat, when compared to blood. Therefore, it can be deduced that the

detection of varying concentration of foreign material in a suspending medium is highly

dependent on the permittivity contrast of the two. However, in the real environment where such

sensors will be used for plaque characterization, more than the concentration of fat and calcium

in blood, the composition of plaque would vary. Therefore, detection of high calcium and fat

content in blood is a sufficient criterion for the feasibility of the sensor system.

4.3.4 Dielectric imaging of biomaterials

There is an increasing need for spatially localized characterization of biomaterials for effective

analysis of test samples, for example, analysis of different protein molecules on same test plate.

Thus spatial imaging of permittivity for an area of a biological test sample would provide an

accurate understanding of the properties of the test sample. This section of the chapter is

dedicated towards establishing a frequency shift biosensor array for accurate spatial dielectric

imaging of a given area of a biological test sample. In addition, sensor arrays allow

characterization of biomaterials without need of precise positioning of the sensor or of the

biological sample, as a complete area will be scanned.

Figure 4.17 Variation of the resonant frequency of the oscillator with varying fraction of fat and calcium

in water. The resonant frequency scales up with increasing concentration.

_________________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

The CMOS sensor oscillator described above has been used in switched array configuration, for

dielectric imaging of biomaterials. Fig. 4.18 shows the architecture of switched four element

sensor array system: a common current source is used for all the oscillators and is connected to

one oscillator at a given point of time with the switches shown in the figure.

In this work pMOS switches were used. The switches were turned off with the bias voltage of

2.5 V applied to the gates of the pMOS transistors. Individual sensor units were activated by

applying 0 V bias to the gates of the corresponding pMOS transistors. The corresponding chip

micrograph is shown in Fig. 4.19.

__________________________________________________________________________________________

Figure 4.18 Four unit switched sensor array. PMOS transistors are used as switches to a common current

source supplying the sensor oscillators. A digital control for the switches is shown.

Figure 4.19 Chip micrograph of four unit sensor array.

_________________________________________________________________________________________

Chapter 4 Detection of Analyte Concentration

As a prototype a four unit sensor array was demonstrated. The sensors are marked on the chip as

S1, S2, S3 and S4. Sensors 1 and 2 (S1 and S2) operate at the same frequency of 8.28 GHz and

sensor 3 and 4 (S3 and S4) operate at 7.8 GHz with power levels of -6 dBm. The chip was

fabricated in 0.25 µm BiCMOS process and the IDC was fabricated on the TM1 of the BEOL

stack; therefore, the sensitivity is of the same order as 6 GHz sensor, of 22 MHz/permittivity. As

an outlook the sensor array is proposed to be used for spatial imaging of immobilized molecules.

Therefore, for future surface chemistry need, the sensor is fabricated in TM1. Three different

materials (glue, air, saturated glucose solution) have been used to demonstrate the functionality

of the imaging approach. In a first step, each of these materials was put on all the sensors

simultaneously. After these first experiments, the sensor array was further used to map different

materials on top of different cells of the sensor array. Fig. 4.21 shows the dielectric mapping of

the biomaterials.

_____________________________________________________________________________________________

Figure 4.20 Imaging of dielectric distribution as tabulated in table 1; Green: glue, Orange: honey, Blue: air.

Chapter 4 Detection of Analyte Concentration

Fig. 4.20 shows the mapping scheme. The materials on one sensor have negligible influence on

the sensitivity of the neighboring sensors. The four cases shown in Fig. 4.21 have been

summarized in table 4.1.

Table 4.1: Four sensor imaging scheme

Case

Senosr1

Sensor 2

Sensor 3

Sensor 4

Glue

8.137 GHz

Air

8.28 GHz

Air

7.79 GHz

Air

7.8 GHz

Glue

8.137 GHz

Glue

8.137 GHz

Air

7.79 GHz

Air

7.794

GHz

Glue

8.132 GHz

Glucose

8.03 GHz

Air

7.78 GHz

Glucose

7.6 GHz

Glue

8.13 GHz

Glucose

8.03GHz

Glue

7.74 GHz

Glucose

7.6 GHz

Careful spotting of the sensors using needles with diameter of the order of 50 µm was used to

spot the sensors with respective materials. The inductors as seen from the chip photograph have

been placed far away from the sensors in the layout. Therefore, the spotting of the sensors does

not influence the inductors. In terms of lateral resolution, the first results show that the resolution

is of the order of the sensor dimension (in order of 50 µm). The sensor array can be further

extended to multi-element array; however, the output architecture should be completely

decoupled and a corresponding frequency counter for the oscillator sensors will be an ideal

architecture.

4.4 Conclusion

In this chapter high-frequency CMOS compatible sensors are demonstrated for various

biomedical applications. The utilization of such a sensor in the detection of concentration of

glucose in a homogeneous glucose solution is demonstrated at two frequency range namely 12

GHz and 6 GHz. The sensitivity of the sensor is shown to be dependent on the proximity of the

sensor to the analyte sample to be probed. Post processing (chemical mechanical polishing) step

to bring the sensor close to the fluid samples has been demonstrated. Such post-processed sensor

chips have been shown to operate in conjunction with polymer based microfluidic systems,

therefore, demonstrating hybrid CMOS/Polymer based sensing platform. The sensor systems

were shown to have sufficient sensitivity and resolution even without the post-processing or

polishing step. When the sensor is placed at a sufficient distance from the fluid sample, (IDC

fabricated on TM1 for 6 GHz sensor) the sensor could distinguish 10% increase in glucose

Chapter 4 Detection of Analyte Concentration

concentration in the solution. Therefore, the sensing platform can be used with or without post

processing steps.

Cytometric applications especially characterization of cells and molecules in a suspension is of

prime focus of today’s biosensors. In this chapter the 12 GHz sensor system has been used to

detect the concentration of polystyrene microbeads in a suspension. The advantages of using

high-frequency sensor approach for detection of concentration of particles in a biological

suspension have been shown in this chapter. Therefore, CMOS compatible high-frequency

sensors can be used for cytometric applications, for e.g. distinguishing living cells from dead

cells in suspensions, detection of concentration of white or red blood corpuscles in serum etc.

There is an endeavor to establish microwave sensor modalities for minimally invasive detection

of plaques in arteries in order to offer a cheap and robust alternative to the established optical

approaches. A section of this chapter deals with the establishment of the proof of concept of

microwave sensors that can be developed in the future to detect plaque in arteries. In this chapter

calcium and fat in their liquid phase have been detected in blood. Various concentrations of

calcium and fat in blood have been precisely sensed. The measurement was verified by

extracting the permittivity of the mixtures from the measurements and fitting it to the analytically

obtained values. The fit of the measurement with the analytical results demonstrates the

feasibility of establishment of microwave sensors for intravascular imaging modalities. In the

final application of the high-frequency biosensors, an array of such sensors applied to dielectric

imaging of biomaterial is demonstrated. With the sensor prototype described in this work,

precise discrimination of biomaterials based on their permittivity values is demonstrated.

All in all the main focus of the chapter was to demonstrate the capabilities of CMOS compatible

high-frequency sensors for various biological and medical diagnostic applications. Such sensors

are fast, require no labeling and are easy to handle because of simple front-end circuit platform;

therefore, they are termed as the next generation biosensors ahead of the established optical

sensor platforms.

Chapter 5 Towards Particle Counting

TOWARDS PARTICLE COUNTING

In this chapter* we propose a sensor architecture and a corresponding read-out technique for

detection of dynamic capacitance change that can be applied to rapid particle counting and single

particle sensing in a fluidic system. The sensing principle is similar to the previous chapters and

is based on capacitance variation of an interdigitated capacitors (IDC) structure embedded in an

oscillator circuit. The capacitance scaling of the IDC results in frequency modulation of the

oscillator. A demodulator architecture is employed to read-out the frequency modulation caused

by the capacitance change. A self-calibrating technique is employed at the read-out amplifier

stage. The capacitance variation of the IDC due to particle flow causing frequency modulation

and the corresponding demodulator read-out has been analytically modelled. Experimental

verification of the established model and the functionality of the sensor chip were shown using a

modulating capacitor independent of fluidic integration. Initial results show that the sensor is

capable of detecting frequency changes of the order of 70 parts per million (PPM) which

translates to a shift of 1 MHz at 14.3 GHz operating frequency. It is also shown that a

capacitance change every 3 µs can be accurately detected.

5.1 Introduction

Detection and analysis of single cell and particle in aqueous solution is of high relevance in

biosensing applications. Single cell based biosensors have come into prominence primarily due

to emerging field of POCT in the area of disease monitoring and control and medical diagnostics.

The capability of the inclusion of transducers, sensors, detection circuits and microfluidic

platform provides unprecedented advantages to POC devices and has been shown in chapter 3

and chapter 4. An alternative CMOS “all-electrical” approach for standard ELISA based

immunosensors has been shown in chapter 3. On the same lines, CMOS/BiCMOS based

biosensors applied to single particle (cell) analysis can be an attractive alternative to detection

methods based on fluorescence, acoustic, surface plasmon resonance based, amperoemetric, etc.

From the circuit design aspect, designing of CMOS/BiCMOS compatible single particle sensors

requires highly sensitive sensing circuit and at the same time flexible read-out approach. Hybrid

integrated circuits for easy read-out have been demonstrated by research groups [151].

 Parts of this chapter have been published as “Self-calibrating highly sensitive dynamic capacitance sensor towards

rapid sensing and counting of particles in laminar flow systems”, Analyst, 2015, 140, 3262-3272 DOI:

10.1039/C5AN00187K

Chapter 5 Towards Particle Counting

In this chapter, a compact BiCMOS high sensitive capacitive sensor approach is proposed, where

the sensing principle exploits microwave frequencies and at the same time provides a pseudo DC

output. The sensor system shows very high sensitivity of the order of 70 parts per million (PPM)

and is in the range of detection single particles in a suspension as well as depicts excellent

flexibility in handling due to DC output. An analytical model is established to depict the

operation of the capacitive sensor in conjunction with a flow assisted fluidic system. The

functioning of the sensor system is further demonstrated using a modulating capacitor emulating

the flow of particles in a fluid system. Therefore, the proposed system is suitable for particle

counting and single particle sensing applications, as the capacitance modulation due to particle

flow in a fluid system is analogous to modulating capacitor used in this work. The sensing

principle is based on the previously demonstrated (chapter 3, chapter 4) capacitive sensor

embedded in an oscillator circuit. The operating frequency of the sensor is in the range of 12

GHz to 14.5 GHz, thus exploiting the advantages of high-frequency sensing approach. The

frequency modulation of the oscillator due to the capacitance change is read out using a

demodulator circuit. Therefore, the output of the sensor is a pseudo DC (few KHz) signal, thus

making handling of the sensor extremely flexible. All in all, the proposed sensor system adds the

advantages of high-frequency detection technique, miniaturization and simultaneously keeps the

output handling capability simple. Moreover, the topology opens the possibility of integrating

functionalities such as in-situ signal processing, making these chips even more lucrative. The

measurement time of the sensor is dependent on the settling time of on-chip circuit blocks and

can be reduced to the order of few micro seconds. Therefore, the measurement time can be

reduced considerably compared to other aforementioned techniques.

The theory has been further extended to address the problem of noise in such integrated

microfluidic systems. Noise from the sensor circuit and also from the external biological

environment plays a crucial role in such devices. Noise can be eliminated by using a correlation

technique using two such demodulator architectures with the same integrated system. The recent

integration possibilities of such sensor chips with MEMS-based microfluidic systems add more

relevance to such sensors being used in biosensing [98, 152]. Therefore, high-frequency

microelectronics-based fluidic sensor circuits with DC output handling can be suggested as a

promising tool for the miniaturization of conventional biological cell detection techniques.

5.2 System dynamics analysis

The capacitive sensors demonstrated so far in this thesis work were suited for detecting dielectric

variation in the IDC environment, however, in a static condition. When embedded in an

oscillator circuit, the capacitance changes of the IDC resulted in the resonance frequency shift of

the oscillator and was used to determine the concentration of biomarkers like creatinine,

concentration of glucose or concentration of suspended particles in an aqueous solution and

more. As mentioned above the sensors were operated in a static condition, even with the aid of

Chapter 5 Towards Particle Counting

microfluidic systems, average concentration of particles in the suspension was detected. In this

work, we propose an advanced circuitry and an analytical theory to extend the capacitive sensing

technique based on frequency shift sensor towards a flow assisted fluidic system.

Applications of flow assisted systems range from analysis of single particles (for e.g.,

cells) to counting of particles in a solution. The extension to a dynamic approach is

brought by the inclusion of a demodulator circuitry to detect the frequency modulation

that would be caused by the dynamic capacitance shift due to flow of particles in the fluid

system. The sensor is designed to operate in the frequency range of 12 GHz to 14.3 GHz,

with the demodulator output in the range of few kHz.

The system is modelled in two steps: in the first step the dynamic capacitance change of the IDE

due to particle flow in an aqueous solution is modelled and simulated. In the subsequent step the

demodulator circuitry for detection of the dynamic particle flow is mathematically modelled and

simulated.

5.2.1 Modelling of dynamic capacitance sensor

The modelling of the capacitive sensor in a fluid flow environment where the flow of particles

causes capacitance modulation of the IDC is done using a long fluid channel approximation. In

the long channel approximation, the sensor is assumed to be considerably far from the inlets and

the outlets of the fluid system. Such a sensor configuration is fabricated for particle

concentration analysis in chapter 4. Fig. 5.1 shows a test structure of the same.

_________________________________________________________________________________________

Figure 5.1 Fabricated sensor chip with long channel microfluidic system integration. a) High-frequency

sensor chip showing the sensor arrangement. b) A long channel microfluidic channel is aligned on top of

the sensor. The two conditions depict the channel with and without the fluid.

a.)

b.)

Chapter 5 Towards Particle Counting

In such a condition, the suspended particles in a laminar flow of the aqueous solution are in a

steady state when they reach the sensor. The velocity of the particles can be approximated to the

mean velocity of the fluid on top of the sensor. This is significant in order to determine the flow

rate of the particles.

The inflow and outflow of particles on top of the sensor creates a capacitance modulation. Fig.

5.2 (a) shows the model for the flux of the particles on top of the IDC sensor, in the fluid system.

The permittivity contrast between the aqueous medium and the particles determine the height of

the capacitance modulation. The 2D geometry of the IDC sensor structure along with the

simulated variation of its capacitance due to a particle flowing on top of it is shown in Fig. 5.2

(b). IDC sensor employed in this work to design the sensor oscillator has finger width equal to

finger spacing of 5 µm. A particle of diameter 8 µm (diameter of a standard yeast cell) and

permittivity 20 is considered, flowing in an aqueous solution (solution of water) of permittivity

60. In the previous chapter the permittivity of water with respect to frequency was shown. At the

frequency range of operation of the sensor system (12 GHz-14.5 GHz), the aqueous solution of

water should be around 40. Therefore, the assumed permittivity values are pragmatic, as the

aqueous solution which is generally a solution of water.

__________________________________________________________________________________

_____

Figure 5.2 a) Schematic depiction of particle flow in a long channel fluid system aligned on top of

the sensor. b) Geometry of IDE sensor considered in this work. Simulated variation of sensor

capacitance due to flow of particles. The capacitance of variation is plotted with respect to

position of particle on top of the sensor.

a.)

b.)

Chapter 5 Towards Particle Counting

As the particle migrates on top of the sensor, the capacitance reduces as shown in Fig. 5.2

(b). This can be attributed to the lower permittivity of the particles compared to the

suspending aqueous solution. The IDC regains its capacitance value once the particle

moves away from the sensor. This is defined as the capacitance modulation. A steady

flow of such particles will, therefore, cause capacitive pulses. The simulation was carried

out on COMSOL multi-physics software. The irregularities in the simulated curve come

from the meshing of the structure with the moving particle.

Embedded in the oscillator these capacitive pulses will translate to resonance frequency

modulation of the oscillator circuit. The fluid velocity in the channel and the

concentration of particles in the fluid determine the modulation rate. From the sensing

aspect, detection of this frequency modulation will enable particle counting. This dynamic

behaviour of the capacitive sensor based on the particle flow can be sensed using an

integrated phased-locked loop (PLL) demodulator in conjunction with the sensor

embedding oscillator circuit.

A typical PLL circuit stabilizes the resonance frequency of a voltage-controlled oscillator (VCO)

using a reference, typically a crystal oscillator [110]. In the designed sensor system, a

permittivity-controlled oscillator replaces the VCO in the PLL, where the variable capacitor

(IDC sensor) in the oscillator is a function of permittivity instead of voltage, as explained in

previous chapters. When the resonant frequency of the oscillator is modulated by a moving

particle, or by particles of different type the PLL output frequency is stabilized by a control

voltage, which serves as the demodulator output. The significant aspect of the PLL used in the

sensor system is the constant gain; this enables a self-calibrating feature of the sensor

architecture enabling detection of extremely minute capacitance change. A detailed analysis of

the self-calibrating feature of the sensor system is done in the subsequent sections.

5.2.2 Design of sensor circuit

The oscillator sensor circuit has the same topology as was described in the previous chapters; a

cross-coupled CMOS oscillator using the IDC sensor as the variable capacitor. However, for the

design of the complete sensor system additional variable capacitors are used as shown in Fig.

5.3. The CMOS cross-coupled oscillator is further embedded in a PLL to demonstrate the

proposed technique of frequency modulation-demodulation. The sensor IDC is employed along

with three variable capacitors (varactors) in order to modify the oscillation frequency. The

varactors consist of two pMOSFETs, the sources and drains of which are connected to the

control voltage.

Chapter 5 Towards Particle Counting

A large coarse-tuning capacitor Ccoarse is responsible for compensating Process-Voltage-

Temperature (PVT) variations. The MOSFETs used for the Ccoarse is 8 fingers, with W/L ratio of

12 for each gate finger. The value of the on chip Ccoarse is 137 fF. PVT variations are brought

about by uncontrolled variations in the CMOS processing altering the properties of the active

devices (transistors), minor variations in the supply voltage altering the operating point of the

circuits or variation of on chip temperature varying the circuit conditions. Hu et al [153] deals

with the influence of PVT variations on circuit performance and corresponding compensation

techniques. A small fine-tuning capacitor Cfine is used to detect small frequency changes. For the

Cfine, single finger MOSFETs has been used with the same W/L ratio of 12. The value of the Cfine

is 3.2 fF. Moreover, an additional small varactor, Cmod is used to emulate the dynamic

capacitance change for the initial measurements independent of an integrated microfluidic

channel. The Cmod has the same dimension and value of the Cfine. The buffer stage is used to

isolate the oscillator from the subsequent circuit chain following the oscillator. The following

table shows the values of the individual components used in the sensor circuit design.

_______________________________________________________________________________________

Figure 5.3 Schematic of the sensor circuit. The sensor is embedded in the oscillator circuit. The

variable capacitor Cmod is used for experiments without fluid integration.

Chapter 5 Towards Particle Counting

Name of the component

Value

Inductor (L)

320 pH

Fine Capacitor (Cfine)

3.2 fF

Coarse Capacitor (CCOARSE)

137 fF

Modulating capacitor (Cmod)

3.2 fF

Sensor Capacitor (IDC): simulated

50 fF

The rationale behind this topology of the sensor oscillator is to use a relatively slow coarse

tuning loop with a high gain together with a fast fine tuning loop with a relatively small gain.

This makes the detection of fast dynamic changes in the sensor capacitor easy. This will be

explained in the subsequent section.

5.2.3 Frequency demodulator architecture

Figure 5.4 shows the frequency demodulator architecture with the sensor circuit embedded in the

demodulator architecture.

The readers should refer to works of Hu et al and Herzel et al [153, 154], for detailed

understanding of the circuit blocks used in the frequency demodulator architecture. However, in

this section of the chapter a brief description of the circuit blocks will be presented. The main

focus is on the derivation of the analytical equations governing the frequency demodulation

architecture.

__________________________________________________________________________________________

Figure 5.4 Demodulator architecture block diagram. The output of the demodulator is taken from the

fine-tuning loop containing C1 and R. The VCO shown in the block represents the oscillator with sensor.

Table 5.1 Component parameters for sensor circuit

Ccoarse

Chapter 5 Towards Particle Counting

A second-order charge pump (CP) PLL is considered as was shown in the work of Herzel et al

[155]. The oscillator circuit is controlled by two sets of tuning voltages: V1 supplying the Cfine

varactor of the oscillator circuit also referred to as fine-tuning voltage; V2 supplying the Ccoarse

varactor and is referred to as the coarse tuning voltage. The voltages are controlled by the two

CPs shown as CP1 and CP2 in Fig. 5.4. The CPs are current sources with switch supplying the

control voltage to the oscillator, based on the input from the phase frequency detector (PFD).

The UP input of the PFD is driven by a reference signal with phase φ (t) in the range of MHz.

The oscillator output frequency is divided by the appropriate dividing ration (N), shown by the

1/N block. This is done to equate the frequency output of the oscillator to the reference input at

the normal condition (when no frequency tuning of the oscillator occurs). The divider output is

connected with the DN input of the PFD. The PFD compares the frequencies of the divided

oscillator output and the reference signal. When the oscillator output is tuned due to the variation

of permittivity of the IDC sensor, the PFD will generate the required signal for the CPs to

generate the appropriate voltage for the restoration of the original resonance frequency of the

oscillator.

The fundamental idea of designing the PLL are:

- Large gain of the detector

- Constant gain of the detector

The detector gain is inverse of the oscillator gain. The above requirements translate to the fact

that that the oscillator gain should be small and constant. This requires the capacitor Ccoarse to be

sufficiently large such that the coarse tuning loop has a weak influence on the loop dynamics.

Since the detector gain is basically the inverse VCO gain [105], the FM detector is highly linear.

The settling time of the overall loop should be fast enough in order to detect the dynamic

capacitive changes. The bias voltage on V1 (t) which stabilizes the oscillator frequency by tuning

the Cfine varactor, will be taken as the output of the demodulator. This output is fed to an

operational amplifier for the amplification of the output signal.

We can now proceed with the mathematical analysis of the frequency demodulator architecture.

We consider a PLL with an FM input signal

𝜔𝑅𝐸𝐹(𝑡)= 𝜔0+𝑚𝜔0sin(𝜔𝑚𝑡) (5.1)

where ωm is the modulation angular frequency, ω0 is the mean reference angular frequency, and

m is the modulation index. We define the phase error at the phase frequency detector (PFD) input

by,

𝜑𝑒(𝑡)= 𝜑𝑅𝐸𝐹(𝑡)−𝜑(𝑡) (5.2)

Its first derivative is given by

Chapter 5 Towards Particle Counting

𝑑𝜑𝑒(𝑡)

𝑑𝑡 = 𝜔𝑅𝐸𝐹(𝑡)−𝑑𝜑(𝑡)

𝑑𝑡 (5.3)

Substituting eq. 5.1 in eq. 5.3 we obtain the second derivative given by

𝑑2𝜑𝑒(𝑡)

𝑑𝑡2= 𝑚𝜔0𝜔𝑚cos(𝜔𝑚𝑡)−𝑑2𝜑(𝑡)

𝑑𝑡2 (5.4)

This equation will be useful to eliminate φ (t) from the differential equation describing the

dynamics of the demodulator architecture.

In the following, we consider a linear, time-invariant continuous-time model (CTM) to keep the

analysis of the FM-induced phase error simple. Describing the governing equations of the other

blocks in the demodulator architecture is necessary in order to derive the output voltage of the

demodulator. Considering Ccoarse tending to infinity, the PLL corresponds to a single-loop

operation as far as the small signal behavior is considered. In that condition, the gain of the PFD

is defined as,

𝐾𝑃𝐹𝐷1 = 𝐼𝐶𝑃1

2𝜋 (5.5)

where Icp1 is the charge pump (CP) current in the fine tuning loop containing C1 in the ON state.

The average CP current is obtained as

𝐼1(𝑡)= 𝜑𝑒(𝑡)𝐾𝑃𝐹𝐷1 (5.6)

The resulting voltage across the R-C1 filter in the fine-tuning loop is given as,

𝑉1(𝑡)= 𝑅𝐼1(𝑡)+1

𝐶1∫𝐼1(𝜏)𝑑𝜏

𝑡

0+ 𝑐𝑜𝑛𝑠𝑡. (5.7)

Here, we used the first-order loop filter composed of C1 and R in order to simplify the analysis.

A more detailed analysis would include the biasing resistors and bypass capacitors. The loop

filter has been described following the overall analysis of the demodulator. However, the

simplification does not imply loss of generality in the derived equation. The derivative of (5.7) is

obtained as,

𝑑𝑉1(𝑡)

𝑑𝑡 = 𝑅𝑑𝐼1(𝑡)

𝑑𝑡 + 𝐼1(𝑡)

𝐶1 (5.8)

The equation governing the oscillator output is given as,

𝑑𝜔(𝑡)

𝑑𝑡 = 2𝜋𝐾1𝑑𝑉1(𝑡)

𝑑𝑡 (5.9)

Where K1 is the oscillator gain. Substituting (5.8) into (5.9) we obtain,

Chapter 5 Towards Particle Counting

𝑑𝜔(𝑡)

𝑑𝑡 = 2𝜋𝐾1𝑅𝑑𝐼1(𝑡)

𝑑𝑡 +2𝜋𝐾1𝐼1(𝑡)

𝐶1 (5.10)

The PFD input phase obeys

𝑑2𝜑(𝑡)

𝑑𝑡2= 1

𝑁𝑑𝜔(𝑡)

𝑑𝑡 (5.11)

Substituting (5.10) into (5.11) we obtain

𝑑2𝜑(𝑡)

𝑑𝑡2= 2𝜋𝐾1𝑅

𝑁 𝑑𝐼1(𝑡)

𝑑𝑡 + 2𝜋𝐾1𝐼1(𝑡)

𝑁𝐶1 (5.12)

Replacing the average value of I1 (t) obtained in (5.6) into (5.12) we obtain

𝑑2𝜑(𝑡)

𝑑𝑡2= 2𝜋𝐾𝑃𝐹𝐷1𝐾1𝑅

𝑁 𝑑𝜑𝑒(𝑡)

𝑑𝑡 + 2𝜋𝐾𝑃𝐹𝐷1𝑘1𝜑𝑒(𝑡)

𝑁𝐶1 (5.13)

φ (t) can be eliminated from the above equation by utilising (5.4) resulting in

𝑑2𝜑𝑒(𝑡)

𝑑𝑡2+ 2𝜋𝐾𝑃𝐹𝐷1𝐾1𝑅

𝑁 𝑑𝜑𝑒(𝑡)

𝑑𝑡 + 2𝜋𝐾𝑃𝐹𝐷1𝑘1𝜑𝑒(𝑡)

𝑁𝐶1 = 𝑚𝜔0𝜔𝑚cos(𝜔𝑚𝑡) (5.14)

Now we incorporate the coarse tuning loop consisting Ccoarse in the analysis. According to the

block diagram of the demodulator architecture shown there is no additional resistor as was

present in the fine-tuning loop. Therefore, the obtained voltage equation at the coarse tuning

node corresponding to equation (5.8) is

𝑑𝑉2(𝑡)

𝑑𝑡 = 𝐼2(𝑡)

𝐶𝑐𝑜𝑎𝑟𝑠𝑒 (5.15)

It can be seen from the demodulator architecture, the two charge pumps are driven by the same

PFD such that the waveforms V1 (t) and V2 (t) are the same, except for the constant factor given

by the ratio of the charge pump currents in the ON state. Therefore, the voltage equation at the

coarse tuning loop is given as

𝑑𝑉2(𝑡)

𝑑𝑡 = 𝐼1(𝑡)

𝐶1 𝐼𝐶𝑃2𝐶1

𝐼𝐶𝑃1𝐶𝑐𝑜𝑎𝑟𝑠𝑒 (5.16)

The oscillator frequency is the sum of the control voltages weighted by the gains of the oscillator

for individual loops,

𝑑𝜔(𝑡)

𝑑𝑡 = 2𝜋𝐾1𝑑𝑉1(𝑡)

𝑑𝑡 + 2𝜋𝐾2𝑑𝑉2(𝑡)

𝑑𝑡 (5.17)

Including the above constraints of two loops, we obtain a similar equation as (13), and is given

as,

𝑑2𝜑𝑒

𝑑 (𝑡)

𝑑𝑡2+2𝛾𝑑𝜑𝑒

𝑑(𝑡)

𝑑𝑡 + 𝜔𝑛

2𝜑𝑒

𝑑(𝑡)= 𝐹𝑐𝑜𝑠(𝜔𝑚𝑡) (5.18)

Chapter 5 Towards Particle Counting

where we introduced the following abbreviations

𝛾 = 𝐼𝐶𝑃1𝐾1𝑅

2𝑁 (5.19)

𝜔𝑛

2= 𝐼𝐶𝑃1𝐾1

𝐶1𝑁(1 + 𝐼𝐶𝑃2𝐾2𝐶1

𝐼𝐶𝑃1𝐾1𝐶𝑐𝑜𝑎𝑟𝑠𝑒) (5.20)

𝐹 = 𝑚𝜔0𝜔𝑚 (5.21)

Equation 5.18 is a well-known differential equation describing a damped harmonic oscillator

driven by external force. In our case, the driving force is the variation of the capacitance due to

flow of particles on top of the sensor. The solution of such a differential equation has been well

discussed and can be used further to obtain the demodulator output voltage, which serves as the

output of our sensor system.

The solution of the differential equation yields,

𝑉1(𝑡)= 𝑉𝑑𝑒𝑚 cos(𝜔𝑚𝑡+ 𝜑1) (5.22)

Solution for Vdem shows that it can be expressed as

𝑉𝑑𝑒𝑚 = 𝑚𝜔0

2𝜋(𝐾1

𝑁) (5.23)

The Vdem output is fed to the operational amplifier as shown in Fig. 5.5. A resistance Rin of value

5 MΩ was used and the corresponding RC biasing of the referenced input results in the same DC

values of both the inputs to the amplifier. The Vdem output serves as one of the inputs to the

amplifier, while the other input is the time averaged value of Vdem due to a very large value of

the biasing capacitance used. This technique eases the amplification of very small signal changes

at the demodulator output regardless of PVT variations. The transistors were sized as 20 µm for

the amplifier stage. The device mismatch between the two transistors of the amplifier does not

influence the operation of the amplifier. It requires no further external calibration often needed

_______________________________________________________________________________________

Figure 5.5 Differential operational amplifier with RC biasing for self-calibration.

Chapter 5 Towards Particle Counting

for differential amplifiers, therefore, depicting the self-calibrating feature of the sensor

architecture.

It is important to understand the filter that is used in conjunction with the charge pump. Fig. 5.6

shows the architecture of the loop filter. The coarse capacitor Ccoarse is large enough and as

mentioned above has no influence on the loop dynamics. Therefore, the coarse loop can be

ideally considered not active, and the loop filter then is a third order loop filter. The transfer

function of the loop filter is given as,

𝐺(𝑠)= 1

𝑠𝐶3

1 𝑍2

⁄

1 𝑍2

⁄+1 𝑍1

⁄ (5.24)

where,

𝑍1= (𝑅 + 1

𝑠𝐶1) ∥ 1

𝑠𝐶2∥ 𝑅𝑏𝑖𝑎𝑠1 ∥ 𝑅𝑏𝑖𝑎𝑠2 (5.25)

𝑍2= 𝑅3+1

𝑠𝐶3 (5.26)

The voltage divider incorporated with with the resistors Rbias1 and Rbias2 are significant in terms

of the dual loop operation of the PLL. This arrangement keeps the fine tuning voltage at a value

where the gain is constant. A large capacitor used in the coarse tuning loop (Ccoarse) acts as the

filter for the coarse tuning loop. For the approximation of the large coarse capacitor to be valid,

the value of Ccoarse should be

𝐶𝑐𝑜𝑎𝑟𝑠𝑒 ≫ 𝐼𝐶𝑃2𝐾𝑃𝐹𝐷2

𝐼𝐶𝑃1𝐾𝑃𝐹𝐷1 𝐶1 (5.27)

Figure 5.6 Loop filter showing the coarse and fine tuning loop.

________________________________________________________________________________________

Chapter 5 Towards Particle Counting

The primary filter parameters R and C1 are chosen for this operation as, R = 1K and C1 is

chosen to be 1 nF and is implemented on board. The loop bandwidth can be obtained by dividing

eq. 5.19 by 2.

𝑓𝐵𝑊 =𝐼𝐶𝑃1𝐾1𝑅

2𝜋𝑁 (5.28)

The charge pump current in the fine tuning loop is set as 4 mA. The divider ratio is 64. The gain

of the oscillator K1, 100MHz/V, is kept low due to high detector gain, as was mentioned

previously. Therefore, R is the tuning parameter for the loop bandwidth. The loop bandwidth was

designed for 300 KHz. This has been done for the dynamic detection of capacitive change every

3 s.

The charge pump architecture for the fine tuning loop is shown in Fig. 5.7. The pMOS transistor

delivers the UP current and nMOS transistor delivers the DOWN current. This architecture is

used in order to have a low noise charge pump. For a particular charge pump current moderately

sized transistors are used with high gate-source voltage [154]. For the course tuning loop a

traditional low-current charge pump is used with a large capacitor to block the noise.

The voltage divider network comprising of the resistors Rbias1 and Rbias2 at the output of the

charge pump stabilizes the DC output voltage at a desired value of Vdd x (Rbias2/Rbias1 + Rbias2).

At the same time the fine tuning loop gain of the oscillator is kept constant against PVT

variations. The value of Rbias1 and Rbias2 is 1 K.

Figure 5.7 Charge pump architecture for fine tuning loop.

________________________________________________________________________________________

Chapter 5 Towards Particle Counting

The VCO sensitivity is described as the voltage shift caused by the frequency response due to a

single particle on the sensor. This voltage shift should be higher than the standard deviation of

the voltage caused by the noise. The second order PLL results in a flat voltage spectrum with

decay depending on 1/f2 at high frequency. This voltage spectrum is low-pass filtered by a

relatively slow operational amplifier shown in Fig. 5.5, implying the loop bandwidth should be

small for a good resolution. As mentioned above, the loop bandwidth is designed for 300 KHz.

The oscillator sensitivity in detection of capacitive pulses is discussed in the results section.

5.3 Results and discussion

The sensor system is fabricated in standard 0.13 µm SiGe:C BiCMOS process as explained in

the previous chapter. The BEOL stack with the metallization was discussed in the previous

chapters. The fabricated chip is shown in Fig. 5.8. The total area of the chip is 2.4 mm2.

For the electrical characterization of the chip, the chip was mounted on a FR4 board using

dielectric glue. The operating frequency range of the sensor system is 12 GHz to 14.5 GHz. The

test chip has an RF output from the sensor oscillator in order to characterize the sensor. Wire

bonding technique is used to make the electrical connections from the chip to the FR4 board. The

operating frequency range of the sensor makes wire bonding a feasible approach for electrical

connections. The total package size of the system is 5 cm x 2 cm. As mentioned above, the

sensing principle being based on high-frequency dielectric detection requires no additional

reference electrodes for measurements. Therefore, no additional bulky test-benches are required

for the measurement setup, restricting the overall size of the system to the package size of the

chip. This small size of the chip therefore makes the chip lucrative for LOC application.

________________________________________________________________________________________

Figure 5.8 Chip photograph of the sensor and demodulator architecture.

Chapter 5 Towards Particle Counting

The fastest frequency modulation rate that the sensor system is capable of determining, defines

the fluid pressure or velocity that can be approximately used with such a sensor system. This is

in turn determined by the settling time of the PLL. The PLL settling time, the higher is the

velocity of the fluid system that can be used. The value of the capacitor Ccoarse defines the settling

time of the PLL. Fig. 5.9 shows the simulated settling time of the PLL as a function of the value

of Ccoarse..

For higher value of Ccoarse the PLL has a faster settling time and a value of 10 nF gives a settling

time of approximately 3 µs. However, physically integrating of 10 nF capacitor is impossible on

the chip due to huge area constraint. Therefore, the 10 nF capacitor can be implemented

externally on the board where the chip is mounted. This gives the flexibility to adjust the settling

time of the PLL and in turn the sensor system in accordance with the intended application. This

makes the chip suited for a wide range of applications requiring different fluid velocities in the

microfluidic system. In the present analysis, the settling time of the PLL shows the minimum

required measurement time of the system could be as low as 3 µs to 5 µs. Therefore, extremely

fast measurements can be performed as compared to the established sensor platforms.

Fig. 5.10 shows the simulated demodulator output voltage for a sinusoidal frequency modulation.

The modulation period chosen is 100 µs and the modulation index is 0.0001 (100 parts per

million). From the sensing aspect, these simulation conditions translate to capacitance change

due to particle flow every 100 µs. Therefore, the capacitive modulation pulses discussed in Fig.

5.3 (b), will the period of 100 µs. This kind of period of capacitance change depicts extremely

low solute or particle concentration in a solution. The modulation index relates to the change in

________________________________________________________________________________________

Figure 5.9 Simulated settling time of the PLL as a function of the coarse loop filter capacitance. The

settling time obtained is 3 µs. Capacitance change every 3 µs can be accurately detected.

Chapter 5 Towards Particle Counting

resonant frequency of the oscillator to the presence of a particle on top of the embedded IDC

sensor. With the closed loop operating frequency of 14.3 GHz, the modulation index of 0.0001

translates to a change of 1.43 MHz. Therefore, the sensor shows high order of sensitivity and

detection resolution along with very fast response time.

As seen from the simulation results, the demodulator output voltage follows the modulating

voltage. This can be attributed to the fact, that the modulation period is much slower compared to

the PLL settling time and therefore, any modulation of the frequency due to capacitance change

is accurately followed.

The electrical measurement of the chip shows that the overall DC current drawn by the chip from

a 3.3 V supply is 80 mA. A measurement of the process variation was conducted to deduce the

reproducibility of the chip. Several chips were measured from the same wafer and the output

characteristics were not seen to vary more than 0.2%. The resonant oscillator circuit was

characterized and measured to determine the operating frequency of the sensor. Fig. 5.10 shows

the output spectrum of the closed loop resonant oscillator circuit. The tuning range of the PLL is

from 12.6 GHz to 14.3 GHz as was measured by tuning the bias voltage of the on chip Ccoarse

varactor. The output power is -6.7 dBm and the reference spur level is below -62 dBm. From the

output spectrum shown in Fig. 5.10, the noise level compared to the signal output is shown; this

noise floor is low enough to allow the locking of the PLL. Additionally, high order low pass

filter is employed for a smooth detector output at a given detector gain. As mentioned in the

previous sections in order to determine the sensitivity of the demodulator independently of

__________________________________________________________________________________________

Figure 5.10 Simulated demodulator output for input modulating voltage of period 100 µs. The period of

the voltage being much higher than the settling time of PLL, it is accurately followed by the demodulator.

Chapter 5 Towards Particle Counting

fluidic system, a small sinusoidal signal Vmod was applied to the modulating capacitor. The

output voltage of the demodulator is taken as mentioned in the block diagram of the demodulator

architecture and is fed to an operational amplifier for further amplification. The modulation input

of the VCO has a gain of 100 MHz/V at a DC level of 1.25 V. By adding a sinusoidal low-

frequency modulation signal of 10 mV peak-to-peak amplitude to a DC voltage of 1.25 V the

oscillator frequency changes by 1 MHz in open loop condition.

This change of 1 MHz translates to a modulation index of 0.00007 or 70 parts per million. In

closed-loop operation, the oscillator frequency is kept constant, while the fine-tuning voltage is

modulated. The demodulation sensitivity is obtained by changing the modulation frequency and

measuring the rms value of the demodulator output voltage.

________________________________________________________________________________________

Figure 5.11 The output spectrum of the sensor oscillator. The operating frequency is 14.272 GHz.

_________________________________________________________________________________________

Figure 5.12 Demodulator output voltage as a function of the modulation period. The demodulator

voltage follows the input period till 300 KHz (3.3 µs).

Chapter 5 Towards Particle Counting

Fig. 5.12 shows the demodulator output as the function of the period of modulation. The

applied DC voltage is 10 mV peak to peak. The PLL is unable to follow the signal above

the loop bandwidth of 300 KHz (time period 3 µs). In such a case the demodulator output

voltage is reduced. In terms of the sensing aspect, modulating frequency 300 kHz relates

to a measurement speed of 3 µs. Therefore, following the fluidic integration, every 3 µs a

capacitance change due to flow of particle on top of the senor can be accurately detected.

This measurement time is sufficiently small when compared to the state-of-the-art particle

sensing. The proposed architecture can, therefore, sufficiently increase the time efficiency

of such microelectronics integrated fluidic systems. However, slower fluid flow can be

detected using a larger value for the C

coarse

capacitor. In order to detect changes of the

order of milliseconds a higher coarse tuning filter capacitor C

coarse

is required. The lower

limit for the C

coarse

value of 100 nF is 50 KHz corresponding to 20 µs. This lower limit

can be further increased as seen in Fig. 5.12, where the C

coarse

value of 2 µF extends the

lower limit of measurement to 20 KHz. This accounts for a measurement speed where a

change of capacitance up to every 50 µs can be detected. Therefore, based on the

application and the fluid velocity required, appropriate C

coarse

capacitor can be used on the

board. This makes the overall sensor system suited for wide range of fluid velocities.

The sensitivity of the sensor obtained from the electrical characterization using the

modulating capacitor, is of the order of 70 ppm. For the closed loop operation at 14.3

GHz, this resolution translates to the detection limit of 1 MHz. For the modulating

capacitor used in this work, this renders a change of 60 aF for initial capacitance value of

18 fF. From the aspect of frequency shift with respect to permittivity ambient of the IDE,

this ultra-low modulation index detection capability shows a change of 0.25 in the

absolute permittivity value in the dielectric ambient of the IDC. With the measured

sensitivity and resolution, the sensor is highly suited for sensing extremely low particle

concentrations.

The capacitive detection technique is also independent of the polarity of the particles in

the fluidic system. This is primarily due to the sensing principle being based on the

dielectric contrast between the particles and the suspending medium. Therefore, the

sensor system can be ideally used for charged and uncharged species. As mentioned

previously, the measurement with the modulating capacitor is analogous to the

capacitance modulation caused by the particle flow in a fluidic system. Therefore, the

above measurements show that the established model is highly suitable for particle

detection in fluidic systems. Another important aspect of LOC systems is the feasibility of

the same outside laboratory conditions [156, 157], where the difficulty stems out due to

external conditions, like temperature variations mechanical stress etc. The working of the

established prototype sensor system in such conditions will be dependent on the

packaging. However, the external condition will have negligible influence on the sensing

concept due to on-chip stabilisation and configuration capabilities of the chip. In the

Chapter 5 Towards Particle Counting

subsequent sections the capability of correlation technique to eliminate the external noise

is also shown. Therefore, the sensor system can be ideally used outside laboratory

conditions as well. The sensor enhances the measurement time and also possesses self-

calibrating and reconfigurable features, which can be utilized for different applications

based on different fluid flow rates. The stability of the sensor circuit is obtained by the

voltage divider at the charge pump output. This keeps the oscillator gain and the detector

gain constant with respect to PVT variations.

5.4 Dual demodulator architecture

In this section of the chapter dual demodulator architecture is proposed; the dual

demodulator has two sensors that can be aligned on the same microfluidic channel. Such

an architecture is used to counteract the problem of noise in the fluid based sensor system.

5.4.1 Elimination of noise by time-averaging

The significant effect of noise in the sensing system is the limitation in the accuracy of

the particle counting process. A long-term measurement technique with time averaging

capabilities can be a potential solution to the problem. In this case we use two

demodulators on the same chip as described above. The corresponding chip is shown in

Fig. 5.13.

________________________________________________________________________________________

Figure 5.13 Chip photograph showing dual demodulator architecture.

Sensor 1

Sensor 2

Demodulator 1

Demodulator 2

Chapter 5 Towards Particle Counting

The two demodulator architectures are decoupled with different power supplies supplying

the demodulators. Such a system removes any correlated noise that could have otherwise

affected the system if same power supply was used. The dual demodulator architecture is

now modelled and the capability of eliminating noise from the system is depicted.

For this modelling purpose, we consider two demodulator detectors with sensors located

at different positions of a stream line in a fluidic channel. For simplicity, we assume that

the momentary frequencies of the free-running sensor embedded oscillators represent a

chain of rectangular pulses with random position. The corresponding demodulator outputs

are shown in Fig. 5.14.

The demodulator output has the same waveform as the frequency output from the

oscillators, since the PLL settling is fast compared to the frequency modulation. It can be

calculated by multiplying the frequency change with the FM detector gain.

The cross-correlation between the two detector output voltages is defined by

𝐶(𝑡,𝜏)= < 𝑉

(𝑡+𝜏)𝑉

(𝑡)>

(5.29)

where the brackets denote the stochastic average. In steady state, the stochastic average

can be calculated by time averaging over a long period of time T

max

𝐶(𝜏)=

𝑇

𝑚𝑎𝑥

−𝜏

∫𝑉

(𝑡+𝜏)𝑉

(𝑡)𝑑𝑡

𝑇

𝑚𝑎𝑥

−𝜏

(5.30)

If the time is sampled with the step width T

, we can define

________________________________________________________________________________________

Figure 5.14 Pulse train emulating the signals from 2 VCOs which are delayed by time Δt.

Chapter 5 Towards Particle Counting

𝑡

𝑛

= 𝑛𝑇

𝑠

, 𝑛 = 0,1,2,…….𝑁

(5.31)

and

𝜏

𝑚

= 𝑚𝑇

𝑠

, 𝑚 = 0,1,2,……𝑀

(5.32)

The cross-correlation is then given by

𝐶

𝑚

𝑁−𝑚

∑𝑉

𝑛

(2)

𝑉

𝑛−𝑚

(1)

𝑁

𝑛=𝑚

(

5.33

)

For the chain of pulses depicted in Fig. 5.14, the cross-correlation is given shown in Fig.

5.15. The peak maximum of the triangle gives the variance of the voltage, and the peak

position gives the delay between the two detectors.

The main advantage of the correlation method is the fact that non-correlated noise

voltages v

and v

added to the ideal detector outputs V

and V

will be eliminated,

provided that the number N of data points is sufficiently large. In order to illustrate the

noise reduction capability, we added strong random noise to the demodulator output

signals.

________________________________________________________________________________________

Figure 5.15 Correlation between the two demodulator voltage outputs.

Chapter 5 Towards Particle Counting

A correlation between the two pulse sequences infested with random noise signals

demonstrates the elimination of the non-correlated noise shown in Fig. 5.16. It is evident

from Fig. 5.16, non-correlated noise voltages on the two detector outputs can be ideally

eliminated by the principle of time averaging. Device noise, thermal noise and 1/f noise

accounts for this kind of non-correlated noise.

As mentioned above another type of noise in silicon chips arises from the power supply or

termed as supply noise [158]. This type of noise may result in strongly correlated noise in

the two demodulators, especially, if they are integrated on the same chip. Since correlated

noise will not be eliminated by time averaging, noise coupling between the demodulators

through supply or substrate should be minimized. This entails for separate biasing of the

two demodulators and is discussed in [159]. Sufficient distance between noise aggressors

and noise victims, and the use of guard bands around critical circuit blocks are of

advantage as well.

Moreover, electromagnetic coupling through close and parallel bond

wires must be avoided.

Another type of environmental noise is temperature noise. This type of noise was

discussed in the context of oscillator-based reactance sensors [160], where environmental

noise was reduced by noise cancellation and filtering. Since temperature changes are

correlated noise sources for the two sensor capacitances, our approach cannot eliminate

this type of noise. However, if the temperature changes are much slower than the total

measurement time, they have a small effect on the demodulator sensitivity. Moreover,

bandgap references for each of the two demodulators can be used to stabilize the supply

voltages with respect to temperature variations.

________________________________________________________________________________________

Figure 5.16 Pulse trains showing frequency pulses from two oscillators covered with random noise. Cross

correlation between the pulses shows the delay time equal to the one obtained in Fig. 5.15.

Chapter 5 Towards Particle Counting

5.4.2 Particle concentration and flow-rate

In order to detect the concentration of particles and flow rate in the laminar flow system using

the two demodulator architecture, the dynamics of the individual demodulator has to be

optimized while V1(t) in Fig. 4 is the individual demodulator output signal. The two tuning loops

of the demodulator comprising of Ccoarse and Cfine have time constants τcoarse and τfine respectively.

τcoarse determines the sensitivity of the detector system and has to be considerably large compared

to the delay between the “frequency change” events at the two oscillators due to the flow of

particle on top of the respective sensors.

𝜏𝑐𝑜𝑎𝑟𝑠𝑒 > 𝛥𝑡 (5.34)

Δt is the delay between the sensors. From the demodulator architecture shown in Fig. 4, and

analysis of dual loop PLL [162] it is known that a frequency variation of the oscillator is restored

by the coarse tuning loop and the time constant is given by,

𝜏𝑐𝑜𝑎𝑟𝑠𝑒 =𝐶𝑐𝑜𝑎𝑟𝑠𝑒𝛥𝑉2

𝐼𝐶𝑃2 (5.35)

where ΔV2 is the voltage change on the coarse tuning loop due to frequency modulation as

shown in Fig. 4. ICP2 has been described above, is the charge pump current. The condition

mentioned in equation (29) for highly sensitive architecture, requires a high value of τcoarse; this

can be achieved by lowering the ICP2 in conjunction with a high Ccoarse. In the case where the

τcoarse is smaller than Δt, V2(t) in Fig. 4 can be taken as the output of the individual demodulator.

Such a condition arises for extremely slow flow rate or very low solute concentration which

causes the “frequency change” event at the two oscillators to be widely spaced. A similar

mathematics done for V2 (t), as was done for V1 (t) would show a loss of sensitivity in such a

situation. However, V2(t) can be used as an output by sacrificing the sensitivity as the delay is

very large and the output voltage pulses are far apart from each other. In that case the Ccoarse

value should be small for fast settling of V2 (t). Thus, a self-calibration for different flow rates is

seen in the dual demodulator approach as well.

In order to obtain the concentration of particles in the suspension we assume that the frequency

pulses obtained from the two sensors are proportional to the particle density. This assumption is

valid for low to medium solute concentration in the suspension, which is typically the case in

fluidic systems. As mentioned in a previous section the temperature and process variation have

minimum influence on our demodulator architecture, which implies that the output voltage of the

two demodulator sensor is only proportional to the frequency changes in the oscillator.

Therefore, the concentration of particles in the solution can be obtained from the cross-

correlation of the two output signals and can be given as,

𝑛𝑝𝑎𝑟𝑡𝑖𝑐𝑙𝑒 = 𝛼𝜎𝑣

2 (5.36)

Chapter 5 Towards Particle Counting

100

where σv is the magnitude of the correlation peak and α is a proportionality constant.

From the analysis it is seen that there is no theoretical limitation of particle concentration

detection, as the correlation peak will grow with time and can be estimated. Therefore, any

concentration of solute in a suspension can be estimated. However, if the measurement

conditions (for example temperature) change during the measurement time, detection of the real

concentration can be affected and such a condition can be avoided using bandgap references as

mentioned above.

The delay time of the correlation peak can be used to obtain the flow rate of the particles. If the

sensors are separated by a distance s and the peak of the correlation occurs at Δt, the flow rate

can be written as,

𝑣𝑝𝑎𝑟𝑡𝑖𝑐𝑙𝑒 = 𝑠

𝛥𝑡 (5.37)

In order to detect particles with different dielectric characteristics the voltage pulses would be

used. For particles with different dielectric permittivity the height of the output voltage pulse will

be different for different particles as is shown in Fig. 5.14. In order to detect the concentration of

different particles in the suspension the height of the voltage pulses should be analyzed.

However, this requires time recording of the output pulses which in turn would require excessive

data processing and increase the complexity and area of the chip.

5.5 Conclusion

We have presented a highly sensitive PLL demodulator architecture in conjunction with a

capacitance based frequency shift sensor for detection of dynamic capacitance change. The

sensor system can be employed towards particle counting in a flow assisted fluid system. A

sensitivity of 70 ppm was experimentally measured using a modulating capacitor. This

sensitivity allows a sensing capability of 1 MHz frequency shift for 14.3 GHz oscillator sensor.

From the frequency shift sensor aspect this translates to the detection capability of 0.25 in the

absolute permittivity value. Therefore, in the context of flow based sensors with very low

concentration of particles in the suspension this technique offers extremely high sensitivity. The

second significant property of the sensor is its self-calibration capability based on the fluid flow

rate. Capacitance change as fast as every 3 µs can be accurately detected by the sensor system

and has been shown. The fast measurement approach reduces the measurement time

considerably. Owing to the high operating frequency of the sensor, low-frequency dispersion

mechanisms can be avoided while utilizing the sensor for biological suspensions. On the other

hand, the sensor has a very low-frequency (few KHz) output making the handling of the sensor

highly simple. A configuration of two such detectors in a stream of particles in a microfluidic

channel is proposed, where the system noise is suppressed by time averaging. After calibration,

this method will provide particle density, mean velocity and fluid flow rate for a laminar flow in

a microfluidic device.

Chapter 6 Conclusion and Outlook

101

CONCLUSION and OUTLOOK

In this thesis, “More Than Moore” strategy is employed to establish integrated CMOS high-

frequency biosensors. From the aspect of the development of POC diagnostic systems or typical

LOC devices, this single chip solution based on “More Than Moore” approach, provides the

right platform for miniaturized “easy to handle” hand-held devices. The single chip solution not

only eases the sensing operation by making it an “all-electrical” approach, but, with easier data

acquisition capabilities and measurement techniques, reduces the size of the measurement test-

bench and the overall area of the sensor system. The miniaturization of the overall system is one

of the key advantages of the “all-electrical” single chip solution when compared to the state-of

the art optical, electrochemical techniques. The limitations of optical or electrochemical sensors

were outlined in the thesis and the clear advantage of single chip solution was presented by

considering various biosensing applications. In one of the sections of the thesis, it was shown

that the data output from the sensor system can be a simple DC voltage value, thus, elucidating

why complex measurement system is not needed for this kind of sensors. Another important

aspect of single chip solution is the high sensitivity of the sensor system which stems from the

fact that the data acquisition circuit is very close to the sensor owing to the single chip solution.

The cost of the overall sensor system is reduced considerably as well, due to the following

aspects:

- ease of fabrication due to the know process technology

- the “all-electrical” sensing scheme requires no biomarkers

- due to extremely small sample volume, the cost of analyte is less

The other aspect of the thesis is the use of high-frequency permittivity detection as the sensing

technique. The first immediate advantage of use of high-frequency technique is the

miniaturization of the sensor. This was seen all through-out the thesis work where IDC was used

as the sensor and the overall size of the sensor is of the order of few hundreds of micrometers.

This miniaturized sensor size is of the order of biomaterials like biological cells. It also reduces

the volume of the analyte sample needed for sensing. As mentioned above, reduction in the

analyte volume reduces the cost of the overall sensor application. This was observed in the

establishment of the immunosensor for creatinine. Solution volume as low as 2 µl was measured

using the sensor systems. The other advantage of high-frequency dielectric measurement stems

from the aspect of dielectric dispersion of biological suspensions. At lower frequencies

biological suspensions show dispersion mechanisms based on the parameters of the solute

present in the suspension and also based on the solution and sensor electrode interface. At higher

frequencies these dispersion mechanisms are no longer dominant and thus the sensor data

analyzing and processing become considerably easier. The unwanted surface chemistry of

Chapter 6 Conclusion and Outlook

102

electrode and solution interface or the other analyte parameter dependent dispersion mechanisms

no longer play any role in the sensor data output. High-frequency biosensors based on

permittivity detection also evades from the use of reference electrodes, which is commonly used

for other electrical sensors like electrochemical sensors and low-frequency impedance sensors.

This aids in further miniaturization of the overall sensor system. The high-frequency BiCMOS

biosensors based on permittivity detection also provide the additional advantage of negligible

incubation time for sensing. This is due to the capability of measuring extremely small changes

in permittivity as depicted by the developed sensors.

The developed sensor system was applied for various applications like immunosensing, glucose

sensing, detection of analyte concentrations in suspensions, etc. Standard BiCMOS technology

was used for the fabrication of the sensor system. This chapter summarizes the technology

platform along with the established applications of the biosensor platform.

Design and Integration

Standard BiCMOS technology (0.25 µm and 0.13 µm) of IHP was used for the fabrication of the

biosensors. The topmost metal layers of BEOL metal stack (TM1 and TM2) were utilized for the

design of the sensor. The top most metal layer ensures closest proximity of the sensor to the

biomaterial. The biomaterials for e.g. protein molecules are immobilized on top of the

passivation layer above the topmost metal layer. The polymer based microfluidic system is as

well bonded on top of the TM2 metal layer. The top metal layers also provide high quality factor

of the sensor due to high thickness of the metal layers. The quality factor of the sensor was

shown to be a very significant parameter of the sensor design. The sensor being used as a

resonator in conjunction with a pair of inductors, determined the overall quality factor of the

resonator. Thus, the design of the active circuit (CMOS oscillator) driving the oscillations of the

resonator is determined by the quality factor of the overall resonator, as the losses in the

resonator is compensated by a negative resistance mechanism in the active devices (transistors).

In case of the designed biosensor, the quality factor of the resonator is degraded by the loss

factor of the biomaterial (for e.g. biological suspension). Therefore, design of the active circuit

calls for special attention as compared to the standard communication circuits. The highest loss

factor that can be incurred for a specific application had to be taken into account for the design of

the active oscillator circuit. The influence of the biomaterials on the quality factor of other

passive RF components (inductors) in the circuit results in the increase in area budget. However,

owing to the single chip solution, the data acquisition circuits being on the same chip, the overall

sensor system size is smaller when compared to the optical or hybrid amperometric techniques

[55, 58, 64].

IDC structure was used as the sensor in the thesis. At the given frequency range of operation the

designed IDCs were shown to be purely capacitive- structures. The IDCs were shown to obey the

quasi-static approximations of the Maxwell equations. The sensing concept was based on the

variation of the fringing electric field between the fingers of the IDC due to the presence of

Chapter 6 Conclusion and Outlook

103

materials of different values of permittivity on top of it. The intensity of the electric field of the

IDC decreases exponentially in the normal direction of the IDC. The depth to which the electric

field penetrates in the material under test (biomaterials) is determined by the geometry of the

IDC. The capacitance density can also be tuned based on the geometry of the IDC. Thus, based

on the application the IDC can be designed.

The influence of the passivation layer thickness was shown to play an important role in the

sensitivity of the sensor system. As the field degrades exponentially in the perpendicular

direction of the IDC, the thickness of the passivation layer was shown to influence the amount of

electric field penetrating into the biomaterial on top of the passivation layer and in turn affecting

the sensitivity of the system. The sensor was shown to be sensitive up to a passivation thickness

of 10 µm.

Thus with careful consideration of the technology and other design aspects, single chip

biosensors designed for various applications like immunosensing, glucose sensing etc., were

shown.

Dielectric Immunosensor

One of the main applications of the biosensors developed in this thesis was to employ the sensors

in immunosensing application for detection of creatinine. Creatinine concentration is one of the

most frequently detected parameters in clinical diagnostics, as it is the index for renal glomerular

infection. The concentration of creatinine in serum and urinary excretion is primarily unaffected

by dietary changes such as intake of a creatinine-free diet. Therefore, undoubtedly it is one of the

safest detected parameters.

In lieu with the target of the thesis to establish single chip biosensor solutions, one of the main

goals while developing the immunosenor was to accomplish immobilization of creatinine

molecules on the standard passivation layer (Si3N4) of CMOS/BiCMOS technology. This was

achieved by a modified surface chemistry as described in chapter 3 of the thesis. The ability to

immobilize creatinine molecules on the surface of Si3N4 ensured no additional post-processing

step of the sensor chip and therefore, lays the foundation for next generation single chip

immunosensors as opposed to the hybrid ones published in the literature [30, 31, 44, 40, 41].

Established competitive ELISA approach was used for detection of the creatinine concentration.

Incubated anti-creatinine antibodies in different concentrations of creatinine solution were made

to interact with the chip bound creatinine molecules. Different amount of antibodies binding to

the chip bound creatinine molecules gave the measure of the creatinine concentration used in the

incubation phase. The results obtained show that the concentration of creatinine can be detected

in clinically relevant range of 0.88 µM to 88 µM and is comparable to amperometric or optical

techniques used presently. On the other hand, a better dynamic range in case of electrical

Chapter 6 Conclusion and Outlook

104

measurements was observed in comparison to the optical technique with approximately equal

sensitivity.

A frequency shift of 35 MHz was observed with every 10 fold increase in the concentration of

creatinine used in the incubation phase. This sensitivity was measured to be considerably higher

than the process and measurement variation of approximately 4 MHz. Thus, the high-frequency

immunosensor established in this thesis work was shown to be quite robust and there was high

repeatability of the measurement results. A lot of published literature show that creatinine

enzyme immunoassays can specifically measure the concentration of creatinine in real samples

like serum. Therefore, it can be stated that the established high-frequency sensor is capable of

measuring the cretainine concentration in real samples as well. The other aspect of the

immunosensor is the ability pf measuring concentration in the nano-molar range. Therefore, such

an immunosensor can be adapted for other clinically relevant analytes as well.

Detection of Analyte Concentration

Detection of analyte concentration is a major area of research in biosensor applications.

Determination of concentration of cells like white blood corpuscles or red blood corpuscles in

serum or detection of living and dead cells in a suspension, determination of glucose

concentration etc., are few major applications of detection of analyte concentration in a

suspension. In this thesis as one of the applications of the high-frequency biosensor, detection of

concentration of various anlyte in bio-suspension was focused on. One of the main goals in this

part the thesis was to integrate microfluidic system with the BiCMOS sensor chip. Two possible

approaches for the microfluidic integration were explored. One was to construct a non-

conducting wall around the sensor. This approach was relatively straightforward and required no

additional post-processing step. The second approach was based on the bonding of polymer

(PDMS) based microfluidic systems with the sensor chip. The PDMS microfluidic system was

fabricated using a soft lithography technique. For the bonding of the microfluidic system to the

sensor chip an additional post-processing step of chemical mechanical polishing of the BiCMOS

chip was conducted in order to meet the planarity requirements for the accurate bonding of the

microfluidic system and the BiCMOS chip.

Various analyte concentrations in suspensions were detected using various sensor systems. The

sensors operating at 6 GHz and 12 GHz were used as glucose concentrations. The sensitivity of

the sensors when fabricated on TM1 and TM2 of the BiCMOS stack was compared. The sensor

fabricated on TM1 metal layer (sensor operating at 6 GHz) demonstrated considerable sensitivity

and could distinguish 10 % increase in the concentration of glucose in the suspension. This

matches with simulation showing the influence of passivation layer on the sensitivity of the

system. It should be mentioned here that the sensor on TM1 has two passivation layers (SiO2 and

Si3N4) between its surface and the glucose solution.

Chapter 6 Conclusion and Outlook

105

The 12 GHz sensor system was further used to demonstrate the capability of detection of

concentration of particles in a suspension. Concentration of micro-beads suspended in a solution

was detected using this sensor system. Therefore, such a sensor architecture can be applied to

various cytometric applications. The theoretical basis of working of the sensor system in

cytomteric application was explored and was shown that the average permittivity of the

suspension is dependent on the orientation polarization of water molecules in the suspension.

The presence of particles influences the orientation of water molecules based on Einstein Stoke’s

equation and therefore, causes a change in the overall permittivity based on their concentration.

This matches with the theory proposed for high frequency sensors, where other dispersion

mechanisms have negligible influence on the sensor output. The 12 GHz sensor system was also

used to establish proof of principle for establishment of minimally invasive plaque sensors. The

sensor was utilized to detect the concentration of fat and calcium in aqueous phase.

Another important aspect of the sensor system investigated in this part of the thesis was the

capability of imaging of biomaterials based on the permittivity distribution. Two sensors

operating at 7 GHz and 8 GHz were investigated for the imaging applications. The sensor

systems were shown to be able to accurately image the spatial distribution of permittivity. In the

work shown, a lateral resolution of the order of µm was shown. Such a sensor can be applied for

near field imaging of cancerous tissues. Thus, high-frequency BiCMOS biosensors with versatile

applications were demonstrated in this section of the thesis.

Towards Particle Counting

Single particle sensing and counting is another significant research avenue in the area of

biosensors. To this aspect, in this thesis a highly sensitive capacitive sensor system was designed

and electrically characterized, depicting the ability to measure capacitance change of the order of

auto Farrads. Such small change in capacitance can be related to extremely low solute or particle

concentration in suspensions or single particle. The sensor system is suited for dynamic fluid

system. Ultra-high-speed measurement capability of the order of 3 µs was demonstrated with the

sensor system.

A dual loop PLL demodulator architecture was used for the development of the high precision

sensor system. The flow of particles on top if the sensor embedded in an oscillator was modeled

as the oscillation frequency modulation of the oscillator. A self-calibrating feature was employed

in the PLL architecture for detection of very small capacitive changes. A theoretical formulation

of the working of the demodulator architecture was established. A modulating capacitor was

incorporated in the design for electrical characterization of the sensor system. A change of 1

MHz at the operating frequency of 14.3 GHz was detected accurately. In terms of capacitance

change this translates to a change of 60 aF at the starting capacitance of 18 fF. The key feature of

Chapter 6 Conclusion and Outlook

106

the sensor system is the DC output. This makes the sensor architecture extremely flexible and

easy to handle.

Noise, which is a significant source of major problems in biosensors, was addressed in this

section of the thesis. Dual demodulator architecture was proposed and the technique of cross-

correlation was employed to remove the non-correlated noise form the system. The sensor

system can be accurately used to count the particles and also extract the flow rate and overall

concentration of the particles.

Outlook

The thesis shows the feasibility of establishing integrated CMOS/ BiCMOS biosensors suitable

for a variety of applications, like immunosensors, glucose sensors, cytomteric sensors and more.

This single chip solution is the first step towards producing cheap and easy to handle POC

diagnostic devices. Most of the sensor systems shown in this thesis have high-frequency outputs.

The next viable step is the complete system integration with digital output. The prior requirement

in order to establish the complete sensor system is the conversion of the high-frequency output to

DC output. One such sensor system with DC output was shown in chapter 5 of the thesis.

However, the system was suited primarily for biosensors assisted with flow based fluidic

systems. For more static applications, methods of conversion of high-frequency sensor signals to

DC values need to be implemented. These methods could involve use of PLL applicable for

static approaches, or implementation of frequency counters. Thus, conversion of the high-

frequency sensor signals to DC output is the most primary and significant step for system

integration. The next step towards system integration with digital outputs is the implementation

of analog to digital converters (ADCs). With digital outputs the sensor systems will become

more user-friendly and easy to handle. Therefore, gradual steps should be taken to establish a

complete biosensor module from the established sensor architectures. Being a single chip

solution, these sensor architectures provide the freedom to develop on chip digital circuitry for

signal processing. The signal processing circuits being very close to the sensor system will also

provide maximum sensitivity. Due to the close proximity no degradation of signal will take place

unlike the hybrid integrated biosensors.

The next enhancement of the sensor systems is the development of parallel sensing schemes. To

this aspect, sensor array was demonstrated in chapter 4 of the thesis work. The rational step

towards establishing parallel sensing schemes is the enhancement of this array into more number

of units or pixels. Standard microtitre plates can be take as reference examples for the

establishment of this parallel sensing approach. However, parallel sensing scheme with multiple

sensors can be a complex system to handle due to routing of a large number of high frequency

signal paths. Considerable amount of research work has to be invested in order to develop the

right strategy for parallel sensing. This kind of sensing approach would also require a suitable

Chapter 6 Conclusion and Outlook

107

automation of the sensor system in order to control the sensor operations, for e.g., with switches.

The digital control of switches will require automated control using for e.g., a microcontroller.

Therefore, automated parallel sensing schemes should be looked at to increase the throughput of

such sensor architectures.

From the biological aspects more applications should be targeted at using the developed

BiCMOS biosensors. Some significant applications like immunosensing, glucose sensing, and

detection of concentration of particles in a suspension have already been addressed in this thesis.

One such example of diversification of application of these sensor systems is the establishment

of an “all-electrical” alternative for the established micro-titre plates used for the growth culture

of cells in an aqueous environment. The sensor systems have already been shown to be able to

detect small variations of concentration of particle s in a suspension. Establishment of such

intelligent micro-titre plates for growth culture would be an extension of the concentration

detection sensor. Other applications can involve diversification of the immunosensor for various

proteins, specific glucose molecule sensor, etc. Establishing spectroscopy technique is another

outlook to the work. If an overall sensor system is designed for the frequency range covering

from MHz to GHz ranges, the same sensor chip can be used for understanding the low frequency

attributes of biological samples as well as their concentration which can be detected better at

high frequency.

All in all, in this thesis a new approach towards biosensors has been demonstrated with the

establishment of complete BiCMOS integrated biosensor platform. This approach has the

potential for establishing new POC devices for diverse applications and at the same time with

extreme ease of handling and reduced cost.

Bibliography

108

BIBLIOGRAPHY

[1] C. D. Chin, V. Linder and S. K. Sia, “Lab-on-a-chip devices for global health: past studies and future

opportunities”, Lab chip, 2006, 7, 41-47

[2] C. Wilmott and J. E. Arrowsmith, “Point-of-care testing”, Surgery, 2010, 28, 159-160

[3] E. Petryayeva and W. Russ Algar, “Toward point-of-care diagnostics with consumer electronic

devices: the expanding role of nanoparticles”, RSC Advances, 2015, 5, 22256-22282

[4] C. D. Chin, V. Linder and S. K. Sia, “Commercialization of microfluidic point-of-care diagnostic

devices”, Lab Chip, 2012, 12, 2118-2134

[5] J. S. Daniels and N. Pourmand, “Label-free impedance biosensors: opportunities and challenges”,

Electroanalysis, 2007, 19, 1239-1257

[6] N. M. M. Pires, T. Dong, U. Hanke and N. Hoivik, “Recent developments in optical detection

technologies in lab-on-a-chip devices for biosensing applications”, Sensors, 2014, 14, 15458-15479

[7] S. Lopez, “Pulse Oximeters: Fundamentals and Design”, NXP, FreeScale Semiconductors Document

[8] M-I Mohammed and M. P. Y. Desmulliez, “Lab-on-a-chip based immunosensor principles and

technologies for the detection of cardiac biomarkers: a review”, Lab Chip, 2011, 11, 569-595

[9] F. Rusling, C.V. Kumar, J. S. Gutkind and V. Patel, “Measurement of biomarker proteins for point-of-

care early detection and monitoring of cancer”, Analyst, 2010, 135, 2496-2511

[10] A. M. Foudeh, T. F. Didar, T. Veres and M. Tabrizian, “Microfluidic designs and techniques using

lab-on-a-chip devices for pathogen detection for point-of-care diagnostics”, Lab Chip, 2012, 12, 3249-

3266

[11] N. Wongkaew, P. He, V. Kurth, W. Surareungchai and A. J. Baemner, “Multi-channel PMMA

microfluidic biosensor with integrated IDUAs for electrochemical detection”, Anal. Bional. Chemistry,

2013, 405, 5965-5974

[12] M. Ferrari, “Cancer nanotechnology”, Nat. Rev. Cancer, 2005, 5, 161-171

[13] S. Pennathur and D. K. Fygenson, “Improving fluorescence detection in lab on chip devices”, Lab

Chip, 2008, 8, 649-652

[14] L. M. Lee, X. Cui and C. Yang, “The application of on-chip optofluidic microscopy for imaging

Giardia lamblia trophozoites and cysts”, Biomedical Microdevices, 2009, 11, 951-958

[15] N. Ramalingam, Z. Rui, H. B. Liu, C. C. Dai, R. Kaushik, B. M. Ratnaharika and H. Q. Gong, “Real-

time PCR-based microfluidic array chip for simultaneous detection of multiple waterborne pathogens”,

Sensors and Actuators B Chemical, 2010, 145, 543-552

[16] R. Ishimatsu, A. Naruse, R. Liu, K. Nakano, M. Yahiro, C. Adachi and T. Imato, “An organic thin

film photodiode as a portable photodetector for the detection of alkylphenol polyethoxylates by a flow

fluorescence-immunoassay on magnetic microbeads in a microchannel”, Talanta, 2013, 117, 139-145

Bibliography

109

[17] N. Yildirim, F. Long, C. Gao, M. He, H.C. Shi and A. Z. Gu, “Aptamer based optical biosensor for

rapid and sensitive detection of 17 β-estradiol in water samples”, Environmental Science and Technology,

2012, 46, 3288-3294

[18] M. Wolf, D. Juncker, B. Michel, P. Hunziker and E. Delamarche, “Simultaneous detection of C-

reactive protein and other cardiac markers in human plasma using micromosaic immunoassays and self-

regulating microfluidic networks”, Biosensors and Bioelectronics, 2004, 19, 1193-1202

[19] J. Ziegler, M. Zimmermann, P. Hunziker and E. Delamarche, “High-performances immunoassays

based on through stencil patterned antibodies and capillary systems” Analytical Chemistry, 2008, 80,

1763-1769

[20] C. Joensson, M. Aronsson, G. Rundstroem, C. petterson, I. Medel-Hartvig, J. Bakker, E. Martinsson,

B. Liedberg, B. MacCraith, O. Oehman and J. Melin, “Silane-dextran chemistry on lateral flow polymer

chips for immunoassays”, Lab Chip, 2008, 8, 1191-1197

[21] X. L. Su and Y. Li, “Quantum dot biolabeling coupled with immunogenetic separation for detection

of Escherichia coli” Analytical chemistry, 2004, 76, 4806-4810

[22] M. Nichkova, D. Dosev, A. E. Davies, S. J. Gee, I. M. Kennedy and B. D. Hammock, “Quantum

dots as reporters in multiplexed immunoassays for biomarkers of exposure to agrochemicals”, Analytical

letters, 2007, 40, 1423-1433

[23] L. J. Chabonniere and N. Hildebrandt, “Lanthanide complexes and quantum dots: a bright wedding

for resonance energy transfer” European journal of Inorganic Chemistry, 2008, 3241-3251

[24] R. C. Stringer, D. Hoehn and S. A. Grant, “ Quantum dot based biosensor for detection of human

cardiac troponin 1 using a liquid core waveguide” IEEE Sensors Journal, 2008, 8, 295-300

[25] L. Gervais and E. Delmarche, “Toward one step point of care immunodiagnostics using capillary-

driven microfluidics and PDMS substrates”, Lab Chip, 2009, 9, 3330-3337

[26] K. E. Sapsford, T. Pons, I. L. Medintz and H. Mattoussi, “Biosensing with luminescent

semiconductor quantum dots”, Sensors, 2006, 6, 925-953

[27] K. Kerman, T. Endo, M. Tsukamoto, M. Chikae, Y. Takamura and E. Tamiya, “Quantum dot-based

immunosenosr for the detection of prostate-specific antigen using fluorescence microscopy”, Talanta,

2007, 71, 1494-1499

[28] M. Zimmermann, P. Hunziker and E. Delmarche, “Autonomous capillary system for one step

immunoassays”, Biomedical Microdevices, 2009, 11, 1-8

[29] A. Xiang, F. Wei, X. Lei, Y. Liu and Y. Guo, “A simple and rapid capillary chemiluminesce

immunoassay for quantitatively detecting human serum HbsAg”, European Journal of Clinical

Microbiology Infect. Dis., 2013, 32, 1557-1564

[30] I. H. Cho, E. H. Paek, Y. K. Kim, J. H. Kim and S. H. Paek, “ Chemiluminometric enzyme-linked

immunosorbent assays (ELISA)-on a chip biosensor based on cross-flow chromatography”, Analytical

Chim. Acta., 2009, 632, 247-255

[31] H.Huang, X. L. Zheng, J. S. Zheng, J. Pan and Y. P. Pu, “Rapid analysis of alpha-fetoprotein by

chemiluminescence microfluidic immunoassay system based on super-paramagnetic microbeads”

Biomedical Microdevices, 2009, 11, 213-216

Bibliography

110

[32] M. Yang, S. Sun, Y. Kostov and A. Rasooly, “An automated point-of-care system for

immunodetection of taphylococcal enterotoxin B”, Analytical Biochemistry, 2011, 416, 74-81

[33] D. Caputo, G. de Cesare, L. S. Dolci, M. Mirasoli, A. Nascetti, A. Roda and R. Scipinotti,

“Microfluidic chip with integrated a-Si.H photodiodes for chemiluminescence-based bioassays”, IEEE

Sensors Journal, 2013, 13, 2595-2602

[34] M. Hao and Z. Ma, “An ultrasensitive chemiluminescence biosensor for carcinoembryonic antigen

based on autocatalytic enlargement of immunogold nanoprobes”, Sensors, 2012, 12, 17320-17329

[35] M. Yang, Y Kostov, H. A. Bruck and A. Rasooly, “Carbon nanotubes with enhanced

chemiluminescence immunoassay for CCD-based detection of staphylococcal enterotoxin B in food”

Analytical Chemistry, 2008, 80, 8532-8537

[36] F. Darain, P. Yager, K. L. Gan and S. C. Tjin, “On chip detection of myoglobin based on

fluorescence”, Biosensors and Bioelectronics, 2009, 24, 1744-1750

[37] A. Bhattacharyya and C. M. Klapperich, “Design and testing of a disposable microfluidic

chemiluminescent immunoassay for disease biomarkers in human serum samples”, Biomedical

Microdevices, 2007, 9, 245-251

[38] D. M. Vykoukal, G. P. Stone, P. R. C. Gascoyne, E. U. Alt, J. Vykoukal, “Quantitative detection of

bioassays with a low-cost image-sensor array for integrated microsystems”, Angew. Chem. Int. Ed., 2009,

121, 7785-7790

[39] C. C. Lin, F. H. Ko, C. C. Chen, Y. S. Yang, F. C. Chang and C. S. Wu, “Miniaturized metal

semiconductor metal photocurrent system for biomolecular sensing via chemiluminescence”,

Electrophoresis, 2009, 30, 3189-3197

[40] J.R. Wojciechowski, L. C. Shiver-lake, M. Y. Yamaguchi, E. Fuereder, R. Pieler, M. Schamesberger,

C. Winder, H. J. Prall, M. Sonnleitner and F. S. Ligler, “Prganic photodiodes for biosensor

miniaturization”, Analytical Chemistry, 2009, 81, 3455-3461

[41] O. Krupin, H. Asiri, C. Wang, R. N. Tait and P. Berini, “Biosensing using straight long-range surface

plasmon waveguides”, Optics Express, 2013, 21, 698-709

[42] A.M Foudeh, J. T. Daoud, S. P. Faucher, T. Veres and M. Tabrizian, “Sub-femtomole detection of of

16s rDNA from legionella pneumophila using surface plasmon resonance imaging”, Biosensors and

Bioelectronics, 2014, 52, 129-135

[43] E. Ouellet, C. Lausted, T. Lin, C.W.T. Yang, L. Hood and E. T. Lagally, “Parallel microfluidic

surface plasmon resonance imaging arrays”, Lab Chip, 2010, 10, 581-588

[44] S. Unser, I. Bruzas, J. He and L. Sagle, “Localized Surface plasmon resonance biosensing: Current

challenges and approaches”, Sensors, 2015, 15, 15684-15716

[45] B. N. Feltis, B. A. Sexton, F. L. Glenn, M. J. Best, M. Wilkins and T. J. Davis, “A hand-held surface

plasmon resonance biosensor for the detection of ricin and other biological agents”, Biosensors

Bioelectronics, 2008, 23, 1131-1136

[46] T. M. Chinowsky, M. S. Grow, K. S. Johnston, K. Nelson, T. Edwards, E. Fu and P Yager,

“Compact high performance surface plasmon resonance imaging systems”, Biosensors Bioelectronics,

2007, 22, 2208-2215

Bibliography

111

[47] T. M. Chinowsky, J. G. Quinn, D. U. Bartholomew, R. Kaiser and J. L. Elkind, “ Performance of the

spreeta 2000 integrated surface plasmon resonance affinity sensor”, Sensors and Actuators B: Chemical,

2003, 91, 266-274

[48] S. Wang, X. Zhao, I. Khimji, R. Akbas, W. Qiu, D. Edwards, D. W. Cramer, B. Ye and U. Demirci,

“Integration of cell phone imaging with microchip ELISA to detect ovarian cancer HE4 biomarker in

urine at the poibt-of-care”, Lab Chip, 2011, 11, 3411-3418

[49] J. C. Jokerst, J. A. Adkins, B. Bisha, M. M. Mentele, L. D. Goodridge and C. S. Henry,

“development of a paper-based analytical device for calorimetric detection of selsect foodborne

pathogens”, Analytical Chemistry, 2012, 84, 2900-2907

[50] A. Rasooly, H. A. Bruck and Y. Kostov, “An ELISA lab-on-a-chip (ELISA-LOC)”, Methods

Molecular Biology, 2013, 949, 451-471

[51] F. Zhou, M. Wang, L. Yuan, Z. Cheng, Z. Wu and H. Chen,“Sensitive sandwich ELISA based on a

gold nanoparticle layer for cancer detection”, Analyst, 2012, 137, 1779-1784

[52] F. Li, L. Jiang, J. Han, Q. Liu, Y. Dong, Y. Li and Q. Wei, “A label-free amperometric

immunosensor based for the detection of carcinoembryonic antigen based on novel magnetic carbon and

gold nanocomposites”, RSC Advances, 2015, 5, 19961-19969

[53] C. Chen, G. Xie, D. Yang, H. Xiao, Y. Fu, Y. Tan and S. Yao, “ Recent advances in electrochemical

glucose biosenosrs:a review”, RSC Advances, 2013, 3, 4473-4491

[54] J. Hajdukiewicz, S. Boland, P. Kavanagh, A. Nowicka, Z. Stojek and D. Leech, “Enzyme-amplified

amperometric detection of DNA using redox mediating films on gold microelectrodes”, Electroanalysis,

2009, 21, 342-350

[55] A. Ramanaviciene and A. Ramanavicius, “Pulsed amperometric detection of DNA with an

ssDNA/polypyrrole-modified electrode”, Analytical Bioanalytical Chemistry, 2004, 379, 287-293

[56] X. Zhu, K. Ino, Z. Lin, H. Shiku, G. Chen and T. Matsue, “Amperometric detection of DNA

hybridization using multi-point, addressable electrochemical device”, Sensors and Actuators B:

Chemical, 2011, 160, 923-928

[57] N.J. Forrow, G. S. Sanghera, S. J. Walters and J. L. Watkin, “Development of a commercial

amperometric biosensor electrode for the ketone β-3-hydroxybutyrate”, Biosensors Bioelectronics, 2005,

20, 1617-1625

[58] J. Wang, “Electrochemical glucose biosensors”, Chemical Reviews, 2008, 108, 814-825

[59] S. Kang, A. Nieuwenhuis, K. Mathwig, D. Mampallil and S. G. Lemay, “ Electrochemical single-

molecule detection in aqueous solution using self-aligned nanogap transducers”, 2013, ACS Nano, 2013,

7, 10931-10937

[60] S. Ko, B. Kim, S. S. Jo, S. Y. Oh and J. K. Park “Electrochemical detection of cardiac troponin I

using a microchip with the surface functionalized polydimethylsiloxane channel”, Biosensors

Bioelectronics, 2007, 23, 51-59

[61] D. Purvis, O. Leonardova. D. Farmakovsky and V. Cherkasov, “An ultrasensitive and stable

potentiometric immune sensor”, Biosensors Bioelectronics, 2003, 18, 1385-1390

Bibliography

112

[62] A.G. Gehring, D. L. Patterson and S. I. Tu, “Use of a light-addressable potentiometric sensor for the

detection of Escherichia coli O157”, Analytical Biochemistry, 1998, 258, 293-298

[63] K. Chumbimuni-Torres, J. Wu, C. Clawson, M. Galik, A. Walter, G. Flechsig, E. Bakker, L. Zhang

and J. Wang, “Amplified potentiometric transduction of DNA hybridization using ion-loaded liposomes”,

Analyst, 2010, 135, 1618-1623

[64] E. Ogabiela and S. B. Adeloju, “A potentiometric phosphate biosensor based on entrapment of

pyruvate oxidase in a polypyrrole film”, Analytical Methods, 2014, 6, 5290-5297

[65] J. G. Ayenimo and S. B. Adeloju, “Improved potentiometric glucose detection with ultra-thin

polypyrrole-glucose oxidase films”, Analytical Methods, 2014, 6, 8996-9006

[66] S. Viswanathan, H. Radecka and J. Radecki, “Electrochemical biosensors for food analysis”,

Monatsh. Chem., 2009, 140, 891-899

[67] Z. Shi, Y. Tian, X. Wu, C. Li and L. Yu, “A one-piece lateral flow impedimetric test strip for label-

free clenbuterol detection”, Analytical Methods, 2015, 7, 4957´4964

[68] A. H. Loo, A. Bonanni and M. Pumera, “Impedimetric thrombin aptasensor based on chemically

modified graphenes”, Nanoscale, 2012, 4, 143-147

[69] X. Munoz-Berbel, N. Vigues, M. Cortina-Puig, R. Escude, C. Garcia-Aljaro, J. Mas and F. X.

Munoz, “Impedimetric approach for monitoring bacterial cultures based on the changes in the magnitude

of the interface capacitance”, Analytical Methods, 2010, 2, 1036-1042

[70] Abott Point of care, USA

[71] L. Rassaei, E. D. Goluch and S. G. Lemay, “Substrate dependent kinetics in tyrosinase-based

biosensing: Amperometry vs Spectrophotometry”, Analytical and Bioanalytical Chemistry, 2012, 493,

1577-1584

[72] S. Goh, And R. J. Ram, “Impedance spectroscopy for in situ biomass measurements in

microbioreactor”, 14th International Conference on Miniaturized Systems for Chemistry and Life Science,

Netherlands, 2010.

[73] E. E. Kommenhoek et al, “Monitoring of yeast cell concentration using micromachined impedance

sensor”, Sensors and Actuators B: Chemical, 2006, 115, 384-389

[74] A. Faenza, M. Bocchi, N. Pecorari, E. Franchi and R. Guerrieri, “Impedance measurement technique

for high sensitivity cell detection in micro structures with non-uniform conductivity distribution”, Lab

Chip, 2012, 12, 2046-2052

[75] Micronit microtechnologies, Electrochemical Impedance spectroscopy chips (EIS Chips)

[76] Gamry Instruments, Electrochemical Impedance spectroscopy

[77] M. Nikolic-Jaric, S. F. Romanuik, G. A. Ferrier, G. E. Bridges, M. Butler, K. Sunley, D. J. Thomson

and M. R. Freeman, “Microwave frequency sensor for detection of biological cells in microfluidic

channels”, Biomicrofluidics, 2009, 3, 1-15

[78] G. E. Moore, “Cramming more components onto integrated circuits”, Electronics, 38, 1965

Bibliography

113

[79] W. Arden, M. Brillouet, P. Cogez, M. Graef, B. Huizing and R. Mahnkopf, “More-Than-Moore”:

White Paper

[80] K. H. Lee, J. O. Lee, S. Choi, J. B. Yoon and G. H. Cho, “A CMOS label-free DNA sensor using

electrostatic induction of molecular charges”, Biosensors Bioelectronics, 2012, 31, 343-348

[81] C. Stagni, C. Guiducci, L. Benini, B. Ricco, S. Carrara, B. Samori, C. Paulus, M. Schienle, M.

Augustyniak and R. Thewes, “ CMOS DNA sensor array with integrated A/D conversion based on label-

free capacitance measurement”, IEEE Journal of Solid-State Circuits, 2006, 41, 2956-2964

[82] E. Anderson, J. Daniels, H. Yu, T. Lee and N. Pourmand, “ A label-free CMOS DNA microarray

based on charge sensing”, IEEE Instrumentation and Measurement Technology Conference, Canada,

2008

[83] A. Hassibi and T. H. Lee, “A programmable 0.18 µm CMOS electrochemical sensor microarray for

biomolecular detection”, IEEE Sensors Journal, 2006, 6, 1380-1388

[84] K. Lee, J. Nam, S. Choi, H. Lim, S. Shin and G. Cho, “A CMOS impedance cytometer for 3D

flowing single-cell real time analysis with ΔΣ error correction” IEEE, International Solid-State Circuits

Conference Digest of Technical Papers (ISSCC), USA, 2012, 304-306

[85] Y. Adiguzel and H. Kulah, “CMOS cell sensors for point-of-care diagnostics”, Sensors, 2012, 12,

10042-10066

[86] E. Ghafar-Zadeh and M. Sawan, “CMOS based capacitive detection lab-on-chip: design and

experimental results”, IEEE Custom Integrated Circuits Conference (CICC), USA, 2007

[87] S.S. Stuchly, C. E. Bassey, “Microwave coplanar sensors for dielectric measurements”,

Measurement Science technology, 1998, 9, 1324-1329

[88] A. I. Gubin, AA. Barannik, N. T. Cherpak, S. Vitusevich, A. Offenhaeusser and N. Klein,

“Whispering-gallery mode resonator technique for charactrization of small volumes of biochemical

liquids in microfluidic channel”, IEEE European Microwave Conference Proceedings, UK, 2011, 615-

618

[89] D. J. Rowe, A. Porch, D. A. Barrow, C. J. Allender, “Microfluidic device for compositional analysis

of solvent systems at microwave frequencies”, Sensors and Actuators B: Chemical, 2012, 169, 213-221

[90] J.C. Booth, J. Mateu, M. Janezic, J. Baker-Jarvis and J. A. Beall, “Broadband permittivity

measurements of liquid and biological samples using microfluidic channels”, IEEE/MTT-S, International

Microwave Symposium, USA, 2006, 1750-1753

[91] K. Grenier and D. Dubuc, “Integrated broadband microwave and microfluidic sensor dedicated to

bioengineering”, IEEE trans. on Microwave Theory and Techniques, 2009, 57, 3247-3253

[92] K. Grenier, D. Dubuc, T. Chen, F. Artis, T. Chretiennot, M. Poupot and J-J. Fournie, “Recent

advances in microwave-based dielectric spectroscopy at the cellular level for cancer investigation”, IEEE

Trans. on Microwave Theory and Techniques, 2013, 61, 2023-2030

[93] M. Nikolic-Jaric, S. F. Romanuik, G. A. Ferrier, G. E. Bridges, M. Butler, K. Sunley, D. J. Thomson

and M. R. Freeman, “Microwave frequency sensor for detection of biological cells in microfluidic

channels”, Biomicrofluidics, 2009, 3, 1-15

Bibliography

114

[94] Y. Yang, H. Zhang, J. Zhu, G. Wang, T.-R. Tzeng, X. Xuan, K. Huang and P. Wang,

“Distinguishing the viability of a single yeast cell with an ultra-sensitive radio frequency sensor“, Lab

Chip, 2010, 10, 553-555

[95] J. Chien, M. Anwar, E. Yeh, L. P. Lee and A. M. Niknejad, “A 6.5/17.5 GHz dual-channel

interferometer-based capacitive sensor in 65 nm CMOS for high-speed flow cytometry“, IEEE,/MTT-S

International Microwave Symposium, USA, 2014

[96] H. Wang, C. Sideris and A. Hajimiri, “A frequency-shift based CMOS magnetic biosensor with

spatially uniform sensor transducer gain”, IEEE Custom Integrated Circuits Conference (CICC), USA,

2010

[97] A. Manz, “Miniaturized total chemical-analysis systems: a novel concept for chemical sensing”,

Sensors and Actuators B: Chemical, 1990, 1, 244-248

[98] M. Kaynak, M. Wietstruck, C. Kaynak, S. Marschmeyer, P. Kulse, K. Schulz, H. Silz, A. Krueger,

R. Barth, K. Schmalz, G. Gastrock and B. Tillack, “BiCMOS integrated microfluidic platform for bio-

MEMS applications”, IEEE/MTT-S, International Microwave Symposium, USA, 2014

[99] C. B. Kayanak, M. Kaynak, M. Wietstruck, St. Marschmeyer, P. Kulse, K. Schulz, H. Silz, a.

Krüger, R. Barth, K. Schmalz and B. Tillack, “Modeling and characterization of BiCMOS embedded

microfluidic platform for biosensing applications”, Silicon Monolithic Integrated Circuits in RF Systems

(SIRF – Biowireless 14), USA, 2014

[100] S. K. Sia and G. M. Whitesides, “Microfluidic devices fabricated in poly (dimethylsiloxane) for

biological studies”, Electrophoresis, 2003, 24, 3563-3576

[101] G.M. Whitesides “Overview the origins and the future of microfluidics”, Nature, 2006, 442, 368-

373

[102] B. Eversmann, M. Jenkner, F. Hofmann, C. Paulus, R. Brederlow, B. Holzapfl, P. Fromherz, M.

Merz, M. Brenner, M. Schreiter, R. Gabl, K. Plehnert, M. Steinhauser, G. Eckstein, D. Schmitt-landsiedel

and R. Thewes,“ A 128 x 128 CMOS biosensor array for extracellular recording of neural activity”, IEEE

Journal of Solid-State Circuits, 2003, 38, 2306-2317

[103] L. Lin, L. Xiaowen and A-J. Mason, “Die-level photolithography and etchless parylene packaging

process for on-CMOS electrochemical biosensors”, IEEE International Symposium on Circuits and

Systems (ISCAS), 2012, Korea

[104] B. Heinemann, H. Ruecker, R. Barth, J. Drews, O. Fursenko, T. Grabolla, R. Kurps, St.

Marschmeyer, A. Scheit, D. Schmidt, A. Trusch, D. Wolansky and Y. Yamamoto, “ SiGe p-n-p HBTs

with 265 GHz fmax, 175 GHz ft and 3.65-ps gate delay”, IEEE Electron Device Letters, 2014, 35, 814

[105] LM. Silva, AM. Salgado and MA. Coelho, “Development of an amperometric biosensor for phenol

detection”, Environmental Technology, 2011, 32, 493-497

[106] T. H. Lee, “The design of CMOS radio-frequency integrated circuits”, 2nd Edition, Cambridge

University press, 2003

[107] C. Kittel, “Introduction to solid state physics”, 8th Edition, 2004

[108] H. P Schwan, “Electrical properties of tissue and cell suspensions”, Advances in Biological and

Medical Physics, 1957

Bibliography

115

[109] R. Pethig and D. B. Kell, “The passive electrical properties of biological systems: their significance

in physiology, biophysics and biotechnology”, Phys. Med. Biology, 1987, 32, 933-970

[110] B. Razavi, “RF Microelectronics”, 2nd Edition, Prentice Hall Communication, 2011

[111] C. N. Chang, Y. C. Wong, and C. H. Chen, “Full-wave analysis of coplanar waveguides by

variational conformal mapping technique”, IEEE Trans. Microwave Theory and Techniques, 1990, 38,

1339-1344

[112] G R. Garg, “Analytical and Computational Methods in Electromagnetics”, Artech House

[111] K. Sundara-Rajan, A. V. Mamishev and M. Zahn, “Fringing electric and magnetic field sensors”,

Encyclopedia of Sensors, 2006, 1-12

[112] G. D. Alley, “Interdigital capacitors and their application to lumped-element microwave integrated

circuits”, IEEE Transactions Microwave Theory and Techniques, 1970, 18, 1028-1033

[113] JS. Wei, “Distributed capacitance of planar electrodes in optic and acoustic surface wave devices”,

IEEE Journal of Quantum Electronics, 1977, 13, 152-158

[114] R. Igreja and C. J. Dias, “Analytical evaluation of the interdigital electrodes capacitance for a

multilayered structure”, Sensors and Actuators A: Physical, 2004, 112, 291-301

[115] A. Warsinke, “Point-of-care testing of proteins”, Analytical Bioanalytical Chemistry, 2009, 5 1393-

1405

[116] S. Feng, R. Caire, B. Cortazar, M. Turan, A. Wong and A. Oczan, “ Immunochromatographic

diagnostic test analysis using google glass”, ACS Nano, 2014, 8, 3069-3079

[117] P. Sithigorngul, S. Rukpratanporn, N. Pecharaburanin, P. Suksawat, S. Longyant, P.

Chaivisuthangkura and W. A. Sithigorngul, “A simple and rapid immunochromatographic test strip for

detection of pathogenic isolates of Vibrio harveyi”, Journal of Microbiological Methods, 2007, 71, 256-

264

[118] T. Peng, W. Yang, W. Lai, Y. Xiong, H. Wei and J. Zhang, “Improvement of the stability of

immunochromatographic assay for the quantitative detection of clenbuterol in swine urine”, Analytical

Methods, 2014, 6, 7394-7398

[119] F. F. Bier and S. Schumacher, “Integration in bioanalyis: technologies for Point-of-Care testing“,

Adv. Biochem. Eng. Biotechnol. 2013, 133, 1-14

[120] F. Ricci, G. Adornetto and G. Palleschi, “A review of experimental aspects of electrochemical

immunosensors”, Electrochimica Acta, 2012, 84, 74-83

[121] A. Warsinke, “Electrochemical biochips for protein analysis”, Adv. Biochem. Engin/Biotechnol.,

2008, 109, 155-193

[122] G.Gauglitz, “Direct optical detection in bioanalysis: an update”, Analytical Bionalaytical

Chemistry, 2010, 398, 2363-2372

[123] Y. Wan, Y. Su, X. Zhu, G. Liu and C. Fan, “Development of electrochemical immunosenosrs

towards point of care diagnostics”, Biosensors Bioelectronics, 2013, 47, 1-11

Bibliography

116

[124] V. Mani, B. Chikkaveeraiah, V. Patel, J. S. Gutkind and J. F. Rusling, “Ultrasensitive

immunosensor for cancer biomarker proteins using gold nanoparticle film electrodes and multienzyme-

particle amplification”, ACS Nano, 2009, 3, 585-594

[125] C. Gabriel, S. Gabriel and E. Corthout, “The dielectric properties if biological tissues: Literature

survey”, Physical Medical Biology, 1996, 41, 2231-2249

[126] T. Hanai, N. Koizumi and A. Irimajiri, “A method for determining the dielectric constant and the

conductivity of membrane-bounded particles of biological relevance”, Biophys. Struct. Mech., 1975, 1,

285-294

[127] D. Johnson, in Clinical Chemistry (Ed. E. H. Taylor), Wiley, 1989, 55-82

[128] M. Z. Jaffe, “Uber den Niederschlag, welchen pikrinsäure in normalem Harn erzeugt und über eine

neue Reaction des Kreatinins”, Zeitschrift für Physiologische Chemie, 10 (1886)

[129] P. Fossati, L. Prencipe and G. Berti, “Enzymic creatinine assay: A new calorimetric method based

on hydrogen peroxide measurement” Clin. Chem., 1983, 29, 1494-1496

[130] A. Benkert, F. Scheller, W. Schoessler, C. Hentschel, B. Micheel, O. Behrsing, G. Scharte, W.

Stoecklein and A. Warsinke, “Development of a creatinine ELISA and an amperometric antibody-based

creatinine sensor with a detection limit in the nanomolar range", Analytical Chemistry, 2000, 72, 916-921

[131] L. Li, C. Li, Z. Zhang and E. Alexov, “On the dielectric constant of proteins: smooth dielectric

function for macromolecular modeling and its implementation in Delphi”, Journal of Chem. Theor.

Comput., 2013, 9, 2126-2136

[132] P. Kukic, D. Farrell, L. P. MacIntosh, B. Garcia-Moreno, E. K. S. Jensen, Z. Toleikis, K. Teilum

and J. E. Nielsen, “Protein dielectric constants determined from NMR chemical shift perturbations” J.

Am. Chem. Soc., 2013, 135(45), 16968–16976.

[133] C. Spegel, A. Heiskanen, L. H. D. Skjolding and J. Emneus, “Chip based electroanalytical systems

for cell analysis”, Electroanalysis, 2008, 20, 680-702

[134] A. Han and A. B. Frazier, “Ion channel characterization using single cell impedance spectroscopy”,

Lab Chip, 2006, 6, 1412-1414

[135] K.-H. Han, A. Han and A. B. Frazier, “Microsystems for isolation and electrophysiological analysis

of breast cancer cells from blood”, Biosensors Bioelectronics, 2006, 21, 1907-1914

[136] K. Cheung, S. Gawad and P. Renaud, “Impedance spectroscopy flow cytometry on chip label-free

cell differentiation” Cytometry, Part A, 2005, 65, 124-132

[137] D. D. Carlo and L. P. Lee, “Dynamic single-cell analysis for quantitative biology” Analytical

Chemistry, 2006, 78, 7918-7925

[138] M. Lisker, A. Trusch, A. Krueger, M. Fraschke, P. Kulse, St. Marschmeyer, J. Schmidt, Ch.

Meliani, B. Tillack, T. Kraemer, N. Weimann, I. Ostermay, O. Krueger, T. Jensen, T. Al-Sawaf, V.

Krozer and W. Heinrich, “ Combining SiGe BiCMOS and InP processing in an on-top chip integration

approach”, ECS Transactions, 2014, 64, 177-194

[139] C. McDonald and G. M. Whitesides, “Polydimethylsiloxane as a material for fabricating

microfluidic devices”, Accounts of Chemical Research, 2002, 35, 491-499

Bibliography

117

[140] M. D. Belrhiti, S. Bri, A. Nakheli, M. Haddad and A. Mamouni, “Dielectric constant determination

of liquid using regular waveguide structure combined with EM simulation”, Journal of Mater. Environ.

Sci., 2012, 3, 575-584

[141] W. Kuang and S. O. Nelson, “Dielectric relaxation characterization of fresh fruits and vegetables

from 3 to 20 GHz”, Journal of Microwave power and electromagnetic energy, 1997, 32, 114-122

[142] V. V. Meriakri, E. E. Chigrai, D. Kim, I. P. Nikitin, L. I. Pangonis, M. P. Parkhomenko and J. H.

Won, “Dielectric properties of glucose solutions in the milimitre-wave range and control of glucose

content in blood”, IOP Measurement Science and Technology, 2006, 18, 1-6

[143] T. H. Basey-Fisher, S. M. Hanham, S. A. Maier, M. M. Stevens, N. M. Alford and N. Klein,

“Microwave relaxation analysis of dissolved proteins: towards free solution biosensing”, Applied Physics

Letters, 2011, 99, 233703

[144] M. Y. Koledintseva, R. E. DuBroff and R. W. Schwartz, “A Maxwell-Garnett model for dielectric

mixtures containing conducting particles at optical frequency”, Progress in Electromagnetics Research,

Pier, 2006, 63, 223-242

[145] K. K. Kaerkkaeinen, A. H. Sihvola and K. I. Nikoskinen, “Effective permittivity of mixtures:

Numerical validation by the FDTD method”, IEEE transactions on Geoscience and Remote Sensing,

2000, 38, 1303-1308

[146] P. L. Weissberg, “Atherogenesis: Current Understanding of the Causes of Atheroma,” Heart, 2000,

83, 247-252

[147] R. Ross, "The Pathogenesis of Atherosclerosis: A Perspective for the 1990s". Nature, 1993, 362,

801-809

[148] B. N. Potkin, et al., “Coronary Artery Imaging with Intravascular High-frequency Ultrasound”,

Circulation, 1990, 81, 1575-1585

[150] D. Huang, et al., “Optical Coherence Tomography,” Science, 1991, 254, 1178-1180[

[151] N. Haandbaek, O. With, S. C. Buergel, F. Heer and A. Hierlemann, “Microfluidic sensor using

resonance frequency modulation for characterization of single cells”, 17th International Conference on

Miniaturized Systems for Chemistry and Life Science, Germany, 2013

[152] H. Lee, Y. Liu, R. M. Westervelt and D. Ham, “IC/microfluidic hybrid system for magnetic

manipulation of biological cells”, IEEE Journal of Solid State Circuits, 2006, 41, 1471-1480

[153] K. Hu, S. A. Osmany, J. C. Scheytt and F. Herzel, “An integrated 10 GHz low-noise phased lock

loop with improved PVT tolerance”, Analog Integrated Circuits and Signal Processing, 2011, 67, 319-

330

[154] F. Herzel, S.A. Osmany and J. C. Scheytt, “Analytical phase noise modeling and charge pump

optimization for fractional-PLLs”, IEEE Transactions on Circuits and Systems I, Regular papers, 2010,

57, 1914-1924

[155] F. Herzel, G. Fischer and H. Gustat, “An integrated CMOS RF Synthesizer for 802.11a Wireless

LAN”, IEEE Journal of Solid-State Circuits, 2003, 38(10), 1767-1770

[156] G. Konvalina and H. Haick, “Sensors for breath testing: from nanomaterials to comprehensive

disease detection”, Acc. Chem. Res., 2014, 47, 66-76

Bibliography

118

[157] H. Haick, Y. Broza, P. Mochalski, V. Ruzsanyi and A. Amann, “Assessment origin and

implementation of breath volatile cancer markers”, Chem. Soc. Rev. 2014, 43, 1423-1449

[158] F. Herzel and B. Razavi, “A study of oscillator jitter due to supply and substrate noise”, IEEE

Transactions on Circuits and Systems II: Analog Digital Signal Processing, 1999, 46, 56-62

[159] S.A. Osmany, F. Herzel, J. C. Scheytt, “Analysis and minimization of substrate spurs in fractional-

N frequency synthesizers”, Analog Integrated Circuits and Signal Processing, 2013, 74, 545-556

[160] H. Wang, C. C. Weng and A. Hajimiri, “Phase noise and fundamental sensitivity of oscillator-based

reactance sensors”, IEEE Transactions on Microwave Theory and Techniques, 2013, 61, 2215-2229

List of Related Publications

119

LIST of RELATED PUBLICATIONS

Conference Proceedings

[1] S. Guha et al, “Integrated high-frequency sensors in catheters for minimally invasive plaque

characterization”, European Microelectronics and Packaging Conference and Exhibition, September

2015, Friedrichshafen, Germany

[2] S. Guha et al., “12 GHz CMOS MEMS lab-on-chip system for detection of concentration of

suspended particles in bio-suspensions”, Biodevices, January 2015, Lisbon, Portugal

[3] S. Guha et al., “An 8 GHz CMOS near field bio-sensor array for imaging spatial permittivity

distribution”, IEEE-MTT-S International Microwave Symposium, May 2014, Tampa, USA

[4] S. Guha et al., “CMOS lab on chip device for dielectric characterization of cell suspensions based on a

6 GHz Oscillator”, IEEE European Microwave Conference, September 2013, Nuremberg, Germany

[5] (Invited) S. Guha et al., “High frequency biomedical sensors integrated with polymer microfluidic

systems”, Biomedical Workshops, IEEE European Microwave Conference, September 2013, Nuremberg,

Germany

[6] S. Guha et al, “CMOS based sensor for dielectric spectroscopy of biological cell suspensions”,

International Conference on Electrical Bioelectrical Impedance (ICEBI), May 2013, Heiligensatdt,

Germany

[7] S. Guha et al, “CMOS MEMS microfluidic systems for cytometry at 5 GHz”, International

Conference on Microfluidic Handling Systems (MFHS), October 2012, Enschede, Netherlands

[8] F.I. Jamal, S. Guha and C. Meliani, “A SiGe BiCMOS dielectric sensor utilizing an open-ended

microstrip line in a 28 GHz Colpitt’s Oscillator”, IEEE European Microwave Conference,September

2014, Rome, Italy

[9] S. Vehring, S. Guha, F. I. Jamal and C. Meliani, “Permittivity sensor based on 60 GHz patch

antenna“, German Microwave Conference (GEMIC), April 2015, Nuremberg, Germany

[10] F. I. Jamal, S. Guha, S. Vehring, D. Kissinger and C. Meliani, “K-Band BiCMOS based near field

biomedical dielectric sensor for detection of fat and calcium in blood”, IEEE European Microwave

Conference, September 2015, Paris, France

[11] C. Meliani, S. Guha, F. I. Jamal, M. Eissa and S. Vehring, “Integrated mm-wave near-field sensor

concepts for biomaterial characterization”, European Microwave Week, September 2015, Paris, France

List of Related Publications

120

[12] F. I. Jamal, S. Guha, M.H. Eissa, J. Boerngaber, D. Kissinger and J. Wessel, “Comparison of

Microstrip Stub Resonators for Dielectric Sensing in Low-Power K-band VCO”, IEEE Topical

Conference, Biowireless, Radio Wireless Week, January 2016, Austin, Texas, USA.

[13] F.I. Jamal, S. Guha, M.H. Eissa, Ch. Meliani, D. Kissinger, J. Wessel “A 24 GHz Dielectric Sensor

Based on Distributed Architecture”, Proc. German Microwave Conference (GeMiC), 2016, 173

[14] M.H. Eissa, F.I. Jamal, S. Guha, Ch. Meliani, D. Kissinger, J. Wessel “Low-Power Planar Complex

Dielectric Sensor with DC Readout Circuit in a BiCMOS Technology”, IEEE MTT-S International

Microwave Symposium (IMS), 2016, San Francisco, USA

[15] F.I. Jamal, S. Guha, M.H. Eissa, Ch. Meliani, H.J. Ng, D. Kissinger, J. Wessel “A Fully Integrated

Low-Power K-Band Chem-Bio-Sensor with On-Chip DC Read-out in SiGe BiCMOS Technology”,

European Microwave Conference (EuMC) 2016, London, UK

[16] D. Wagner, F.I. Jamal, S. Guha, Ch. Wenger, J. Wessel, D. Kissinger, D. Ernst, K. Pitschmann, B.

Schmidt, M. Detert, “Packaging of a BiCMOS Sensor on a Catheter Tip for the Characterisation of

Atherosclerotic Plaque”, 6th Electronics System-Integration Technology Conference (ESTC), 2016,

Grenoble, France

Journal Articles

[1] S. Guha et al, “Label free sensing of creatinine using a 6 GHz CMOS near-field dielectric

immunosensor”, Analyst, 2015, 140, 3019-3027

[2] S. Guha et al., “Self-calibrating highly sensitive dynamic capacitance sensor towards rapid sensing

and counting of particles in laminar flow systems”, Analyst, 2015, 140, 3262-3272

[3] S. Guha et al, “CMOS based sensor for dielectric spectroscopy of biological cell suspension”, IOP

Journal of Physics, 2013, 434

[4] F.I. Jamal, S. Guha, M.H. Eissa, J. Borngräber, Ch. Meliani, H.J. Ng, D. Kissinger, J. Wessel, “Low-

Power Miniature K-Band Sensors for Dielectric Characterization of Bio-Materials”, IEEE Transactions

on Microwave Theory and Technique, 2017, 65 (3), 1012-1023

Patents Filed

[1] “Inhomogene Übertragungsleitung zur positionsaufgelösten Permittivitätsbestimmung”,

IHP.413.PCT-Anmeldung, am 29.01.206, AZ: PCT/EP2016/051887 (European Patent)

[2] “Schaltbarer Messträger zur positionsaufgelösten Permittivitätsbestimmun”

IHP.408.14, DE-Patentanmeldung am 02.02.2015, AZ: 10 2015 201 771.0 (German Patent)

List of Figures

121

LIST of FIGURES

Figure 1.1 Steps for typical clinical diagnostic. The testing and results delivery

steps are highly time consuming…………………………………………………………………………01

Figure 1.2 Steps for point of care diagnostics. The time constraints are

considerably reduced…………………………………………………………………………………………02

Figure 1.3 Schematic of optical immunosensors based on fluorescence

detection……………………………………………………………………………………………………………03

Figure 1.4 Schematic of optical immunosensors based on surface plasmon

resonance…………………………………………………………………………………………………………..05

Figure 1.5 Schematic of optical immunosensors based on amperometric detection

technique…………………………………………………………………………………………………………..06

Figure 1.6 Technology road-map showing More Than Moore concept..……………..09

Figure 1.7 CMOS single chip biosensor approach……………………………………………….10

Figure 1.8 Passive microwave sensors for detection of concentration of

cells…………………………………………………………………………………………………………………….10

Figure 1.9 Interferrometric approach based on passive structures for

biosensors………………………………………………………………………………………………………….11

Figure 1.10 Sensor architecture used in this thesis. Capacitive sensor is embedded

in a CMOS oscillator circuit…………………………………………………………………………………12

Figure 2.1 . Cross section schematic of BiCMOS Back-end-of-line stacks of IHP. a)

SG25H1 BiCMOS process with five metal layers. b) SG13S BiCMOS process with

seven metal layers………………………………………………………………………………………………18

List of Figures

122

Figure 2.2 New sensor approach a) Previous generation biosensors with hybrid

integration scheme b) Integrated CMOS sensor scheme……………………………………19

Figure 2.3 LC resonant tank circuit with the capacitor acting as the sensor b)

quality factor of the LC tank circuit with and without loss………………………………….20

Figure 2.4 Damped oscillation due to loss. The losses are incurred due to the

series resistance accompanying the inductors as well as the imaginary part of the

permittivity of the biomaterial……………………………………………………………………………21

Figure 2.5 Negative resistance compensation done by the active circuit in order to

sustain the oscillation of the resonance tank……………………………………………………..22

Figure 2.6 Typical layout of a sensor system. The distance of the inductors from

the sensors is 500 µm. This evades the interaction of biomaterials with the

inductors…………………………………………………………………………………………………………….23

Figure 2.7 Coplanar transmission line sandwiched between two

dielectrics.………………………………………………………………………………………………………….24

Figure 2.8 Partial capacitances of the coplanar transmission line showing the

contribution of the individual dielectric layers …………………………………………………..25

Figure 2.9 Geometry of the Interdigitated capacitor showing the spacing and

width of the fingers…………………………………………………………………………………………….26

Figure 2.10 Cross-sectional schematic of the IDC fabricated on TM2 metallization

layer of BEOL stack. The fringing fields penetrate into the biomaterial placed on

top of the passivation layer………………………………………………………………………………..27

Figure 2.11 Model of IDC for evaluation of the capacitance fabricated on TM2 of

the BiCMOS stack. The IDC electrodes have Si3N4 and MUT on top and SiO2 at the

bottom. The equipotential surfaces are marked with vertical dotted lines. The two

sets of electrodes are at the potential V and –V…………………………………………………28

Figure 2.12 The partial capacitance components of the IDC for individual dielectric

layers………………………………………………………………………………………………………………….29

List of Figures

123

Figure 2.13 (a) Self resonance phenomenon of a typical IDC structure. The self-

resonance frequency (SRF) is around 150 GHz (b) Variation of the capacitance of

the IDC with respect to permittivity at the operating frequency range of 6 GHz –

12 GHz………………………………………………………………………………………………………………..30

Figure 2.14 The sensor circuit with IDC sensor embedded in a cross coupled CMOS

oscillator…………………………………………………………………………………………………………….31

Figure 2.15 (a) Model for negative transconductance cross coupled oscillator

without considering dielectric losses. (b) Model for negative transconductance

cross coupled oscillator with considering dielectric losses………………………………….32

Figure 3.1 Schematic of a basic immunosensor. The first step includes

immobilization of antibodies/antigens on a transducer surface followed by the

binding of corresponding antigens/antibodies. A sensing scheme is employed to

detect the antigen-antibody pair. A signal processing front-end circuit converts

the detected signal to electrical signal………………………………………………………………..35

Figure 3.2 IDC sensor on BiCMOS back-end-of-line (a) Schematic of BiCMOS back-

end-of-line stack with seven metallization layers. The top two metallization layers

are thick and are less resistive. (b) Geometrical schematic of the IDC showing the

length, spacing and width of the fingers. (c) Schematic of the immobilized

creatinine on the sensor surface…………………………………………………………………………38

Figure 3.3 Scanning electron microscopy (SEM) image of the sensor chip showing

the inductors on topmost metal layer. A focused ion beam (FIB) cutting is

performed to expose the IDC sensor surface………………………………………………………39

Figure 3.4 Sensor operation of the dielectric immunosensor. Creatinine molecules

had been immobilized on the passivated surface of the sensor. Anti-creatinine

antibodies were incubated in four different concentrations of creatinine

molecules (pre-treatment phase). The four different antibody solutions are

allowed to bind to the immobilized creatinine molecules. Antibody samples

incubated with higher concentration of creatinine have less free antibodies left to

bind to immobilized creatinine molecules……………………….....................................40

List of Figures

124

Figure 3.5 Chemical structure of creatinine and Chemical structure of crea-BSA

molecule…………………………………………………………………………………………………………….42

Figure 3.6 Typical variation of IDC capacitance as a function of concentration of

creatinine molecules used in the incubation of antibodies. The capacitance

increases with increase in creatinine molecule concentration used during

incubation. The capacitance variation is strong when the surrounding medium for

experiment is water while with air the variation is negligible……………………………..44

Figure 3.7 Chip photograph of Si3N4/Si test chip for optical measurement. Chip

size is 1 cm x 1 cm………………………………………………………………………………………………45

Figure 3.8 Indirect competitive assay principle for optical creatinine determination

with creatinine-modifies Si3N4 test chips…………………………………………………………….46

Figure 3.9 Optical measurement of creatinine concentration. The response slope

of the optical measurement in the range 0.88 to 88 µM shows the dynamic range

of the standard measurement technique……………………………………………………………46

Figure 3.10 Chip photograph of dielectric sensor. The sensor area (IDC) is marked

in red ……………………………………………………………………………………………………………......47

Figure 3.11 Calibration of sensor circuit with glucose solution. The red curve

shows the simulation and the black triangles are the measurement results. The

resonant frequency up-shifts with increasing glucose concentration………………….48

Figure 3.12 a) Resonant frequency peak for chip treated with antibodies

incubated with 0.88 µM creatinine. (b) Resonant frequency peak for chip

treated with antibodies incubated with 88 µM creatinine………………………………….49

Figure 3.13 Measured variation of resonant frequency as a function of creatinine

concentration used in the incubation of antibodies. The black curve shows the

resonant frequency for four samples with increasing concentration of creatinine

used in incubation while the experiment was done in aqueous (water)

environment. The resonant frequency downshifts with increasing creatinine

concentration. The red curve shows the same experiment done with air as the

surrounding medium………………………………………………………………………………………….50

List of Figures

125

Figure 3.14 Error bar measurement for two sets of chips. The maximum frequency

drift does between two chips does not exceed 4 MHz……………………………………….52

Figure 4.1 Typical lab on a chip system showing the microfluidic and detection

system. The same platform houses the microcontroller and the detection

circuits………………………………………………………………………………………………………………..55

Figure 4.2 Variation of capacitance of the IDC with respect to permittivity of MUT

for different thicknesses of passivation layer on top of the IDC. The variation of

capacitance reduces with increasing thickness of passivation layer…………………...57

Figure 4.3 Planarization of the BiCMOS stack for microfluidic integration. (a)

Schematic of the back-end-of-line stack. (b) Schematic of the stack followed by

planarization (c) SEM image of a typical planarized chip…………………………………….58

Figure 4.4 Schematic of the microfluidic integration with the CMOS sensor

chip…………………………………………………………………………………………………………………….59

Figure 4.5 Fabrication and bonding of PDMS microfluidic channel with the CMOS

sensor chip. The PDMS microfluidic system is fabricated using soft lithography

approach. Oxygen plasma bonding is used to bond the PDMS microfluidic channel

to the silicon chip……………………………………………………………………………………………….60

Figure 4.6 Permittivity of water with respect to frequency. The static permittivity

of water is 78 while the infinite frequency permittivity id 4. The characteristic

frequency of the Debye relaxation process is 18 GHz…………………………………………61

Figure 4.7 Glucose sensor with non-conductive wall around the senor for fluid

handling. (a) Typical sensor chip (b) Sensor mounted on board with non-

conductive wall…………………………………………………………………………………………………..62

Figure 4.8 Variation of the resonance frequency of the oscillator with materials of

different permittivities. The oscillating frequency downshifts with increasing

permittivity…………………………………………………………………………………………………………63

Figure 4.9 Variation of oscillating frequency of the sensor with increasing

concentration of water in glucose solution…………………………………………………………64

List of Figures

126

Figure 4.10 Variation of the oscillating frequency for different organic alcohols.

The resonance frequency downshifts with increasing permittivity and has a

sensitivity of 100 Mhz/permittivity……………………………………………………………….......65

Figure 4.11 Variation of oscillating frequency of the sensor with increasing

concentration of water in glucose solution…………………………………………………………66

Figure 4.12 Dielectric dispersion curve for biological cell suspension…………………67

Figure 4.13 Variation of oscillating frequency of the sensor with varying

concentration of microbeads in acetone…………………………………………………………….68

Figure 4.14 Calibration of the sensor system. Resonant frequency downshifts with

increasing permittivity of alcohol……………………………………………………………………….70

Figure 4.15 Variation of the resonant frequency of the oscillator with varying

fraction of fat and calcium in blood. The resonant frequency scales up with

increasing concentration…………………………………………………………………………………….70

Figure 4.16 Permittivity of binary mixture: fat and calcium in blood. The extracted

permittivity values are fit to the analytically calculated permittivity values………..71

Figure 4.17 Variation of the resonant frequency of the oscillator with varying

fraction of fat and calcium in water. The resonant frequency scales up with

increasing concentration…………………………………………………………………………………….72

Figure 4.18 Four unit switched sensor array. PMOS transistors are used as

switches to a common current source supplying the sensor oscillators. A digital

control for the switches is shown……………………………………………………………………….73

Figure 4.19 Chip micrograph of four unit sensor array………………………………………..73

Figure 4.20 Imaging of dielectric distribution as tabulated in table 1; Green:glue,

Orange:honey,Blue:air………………………………………………………………………………..........74

Figure 5.1 Previously fabricated sensor chip with long channel microfluidic system

integration. a) High-frequency sensor chip showing the sensor arrangement. b) A

List of Figures

127

long channel microfluidic channel is aligned on top of the sensor. The two

conditions depict the channel with and without the fluid…………………………………..79

Figure 5.2 a) Schematic depiction of particle flow in a long channel fluid system

aligned on top of the sensor. b) Geometry of IDE sensor considered in this work.

Simulated variation of sensor capacitance due to flow of particles. The

capacitance of variation is plotted with respect to position of particle on top of

the sensor…………………………………………………………………………………………………………..80

Figure 5.3 Schematic of the sensor circuit. The sensor is embedded in the

oscillator circuit. The variable capacitor Cmod is used for experiments without fluid

integration………………………………………………………………………………………………………….82

Figure 5.4 Demodulator architecture block diagram. The output of the

demodulator is taken from the fine-tuning loop containing C1 and R. The VCO

shown in the block represents the oscillator with sensor……………………………………83

Figure 5.5 Differential operational amplifier with RC biasing for self-

calibration………………………………………………………………………………………………………….87

Figure 5.6 Loop filter showing the coarse and fine tuning loop…………………………..88

Figure 5.7 Charge pump architecture for fine tuning loop………………………………….89

Figure 5.8 Chip photograph of the sensor and demodulator

architecture………………………………………………………………………………………………………..90

Figure 5.9 Simulated settling time of the PLL as a function of the coarse loop filter

capacitance. The settling time obtained is 3 µs. Capacitance change every 3 µs can

be accurately detected……………………………………………………………………………............91

Figure 5.10 Simulated demodulator output for input modulating voltage of period

100 µs. The period of the voltage being much higher than the settling time of PLL,

it is accurately followed by the demodulator……………………………………………………..92

Figure 5.11 The output spectrum of the sensor oscillator. The operating frequency

is 14.272 GHz……………………………………………………………………………………………………..93

List of Figures

128

Figure 5.12 Demodulator output voltage as a function of the modulation period.

The demodulator voltage follows the input period till 300 KHz (3.3 µs)……………..93

Figure 5.13 Chip photograph showing dual demodulator architecture……………….96

Figure 5.14 Pulse train emulating the signals from 2 VCOs which are delayed by

time Δt……………………………………………………………………………………………………………….96

Figure 5.15 Correlation between the two demodulator voltage

outputs...................................................................................................................97

Figure 5.16 Pulse trains showing frequency pulses from two oscillators covered

with random noise. Cross correlation between the pulses shows the delay time

equal to the one obtained in Fig. 5.15....................................................................98

List of Tables

129

LIST of TABLES

Table 2.1 Performance parameters of SG25H1…………………………………………………..17

Table 2.2 Performance parameter of SG13S……………………………………………………….18

Table 4.1 Four sensor imaging scheme……………………………………………………………….75

Table 5.1 Component parameters for sensor circuit…………………………………………..83

List of abbreviations

130

LIST of ABBREVIATIONS

IHP Microelectronic – Innovations for High Performance Microelectronics, Leibniz

institute for innovative Microelectronics

SiGe: Silicon Germanium

MOSFET – Metal oxide semiconductor field effect transitor

POC – Point of Care

LOC – Lab on chip

DNA – Deoxy ribonucleic acid

CCD – Charge coupled device

ELISA – Enzyme linked immunosorbent assay

SPR – Surface plasmon resonance

CMOS – Complementary metal oxide semiconductor

BiCMOS – Bipolar Complementary metal oxide semiconductor

MEMS – Micro-electro-mechanical system

BEOL – Back end of line

RF – Radio frequency

IDC – Interdigitated capacitor

PDMS – Polydimethysiloxane

HBT – Heterojunction bipolar transistor

MIM – Metal insulator metal

List of abbreviations

131

TM – Top metal

FEOL – Front end of line

SRF – Self resonating frequency

MUT – Material under test

MOS – Metal oxide semiconductor

POCT- Point of care testing

Pd – Penetration depth

SEM – Scanning electron microscopy

FIB – Focused ion beam

Crea-BSA – Creatinine bovine serum albumin

PBS – Phosphate buffer solution

ADS – Agilent design system

CMP – Chemical mechanical polishing

CVD – Chemical vapor deposition

RIE – Reactive ion etching

PMMA – Polymethylmethacrylate

IVUS – Intra vascular ultrasound

OCT – Optical coherence tomography

PPM – Parts per million

PLL – Phase locked loop

VCO – Voltage controlled oscillator

PVT – Process voltage temperature

List of abbreviations

132

CP – Charge pump

FM – Frequency modulation

PFD – Phase frequency detector