Label-Free Biochemical Recognition Using MEMS Resonators for Microarray Technology

University of Denver Digital Commons @ DU

Electronic Theses and Dissertations Graduate Studies

1-1-2011

Babak Tousifar University of Denver

Follow this and additional works at: https://digitalcommons.du.edu/etd

Part of the Biomechanical Engineering Commons

Recommended Citation Tousifar, Babak, "Label-Free Biochemical Recognition Using MEMS Resonators for Microarray Technology" (2011). Electronic Theses and Dissertations. 655. https://digitalcommons.du.edu/etd/655

This Thesis is brought to you for free and open access by the Graduate Studies at Digital Commons @ DU. It has been accepted for inclusion in Electronic Theses and Dissertations by an authorized administrator of Digital Commons @ DU. For more information, please contact [email protected],[email protected]. LABEL-FREE BIOCHEMICAL RECOGNITION USING MEMS RESONATORS FOR

MICROARRAY TECHNOLOGY

______

A Thesis

Presented to

the Faculty of Engineering and Computer Science

University of Denver

______

In Partial Fulfillment

of the Requirements for the Degree

Master of Science

______

Babak Tousifar

November 2011

Advisor: Dr. Siavash Pourkamali

©Copyright by Babak Tousifar 2011

Author: Babak Tousifar Title: LABEL-FREE BIOCHEMICAL RECOGNITION USING MEMS RESONATORS FOR MICROARRAY TECHNOLOGY Advisor: Dr. Siavash Pourkamali Degree Date: November 2011

Abstract

Highly sensitive biosensors capable of detecting and characterizing smallest

quantities of cellular and molecular targets are needed in pharmaceutical and medical

diagnostics industries. In this work, the importance of biological target recognition

specifically proteins through microarray technologies has been discussed and the most

successful tools and techniques have been studied. Moreover, a thermally actuated Micro

Electro-Mechanical Systems (MEMS) resonator has been demonstrated and fabricated in

this work as an accurate, reliable and low cost biotechnology tool.

For the proof of concept, amine coupling of octadecylamine chain to the epoxide

reactive groups of the functionalized silicon dioxide surfaces has been shown through

resonator frequency monitoring. The frequency deviation of the sensors implied a

meaningful surface coverage after target detection. Furthermore, X-ray Photoelectron

Spectroscopy (XPS) analysis of the devices at different stages of the surface modification

supported that the frequency deviations are due to receptor and target attachments.

In addition, two surface passivation techniques against non-specific adsorption of proteins have been investigated on silicon surfaces and fluorescent microscopy was

performed for quality control and validation of experiments.

Acknowledgements

Words fall short to convey my gratitude to those who helped and supported me

during my Master studies. I would like to take this opportunity to express my individualized thanks to all of those, even though the list is long and the space allocation for this purpose is short.

My special thanks and heartfelt gratitude go to my dear academic advisor Dr. Siavash

Pourkamali and research co-adviser Dr. Byron W. Purse for their invaluable advice, patience,

guidance, and consistent support throughout my studies. The accomplishment of this work

would have been impossible without their help and support.

I would like to extend my sincere appreciation to my dear professor Dr. Corinne

Lengsfeld for her kind supports. I am also grateful of Professor Daniel L. Armentrout for his

thoughtful advice in a section of this thesis. I would like to thank the Department of

Mechanical and Material Engineering of University of Denver for its financial support and

providing a friendly and progressive work atmosphere. I also appreciate the help and

cooperation of Department of Chemistry at Colorado State University for providing the XPS

facility. Furthermore, I would like to thank my colleagues and friends and at University of

Denver Specifically Dr. Mirek Kvanisca and Jennifer Chapin for their valuable dedications.

Finally, the deepest gratitude goes to my ever companion Saba Bakhshi, my dear

family and friends for their eternal love and support.

iii

Table of Contents

List of Figures ...... vi

List of Tables ...... x

Chapter One: Introduction ...... 1 1.1 Proteomics and Microarrays ...... 1 1.2 Gene Expression Review ...... 2 1.2.1 Transcription ...... 3 1.2.2 Translation ...... 3 1.3 Protein Microarray Applications ...... 7 1.3.1 Protein Therapeutics ...... 7 1.3.2 Medical Diagnostics...... 8 1.3.3 National Safety ...... 9 1.4 Microarray Sensing Mechanisms ...... 9 1.4.1 Sensor Interface and Substrates ...... 10 1.4.2 Molecular Recognition Elements ...... 11 1.4.3 Transduction Mechanisms ...... 16 1.5 Microarray Challenges and its Prospective ...... 24

Chapter Two: MEMS Biosensing Method and Mechanism ...... 27 2.1 Thermal-Piezoresistive Resonators ...... 27 2.1.1 Thermal Actuation ...... 27 2.1.2 Device Concept and Operation Principles ...... 28 2.1.3 Fabrication Process ...... 30 2.1.5 Detection Sensitivity Limit ...... 33 4.4 Resonator Static Temperature and its Effect on Biomolecules ...... 35 2.1.7 Biochemical Sensing Mechanism ...... 37

Chapter Three: Biochemical Sensing Results and Discussion ...... 39 3.1 Octadecylamine Sensing ...... 39 3.1.1 Surface Functionalization ...... 39 3.1.2 Octadecylamine Couplings ...... 41 3.1.3 Frequency Measurements ...... 41 3.1.4 Surface Coverage Analysis ...... 44 3.1.5 XPS Surface Analysis ...... 45 3.2 Biotin-Avidin Conjugation Sensing ...... 49 3.2.1 Surface Biotinylation ...... 50

3.2.2 Avidin Conjugation ...... 53 3.3 Surface Passivation Methods and Characterization ...... 57 3.3.1 Polymer Chemistry ...... 57 3.3.2 Surface Chemistry ...... 60

Chapter Four: Conclusion, Discussion and Future Work ...... 64 4.1 Real Time Biochemical Sensing ...... 64 4.2 Array Implementation ...... 66 4.3 Encapsulation in Micro-Fluidic Channels ...... 67

References ...... 69

List of Figures

Figure 1.1 (a) Gene locations in Eukaryotic Cells in the form of DNA double helix strand, (b) DNA double strand macromolecule structure and its biochemical components. Figure 1.2 Transcription process in the cell nucleolus starts from attachment of an RNA Polymeras to the specific genes and encode a single strand molecule (RNA) from the template strands. Figure 1.3 Three-nucleotide based Codons for amino acids. UAA, UAG and UGA are stop codons which terminate the translation process. Figure 1.4 Gene expression process in eukaryotic cells and detailed description of its steps from DNA transcription to Protein (polypeptide) translation. Figure 1.5 Biosensor schematic and its components including recognition element, interface and transducer. Figure 1.6 Microarray substrate with the grown interfaces to immobilize the molecule of interest i.e. protein. Figure 1.7 (a) Double Strand DNA (dsDNA) denatures in higher temperatures in its melting temperature (Tm) to form ssDNAs. (b) ssDNA can be used as both probe and target for biosensing application based on the nucleotide sequential complimentary matches (hybridization process). Figure 1.8 (a) Antibody (Immunoglobulin) structure, (b) Antigen binding site (c) Formation of antibody/antigen complexes on a virus and globular protein surfaces. Figure 1.9 MREs Surface Immobilization Methods. (a) non-covalent/non-specific (b) covalent/non specific, (c) non-covalent/specific and (d) covalent/specific. Figure.1.10 Nucleic Acid Programmable Protein Array (NAPPA) technique illustration. Figure 1.11 Protein microarray protocol via sample labeling. Figure 1.12 Scanning systems in microarray technology via labeling (a) PhotoMultiplier Tube (PMT) (b) Charge Coupled Devices (CCD) Figure 1.13 Surface Plasmon Resonance (SPR). The detection angle in respect to the intensity change at a fixed angle (non-reacted receptor) has been illustrated. Figure 1.14 (a) Mass Spectrometry (MS) detection principles. (b) Acquired spectrum example

Figure 1.15 (a) Atomic Force Microscopy (AFM) Detection Principals (b) AFM image of the surface of OmpF channel proteins, (b) closed channels (c) open channels (Müller and Engel). (d) Three components of the OmpF channel protein (Cowan et al.). Figure 1.16 A built-in microfluidic channel in a microcantiliver structure implemented as a biosensor. The additional mass from the biochemical reaction results in changing of the cantilever reference frequency. (b) SEM image of an array microcantilevers 750µm long, 100µm long and 0.9µm thick Figure 1.17 (a) QCM device configuration (b) Sensing principals based on the reference frequency set point Figure 1.18 Electrochemical detection in a glucose eeter. Figure 2.1 Top view schematic diagram of a thermally actuated I2-BAR showing the current flow and the qualitative distribution of AC temperature fluctuation amplitude (red being the maximum and blue minimum). Figure 2.2 COMSOL finite element analysis of a thermally actuated I-shaped resonator showing the vibration direction and the qualitative distribution of AC temperature fluctuation amplitude (red being the maximum and blue minimum). Figure 2.3 (a) Single-mask resonator microfabrication process. The blue, yellow and gray colors imply thermally grown silicon dioxide, built-in SIO silicon dioxide layer and silicon substrate (b) SEM view of the micro fabricated resonators in arrays. (c) Single I2- BAR Resonators with thermal actuators in the middle of the structure (d) Resonator’s mass sensing part (e) a 31MHz 10µm thick resonator etched using a custom cyclic RIE recipe, and (f) a 61MHz 15µm thick I2-BAR etched by DRIE. (ICP + Bosch process) (g) Rotational mode disc resonator capable of working in the liquid Figure 2.4 Measured frequency responses for the thermally actuated 61MHz resonator of Fig. 2c with different bias currents. Red and blue plots refer to vacuum and Air testing conditions respectively. Current range = 45-100mA in vacuum and 55-100mA in air. Figure 2.5 (a) Change in the measured resonance frequency for the resonator of Fig. 2.3.f as a function of the overall exposure time to flow of particles. The plot shows four consecutive reduction steps with one of the steps having a slope twice the rest showing that two particles have been deposited during that specific interval (10-15min); (b) SEM view of the same resonator after particle deposition showing 5 spherical particles deposited on its two sensing plates. The particle on the support has negligible effect on the resonant frequency. Figure 2.6.COMSOL finite element thermal analysis of the resonator of Fig. 1, by applying DC and AC currents of 30mA and 5mA between the two pads on the two sides of the resonator. The color code shows static temperature of the structure resulting from vii

the DC bias current that has a maximum value of only ~50ºC. The numbers show temperature fluctuation amplitude (Tac) at different points along the length of the actuator. Figure 2.7 Biochemical Sensing Principals of the fabricated MEMS Resonator. (a) Functionalized resonator with antibody, (b) Specific target detection, (c) Frequency deviation of the second resonator due to the additional mass Figure 3.1 Surface functionalization with an epoxide allows for the attachment of a variety of MREs, R–NH2. Figure 3.2 Biosensors surface linkage synthetic scheme used for surface activation, functionalization with epoxide groups and octadecylamine coupling through amine- epoxide reaction. Figure 3.3.(a) SEM view of a 3-bar I-shaped resonator with different resonator dimensions labeled on it. (b) Consistent frequency shifts for different resonator bias currents before and after octadecylamine attachment. Figure 3.4 Each epoxide group links to the surface by bonding to three silicon open bonds on the surface Figure 3.5 (a) Model of a silicon crystal as seen on the 100 surfaces, the area of every single silicon crystal is equal to 0.3nm2; (b) 8 silicon atoms on the surface with open bonds in every single silicon crystal unit available for surface linking. Figure 3.6 Different surface chemistry analysis techniques in respect to the detection range, analytical area. Figure 3.7 X-ray Photoelectron Spectroscopy (XPS) (a) Facility at University of Maryland (b) Instrumentation principles implemented for elemental composition analysis of our biosensors Figure 3.8 XPS surface analysis results of the resonators surfaces confirms addition of the carbon and nitrogen containing amine groups to the surface in different reaction steps. (a) SiO2, (b) Epoxide, (c) Octadecylamine Figure 3.9 (a) Biotin Structure (b) Avidin tertiary structure with its four subunits each holding a biotin (c) Schematic surface linkage synthesis for Biotin/Avidin Conjugation Figure 3.10 Biosensors surface modification processes used for functionalization with epoxide reactive groups to graft PEG2-Biotin (amin-activated biotin) as the receptor Figure 3.11.a XPS surface analysis of the biosensors before and after surface biotinylation process shows considerable amount of nitrogen, carbon and sulfur after biotin grafting. The lower Si2p implied a chemical coverage on top of the silicon dioxide surfaces. viii

Figure 3.11.b Elevated amount of detected Sulfur shows the reaction optimization performed at 800C Figure 3.12 Surface linkage synthetic schemes used for Biotin-Avidin conjugation process. In this case avidin as our target couples with biotin as the receptor chemically grafted on the surface Figure 3.13 XPS analysis for the tested resonators before and after rinsing process Figure 3.14 XPS surface analysis of the silicon dioxide surfaces before and after immersion in avidin solution shows a significant amount of elevated nitrogen, carbon and sulfur. Figure 3.15 PLL-PEG Structure used for surface passivation of the biosensor against proteins physical adsorption Figure 3.16 PLL-g-PEG schematic function as covers the negatively charged silicon dioxide surfaces against protein non-specific adsorption. As shown avidin only binds specifically to its receptor, biotin and the copolymer repels all other kinds of proteins. Figure 3.17 Fluorescent microscopy of the prepared samples based on the table 3.3. The intensity of 488nm emitted light increased due to the presence of Alexa fluor® 488 Avidin conjugated with biotin and increased in samples with higher ratios of PLL-PEG- Biotin in non-biotinylated copolymer i.e. TRIC labeled PLL-PEG (red). Figure 3.18 Heterobifunctional crosslinkers surface linkage synthesis used for surface biotinylation and passivation at the same time. Figure 3.19 Fluorescent Microscopy of the: (a) Control Experiments (b) Passivation Experiment Figure 4.1 a) SEM view of a fabricated rotational mode disk resonator; b) COMSOL finite element modal analysis of the same resonator. Red and blue show the maximum and minimum vibration amplitudes respectively; c) Measured frequency response for the same resonator both in air and in water showing an unprecedented Q of 304 in liquid. Figure 4.2 (a) Schematic view of a disk resonator along with its electrical connections encapsulated in a microfluidic channel. (b)Schematic view of the proposed 2-dimensional resonator arrays with isolated vertical and horizontal electrical interconnects and switches for resonator selection. Figure 4.3 SEM view of an encapsulated capacitive resonator resonator, b) cap openings covered with PECVD silicon dioxide, c) openings etched in oxide for electrical contacts, d) frequency response of the encapsulated resonator.

List of Tables

Table 1.1 Transduction and detection mechanisms comparison table Table 2.1 Summary of the measurement results obtained from I2-BARs with different dimensions showing increasing K as the same resonator is scaled down. The 15.9MHz and 31MHz resonators are fabricated on a 10µm thick device layer etched using a custom cyclic RIE recipe, the rest of the devices are fabricated on a 15µm thick device layer etched by silicon DRIE Table 3.1 Summary of measurement results for different functionalized resonators Table 3.2 Resonators frequency measurements after avidin conjugation and DI water rinsing processes Table 3.3 PLL-PEG deposition characterization plan

Chapter One: Introduction

1.1 Proteomics and Microarrays

Understanding complex cellular systems will require the identification and

analysis of each of its components and determination of their function together. Now that

more than 40 genomes including the human genome have been fully sequences, the next

daunting challenge will be to connect genes to function. A DNA microarray highlights

those genes which are expressed, but the translation process and its protein product are

further regulated by gene silencing via RNAi [1]. Therefore, a critical step in this process is to determine the biochemical activities of the proteins and how these activities themselves are controlled and modified by other proteins.

On the other hand, proteins are complex macromolecules both in terms of

structure and function and their chemical structure has been developed and modified

during human evolution history [2]. Unlike the genome which contains a static number of genes, the levels of protein within the cells are extremely dynamic. Proteins are continuously processed within the body in response to external stimuli and go through a wide range of post-translational modifications [3,7]. Therefore, it is hard to accurately determine the exact number or quantities of proteins which are present within humans, but it is predicted that the human proteome expressed by 40,000 genes exceeds 1 million proteins [4]. In the other word, Protein families are remarkably diverse and have

considerable differences in their physical sizes, chemical and structural properties,affinity

constants and relative abundance within the cells [15]. As a result of such diversity,

protein mapping and accurately characterizing of its interactions is extremely challenging

and requires high throughput techniques.

Typical antibody-based methods such as Enzyme-linked Immunosorbent Assay

(ELISA) have been elucidated by studying single molecules, i.e. one experiment at a time. However, these processes are slow, laborious and expensive. For example measuring the affinity binding constant of the protein-protein reaction which is desirable to study the disease biochemical pathways needs several millions of such assays performed over several months [5]. In contrast high throughput screening (HTPS) methods which will be discussed in section 2.2 have been developed in the last decade to optimize the study of large numbers of molecules, including DNA, proteins and metabolites [6,7,8].

1.2 Gene Expression Review

Gene expression is a cellular procedure by which the information in genome is used and processed into biosynthesis of functional products, proteins. This procedure is highly similar in all living things from eukaryotes to even viruses. Gene expression consists of three major steps including transcription, translation and post-translational modifications. The genetic information known as genome is preserved in the cell chromosome in the form of DNA (deoxyribonucleic acid), composed of two long polynucleotide strands with nucleotide subunits. Each subunit consists of a deoxyribose sugar, a nitrogen containing base attached to the sugar, and a phosphate group. There are

four different types of nucleotides including adenine (A), guanine (G), cytosine(C) and

thymine (T) in a DNA strand which are different in their nitrogenous base (Figure 1.1).

1.2.1 Transcription

In transcription, a single strand polynucleotide called RNA is transcribed from a

DNA strand which represents at least one gene. The transcription process starts from the

5’ side of the coding strand by attachment of an enzyme called RNA polymerase (RNAP)

and continues by adding RNA nucleotides to the 3’ side of the daughter strand. During

the transcription nucleotide thymine (T) is replaced with uracil (U) in the new RNA

strand. Transcription process leads to generate different types of RNA including

messenger RNA or mRNA (a coding molecule) and other non-coding molecules such as

transfer RNA (tRNA), ribosomal RNA (rRNA) and small nuclear RNA (snRNA) can be

produced as shown in Figure 1.2.

1.2.2 Translation

Translation is a following process in gene expression by which proteins are being synthesized. Proteins consist of one or more single linear polymer chains called polypeptide folded into globular and fibrous forms guided in part by hydrogen bonding.

A polypeptide itself consists of various amino acids bonded covalently together by peptide bonds which are between the carboxyl and amino groups of adjacent amino acid

residues. There are totally 20 amino acids and their sequence in a protein polypeptide

chain is defined by the groups of three consecutive nucleotides on mRNA called codons.

Translation takes place in the cell cytoplasm in the presence of mRNA, ribosome, tRNA

(an adapter molecule) and amino acids as illustrated in Figure 1.3 and 1.4.

(a)

(b)

Figure 2.1. (a) Gene locations in Eukaryotic Cells in the form of DNA double helix strand, (b) DNA double strand macromolecule structure and its biochemical components. [9],[10] 4

Figure 1.2. Transcription process in the cell nucleolus starts from attachment of an RNA Polymeras to the specific genes and encode a single strand molecule (RNA) from the template strands. [10]

Figure 1.3. Three-nucleotide based Codons for amino acids. UAA, UAG and UGA are stop codons which terminate the translation process. [11]

Figure 1.4. Gene expression process in eukaryotic cells and detailed description of its steps from DNA transcription to Protein (polypeptide) translation.

[9]

1.3 Protein Microarray Applications

In general, based on the array surface formulations and detection techniques, two

trends of protein microarrays are currently used to study the biochemical activities of

proteins: analytical and functional microarrays. Analytical microarrays detect the abundance of proteins in a complex mixture in order to measure the binding affinities or the protein expression levels. The most widely use of such microarrays are antibody microarrays with application in medical diagnostics [12,13]. However, Functional microarrays are used to study the biochemical activities of an entire proteome in a single experiment and investigate various enzymatic activities and protein interactions such as protein–protein, protein– nucleic acid, protein–lipids interactions [14].

Therefore, the abovementioned methods can play an important role in protein therapeutics, medical diagnostics and nation safety.

1.3.1 Protein Therapeutics

Using antibody-based therapeutics in drug discovery is going to expand during this decade [15]. Currently 20% of the marketed biological products are allocated to antibodies and numerous therapeutics products are being monitored for clinical trials

[16]. It is predicted that several tens of monoclonal antibodies will be present in the market within the next several years [17]. Therefore, screening these high affinity macromolecules increases the need to develop a successful and efficient microarray technology and integration of such into the drug discovery pipeline.

1.3.2 Medical Diagnostics

Biochemical changes in the patient's blood can be a signal from a damaged or dysfunctional organ in advance, prior to microscopic cellular damages or other symptoms

that can be observed. Early, rapid, and sensitive detection of most diseases is vital for

successful clinical treatment. Infectious diseases such as HIV, herpes and hepatitis C

Virus (HCV) are increasing rapidly around the world, with several million new cases in each disease category annually [18]. According to World Health Organization (WHO), in the case of HCV [19], it is estimated that there are over 170 million people affected chronically globally and the number of new cases is increasing at a rate of 3 to 4 million annually [15]. Moreover, the chronic stage of the infection which causes liver inflammations and cancer kill approximately 300,000 people every year. This is reaching an epidemic state among both developed and developing nations, and is affecting all sections of society. Today there is no vaccine against HCV and only 40-50 % of the patients manage to clear the virus permanently from their bodies after up to 48 weeks of combination therapy with pegylated interferon alpha and ribavirin [20]. Due to the silent character of the infection and infrequent routine testing for HCV, the disease often remains unnoticed for years and treatment does not start until irreversible liver damage has occurred. Most diagnostic assays which are based on Enzyme-linked immunosorbent assay (ELISA) or Polymerase Chain Reaction (PCR) formats are costly due to the labwares disposal and require lots of time and labor. Therefore, Microarray-based platforms containing ligands for a variety of disease and running with small samples can

reduce the cost of diagnostics in infectious diseases alongside with its high accuracy and reliable detection.

1.3.3 National Safety

Detection of pathogens in the environment, food and battlefield conditions is becoming a requirement [21]. Aptamers or synthetic ligands which mimic protein characteristics are more suitable for extreme environments where proteins cannot survive using natural receptors in the conventional microarray surface immobilization techniques

[22]. Moreover, Assembly of proteins on solid state substrates integrating with microfluidics can play an important role to detect bacteria and other toxins rapidly. All of these criteria impact the development of protein microarray technology and additional funding in this sector will encourage research and development for novel sensing tools

[15].

1.4 Microarray Sensing Mechanisms

The protein microarrays like any type of biosensing system is defined by the

International Union of Pure and Applied Chemistry (IUPAC) to be a “device that uses specific biochemical reactions mediated by isolated enzymes, immunosystems, tissues, organelles, or whole cells to detect chemical compounds usually by electrical, thermal, or optical signals” [23]. Therefore biosensors consist of three components (Figure 1.5):

1. A substrate or interface able to immobilize the biomolecules

2. A biological recognition element able to interact specifically with a target

3. A transducer able to convert the recognition event into a measurable signal

Figure 1.5. Biosensor schematic and its components including recognition element, interface and transducer. [24]

1.4.1 Sensor Interface and Substrates

Generally, all protein arrays are composed of a substrate to attach the chemical

and biological interfaces as the sensor’s receptor (Figure 1.6). Preferred materials include

glass, silicon, plastic and synthetic polymers such as polystyrene,

polyvinylidenedifluoride, (PVDF), nitrocellulose [33] Planar glass or silicon chips were the first substrates used in protein microarrays due to their low cost, low surface roughness for accurate array scanning and low fluorescent background to obtain higher signal to noise ratio [25, 26]. PVDF (polyvinylidene fluoride) membranes are well known for high protein binding affinity without additional surface modifications; however suffer from lack of specific control in orientation [27]. Nitrocellulose, is among the most popular substrates due to its low cost production and high binding surface area, although non-specific binding and high fluorescent background problems has been reported [28]. Gold surfaces are also common to be used in various detection methods such as Fluorescence, Surface Plasmon Resonance SPR and electrochemistry due the

surface chemistry compatibility [29,30,31]. However depositing of such coatings requires elaborate and expensive thin film deposition techniques such as physical vapor deposition (PVD) or chemical vapor deposition (CVD) [32] and sometimes a suitable interlayer adhesive such as chromium is required to fix the deposited coating to the substrate [33].

Figure 1.6. Microarray substrate with the grown interfaces to immobilize the molecule of interest i.e. protein. [33]

1.4.2 Molecular Recognition Elements

Molecular Recognition Elements (MREs) are immobilized on the sensor surface and bind specifically to their target molecules (ligand) as in the lock-and-key principal.

There are different sorts of molecules to be used as recognition elements, however DNA and proteins are among the most common types. In the case of DNA the recognition mechanism is based on DNA hybridization between immobilized probe and the desired target strand with a complementary sequence. In fact, double helix DNA strand at its 11

melting point denatures and form a single strand. Therefore adjusting the temperature in

the sensing mode cause the double helix to reform (renaturation) at regions where the

complimentary sequences are aligned (Figure 1.7).

(a) (b)

Figure 1.7. (a) Double strand DNA (dsDNA) denatures in higher temperatures in its melting temperature (Tm) to form ssDNAs. (b) ssDNA can be used as both probe and target for biosensing application based on the nucleotide sequential complimentary matches (hybridization process). [34,35]

Protein-based biosensors include antibody/antigen and enzyme/substrate that

works based on high specific, versatile and strong bonding of the recognition elements to

its target. In the other word, antibodies as a receptor have a specific binding site which a

defined antigen can be matched with it (Figure 1.8). Although the most commonly used recognition molecules are antibodies (immunosensors), other molecules including aptamers (protein-binding nucleic acids) and peptides have been utilized in microarrays.

(a) (b)

(c)

Figure 1.8. (a) Antibody (Immunoglobulin) structure (b) Antigen binding site (c) Formation of antibody/antigen complexes on a virus and globular protein surfaces. [36]

1.4.2.1 MREs Immobilization

A number of various methods are being used to facilitate immobilization on the surface. They can broadly categorize as non-covalent/non-specific, covalent/non-specific and non-covalent/specific and covalent/specific (Figure 1.9). The common procedure for the first category is protein physical adsorption to the surface via electrostatic, hydrophobic or polar interactions (Figure 1.9.a) [37,44]. The drawback of this technique remains to be its random and uncontrolled recognition element orientation, week and 13

reversible immobilization and the receptor instability itself due to the high non specific

bindings. For stronger immobilization the covalent binding to the SAM (Self Assembled

Monolayer) with functionalized groups such as amine and epoxide can be used, although

the couplings are potentially non-specific and non-oriented (Figure 1.9.b). In another approach an intermediate/tag molecules with a relatively high binding affinity may be used to avoid the non-specific bindings with the immobilized protein. In this method such as avidin-biotin, DNA directed ( ssDNA to ssDNA/Protein hybridization), His-tag, protein A or G (used to immobilize IgG) and GST (glutathione S-transferase which binds to glutathione) may be used for immobilization [38,39, 40, 41] (Figure 1.9.c). However the binding strength in this method can be low due to the weak intermolecular interactions. To obtain both binding strength and specificity, SAM can be functionalized with maleimido/thiolated or Heterobifunctional Crosslinkers such as N- hydroxysuccinimide ester reagents for coupling with protein cysteine residues or active sites (Figure 1.9.d) [42,43, 44].

Another alternative approach is in situ proteins synthesis and on-site immobilization system called nucleic acid programmable protein array (NAPPA). This method is based on the cell-free gene expression and directly sensing the product. The chips are prepared with a pre-array of immobilized plasmid (an autonomous self replicating DNA in the form of loop). Then the mammalian cell-free gene expression transcripts and translates the plasmid gene. As a result of this self-expression system, the recognition element such as GTS-tagged protein is encoded and immobilized by a previously localized anti-GST antibody (Figure 1.10) [44].

(b) (d)

(a) (c)

Figure 1.9. MREs Surface Immobilization Methods. (a) non-covalent/non-specific (b) covalent/non specific, (c) non-covalent/specific and (d) covalent/specific. [44]

Figure.1.10. Nucleic Acid Programmable Protein Array (NAPPA) technique illustration.

44]

1.4.3 Transduction Mechanisms

Microarray classifications based on their detection methods are direct and indirect

recognition. In each group, there are several types of transducers including optical,

electrochemical, and acoustic, mechanical, magnetic and thermal. Generally

characteristics including nature of application, the label molecule (in direct detection),

sensitivity, cost and speed of the detection play an important role in overcoming the new

technologies.

1.4.3.1 Sample Labeling (Indirect) Technique:

The majority of medical diagnostics method involving antibody microarrays

reported to date, including those profiling the protein composition of serum in patients

with prostate, pancreatic, bladder and lung cancers are based on direct sample labeling

[45]. To detect proteins on such microarrays, targets get labeled with fluorescent dyes.

Also, a variety of photochemical, radioactive or affinity tags are used as an intermediate

molecule to track down the target during the process. After binding the labeled target to

its specific receptor, dye molecules get excited by a high power laser beam.

Subsequently the emission lights collected by Charge Coupled Devices (CCD) or

PhotoMultiplier Tube (PMT) generate images from the reacted surfaces. These images are analyzed by image processing and clustering techniques (Figure 1.11). Despite of being the most widely used diagnostic tools both in industrial and academic laboratories

this technique suffer from a relatively high price and long incubation time [46]. The detection of accuracy is a single dye in 10µm2 [47]. (Figure 1.11)

(a)

Figure 1.11. Protein microarray protocol via sample labeling. [48],

1.4.3.1.1 Microarray scanners

The mostly utilized scanners in labeling techniques are laser and CCD scanner. In the laser scanner know as PMT the laser light (dotted lines in figure 1.12.a) passed through a beam splitter through an objective lens where it is focused at the array surface.

The light reflections (solid lines) take place form a mirror and after traveling through an emission filter, focused into photomultiplier tube via a pinhole. On the other hand CCD scanner works based on high intensity white light (dotted line in figure 1.12.b). A desired wavelength is generated by passing the white light through an excitation filter. After

traveling inside a fiber optic guided through a fiber optic ring, it illuminated the area of

interest. The reflected lights from the surface (solid line) passes through an emission filter

where they are collected by the CCD camera. The amount of light enters into the camera

can be adjusted by the shutter.

(a) (b)

Figure 1.12. Scanning systems in microarray technology via labeling (a) PhotoMultiplier Tube (PMT) (b) Charge Coupled Devices (CCD)

[49]

1.4.3.2 Label-free (Direct) Platforms

1.4.3.2.1 Surface Plasmon Resonance (SPR)

The most frequently used label-free technique for detection of proteins is Surface

Plasmon Resonance (SPR) was introduced in 1983 [50]. This technology is based on the an optical surface sensing which monitors the refractive index of the dielectric layer formed on top of a metal film (Figure 1.13). The advantage of this method is the capability of real time analysis of the sample to obtain the kinetic and thermodynamic information of the protein/target interaction. However, the disadvantage of this method is

underlying on its relatively high equipment cost and detection limit

[44].

Figure 1.13. Surface Plasmon Resonance (SPR). The detection angle in respect to the intensity change at a fixed angle (non-reacted receptor) has been illustrated.

[51,52]

1.4.3.2.2 Mass Spectrometry (MS)

This analytical method is based on measuring the mass-to-charge ratio of the

charged particles [53]. Protein macromolecules, in the case of protein microarray, are

activated by incident of a laser beam, ionized, accelerated through an electric field and finally get separated based on their mass-to-charge ratios (Figure 1.14). MS has been

utilized for the advancement of proteomics during the last decade and it has provided

chemical and structural information not achievable through other methods [54,55]. MS is

simple, fast and sensitive for detection of small protein modifications (absent or oxidised 19

amino acids) and picomolar levels of cancer markers [56]. However, the experimental procedure and data analyzing remain challenging [57].

(a)

(b)

Figure 1.14. (a) Mass Spectrometry (MS) detection principles. (b) Acquired spectrum example [58] 1.4.3.2.3 Atomic Force microscopy (AFM)

AFM is one of the most established methods used for surface characterization since 1980s. It works based on the intermolecular force interactions between an integrated nano-scaled tip (cantilever) and sample interface [59]. As the probe scan the surface, deflection in the cantilever beam is detected through a photodiode sensor and generates a 3-dimentional image in the range of angstrom to illustrate the height of the sample. In both tapping and contact modes the height of the cantilever is adjusted by a feedback system (Figure 1.15). The high resolution of images in AFM can compete with 20

other techniques, although its relatively high operation cost, slow speed and expertise

requirements to run the instrument makes it more useful for design approaches and

protein-protein interactions [60] rather than microarray development [44] (see Figure

1.14).

(a) (d)

(c)

Figure 1.15. (a) Atomic Force Microscopy (AFM) Detection Principals (b) AFM image of the surface of OmpF channel proteins, (b) closed channels (c) open channels (Müller and Engel). (d) Three components of the OmpF channel protein (Cowan et al.).

[61,62,63]

1.4.3.2.4 Microcantilever

A derived technology based on AFM microcantilever deflection has been developed for sensitive and multianalyte recognitions. In such approach up to 50,000 cantilevers are fabricated in a 1cm 2 array [64,65,66] (Figure 1.16.b) and the detection principal is based on the analyte binding-induced changes in either mechanical stiffness or resonance frequency of the cantilever [67,68,69]. The detection limit as low as picomolar has been reported using built-in microfluidics channels (Figure 1.16) [70].

(a)

(b)

Figure 1.16. A built-in microfluidic channel in a microcantiliver structure implemented as a biosensor. The additional mass from the biochemical reaction results in changing of the cantilever reference frequency [70]. (b) SEM image of an array microcantilevers 750µm long, 100µm long and 0.9µm thick [15].

1.4.3.2.5 Quartz Crystal Microbalance (QCM)

QCM is one the earliest and most common acoustic-wave sensors for direct bio- detection [71,72]. Based on the piezoelectric properties of its crystalline structure, by applying voltage an acoustic wave propagates due to the structural deformation. In a typical application, the mechanical resonant device is embedded in an oscillator circuit, and the frequency shift resulting from bio-molecular reactions is recorded (Figure 1.17).

(a) (b)

Figure 1.17. (a) QCM Device Configuration (b) Sensing Principals based on the reference frequency set point [73]

A figure of merit for an acoustic mass sensor is the change in its frequency caused

by a change in the added mass per sensor unit area [73]. Typical measured mass sensitivities for such sensors are in the order of 10-2000 Hz/g in every square cm of the sensor and increases proportionally with the square of resonance frequency. Experimental frequency resolutions achieved for oscillators engaging such devices is 0.01 to 1Hz resulting in minimum detectable mass densities in the order of ng/cm2 [74].

1.4.3.2.6 Electrochemical

In these types of biosensors such as glucose meters a conductive polymer containing glucose oxidase covers the platinum electrode as illustrated in Figure 1.18. In

the presence of glucose provided by the analyte e.g. blood sample, hydrogen proxide is

produced and the flow of electrons passes through the device circuit. Therefore the generated current by such electrochemical reactions can be used to determine the glucose

concentration in the defined volume unit. 23

Figure 1.18. Electrochemical detection in a glucose meter.

[75]

In addition to the discussed label-free methods other approaches based on thermal approaches, nanotube/nanowire transistors [76], nanohole optical arrays [77], Kelvin

nanoprobes [78], surface enhanced Raman scattering [79] Localized Surface Plasmon

Resonance [80] and ellipsometry [81,82] are under investigation.

1.5 Microarray Challenges and its Prospective

Although various tools and techniques have been developed during the last decade, not all the detection techniques are amenable to miniaturization and Micro

Electro-Mechanical Systems (MEMS). Therefore their potentials for developing a high- throughput, cost-efficient, fast and sensitive bioanalytical microarray technology remain a concern.

Labeling (Indirect) methods are one the most sensitive biorecognition platforms

(in molecular level) [83,84], though the major drawback of such approach is associated with the labeling procedure which is time consuming, labor intensive and rarely equally

consistent for all proteins in a given sample which makes acquiring quantitative information more difficult. [85]. Moreover, labeling process has the potential to alter the tertiary structure (folding configuration) of the proteins and affect its function and binding properties. [86] Also due to the complex nature of the experiments and labeling process, the accurate detection can be affected.

On the other hand, the direct biochemical analysis methods have some limitations shown in table 1.1. For example, optical platforms such as SPR suffer from high cost of optical setups and low compatibility of being utilized for the high-throughput applications. Also electrochemical biosensors have the restrictions of being miniaturized and implemented in microarray platforms.

However, Micro electro-mechanical resonant chemical and biological sensing techniques are among the popular and commercially available techniques. Typically sensing based on measurement of mass-absorption induced frequency shift in a high quality factor mechanical resonator provides the highest sensitivity and resolution among other sensing mechanisms. While cheap to be fabricated, they are easy to be handled with a relatively fast sample analyzing. Also the sensors micron-sized configurations make them amenable for microarray platforms. Moreover protein bioaffinity characterizations for proteomics and drug design research can be when integrating with microfluidics.

Table 1.1. Transduction and Detection Mechanisms Comparison Table

Data Transduction Typical Microarray Affinity Time and Cost Analysis Platform Method Capability (K /K ) Labor on Off Difficulty

Labeling Fluorescent YES NO High High High ndirec I

Optical SPR NO YES High High Low

Electrochemical Amperometric NO NO Low Low Low Direct Acoustic and QCM, YES YES Low Low Low Mechanical Cantilever

Chapter Two: MEMS Biosensing Method and Mechanism

2.1 Thermal-Piezoresistive Resonators

As the potential emerging technology for next generation integrated frequency references and resonant sensors, MEMS resonators have received a lot of attention over the past decade [87,88,89,90,91]. Most of the resonators use piezoelectric [92,93] or

electrostatic (capacitive) [94,95,96] electromechanical transduction each having its

advantages and disadvantages. Piezoelectric resonators are mainly limited by their

material integration requirements and relatively low quality factors, while the major

bottle-neck for capacitive resonators is the weak electromechanical coupling leading to

the need for deep submicron transduction gaps and associated fabrication challenges.

2.1.1 Thermal Actuation

Thermal actuation is a well known mechanism that can be implemented at

microscale without any fabrication challenges or the need for sophisticated material

integration. Furthermore, it has the advantage of being an internal actuation mechanism,

i.e. the actuation force is generated inside the monolithic structure and the actuator is part

of the structure itself. We have demonstrated promising results using

thermalpiezoresistive resonators for sensory applications [98,99,100]. This significantly

improves the robustness and reliability of such resonators especially for sensory

applications where the resonator has to be in contact with the surrounding their

transduction gaps), but also penetration of the slightest amounts of contaminants into

their narrow transduction gaps can result in a complete failure of the resonator. In

addition, thermal actuators have great properties such as large actuation force, low

operating voltage and simplicity of design and integration. On the downside their power

consumption and high body temperature limits their application in some cases. Thermal

actuators are usually considered and referred to as slow actuators only suitable for DC or

very low-frequency applications. This is mainly due to the time delay for the temperature

of a heating element to reach the desired level and generate the expected force. However,

it has been shown that thermal actuators become more efficient as the resonator

dimensions are scaled down to reach higher resonant frequencies in the range of MHz

[98]. This becomes very clear when compared to the air gap electrostatic resonators that

not only cannot operate with a high-Q in a liquid environment (due to penetration of

liquid in

2.1.2 Device Concept and Operation Principles

Thermally actuated single crystalline silicon resonators with piezoresistive

readout have been used in this work as highly sensitive mass sensors. The simplicity and

monolithic structure of such devices makes them very robust and durable and therefore

suitable for sensory applications where the device needs to be in direct contact with the

surroundings. Besides, operation in liquid is generally a major challenge for microscale

resonators due to large surface to volume ratio and excessive viscous damping from the

liquid. However, disk-shaped thermal-piezoresistive resonators can be designed for operation in liquid with relatively high quality factors [97]. Although all the

measurements in this work have been performed in air after drying the resonators, the

eventual goal is to perform direct analysis with resonators while immersed inside a

biological solution.

Figure 2.2 shows COMSOL finite element analysis of a thermally actuated I- shaped resonator used in this work referred to as I2-Bulk Acoustic wave Resonators (I2-

BARs)[90].

Figure 2.2. COMSOL finite element analysis of a thermally actuated I-shaped resonator showing the vibration direction and the qualitative distribution of AC temperature fluctuation amplitude (red being the maximum and blue minimum).

Thermal actuation occurs by passing a current between the two pads on the two sides of the structure resulting in an ohmic power loss in the structure. By passing a combination of a DC and an AC current, the power loss will have a component at the

same frequency as the applied AC current: Pac = 2ReIdciac, where Re is the electrical

resistance between the two pads and Idc and iac are the applied DC and AC currents respectively. The fluctuating power loss results in a fluctuating temperature gradient and therefore fluctuating thermal expansion of the structure. Due to their higher electrical resistance, most of the heat gradient is generated in the narrow pillars in the middle of the

structure. The AC extensional force resulting from the fluctuating temperature in the

pillars can actuate the resonator in its in-plane extensional resonance mode. As the

resonator vibrates, the alternating tensile and compressive stress in the pillars results in

fluctuations in their electrical resistance (due to the piezo-resistive effect). This results in

fluctuations in the DC current passing through the resonator that represents the vibration

amplitude (output signal) of the resonator.

2.1.3 Fabrication Process

A single mask process [98] was used to fabricate the resonators on an N-type

Silicon on Insulator (SOI) substrate with device layer thickness of 5μm and buffer oxide

(BOX) thickness of 5μm. The fabrication process starts by thermally growing a thin

(~200nm) layer of silicon dioxide on the device layer that will serve as a hard mask for silicon etching. The silicon dioxide layer is patterned to define the resonator structures.

The structures are then carved into the SOI device layer all the way down to the BOX by plasma etching. Finally, the structures are released by etching the underlying BOX layer in hydrofluoric acid (HF). At the same time the remaining oxide mask on top of the structures is also etched away (Figure 2.3).

(a) (b)

(d) (c)

(e) (f) (g)

Figure 2.3. (a) Single-mask resonator microfabrication process. The blue, yellow and gray colors imply thermally grown silicon dioxide, built-in SIO silicon dioxide layer and silicon substrate (b) SEM view of the micro fabricated resonators in arrays. (c) Single I2-BAR Resonators with thermal actuators in the middle of the structure (d) Resonator’s mass sensing part (e) a 31MHz 10µm thick resonator etched using a custom cyclic RIE recipe, and (f) a 61MHz 15µm thick I2-BAR etched by DRIE. (ICP + Bosch process) (g) Rotational mode disc resonator capable of working in the liquid

-50 Tunning range=%1.39 -55

-60 dB -65 Tunning range =%0.91 -70

-75 60.8 60.98 61.16 61.34 61.52 61.7 Frequency (MHz)

-52 -63

-56 -66 dB dB -60 -69 Q=12000 Q=14000 -64 -72 Current=100mA Current=60mA -68 -75 60.825 60.84 60.855 60.87 61.617 61.632 61.647 61.662 Frequency (MHz) Frequency (MHz)

Figure 2.4, Measured frequency responses for the thermally actuated 61MHz resonator of Fig. 2c with different bias currents. Red and blue plots refer to vacuum and Air testing conditions respectively. Current range = 45-100mA in vacuum and 55-100mA in air.

Resonant frequency of ~61MHz with quality factors ranging from 12,000-14,000 were measured for this resonator under vacuum. Under armonspheric pressure, the quality factor dropped to 6,000-8,000. As expected, as the DC bias current increases the output signal level increases while due to the higher static temperature and softening of the structural material, the resonant frequency decreases. Table 2.1 presents measured data for different resonators under different bias current and pressure conditions.

Table. 2.1. Summary of the measurement results obtained from I2-BARs with different dimensions showing increasing K as the same resonator is scaled down. The 15.9MHz and 31MHz resonators are fabricated on a 10µm thick device layer etched using a custom cyclic RIE recipe, the rest of the devices are fabricated on a 15µm thick device layer etched by silicon DRIE.

Resonator Dimensions Scale Freq. Current K (µm) Q -1 Cond. Factor (MHz) (mA) gm (µA/V) (×10-7W ) a b c L W 59000 7.95 11.2 2.7 3.65 Vac. 1X 274 30 67 31 4 37000 7.75 40.5 53 8.67 Vac. 24000 10.87 26 20.4 12.57 Vac. 0.7X 193 21 47 22 2.7 16000 10.63 30.5 22.2 14.93 Vac. 12000 13.94 21.3 11.2 20.63 Vac. 0.5X 144 16 35 16 2 7000 13.63 30.9 24.4 36.49 Vac.

28500 15.89 55 15.4 1.78 Vac. 29000 15.65 80 55.5 3 Vac. 1X 80 53 16 64 17.4 10800 15.91 50 4.2 1.80 Atm. 10900 15.73 90 30.3 3.43 Atm. 43000 31.19 10 0.56 1.29 Vac. 38500 30.72 65 83.3 5.12 Vac. 0.5X 39 25 8.5 32 8.5 13900 31.19 20 0.91 1.64 Atm. 12000 30.90 70 20.4 3.47 Atm.

14000 61.64 60 7.7 1.53 Vac. 12000 60.85 100 29.4 2.45 Vac. NA* 22 15 4.4 18 5 7500 61.65 60 4.54 1.68 Atm. 7700 61.11 100 17.8 2.32 Atm. * Unlike the other two resonators in this group, this device has been fabricated on a thicker substrate and etched in an ICP system. Therefore it has wider and taller pillars resulting in lower K.

2.1.5 Detection Sensitivity Limit

The minimum detectable added mass per unit area of a resonant sensor depends on the resonator mass sensitivity and the minimum measurable shift in the output frequency which depends on the noise level of the oscillator circuit (oscillator phase noise). The resonator frequency shift caused by addition of a small mass can be calculated as:

ƒ = → ƒ = , ΔM<

where K, M and f are the effective stiffness, mass and resonance frequency of the resonator respectively, ΔM is the added mass Δƒ is the resulting frequency shift.

Theoretically, a disk resonator similar to that of Figure 2.3.g with diameter of 10μm, and

thickness of 500nm, will have a resonance frequency of 91MHz. Addition of a mass

density of 1 µg/cm2 on the surfaces of such resonator will result in a frequency shift of

~1.6MHz. Considering that typical measured mass sensitivities for quartz resonators is in

2 the order of 10-2000 Hz/ g/cm , the abovementioned silicon resonator has three to five

orders of magnitude larger mass sensitivity.

Also, it has been demonstrated experimentally that the developed

thermalpiezoresistive resonators are suitable for sensory applications with a relatively

high sensitivity [99,100,101]. Figure 2.5 shows the SEM view and frequency drift

characteristics of a similar 61MHz resonator subjected to bombardment by artificially

generated 1μm diameter particles. The extremely high mass sensitivity of 1.2 kHz/pg for

the resonator results in distinct frequency shift of ~850Hz for each particle sitting on the

resonator body.

Such mass sensitivity is more than an order of magnitude higher than that of the

most sensitive quartz resonators. In addition, there is plenty of room to conveniently

increase the mass sensitivity of such resonators by another 2-3 orders of magnitude by

shrinking their sizes.

Figure 2.5. (a) Change in the measured resonance frequency for the resonator of Fig. 2.3.f as a function of the overall exposure time to flow of particles. The plot shows four consecutive reduction steps with one of the steps having a slope twice the rest showing that two particles have been deposited during that specific interval (10-15min); (b) SEM view of the same resonator after particle deposition showing 5 spherical particles deposited on its two sensing plates. The particle on the support has negligible effect on the resonant frequency.

2.1.6 Resonator Static Temperature and its Effect on Biomolecules

The only concern expressed by the committee on this work is the possible effect

of the elevated static temperature of the resonators on the biochemical reactions and

species. Our response to this concern is Three-fold:

1. The resonators will be dormant (not actuated) during the biomolecule absorption process and are only turned on for readout after sufficient time has elapsed for the chemical reactions to take place;

2. Finite element thermal analysis presented in Figure 2.6 show that for the typical

operating conditions of thermal-piezo devices, the static temperature of the resonators are only elevated by 20-30 0C compared to the surroundings. The other issue that needs to be addressed is the harmful effect of the electrical signals required for operation of the

devices on the surrounding solution that can cause electrolysis of water and decomposition of biomolecules. Such temperatures can be easily tolerated by the absorbed molecules and do not present a potential risk for implementation of the demonstrated sensors.

Figure 2.6.COMSOL finite element thermal analysis of the resonator of Fig. 1, by applying DC and AC currents of 30mA and 5mA between the two pads on the two sides of the resonator. The color code shows static temperature of the structure resulting from the DC bias current that has a maximum value of only ~50ºC. The numbers show temperature fluctuation amplitude (Tac) at different points along the length of the actuator.

3. A very effective and convenient temperature compensation technique for silicon-based resonators is using degenerate doping levels (very high doping level) in the structural material. Resulting changes in the band structure of the semiconductor leads to a significantly suppressed temperature dependency for resonant frequency [102].

Furthermore, it has been observed that the temperature coefficient of frequency

(TCF) for thermally actuated resonators can be tuned by changing their operating bias current [103]. Using a combination of degenerate doping and bias current tuning, we 36

have achieved unprecedented and impressive TCF values as low -0.05ppm/0C thermalpiezoresistive resonators [103]. This would be equivalent to a frequency uncertainty of only ~4.5Hz if temperature control with an accuracy of 1ºC is provided for the fluidic sample, which is quite common for biological experiments.

Moreover, due to the tunability of resonator TCF with bias current, it is expected that even lower resonator TCF values can be realized via further characterization and optimization, relieving the temperature stability requirements. As part of this proposed effort we will target lower resonator TCF down to at least 0.01ppm/0C. Finally, for applications where very aggressive temperature compensation is required, having an integrated non-functionalized resonator next to the active resonator and monitoring the frequency difference between the two will highly suppress the effect of temperature fluctuations.

The other issue that needs to be addressed is the harmful effect of the electrical signals required for operation of the devices on the surrounding solution that can cause electrolysis of water and decomposition of biomolecules. Such temperatures can be easily tolerated by the absorbed molecules and do not present a potential risk for implementation of the proposed sensors. All the resonator conductive surfaces will be covered with a thin layer (20-100nm) of insulating silicon dioxide to completely eliminate such effects

2.1.7 Biochemical Sensing Mechanism

As illustrated in Figure 2.6.a a rotational mode disc resonator is functionalized with an antibody as the receptor Upon exposure of the surface to a biological sample 37

containing a variety of targets, only matching ones attach to the antigen binding site of immobilized antibody (Figure 2.7.b). The sensing mechanism is based on a mass-induced frequency shift due to the attached biochemical compounds (target) on the resonator surfaces (Figure 2.7.c).

(Receptor) (Target)

(a)

(c)

(b)

Figure 2.7. Biochemical Sensing Principals of the fabricated MEMS Resonator. (a) Functionalized resonator with antibody, (b) Specific target detection, (c) Frequency deviation of the second resonator due to the additional mass

Chapter Three: Biochemical Sensing Results and Discussion

3.1 Octadecylamine Sensing

As a proof of concept that fabricated thermally actuated MEMS resonators are capable of sensing label-free biochemical samples, octadecylamine has been used as the target. Octadecylamine is a straight chain of 18 carbon atoms with an amine-terminated active site. The amine group facilitates the amine-epoxide coupling process. Therefore, in this approach biosensor surfaces is functionalized with active epoxide groups as MRE

(receptor) first and octadecylamine (target) attachments take place after amine-epoxide coupling. As a result of these couplings, the frequency deviations of the sensors in respect to their reference frequency point proved the detection capability of the sensors due to the additional mass.

3.1.1 Surface Functionalization

In order to realize the compatibility of the mentioned sensing technique in section

2.2 for biorecognition applications, the surface of the resonators must be functionalized with a molecular recognition element (MRE). As discussed in section 1.4.2.1 the simplest method for deposition of a MRE onto a surface is simple adsorption, but there is always a desorption process which causes mass reduction from the surface. To eliminate this problem, we used chemical grafting, whereby the MRE is irreversibly attached to the

surface by chemical bonds. For silicon oxide surfaces, the needed chemistry is well established [104].

Silicon dioxide surfaces are best functionalized when using Si–O–Si bond- forming reactions of a variety of commercial precursors [105]. The most common method for chemical grafting of silanes to silicon oxide is to use functional trimethoxysilane derivatives such as 3-glycidoxypropyltrimethoxysilane, which has the reactive epoxide that couples with amino groups (Figure. 3.1) [106].

Figure 3.1. Surface functionalization with an epoxide allows for the attachment of a variety of MREs, R–NH2.

In order to surface linkage synthesis for octadecylamine sensing, the previously fabricated device surfaces were activated by oxidation in 1N HNO3 for 30-60 min followed by a rinse in DI water, acetone, and ethanol for 5 min each. Surface functionalization for growing the epoxide groups on silicon surfaces were done by soaking the chips in a solution consisting of 24 parts xylene, 8 parts 3- glycidoxypropyltrimethoxysilane and 1 part N,N-diisopropylethylamine at ~800C for 5 hours. (Figure 3.2.a). Excess reagents were removed by three washes in acetone and

ether. The first frequency measurement was performed in this step after drying the

resonators.

3.1.2 Octadecylamine Couplings

Therafter the functionalized biochips coupled with amine groups of

Octadecylamine MREs. The attachment was performed by immersing the biochips in

5mM Octadecylamine in ethanol at room temperature overnight (Figure 3.2.b). To

remove unreacted materials rinsed off in acetone for 5 minutes, consequently wth ether

(99.999 %) for 30 seconds and dried by ether quick evaporation.

3.1.3 Frequency Measurements

Device resonance frequencies were measured before and after octadecylamine

reaction in the air. Comparing the measured frequencies shows a clear and consistent

frequency shift of ~100ppm for 3 out of the 4 tested resonators in respect to the reference

frequency point (Table 3.1, Figure 3.3). The 4th resonator shows a larger frequency shift of 140ppm, which is suspected be due to addition of contaminants during the several chemical synthesis and rinsing steps, or absorption of airborne particles during exposure of the sample to air (e.g. during testing or transferring from one place to another).

Figure 3.3 shows the measured frequencies for a 30.4MHz resonator before and after octadecylamine attachment at different bias currents showing a consistent shift at each bias current due to the added mass.

3-Glycidoxypropyltrimethoxysilane Recognition Element (Receptor)

(a)

i-Pr2NEt, xylene , 800C

Activated Si Surface in HNO3 Epoxide Formation

(b) Octadecylamine in Ethanol, room temperature, 18hr Target

Addition of 4.5 × 10 g for each reaction with−22 Octadecylamine

Surface with increased tethered mass

Figure 3.2. Biosensors surface linkage synthetic scheme used for surface activation, functionalization with epoxide groups and octadecylamine coupling through amine- epoxide reaction.

Table 3.1 Summary of measurement results for different functionalized resonators

L , L W , ƒ (MHz) ƒ (MHz) 1 2, 1 1 2 Δƒ (Hz) Δƒ (ppm) W2,W3 (μm) Functionalized Octadecylamine 23,36,33,4,2 30.437427 30.434413 3014 99 23,36,33,4,2 29.891373 29.888322 3051 100

23,36,33,--, 2 23.809419 23.805919 3400 140 23,36,33,--, 2 27.001593 26.998914 2679 99

Figure 3.3. (a )SEM view of a 3-bar I-shaped resonator with different resonator dimensions labeled on it., (b) Consistent frequency shifts for different resonator bias currents before and after octadecylamine attachment.

3.1.4 Surface Coverage Analysis

Attachment of each epoxide group requires linking to three silicon atoms on the

surface (Figure 3.4). On the other hand during the surface fictionalization, each crystal

with a lattice constant of 5.43A0 consists of eight open bonds on 100 silicon surface

exposed to reaction reagents (Figure 3.5.a). This means that if all the available bonds are

used for linking with 100% packing, 8.8 epoxide groups will be available in every nm2 of silicon surface (Figure 3.5.b).

Figure 3.4. Each epoxide group links to the surface by bonding to three silicon open bonds on the surface

(a) (b)

Figure 3.5. (a) Model of a silicon crystal as seen on the 100 surfaces, the area of every single silicon crystal is equal to 0.3nm2; (b) 8 silicon atoms on the surface with open bonds in every single silicon crystal unit available for surface linking.

On the other hand, according to the device mass, reference frequency value and frequency shift after octadecylamine attachment, the added mass for the 1st measured resonator with the resonance frequency of 30.4 MHz is:

1 = = 1.0 2 ∆푓 ∆푚 − → ∆푚 푝푔 Therefore, considering푓 octadecylamine푚 molecular weight, the added 1.0pg of octadecylamine is equal to 2.2x109 added molecules on the surface. Based on the resonator surface area of 3.6x10-9m2, we will end up having 0.61 molecules in every nm square which is equivalent to 6.93% of the calculated theoretical limit of the added molecules in 1nm2.

3.1.5 XPS Surface Analysis

To make sure that the surface linking has resulted from the attachment of the expected molecules and groups to the silicon surface and that the measured resonator frequency shifts are not a result of measurement errors or contamination, a suitable surface analysis needed to be considered. According to our samples analytical area and desired detection range X-ray Photoelectron Spectroscopy (XPS) was utilized in every step of the surface chemistry (Figure 3.6). XPS is an analytical tool to quantitatively analyze the chemical and electronic states and elemental compositions of the materials. In this technique the output data forms a spectra obtained by irritating the top 1-10nm of the material surfaces with an X-ray beam and collecting the kinetic energy and number of electrons in an ultra high vacuum condition (Figure 3.7). In our case, the elemental

compositions on the surface support the addition of the desired molecules after each

reaction. After the XPS analysis, in silicon dioxide surfaces a considerable amount of

carbon was detected most likely due to the organic contaminants or residues (e.g.

solvents used to clean the surface or airborne particles) (Figure. 3.8.a). The carbon

concentration on the surface increased significantly after the surface modification by

linking the epoxide groups (Figure. 3.8.b). After octadecylamine attachment a small

amount of nitrogen (from the amine groups) could be detected on the surface while the carbon concentration further increased (Figure. 3.8.c).

Figure 3.6. Different surface chemistry analysis techniques in respect to the detection range, analytical area. [107] 46

(a)

(b)

Figure 3.7. X-ray Photoelectron Spectroscopy (XPS) (a) Facility at University of Maryland (b) Instrumentation principles implemented for elemental composition analysis of our biosensors

[108] 47

(a) SiO2

(b) Epoxide

Figure 3.8. XPS surface analysis results of the resonators surfaces confirms addition of the carbon and nitrogen containing amine groups to the surface in different reaction steps. (a) SiO2, (b) Epoxide, (c) Octadecylamine

3.2 Biotin-Avidin Conjugation Sensing

In the previous chapter it has been shown that the developed thermally actuated

MEMS resonators are capable of sensing biochemical analytes based on the proposed

surface chemistry. Since the ultimate goal is implementing the devices as microarray

platforms, sensing biomolecules such as protein becomes critical due to their large size,

low stability and various intermolecular forces. For this purpose avidin was considered for its high binding affinity to its ligand, biotin. The dissociation constant of avidin to

biotin is measured to be 10−15 M and it is one of the strongest known non-covalent bonds

[109]. Avidin is a tetrameric protein which contains four identical subunits each of which

can bind to biotin (Vitamin B7). (Figure. 3.9).

(a)

(b) (c)

Figure 3.9. (a) Biotin Structure (b) Avidin tertiary structure with its four subunits each holding a biotin (c) Schematic surface linkage synthesis for Biotin/Avidin Conjugation

3.2.1 Surface Biotinylation

The biochip surfaces were functionalized with epoxide groups through the same

method discussed in chapter 2.3. Thereafter the biotinylation were done in 0.5mM biotin

in ethanol overnight and samples were rinsed with ethanol for 5 minutes followed by a

minute rinsing in ether (Figure 3.10).

Figure 3.10. Biosensors surface modification processes used for functionalization with epoxide reactive groups to graft PEG -Biotin (amin-activated biotin) as the receptor 2

3.2.1.1 XPS Surface Analysis

For validating the surface chemistry and confirming the successful biotin-epoxide couplings, XPS analysis was performed. It was observred that the carbon and nitrogen peaks increased significanty for the biotinylated surfaces compared to the non- biotinylated blank SiO2. Also a sulfur peak was detected for the biotinylated samples due

to the presence of sulfur in the biotin chemical structure. A significant decrease in the

amount of silicon showed a surface coverage due to the chemical grafting and grown

molecules on the surfaces. (Figure 3.11.a)

SiO2 (Blank)

Biotinylated SiO2

Figure 3.11.a XPS surface analysis of the biosensors before and after surface 51 biotinylation process shows considerable amount of nitrogen, carbon and sulfur after biotin grafting. The lower Si2p implied a chemical coverage on top of the silicon dioxide surfaces.

3.2.1.2 Reaction Optimization

For optimizing the biotin reactivity toward epoxide, we had two functionalized

samples react with biotin in temperatures of 500C and 800C, overnight. The counted

atoms of Nitrogen, Carbon and Sulfur with XPS showed a significant improvement

regarding the biotin attachment in 800C (Figure 3.11.b).

Biotinylated SiO2

Epoxy-Biotin c/s Coupled 80 0C

Epoxy-Biotin Coupled 50 0C

S2s: 228.2(ev)

Figure 3.11.b Elevated amount of detected Sulfur shows the reaction optimization performed at 800C

3.2.2 Avidin Conjugation

The conjugation process for the biotinylated sensors was done in 0.5 µM avidin

solution in 10 mM HEPES buffer (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) pH 7.4 (Figure 3.12).

Figure 3.12 Surface linkage synthetic schemes used for Biotin-Avidin conjugation process. In this case avidin as our target couples with biotin as the receptor chemically grafted on the surface

3.2.2.1 Frequency Measurements

The frequency measurements were done before and after avidin conjugation. The results have shown non-consistent positive frequency deviations for three tested resonators. Also, after two rinsing process in DI water for 24 hours, the resonators frequency were measured for a second time and different frequency shifts obtained compared to the first set of acquired data. (Table3.2).

Table 3.2 Resonators frequency measurements after avidin conjugation and DI water rinsing processes

3.2.2.2 XPS Surface Analysis

XPS analysis was performed for the tested resonators before and after rinsing process. It has been observed that a considerable amount of nitrogen and carbon has been removed from the surface during the rinsing process (Figure 3.13).

Also, in a separate control experiment, XPS analysis for the native oxide silicon surfaces immersed in 0.5µM avidin solution showed a considerable peak of nitrogen and carbon. Also XPS from the resonator surfaces before and after rinsing process showed a significant decline in the nitrogen peak (Figure 3.14). By comparing the frequency measurements and XPS results, it was concluded that avidin electrostatic adsorption takes place on the biochip surfaces.

Avidin-Resonators

N1s C1s 7.3% 56.9%

Avidin-Resonators after Rinsing Process

N1s C1s 5.3% 54.9%

Figure 3.13. XPS analysis for the tested resonators before and after rinsing process 55

SiO2 (Control)

C1s 14.3%

SiO2- Avidin

C1s S2s N1s 48% 11% 0.1%

Figure 3.14. XPS surface analysis of the silicon dioxide surfaces before and after immersion in 56 avidin solution shows a significant amount of elevated nitrogen, carbon and sulfur.

3.3 Surface Passivation Methods and Characterization

According to the surface chemistry analysis and frequency measurements results,

biochip surfaces needs to be passivated against avidin physical adsorption. For this propose, two different techniques have been studied and characterized based on PLL-g-

PEG copolymer and non-reactive chemical groups such as propyle.

3.3.1 Polymer Chemistry

It has been shown that using a graft co-polymer of poly (L-lysine) and poly

(ethylene glycol), PLL-g-PEG, adsorbed on silicon dioxide surfaces generates a protein resistant coating [110]. This resistance depends on the molecular weight of PEG chains and the grafting ratio which is lysine units to PEG chains [111]. Figure 3.15 and 3.16 shows the molecular structure and passivation function of the copolymer.

Figure 3.15. PLL-PEG Structure used for surface passivation of the biosensor against proteins physical adsorption. 57

Figure 3.16. PLL-g-PEG schematic function as covers the negatively charged silicon dioxide surfaces against protein non-specific adsorption. As shown avidin only binds specifically to its receptor, biotin and the copolymer repels all other kinds of proteins. [112]

3 .3.1.1 Deposition Characterization Experiments:

PLL (20)-g[3.5]-PEG(2)/PEG(3.4)-Biotin (50%) [PLL (20 kDa) grafted with PLL

(2 kDa) and PEG-Biotin (3.4 kDa)], PLL (20)-g[3.5]- PEG(2)/TRIC (fluorescent red

label) from Surface Solution (Switzerland) and Alexa fluor® 488 Avidin from Invitrogen

(Carlsbad, CA) have been used in the experiments.

Preliminary experiments were done by choosing two types of silicon surfaces i.e.

native oxide and grown oxide. Fluorescent Microscopy showed an ineffective amount of

adsorbed polymer on the native-oxide silicon chips, however vice versa in the 1µm

grown oxide chips. Also the effect of surface cleaning in piranha solution (3:1 Sulfuric

Acid and Hydrogen Peroxide) and oxygen plasma treatment with Ion Reactive Etching

(RIE) before the polymer adsorption concluded to be positive. Therefore, an experimental

design was done based on the preliminary polymer characterization and fluorescent

analysis. The silicon chips had been wet oxidized for 2 hours in 11000C and thoroughly

cleaned with piranha solution for 30 minutes and treated by RIE oxygen plasma for 4

minutes before the polymer surface adsorption. To monitor both passivation and

conjugation process, different ratios of TRIC labeled and biotinylated PLL-g-PEG

mixtures were prepared (Table 3.3). The concentrations of the mixtures were considered

to be constant at 0.1mg/ml in 10 mM HEPES buffer (4-(2-hydroxyethyl)-1- piperazineethanesulfonic acid) pH 7.4 with the deposition time equal to 60 minutes. Also

Alexa fluor® 488 Avidin conjugation was completed in 30 minutes right after the

functionalization. All the experiments before fluorescent analysis were done in the clean room to prevent any type of contamination.

Table 3.3. PLL-PEG deposition characterization plan

3.3.1.2 Fluorescent Microscopy

Fluorescent microscopy was done for the TRIC and Alexa fluor® 488 labels with

the emission wavelength of 625nm and 488nm respectably. For two set of samples a 59

blank (non-reacted silicon chip) were used for imaging calibration. The imaging results

revealed a successful surface passivation in the sample functionalized with the 100%

TRIC-PLL-g-PEG. Also an increase in the amount of conjugated avidin has been

observed with adding the ratio of biotiniylated PLL-PEG to the mixtures (Figure 3.17).

TRIC labeled PLL-PEG (red) and PLL-PEG-Biotin Mixtures Passivation Blank 1:0 3:1 1:1 1:3 0:1

Alexa fluor® 488 Avidin (green) Blank 1:0 3:1 1:1 1:3 0:1

Figure 3.17. Fluorescent microscopy of the prepared samples based on the table 3.3. The intensity of 488nm emitted light increased due to the presence of Alexa fluor® 488 Avidin conjugated with biotin and increased in samples with higher ratios of PLL-PEG- Biotin in non-biotinylated copolymer i.e. TRIC labeled PLL-PEG (red).

3.3.2 Surface Chemistry

As discussed in the section 1.4.2.1 MREs immobilization were formed using the non-covalent attachment method were susceptible to removal from the surface under extreme conditions. However, developing covalent linkages using to self-assembled aminosilane monolayer films and heterobifunctional crosslinkers have been established

[65]. The reactive groups of the crosslinkers such as N- hydroxysuccinimidyl esters facilitate the reaction toward amine functional groups on the surface. Therefore

immobilization process consists of silanization (functionalization the surface with amine

groups), biotinylated crosslinker couplings and protein conjugation (Figure 3.18)

(a)

(b)

Figure 3.18. Heterobifunctional crosslinkers surface linkage synthesis used for surface biotinylation and passivation at the same time.

This reaction allows us to use non-reactive groups during the surface silanization to obtain the desired surface passivation. Therefore in the experiments, propyl groups have been used as a non-reactive group and different ratios of 0%, 0.1%, 1%, and 5% with amino reactive groups were obtained on the surface.

As a more detail description of the reactions, silicon dioxide (native oxide)

substrates were cleaned by immersion in 3:1 concentrated H2SO4:H2O2 for 30 minutes, 61

rinsed in deionized H2O and RIE oxygen plasma cleaned prior to silanization.

Thereafter, silanization of substrates was performed using 0, 0.1, 1 and 5% solutions of

distilled trimethoxysilylpropyldiethylenetriamine [DETA; Fisher Scientific] in 1 mM

acetic acid in DI water for 20 min at room temperature. Two sets of experiments were

designed to control the reaction steps and examine the surface passivation satisfaction. In

the control test, Alexa Fluor® 488 carboxylic acid succinimidyl ester was used instead of the crosslinker to monitor the reactivity of the silanized surfaces (Figure 3.19.a).

Fluorescent microscopy revealed that by increasing the amine concentrations in the

amine/propyl mixtures the intensity of the emitted Alexa Fluor® 488 dyes enhanced and

vice versa. This means the more amine groups grow on the surface, the more crosslinkers

are ready for the conjugation process. Therefore experiment (a) could proved the

reactivity of the amine toward the crosslinkers successfully.

In the surface passivation examination (experiment (b)) Alexa Fluor® 488

carboxylic acid succinimidyl ester replaced with Biotin N-hydroxysuccinimide ester

(Biotin-NHS). 1mM of the Biotin-NHS dissolved in 100 ml DMSO and diluted to 30 ml

in DMF. The substrates were silanized as in the experiment (a) and were immersed in the

crosslinker solution for 2 h at room temperature. Then it followed by rinsing in

DMSO/DMF and dried with nitrogen flow. The samples were conjugated in 0.5µM

Alexa Fluor® 488 Avidin solution for 30 minutes. After completion of the reaction,

fluorescent microscopy was done for the samples. The images showed the expected

emitted light due to the presence of Alexa Fluor® 488 Avidin conjugated with Biotin-

NHS. Also Images from all the samples showed an increase in the light intensity due to

increase amount of reactive amine groups in the mixture ratios of propyl:amine used for silanization. (Figure 3.19.b)

(a) Amin + carboxylic acid succinimidyl ester

Receptor Target

NH2:CH3 Blank 0% 0.1% 1% 5%

(b) Biotin N-hydroxysuccinimide ester +Alexa Fluor® 488 Avidin

Receptor Target Conjugated Avidin

Passivator Element

Avidin Conjugation

NH2:CH3 Blank 0% 0.1% 1% 5%

Figure 3.19. Fluorescent Microscopy of the: (a) Control Experiments (b) Passivation Experiment

Chapter Four: Conclusion, Discussion and Future Work

This work demonstrated formation of molecular monolayers on silicon surfaces of

micromechanical resonant devices and measurement of the mass of the deposited

monolayer using such. Thermally actuated single crystalline silicon micromechanical

resonators with resonant frequencies in the 20-30 MHz range were used as highly

sensitive mass sensors. Monolayers of long chain amine molecules were formed by

reacting amine groups with epoxide groups covalently bonded to the silicon surfaces. As

a result a consistent negative frequency shift of ~100ppm was measured for the resonators. Mass calculations based on the measured frequency shift show a 6.9% surface coverage for the monolayer. XPS surface analysis shows existence of nitrogen atoms on the resonator surface confirming that the measured frequency shifts are due to the addition of the amine molecules. For controlled immobilization of larger biomolecules such as proteins on the resonators, two surface passivation methods based on the polymer and surface chemistry were investigated. For the characterization and optimization of the proposed methods fluorescent microscopy was utilized.

4.1 Real Time Biochemical Sensing

For real time target sensing and characterizing, the same characterized methods can be applied to the resonators capable of working in the liquid. However, the main challenge for direct chemical detection in liquid media is the significant drop in the

resonator quality factors caused by the excessive viscous damping induced by the surrounding liquid. While quality factors in the few tens of thousands under vacuum and

few thousands in air can easily be achieved for such resonators, typical measured Q

values in liquid media are below 30 [113,114]. The highest micro-resonator Q value in

water reported so far is 94 [115]. We have recently demonstrated rotational mode disk

resonators capable of operation in liquid with quality factors as high as 304 [116] which

makes it reliable for measurements in the liquid.

(a) (b)

(c)

Figure 4.1. a) SEM view of a fabricated rotational mode disk resonator; b) COMSOL finite element modal analysis of the same resonator. Red and blue show the maximum and minimum vibration amplitudes respectively; c) Measured frequency response for the same resonator both in air and in water showing an unprecedented Q of 304 in liquid.

4.2 Array Implementation

The resonators can be encapsulated in 2D array structures with two electrical

interconnect layers to prevent the required large number of electrical connections (Figure

4.2.a). One of the challenges facing large sensor arrays with electrical readout is the large

number of electrical connections required to allow access to individual sensors in a large

array. To alleviate this issue, 2-dimentional array structures with two electrical interconnect layers can be used (Figure 4.2.b).

(a) (b)

Figure 4.2. (a) Schematic view of a disk resonator along with its electrical connections encapsulated in a microfluidic channel.

(b)Schematic view of the proposed 2- dimensional resonator arrays with isolated vertical and horizontal electrical interconnects and switches for resonator selection.

The patterned SOI device layer that also constitutes the resonator bodies will be

used as one of the layers and an isolated metal or doped polysilicon layer bridging over

the silicon interconnects will serve as the second layer. By using the proposed

configuration, an array of 2,500 resonators can be controlled using 100 control signals

each controlling one of the switches (external transistor switches on the readout circuitry,

or integrated MEMS switches).

4.3 Encapsulation in Micro-Fluidic Channels

To enhance the interaction between resonator surfaces and the solution, and to

protect the resonators from contaminants and particulates, encapsulation of the resonators

in each array in a single meander shaped micro-fluidic channel passing over all the resonators in the array can be effective (Figure. 4.2.b).

(a) (b)

Figure 4.3. SEM view of an encapsulated capacitive resonator resonator, b) cap openings covered with PECVD silicon dioxide, c) openings etched in oxide for electrical contacts, d) frequency response of the encapsulated resonator.

Injecting a sample solution through such channel provides the opportunity for every

molecule in the solution to come in close vicinity of every single resonator in the array

potentially getting absorbed by the immobilized matching MRE on any of the resonators.

A wide variety of well established micro-fluidic channel fabrication processes and materials are available [117,118,119] that will be considered for this task. Furthermore, a polysilicon-silicon dioxide encapsulation process for packaging of electrostatic silicon resonators (Figure 4.3) [120,121]. The same fabrication process can be used to form polysilicon caps, with small openings on top of the silicon thermal resonators. After functionalization of the resonators, the small openings can be covered by spin-coating with a viscous polymer. The polymer has to be viscous enough so that it does not diffuse into the channel while covering the narrow openings.

References

[1] A. de Fougerolles, H. P. Vornlocher, J. Maraganore and J. Lieberman, Nat. Rev. Drug Discov., 2007, 6, 443.

[2] A, B, (2002), Molecular Biology of the Cell, Garland Science,

[3] K. Y. Tomizaki, K. Usui and H. Mihara, Chembiochem, 2005, 6, 783.

[4] L. Melton, Nature, 2004, 429, 101.

[5] E. Jacoby, R. Bouhelal, M. Gerspacher, K. Seuwen, ChemMedChem 1, 761 (2006).

[6] G. MacBeath and S. L. Schreiber, Science, 2000, 289, 1760.

[7] H. Zhu, M. Bilgin, R. Bangham, D. Hall, A. Casamayor, P. Bertone, N. Lan, R. Jansen, S. Bidlingmaier, T. Houfek, T. Mitchell, P. Miller, R. A. Dean, M. Gerstein and M. Snyder, Science, 2001, 293, 2101.

[8] J. LaBaer and N. Ramachandran, Curr. Opin. Chem. Biol., 2005, 9, 14.

[9] http://research.ncku.edu.tw/re/commentary/e/20100108/1.html

[10] http://www.mun.ca/biology/desmid/brian/BIOL2060/BIOL2060-18/CB18.html

[11] http://evolution-textbook.org/content/free/figures/02_EVOW_Art/26_EVOW_CH02

[12] M. Dufva and C. B. V. Christensen, Expert Rev. Proteomic, 2005, 2, 41.

[13] M. M. Ling, C. Ricks and P. Lea, Expert Rev. Mol. Diagn., 2007, 7, 87.

[14] Hall et al., 2004; Zhu et al., 2001

[15] K. D. (2004). Microarray Technology. WILEY-VCH Verlag GmbH & Co. KGaA

[16] PhRMA 2002 (http:/ jwww.phrma.org)

[17] Business Week 2000 and PhRMA, 2002.

[18] NIH Consensus Statement on management of hepatitis C: 2002. NIH Consens State Sci Statements, 2002; 19: 1-46.

[19] World Health Organization, Hepatitis C Factsheet (http:/ jwww.who.intjinffsjenjfact164.html)

[20] NIH Consensus Statement on management of hepatitis C: 2002. NIH Consens State Sci Statements 2002; 19: 1-46.

[21] Delehanty, j. B., and Ligler, F. S., Anal. Chern. 2002, 74, 5681-5687.

[22] Materials Research to Meet 21" Century Defense Needs -Interim Report, National Research Council, USA. (http:/ jwww.nap.edujhtmljmaterials_researc hjinterim_report.pdf)

[23] IUPAC Compendium of Chemical Terminology 2nd Edition (1997). (1992), International Union of Pure and Applied Chemistry: Research Triangle Park, NC.

[24] James P. Chambers, et al, “Biosensor Recognition Elements”, Curr. Issues Mol. Biol. 10: 1–12.

[25] A. Wolter, R. Niessner and M. Seidel, Anal. Chem., 2007, 79, 4529.

[26] A. J. Nijdam, M. M. C. Cheng, D. H. Geho, R. Fedele, P. Herrmann, K. Killian, V. Espina, E. F. Petricoin, L. A. Liotta and M. Ferrari, Biomaterials, 2007, 28, 550.

[27] H. Zhu and M. Snyder, Curr. Opin. Chem. Biol., 2003, 7, 55.

[28] C. Steinhauer, A. Ressine, G. Marko-Varga, T. Laurell, C. A. K. Borrebaeck and C. Wingren, Anal. Biochem., 2005, 341, 204.

[29] B. T. Houseman, J. H. Huh, S. J. Kron and M. Mrksich, Nat. Biotechnol., 2002, 20, 270.

[30] J. C. Love, L. A. Estroff, J. K. Kriebel, R. G. Nuzzo and G. M. Whitesides,Chem. Rev., 2005, 105, 1103.

[31] L. Yan, W. T. S. Huck and G. M. Whitesides, J. Macromol. Sci.-Polym. Rev, 2004, C44, 175.

[32] Gates SM: Surface chemistry in the chemical vapor deposition of electronic materials. Chem. Rev. 96, 1519-1532 (1996).

[33] A. Talapatra, R. Rouse and G. Hardiman, “Protein microarrays: challenges and promises”, Pharmacogenomics (2002)

[34] http://www.patentlens.net

[35] D. Ham and R. M. Westervelt, “The Silicon that Moves and Feels SmallLiving Things”Harvard University, Cambridge, MA 02138

[36] http://www.tutorvista.com

[37] F. Rusmini, Z. Y. Zhong and J. Feijen, Biomacromolecules, 2007, 8,1775.

[38] R. Y. P. Lue, G. Y. J. Chen, Y. Hu, Q. Zhu and S. Q. Yao, J. Am. Chem.Soc., 2004, 126, 1055.

[39] K. Kato, H. Sato and H. Iwata, Langmuir, 2005, 21, 7071.

[40] A. Savchenko, E. Kashuba, V. Kashuba and B. Snopok, Anal. Chem.,2007, 79, 1349.

[41] T. H. Ha, S. O. Jung, J. M. Lee, K. Y. Lee, Y. Lee, J. S. Park and B. H. Chung, Anal. Chem., 2007, 79, 546.

[42] K. L. Brogan and M. H. Schoenfisch, Langmuir, 2005, 21, 3054.

[43] J. J. Davis, J. Tkac, S. Laurenson and P. K. Ferrigno, Anal. Chem., 2007, 79, 1089.

[44] D.J. (2009), Engineering the Bioelectronic Interface Applications to Analyte Biosensing and Protein Detection, Oxford, UK, Royal Society of Chemistry

[45] W. M. Gao, R. Kuick, R. P. Orchekowski, D. E. Misek, J. Qiu, A. K. Greenberg, W. N. Rom, D. E. Brenner, G. S. Omenn, B. B. Haab and S. M. Hanash, BMC Canc., 2005, 5.

[46] S. M (2005).Protein Microarrays. London.Jones and Bartlett Publisher

[47] www.affymetrix.com

[48] M. F. Templin, et al, “protein microarray technology”, TRENDS in Biotechnology Vol.20 No.4 April 2002

[49] R.S. (2009), Microarray Technology in Practice, Cambridge, U, Elsevier Inc.

[50] B. Liedberg, C. Nylander and I. Lundstrom, Sens. Actuat., 1983, 4, 299.

[51] Matthew A. Cooper “Optical biosensors in drug discovery” Nature Reviews Drug Discovery 1, 515-528 (July 2002)

[52] Cooper, M. A., Optical Biosensors in Drug Discovery.Nature Reviews, 2002. 1: pp. 515–528.

[53] Sparkman, O. David (2000). Mass spectrometry desk reference. Pittsburgh: Global View Pub. ISBN 0-9660813-2-3.

[54] X. B. Yu, D. K. Xu and Q. Cheng, Proteomics, 2006, 6, 5493.

[55] K. R. Kozak, M. W. Amneus, S. M. Pusey, F. Su, M. N. Luong, S. A. Luong, S. T. Reddy and R. Farias-Eisner, Proc. Natl. Acad. Sci. USA, 2003, 100, 12343.

[56] L. Favre-Kontula, Z. Johnson, T. Steinhoff, A. Frauenschuh, F. Vilbois and A. E. I. Proudfoot, J. Immunol. Methods, 2006, 317, 152.

[57] X. B. Yu, D. K. Xu and Q. Cheng, Proteomics, 2006, 6, 5493.

[58] http://www2.chemistry.msu.edu/faculty/reusch/VirtTxtJml/Spectrpy/MassSpec/mas pec1.htm

[59] G. Binnig and H. Rohrer, Rev. Mod. Phys., 1987, 59, 615.

[60] M. Lee, D. K. Kang, H. K. Yang, K. H. Park, S. Y. Choe, C. Kang, S. I. Chang, M. H. Han and I. C. Kang, Proteomics, 2006, 6, 1094.

[61] http://www.cds.caltech.edu/~murray/amwiki/index.php/Atomic_force_microscope

[62] Daniel J. MuÈller and Andreas Engel “Voltage and pH-induced Channel Closure of Protein OmpF Visualized by Atomic Force Microscopy” , J. Mol. Biol. (1999) 285, 1347-1351

[63] http://samlab.epfl.ch/page-15480-en.html

[64] K. Salaita, Y. H. Wang, J. Fragala, R. A. Vega, C. Liu and C. A. Mirkin, Angew. Chem.-Int. Ed., 2006, 45, 7220.

[65] C. N. LaFratta and D. R. Walt, Chem. Rev., 2008, 108, 614.

[66] D. J. Muller and Y. F. Dufrene, Nat. Nanotechnol., 2008, 3, 261.

[67] K. M. Goeders, J. S. Colton and L. A. Bottomley, Chem. Rev., 2008,108, 522.

[68] J. Fritz, M. K. Baller, H. P. Lang, H. Rothuizen, P. Vettiger, E. Meyer, H. J. Guntherodt, C. Gerber and J. K. Gimzewski, Science, 2000, 288, 316.

[69] G. H. Wu, R. H. Datar, K. M. Hansen, T. Thundat, R. J. Cote and A. Majumdar, Nat. Biotechnol., 2001, 19, 856.

[70] Thomas P. Burg, et al, “Weighing of biomolecules, single cells and single nanoparticles in fluid”, Vol 446-26 April 2007

[71] Mannelli, I.; Minunni, M.; Tombelli, S.; Mascini, M., “Bulk acoustic wave affinity biosensor for genetically modified organisms detection,” IEEE Sensors Journal, Vol. 3, Issue 4, Aug. 2003, pp.369-375.

[72] York, C.; French, L.A.; Millard, P.; Vetelino, J.F., “A lateral field excited acoustic wave biosensor,” 2005 IEEE Ultrasonics Symposium, Vol. 1, pp.44-49.

[73] Q-Sense Basic Training, April 4-5, 2006

[74] Rodriguez-Pardo, L.; Rodriguez, J.F.; Gabrielli, C.; Perrot, H.; Brendel, R., “Sensitivity, noise, and resolution in QCM sensors in liquid media,” IEEE Sensors Journal, Vol. 5, Issue 6, Dec. 2005, pp.1251–1257.

[75] http://nanotechweb.org

[76] Y. Cui, Q. Q.Wei, H. K. Park and C. M. Lieber, Science, 2001, 293, 1289.

[77] J. Ji, J. G. O’Connell, D. J. D. Carter and D. N. Larson, Anal. Chem., 2008, 80, 2491.

[78] A. K. Sinensky and A. M. Belcher, Nat. Nanotechnol., 2007, 2, 653.

[79] A. E. Grow, L. L. Wood, J. L. Claycomb and P. A. Thompson, J. Microbiol. Methods, 2003, 53,221.

[80] E. Hutter and J. H. Fendler, Adv. Mater., 2004, 16, 1685.

[81] Z. H. Wang, Y. H. Meng, P. Q. Ying, C. Qi and G. Jin, Electrophoresis, 2006, 27, 4078.

[82] M. Schaferling and S. Nagl, Anal. Bioanal. Chem., 2006, 385, 500.

[83] J. Wang, Small, 2005, 1, 1036.

[84] M. M. C. Cheng, G. Cuda, Y. L. Bunimovich, M. Gaspari, J. R. Heath, H. D. Hill, C. A. Mirkin, A. J. Nijdam, R. Terracciano, T. Thundat and M. Ferrari, Curr. Opin. Chem. Biol., 2006, 10, 11.

[85] X. B. Yu, D. K. Xu and Q. Cheng, Proteomics, 2006, 6, 5493.

[86] N. Ramachandran, D. N. Larson, P. R. H. Stark, E. Hainsworth and J. LaBaer, FEBS J., 2005, 272, 5412. 73

[87] B. P. Harrington, and R. Abdolvand, “Q-enhancement through minimization of acoustic energy radiation in micromachined lateral-mode resonators,” Proceedings, Transducers 2009, pp. 700-703.

[88] M. Rinaldi, C. Zuniga, C. Zuo, and G. Piazza, “Ultra-thin Super High Frequency two-port ALN contour-mode resonators and filters,” Proceedings, Transducers 2009, pp. 577-580.

[89] C. Nguyen, "MEMS technology for timing and frequency control," IEEE Trans. Ultrasonics, Ferroelectrics and Frequency Control, vol. 54, no. 2, pp. 251-270, Feb. 2007.

[90] G.K. Ho, K. Sundaresan, S. Pourkamali and F. Ayazi, "Low impedance, highly tunable, I2-resonators for temperature compensated reference oscillators,” proceedings, MEMS’05, pp. 116-120.

[91] S. Pourkamali, G. K. Ho, and F. Ayazi, “Low-Impedance VHF and UHF Capacitive Silicon Bulk Acoustic-Wave Resonators—Part II: Measurement and Characterization,” IEEE Transactions on Electron Devices, Volume 54, Issue 8, Aug. 2007, pp. 2024-2030.

[92] G.K. Ho, K. Sundaresan, S. Pourkamali and F. Ayazi, "Low impedance, highly tunable, resonators for temperature compensated reference oscillators,” proceedings, MEMS’05, pp. 116-120.

[93] S. Pourkamali, G. K. Ho, and F. Ayazi, “Low-Impedance VHF and UHF Capacitive Silicon Bulk Acoustic-Wave Resonators—Part II: Measurement and Characterization,” IEEE Transactions on Electron Devices, Volume 54, Issue 8, Aug. 2007, pp. 2024-2030.

[95] G.K. Ho, K. Sundaresan, S. Pourkamali and F. Ayazi, "Low impedance, highly tunable, I2-resonators for temperature compensated reference oscillators,” proceedings, MEMS’05, pp. 116-120.

[96] S. Pourkamali, G. K. Ho, and F. Ayazi, “Low-Impedance VHF and UHF Capacitive Silicon Bulk Acoustic-Wave Resonators—Part II: Measurement and Characterization,” IEEE Transactions on Electron Devices, Volume 54, Issue 8, Aug. 2007, pp. 2024-2030.

[97] A. Rahafrooz, and S. Pourkamali, “Characterization of Rotational Mode Disk Resonator Quality Factors in Liquid,” 2011 IEEE International Frequency Control Symposium (IFCS 2011)

[98] A. Rahafrooz, A. Hajjam, B. Tousifar, S. Pourkamali, “Thermal actuation, a suitable mechanism for high-frequency electromechanical resonators,” proceedings, IEEE MEMS conference, 2010, pp. 200-203.

[99] A. Hajjam, A. Rahafrooz, and S. Pourkamali, “Thermally actuated low impedance MEMS Resonators for Mass Sensing Applications,” proceedings, IEEE Electron Devices and Solid State Circuits Conference (EDSSC), 2009.

[100] A. Hajjam, J. C. Wilson, A. Rahafrooz and S. Pourkamali, “Fabrication and characterization of resonant aerosol particle mass sensors,” proceedings, IEEE MEMS Conference, 2010, pp.863-866.

[101] A. Hajjam, A. Rahafrooz, J. C. Wilson, and S. Pourkamali, “Thermally actuated MEMS resonant sensors for mass measurement of micro/nanoscale aerosol particles,” proceedings, IEEE Sensors Conference, 2009.

[102] A. Samarao, and F. Ayazi, “Temperature Compensation of Silicon Micromechanical Resonators via Degenerate Doping,” Proc., IEEE International Electron Device Meeting, pp. 789-792, Dec. 2009.

[103] A. Hajjam, A. Rahafrooz, and S. Pourkamali, “Sub-100ppb/ºC Temperature Stability in Thermally Actuated High Frequency Silicon Resonators via Degenerate Phosphorous Doping and Bias Current Optimization,” proceedings, IEEE International Electron Device Meeting (IEDM), 2010.

[104] Van Der Voort, P., and Vansant, E. F. (1996) Silylation of the silica surface a review, Journal of Liquid Chromatography & Related Technologies 19, 2723- 2752.

[105] Yoshida, W., Castro, R. P., Jou, J. D., and Cohen, Y. (2001) Multilayer alkoxysilane silylation of oxide surfaces, Langmuir 17, 5882-5888.

[106] Ducker, R. E., Montague, M. T., and Leggett, G. J. (2008) A comparative investigation of methods for protein immobilization on self-assembled monolayers using glutaraldehyde, carbodiimide, and anhydride reagents, Biointerphases 3, 59-65.

[107] http://www.eaglabs.com/solutions/application/thinfilmanalysis.php

[108] http://www.chem.umd.edu/research/sharedinstrumentation/ xrayphotoelectronspectroscopy

[109] Green, N. (1963). The Use of [14C] Biotin for Kinetic Studies and for Assay. Biochem J, 89, 585-591 75

[110] Elbert, D.L., “Self-assembly and steric stabilization at heterogeneous, biological surfaces using adsorbing block copolymers”, Chemistry and Biology 1998, 5 (3), 177-183

[111] Kenausis, G.L. et al., “Poly(L-lysine)-g-poly(ethylene glycol) layers on metal oxide surfaces: Attachment mechanism and effects of polymer architecture on resistance to protein adsorption”, J. Phys. Chem. B 2000, 104, 3298-3309.

[112] Ruiz-Taylor L A et al. PNAS 2001;98:852-857

[113] C. Vancura, et al, “Fully integrated CMOS resonant cantilever sensor for biochemical detection in liquid environments,” proceedings, Transducers 2005, pp. 640-643.

[114] O. Ducloux, et al, “MEMS resonators actuated by TbCo/ FeCo nanostructured magnetostrictive multilayers in liquid environments,” proceedings, Transducers 2009.

[115] J. H. Seo, and O. Brand, “High Q-factor in-plane-mode resonant microsensor platform for gaseous/liquid environment,” Journal of Micro Electro Mechanical Systems, Vol. 17, No. 2, April 2008, pp. 483-493.

[116] A. Rahafrooz, and S. Pourkamali, “Rotational Mode Disk Resonators for High-Q Operation in Liquid,” IEEE Sensors Conference, 2010.

[117] S. M. Kim, S. H. Leeb and K. Y. Suh, “Cell research with physically modified microfluidic channels: A review”, Lab Chip, 2008, 8, 1015–1023 | 1015.

[118] E. Kjeang, N. Djilali, D. Sinton, “Microfluidic fuel cells: A review”, Journal of Power Sources 186 (2009) 353–369.

[119] B. D. Iverson, S. V. Garimella, “Recent advances in microscale pumping technologies: a review and evaluation”, Microfluid Nanofluid (2008) 5:145–174.

[120] Siavash Pourkamali, “High frequency capacitive single crystal silicon resonators and coupled resonator systems,” PhD Thesis, Georgia Institute of Technology, August 2006.

[121] S. Pourkamali, and F. Ayazi, “Wafer-Level Encapsulation and Sealing of Electrostatic HARPSS Transducers,” proceedings, 2007 IEEE Sensors Conference, pp. 49-52.