On-chip biosensing platforms based on gold and silicon optical nano-resonators

On-Chip Biosensing Platforms based on Gold and Silicon Optical Nano-Resonators Özlem Yavaş Supervisor: Prof. Romain Quidant Co-supervisor: Dr. Vanesa Sanz Beltran Plasmon Nano-Optics Group This dissertation is submitted for a degree of Philosophy Doctor UPC-PMT, Castelldefels February 2019 ii iii to my family iv v Acknowledgements "If I have seen further it is by standing on the shoulders of Giants." Isaac Newton Taking credit for this PhD thesis all by myself would be tantamount to hypocrisy. There are many people listed below, who have been crucially important for the success of the projects I finished during my PhD. In my opinion, if you talk about success, you have to also talk about luck. Below lies the proof of how lucky I was. First of all, I want to thank Prof. Romain Quidant for being an extremely positive, motivating, understanding and trusting group leader. With his support, even the hardest problems ahead became achievable. His kindness and optimism always inspired me (and it still does) and I consider myself extremely lucky to have worked with Romain as my PhD supervisor. Thank you! I want to also thank my co-supervisor Dr. Vanesa Sanz who has always been supportive and comforting in the times of desperation in the lab. She developed the surface functionalization protocols I used for LSPR experiments and I am very happy to have had her around during my PhD and learn from her. She set a great example of hardworking, dedicated scientist for me. Besides my official supervisors Romain and Vanesa, I was extremely lucky to have Dr. Mikael Svedendahl’s guidance during the last years of my PhD. All the Comsol simulations presented in this thesis have been done by Mikael, but that is only the tip of the iceberg showing his support and help. He is an exceptionally smart, efficient and kind person who always supported and encouraged me. I have to mention that if I had seen further, it is by standing on the shoulders of Dr. Maria Alejandra Ortega, Dr. Srdjan Acimovic, Dr. Johann Berthelot and other former members of PNO who worked on the biosensing projects. Their efforts in the project before I have started my PhD made my work easier. I had the privilege to build on what they have started. I thank to Mariale for training me on her last weeks at ICFO as a PhD student. I was lucky to have her around in the very beginning. I thank Srdjan, for all the useful discussions on Skype and always motivating and encouraging me about the LSPR project. I thank Dr. Paulina Dobosz for keeping my curiosity alive and exciting me for going back to the lab, whenever I needed a push. I learned a lot from her. She had been a great friend and supportive colleague. I also owe her all of my 4 hangovers in Barcelona during the last 4 years. I also thank Dr. Luis Miguel Fidalgo for all the useful discussions on micro fabrication. He was always able to see the big picture and motivate me. I thank Jose Garcia Guirado, my officemate and labmate for his help and training in the cleanroom. I appreciate the entertainment he provided by his ‘’only jose’’ stickers he puts on everything. I am also thankful to him for keeping me young and fresh in a freezing office. There was no winner of the AC wars during the last 5 years, only constant sore throat on my side and constant sweating on Jose’s. Acknowledgements vi I need to express my gratitude to the whole PNO group for keeping a functional work environment and their support. I especially thank Alexia Stollman for reading my thesis rigorously and helping me to improve it. I thank to Dr. Jil Schwender for being the nicest and sweetest alarm clock I have ever had. Starting every day with a friendly chat and coffee together made getting up easier. She had been an incredible friend the whole time and we laughed and cried together along the way. Thank you girl! Dr. Michal Tomza deserves to be mentioned here as he had been a great friend, a psychologist, a cook, a running trainer, a tour guide, and a tubing cutter for me! He had been my neighbor, my best friend and he even acted like a parent when I needed one around. He deserves to be proud of this thesis. And yes, at times, he cut tygon tubings in the lab for me, so that I can do my experiments without focusing on this menial task. He claimed to do it because it helped him take a break from his work, but I think he actually liked our long gossips. Out of all of the people I met at ICFO, I have to thank Adeel Afridi the most, because he is a jealous and possessive friend who would not forgive me if I don’t. I don’t know how he will continue his PhD without me. I am so grateful for our ridiculous discussions about very philosophical subjects. I honestly don’t know who will answer my hardest questions in life if I don’t have him around on a daily basis. My friends, scattered to different countries, receive my constant gratitude not only for the support with my PhD thesis, but also for their love and friendship. I don’t have any siblings, but I know that if I had sisters, they would be just like Ece, Özgün and Burcu! They always had time for me when I needed to chat. By being there for me no matter what, they made distances small, goals achievable and life brighter. I thank all of them one by one. Finally and most importantly, I am grateful to my wonderful family. I thank my parents for bearing with me for 30 years, for being supportive, loving and caring. I thank my mom for being the most wonderful and giving person that she is, for being my best friend, and for all the wonderful food that she stuffed in my freezer whenever she visited me! I thank my father for only caring about my happiness; always giving the priority to me being happy. I know that me getting a PhD in physics means more to him than to anybody else in the world. However, if today I decided to quit, he would support me regardless. Knowing this keeps me going. Teşekkürler bikilerim! I thank my partner, Tomek, for the warmth he continuously provides. I am grateful to him for believing in my success, contributing to my growth and bringing food to ICFO on the days I had to work until late. I don’t know how I got so lucky to have him and his loving family by my side. You see, I was not alone on this path and I am very grateful for all those wonderful people and many more that I could not refer here. I am the luckiest person in the world. And I also worked hard for this PhD. ;) vii Abstract Point-of-care (POC) devices are compact, mobile and fast detection platforms expected to advance early diagnosis, treatment monitoring and personalized healthcare, and revolutionize today’s healthcare system, especially in remote areas. The need for POC devices strongly drives the development of novel biosensor technology. Building a small, fast, simple, and sensitive platform for biomolecule detection is a challenge that relies on the integration of multiple fields of expertise and engineering. Optical nanoresonators have shown great promise as label-free biosensors because of direct light coupling and sub-wavelength sensing modes. Metallic nanoresonators with localized surface plasmon resonances (LSPR) are already well studied and were proven a solid alternative to the commercialized surface plasmon resonance (SPR) sensors. More recently, dielectric nanoresonators have also gained traction due to the reduced losses and the ability to manipulate both the electric and magnetic components of the incident light. In this thesis, we advance the field of biosensing and use optical nanoresonators as operative platforms relevant for disease diagnosis and treatment monitoring. By combining different optimized optical nanoresonators, both metallic and dielectric, with state-of-the-art microfluidics and surface chemistry, we have developed and tested several detection platforms. We first focused on developing a microfluidic lab-on-chip device for multiplexed biosensing utilizing the LSPR of gold nanoresonator arrays. By simultaneously tracking the extinction of 32 sensor arrays, we demonstrated multiplexed quantitative detection of four breast cancer markers in human serum. We showed that with well-optimized immunoassays, a low limit of detection (LOD) can be reached, paving the way towards clinically-relevant POC devices. Additionally, we implemented silicon nanoresonators supporting Mie resonances into functional and clinically-relevant applications. By integrating several arrays of Si nanoresonators with state- of-the-art microfluidics, we demonstrated their ability to detect cancer markers in human serum with high sensitivity and high specificity. Furthermore, we showed that the fabrication of Si nanoresonator array using low cost and scalable projection lithography leads to sufficiently low limits of detection, while enabling cheaper and faster sensor production for future POC applications. We also investigated the respective role of electric and magnetic dipole resonances and showed that they are associated with two different transduction mechanisms: resonance redshift and extinction decrease. Our work advances the development of future point-of-care sensing platforms for fast and low cost health monitoring at the molecular scale. viii ix Table of Contents Acknowledgements .......................................................................................................................... v Abstract .......................................................................................................................................... vii List of Figures ................................................................................................................................... xi List of Tables .................................................................................................................................. xiii INTRODUCTION ................................................................................................................................ 1 1 CONCEPTS AND BACKGROUND ................................................................................................ 5 1.1 Light-nanoparticle interaction ................................................................................................ 5 1.1.1 Resonances of metallic nanoresonators ............................................................................. 8 1.1.2 Resonances of dielectric nanoresonators ..................................................................... 10 1.2 Sensing with optical nanoresonators ................................................................................... 12 1.2.1 Sensing with metallic nanoresonators (LSPR) ............................................................... 14 1.2.2 Sensing with dielectric nanoresonators ........................................................................ 15 1.3 Microfluidics for Lab-on-chip biosensing platforms ............................................................. 16 1.4 Surface Chemistry ................................................................................................................. 19 1.4.1 Chip binding ................................................................................................................... 19 1.4.2 Sensor functionalization ................................................................................................ 20 2 METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS .......................... 23 2.1 Fabrication of Nanoresonators for Biosensing ..................................................................... 23 2.1.1 Fabrication of gold NRs ................................................................................................. 24 2.1.2 Fabrication of silicon sensors ........................................................................................ 26 2.1.3 Sensor preparation and surface chemistry ................................................................... 28 2.2 Microfluidic chip fabrication ................................................................................................ 29 2.2.1 UV photolithography for mold fabrication .................................................................... 29 2.2.2 Multilayer Soft Lithography for PDMS microfluidics ..................................................... 30 2.3 LOC Assembly ....................................................................................................................... 31 2.4 Opto-fluidic setup ................................................................................................................. 32 2.5 Immunoassays for on-chip biosensing ................................................................................. 34 2.5.1 Cancer marker detection with Au NR sensors ............................................................... 35 2.5.2 Cancer marker detection with Si NR sensors ................................................................ 35 2.6 ELISA protocol ....................................................................................................................... 36 2.7 Data processing and Analytics .............................................................................................. 37 INTRODUCTION 2 complex problem requires researchers to combine multiple disciplines and technologies. Especially, keeping in mind the necessity of performing diagnostics fast and in the vicinity of the patients in critical cases, lots of effort needs to be put into this research with the goal of moving towards specified POC devices for certain types of samples and diseases. Lab-on-chip (LOC) technology is one of the key elements that reduces the cost, size and reagent volume of POC devices. It enables the automation of the multiple analytical steps of a diagnostic assay and offers user-friendly and mobile devices. It is a strong enabling tool, developed over the last few decades, both in research and for the industry. It has already achieved successful commercialization for certain applications such as ELISA with integrated sample processing and glucose monitoring. Some microfluidic chips that can perform sample preparation, cell lysis, purification and many other laborious analytical steps already exist, proving that an entire laboratory can fit into of a few square centimetres. The development of such industrial chips, which can perform tens of different functions, relies on the development of small modules with separate single functions that are then combined to form a larger more complex system. This has the significant advantage of allowing researchers to develop and test a single function at a time. Working on simpler modules allows faster design iteration, with chips tested continuously in a chip-in-a-lab manner, i.e. placing them in bulky custom-made control and read-out set-ups for quick, customized and specific testing. Then, these separately-tested individual modules are integrated to form the complex and ultimately multifunctional industrial chips for direct use in POC devices. Therefore, even though current academic efforts in the direction of POC LOC system developments look more like chip-in-a-lab then labon-a-chip, these steps are crucial in the development of this technology for real life applications. Another key element of POC devices is the biosensing mechanism, which is responsible for the detection of the specifically targeted biomarkers of the POC device, and ultimately determines the device sensitivity. It essentially transduces the biochemical information into a detectable signal. For clinical diagnostics, the gold standard biosensing technique is enzyme linked immunosorbent assay (ELISA). This method, used both in research and by large analytical laboratories, requires using bulky well-plates and large amounts (microliters to millilitres) of samples and reagents as well as cumbersome plate readers. It is clearly not a method suitable for POC diagnostics. The chemistry used for ELISA relies on multiple incubation steps of labelled or conjugated antibodies and substrate solutions. The available biosensing transduction methods that can be combined with LOC technologies are based on fluorescence microscopy or impedance measurements. Optical detection schemes INTRODUCTION 3 hold great advantages, such as the possibility of label-free measurements, ease of performance, and scalability. Surface Plasmon Resonance (SPR) is an optical phenomenon that is highly sensitive to events on the surface of the metals, making it an interesting tool for biosensing. SPR sensing is based on exciting surface plasmon resonances at the interface of the metal and its surrounding environment, and therefore requires a specific optical configuration. This requirement is not optimal with the portability and small size expected from POC devices. A simpler method involves using a localised surface plasmon resonance (LSPR). This resonance is excited by incident light interacting with subwavelength scale metallic nanostructures, as opposed to thin films used for SPR-based sensors. LSPR is very sensitive to changes in its environment, allowing a simpler sensing setup. It is therefore not surprising that LSPR phenomenon has already been extensively studied and has already been used for biosensing applications. Its integration with microfluidic chips however, was relatively immature at the beginning of this work since most studies lacked clinically-relevant results and/or fully functional, practical and an automatable chip design for multiple analyte detection. More recently, dielectric nanostructures with a high refractive index have attracted attention due to their ability to manipulate the electric and magnetic components of the incident light. They are proposed as an alternative to metallic nanostructures for SPR sensing because they do not suffer from ohmic losses and provide better quality factor resonances with unique optical properties allowing the engineering of meta-properties. All-dielectric nanophotonics is a rapidly growing field for a number of different applications. However, there are only a few studies, so far, in the biosensing field. They all lack realistic and clinically-relevant experiments, leaving a wide gap in the literature, with a significant potential impact. The goal of this thesis work is to develop on-chip biosensing platforms that integrate microfluidics, optical nanoresonators, and relevant surface chemistry achieving robust, sensitive, specific, rapid, and real-time detection of clinicallyrelevant biomarkers. The first objective of the project is to develop an LSPR based microfluidic platform for the multiplexed detection of cancer markers in human serum. The second objective is to explore and develop sensing with dielectric nanoresonators, as opposed to the classically-used metal nanostructures for onchip biosensing applications. The multidisciplinary nature of this work integrates different disciplines of science and technology. In Chapter 1, the different concepts involved in this research are INTRODUCTION 4 explained. These include optical nanoresonators, microfluidics, and the surface chemistry used in chip fabrication and for biosensing. Chapter 2 presents the experimental techniques that are used to manufacture and operate the lab-on-chip platforms, integrate the sensors, and interpret the experimental results. Chapter 3 focuses on the development of the LOC platform with LSPR based sensors for multiplexed and fully-automated biosensing. The detection of four breast cancer markers in human serum is demonstrated. In addition, the chip design and system optimization are shown, as well as a comparison with the gold standard ELISA technique. Chapter 4 presents the results of on-chip sensing experiments with periodic silicon nanoresonator arrays fabricated by electron beam lithography. The electric dipole resonances of the nanoresonators and the diffractive modes of the periodic nanoresonator arrays are demonstrated to be effective for sensing. Results from cancer marker detection experiments are also presented and compared with the results obtained for the same bioassays using LSPR sensors. Chapter 5 focuses on semi-random silicon nanoresonator arrays to explore the different transduction mechanisms of the nanoresonators. These resonators are no longer periodic and do no longer need to be fabricated by long and costly electron beam lithography. The different transduction mechanisms are associated with different resonance modes through simulations and their sensing performance is then demonstrated for cancer biomarker detection. 5 1 CONCEPTS AND BACKGROUND The sensing platforms developed during this PhD work are highly multidisciplinary, integrating fields from nanophotonics and microfluidics to surface and biochemistry. The plasmonic nanostructures are used for optical sensing. Microfluidics is used to deliver samples and target molecules to the sensors, as well as sensor preparation steps in a controlled environment. Surface chemistry is used to prepare the sensors to capture the target molecules and run the immunoassays. In this chapter, I introduce the basic concepts that are involved in this dissertation, starting with presenting the basics of resonant light scattering nanostructures in both metallic and dielectric nanoparticles. Next, I focus on microfluidics, followed by the surface chemistry that was used in our experimental scheme. 1.1 Light-nanoparticle interaction Light scattering by small particles is a well-known and studied phenomenon, historically originating from the work of Lord Rayleigh in 1871.1 When light is incident on a small particle, the electric charges in the particle are displaced by the electric field of the incident light, setting them into oscillatory motion. These accelerating charges radiate electromagnetic energy, which is known as scattering. In addition to scattering, some of the incident energy is transformed into heat, resulting in absorption. Therefore, the incident light is extinct due to both scattering from and absorption by the particle. Both of these contributions depend significantly on the properties of the particle material and the surrounding media, as will be discussed in this chapter. Mie theory, developed by Gustav Mie in 1908, describes the electromagnetic field scattered and absorbed by a homogeneous and isotropic sphere of an arbitrary radius a, due to an incident plane wave of wavelength λ. Mie theory involves expanding electromagnetic fields into spherical harmonics and determining their expansion coefficients using boundary conditions. According to Mie theory2 , the CONCEPTS AND BACKGROUND 6 scattering, extinction and absorption cross-sections of the particle can be expressed as follows: 𝐶𝑠𝑐 𝑀𝑖𝑒 =𝜆2 2𝜋∑(2𝑛+1){|𝑎𝑛|2+|𝑏𝑛|2} ∞ 𝑛=1 , 𝐶𝑒𝑥𝑡 𝑀𝑖𝑒 =𝜆2 2𝜋∑(2𝑛+1)𝑅𝑒{𝑎𝑛+𝑏𝑛}, ∞ 𝑛=1 𝐶𝑎𝑏𝑠 𝑀𝑖𝑒 =𝐶𝑒𝑥𝑡 −𝐶𝑠𝑐, (1) where the expansion coefficients a and b are2: 𝑎𝑛=𝑚𝜓𝑛(𝑚𝑥)𝜓𝑛 ′(𝑥)−𝜓𝑛𝜓𝑛 ′(𝑚𝑥) 𝑚𝜓𝑛(𝑚𝑥)𝜉𝑛 ′(𝑥)−𝜉𝑛(𝑥)𝜓𝑛 ′(𝑚𝑥) 𝑏𝑛=𝜓𝑛(𝑚𝑥)𝜓𝑛 ′(𝑥)−𝑚𝜓𝑛𝜓𝑛 ′(𝑚𝑥) 𝜓𝑛(𝑚𝑥)𝜉𝑛 ′(𝑥)−𝑚𝜉𝑛(𝑥)𝜓𝑛 ′(𝑚𝑥) (2) 𝜓 and 𝜉 are related to Bessel functions of the first kind. 𝑚 is the relative refractive index 𝑚=𝑛𝑝𝑎𝑟𝑡𝑖𝑐𝑙𝑒 𝑛𝑚𝑒𝑑𝑖𝑢𝑚, and 𝑥 is the size factor 𝑥=𝑘𝑎, where k is the wave vector in the medium. These cross sections are valid for particles of all materials and sizes. Let’s consider the example where a homogeneous subwavelength sphere with dielectric function 𝜖(𝜔) and radius 𝑎 (𝑎≪𝜆) is placed into a homogeneous medium with dielecric constant εm. As𝜖𝑚. As the particle size is much smaller than the wavelength of the incident light, the phase of the harmonically oscillating electromagnetic field can be considered constant across the particle volume (𝐸 = 𝐸0𝑧). This reduces the problem into one with a particle in an electrostatic field (Figure 1-1). CONCEPTS AND BACKGROUND 7 Figure 1-1 Sketch of a spherical nanoparticle placed into an electrostatic field (modified from S.A.Maier3). Solving the Laplace equation and applying the boundary conditions, the electric field inside and outside the particle is found to be: 𝐸𝑖𝑛 =3𝜖𝑚 𝜖+2𝜖𝑚𝐸0, 𝐸𝑜𝑢𝑡 =𝐸0+3𝑛(𝑛 .𝑝)−𝑝 4𝜋𝜖0𝜖𝑚(1 𝑟3), (3) where 𝑛 is the normal unit vector pointing the location where the field is calculated, 𝑝= 𝜖0 𝜖𝑚𝛼𝐸0 is the dipole moment of the particle and 𝑟 is the distance from the center of the sphere to the point where the field is calculated. The polarizability of the particle is then3: 𝛼=4𝜋𝑎3𝜖−𝜖𝑚 𝜖+2𝜖𝑚, (4) From this equation one can see that when |𝜀+ 2 𝜀𝑚| reaches a minimum, the polarizability experiences a resonant enhancement. The scattering and absorption cross sections of the particle are given by3: 𝐶𝑠𝑐 =𝑘4 6𝜋|𝛼|2=8𝜋 3𝑘4𝑎6|𝜖−𝜖𝑚 𝜖+2𝜖𝑚|2, 𝐶𝑎𝑏𝑠 =𝑘𝐼𝑚[𝛼]=4𝜋𝑘 𝑎3𝐼𝑚[𝜖−𝜖𝑚 𝜖+2𝜖𝑚]. (5) Note that the same results in Eq. 5 can be derived from the general Mie theory cross sections by retaining only the first term (by coefficients 𝑎1 and 𝑏1), which represents the dipole mode and neglecting the other terms. 𝜖𝑚𝑎 𝜖 𝐸0𝑧 𝑟 CONCEPTS AND BACKGROUND 8 It is also interesting to note that for small particles absorption is the dominating mechanism, with its cross-section, scaling with 𝛼3, as opposed to 𝑎6 for scattering. As the size of the particle increases, the dipole resonance redshifts due to the weakening of the restoring force caused by polarized charges in the particle. The bigger the particle, the more separated the charges at the opposite ends of the particle are. This results in a decrease of their interaction, shifting the resonance to longer wavelengths. The resonance behavior of a spherical particle is very well known and used as an example for its simplicity. The shape of the particle however also heavily influences this resonant behavior. The more complex the shape of the nanostructure, the more complex its optical response will be. These responses are then usually simulated numerically, using models and different software, because there is no analytical solution to Maxwell’s equations for them. It can be seen from Eq. 5 that the extinction (𝐶𝑒𝑥𝑡 =𝐶𝑠𝑐 +𝐶𝑎𝑏𝑠) experiences a resonance also when the polarizability is resonant, thus, it also depends on the dielectric function of the medium. This is the reason why these nanostructures make good sensors and the key takeaway message of this section: any changes in a nanostructures’ dielectric environment can be directly correlated to changes in its’ optical behavior. So far, the equations in this section are valid for nanoresonators of any material. From now on, we will focus on metallic and dielectric nanoresonators only, as these are the materials used in the sensors developed in this thesis. The difference between the dielectric and metallic nanostructures stems from their different dielectric functions and the existence of the free electrons in metals as opposed to dielectric materials. 1.1.1 Resonances of metallic nanoresonators The free electrons in the metals oscillate around the fixed positively charged background due to the electric field of the incident light, resulting in electromagnetic excitations confined on the surface of the metal called surface plasmons. For propagating plasmons, matching the momentum of the incident field to the one of the surface plasmons requires special geometries4–6 which can be complex, whereas in subwavelength nanoparticles direct illumination is able to excite resonances once the resonance condition, shown above, is satisfied. These resonances of the metallic nanostructures are called localized surface plasmon resonances (LSPR). Metallic nanoresonators such as gold and silver can exhibit CONCEPTS AND BACKGROUND 9 their resonances in the visible light spectrum, making them attractive and practical candidates for many applications in biomedicine and sensing.7–10 Besides this, these nanoparticles highly localize and enhance the electromagnetic fields around them in the spatial ranges that match the sizes of biomolecules such as proteins and antibodies (mode-analyte overlap). This is why these particles are very sensitive to molecules in their vicinity. Drude model approximates the dielectric function of a metal: 𝜖𝐷𝑟𝑢𝑑𝑒(𝜔)=1− 𝜔𝑝 2 𝜔2+𝑖𝛾𝜔, (6) where 𝜔𝑝=√𝑛𝑒𝑒2/(𝑚𝑒𝜖0) is the plasma frequency, 𝑛𝑒 is the electron density of the metal and 𝑒 and 𝑚𝑒 are the charge and mass of an electron. 𝛾 is the damping term and 𝜔 is the frequency of the incident light. The wavelength dependence of the dielectric function of gold is shown in Figure 1-2. One can see that the imaginary part of the dielectric function is non-zero, which is a damping factor of the oscillations, indicating losses. These result in a widening of the resonance peaks, increasing the full width at half maximum (𝑤=𝐹𝑊𝐻𝑀) which corresponds to a lowering of the quality factor (=𝜆𝑟𝑒𝑠/𝑤). Figure 1-2 Dielectric function of gold calculated by Drude model. The values for gold taken from [Johnson and Christy, 1972].11 LSPR is a very well-studied phenomenon which is already being used in some applications in the fields of biosensing, cancer treatment, SERS and imaging.7–10,12 The following sections and chapters of this thesis focus on the biosensing applications of LSPR phenomena. 200 400 600 800 1000 -60 -40 -20 0 1 2 3 4 5 Wavelength(nm) Re Im ϵDrude CONCEPTS AND BACKGROUND 10 1.1.2 Resonances of dielectric nanoresonators Even though it has been so many years since the Mie theory completely described the multipolar resonances of dielectric particles, interest in this field has recently been renewed with the emergence of all-dielectric nanophotonics field, which allow increasing control over sensor design, fabrication and characterization techniques.13,14 Due to the ability of dielectric nanoparticles to support electric displacement currents inside them, they can exhibit magnetic dipole resonances as opposed to their metallic counterparts. This ability to manipulate both the electric and magnetic components of the electromagnetic waves incident upon them combined with their low-loss resonances paves the way to applications such as cloaking12, superlensing15, negative refraction16 and suggest them as an alternative to their metallic counterparts. The high quality factor resonance peaks of dielectric nanostructures, along with their easily engineered resonance modes are what has attracted the attention of the sensing community recently.17–20 The magnetic modes are characterized by the displacement current inside the particle inducing a magnetic dipole moment in the orthogonal direction. Focusing on Si nanocylinders (Si NCs), as they are used in this thesis for sensing applications, Figure 1-3 shows the magnetic and electric dipole (MD and ED) resonances of the Si NCs. As shown in Figure 1-3a, the circular electric displacement current requires sufficient electric field retardation at the bottom and top of the particle, therefore magnetic dipole modes can only be excited in the Si NCs that have the height large enough. This height corresponds to larger than 50-100 nm for silicon.21 The electric dipole resonance is due to the polarization of the charges inside the particle due to the electric field of the incoming light. The oscillation of the electric charges induces a magnetic current loop. This mode can be supported in shallow particles, since the magnetic permittivity of cylinder and the surrounding material is 1, and supports the magnetic current loop. CONCEPTS AND BACKGROUND 11 Figure 1-3 (a) magnetic dipole (MD) and (b) electric dipole (ED) modes shown for dielectric nanocylinders. The circulating displacement currents are indicated inside the particle, inducing magnetic and electric dipole moments. (c) The scattering cross sections for the 100nm tall Si NC with diameter of 150 nm, exhibiting MD and ED and magnetic quadruploe (MQ) modes. Reprinted from Polman et. al.22 The higher order terms in the Mie extinction cross section define the higher order multipole modes of the resonances. In high dielectric constant nanostructures, depending on their size and geometry, electric and magnetic dipoles, quadrupoles and higher order modes are present. A clever engineering of the structures geometry can make some of these modes to appear in the visible range, which is particularly interesting because it significantly simplifies the optics needed for detection. Figure 1-4 Resonances of Si NCs of different radii and heights with fixed interparticle separation 𝑠= 200 nm in the array. 500 600 700 800 900 1000 0.0 0.2 0.4 0.6 0.8 Extinction Wavelength (nm) r=200 r=180 r=160 r=140 r=120 r=100 r=80 r=70 500 600 700 800 900 1000 0.0 0.1 0.2 0.3 0.4 0.5 0.6 Extinction Wavelength (nm) 50nm 100nm CONCEPTS AND BACKGROUND 18 researchers in numerous creative ways and it continues to be an interesting field for engineers and biologists.44–49 Figure 1-8 Micro-mechanical valves in two-layer PDMS chips. (a) Push-down and (b) push-up valve architecture (reprinted from Melin et. al.44). (c) Picture of a valve in a closed state (reprinted from Stanford Microfluidics Foundry webpage50). In our sensing platforms we rely on the push-down design, in order to integrate the optical sensor arrays with the flow layer of the chip (see Chapter 2 for further details). We aim for simultaneous and parallel detection of different biomolecules or samples on a single microfluidic chip with multiple nanosensor arrays located in different channels of the chip. For target molecule detection, we utilize immunoassays (see next section). Therefore, it is important to design channel networks that can introduce common reagents into multiple experimental channels simultaneously, as well as enabling individual access to these channels for measuring different samples in parallel. This is satisfied with a special design of channels, inlets and outlets of these channels, and control valves that regulate the flow on those channels to perform the steps of immunoassays and sensor preparation. (a) (b) (c) CONCEPTS AND BACKGROUND 19 Figure 1-9 Microfluidic flow (blue) and control (red) channel architecture for (a) common treatment of all experimental channels and (b) individual treatment of each experiment channels. Figure 1-9a shows the architecture that allows common reagents to be introduced into all experimental channels simultaneously with the same flow rate. To allow this, the channels need to have the ‘’tree’’ structure, equally dividing the laminar flow of the reagent/sample from the common inlets into the experiment channels. By opening and closing the valves, which are the red areas intersecting the blue flow channels, the flow from different inlets can be directed to the channels. Similarly Figure 1-9b shows the architecture for individual channel treatment, where the inlets and outlets are separately connected to each channel. Those individual outlets are used when the channels are treated with common inlets in our architecture. The full layouts of the microfluidic channel networks used in experiments will be explained in Chapters 3 and 4, as for different experiments, slightly different designs are utilized. 1.4 Surface Chemistry Surface chemistry deals with the chemical changes at the interface of two materials/environments. We need to employ surface chemistry for multiple reasons in our platforms. One is to bind the PDMS chips onto the sensor substrates for performing sensing assays, and the other is to functionalize the sensor surface so that the target molecules can be attached to the surface of the sensors for detection. 1.4.1 Chip binding 8 experiment channels … Common inlets 8 experiment channels … Individual inlets & outlets (a) (b) CONCEPTS AND BACKGROUND 20 Binding the sensor substrates to the PDMS chips requires modifications of both material surfaces. Prior to this, the PDMS chip layers (control and flow) need to be bound covalently so that the chip can endure the pressures that will applied during the measurements. Two PDMS layers form a bond when the surface methyl groups are oxidized with oxygen or UV ozone plasma. When both oxidized layers are brought into contact, covalent bonding occurs in several minutes. Binding the glass substrates with PDMS chips relies on the same method. Glass and PDMS are oxidized by the oxygen plasma and brought into contact for covalent binding. Some applications that require forming organic self-assembled monolayers (SAM) on the sensor surface prior to chip assembly pose a crucial problem, as the plasma activation of the substrate would lead to destruction of the SAM. Therefore, in those cases, for the chip assembly, only the PDMS surface is oxidized and brought to pressure contact with the substrate at lower temperature. This results in slightly weaker but sufficiently strong binding for the sensing experiments. 1.4.2 Sensor functionalization To detect biomolecules, we run sandwich immunoassays on the optical nanoresonators. The key step is to immobilize antibodies on the sensors, without non-specific binding. In some cases, the passive adsorption of molecules onto the sensor surface is not strong enough or even possible, therefore, the functionalization of the sensor surface is necessary for running a robust immunoassay. The SAM of mercaptoundecanoic acid (MUA) is used on gold sensors to act as a link between sensors and the antibodies. The activation of carboxylic acid groups of MUA with EDC (carbodiimide) – NHS (N- hydrocarboxysuccinimide) reaction is used for immobilizing antibodies onto the MUA (Figure 1-10a). Once the antibodies are immobilized, before the target molecule is introduced into the environment, the active MUA sites that are uncovered by antibodies are blocked by ethanolamine (Figure 1-10b). This method is commonly used in plasmonic sensing applications, therefore the protocols are well established in the literature51–55, and only need to be adapted to our sensors. CONCEPTS AND BACKGROUND 21 Figure 1-10 Surface fucntionalization. (a) EDC-NHS chemistry on MUA for antibody immobilization on gold. (b) Ethanolamine blocking of the active MUA sites after antibody immobilization. (reprinted from Bhadra et. al.52) After antibody immobilization and ethanolamine blocking steps, the sandwich immunoassay can be performed on the sensors, described in detail in Chapter 2. 22 23 2 METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS The experimental methods used in the measurements consists of combination of multiple fields and technologies. The work flow of a regular on-chip sensing experiment consists of (i) fabricating the desired type and design of NRs on a glass substrate, (ii) fabricating the microfluidic chip with desired channel networks, (iii) preparing the NR sensor surface for sensing experiments, (iv) assembling the sensors with the microfluidic chip and (v) further chemical preparation of the sensors and (vi) finally the biomolecular sensing steps followed by (vii) data processing. Multiple manufacturing steps are usually prone to failing, due to incompatibilities between materials, fabrication steps and the complexity of the procedures. Many steps of fabrication and preparation procedures have to be followed and combined carefully in order to achieve a successful sensing measurement. The average time of fabrication of a biosensing platform we focused on in the scope of this thesis is between 2 days to 4 days. The most crucial aspect for developing a reliable point-of-care diagnostic device development is reproducibility. For a platform that consists of multiple fabrication steps and materials, it is very important to follow proper fabrication and preparation steps in order to perform successful experiments and obtain reliable results. In this chapter, the methods used throughout the thesis will be described in detail. The fabrication of both gold and silicon nanoresonators (NR), preparation of the NRs for biosensing experiments, fabrication of microfluidic chips and the integration of the chips with sensors will be explained. In the last section, the details of the sensing procedure on-chip will be given. 2.1 Fabrication of Nanoresonators for Biosensing METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 24 Nanoresonators (NRs) are the structures that have at least one dimension that ranges between 1-100 nm. To fabricate the structures in highly controlled and reproducible ways, highly controlled environments such as cleanrooms are needed, equipped with machines for fabrication and characterization of those structures. The nanostructures can be fabricated by top-down or bottom-up techniques, where the former constitutes of the fabrication of the nanostructures from a larger block of material and the latter is based on fabrication of nanostructures using atoms or molecules as the building blocks and bringing them together to form the desired structure.56 The top-down methods are used for the NRs employed in the projects followed in next chapters. Au and Si NR based sensing platforms are in the scope of this thesis. For the LSPR sensing platform that is developed and presented in chapter 3, gold nanorod arrays are employed, whereas for the Si NR based platforms that are presented in chapter 4, Si nanocylinder (Si-NC) arrays are used. Two types of Si-NC arrays are fabricated: periodic and semi-random. The different fabrication procedures are described in following sections. E-beam lithography is a highly controllable method, enabling the fabrication of highly reproducible samples with high resolution in feature size and high accuracy in positioning and aligning.57 It also provides the ability to design and fabricate different nanostructures in parallel with desired optical and mechanical properties. Therefore, we used e-beam lithography for fabricating sensors consisting of periodic and highly ordered structures. Despite its advantages, e-beam lithography is not a cost-effective method and it also suffers from high exposure times for patterning a small area of the sample. Colloidal lithography techniques are implemented to fabricate simpler arrays in a large-scale and cheap manner.58 When periodicity is not crucial colloidal lithography is a more suited fabrication method because it is significantly more efficient. Therefore, we used a colloidal lithography method customized in our lab, in order to fabricate randomly distributed Si-NC arrays for the part of the results presented in Chapter 5. 2.1.1 Fabrication of gold NRs The fabrication procedure of gold nanorod arrays is based on electron beam (ebeam) lithography with negative resist, followed by a reactive ion etching (RIE) METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 25 process. Figure 2-1 shows the fabrication procedure for gold nanorod arrays. The steps are as follows: Cleaning: The glass substrate (PGO MEMPAX) of 25x25 mm with thickness of 0.4 mm is cleaned with soap and water, followed by aceton and IPA rinsing. Metallization: The cleaned substrate is coated with 2 nm of Ti layer for adhesion, followed by 50 nm of Au layer in the Lesker Lab18 evaporator with the coating rate of 1 A/s. Resist coating: The metallized sample is then spin coated with negative e-beam resist ARN-7500-08 at 8000 rpm and baked at 85 C for 1 min. E-beam writing: In the negative resist e-beam lithography, the unexposed regions of the resist are removed by developer. The e-beam exposure of the desired pattern is realized by CRESTEC CABL writer. The pattern design is prepared in the custom software of the CRESTEC CABL. The optimization of parameters of exposure such as exposure time (dose), the current, etc. are crucial for obtaining high quality structures and it is important to test and fix those parameters by running test samples. The exposure time for one sample is around 4 hours regarding the number and the size of arrays needed to perform the experiments that will be described in Chapter 3. Developing: The exposed sample is developed in 1:4 AR-300-47 developer for 3 minutes and rinsed with water and dried with N2. The sample is baked at 85 C for 1 min to obtain resist contrast for RIE step. RIE: The gold layer is etched using directional argon plasma at the rate of 15 nm/sec using the ARN resist as etch mask. Cleaning: The mask is then cleaned by either O2 plasma or dipping the substrate into 1:3 piranha solution (Hydrogen peroxide: Sulfuric acid) for 10 seconds. The latter worked better for cleaning the sensor surface, eventhough requires careful handling of the acid and longer than 10 sec washing causes the Ti layer to be etched away, therefore the dissociation of the gold nanorodos from the substrate surface. METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 26 Figure 2-1 Fabrication procedure of gold nanostructures with negative e-beam resist. Following the cleaning of the substrate, the gold sensor surface is functionalized as will be described later in this chapter. 2.1.2 Fabrication of silicon sensors periodic Si nanodisk arrays Periodic Si nanodisk arrays were fabricated using standard e-beam lithography (EBL) followed by a reactive ion etching (RIE) step very similar to the gold nanorod fabrication described in previous section. We used the silicon-coated quartz substrates (Siegert Wafers) of 25x25 mm. The resist coating, e-beam writing, developing steps are kept the same as in Au NR fabrication. After that, the nanodisk patterns were transferred to the silicon layer by RIE using SF6 and C4F8 gases for etching. After the RIE, the substrate cleaning is done in the same way as described for Au NRs. semi-randomly distributed Si nanocylinder (Si-NC) arrays For the fabrication of the semi-randomly distributed Si-NCs with no long-range order but with a typical interparticle distance the fabrication procedure is as follows: glass Au layer (50 nm) Ti layer (2 nm) metallization e-beam resist resist coating e-beam exposure develop RIE piranha cleaning cleaning METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 27 Figure 2-2: Fabrication steps of randomly distributed silicon nanocylinder (Si-NC) arrays. (a) Fabrication steps of Si-NC arrays. (b) Dark field microscopy image of the sulfate latex beads dropcasted on the gold layer. (c) The SEM image of the substrate after the tape stripping step showing the stripped regions with gold nanodisks (above the dashed line) and unstripped regions with the beads on top of gold nanodisks (below the dashed line). (d) The SEM image of the substrate after etching the Silicon layer by RIE, using the gold disks as etching mask. The gold mask is seen on the Si- NCs. Cleaning: The Si-coated quartz substrates of 25x25 mm that are purchased from Siegert Wafer, GmbH were cleaned by aceton and IPA rinsing. Metallization: After cleaning the substrates, 2nm of Ti layer is coated as an adhesion layer on silicon and then 50 nm of Au layer is evaporated on the thin Ti layer. Bead dropcasting: Then the sample is treated with O2 plasma for 5 seconds at 100 Watts (200 ml/min flow). After the plasma, the sample is incubated for 1 minute in 0.2% Poly-diallyl dimethylammonium chloride (PDDA) solution for surface activation and the surface becomes positively charged, and then washed with water and dried with N2. After the surface activation step, the sample becomes ready for the dropcasting the Sulfate Latex beads (Thermofisher, S37491, 0.2 µm) 2 µm Si layer Quartz Au layer RIE RIE Piranha cleaning 4µm Tape strip (a) (b) (c) (d) Dropcast tape stripped 300 nm METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 34 Figure 2-6 The valve control setup and software. (a) electronic solenoid valves (Pneumadyne) are connected to the chip through tygon tubing filled with water. The pressure regulator is used to control the pressure inside the control channels. (b) The custom home-made software (Matlab) to control the electronic valves. The buttons on the chip map can be clicked to turn the valves on and off. There are also embedded script functions to run a part or the whole of the sandwich assay on the chip. The samples are connected to the flow layer with tygon tubings that are connected to a nitrogen flow at a constant pressure of 3 psi. This allows for the sample to flow into the channels when the control valves are opened. 2.5 Immunoassays for on-chip biosensing As described in the first chapter, the sensing applications in this thesis mainly use sandwich assay type of immunoassays. All the measurements, except for specifically stated ELISA measurements in Chapter 3, take place on chip. The microfluidic channels are used to control the flow of reagents on specific sensing sites. The details of the full microfluidic chip designs and experiments will be discussed in next chapters, while in this section the focus will be on the sandwich immunoassay specifics. In the sandwich assay, the capture antibody, immobilized on the sensor surface, specifically captures the target marker. The resonance response for the small target molecule is typically small and the resolution between different concentrations of the marker is not high enough to obtain a reliable sensing signal. Therefore, an amplification antibody that also recognizes and binds to the target biomolecule is used for signal amplification. In this section, the details of the immunochemistry used for the sensing with both LSPR and silicon nanoresonator sensors are described. (a) (b) METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 35 2.5.1 Cancer marker detection with Au NR sensors In order to immobilize the capture antibodies on the sensor surface coated with MUA, EDC/NHS chemistry is used to prepare amine-reactive esters of carboxylate groups. EDC/NHS reaction takes place in 45 minutes in MES buffer. Right after EDC/NHS reaction, the capture antibodies in 10 mM phosphate buffer (PB) are introduced into the channels for binding on the sensor surface. Once the sensor surface is coated with capture antibodies, the LSPR centroid shift signal saturates. Depending on the concentration of the antibody solution in PB, the saturation time changes slightly between 10 and 30 minutes. For most of the measurements presented in this thesis, the capture antibodies are flown in the channels for 60 minutes unless otherwise is specified. Following the immobilization of the capture antibodies on the specific detection sites on the sensors by the help of microfluidics, to prevent non-specific binding and for the blocking of unreacted NHS-ester groups on the sensors are blocked with ethanolamine solution in PB for 10 minutes. Next, the cocktail of markers prepared in PBS-BSA buffer or 100% unfiltered human serum (Sigma Aldrich) are introduced to the detection sites for an hour, each having different concentrations, in order to obtain calibration curves. In the case of multiplexed detection of multiple markers, each marker is captured by their specific antibody, giving LSPR shifts on the corresponding detection sites. Finally, the amplification antibodies in PBS with 1% BSA with PH of 7.4 to match the human serum PH are introduced through the channels, in order to amplify the LSPR shift. Figure 2-7 Sandwich assay protocol on gold sensor surface. 2.5.2 Cancer marker detection with Si NR sensors For the specific detection of the target protein (prostate specific antigen, PSA) we formed a layer of capture antibody on the sensor surface. The sensors were first flushed with 10 mM phosphate buffer (PB) in order to have a base for tracking the Au EDC/NHS chemistry MUA Activated MUA Capture Antibody Ethanolamine passivation Cancer markers Amplification antibody sandwich METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 36 shifts. The monoclonal antibody (mAb) (BiosPacific, A45160) for PSA was introduced to the channels in 10 mM PB through a common inlet so that the solution flowed through all the channels simultaneously and all the sensors were coated with the capture antibody. Once the capture antibody layer was formed, the resonance shift signal saturated and the channels were then washed with PB shortly to remove the excess unbound antibodies. The signal saturated after only 15 mins after 150 μg/ml mAb solution was introduced to the sensors. The binding kinetics and the saturation time depend on the mAb concentration used. For the proof of concept experiments we have measured the PSA calibration curve in phosphate buffer saline (PBS) with 1% bovine serum albumin (BSA). The BSA was added to the PBS as a blocking agent to prevent the unspecific binding and also to mimic the human serum proteins in this preliminary measurement. The 7 PSA calibration concentrations were prepared and introduced to different sensors through individual inlets of each channel. An extra channel was used as a control channel with no PSA added in the buffer. For the PSA calibration curve measurements in human serum, the calibration concentrations were similarly prepared in 50% diluted human serum instead of PBS with BSA. For the experiment presented in the Fig. 6, the capture antibody concentration used was 300 μg/ml. For the target proteins that are small in size compared to antibodies, the shift due to different concentrations of these proteins is not easy to detect directly. For the signal amplification, we used an amplification antibody that recognizes the target protein. For our measurements with PSA, polyclonal antibody (pAb) (BiosPacific, D63010) for PSA was used as a detection antibody. 100 μg/ml pAb in PBS with 1% BSA was introduced in the channels through common inlets after a short washing step with buffer. After the pAb binding signal was saturated, the channels were again washed with buffer to remove unbound antibodies and to eliminate the bulk refractive index effect due to free pAb solution in the channels. This way the absolute effect of pAb binding on the sensor area was measured. The centroid shifts due to the absolute pAb binding step were extracted to plot the calibration curves presented in this work. 2.6 ELISA protocol METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 37 Enzyme linked immunosorbent assay, ELISA, is a gold standard method for detection of markers in clinical diagnostics. We have performed a few ELISA measurements to compare it with our platform performance. We followed a standard ELISA protocol starting with a non-coated 96 well ELISA plate. To immobilize the capture antibodies, we prepared antibody dilutions in 50 mM carbonate buffer (CB) incubated them in the wells for 18 hours at 4⁰C. Then, the wells were washed 3 times with 10mM PBS-Tween and the markers are introduced into the wells, diluted in PBS-BSA (1%). After an hour of incubation at 37⁰C the wells are again washed 3 times with PBS-Tween and the amplification antibody used on-chip experiments is introduced to the wells to form the sandwich. Similarly to the previous steps, after 1 hour of incubation, the wells were washed and secondary antibody - horseradish peroxidase (HRP) conjugate is added to the wells. It is important to note that the Goat anti-Mouse (GAM) or Goat anti-Rabbit (GAR) antibodies are used depending on the source of the amplification antibody in the previous step. After the incubation of the conjugate, the wells are again washed and the colorimetric substrate, OPD (ophenylenediamine dihydrochloride) is added to the wells. Once the color of the solution in the wells turn yellow (takes around 30 mins) the stop solution (2.5 M H2SO4) is added on the OPD and absorption in the wells are read by the platereader. 2.7 Data processing and Analytics One of the key advantages of the integration of optical sensing with the microfluidic chip technology is to be able to get an instant readout of the sample concentration during the assay steps. In order to achieve that, our transmission microscopy set-up and a dedicated labview program collect the real-time transmission spectra of the sensing sites on the chips. The 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 signal from an area on the chip with no sensors for each sensing site and the 𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑 signal (dark signal) are recorded in the beginning of the measurement and the rest of the readout is ran automatically, scanning the sensing chip with a galvo mirror. The extinction is calculated by the following formula: 𝐸=log(1−𝑇)=log(𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒−𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑 𝑠𝑖𝑔𝑛𝑎𝑙−𝑏𝑎𝑐𝑘𝑔𝑟𝑜𝑢𝑛𝑑 ) (11) METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 38 where 𝑇 is transmission through the sample and 𝑠𝑖𝑔𝑛𝑎𝑙 is the experimental signal readout of the sensors. The extinction centroid is calculated as described in Chapter 1 in real-time and displayed during the measurement, enabling immediate access to the results and quick troubleshooting in case of any problem with the assay or the chip. Figure 2-8 shows the real-time centroid shifts for CA15-3 detection measurement on an LSPR chip. All the surface chemistry and assay steps can be tracked during the experiment, allowing for gathering the maximum information. Figure 2-8 The real-time LSPR centroid shifts due to 8 different CA 15-3 concentration solutions at the 8 separate detection sites on a chip. After the measurements, offline data processing is performed to get more information out of the collected data. The centroid shifts for detection antibody binding over the time is calculated for each sensing site and plotted against corresponding concentrations (See Figure 2-9). This plot is called concentrationresponse curve or calibration curve in analytical chemistry and has a trend of an S- curve. The data is fitted by four parameter logistic function: 𝑦=𝑚𝑎𝑥+𝑚𝑖𝑛−𝑚𝑎𝑥 1+(𝑥 𝐶)𝑠 (12) where; 𝑚𝑎𝑥 = the maximum value that can be obtained (at infinite concentration) 012345 0 2 4 6 361.2 36.2 3.62 1.81 0.36 0.18 0.03 0 EDC/NHS Capture Antibody Marker Amplification Antibody [CA15-3] (U/ml) LSPR centroid shift (nm) time (h) METHODS FOR ON-CHIP BIOSENSING WITH OPTICAL-NANORESONATORS 39 𝑚𝑖𝑛 = the minimum value that can be obtained (at zero concentration) 𝐶 = deflection point of the S-curve, EC50 value (the point halfway between 𝑚𝑎𝑥 and 𝑚𝑖𝑛, 50% Effect Concentration) 𝑠 = Hill’s slope of the curve (steepness of the curve at EC50 value) From the fitted data, the dynamic range and limit of detection (LOD) and limit of quantification (LOQ) of the calibration curve are calculated. Dynamic range is the interval between upper and lower limits of quantitation, which is the range that the curve is linear and therefore the measurement is expected to be precise and accurate, also called as the working range of the sensors. This range is taken to be the range between EC20 and EC80 values of the calibration curve. LOD of the sensor is the smallest analyte concentration that can be detected but not necessarily quantified exactly. Commonly it is defined as the 3 times the standard deviation on the blank measurement (measurement of the noise level in the assay when no analyte is present). For our measurements we took the EC10 value of the fitted curve as the LOD, as a convention. LOQ is the lowest concentration that can be quantified exactly. It is defined as 10xLOD. The sensitivity is defined as the assay response per unit analyte concentration. However this parameter is often defined differently by researchers and sometimes in conflicting senses. Some researchers define sensitivity as the LOD or LOQ or the resolution (smallest concentration difference that can be determined with confidence). Therefore, proper assessment of dynamic range and LOD is what we focus on in our measurements. For sensitivity we present the EC50 value of our sensors in the following chapters. 40 Figure 2-9 Obtaining calibration curves. (a) the LSPR shifts for the amplification antibody step from Figure 2-8. (b) the calibration curve obtained. The line is the 4-parameter logistic curve fit and the dashed region is the dynamic range of the sensor. LOD of 0.11 ng/ml is marked with a star. 3.5 4.0 4.5 5.5 6.0 6.5 7.0 Time (h) 7.0 6.5 6.0 5.5 3.5 4.0 4.5 LSPR centroid shift (nm) 0.01 0.1 1 10 100 1000 0.0 0.2 0.4 0.6 0.8 1.0 normalized LSPR shift [CA15-3] (U/ml) LOD=EC10 0.11 ng/ml (a) (b) 41 3 ON-CHIP MULTIPLEXED LSPR SENSING The results presented in this chapter as well as part of text and figures are published in ACS Sensors during the course of my PhD studies.59 Yavas, O.; Acimovic, S. S.; Garcia-Guirado, J.; Berthelot, J.; Dobosz, P.; Sanz, V.; Quidant, R. Self-Calibrating on-a-Chip LSPR Sensing for Quantitative and Multiplexed Detection of Cancer Markers in Human Serum. ACS Sensors 2018, 3 (7), 1376–1384. Abstract - The need for point-of-care (POC) devices capable of early detecting diseases and monitoring their status, out of a lab environment, has stimulated the development of compact biosensing configurations. While Localized Surface Plasmon Resonance (LSPR) sensing integrated into a state-of-the-art microfluidic chip stands as a promising approach to meet this demand, its implementation into an operating sensing platform capable to detect quantitatively a set of molecular biomarkers in an unknown biological sample is only at its infancy. In this chapter, we present an on-chip LSPR chip capable to perform automatic, quantitative and multiplexed, screening of biomarkers. We demonstrate its versatility by programming it to detect and quantify in human serum four relevant human serum protein markers associated with breast-cancer. Early diagnosis and accurate monitoring of disease progression is required to determine an optimum treatment and increase the recovery rate of patients. The complex nature of biological processes and pathways leads to significant changes in the levels of multiple molecular markers in patient’s blood over the course of a disease.60–62 Therefore, it is highly relevant to be able to track simultaneously the levels of a set of these markers, in order to spot the actual disease status. Enzymelinked immunosorbent assay, ELISA is one of the most commonly used detection technique in clinics and research laboratories.63 Its reliability, low limit of detection and commercial availability are the biggest advantages. However, the bulky wellplates and readers, long assay times and the large reagent volumes are not compatible with on-site, quick and multiplexed measurements. The ultimate goal ON-CHIP MULTIPLEXED LSPR SENSING 42 of developing point-of-care (POC) devices that allow for on-site, multiplexed measurement of analytes in a fast and cost-effective manner has motivated the development of a variety of biosensors based on different transduction mechanisms.64–67 Among them, optical sensors are particularly attractive owing to their fast response and compatibility with miniaturization and parallel sensing.34,68,69 Surface Plasmon Resonance (SPR) sensing is one of the welldeveloped and commercially available optical sensing schemes. A competitive approach to optical sensing relies on Localized Surface Plasmon Resonance (LSPR) supported by noble metal nanoparticles. Unlike SPR-sensing based on flat metal films, LSPR-sensing enables engineering the sensing volume down to the subwavelength (molecular) scale and benefits from direct coupling with propagating light. LSPR-based biosensing has shown to be a powerful approach to compact and simple platforms, especially attractive for POC applications.34,69–75 Yet, despite its great potential, its implementation into operating sensing devices capable to quantitatively assess the analyte concentration from unknown biological samples, remains little advanced.29,76 This is in part due to the complexity of implementing a working platform, where several disciplines like physics, surface chemistry, fluidics and electronics have to be optimally combined. Furthermore, while most efforts have so far focused on the detection of single analytes, reliable detection and monitoring of diseases generally requires multiplexed detection of several biomarkers. To this end, earlier works proposed employing colloidal metallic nanoparticles with different LSPR frequencies, each functionalized with a different receptor.77 While promising, this solution-based approach currently faces several drawbacks including aggregation of nanoparticles and optical signal fluctuations due to multiple washing steps. Another related configuration for multiplexing consists in immobilizing on a substrate antibody-functionalized gold nanorods of different aspect ratios78 that exhibit different optical properties allowing the selection of multiple working wavelengths. Chen et al. developed a multiplex serum cytokine immunoassay using nanoplasmonic biosensor microarrays. In this configuration, the solution based gold nanorods were immobilized on a substrate using a microfluidic channel network which was then removed and replaced by another microfluidic network used for the sensing assay.69 To date, none of the proposed LSPR-based schemes has enabled multiplexed and quantitative detection of several analytes in a biological sample. As a first step towards this goal, we recently presented a strategy34 that combines top-down engineered gold nanoparticle arrays with state-of-the-art microfluidics19–23 comprising micromechanical valves.46 This unique combination, which provides a controlled environment to the sensors, accurate delivery of ON-CHIP MULTIPLEXED LSPR SENSING 43 reagents and sample as well as automated assay operation, was successfully used to detect protein cancer markers in human serum with high sensitivity and specificity. As a proof of principle, detection of prostate specific antigen (PSA) and alpha-fetoprotein (AFP) in a clinically relevant level range were successively obtained.34 Here, by leveraging on the developed toolbox, we report on the first implementation of quantitative and multiplexed LSPR sensing on-a-chip. Our LSPR chip enables us to perform simultaneously self-calibrating, automated and multiplexed real-time detection of four breast cancer protein markers in human serum. These sensing performances combined with the long shelf-life of the chips, brings the LSPR based sensing one step closer to real-life operating POC devices. In the next sections, we discuss the clinical motivation, the platform design, and the results related to the sensing applications. 3.1 Clinical Motivation Clinical studies show that accurate detection of serum protein markers for breast cancer are pivotal to treatment monitoring towards a better prognosis.84–92 The most common markers are CA (cancer antigen) 15-3 and CEA (carcinoembryonic antigen).86,87,92 Also high serum ErbB2 (HER-2/Neu) concentrations are shown to be of use to monitor the response to specific treatment types.85,91 In addition to these antigens, the CA 125, which is a serum marker for ovarian cancer and some other diseases, is also shown to be a predictive marker of metastasis in breast cancer patients.84 More serum protein markers are relevant to breast cancer; extensive reviews can be found in the literature.88,92,93 Since tracking the level of a single marker alone may not be sufficiently conclusive in most cases, one needs to be able to monitor multiple markers in parallel. In our multiplexed measurements, we have focused on the four aforementioned molecules to demonstrate that our platform enables us to reliably quantify the concentrations of four relevant molecules with high specificity and reduced cross-reactivity. To assess our sensing performance, we also list the clinical cut-off concentrations of the selected markers with our results later in Table 2. The clinical cut-off concentration of a marker is the maximum concentration that a sample can contain to be considered healthy. ON-CHIP MULTIPLEXED LSPR SENSING 50 limit of detections of these individual marker measurements can be found in Table 2. For ErbB2 protein, the direct protein detection without amplification antibody signal was sufficient to obtain a sensitive enough calibration curve. This effect can be due to the possible higher affinity constant of the selected antibody for ErbB2 compared to the other proteins or effects related to the 3D structure of the protein or steric hindrances when the antigen is immobilized on the sensor surface. ON-CHIP MULTIPLEXED LSPR SENSING 51 Figure 3-5 Antibody concentration optimization experiments for (a,b) CA 15-3, (c,d) CA125, (e,f) CEA and (g,h) ErbB2 markers. 10 100 1000 10000 0.0 0.1 0.2 0.3 0.4 0.5 0.6 150 100 25 10 LSPR Shift (nm) [CA125] (U/ml) Capture antibody concentrations (g/ml) 110 100 1000 10000 0.0 0.1 0.2 0.3 0.4 0.5 0.6 150 100 50 25 LSPR Shift (nm) [CEA] (ng/ml) Capture antibody concentration (g/ml) 10 100 1000 10000 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 200 150 100 50 [CA 125] (U/ml) Amplification antibody concentrations (g/ml) 110 100 1000 10000 0.0 0.1 0.2 0.3 0.4 0.5 0.6 300 200 100 50 [CEA] (ng/ml) Amplification antibody concentration (g/ml) 0 1 10 100 1000 10000 1.5 2.0 2.5 3.0 3.5 4.0 150 100 [ErbB2] (ng/ml) Capture antibody concentration (g/ml) 10 100 1000 2.0 2.5 3.0 3.5 4.0 4.5 5.0 100 25 10 5 LSPR Shift (nm) [ErbB2] (ng/ml) Capture antibody concentration (g/ml) (a) (b) (c) (d) (e) (f) 0.01 0.1 1 10 100 1000 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 150 100 50 25 [CA15-3] (ng/ml) Amplification antibody concentration (g/ml) 0.01 0.1 110 100 1000 0.0 0.2 0.4 0.6 0.8 1.0 1.2 100 25 12.5 LSPR shift (nm) [CA 15-3] (U/ml) Capture antibody concentration (g/ml) (g) (h) ON-CHIP MULTIPLEXED LSPR SENSING 52 Table 1 Optimum antibody concentrations determined from the optimization experiments. Figure 3-6 Individual calibration curves for the four biomarkers obtained for optimized antibody concentrations. Error bars reflect the deviation between replicas on different chips for (c) and (e), and on the same chip for (d) and (f). In order to have a clear comparison between different measurements and different markers, the normalized calibration curve data is plotted in Figure 3-6. The error bars on the calibration curve data represent the variation between replicas from different chips for CA 15-3 and CA 125 and on the same chip for CEA and ErbB2. The relative standard deviation (standard deviation/mean signal) 0.01 0.1 110 100 1000 10000 0.0 0.2 0.4 0.6 0.8 1.0 normalized LSPR shift [CA15-3] (U/ml) 10 100 1000 10000 100000 [CA125] (U/ml) 0.1 110 100 1000 10000 [ErbB2] (ng/ml) 110 100 1000 10000 0.0 0.2 0.4 0.6 0.8 1.0 1.2 normalized LSPR shift [CEA] (ng/ml) (a) (c) (b) (d) ON-CHIP MULTIPLEXED LSPR SENSING 53 between replicas in the dynamic range is ~14% for interchip measurements and ~19% for intrachip measurements, suggesting that the reproducibility on different chips is as high as on the same chip. Owing to the self-calibration capacity of the platform, the sample quantification performance of the platform is independent of such small variations. The detection sensitivity extracted from the calibration curves for CA 15-3, Erb2 and CEA are compatible with clinical cut-off values84,85,87,92. For CA 125, the not so good limit of detection is attributed to the low quality of available antibodies. Shelf-life of the device It is important to know the shelf-life of the fabricated platform for a step towards POC device development. We have tested and found that once assembled, the chip can be stored at room temperature, in dark, up to 3 months without any significant alteration of its performance. Figure 3-7, shows the antibody binding performance of 1-week old and 3-months old chips, as well as associated calibration curves obtained by sandwich assay for CA15-3 protein. Figure 3-7 Shelf life of the fabricated chips. (a) the real time LSPR shift signals of the capture antibody binding on the sensors tested by 1 week old and 12 week old chips.(b) the CA15-3 calibration curves obtained on the respective chips in (a). The slope of the curves are shown on the graph. Data in (b) is normalized for fair comparison of shifts on different chips. Assay time The assay time for the on-chip immunoassays is another parameter to optimize, in order to achieve an efficient and fast detection scheme. In our experiments, we have used 1 hour of antibody and sample flow time, to keep consistency between previous measurements34 and our experiments and keeping this parameter constant, we optimized for antibody concentrations. However, the capture antibody immobilization time can be optimized to achieve faster sensor 40 60 80 100 120 140 160 0 2 4 6 8 LSPR shift (nm) time (min) 1 week old chip 12 weeks old chip 0.1 110 100 1000 0.0 0.2 0.4 0.6 0.8 1.0 1 week old chip 12 weeks old chip normalized LSPR shift [CA15-3] (U/ml) slope=0.70 slope=0.79 0 2 4 6 8 LSPR shift (nm) edc/nhs PB CA15-3 capture ab PB Ethanolamine blocking (a) (b) ON-CHIP MULTIPLEXED LSPR SENSING 54 preparation. As can be seen from Figure 3-7a, the LSPR shift saturates quickly after the capture antibody is introduced into the channels. 10-15 minutes of minimum antibody flow time is sufficient for coating the sensor surface with capture antibodies. Similarly, the real time LSPR sensing data (see Figure 2-8) suggests that the calibration curves obtained during the detection antibody flow, can be shortened down to 15 minutes, as the LSPR shift signal saturates after 15 minutes. A careful conduction of assay time optimization experiments can provide the minimum assay time with maximum sensitivity and dynamic range. 3.3.2 Cross reactivity Nonspecific antibody-protein cross-reaction results in inaccuracy and unreliable sensor response, interfering with the purpose of multiplexing95,96. In order to verify the specificity of the platform, we conducted a cross-reactivity control experiment (Figure 3-8a), where we immobilized the capture antibodies against CA 15-3, CA125, CEA and Erb2 on four sensor arrays (flow mode iii) and we flowed high concentrations of the four proteins individually (flow mode ii). Concentrations of proteins were selected to be the maximum ones of the dynamic ranges of their respective calibration curves. LSPR shifts obtained after sandwich formation for CA 15-3, CA125 and CEA channels and during the sample flow for ErbB2 channels are presented in Figure 3-8b. No cross-reactivity was observed between different species despite the high concentrations of proteins used. Conventionally, the cross-reactivity between two molecules are analysed by comparing the calibration curves obtained alone and in presence of the two molecules and calculated as shown in ref 95. To test our platform and antibodymarker pairs in a conventional way, in addition to the cross-reactivity experiment described above and in main text, we have selected the CA15-3 antibody and ErbB2 protein pair and performed a conventional cross-reactivity control experiment. Figure 3-9 show the calibration curve of CA 15-3, obtained alone and in presence of ErbB2 molecule. Black data points are corresponding to a measurement of CA15-3 and ErbB2 mixture at the linear range of CA15-3 calibration curve obtained alone (yellow data points). The raw LSPR shifts for both experiments are shown with no normalization. The agreement between two sets of data show that there is no effect of the presence of ErbB2 molecule on the CA15-3 calibration curve obtained by sandwich assay. ON-CHIP MULTIPLEXED LSPR SENSING 55 Similarly, the data on multiplexed detection in PBS buffer presented in the next section, compared with the individually obtained calibration, show that there is no cross-reactivity between any of the antigen-antibody pairs (Figure 3-10). Our data show that the platform ensures limited cross-contamination between the channels and the antibody selections prevent any significant cross-reactivity. Figure 3-8 Cross-reactivity experiments. (a) Sketch of the experimental steps (For sake of clarity, the control layer of the chip is not shown). Capture antibodies are immobilized and the proteins are flowed separately in different channels. Four replicas of the controls with no proteins is flowed also to check for channel-to-channel variation of the signals. Every intersection of the orthogonal flow channels is a sensing region. The sensing regions corresponding to each matching antibody-protein pair is marked with a star. (b) Corresponding LSPR shifts on the sensors. Shifts for the control channels are merged and the error bar is associated to the standard deviation of four replicas for the control measurement with each antibody pair. The protein concentrations are 18U/ml, 9000kU/ml, 3200ng/ml and 600ng/ml for CA 15-3, CA 125, CEA and ErbB2 respectively. CA15-3 CA125 CEA 0.0 0.2 0.4 0.6 0.8 1.0 LSPR shift (nm) Antibody pairs CA15-3 CA125 ErbB2 control CA15-3 protein CA125 protein CEA protein ErbB2 protein Sandwich assay Direct detection x3.5 * ** * control CA 15-3 control control control CA 125 CEA ErbB2 Step 1: Capture antibodies Step 2: Proteins Step 3: Detection antibodies * * * * (a) (b) ON-CHIP MULTIPLEXED LSPR SENSING 56 Figure 3-9 The cross-reactivity control experiment where the CA15-3 calibration points are obtained individually and in presence of ErbB2 marker. No cross-reactivity is observed. The calibration data obtained with solutions prepared in PBS with 1% BSA. 3.3.3 Multiplexed detection As a proof of principle experiment of multiplexed detection of the 4 biomarkers, we performed calibration curve measurements in 10mM PBS buffer with 1% BSA (bovine serum albumin) as a blocking agent. Each of the four capture antibodies was immobilized on a different sensor array (utilizing mode iii and step 1 described in Figure 3-4) and 8 cocktails of proteins with varying concentrations were flowed into the chip (utilizing mode ii and step 2 described in Figure 3-4) before introducing the amplification antibodies (Figure 3-4b, step3). Normalized calibration curves obtained for each protein marker are presented in Figure 3-10. Multiplexed measurements exhibit similar response compared to individually obtained calibration curves (dashed lines in Figure 3-10), showing no crossreactivity between the four proteins. A detailed comparison of the sensitivity (half maximal effective concentration, EC50), limit of detection (LOD at EC10), and the dynamic ranges of the curves (EC20-EC80) is presented in Table 1. 0.01 0.1 110 100 1000 0.0 0.2 0.4 0.6 0.8 1.0 1.2 in presence of Erb2 alone LSPR shift (nm) [CA15-3] (u/ml) ON-CHIP MULTIPLEXED LSPR SENSING 57 Figure 3-10 Multiplexed detection of four molecules in PBS. Solid lines are the four-parameter logistic equation fit to the data. The dashed lines are the fit to the individually obtained calibration curves. 3.3.4 Detection in human serum and sample quantification Following the multiplexed measurement in buffer, we demonstrate here multiplexed detection in 100% human serum. Figure 3-11a demonstrates the calibration curves obtained simultaneously from a single chip. Table 1 summarizes the LOD, sensitivity and dynamic range of the four markers measured in human serum. Due to the complex matrix, the curves are shifted towards higher concentrations compared to the measurements in PBS-BSA(1%) (dashed lines in Figure 3-11a) and therefore the LOD and EC50 values are slightly higher. Measurements were repeated 3 times to demonstrate their reproducibility. Data, presented in Figure 3-11b, exhibit very similar characteristics, suggesting high interchip reproducibility, especially in the sandwich assays for CA 15-3, CEA and CA125. The calibration curves for ErbB2 show larger dispersion, mostly because, for this molecule, no amplification antibody was used. Therefore for this application, the sandwich assay approach with detection antibodies provided more reproducible curves in complex media. 0.1 110 100 1000 10000 0.0 0.2 0.4 0.6 0.8 1.0 CA15-3 (U/ml) CA125 (U/ml) CEA (ng/ml) ErbB2 (ng/ml) normalized LSPR shift protein concentration ON-CHIP MULTIPLEXED LSPR SENSING 58 Figure 3-11 Multiplexed detection in human serum. (a) Calibration curves simultaneously obtained in 100% human serum (solid lines), and in PBS buffer from Figure 3-10 (dashed lines). (b) The calibration curves obtained on 3 different chips (solid, dashed and dotted lines). As a final step, towards real multiplexed sensing experiment on an unknown clinical sample, we demonstrate multiplexed sample recovery. To this end, we prepared six cocktails with varying biomarker concentrations for calibration curve. Then for recovering purpose, we spiked two samples with a mixture of targeted markers at different levels in human serum to mimic an unknown clinical sample. The spiked sample concentrations were interpolated from the calibration curves that were simultaneously acquired. Figure 6b shows an example recovery measurement in whole human serum for four molecules obtained for various replicas. The recovery rates (R) are listed in the table. Two of the sample concentrations, being CA15-3 and CA 125, were found to be significantly different but not incoherent (120% < R ≤ 130%). The recovery value for CEA was underestimated (R < 80%). ErbB2 concentration in the spiked sample is below the linear range of the calibration curve, in order to demonstrate the sensing performance below the LOD. The prepared and quantified ErbB2 concentrations were 15 ng/ml and 20±20.2 ng/ml respectively. The uncertainties on the recovery are mostly attributed to spiking errors during sample preparation. 0.01 0.1 1 10 100 1000 10000 0.0 0.2 0.4 0.6 0.8 1.0 1.2 CA 15-3 (U/ml) CEA (ng/ml) CA 125 (U/ml) ErbB2 (ng/ml) normalized LSPR shift protein concentration 0.1 1 10 100 1000 10000 0.0 0.2 0.4 0.6 0.8 1.0 1.2 CA 15-3 (U/ml) CEA (ng/ml) CA 125 (U/ml) ErbB2 (ng/ml) normalized LSPR shift protein concentration (a) (b) ON-CHIP MULTIPLEXED LSPR SENSING 59 Marker Cut-off concentratio n Measurement LOD Sensitivity Dynamic Range CA 15-3 25-40 U/ml 92 Individual 0.11 U/ml 2.0 U/ml 0.30-8.9 U/ml Multiplexed in PBS 0.21 U/ml 3.88 U/ml 0.6 – 19.2 U/ml Multiplexed in HS 1.85 U/ml 17.9 U/ml 4.2 – 75.4 U/ml CA 125 35 U/ml 84 Individual 0.138 kU/ml 4.5 kU/ml 0.524 – 100 kU/ml Multiplexed in PBS 0.139 kU/ml 3.9 kU/ml 0.580 – 30 kU/ml Multiplexed in HS 1.021 kU/ml 12.522 kU/ml 2.575 – 6.089 kU/ml CEA 2 ng/ml 87 5 ng/ml 84 Individual 14.7 ng/ml 192.0 ng/ml 35.2 – 645.7 ng/ml Multiplexed in PBS 16.3 ng/ml 220.5 ng/ml 44.5 – 1191.3 ng/ml Multiplexed in HS 76.19 ng/ml 635.4 ng/ml 166.7-2422.3 ng/ml ErbB2 15 ng/ml 85 Individual 3.5 ng/ml 99.8 ng/ml 12.07 – 824.5 ng/ml Multiplexed in PBS 3.9 ng/ml 145.0 ng/ml 11.75– 1742.6 ng/ml Multiplexed in HS 31.9 ng/ml 391.7 ng/ml 80.5 – 1904.8 ng/ml Table 2 The analytical parameters of the assays on chip summarized for individual marker detection (Figure 3-6), multiplexed detection in PBS buffer (Figure 3-10) and multiplexed detection in human serum (Figure 3-11a). PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 66 concentrations for cancer screening. We first study the optimal structural design of Si nanodisks for molecular sensing. Then, we demonstrate detection of PSA (prostate specific antigen) in buffer with a limit of detection (LOD) that is comparable to gold standard immunoassay techniques. Finally, to validate its operation in clinical conditions, the platform is tested in human serum. To the best of our knowledge, this is the first report combining silicon nanoresonators and microfluidics to perform clinically relevant immunoassays. 4.1 Detection chip Our platform consists of silicon nanodisk arrays on a quartz substrate integrated with a PDMS microfluidic chip including micromechanical valves (Figure 4-1). We fabricate the silicon nanodisk arrays using standard negative resist e-beam lithography followed by a reactive ion etching step on commercial amorphous silicon coated quartz samples (see Chapter 2). The nanodisk arrays have a fixed height h=50 nm. We choose to tune the disk radius r and inter-particle distance s to assess the optimum nanosensor parameters. The extinction spectra of the nanodisk arrays are measured using our homemade transmission microscopy setup34. Our optical detection enables us to interrogate up to 32 regions in parallel throughout the chip for real-time resonance tracking of different sensor arrays as described in previous chapters. Figure 4-1 The sensing chip with 32 arrays. (a) Picture of an assembled chip with 8 sensing channels. (b) Close-up picture of the 8 sensing channels showing silicon nanodisk arrays with different parameters. (c) SEM micrograph of a small portion of a silicon nanodisk array with h=50nm, r=140 nm and s=200 nm. 500 nm (a) (b) (c) Si quartz 500 μm PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 67 The PDMS chip is fabricated by multilayer soft lithography (see Chapter 2) leading to 8 sensing channels that are all individually and simultaneously addressable.34 The sample flow on the experiment channels is controlled by the micromechanical valves.46 The map of flow and control channel networks are shown in Figure 4-2 in blue and red, respectively. A common reagent can be flown in all sensing channels through a common inlet (labeled yellow), or individual samples can be flown through different sensing channels through the individual inlet (labeled green). This enables the sandwich assay formation steps for a full calibration curve with 8 different concentrations. Unlabeled blue channels are the outlets of the chip. Figure 4-2 The microfluidic flow (blue) and control (red) channel network with 8 sensing channels. The fluids can be flown into all the 8 channels through a common inlet (yellow), or individually through individual inlets (green). Unlabelled blue channels are the waste outlets of the chip. The detection of the biomolecules is based on a standard sandwich assay scheme. The capture antibody is immobilized on the silicon sensors by passive adsorption similarly to ELISA and other immunoassay techniques.124 The details of the sensing protocol will be explained in the following sections. Figure 4-3 schematically summarizes the sensing steps with the corresponding real-time resonance shifts. Control layer Flow layer Individual inlet Common inlet PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 68 Figure 4-3 Schematics of the sensing protocol (left) and the evolution of the nanodisk resonance during the different steps of the sandwich assay (right). The sample (100 ng/ml PSA) and control (no PSA) experiments are in grey and purple, respectively. 4.2 Optical characterization The extinction spectrum of individual h=50 nm Si nanodisks is mainly dominated by their electric dipole resonance (See Finite Element Method (FEM) simuations in Figure 4-4).21,125 Here, for sensing purposes, we aim at exploiting the strong collective resonance arising from coherent far field dipole coupling due to the periodicities of the arrays. These diffractive modes only arise when each individual particle support an electric dipole resonance.121 The array resonance is optimized by changing the disk radius (r) and separation (s). Figure 4-5 shows the measured extinction spectra of the different arrays along with the corresponding Finite Element Method (FEM) simulations. The resonances are tuned by changing the periodicity at fixed radius (Figure 4-5a), or conversely, changing the nanodisk radius at constant disk separation (Figure 4-5c). The corresponding FEM simulations on infinitely large arrays are in good agreement with the measured data (Figure 4-5 b and d). The amplitude and width of each of these resonances vary for different nanodisk arrays. While these properties are important for the detectability of spectral shifts, also the refractive index sensitivity is expected to vary with the array parameters. : capture antibody : target protein : detection antibody i ii iii i ii iii [PSA] controlsample 0 1 2 3 0 1 2 3 100 ng/ml 0 ng/ml Resonance Shift (nm) time(h) PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 69 Figure 4-4 FEM Simulations of infinite Si nanodisk arrays. Electric field enhancement (a) on the nanodisk and glass surface, plotted in a periodic manner to visualize the interparticle couplings, and (b) in a cross section of one unit cell, 25 nm from the glass surface. (c) The near field exponential decay by the distance from the nanodisk in the x-direction. The nanodisk array was placed on a glass/water interface with parameters r=140 nm and s=300 nm. (d) The extinction spectra of the nanodisk arrays with varying periodicity in x-direction with fixed nanodisk separation in y-direction (Sy=300 nm). (e) The extinction spectra of the nanodisk arrays with varying periodicity in y-direction with fixed nanodisk separation in x-direction (Sx=300 nm). (f) Single particle extinction, scattering and absorption spectra for r=140 nm and h=50 nm disk. PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 70 Figure 4-5 Resonance tuning of Si-nanodisk arrays. Experimental extinction spectra of silicon nanodisk arrays in air: (a) Influence of the disk separations at fixed radius r=140 nm and (b) Influence of the disk radius at fixed disk separation of s=200 nm. (c, d) Corresponding FEM simulations. Inset in (a) shows the geometry of the disks and the incident light polarization. 4.3 Bulk refractive index sensitivity characterization In order to identify the structural parameters (r and s) that provide the highest sensitivity to the surrounding media we performed bulk refractive index sensitivity (BRIS) experiments. In these experiments, we fabricated sensor arrays with 4 different radii (r=120, 140, 160 and 180 nm) and with disk separations varying from 100 nm to 450 nm with 50 nm increments. Once integrated to the microfluidics, the fabricated sensor arrays are exposed to increasing concentrations of glucose solutions in ultra-pure water (Figure 4-6a). 500 600 700 800 900 0.0 0.5 1.0 100 150 200 250 300 500 600 700 800 900 0.0 0.2 0.4 0.6 500 600 700 800 900 0.0 0.2 0.4 0.6 (b) (a) 500 600 700 800 900 0.0 0.5 1.0 200 180 160 140 120 100 80 r = 140nm (c) (d) E k Extinction Wavelength (nm) r (nm) s (nm) Measurements Simulations s s s = 200 nm r PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 71 Figure 4-6 Bulk refractive index sensitivity (BRIS). (a) Schematics of the BRIS experiment in which the 8 microfluidic channels are used to flow different water/glucose mixtures on Si nanodisk arrays with different s and r. (b) Evolution of the extinction spectra of a nanodisk array (r=140 nm, s=300 nm) exposed to 6 different glucose-water mixtures. For illustration, Figure 4-6b shows the resulting redshift in the extinction spectra for r=140 nm and s=300 nm. Our automated parallel acquisition enables us to track in real time the spectrum of each of the different arrays on the chip and extract the corresponding shifts in the main peak centroid. Figure 4-7 shows the peak and centroid shifts of two sets of sensors where different glucose concentrations are flowed in the channels sequentially with a step of washing with water in between. We observe instantaneous shifts as the refractive index of the surrounding media changes. The redshifted signal returns back to the baseline value for the washing steps with water ensuring that there is no irreversible modification of the sensors and the shifts are indeed due to bulk refractive index changes. For asymmetric extinction peaks like the one considered here, peak centroid tracking was found to be more sensitive than standard peak tracking (Figure 4-7).26 % glucose s 650 700 750 800 850 900 950 0.0 0.1 0.2 0.3 Extinction Wavelength (nm) 0 0.75 1.5 3 6 12 830 840 850 860 0.20 0.25 % glucose (a) r (nm) (b) r = 140 nm s = 300 nm PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 72 Figure 4-7 Comparison of the shifts of the centroid and peak positions of the resonances. Real-time centroid and peak position shifts of the resonances due to changing bulk refractive index of the surrounding media (glucose percentage concentrations) for (a) r=140 nm and s=250 nm and (b) r=160 nm and s=250 nm nanodisk arrays. The shifts of the peak positions (grey lines in both plots), are slightly lower than the shifts of the centroid positions. Figure 4-8 shows the evolution of the solution of the peak centroid with the refractive index for r=140 nm and s=300 nm nanodisk array. From the slope of the linear fit we extract the bulk refractive index sensitivity (BRIS) of the sensor. Figure 4-9 shows the centroid shifts of all the different arrays fabricated on the same chip. The BRIS values for sensors with different disk separation and radius are gathered in Figure 4-10. The arrays with separations larger than 300 nm were not considered as they either exhibit very low extinction, due to a low particle density, or resonances that were out of the spectral range of our set-up. Within the considered parameter range, we found that the BRIS values increased with increasing nanodisk separations. The highest BRIS value, 227 nm/RIU, was measured for the array with r=140 nm and s=300 nm featuring a resonance centered at 844nm with a quality factor of 20. Despite the simplicity of our structure, this BRIS value is only slightly lower than the previously reported BRIS values of more complex silicon nanostructure arrangements.40,119 (b)(a) 0.0 1.5 3.0 4.5 6.0 7.5 9.0 0 1 2 3 shift (nm) time (min) centroid position peak position r=140 nm s=250 nm 0.75 % 1.5 % 3 % 6 % 12 % 0 % 0.0 1.5 3.0 4.5 6.0 7.5 9.0 0 1 2 3 4 centroid position peak position shift (nm) time (min) r=160 nm s=250 nm 0.75 % 1.5 % 3 % 6 % 12 % 0 % PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 73 Figure 4-8 BRIS of the Si nanodisk array with r=140, s=300 nm. The centroid positions extracted from the extinction spectra in Figure 4-6b as a function of the refractive index of the glucose solutions. The inset is the real time centroid shifts during the sequential flow of varying glucose concentrations separated by rinsing step. Figure 4-9 The BRIS measurements for all the presented arrays with varying radius r, and nanodisk separation s. 1.330 1.335 1.340 1.345 1.350 844 845 846 847 848 Centroid wavelength (nm) Refractive index 02468 0 2 4 Centroid shift (nm) Time (min) r = 140 nm s = 300 nm 1.332 1.336 1.340 1.344 1.348 1.352 0 1 2 3 4 222.04 3.66 nm/RIU 188.09 3.31 nm/RIU 125.15 2.29 nm/RIU s = 250 nm s = 200 nm s = 150 nm s = 100 nm 65.32 1.15 nm/RIU r = 160 nm 1.332 1.336 1.340 1.344 1.348 1.352 0 1 2 137.50 +- 2.02 nm/RIU s = 150 nm s = 100 nm 83.61 +- 1.47 nm/RIU r = 180 nm 1.332 1.336 1.340 1.344 1.348 1.352 0 1 2 3 4 50.75 0.67 nm/RIU 100.56 1.50 nm/RIU 146.63 2.37 nm/RIU 192.42 3.39 nm/RIU s = 300 nm s = 250 nm s = 200 nm s = 150 nm s = 100 nm 229.55 2.65 nm/RIU r = 140 nm 1.332 1.336 1.340 1.344 1.348 1.352 0 1 2 3 4 31.85 + 1.70 nm/RIU 154.43 + 2.12 nm/RIU 101.86 + 1.79nm/RIU s = 250 nm s = 200 nm s = 150 nm s = 100 nm 210.09 + 3.01 nm/RIU r = 120 nm Resonance centroid shift (nm) Refractive index PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 74 Figure 4-10 Summary of the BRIS values obtained for different arrays with different radii and disk separations from Figure 4-9. The error bars are smaller than the data points in the plot. 4.4 Cancer marker detection 4.4.1 Detection in PBS-BSA buffer For the target analyte sensing proof of concept experiments, we selected the two sensors exhibiting the highest BRIS values (with radii of 140 and 160 nm and disk separations of 300 and 250 nm, respectively). The experiment consists of flowing different concentrations of the target molecule in each of the individual channels in order to obtain an 8-point calibration curve. We here focused on the detection of Prostate Specific Antigen (PSA). PSA is a 34 kDa protein whose high concentration in blood (greater than 4-10 ng/ml) can be associated to prostate cancer or other prostate disorders.126,127 For the detection of PSA, the capture antibody is first immobilized on the sensor surface by passive adsorption by flowing the antibody solution in phosphate buffer through all 8 channels. The sensors were first flushed with 10 mM phosphate buffer (PB) in order to have a base for tracking the shifts of the monoclonal antibody (mAb) (BiosPacific, A45160) for PSA which was introduced to the channels in 10 mM PB through a common inlet so that the solution flowed through all the channels simultaneously and all the sensors were coated with the capture 100 150 200 250 300 0 50 100 150 200 250 r = 120 nm r = 140 nm r = 160 nm r = 180 nm BRIS (nm/RIU) Disk separation (nm) PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 75 antibody. Once the capture antibody layer was formed, the resonance shift signal saturated and the channels were then washed with PB shortly to remove the excess unbound antibodies. The signal saturated after only 15 mins after 150 μg/ml mAb solution was introduced to the sensors. The binding kinetics and the saturation time depend on the mAb concentration used. For the proof of concept experiments we have measured the PSA calibration curve in phosphate buffer saline (PBS) with 1% bovine serum albumin (BSA). The BSA was added to the PBS as a blocking agent to prevent the unspecific binding and also to mimic the human serum proteins in this preliminary measurement. The 7 PSA calibration concentrations were prepared and introduced to different sensors through individual inlets of each channel. The 8th channel was used as a control channel with no PSA added in the buffer. For the target proteins that are small in size compared to antibodies, the shift due to different concentrations of these proteins is not easy to detect directly. For the signal amplification, we used an amplification antibody that recognizes the target protein. For our measurements with PSA, polyclonal antibody (pAb) (BiosPacific, D63010) for PSA was used for this purpose. 100 μg/ml pAb in PBS with 1% BSA was introduced in the channels through common inlets after a short washing step with buffer. After the pAb binding signal was saturated, the channels were again washed with buffer to remove unbound antibodies and to eliminate the bulk refractive index effect due to free pAb solution in the channels. This way the absolute effect of pAb binding on the sensor area was measured. The centroid shifts due to the absolute pAb binding step were extracted to plot the calibration curves presented in this work. Figure 4-11 Cancer marker sensing in PBS. (a) Real time resonance shifts of the silicon NC arrays (r=140nm, s=300nm) due to the detection of different concentration of PSA. The inset shows the calculated near field distribution for one nanodisk from an infinite array. The white dashed line 100 25 10 5 2.5 1 0.5 0 [PSA] (ng/ml) 015 30 45 60 75 0.0 0.5 1.0 1.5 Resonance Shift (nm) time (min) (b)(a) 100 25 10 5 2.5 1 0.5 0 [PSA] (ng/ml) E 0.1 1 10 100 0.0 0.2 0.4 0.6 0.8 1.0 r=140 nm s=300 nm r=160 nm s=250 nm Resonance Shift (nm) [PSA] (ng/ml) PERIODIC SILICON NANORESONATOR ARRAYS FOR ON-CHIP BIOSENSING 82 platform is compatible with detection of small biomolecules in complex matrices for clinical applications. The reported sensing performance enables us to detect clinically relevant concentrations of PSA. We have also compared our platform with a well-developed LSPR-based sensing protocol and shown that the sensitivity, LOD and the dynamic ranges are comparable. (i) Besides the similar sensing performance, one of the advantages of the silicon based sensors is the significantly longer decay length of the surface field over LSPR modes in metal nanoantennas (Figure S1), which can be beneficial for multilayer assays involving detection of molecules relatively far from the surface. This suggests that in comparison with LSPR sensors, different assay types with multiple layers of antibodies can be efficiently monitored using the silicon nanoresonators which in practice enables more practical and faster detection of the target analyte. (ii) While the extinction resonance of the Si nanoresonators considered here is much weaker compared to gold nanoantennas (inset Figure 4-15), its quality factor is substantially higher (20 versus 10). We foresee further engineering of the Si nanoresonators including arrangement in dimers and oligomers could improve the sensing performance. (iii) While the location of the adsorbed antibodies is not controlled in the present study, a Si-selective surface chemistry is foreseen to compare the sensor performance. (iv) It is also noteworthy mentioning that Si nanoresonators feature higher stability in solution compared to their gold counterparts. While gold requires stabilization with a passivation layer (e. g. using mercaptoundecanoic acid), we did not observe any significant degradation of bare Si cylinders, even after long time exposure to both water and buffer. (v) Finally, the quality of the sandwich assay, hence the LOD and sensitivity, can further be improved by additional optimizations of the capture and detection antibody concentrations. 83 84 5 SEMI-RANDOM SILICON NANORESONATOR ARRAYS FOR BIOSENSING The results presented in this chapter as well as part of the text and figures are published in ACS Nano during the writing of the PhD thesis.129 Yavas, O.; Svedendahl, M.; Quidant, R. Unravelling the role of electric and magnetic dipoles in biosensing with Si nanoresonators. ACS Nano 2019, 13 (4), pp 4582– 4588. Abstract: High refractive index dielectric nanoresonators are attracting much attention due to their ability to control both electric and magnetic components of light. Combining confined modes with reduced absorption losses, they have recently been proposed as an alternative to nanoplasmonic biosensors. In this context, we study the use of semi-random silicon nanocylinder arrays, fabricated with simple and scalable colloidal lithography for the efficient and reliable detection of biomolecules in biological samples. Remarkably, electric and magnetic dipole resonances are associated to two different transduction mechanisms: resonance redshift and extinction decrease. By contrasting both observables, we identify clear advantages in tracking changes in the extinction magnitude. Our data suggest that, despite its simplicity, the proposed platform is able to detect prostate specific antigen (PSA) in human serum with limits of detection meeting clinical needs. In the previous chapter, for an on-chip biosensing platform, we have focused on periodic Si nanodisk arrays of 50 nm height, where the main resonance modes were the bragg diffraction modes and electric dipole excitations of the silicon nanodisks and performed real-time detection of cancer biomarkers in human serum.103 While at that stage, Si-based nano-optical sensors have already reached comparable performances to LSPR counterparts, further developments are required to fully exploit their potential. In particular, there is a need to further understand how the control over both electric and magnetic dipoles could benefit SEMI-RANDOM SILICON NANORESONATOR ARRAYS FOR BIOSENSING 85 the detection sensitivity. Furthermore, especially within the context of point-of- care applications, one needs to identify cost effective strategies to fabricate high refractive index dielectric nanoresonators (HRDN) sensors over large areas. In this chapter, we present a novel platform which contributes to both objectives. A semi-random array of silicon nanocylinders (Si-NCs), fabricated by low-cost and scalable colloidal lithography, is integrated into a microfluidic environment to perform prostate specific antigen (PSA) detection through two different transduction mechanisms: resonance redshift and extinction reduction. Remarkably, we find these observables are associated to electric and magnetic dipole resonances, respectively. Through real-time tracking of both signals, we demonstrate that extinction reduction leads to better sensing performances. 5.1 Semi-random Si-NC arrays Our sensing chip consists of a semi-random array of Si-NCs integrated into a microfluidic environment. Similar arrays have previously been used in various LSPR biosensing schemes.130–134 In order to achieve fast, cheap and large-scale fabrication, we used colloidal lithography and fabricated semi-randomly distributed Si-NC of height of 130 nm and radius of 140 nm with electric and magnetic dipole resonance at 900 nm in aqueous environment (Figure 5-1). We first coated the Si-on-quartz substrates with a 50 nm gold layer. Then, we dropcasted the sulfate latex beads which are charged and therefore repel each other and form a semi-random array, that is, without any long-range order but with a typical nearest neighbour distance.135 We used them as a reactive ion etching mask for etching the gold layer in RIE chamber. Next, we removed them by an adhesive tape and used the patterned gold mask for etching the silicon layer with RIE. Finally, we cleaned the substrate by piranha solution, which removed the gold mask layer by etching away the Ti layer below (see Chapter 2 for the detailed fabrication protocol). SEMI-RANDOM SILICON NANORESONATOR ARRAYS FOR BIOSENSING 86 Figure 5-1 Semi-random Si NC arrays. (a) SEM image of the Si-NCs on quartz. (e) The measured extinction spectra of the silicon NCs in air (black) and in water (grey). The SEM image of the semi-randomly distributed Si-NCs are shown in Figure 5-1a. With this EBL-free method, the whole sample area can be patterned with nanostructures simultaneously without altering the fabrication time or cost. The areas on the substrate to be patterned can be selected by tape stripping the beads away before using them as an etch mask. More precise bead stripping method is described by Acimovic et. al for patterning the sample surface with precision of few micrometers by using a homemade PDMS stripping tape.33 For our sensing device, we tape stripped the edges of the sample, leaving the NCs only at the central region of 0.5 cm2 on the chip. The extinction spectra of the semi-random Si-NC array measured in air and in water are shown in Figure 5-1b. The resonance position in air and in water was 870 and 900 nm, respectively, showing a clear redshift due to the large refractive index change of the local environment. Furthermore, the extinction amplitude is reduced in water compared to the spectrum in air, which is in-line with previous reports.17,20,120 To test the sensing performance of the fabricated Si-NC arrays, we integrated it with a multilayer microfluidic network of PDMS,34 which enables the control of the sample flow on the sensing regions that are separated by microfluidic channels (Figure 5-2). The microfluidic chip design and the operation principles, as well as the fabrication procedure are described in the previous chapters. This configuration is crucial for rapid and practical execution of complex immunoassay steps in a highly controlled environment. 800 850 900 950 1000 0.2 0.4 0.6 0.8 1.0 in air in water Extinction Wavelength,  (nm) (a) (b) 300 nm SEMI-RANDOM SILICON NANORESONATOR ARRAYS FOR BIOSENSING 87 Figure 5-2 The integrated chip. The sensor area is visible as a darker area on the substrate under the PDMS microfluidic channels. 5.2 Bulk refractive index sensitivity and transduction mechanisms Initially, we evaluated the bulk refractive index sensitivity (BRIS) of the Si-NCs by sequentially flowing different percentage glucose solutions through the channels and tracking the centroid position and the extinction amplitude. The centroid shift of the extinction peak with respect to the refractive index of the glucose solution is shown in Figure 5-3a-b. The semi-random Si-NC arrays exhibited a BRIS of 86 nm/RIU by the conventional centroid tracking method. Previously, the BRIS of periodic silicon nanodisk arrays of 50 nm height was reported to be 227 nm/RIU, exhibiting much higher BRIS value, due to enhancement of sensitivity by the diffractive modes induced by the periodicity.103 SEMI-RANDOM SILICON NANORESONATOR ARRAYS FOR BIOSENSING 88 Figure 5-3 Bulk refractive index sensing experiments with semi-random Si-NC arrays. The two different transduction mechanisms analysed. (a) The extinction spectra of the sensors in glucose solutions of varying concentrations. (b) The respective centroid shifts from (a). The slope of the linear response of centroid position to changing refractive index is the BRIS. (c) The real time traces of the extinction amplitude response to changing refractive index by sequential flow of different percentage glucose concentrations. The inset shows the wavelength positions the three measurements: the peak wavelength (black line), at 900 nm (red line) and at 890 nm (blue line). (d) The extinction reduction sensitivity as a function of wavelength. The extinction reduction (ER) is the negative change of extinction signal with respect to the refractive index. The highest sensitivity was reached at 890 nm (blue). The change in extinction with respect to the refractive index, n, of the surrounding medium is shown in the inset for 890 nm, 900 nm and at the peak position of the extinction. In addition to the centroid shift, we notice that the extinction is reduced while increasing the surrounding refractive index, as seen in Figure 5-3a. Based on this observation, we evaluate the sensing performance of our sensors by tracking the extinction reduction in Figure 5-3c-d. We define extinction reduction as the negative change in the extinction signal. In Figure 5-3c, the real-time response of the extinction amplitude to the sequential flow of the distinct glucose concentrations, with washing steps in between, is presented for three different cases. We have tracked the extinction amplitude change at the peak maxima, at 900 nm and at 890 nm, which was found to exhibit the highest extinction reduction sensitivity. The whole wavelength range scanned for the maximum 1.33 1.34 1.35 0.0 0.5 1.0 1.5 Centroid shift (nm) Refractive index, n BRIS= 86 nm/RIU 800 850 900 950 1000 0.2 0.4 0.6 0.8 Extinction  (nm) 1.333 1.335 1.337 1.341 1.350 Refractive index 890 900 910 920 0.62 0.66 0.70 (a) 0 2 4 6 8 0.00 0.01 0.02 0.03 ER time (min) (c) 0% 1.5% 3% 6% 12% 880 900 920 0.6 0.7  (nm) Ext. 890 nm 900 nm peak 850 900 950 1000 -0.5 0.0 0.5 1.0 1.5 ER/n (a.u./RIU)  (nm) 1.335 1.345 0.00 0.01 0.02 0.03 890nm 900nm peak Ext. RI, n (b) (d) SEMI-RANDOM SILICON NANORESONATOR ARRAYS FOR BIOSENSING 89 extinction reduction sensitivity is shown in Figure 5-3d. These results suggest that the optical platform could be rearranged to detect only the transmission amplitude instead of the full spectra, in order to perform biosensing in a cheaper and simpler way. To back up our experimental observations and to study the origin behind the two different transduction mechanisms, we performed extensive numerical simulations using COMSOL FEM solver (See Appendix A for the semianalytical calculations by island film theory136,137). First, we simulated and compared the single and ensemble of Si-NCs. Figure 5-4a and c show the resonances of an isolated Si-NC in aqueous solution while Figure 5-4b and d demonstrate a small part of a semirandom array of Si-NCs under identical conditions. As seen in Figure 5-4a, the resonance position of an isolated Si-NC redshifts and the extinction amplitude decreases as the surrounding refractive index increases. However, this effect is enhanced as more Si-NCs are assembled in a semi-random array. As can be seen in Figure 5-4f, the resonance shift for the array is about twice larger than for an isolated Si-NC. Also, the extinction reduction is increased by a similar amount (Figure 5-4g). These results are likely due to increased shielding effects on the interparticle electromagnetic coupling induced by the increased refractive index of the surrounding medium. While plasmonic metal nanostructures in similar arrangements have shown negligible coupling,136 the mode extension for Si-NCs is significantly greater.103 Furthermore, due to limited computation power, we only modelled 10 nanostructures in the semi-random array assembly. We foresee the observed effects to increase further for a larger array and to reduce the mismatch with experimental observations. For the parameters of the fabricated Si-NCs, both magnetic and electric fields are enhanced at resonance. Interestingly, by separating the extinction cross section into electric and magnetic dipolar components (Figure 5-4e), the underlying mechanisms of the resonance shift and the extinction reduction can be unveiled. From Figure 5-4f and g, it appears that the magnetic resonance is responsible for the resonance redshift, while the extinction reduction is related to the electrical dipole. This observation explains the sources of the experimentally measured BRIS results. We can therefore associate the measured resonance centroid redshift and the extinction reduction with the respective Mie resonance modes. SEMI-RANDOM SILICON NANORESONATOR ARRAYS FOR BIOSENSING 90 Figure 5-4 FEM simulations of Si-NCs. The extinction cross section of (a) a single Si-NC and (b) the average response from a semi-random Si-NC array of 10 particles, surrounded by media with different refractive indices. The E-field and H-field enhancements of (c) the single Si-NC and (d) the semirandom Si-NC array, around the resonance wavelength (940nm). (e) The electric and magnetic dipolar (e.d. and m.d.) components of the extinction cross section of the Si-NC array. (f) The extinction peak shifts due to changing refractive index of surrounding medium. The m.d. resonance of the Si-NC array (red circles) and the total extinction of the Si-NC array (blue squares) and total extinction of a single Si-NC (black triangles) are analysed and compared separately. (g) The wavelength dependence of the extinction cross section reduction (ERσ) due to changing refractive index. 5.3 Cancer marker detection In order to demonstrate the biomolecule detection capability of our platform and compare both aforementioned transduction mechanisms, we focused on the SEMI-RANDOM SILICON NANORESONATOR ARRAYS FOR BIOSENSING 91 detection of prostate specific antigen, PSA. PSA is a protein cancer marker whose concentration in serum tends to overpass its normal level (4-10 ng/ml) for patients affected by prostate cancer.126,127 We used a sandwich immunoassay scheme (Figure 5-5a) for detecting PSA on the chip. The binding events are observed in real-time as a redshift of the resonance centroid and a reduction of the extinction. First, a selective monoclonal capture antibody for PSA is immobilized on the sensor surface in all the eight channels by a common inlet, through passive adsorption, similarly to clinically used enzyme linked immunosorbent assay (ELISA). Then, through individual inlets, eight calibration samples in PBS-BSA (Phosphate Buffer Saline-Bovine Serum Albumin, 1%) buffer with different PSA concentrations is flowed into the distinct channels and the PSA is captured by the antibody on the sensor surface, leading to additional adsorption signals. One of the eight channels was used as a control channel, with only PBS-BSA(1%) buffer flowing and no PSA. Following the PSA capture step, a polyclonal antibody is then introduced in all the channels as an amplification antibody, binding to PSA, resulting in larger and more detectable signals as well as a higher selectivity of the assay. Each step of the sandwich assay is adjusted to be 1 hour to have saturated signal shifts for each channel. Figure 5-5 PSA detection results with semi-random Si-NC arrays.. (a) The sketch of the sandwich assay steps for antigen detection. First the capture antibody is immobilized on the sensor surface by passive adsorption (i), then the antigen is captured by the capture antibody (ii) and finally the signal is amplified by an amplification antibody (iii). (b-c) The calibration curve by the (b) centroid shifts and (c) extinction reduction due to the amplification antibody step obtained from the eight channels of the chip. Error bars represents the replicas of the measurement on the same chip. The PSA calibration curves obtained by tracking the resonance centroid redshift and by tracking the extinction reduction at 890 nm are displayed in Figure 5-5b and c, respectively. The control channel shows no binding signal, suggesting a high specificity. The limit of detection (LOD), calculated conventionally as the EC10 value of the four-parameter logistic curve fit, reached by centroid shift tracking was 1.55 ng/ml, which is below the clinical cut-off concentration of PSA for 110 100 1000 0.2 0.4 0.6 Centroid shift (nm) [PSA] (ng/ml) (a) (b) (c) (i) (ii) (iii) LOD: 1.55 ng/ml LOD: 0.83 ng/ml 110 100 1000 0.000 0.002 0.004 0.006 xtinction@890 nm [PSA] (ng/ml) BIBLIOGRAPHY 98 (13) Krasnok, A.; Makarov, S.; Petrov, M.; Savelev, R.; Belov, P.; Kivshar, Y. Towards All-Dielectric Metamaterials and Nanophotonics. SPIE Opt. Optoelect. 2015, 9502, 950203–950203. (14) Fan, X.; Zheng, W.; Singh, D. J. Light Scattering and Surface Plasmons on Small Spherical Particles. Light Sci. Appl. 2014, 3 (6), e179. (15) Pendry, J. B. Negative Refraction Makes a Perfect Lens. Phys. Rev. Lett. 2000, 85 (18), 3966–3969. (16) Shalaev, V. M. Optical Negative-Index Metamaterials. Nat. Photonics 2007, 1, 41. (17) García-Cámara, B.; Gómez-Medina, R.; Saenz, Jose, J.; Sepulveda, B. Sensing with Magnetic Dipolar Resonances in Semiconductor Nanospheres. Opt. Express 2013, 21 (20), 23007–23020. (18) Semouchkina, E.; Duan, R.; Semouchkin, G.; Pandey, R. Sensing Based on Fano-Type Resonance Response of All-Dielectric Metamaterials. Sensors 2015, 15 (4), 9344–9359. (19) Regmi, R.; Berthelot, J.; Winkler, P. M.; Mivelle, M.; Proust, J.; Bedu, F.; Ozerov, I.; Begou, T.; Lumeau, J.; Rigneault, H.; et al. All-Dielectric Silicon Nanogap Antennas to Enhance the Fluorescence of Single Molecules. Nano Lett. 2016, 16 (8), 5143–5151. (20) Yan, J.; Liu, P.; Lin, Z.; Yang, G. New Type High-Index Dielectric Nanosensors Based on the Scattering Intensity Shift. Nanoscale 2016, 8, 5996–6007. (21) Groep, J. Van De; Polman, a. Designing Dielectric Resonators on Substrates : Combining Magnetic and Electric Resonances. Opt. Express 2013, 21 (22), 1253–1257. (22) Oe-24-3-2047.Pdf.Crdownload. (23) Yu, Y. F.; Zhu, A. Y.; Fu, Y. H.; Luk, B.; Kuznetsov, A. I. High-Transmission Dielectric Metasurface with 2 π Phase. 2015, 418 (4), 412–418. (24) Butakov, N. A.; Schuller, J. A. Designing Multipolar Resonances in Dielectric Metamaterials. Nat. Publ. Gr. 2016, 6 (38487), 1–8. (25) Miller, M. M.; Lazarides, A. A. Sensitivity of Metal Nanoparticle Surface Plasmon Resonance to the Dielectric Environment. J. Phys. Chem. B 2005, 109 (46), 21556–21565. (26) Dahlin, A. B.; Tegenfeldt, J. O.; Hǒǒk, F. Improving the Instrumental Resolution of Sensors Based on Localized Surface Plasmon Resonance. Anal. Chem. 2006, 78 (13), 4416–4423. BIBLIOGRAPHY 99 (27) Singh, P. SPR Biosensors: Historical Perspectives and Current Challenges. Sensors Actuators, B Chem. 2016, 229, 110–130. (28) Haes, A. J.; Stuart, D. A.; Nie, S.; Van Duyne, R. P. Using Solution-Phase Nanoparticles, Surface-Confined Nanoparticle Arrays and Single Nanoparticles as Biological Sensing Platforms. J. Fluoresc. 2004, 14 (4), 355– 367. (29) Haes, A. J.; Chang, L.; Klein, W. L.; Van Duyne, R. P. Detection of a Biomarker for Alzheimer’s Disease from Synthetic and Clinical Samples Using a Nanoscale Optical Biosensor. J. Am. Chem. Soc. 2005, 127 (7), 2264–2271. (30) Haes, a. J.; Hall, W. P.; Chang, L.; Klein, W. L.; Van Duyne, R. P. A Localized Surface Plasmon Resonance Biosensor: First Steps toward an Assay for Alzheimer’s Disease. Nano Lett. 2004, 4 (6), 1029–1034. (31) Kim, S.-W.; Lee, J.-S.; Lee, S.-W.; Kang, B.-H.; Kwon, J.-B.; Kim, O.-S.; Kim, J.- S.; Kim, E.-S.; Kwon, D.-H.; Kang, S.-W. Easy-to-Fabricate and High- Sensitivity LSPR Type Specific Protein Detection Sensor Using AAO Nano- Pore Size Control. Sensors (Basel, Switzerland). April 2017. (32) Song, L.; Zhang, L.; Huang, Y.; Chen, L.; Zhang, G.; Shen, Z.; Zhang, J.; Xiao, Z.; Chen, T. Amplifying the Signal of Localized Surface Plasmon Resonance Sensing for the Sensitive Detection of Escherichia Coli O157:H7. Sci. Rep. 2017, 7 (1), 3288. (33) Acimovic, S. S.; Šípová, H.; Emilsson, G.; Dahlin, A. B.; Antosiewicz, T. J.; Käll, M. Superior LSPR Substrates Based on Electromagnetic Decoupling for On- a-Chip High-Throughput Label-Free Biosensing. Light Sci. Appl. 2017, 6 (8), 1–8. (34) Aćimović, S. S.; Ortega, M. A.; Sanz, V.; Berthelot, J.; Garcia-Cordero, J. L.; Renger, J.; Maerkl, S. J.; Kreuzer, M. P.; Quidant, R. LSPR Chip for Parallel, Rapid, and Sensitive Detection of Cancer Markers in Serum. Nano Lett. 2014, 14 (5), 2636–2641. (35) Chen, S.; Svedendahl, M.; Käll, M.; Gunnarsson, L.; Dmitriev, a. Ultrahigh Sensitivity Made Simple: Nanoplasmonic Label-Free Biosensing with an Extremely Low Limit-of-Detection for Bacterial and Cancer Diagnostics. Nanotechnology 2009, 20 (43), 434015. (36) Aćimović, S. S.; Kreuzer, M. P.; González, M. U.; Quidant, R. Plasmon Near- Field Coupling in Metal Dimers as a Step toward Single-Molecule Sensing. ACS Nano 2009, 3 (5), 1231–1237. (37) Feng, B.; Zhu, R.; Xu, S.; Chen, Y.; Di, J. A Sensitive LSPR Sensor Based on Glutathione-Functionalized Gold Nanoparticles on a Substrate for the BIBLIOGRAPHY 100 Detection of Pb2+Ions. RSC Adv. 2018, 8 (8), 4049–4056. (38) Krasnok, A.; Caldarola, M.; Bonod, N.; Alú, A. Spectroscopy and Biosensing with Optically Resonant Dielectric Nanostructures. 2017, No. January 2018. (39) Bontempi, N.; Chong, K. E.; Orton, H. W.; Staude, I.; Choi, D.-Y.; Alessandri, I.; Kivshar, Y. S.; Neshev, D. N. Highly Sensitive Biosensors Based on All- Dielectric Nanoresonators. Nanoscale 2017, 9 (15), 4972–4980. (40) Yang, Y.; Kravchenko, I. I.; Briggs, D. P.; Valentine, J. All-Dielectric Metasurface Analogue of Electromagnetically Induced Transparency. Nat. Commun. 2014, 5, 5753. (41) Davies, M. J.; Marques, M. P. C. Radhakrishnan, A. N. P. Microfluidics Theory in Practice. In Microfluidics in Detection Science; 2014; p pp 29–60. (42) Sia, S. K.; Whitesides, G. M. Microfluidic Devices Fabricated in Poly(Dimethylsiloxane) for Biological Studies. Electrophoresis 2003, 24 (21), 3563–3576. (43) Reynolds, O. An Experimental Investigation of the Circumstances Which Determine Whether the Motion of Water Shall Be Direct or Sinuous, and of the Law of Resistance in Parallel Channels. Philos. Trans. R. Soc. London 1883, 174, 935–982. (44) Melin, J.; Quake, S. R. Microfluidic Large-Scale Integration: The Evolution of Design Rules for Biological Automation. Annu. Rev. Biophys. Biomol. Struct. 2007, 36 (1), 213–231. (45) Garcia-Guirado, J.; Rica, R. A.; Ortega, J.; Medina, J.; Sanz, V.; Ruiz-Reina, E.; Quidant, R. Overcoming Diffusion-Limited Biosensing by Electrothermoplasmonics. ACS Photonics 2018, 5 (9), 3673–3679. (46) Unger, M. a; Chou, H. P.; Thorsen, T.; Scherer, A.; Quake, S. R. Monolithic Microfabricated Valves and Pumps by Multilayer Soft Lithography. Science (80-. ). 2000, 288 (5463), 113–116. (47) Bhagat, a. a S.; Kuntaegowdanahalli, S. S.; Papautsky, I. Inertial Microfluidics for Continuous Particle Filtration and Extraction. Microfluid. Nanofluidics 2009, 7 (2), 217–226. (48) Gossett, D. R.; Weaver, W. M.; MacH, A. J.; Hur, S. C.; Tse, H. T. K.; Lee, W.; Amini, H.; Di Carlo, D. Label-Free Cell Separation and Sorting in Microfluidic Systems. Anal. Bioanal. Chem. 2010, 397 (8), 3249–3267. (49) Selmi, M.; Gazzah, M. H.; Belmabrouk, H. Optimization of Microfluidic Biosensor Efficiency by Means of Fluid Flow Engineering. Sci. Rep. 2017, 7 (1), 1–11. BIBLIOGRAPHY 101 (50) Stanford University - Stanford Microfluidics Foundry https://web.stanford.edu/group/foundry/Microfluidic valve technology.html. (51) Bhadra, P.; Shajahan, M. S.; Bhattacharya, E.; Chadha, A. Studies on Varying N-Alkanethiol Chain Lengths on a Gold Coated Surface and Their Effect on Antibody-Antigen Binding Efficiency. RSC Adv. 2015, 5 (98), 80480–80487. (52) Ahmad, A.; Moore, E. Electrochemical Immunosensor Modified with Self- Assembled Monolayer of 11-Mercaptoundecanoic Acid on Gold Electrodes for Detection of Benzo[a]Pyrene in Water. Analyst 2012, 137 (24), 5839– 5844. (53) Kumar, A.; Rajesh, P.; Chaubey, A.; Grover, S. K.; Malhotra, B. D. Immobilization of Cholesterol Oxidase and Potassium Ferricyanide on Dodecylbenzene Sulfonate Ion-Doped Polypyrrole Film. J. Appl. Polym. Sci. 2001, 82 (14), 3486–3491. (54) Love, J. C.; Estroff, L. A.; Kriebel, J. K.; Nuzzo, R. G.; Whitesides, G. M. Self- Assembled Monolayers of Thiolates on Metals as a Form of Nanotechnology; 2005; Vol. 105. (55) Kamra, T.; Chaudhary, S.; Xu, C.; Montelius, L.; Schnadt, J.; Ye, L. Covalent Immobilization of Molecularly Imprinted Polymer Nanoparticles on a Gold Surface Using Carbodiimide Coupling for Chemical Sensing. J. Colloid Interface Sci. 2016, 461, 1–8. (56) Biswas, A.; Bayer, I. S.; Biris, A. S.; Wang, T.; Dervishi, E.; Faupel, F. Advances in Top-down and Bottom-up Surface Nanofabrication: Techniques, Applications & Future Prospects. Adv. Colloid Interface Sci. 2012, 170 (1–2), 2–27. (57) Chen, Y. Nanofabrication by Electron Beam Lithography and Its Applications: A Review. Microelectron. Eng. 2015, 135, 57–72. (58) Yang, S. M.; Jang, S. G.; Choi, D. G.; Kim, S.; Yu, H. K. Nanomachining by Colloidal Lithography. Small 2006, 2 (4), 458–475. (59) Yavas, O.; Acimovic, S. S.; Garcia-Guirado, J.; Berthelot, J.; Dobosz, P.; Sanz, V.; Quidant, R. Self-Calibrating on-a-Chip LSPR Sensing for Quantitative and Multiplexed Detection of Cancer Markers in Human Serum. ACS Sensors 2018, 3 (7), 1376–1384. (60) Lee, M. J.; Ye, A. S.; Gardino, A. K.; Heijink, A. M.; Sorger, P. K.; MacBeath, G.; Yaffe, M. B. Sequential Application of Anticancer Drugs Enhances Cell Death by Rewiring Apoptotic Signaling Networks. Cell 2012, 149 (4), 780– 794. BIBLIOGRAPHY 102 (61) Gonzalez-Angulo, A. M.; Hennessy, B. T. J.; Mills, G. B. Future of Personalized Medicine in Oncology: A Systems Biology Approach. J. Clin. Oncol. 2010, 28 (16), 2777–2783. (62) Donzella, V.; Crea, F. Optical Biosensors to Analyze Novel Biomarkers in Oncology. J. Biophotonics 2011, 4 (6), 442–452. (63) Hornbeck, P. V. Enzyme-Linked Immunosorbent Assays. Curr. Protoc. Immunol. 2015, 110 (1), 2–1. (64) McKeating, K. S.; Aubé, A.; Masson, J.-F. Biosensors and Nanobiosensors for Therapeutic Drug and Response Monitoring. Analyst 2016, 141 (2), 429– 449. (65) Vigneshvar, S.; Sudhakumari, C. C.; Senthilkumaran, B.; Prakash, H. Recent Advances in Biosensor Technology for Potential Applications – An Overview. Front. Bioeng. Biotechnol. 2016, 4 (February), 1–9. (66) Spindel, S.; Sapsford, K. E. Evaluation of Optical Detection Platforms for Multiplexed Detection of Proteins and the Need for Point-of-Care Biosensors for Clinical Use. Sensors 2014, 14 (12), 22313–22341. (67) Arlett, J. L.; Myers, E. B.; Roukes, M. L. Comparative Advantages of Mechanical Biosensors. Nat. Nanotechnol. 2011, 6 (4), 203–215. (68) Mukherji, S.; Punjabi, N. Label Free Integrated Optical Biosensors for Multiplexed Analysis. J. Indian Inst. Sci. 2012, 92 (21), 254–293. (69) Chen, P.; Chung, M. T.; Mchugh, W.; Nidetz, R.; Li, Y.; Fu, J.; Al, C. E. T. Multiplex Serum Cytokine Immunoassay Using Nanoplasmonic Biosensor Microarrays. ACS Nano 2015, 9 (4), 4173–4181. (70) Coskun, A. F.; Cetin, A. E.; Galarreta, B. C.; Alvarez, D. A.; Altug, H.; Ozcan, A. Lensfree Optofluidic Plasmonic Sensor for Real-Time and Label-Free Monitoring of Molecular Binding Events Over a Wide Field-of-View. Sci. Rep. 2014, 4, 1–7. (71) Cappi, G.; Spiga, F. M.; Moncada, Y.; Ferretti, A.; Beyeler, M.; Bianchessi, M.; Decosterd, L.; Buclin, T.; Guiducci, C. Label-Free Detection of Tobramycin in Serum by Transmission- Localized Surface Plasmon Resonance. Anal. Chem. 2015, 87 (10), 5278–5285. (72) Ruemmele, J. A.; Hall, W. P.; Ruvuna, L. K.; Van Duyne, R. P. A Localized Surface Plasmon Resonance Imaging Instrument for Multiplexed Biosensing. Anal. Chem. 2013, 85 (9), 4560–4566. (73) Unser, S.; Bruzas, I.; He, J.; Sagle, L. Localized Surface Plasmon Resonance Biosensing: Current Challenges and Approaches. Sensors 2015, 15 (7), BIBLIOGRAPHY 103 15684–15716. (74) Yonzon, C. R.; Jeoung, E.; Zou, S.; Schatz, G. C.; Mrksich, M.; Van Duyne, R. P. A Comparative Analysis of Localized and Propagating Surface Plasmon Resonance Sensors: The Binding of Concanavalin A to a Monosaccharide Functionalized Self-Assembled Monolayer. J. Am. Chem. Soc. 2004, 126 (39), 12669–12676. (75) Anker, J. N.; Hall, W. P.; Lyandres, O.; Shah, N. C.; Zhao, J.; Van Duyne, R. P. Biosensing with Plasmonic Nanosensors. Nat. Mater. 2008, 7 (6), 442–453. (76) Joshi, G. K.; Deitz-McElyea, S.; Liyanage, T.; Lawrence, K.; Mali, S.; Sardar, R.; Korc, M. Label-Free Nanoplasmonic-Based Short Noncoding RNA Sensing at Attomolar Concentrations Allows for Quantitative and Highly Specific Assay of MicroRNA-10b in Biological Fluids and Circulating Exosomes. ACS Nano 2015, 9 (11), 11075–11089. (77) Yu, C.; Irudayaraj, J. Multiplex Biosensor Using Gold Nanorods. Anal. Chem. 2007, 79 (2), 572–579. (78) Wang, Y.; Tang, L. Multiplexed Gold Nanorod Array Biochip for Multi- Sample Analysis. Biosens. Bioelectron. 2015, 67, 18–24. (79) Ouellet, E.; Lausted, C.; Lin, T.; Yang, C. W. T.; Hood, L.; Lagally, E. T. Parallel Microfluidic Surface Plasmon Resonance Imaging Arrays. Lab Chip 2010, 10 (5), 581. (80) Luo, Y.; Yu, F.; Zare, R. N. Microfluidic Device for Immunoassays Based on Surface Plasmon Resonance Imaging. Lab Chip 2008, 8 (5), 694. (81) Sackmann, E. K.; Fulton, A. L.; Beebe, D. J. The Present and Future Role of Microfluidics in Biomedical Research. Nature 2014, 507 (7491), 181–189. (82) Chin, C. D.; Linder, V.; Sia, S. K. Lab on a Chip Commercialization of Microfluidic Point-of-Care Diagnostic Devices. Lab Chip 2012, 12, 2118– 2134. (83) Volpetti, F.; Garcia-Cordero, J.; Maerkl, S. J. A Microfluidic Platform for High-Throughput Multiplexed Protein Quantitation. PLoS One 2015, 10 (2). (84) Norum, L. F.; Erikstein, B.; Nustad, K. Elevated CA 125 in Breast Cancer –A Sign of Advanced Disease. Tumor Biol. 2001, No. 22, 223–228. (85) Esteva, F. J.; Cheli, C. D.; Fritsche, H.; Fornier, M.; Slamon, D.; Thiel, R. P.; Luftner, D.; Ghani, F. Clinical Utility of Serum HER2/Neu in Monitoring and Prediction of Progression-Free Survival in Metastatic Breast Cancer Patients Treated with Trastuzumab-Based Therapies. Breast Cancer Res. 2005, 7 (4), 436–443. BIBLIOGRAPHY 104 (86) Duffy, M. J. Tumor Markers in Clinical Practice: A Review Focusing on Common Solid Cancers. Med. Princ. Pract. 2012, 22 (1), 4–11. (87) Stieber, P.; Untch, M.; Nagel, D.; Seidel, D. Serum CEA and CA 15-3 as Prognostic Factors in Primary Breast Cancer. Br. J. Cancer 2002, 86 (February), 1217–1222. (88) Weigelt, B.; Peterse, J. L.; van ’t Veer, L. J. Breast Cancer Metastasis: Markers and Models. Nat. Rev. Cancer 2005, 5 (August), 591–602. (89) Marella, D. S. Prognostic and Predictive Markers in Early Detection of Different Types of Cancers for Selected Organ Sites. IOSR J. Pharm. Biol. Sci. 2013, 8 (4), 25–42. (90) Harris, L.; Fritsche, H.; Mennel, R.; Norton, L.; Ravdin, P.; Taube, S.; Somerfield, M. R.; Hayes, D. F.; Bast, R. C. American Society of Clinical Oncology 2007 Update of Recommendations for the Use of Tumor Markers in Breast Cancer. J. Clin. Oncol. 2007, 25 (33), 5287–5312. (91) Fehm, T.; Becker, S.; Duerr-Stoerzer, S.; Sotlar, K.; Mueller, V.; Wallwiener, D.; Lane, N.; Solomayer, E.; Uhr, J. Determination of HER2 Status Using Both Serum HER2 Levels and Circulating Tumor Cells in Patients with Recurrent Breast Cancer Whose Primary Tumor Was HER2 Negative or of Unknown HER2 Status. Breast Cancer Res. 2007, 9 (5), 1–8. (92) Duffy, M. J. Serum Tumor Markers in Breast Cancer: Are They of Clinical Value? Clin. Chem. 2006, 52 (3), 345–351. (93) Pultz, B. dos A.; da Luz, F. A. C.; de Faria, P. R.; Oliveira, A. P. L.; de Ara??jo, R. A.; Silva, M. J. B. Far Beyond the Usual Biomarkers in Breast Cancer: A Review. J. Cancer 2014, 5 (7), 559–571. (94) Svedendahl, M.; Chen, S.; Dmitriev, A.; Kall, M. Refractometric Sensing Using Propagating versus Localized Surface Plasmons: A Direct Comparison. Nano Lett. 2009, 9 (12), 4428–4433. (95) Laforte, V.; Juncker, D.; Li, H. Cross-Reactivity in Antibody Microarrays and Multiplexed Sandwich Assays : Shedding Light on the Dark Side of Multiplexing. Curr. Opin. Chem. Biol. 2014, 18, 29–37. (96) Oubina, A.; Gascon, J.; Barcelo, D. Multianalyte Effect in the Determination of Cross-Reactivities of Pesticide Immunoassays in Water Matrices. Anal. Chim. Acta 1997, 347 (97), 121–130. (97) Darwish, I. A. Immunoassay Methods and Their Applications in Pharmaceutical Analysis : Basic Methodology and Recent Advances. Int. J. Biomed. Sciense 2006, 2 (2), 217–235. BIBLIOGRAPHY 105 (98) Otte, M. A.; Estévez, M. C.; Carrascosa, L. G.; González-Guerrero, A. B.; Lechuga, L. M.; Sepúlveda, B. Improved Biosensing Capability with Novel Suspended Nanodisks. J. Phys. Chem. C 2011, 115 (13), 5344–5351. (99) Chung, T.; Lee, S. Y.; Song, E. Y.; Chun, H.; Lee, B. Plasmonic Nanostructures for Nano-Scale Bio-Sensing. Sensors 2011, 11 (11), 10907–10929. (100) Liang, C.-C.; Liao, M.-Y.; Chen, W.-Y.; Cheng, T.-C.; Chang, W.-H.; Lin, C.-H. Plasmonic Metallic Nanostructures by Direct Nanoimprinting of Gold Nanoparticles. Opt. Express 2011, 19 (5), 4768. (101) Pang, S.; Cowen, S. A Generic Standard Additions Based Method to Determine Endogenous Analyte Concentrations by Immunoassays to Overcome Complex Biological Matrix Interference. Sci. Rep. 2017, 7 (1), 1– 10. (102) Kaminski, T.; Gunnarsson, A.; Geschwindner, S. Harnessing the Versatility of Optical Biosensors for Target-Based Small-Molecule Drug Discovery. ACS Sensors 2017, 2 (1), 10–15. (103) Yavas, O.; Svedendahl, M.; Dobosz, P.; Sanz, V.; Quidant, R. On-a-Chip Biosensing Based on All-Dielectric Nanoresonators. Nano Lett. 2017, 17 (7), 4421–4426. (104) Long, F.; Zhu, A.; Shi, H. Recent Advances in Optical Biosensors for Environmental Monitoring and Early Warning. Sensors 2013, 13 (10), 13928–13948. (105) Gavela, A. F.; García, D. G.; Ramirez, J. C.; Lechuga, L. M. Last Advances in Silicon-Based Optical Biosensors. Sensors 2016, 16 (285), 1–15. (106) Passaro, V. M. N.; Dell’Olio, F.; Casamassima, B.; De Leonardis, F. Guided- Wave Optical Biosensors. Sensors 2007, 7 (4), 508–536. (107) Barizuddin, S.; Bok, S.; Gangopadhyay, S. Nanomedicine & Nanotechnology Plasmonic Sensors for Disease Detection - A Review. J. Nanomedicine Nanotechnol. 2016, 7 (3), 1000373. (108) Homola, J. Present and Future of Surface Plasmon Resonance Biosensors. Anal. Bioanal. Chem. 2003, 377 (3), 528–539. (109) Nguyen, H. H.; Park, J.; Kang, S.; Kim, M. Surface Plasmon Resonance: A Versatile Technique for Biosensor Applications. Sensors 2015, 15 (5), 10481–10510. (110) Leung, A.; Shankar, P. M.; Mutharasan, R. A Review of Fiber-Optic Biosensors. Sensors Actuators, B Chem. 2007, 125 (2), 688–703. BIBLIOGRAPHY 106 (111) Caucheteur, C.; Guo, T.; Albert, J. Review of Plasmonic Fiber Optic Biochemical Sensors: Improving the Limit of Detection. Anal. Bioanal. Chem. 2015, 407 (14), 3883–3897. (112) Cao, J.; Sun, T.; Grattan, K. T. V. Gold Nanorod-Based Localized Surface Plasmon Resonance Biosensors: A Review. Sensors Actuators, B Chem. 2014, 195, 332–351. (113) Hammond, J.; Bhalla, N.; Rafiee, S.; Estrela, P. Localized Surface Plasmon Resonance as a Biosensing Platform for Developing Countries. Biosensors 2014, 4 (2), 172–188. (114) Cambiasso, J.; Grinblat, G.; Li, Y.; Rakovich, A.; Cortés, E.; Maier, S. A. Bridging the Gap between Dielectric Nanophotonics and the Visible Regime with Effectively Lossless GaP Antennas. Nano Lett. 2017, acs.nanolett.6b05026. (115) Li, J.; Verellen, N.; Vercruysse, D.; Bearda, T.; Lagae, L.; Van Dorpe, P. All- Dielectric Antenna Wavelength Router with Bidirectional Scattering of Visible Light. Nano Lett. 2016, 16 (7), 4396–4403. (116) Shalaev, M. I.; Sun, J.; Tsukernik, A.; Pandey, A.; Nikolskiy, K.; Litchinitser, N. M. High-Efficiency All-Dielectric Metasurfaces for Ultracompact Beam Manipulation in Transmission Mode. Nano Lett. 2015, 15 (9), 6261–6266. (117) Shcherbakov, M. R.; Vabishchevich, P. P.; Shorokhov, A. S.; Chong, K. E.; Choi, D. Y.; Staude, I.; Miroshnichenko, A. E.; Neshev, D. N.; Fedyanin, A. A.; Kivshar, Y. S. Ultrafast All-Optical Switching with Magnetic Resonances in Nonlinear Dielectric Nanostructures. Nano Lett. 2015, 15 (10), 6985–6990. (118) Wu, A.; Li, H.; Du, J.; Ni, X.; Ye, Z.; Wang, Y.; Sheng, Z.; Zou, S.; Gan, F.; Zhang, X.; et al. Experimental Demonstration of In-Plane Negative-Angle Refraction with an Array of Silicon Nanoposts. Nano Lett. 2015, 15 (3), 2055–2060. (119) Zhang, J.; Liu, W.; Zhu, Z.; Yuan, X.; Qin, S. Strong Field Enhancement and Light-Matter Interactions with All-Dielectric Metamaterials Based on Split Bar Resonators. Opt. Express 2014, 22 (25), 30889–30898. (120) Evlyukhin, A. B.; Novikov, S. M.; Zywietz, U.; Eriksen, R. L.; Reinhardt, C.; Bozhevolnyi, S. I.; Chichkov, B. N. Demonstration of Magnetic Dipole Resonances of Dielectric Nanospheres in the Visible Region. Nano Lett. 2012, 12 (7), 3749–3755. (121) Evlyukhin, A. B.; Reinhardt, C.; Seidel, A.; Luk’Yanchuk, B. S.; Chichkov, B. N. Optical Response Features of Si-Nanoparticle Arrays. Phys. Rev. B - Condens. Matter Mater. Phys. 2010, 82 (4), 1–12. BIBLIOGRAPHY 107 (122) Evlyukhin, A. B.; Eriksen, R. L.; Cheng, W.; Beermann, J.; Reinhardt, C.; Petrov, A.; Prorok, S.; Eich, M.; Chichkov, B. N.; Bozhevolnyi, S. I. Optical Spectroscopy of Single Si Nanocylinders with Magnetic and Electric Resonances. Sci. Rep. 2014, 4, 4126. (123) Habteyes, T. G.; Staude, I.; Chong, K. E.; Dominguez, J.; Decker, M.; Miroshnichenko, A.; Kivshar, Y.; Brener, I. Near-Field Mapping of Optical Modes on All-Dielectric Silicon Nanodisks. ACS Photonics 2014, 1, 794–798. (124) Pei, X.; Zhang, B.; Tang, J.; Liu, B.; Lai, W.; Tang, D. Sandwich-Type Immunosensors and Immunoassays Exploiting Nanostructure Labels: A Review. Anal. Chim. Acta 2013, 758, 1–18. (125) Miroshnichenko, A. E.; Evlyukhin, A. B.; Yu, Y. F.; Bakker, R. M.; Chipouline, A.; Kuznetsov, A. I.; Luk’yanchuk, B.; Chichkov, B. N.; Kivshar, Y. S. Nonradiating Anapole Modes in Dielectric Nanoparticles. Nat. Commun. 2015, 6, 8069. (126) Catalona, W. J., J. P. Richie, F. R. Ahmann, M. A. Hudson, P. T. Scardino, R. C. Flanigan, J. B. deKernion. Comparison of DRE and Serum PSA in the Early Detection of Prostate Cancer: Results of a Multicenter Clinical Trial of 6630 Men. J. Urol. 1994, 151, 1283–1290. (127) Brawer, MK. Chetner, MP. Beatie, J. Buchner, DM. Vessella, RL. Lange, P. Screening for Prostatic Carcinoma with Prostate Specific Antigen. J. Urol. 1992, 147, 841–845. (128) Liu, R.; Zhang, Y.; Zhang, S.; Qiu, W.; Gao, Y. Silver Enhancement of Gold Nanoparticles for Biosensing: From Qualitative to Quantitative. Appl. Spectrosc. Rev. 2014, 49 (2), 121–138. (129) Yavas, O.; Svedendahl, M.; Quidant, R. Unravelling the Role of Electric and Magnetic. ACS Nano 2019, 13, 4582–4588. (130) Feuz, L.; Jonsson, M. P.; Höök, F. Material-Selective Surface Chemistry for Nanoplasmonic Sensors: Optimizing Sensitivity and Controlling Binding to Local Hot Spots. Nano Lett. 2012, 12 (2), 873–879. (131) Dahlin, A. B.; Chen, S.; Jonsson, M. P.; Gunnarsson, L.; Käll, M.; Höök, F. High-Resolution Microspectroscopy of Plasmonic Nanostructures for Miniaturized Biosensing. Anal. Chem. 2009, 81 (16), 6572–6580. (132) Wersäll, M.; Verre, R.; Svedendahl, M.; Johansson, P.; Käll, M.; Shegai, T. Directional Nanoplasmonic Antennas for Self-Referenced Refractometric Molecular Analysis. J. Phys. Chem. C 2014, 118 (36), 21075–21080. (133) Chen, S.; Svedendahl, M.; Duyne, R. P. V.; Käll, M. Plasmon-Enhanced