Utilizing Nanostructures and Nano-Mechanics for Sensitive Analyte Detection Via Surface Enhanced Raman Spectroscopy (SERS) and Micro-Cantilever Sensing Platforms

University of Tennessee, Knoxville TRACE: Tennessee Research and Creative Exchange

Doctoral Dissertations Graduate School

5-2017

Utilizing Nanostructures and Nano-Mechanics for Sensitive Analyte Detection via Surface Enhanced Raman Spectroscopy (SERS) and Micro-Cantilever Sensing Platforms

Ryan Andrew Wallace University of Tennessee, Knoxville, [email protected]

Follow this and additional works at: https://trace.tennessee.edu/utk_graddiss

Part of the Analytical Chemistry Commons

Recommended Citation Wallace, Ryan Andrew, "Utilizing Nanostructures and Nano-Mechanics for Sensitive Analyte Detection via Surface Enhanced Raman Spectroscopy (SERS) and Micro-Cantilever Sensing Platforms. " PhD diss., University of Tennessee, 2017. https://trace.tennessee.edu/utk_graddiss/4434

This Dissertation is brought to you for free and open access by the Graduate School at TRACE: Tennessee Research and Creative Exchange. It has been accepted for inclusion in Doctoral Dissertations by an authorized administrator of TRACE: Tennessee Research and Creative Exchange. For more information, please contact [email protected]. To the Graduate Council:

I am submitting herewith a dissertation written by Ryan Andrew Wallace entitled "Utilizing Nanostructures and Nano-Mechanics for Sensitive Analyte Detection via Surface Enhanced Raman Spectroscopy (SERS) and Micro-Cantilever Sensing Platforms." I have examined the final electronic copy of this dissertation for form and content and recommend that it be accepted in partial fulfillment of the equirr ements for the degree of Doctor of Philosophy, with a major in Chemistry.

Michael J. Sepaniak, Major Professor

We have read this dissertation and recommend its acceptance:

Tessa Calhoun, Jimmy W. Mays, Panos G. Datskos

Accepted for the Council:

Dixie L. Thompson

Vice Provost and Dean of the Graduate School

(Original signatures are on file with official studentecor r ds.) Utilizing Nanostructures and Nano-Mechanics for Sensitive Analyte Detection via Surface

Enhanced Raman Spectroscopy (SERS) and Micro-Cantilever Sensing Platforms

A Dissertation Presented for the

Doctor of Philosophy

Degree

The University of Tennessee, Knoxville

Ryan Andrew Wallace

May 2017 Dedication

To my very intelligent and beautiful wife who has stuck with me and kept me moving forward,

and our soon to be born baby girl.

You are everything to me.

Acknowledgements

First off, I want to thank my research mentor and advisor, Dr. Michael J. Sepaniak, who saw enough potential in me to be a part of his research group. You pushed me when I needed it the most and helped me grow into the scientist that I am today. You have inspired me to work harder and smarter, and I will forever be grateful for your support throughout my graduate career.

Next, I would like to thank the members of my committee, Dr. Tessa Calhoun, Dr. Jimmy

Mays, Dr. Joe Zhuang, and Dr. Panos G. Datskos for challenging me and holding me accountable during the different stages of my graduate career.

I would also like to give additional thanks to Dr. Panos Datskos for allowing me to work with his group at Oak Ridge National Laboratory the last 2 years. Being able to work with other professional scientists and having the exposure to national laboratory protocols has been invaluable for my career. Your guidance during my research and life advice is greatly appreciated.

I would also like to acknowledge Dr. Nickolay Lavrik who helped the Sepaniak group liaise with the Center for Nanophase Materials Science and provide valuable insight into research questions. Thank you for taking the time to help us solve problems and learn different instrumental techniques for micro- and nano-fabrication.

I owe a lot of thanks to the former and remaining members of the Sepaniak group: Tess,

Jennifer, Nahla, Nichole, Danielle, and Rachel. You are the ones that I could bounce ideas off and were not afraid to let me know when I had no idea what I was talking about. Thank you for the group lunches and comradery.

iii

Thank you to my best friends, Casey, Dustin, Kevin, and Alex who have kept me level headed and brought me back down to Earth so I wouldn’t go crazy during this process.

I next want to acknowledge my family that has stood behind me and made the man I am today:

My mother, Karen Wallace, has always been there, through thick and thin, to listen to my problems and offer solutions throughout my life. Watching you work as hard as you have over these years has always inspired me to have the kind of work ethic that would make you proud of me. You are the greatest mother a son could ask for.

My father, Brian Wallace has supported me through the bad times and has been there when I needed him the most. The pep talks and invaluable advice you have given to me over the years will forever be cherished.

My brother Nathan has always looked up to me throughout my life, and he is one of the reasons I have worked so hard to get to where I have gotten. Nathan, I have always wanted to be the example from which to follow, but your dedication and work ethic during your own doctoral research has inspired me to work even harder. Thank you for being there for me during my struggles. I couldn’t ask for a better brother and friend.

My sisters Morgan and Madison have been great throughout my life and research career.

They are not afraid to challenge their older brother, but they also listen. You are the best sisters I could ask for.

My in-laws, Ron and April Campos, for their endless love and support. You have come to my rescue so many times and have always treated me like your own son. Thank you!

Additionally, I want to thank my grandparents Terry and Peggy Wallace and Louis and

Joan Fitzgerald for your infinite knowledge and love. You are the examples I want to follow.

I love you all.

Finally, I must acknowledge my best friend in the whole world, Carol Wallace. You are the greatest thing that has ever happened to me. I wake up every day happier than the day before knowing that you are by my side. The patience and faith you have shown throughout my graduate career is incredible and I want to thank you for that. We have faced many challenges and obstacles in life so far, and I can’t wait to face even more. I can’t wait to see our baby girl enter the world! Thank you for being my rock. I love you so much!

Abstract

The purpose of this dissertation is to present the effective utilization of nano-structures and nano-mechanics in conjunction with surface enhanced Raman spectroscopy (SERS) and micro-cantilever (MC) mechanical sensors for sensitive analytical detection. One of the most important attributes an Analytical Chemist can possess is the ability to develop and efficiently use the tools provided to obtain precise and accurate information that can be effectively communicated. The following is a brief outline of the background concepts and studies that will be present herein.

A discussion of SERS will be presented in which the history and concepts behind the technique will be communicated as it pertains to the work in this dissertation. The theory behind

SERS, including the electromagnetic effect and chemical effect will be conveyed, along with two separate manuscripts that include the use of SERS detection in micro-fabricated pillar arrays that have been enhanced through the presence of silver colloid. Each manuscript included an analytical treatment of the results to ensure a complete study of the system.

A second discussion will include thin-layer chromatography development and theories that can be applied to one of the manuscripts presented. Some of the metrics involved in separations in ultra-thin layer chromatography (UTLC) will be examined, along with a study combining SERS with UTLC to demonstrate an efficient separation and detection platform. The fabrication of the pillar arrays will also be described.

The final discussion and study will involve the use of micro-cantilevers (MCs) as a mechanical sensor. The discussion will include a brief history of MCs and description of the mechanics involved. Detection methods and optimization of the system will also be discussed.

The study will propose the use of a porous silicon oxide active layer on the MCs as a method of detecting trace HF gas sensitively. An analytical treatment of the results is also included.

vii

Table of Contents

Chapter 1 - Overview of Surface Enhanced Raman Spectroscopy (SERS) ...... 1

1.1 Introduction ...... 2

1.2 Raman Spectroscopy ...... 2

1.3 Discovery and Development of SERS ...... 5

1.4 Enhancement ...... 7

1.5.1 Random SERS Substrates ...... 9

1.5.2 Engineered SERS Substrates ...... 11

1.6 Conclusion ...... 11

1.7 References ...... 14

Chapter 2 - Fabrication and Use of Pillar Arrays for Planar Chromatography with SERS Detection ...... 16

2.1 Overview of Chromatography ...... 17

2.2 TLC History and Concept ...... 18

2.3 Van Deemter Equation ...... 19

2.4 Deterministic Pillar Array Fabrication ...... 24

2.5 Stochastic Pillar Array Fabrication ...... 27

2.6 Superhydrophobicity ...... 27

2.7 Conclusion ...... 31

2.8 References ...... 33

Chapter 3 - Overview of Micro-Cantilever (MC) Sensors ...... 34

3.1 Micro-cantilever (MC) Background ...... 35

3.2 MC Measurement Methods – Dynamic Mode ...... 37

3.3 MC Measurement Methods – Static Mode ...... 39

3.4 MC Readout Methods ...... 43

viii

3.5 Conclusion ...... 44

3.6 References ...... 45

Chapter 4 - Superhydrophobic Analyte Concentration Utilizing Colloid-Pillar Array SERS Substrates ...... 46

4.1 Abstract ...... 47

4.2 Introduction ...... 48

4.3 Materials ...... 50

4.4 Preparation of Silver Colloid ...... 50

4.5 Fabrication of Deterministic Pillar Arrays ...... 50

4.6 Fabrication of Stochastic Pillar Arrays ...... 51

4.7 Silver Colloid Delivery to the Pillar Array Surface ...... 52

4.8 Analyte Delivery to the Pillar Array Surface ...... 55

4.9 Droplet Size and Composition ...... 56

4.10 Determination of the Concentration Factor, Dynamic Range, Detection Limit, and Reproducibility ...... 59

4.11 Electro-osmotic Delivery of Analyte for Multiplexing ...... 61

4.12 Concentration of Analyte and Wicking ...... 62

4.13 Conclusions ...... 64

4.14 Acknowledgements ...... 66

Chapter 5 - Ultra-Thin Layer Chromatography with Integrated Silver Colloid-Based SERS Detection ...... 69

5.1 Abstract ...... 70

5.2 Introduction ...... 70

5.3 Materials ...... 72

5.4 Fabrication of Deterministic Pillar Arrays ...... 73

5.5 Preparation of the Silver Colloid-Pillar Array Surface ...... 73

5.6 Analyte Delivery to the Surface and Development ...... 74

5.7 Silver Colloid Stability ...... 74

5.8 Fluorescence UTLC-SERS Separations Utilizing Pillar Array Platforms ... 77

5.9 UTLC-SERS Separations with Pillar Arrays ...... 82

5.10 Conclusions ...... 83

5.11 Acknowledgements ...... 85

5.12 References ...... 86

Chapter 6 - Evaluation of Porous Silicon Oxide on Silicon Micro-Cantilevers for Sensitive Detection of Gaseous HF ...... 88

6.1 Abstract ...... 89

6.2 Introduction ...... 89

6.3 Materials ...... 91

6.4 MC Preparation and PSO Deposition ...... 91

6.5 Controlled Exposure of Hydrogen Fluoride Gas to Porous Silicon Oxide .. 92

6.6 Evaluation of Shielding and Mechanically Distorting MCs and the Subsequent Deposition of Porous Silicon Oxide ...... 92

6.7 Physical Changes in Bending of MCs After Trace HF Gas Exposure ...... 97

6.8 Analytics of Trace HF Gas Exposure to MCs with a Porous Silicon Oxide Layer ...... 98

6.9 Conclusions ...... 101

6.10 Acknowledgements ...... 101

6.11 References ...... 102

Chapter 7 - Concluding Remarks ...... 103

Vita………………………………………………………………………………………………………………106

List of Figures

Figure 1.2.1: Summary of the changes in photon energy for different situations, including infrared absorption, Rayleigh scattering, Stokes Raman scattering, and Anti-Stokes Raman scattering...... 4

Figure 1.2.2: Graphical summary of the optics and detection scheme of the Jobin Yvon Labram Raman Spectrometer used in this work...... 6

Figure 1.4.1: Example of the plasmon resonance that occurs in a metal nanoparticle after irradiation by a light source, such as a laser...... 8

Figure 1.5.2.1: (a) SEM image of the silver films-over nanosphere substrates developed by the Van Duyne group. (b) SEM image of silver colloid deposition within a photolithographic pillar array...... 12

Figure 2.2.1: Diagram of a TLC experiment depicting a separation of a three component mixture...... 20

Figure 2.3.1: A sample plot of the van Deemter equation in which the contributions of the A, B, and C terms can be visualized and compared to the actual plot. The optimum efficiency and velocity are also indicated at the minimum of the actual plot...... 23

Figure 2.4.1: (a-g) Images depicting the several different steps involved in the fabrication of the deterministic pillar arrays. (h) A SEM image of an actual array of the deterministic pillars is included...... 26

Figure 2.5.1: (a-d) Images depicting the different steps involved in the fabrication of the stochastic pillar arrays. (e) A SEM image of an actual array of the stochastic pillars is included...... 28

Figure 2.6.1: (a) The Cassie-Baxter, heterogeneous, wetting state in which the air pockets under the water droplet, within the roughened surface and (b) the Wenzel, homogenous, wetting state in which the water droplet penetrates the roughened surface...... 30

Figure 3.1.1: Image of typical MC defining the thickness, width, and length...... 36

Figure 3.1.2: SEM images of (a) triangular MC made by the Datskos group, (b) traditional rectangular MC purchased from MikroMasch, and (c) another MC geometry made by the Electronics Institute of Technology...... 38

Figure 3.3.1: Diagram of a typical static mode experiment in which the change in position of the laser on the PSD is shown before and after the MC has bent...... 40

Figure 3.3.2: Diagram of the three types of stress on an MC including (a) analyte adsorption onto the surface and subsequent expansion of the MC surface, (b) analyte-induced deformation of a MC in the form of swelling, and (c) analyte-induced deformation of MC due to nanostructured modifying phase.20-21 ...... 42

Figure 4.7.1: SEM images of different colloidal delivery systems including (a) development of Ag colloid in ethanol up the pillar array via capillary action, (b) soaking the deterministic pillar array in the Ag colloid solution with water as the solvent, (c) soaking the deterministic pillar array in the Ag colloid solution with 50% ethanol and 50% water as the solvent, and (d) soaking the stochastic pillar array in the Ag colloid solution with 50% ethanol and 50% water as the solvent...... 54

Figure 4.9.1: (a) Comparison of the contact area of droplets of varying sizes on top of the deterministic and stochastic pillars and (b) comparison of the estimated contact angle of 5 μL droplets consisting of varying amounts of different organic solvents...... 58

Figure 4.10.1: (a) Calibration plot of the averaged MIT peak at 1318 cm-1 vs the concentration of MIT and (b) a comparison of the SERS spectra of concentrated 12 pM MIT vs developed 1.2 nM MIT...... 60

Figure 4.11.1: (a) Schematic diagram of the electro-osmotic delivery system used along with the (b) actual droplets that were spotted and allowed to evaporate on the superhydrophobic substrate. (c) The representative rasters of the droplet spots after evaporation and the largest SERS spectrum found in each droplet spot are presented to demonstrate the ability to multiplex...... 63

Figure 4.12.1: Demonstration of the mass transfer and stacking of R6G at the border between a hydrophilic and hydrophobic zone on a deterministic pillar array. (a) Images of the fluorescence at the border were taken over time, and (b) the intensity of the fluorescence was measured, along

xii with the SERS intensity at 1507 cm-1. (c) A representative SERS spectrum of the R6G is shown as well...... 65

Figure 5.7.1: SEM images of (a) before and (b) after the chromatographic development with the 60% ethanol and 40% water mobile phase, showing the silver colloid on the surface...... 76

Figure 5.7.2: Silver colloid stability studies including (a-b) measuring the SERS activity of MIT on top of Ag colloid that was sprayed on top of a cleaned microscope slide over a period of 2 days and (c-d) measuring the SERS activity of MIT within the superhydrophobic pillar arrays which contained the Ag colloid over a period of 6 days...... 78

Figure 5.8.1: Separation of resorufin and sulforhodamine 640, including (a) a chromatogram using the strongest peak for each component and retardation factor as a measure of distance traveled. In addition, the fluorescence image and the strongest SERS spectrum for each band after separation is included for the original spot after development (b&e), the resorufin band (c&f), and the sulforhodamine 640 band (d&g)...... 80

Figure 5.8.2: Separation of A1, D1, and sulforhodamine 640 including (a) a chromatogram using the strongest peak for each component and retardation factor as a measure of distance traveled. In addition, the fluorescence image and strongest SERS spectrum for each band after separation is included for the original spot after development (b&f), the D1 band (c&g), the A1 band (d&h), and the sulforhodamine 640 band (e&i). Note that the A1 and D1 are distinguished by chromatography not by Raman spectra...... 81

Figure 5.9.1: Separation of five purine and pyrimidine bases, including (a) a chromatogram using the strongest peak for each component and retardation factor as a measure of distance traveled. In addition, the strongest SERS spectrum for each band after separation is included for (b) the original spot after development, (c) adenine, (d) hypoxanthine, (e) guanine, (f) thymine, and (g) cytosine...... 84

Figure 6.5.1: Diagram of the experimental set-up for controlled exposure of diluted HF gaseous samples to PSO modified MCs placed in a sample cell and the subsequent detection of bending changes via optical beam bending...... 93

Figure 6.6.1: SEM images of the (a) top and (b) bottom of a fully distorted MC in which 200 nm of PSO was deposited. SEM images of (c) the end of a MC with a base underneath and before

xiii

200 nm of PSO has been deposited, and (d) the same cantilever after deposition of 200 nm PSO. In addition, (e) profilometry measurements along the length of the cantilever prior to PSO deposition with the base present, after PSO deposition and before the base is removed, and after the base has been removed...... 95

Figure 6.6.2: (a) Images depicting the position of the MC base in reference to the native, halfway bent, and fully bent set of MCs. Average change in signal in the PSD for five different individual MCs exposed to different levels of gases and bending that include (b) a fully distorted set of MCs exposed to saturated air, a native set of MCs exposed to 100 ppm HF gas, a partially distorted set of MCs exposed to 100 ppm HF gas, and a fully distorted set of MCs exposed to 100 ppm HF. (c) Additionally, a comparison of the voltage changes when 1000 ppm HCl gas is added to 100 ppm HF with fully bent MCs...... 96

Figure 6.7.1: Optical profilometry of the MCs showing the changes in the bending before and after exposure to 100 ppm HF gas in 90% humid air for the (a)native, (b)partially distorted, and (c)fully distorted MCs. Average signal from five MCs from the same set, with three different sets being used for each type of bending...... 99

Figure 6.8.1: Calibration curve of the average slope of the initial drop in voltage signal observed via the PSD using varying concentrations of HF gas ranging from 30 ppm to 3000 ppm...... 100

xiv

Abbreviations and Symbols

∅푆 Ratio of Total Area

∗ 휃푤 Apparent Wenzel Contact Angle

휎푠푐 Raman Cross Section

µ Dipole Moment

A Eddy Diffusion

A1 Adriamycin

AFM Atomic Force Microscopy

Ag Silver

AgFON Silver Films-Over Nanosphere

ATP Aminothiophenol

B Longitudinal Molecular Diffusion

CM Resistance to Mass Transfer Mobile Phase

CS Resistance to Mass Transfer Stationary Phase

CNMS Center for Nanophase Materials Science

CV Crystal Violet d Distance Traveled by Analyte

D1 Daunomycin dA Area of Cross Section df Average Film Thickness of Stationary Phase

DM Diffusion Coefficient for Mobile Phase

DOPs Discs-on-Pillars dp Particle Diameter

DS Diffusion Coefficient for Stationary Phase

E Magnitude of Electromagnetic Wave

E Young’s modulus of Micro-cantilever

E0 Peak Amplitude of Electromagnetic Wave

EBL Electron Beam Lithography

EDTA Ethylenediaminetetraactic Acid

ES Energy of Photons After Raman Scattering h Plank’s Constant

H Plate Height

HF Hydrogen Fluoride

HPLC High Performance Liquid Chromatography

I Scattered Light Intensity

K MC Spring Constant k' Retention Factor k’ Partition Coefficient l Length of Micro-cantilever

LOD Limit of Detection

LSPR Localized Surface Plasmon Resonance m* Effective Micro-cantilever Mass mb Mass of the Micro-cantilever

MCs Micro-cantilevers

MIT Mitoxantrone Dihydrochloride n Geometric Parameter

xvi np Number of Photons

PDMS Polydimethylsiloxane

PECVD Plasma-Enhanced Chemical Vapor Deposition

PSD Position Sensitive Detector

PSO Porous Silicon Oxide

PVD Physical Vapor Deposition

R6G Rhodamine 6g

Rf Reference Frequency

Rf Retardation Factor

RIE Reactive Ion Etching

RSD Relative Standard Deviation

SEM Scanning Electron Microscope

SERS Surface Enhanced Raman Spectroscopy t Thickness of Micro-cantilever t Time

TLC Thin-Layer Chromatography tm Time in Mobile Phase tr Elution Time

UTLC Ultra-Thin Layer Chromatography

UV Ultraviolet v Average Linear Velocity vi Frequency of Incident Beam w Width of Micro-cantilever

xvii zmax Bending of Micro-cantilever Tip

α Polarizability of Bond

γLV Interfacial Tension Between Liquid and Vapor

γSL Interfacial Tension Between Solid and Liquid

γSV Interfacial Tension Between Solid and Vapor

ΔEv Difference in Energy of Vibration

Δσ Analyte-Induced Differential Surface Stress

ρ Mass Density

σ One-Fouth Apparent Spot Size

υ Frequency of Light

xviii

Chapter 1

Overview of Surface Enhanced Raman Spectroscopy (SERS)

1.1 Introduction

As an information rich technique, Raman spectroscopy exhibits many analytical advantages, although sensitivity is not among them. However, the small cross sections of the

Raman process are routinely overcome via enormous EM fields located near properly designed plasmonic surface enhanced Raman spectroscopy (SERS) substrates. Unfortunately, the locations of these high field regions on the substrates are sparse and extremely small. Substrate development has been a major focus of SERS research for decades with field enhancements sufficient in the best cases to permit single molecule spectral acquisition.1 An area of SERS research that does not generally receive enough consideration is the development of approaches to direct general classes of polarizable analytes to the substrate and then modifying the region around the plasmonic surface to generate attractive sticking points. A major portion of this dissertation involves the simplified delivery and detection of analytes using substrates that are lithographically engineered pillars, which can be surface modified and/or mated with microfluidic and droplet analyte delivery systems, with enhancement from silver colloid.

1.2 Raman Spectroscopy

Raman scattering was first observed in 1928 by C.V. Raman,1 an Indian physicist who discovered that when light interacts with a molecule the light can transfer a small amount of energy to the molecule, resulting in changes in the color of the scattered light due to energy transmitted into molecular vibrations. These vibrations could then act as a fingerprint of the molecule under observation. This discovery has led to the development of Raman spectroscopy, which is a vibrational spectroscopy technique involving photons impinging on a molecule’s electron cloud resulting in polarization of the molecule. A dipole moment (µ) is generated by the interaction between the oscillating field and the electron distribution of the sample, in which the

2 incident electromagnetic field is directly proportional to the frequency of the light as described by:

휇 = 훼퐸 = 훼퐸0 cos(2휋푣푖푡) 1.2.1 in which 훼 is the polarizability, E is the magnitude of the electromagnetic field that surrounds the analyte, E0 is the peak amplitude of the electromagnetic wave, vi is the frequency of the incident beam, and t is the time.

The polarization causes an excitation of the molecule from a ground state to a virtual energy state. The relaxation of the molecule from the virtual energy state to the ground state is referred to as Rayleigh scattering (elastic scattering). In the case of Raman scattering (inelastic scattering), there is a change in the frequency of the light from the incident radiation. If the photons are scattered to a vibrational frequency that is lower than the incident radiation, then they are referred to as Stokes bands, and if the vibrational frequency is higher than the incident radiation, then they are referred to as anti-Stokes bands,2 as seen in Figure 1.2.1. An energy transfer between the incident photons and the irradiated molecules is induced by the inelastic collisions that occur in Raman scattering when exposed to an electromagnetic field. The energy of the photons after the Raman scattering (Es) occurs can be calculated using:

퐸푠 = ℎ푣 ± ∆퐸푣 1.2.2 in which h is Plank’s constant, v is the frequency of the light, and ΔEv is the difference in the energy of the vibration.

Raman spectrometers are made of three key components, a radiation source, a sampling apparatus, and a detector. For the purposes of this dissertation, the following will briefly describe the setup for the confocal Raman microscope used in this work. The Raman instrument used was a Jobin Yvon Labram Raman Spectrometer comprised of a 633 nm laser. The optical path of the

Figure 1.2.1: Summary of the changes in photon energy for different situations, including infrared absorption, Rayleigh scattering, Stokes Raman scattering, and Anti-Stokes Raman scattering.

4 laser through the microscope and to the CCD detector is summarized in Figure 1.2.2. The digitized signal was then transferred to software on a connected computer for spectral analysis of the acquired signal.

Raman spectroscopy has several advantages including little to no sample preparation, it is non-destructive, and provides unique spectra; however, one of the major disadvantages is the small cross-section of the scattering process. Raman spectroscopy has typical cross sections in the order of 10-29 to 10-31 cm2/molecule which is about 12 to 14 orders of magnitude less than typical fluorescence cross sections. Because of these small cross sections, techniques to enhance the signal of the Raman scattering became necessary for widespread utilization of the technique.

1.3 Discovery and Development of SERS

In the early 1970s, the observation of a monolayer of molecules that had been adsorbed on a solid surface via vibrational spectra was highly sought after.3 However, the idea of using

Raman spectroscopy as a means of detection was considered not feasible due to the weak signal and small cross section characteristic of Raman spectroscopy.1,2 In 1974, Fleischmann et al. was attempting to observe a monolayer of pyridine that was adsorbed from aqueous solution onto a silver electrode, which was roughened by successive oxidation/reduction cycles. The increased

Raman signal observed was credited to the increased surface area due to the roughening of the electrode, and became the first reported observation of SERS. 2,4 However, later experiments found that increased roughness of the electrode lead to a decrease in the signal, so a new explanation was needed. Jeanmaire and Van Duyne were the first to report that the Raman signal enhancement comes from a property of either the electrode surface or the electrode/solution interface.5 Moskovits reported in 1985 that the increase in Raman cross section was due to surface-plasmon excitation from the metallic surface.6 Currently, SERS is an active field of

Figure 1.2.2: Graphical summary of the optics and detection scheme of the Jobin Yvon Labram

Raman Spectrometer used in this work.

6 research in which many different parameters are being adjusted to further enhance Raman signals.

1.4 Enhancement

One of the greatest advantages of SERS is the ability to increase the intensity of the

Raman signal by several orders of magnitude, with typical good enhancements in the range of

106 to 108 with some variance depending on the type of metal substrate being used. SERS causes an increase in the effective Raman cross section to as high as 10-16 cm2/molecule that result in sensitive Raman spectra that sometimes result in additional peaks in the information rich fingerprint region, which cannot be seen with conventional Raman.2

There are two widely accepted theories that contribute to the explanation of the signal enhancement seen in SERS, but the central contributor is the electromagnetic effect.7 The electromagnetic effect is a result of the laser radiation excitation of localized surface plasmons on the roughened metal substrate. Surface plasmons are an oscillation of the free electron density with respect to the fixed ions in a metal.2,8 This oscillation is referred to as the localized surface plasmon resonance (LSPR), which results in an enhancement of the Raman scattering (Figure

1.4.1).9 An enhancement factor of |E|4 can be calculated as a result of the local electric-field enhancement factor at the incident frequency and the corresponding factor at the Stokes-shifted frequency.10 Although the electromagnetic effect plays a central role in the enhancement of the

Raman spectrum, there are other factors involved, such as laser excitation, detection setup, intrinsic properties of the analyte, and analyte adsorption properties.2

The chemical effect in SERS is a smaller contributor to the signal enhancement of

Raman, but it does play a role for molecules that are adsorbed on the metal surface. The enhancement involves a modification of the electronic polarizability of the molecule, which

Figure 1.4.1: Example of the plasmon resonance that occurs in a metal nanoparticle after irradiation by a light source, such as a laser.

8 induces resonant Raman scattering at wavelengths that non-adsorbed molecules would not be resonant.2 The most widely accepted explanation for the modification is the charge-transfer mechanism. The mechanism involves the idea that the Fermi level, or hypothetical energy level of an electron, of the metal is between the molecular ground state and one or more excited state of the molecule, which allows for the movement of electrons between the metal and the molecule at laser excitation strengths typically used in Raman spectroscopy.11,12

1.5 SERS Substrates

A suitable SERS substrate is any surface that has a plasmon-resonance-supporting structure, which is typically found in the gaps, crevices, or sharp features of plasmonic materials

(such as silver, gold, or copper). The surfaces benefit from having little sample preparation, and are homogenous, reproducible, stable, and inexpensive.13 Early SERS experimentation involved using roughened electrodes and other plasmonic surfaces that were highly disordered and resulted in non-reproducible results. Thus, a movement towards ordered surfaces began in order to increase the reproducibility. Although there are many SERS substrates that have been produced, for brevity I choose to focus on substrates relevant to my work and the Sepaniak group, which can be divided into two main categories, random and engineered.

1.5.1 Random SERS Substrates

Colloidal silver consist of aggregated particles of Ag, averaging in size from 25-500 nm in various geometries.14 The benefits of using silver colloid include simple preparation, good stability, ease of characterization of size and shape, and a modest ability to control size and shape. The ability to control/exploit pH, solution ionic strength, and the particle surface chemistry allows for some control of the size, shape, and degree of agglomeration.15

Additionally, silver colloids generate LSPRs that can vary based on shape, size and

9 aggregation.16 Within the nanogaps and junctions between nanoparticles, regions of increased

SERS signal, referred to as “hot spots,” will occur, which can result in enhancement factors larger than 109.17-18 This presents the possibility for detection of analytes at very low concentration, even as low as single molecule detection.19

There are many different methods that have been used to generate Ag colloidal solutions, which include the use of a reductant such as citrate, ethylenediaminetetraactic acid (EDTA), dye

20 molecules, or NaBH4. Also, it has been demonstrated that Ag colloid can be generated by photochemical means via gamma irradiation21, and laser ablation of bulk Ag surfaces.22 The method of preparation of Ag colloid used in my work involves the reduction of AgNO3 with sodium citrate via the method described by Lee and Meisel.23

Silver island films have previously shown modest reproducibility and stability when applied to a surface via physical vapor deposition onto materials such as glass, Teflon, polystyrene, and latex.24-27 Previously, the Sepaniak group has demonstrated the advantage of using silver island films embedded within polydimethylsiloxane (PDMS), which includes an increased surface area caused by the encapsulated nature of the silver within the polymer.28

Another advantage to using Ag-PDMS systems includes a higher resistance to degradation due to temperature, radiation, and high-voltage ionization.29 However, there is not much control over the size of the Ag particles that adsorb onto the surface. Also, there is some variability in the method including the temperature of the substrate during the deposition of silver, the thickness of the silver film, and the use of annealing procedures.25 Silver island films will be used for comparisons in this work.

1.5.2 Engineered SERS Substrates

Different types of engineered SERS substrates have been made in recent years. For example, work completed by Van Duyne showed the utility of silver films-over nanosphere

(AgFON) substrates (Figure 1.5.2.1).30 The Sepaniak group has previously made pillar nanostructures using electron beam lithography (EBL), where densely packed silicon arrays of pillars varying in shape, arrangement, and spacing were fabricated and metalized via physical vapor deposition. This resulted in a pillar array with a discs on top of the pillars, referred to as discs-on-pillars (DOPs).31 An average enhancement factor value was found to be 4 x 109 at optimum height (175 nm), diameter (100 nm), and at the optimum thickness of the Ag films (20 nm).32 These dimensions were optimized for laser excitation at 633 nm. With the optimized pillar dimensions determined, other ways of coating the surface with Ag have been explored.

Another method used by the Sepaniak group is the use of photolithography to create pillar area structures that are not plasmonic by nature, but can incorporate a SERS substrate onto the engineered surface. Photolithography is more advantageous than EBL because it is able to distinguish features with adequate resolution, produce patterns reliably without defects, and allows for high throughput.33 However, it is limited in the sizes of the features to greater than the wavelength of the radiation used.

1.6 Conclusion

SERS is the primary source of chemical identification and quantitation in this dissertation. The enhancement will come from the use of homemade silver colloid nanospheres with approximate sizes around 80 nm. A description of SERS was necessary to understand the utility and sensitivity of the detection technique. The studies described in Chapter 4 and Chapter

Figure 1.5.2.1: (a) SEM image of the silver films-over nanosphere substrates developed by the

Van Duyne group. (b) SEM image of silver colloid deposition within a photolithographic pillar array.

5 use SERS in conjunction with fabricated pillar arrays, in which the fabrication process is detailed in Chapter 2.

1.7 References

1. Haynes, C.L.; Yonzon, C.R.; Zhang, X.; VanDuyne, R.P. J. Raman Spectrosc., 2005, 36, 471-484. 2. Le Ru, E.C.; Etchegoin, P.G. Principles of Surface-Enhanced Raman Spectroscopy and Related Plasmonic Effects. 2008. 3. McQuillan, A. J. Notes Rec. R. Soc., 2009, 63, 105-109. 4. Fleischmann, M.; Hendra, P.J.; McQuillan, A.J. Chem. Phys. Lett., 1974, 26, 163-166. 5. Jeanmaire, D.L.; Van Duyne, R.P. J. Electroanal. Chem., 1977, 84, 1-20. 6. Moskovits, M. Rev. Modern Phys., 1985, 57, 783-826. 7. Ansar, S.M.; Li, X.; Zou, S.; Zhang, D. J. Phys. Chem. Lett., 2012, 3, 560-565. 8. Pines, D. Rev. Mod. Phys., 1956, 28, 184-199. 9. Rojas, R.; Claro, F. Appl. Opt., 1993, 98, 998-1006. 10. Schatz, G.C.; Young, M.A.; Van Duyne, R.P. Surface-Enhanced Raman Scattering: Physics and Applications, 2006, 19-45. 11. Lombardi, J.R.; Birke, R.L.; Lu, T.; Xu, J. J. Chem. Phys., 1986, 84, 4174-4180. 12. Saikin, S.K.; Olivares-Amaya, R.; Rappoport, D.; Stope, M.; Aspuru-Guzid, A. Phys. Chem. Chem. Phys., 2009, 11, 9401-9411. 13. Sharma, B.; Frontiera, R.R.; Henry, A.; Ringe, E.; Van Duyne, R.P. Mater. Today, 2012, 15, 16-25. 14. Sheng, R.; Zhu, L.; Morris, M.D. Anal. Chem., 1986, 58, 1118-1119. 15. Jiang, J.; Oberdӧrster, G. J. Nanopart. Res., 2009, 11, 77-89. 16. Abu-Hatab, N.A.; John, J.F.; Oran, J.M.; Sepaniak, M.J. Appl. Spectrosc., 2007, 61, 1116-1122. 17. Liu, H. et al. Sci. Rep., 2011, 112, 1-5. 18. Tolaieb, B.; Constantino, C.J.L.; Aroca, R.F. Analyst, 2004, 129, 337-341. 19. Lee, C.L.; Kim. H.J.; Karim, M.R.; Lee, M.S. Bull. Korean Chem. Soc., 2006, 27, 545- 548. 20. Bright, R.M.; Musick, M.D.; Natan, M.J. Langmuir, 1998, 14, 5695-5701. 21. Henglein, A. J. Phys. Chem., 1980, 84, 2461-3467. 22. Lee, I.; Han, S.W.; Kim, K. J. Raman Spectrosc., 2001, 32, 947-952. 23. Lee, P.C.; Meisel, D. J. Phys. Chem. 1982, 86, 3391-3395. 24. Alak, A.M.; Vo-Dinh, T. Anal. Chem., 1989, 61, 656-660. 25. Ni, F.; Cotton, T.M. Anal. Chem., 1986, 58, 3159-3163. 26. Vo-Dinh, T.; Hiromoto, M.Y.K.; Begun, G.M.; Moody, R.L. Anal. Chem., 1984, 56, 1867-1870. 27. Enlow, P. D.; Buncick, M.; Warmack, R.J.; Vo-Dinh, T. Anal. Chem., 1986, 58, 1119- 1123. 28. De Jesủs, M.A.; Giesfeldt, K.S.; Sepaniak, M.J. J. Raman Spectrosc., 2004, 35, 895-904. 29. Giesfeldt, K.S. et al. Appl. Spectrosc., 2003, 57, 1346-1352. 30. Zhang, X.; Zhao, J.; Whitney, A. V.; Elam, J. W.; Van Duyne, R.P. J. Am. Chem. Soc. 2006, 128, 10304−10309. 31. Polemi, A.; Wells, S.M.; Lavrik, N.V.; Sepaniak, M.J.; Shuford, K.L. J. Phys. Chem., 2010, 114, 18096-18102. 32. Wells, S.M.; Polemi, A.; Lavrik, N.V.; Shuford, K.L.; Sepaniak, M.J. Chem. Commun.., 2011, 47, 3814-3816.

33. Taylor, L.C. The fabrication and integration of pillar array channels for chip based separations and analysis, 2011, University of Tennesee, Knoxville. Dissertation.

Chapter 2

Fabrication and Use of Pillar Arrays for Planar Chromatography with SERS Detection

2.1 Overview of Chromatography

Chromatography is a method of separating a mixture of compounds for analysis, in which each component in the mixture has a different interaction with the separation interface.1-2 The first reported study involving chromatography was completed by the botanist Tswett in 1903, and it involved the separation of plant pigments with the use of a hydrocarbon solvent and inulin powder.3-4 Tswett’s invention of adsorption chromatography was not furthered until 1931 when several scientists began performing biochemical separations, resulting in adsorption chromatography becoming a staple in biochemistry.3 The fundamental premise of chromatography is the separation of compounds with the strategic use of the stationary and mobile phases. The stationary phase is either a viscous liquid or porous solid layer that is chemically bonded within a capillary tube or on top of a surface and serves as the active layer which directly interacts with the analytes within the mobile phase. The liquid or gas that transports the components of the mixture across or through the stationary phase is referred to as the mobile phase.3,5

A chromatogram is a graphical representation of the detector response to the elution time

(tr), or time it takes for each component in the mixture to be separated and detected. One of the most important figures of merit in a separation is the retardation factor (Rf), as seen in Equation

2.1.1.

1 푅 = 2.1.1 푓 1+푘′

Rf is equal to the inverse of the retention factor (k’) plus one, in which the retention factor is the distance traveled by the analyte from the original spot divided by the distance traveled by the mobile phase from the original spot. Retention factors are used for direct comparison of

17 components in a mixture and between separate mixtures to determine reproducibility in chromatography.

Another important metric to consider in a chromatogram is resolution (Rs), which evaluates the degree of separation of analytes. An ideal separation would have no overlapping analyte bands post-separation. To quantify Rs, Equation 2.1.2 can be used.

2푑 푅푠 = 2.1.2 푤푎+푤푏

Each term is as follows: d is the distance between the centers of two separated analytes, wa and wb are the widths of each band being compared.

There are many different types of chromatography, including adsorption, partition, ion- exchange, molecular exclusion, and affinity chromatography.3 With the variety of chromatographic methods, there are many different applications that can be covered. For the purposes of this dissertation, the focus will be centered around thin-layer chromatography (TLC), an adsorption liquid chromatographic method.

2.2 TLC History and Concept

Izmailov and Schraiber were the first to use TLC in 1938 by separating medicinal compounds on unbound alumina on glass plates.6 By the 1950s, Kirchner and others at the U.S.

Department of Agriculture used TLC the same way as it is used in modern times. The TLC plates were composed of silica gel on top of glass plates and developed via vertical ascension of the mobile phase via capillary action.6

TLC is a common method of separation today for more precise and accurate separations of compounds of interest. TLC plates are made of either glass or plastic with thin layers coated

18 on top to serve as the stationary phase. The mobile phase consists of a binary mixture to further enhance the separation of components across the stationary phase. There are several advantages to using TLC including fast separations, small sample sizes, lower costs, higher sample throughput (due to multiplexing), and high sensitivity.4,7 There are some similarities and differences between TLC and high performance liquid chromatography (HPLC), another common method of liquid chromatography. The similarities include the observed retention factors in both techniques, along with the optimum performance of each technique requiring a Rf of 0.25. HPLC is more advantageous in situations when automation is required, along with systems requiring high mobile phase velocities.

A typical TLC experiment is performed as follows and is seen in Figure 2.2.1. A small sample is deposited on one end of the stationary phase of the TLC plate to form the initial spot.

The TLC plate is then lowered into a closed chamber in which a mobile phase generally containing at a mobile phase mixture at the bottom. The mobile phase is then allowed to develop up the TLC plate, and if the mobile and stationary phases were chosen correctly, a separation of the components in the initial spot will occur.

2.3 Van Deemter Equation

Van Deemter and others developed an equation, including the four major sources of band broadening along with velocity, to explain the kinetics and band broadening seen in chromatographic systems.8 Band broadening is not desirable in chromatographic systems and must be minimized. A metric that can be used to evaluate the amount of band broadening present in a liquid chromatographic system is plate height (H), in which H must be minimized as much as possible. Equation 2.3.1 represents H as a simplified equation.

Figure 2.2.1: Diagram of a TLC experiment depicting a separation of a three component mixture.

퐵 퐻 = 퐴 + + (퐶 + 퐶 )푣 2.3.1 푣 푆 푀

Each term is as follows: H is the plate height, A is eddy diffusion, B is longitudinal molecular diffusion, CS is resistance to mass transfer in the stationary phase, CM is the resistance to mass transfer in the mobile phase, and v is the average linear velocity. The plate height assesses the total band broadening contributions from each term as a function of the average linear velocity.

To maximize the efficiency of the chromatographic system, the A, B, CS, and CM terms must be minimized as much as possible.5

The following is a more in depth analysis of the van Deemter equation that further defines the roles each term plays in band broadening and what can be done to minimize each term. The A term, eddy diffusion, evaluates the random path an analyte travels through a heterogeneous stationary phase, and the term can be minimized by using small, uniform particles that can be tightly packed together. The B term involve molecular diffusion, which describes the behavior of molecules as they diffuse from a region of high concentration to a region of low concentration over time. To minimize the B term, the average linear velocity should be increased via increases in the velocity. The CS term describes the speed at which solute sorption and desorption occurs. Rapid equilibration results in less band broadening, thus the CS term is

5 minimized by using small film thicknesses. The CM term describes the variability in the velocity of the mobile phase as it interacts with an analyte through the chromatographic system. The typical profile of a mobile phase is parabolic, thus resulting in faster mobile phase velocities in the center of flow versus slower velocities near the stationary phase. The CM term can be minimized by decreasing the particle size or internal diameter of the system if column chromatography is used, such that the differences in velocities of the mobile phase are mitigated.

A mathematical summary of the preceding can be seen in equation 2.3.2, which expands the terms of the van Deemter equation.

2 ′푑푓 2 2훾퐷푀 푞푘 휔푑푝 퐻 = 2휆푑푝 + + ( ′ 2 + ) 푣 2.3.2 푣 (1+푘 ) 퐷푆 퐷푀

’ Each term is as follows: dp is the particle diameter, k is the partition coefficient, df is the average film thickness of the stationary phase, Ds is the diffusion coefficient for the stationary phase, DM is the diffusion coefficient for the mobile phase, and the q, λ,γ, and ω terms are independent factors conditional to the packing or ordering of the stationary phase and other parameters.5

A visualization of the van Deemter equation can be seen in Figure 2.3.1, and is referred to as a van Deemter plot. The plate height (H) vs. v is plotted such that the minimum of the curve indicates the optimum velocity for the highest efficiency, or smallest plate height. The plot of the

A, B, CS, and CM terms can also be seen in Figure 2.3.1 in order to visualize their contributions to plate height. Evaluation of a van Deemter plot provides the opportunity to examine the chromatographic system in order to determine appropriate running times through adjustment of the average linear velocity.

The preceding treatment of plate height using the van Deemter equation is based off the idealized chromatographic system, however, this dissertation uses a chromatographic system that is open and comprised of fabricated pillar substrates. Despite this, plate height can still be evaluated, but there are some assumptions that must be made to apply this theory to the pillar array platforms. First, the A term can be excluded from the description if the pillars are in perfect

9 ’ 9 order. The CS term can also be excluded by assuming there is an unretained solute (k = 0).

Further assumptions and treatments by the Sepaniak group of the van Deemter equation has led to a simplified calculation for the relevant plate height and is seen in Equation 2.3.3:

2(0.5)퐷 0.02푑 2푣 퐻 = 푀 + 푝 2.3.3 푣 퐷푀

where H is the relevant plate height, DM is the typical diffusion coefficient, v is the velocity of

9-11 the mobile phase, and dp is the particle diameter. These assumptions allow for an approximation of the plate height in the pillar array platforms such that analysis of separations can be conducted.

2.4 Deterministic Pillar Array Fabrication

From this point forward, the work will involve the use of micro- and nano-fabricated pillar arrays to produce ultra-thin layer chromatography platforms. All of the fabrication for these pillar arrays was completed at the Center for Nanophase Materials Science at Oak Ridge

National Laboratory, and the process for the manufacture of these arrays has been established by the Sepaniak group in conjunction with Nickolay Lavrik.9-10 The first of these pillar array platforms is the deterministic pillar arrays, which consist of identical micro-scale pillars in a uniform array formed through photolithography.

Photolithography is a process in which photons are used to transfer a pattern from a photomask to a light-sensitive “photoresist” on a substrate. This process is the most commonly used method of lithography, especially in the manufacture of nanoelectronics. There are two types of photoresist that can be employed in photolithography. If a resist is exposed to UV light and the resulting interaction creates a chemical reaction that weakens the photoresist, it is referred to as a positive photoresist. If the UV light creates a situation in which the photoresist becomes crosslinked or polymerized, creating a more resistant area, it is referred to as a negative photoresist.12

The deterministic pillar array fabrication process begins with a silicon wafer being coated with

100 nm of silicon dioxide. Two layers of positive photoresist are then spun and deposited on top of the wafer. A mask is placed over the wafer and UV light is exposed to the photoresists such that the pillar array design is written into the resists by weakening areas exposed to the radiation.

A series of chemical washes then remove the weakened sections of the photoresist. A physical vapor deposition (PVD) system is then used to deposit chromium on the wafer to serve as a mask for the newly defined pillar array. A PVD system involves the transfer of a condensed substance

(solid or liquid) to a solid phase in the form of a thin film via vaporization of the condensed substance.13 A series of chemical washes then remove the excess photoresists, leaving only the chromium that was deposited in holes formed by the UV light. The wafer is then placed in a reactive ion etcher (RIE) to undergo a Bosch recipe, which results in scalloped pillar sidewalls which increases surface area and improves pillar stability. The Bosch recipe involves a very fast repetition of etching and passivation steps. The etching step involves the use of isotropic SF6 gas, and the passivation step involves the deposition of C4F8 polymer over the entire wafer surface.

These steps are repeated until the desired height of the pillars is reached (~20 µm). As seen in

Chapters 4 and 5, the pillars are then coated with a layer of porous silicon oxide (PSO) at room temperature via a plasma-enhanced chemical vapor deposition (PECVD) system. This layer of

PSO increased the hydrophobicity of the pillar array such that the surface, and after functionalizing with a C18 phase, the surface became superhydrophobic. This behavior was exploited in the studies summarized in Chapter 4 and Chapter 5. Figure 2.4.1 summarizes the manufacture of deterministic pillar arrays. This process yielded pillars with the following dimensions: 20 µm in height, 1-2 µm in diameter, and a gap size of 1-2 µm.

2.5 Stochastic Pillar Array Fabrication

Stochastic pillar arrays are a randomized array of pillars that begin with a dewetting procedure, which was developed at the Center for Nanophase Materials Science (CNMS).14-15 Stochastic pillar array fabrication is desirable in that it is lower in cost and takes significantly less time to fabricate. As before, the process begins with a silicon wafer with a 100 nm thick layer of silicon oxide. A thin layer of platinum is deposited on the wafer via PVD, with the thickness roughly defining the width of the subsequent pillars. The wafer is then exposed to a rapid thermal processor at maximim power for 8 seconds, heating the wafer to about 900°C. The result is the formation of platinum islands throughout the wafer surface, which act like a hard mask for the etching procedure. The same etching process occurs as described in Section 2.4, in which a

Bosch recipe is used in the RIE. The resulting dimensions of the stochastic pillars are approximate due to the randomized nature of the dewetting procedure. Typical diameters are around 200 nm with heights around 1-2 µm. Figure 2.5.1 shows an overview of the fabrication process.

2.6 Superhydrophobicity

When a droplet of water is placed on a surface, a contact angle at which the droplet rests on the surface can be measured. The contact angle for a smooth surface can be described by the Young equation (Equation 2.6.1).

(훾 −훾 ) cos(휃) = 푆푉 푆퐿 2.6.1 훾퐿푉

The γSV is the interfacial tension between the solid and vapor, the γSL is the interfacial tension

16 between the solid and liquid, and the γLV is the interfacial tension between the liquid and vapor.

Figure 2.5.1: (a-d) Images depicting the different steps involved in the fabrication of the stochastic pillar arrays. (e) A SEM image of an actual array of the stochastic pillars is included.

If the contact angle of the water droplet is less than 90˚, the surface is considered hydrophilic. A contact angle between 90˚ and 150˚ indicates a surface that is hydrophobic. If the contact angle is greater than 150˚, the surface is considered superhydrophobic.16 However, the Young equation cannot fully explain the contact angle of a droplet due to most surfaces not being completely smooth. For example, a hydrophobic surface can be altered to become a superhydrophobic surface by either roughening the surface or generating a particular morphology on the surface.

Heterogeneous wetting refers to the droplet coming into contact with only the top of the roughened surface, such as the tops of pillars, which leaves air bubbles underneath the water droplet, as seen in Figure 2.6.1. When this model is considered, the Cassie-Baxter equation can be used to determine the contact angle of the water droplet (Equation 2.6.2), where rf is the ratio of the actual wetted area to the projected area and ∅푆 is the ratio of the total area of the solid- liquid interface with respect to the total area of solid-liquid and liquid-air interfaces in a plane geometrical area of unity parallel to the rough surface.17

∗ 푐표푠휃퐶퐵 = 푟푓 ∙ ∅푆 ∙ 푐표푠휃 + ∅푆 − 1 2.6.2

Wenzel proposed an equation that could relate the contact angle to surface roughness and surface energies (Equation 2.6.3), where r is the roughness area ratio of the actual surface with respect to

∗ 36 the geometric surface and 휃푤 is the apparent Wenzel contact angle .

∗ 푐표푠휃푤 = 푟 cos 휃 2.6.3

The Wenzel equation assumes that the water penetrates the “grooves” in the roughened surface and is considered homogenous wetting, as seen in Figure 2.6.1.

The mechanism behind the transition between the Cassie-Baxter to the Wenzel state is not well understood, but several theories have been proposed to explain the event. Proposed mechanisms of the transition include changes in the net surface energy levels of the two states,

30 droplet curvature, multi-scale roughness of the surface, change in the balance between the droplet weight and the surface tension force, and the history of the system.18 A recent study has indicated that the mechanism of the transition is a combination of several different mechanisms on the macro-, micro-, and nanoscale, which involves the contact angle/droplet radius, the position of the liquid-vapor interface, and the surface heterogeneity, respectively.18

The use of superhydrophobic surfaces has been recently recognized as a viable method for single molecule detection.19 In order to become a superhydrophobic surface, two key components must be addressed: making a rough surface from a low surface energy material and modifying the rough surface with a material of low surface energy.20 There are many examples of superhydrophobic surfaces that occur in nature including the surface of lotus leaves, which consist of 3-10 μm size protrusions and valleys and a hydrophobic wax material on top of the roughened surface.21 The first reported superhydrophobic surfaces that were made in the lab involved the use of alkylketene dimer, which created a heterogeneous fractal structure on the surface that was also hydrophobic.22 However, the need for ordered nanostructure surfaces is vital for reproducibility. One advantage of utilizing superhydrophobic behavior with pillar array systems is the ability to concentrate an analyte within the water droplet to a level that is detectable by the SERS substrate.

2.7 Conclusion

The fabrication of pillar arrays in conjunction with SERS, a desirable method of detection, will be an effective pairing in that SERS is a non-invasive technique that can provide unique spectra, and the geometries of the pillar arrays can be tailored to optimize separations and subsequent detection. Chapter 4 and Chapter 5 summarizes studies that involved the use of superhydrophobic pillar arrays in conjunction with SERS for chemical identification. In addition

31 to these studies, recent work outside of the Sepaniak group has demonstrated the superhydrophobic nature of pillar array systems effectively makes single molecule detection possible with the use of SERS.23 It is important to note that the following studies intention’s are to demonstrate low limits of detection utilizing SERS, and demonstrating an alternative method of analyte detection after a successful separation.

2.8 References

1. Ettre, L.S. LCGC North Am. 2001, 19, 48. 2. Vickers, A.K.; Decker, D.; Majors, R.E. LCGC North Am. 2007, 25, 616. 3. Harris, D.C. Quantitative Chemical Analysis, 9th ed.; Freeman: New York, 2016. 4. Bricker, C.E.; Taylor, M.A.; Kolb, K.E. J. Chem. Ed. 1981, 58, 48. 5. Miller, J.M. Chromatography: Concepts and Constrasts, 2nd ed.; Wiley: New York, 2005. 6. Sherma, J.; Fried, B. Handbook of Thin-Layer Chromatography, 3rd ed.; Marcel Dekker: New York, 2003. 7. Touchstone, J.C. Practice of Thin Layer Chromatography, 3rd ed.; Wiley: New York, 1992. 8. van Deemter, J. J.; Zuiderweg, F. J.; Klinkenberg, A., Chem. Eng. Sci. 1956, 5, 271. 9. Kirchner, T. B., Hatab, N. A., Lavrik, N. V., Sepaniak, M. J., Anal. Chem. 2013, 85, 11802– 11808. 10. Wallace, R.A.; Lavrik, N.V.; Sepaniak, M.J. Electrophoresis. 2017, 38, 361-367. 11. Lincoln, D.R.; Lavrik, N.V.; Kravchenko, I.I.; Sepaniak, M.J. 2016, 88, 8741-8748. 12. Yeh, W.; Noga, D.E.; Lawson, R.A. J. Vac. Sci. Technol. B. 2010, doi: http://dx.doi.org/10.1116/1.3518136 13. Mahan, J.E. Physical Vapor Deposition of Thin Films, 1st ed.; Wiley: New York, 2000. 14. Charlton, J.J.; Lavrik, N.; Bradshaw, J. A.; Sepaniak, M. J., ACS Applied Materials & Interfaces. 2014, 6 (20), 17894-17901. 15. Kirchner, T. B. The fabrication of micro- and nanoscale deterministic and stochastic pillar arrays for planar separations. University of Tennessee, Knoxville, TN, 2015. 16. Celina, E.; Darmanin, T.; de Givency, E.T.; Amigoni, S.; Guittard, F. J. Colloid Interface Sci., 2013, 402, 1-18. 17. Yan, Y.Y.; Gao, N.; Barthlott, W. Adv. Colloid Interface Sci., 2011, 169, 80-105. 18. Nosonovsky, M.; Bhushan, B. Nano Lett., 2007, 7, 2633-2637. 19. Gentile, F. et al. Microelectron. Eng. 2012, . 20. Ma, M.; Hill, R.M. Curr. Opin. Colloid Interface Sci., 2006, 11, 193-202. 21. Guo, Z.; Liu, W.; Su, B. J. Colloid Interface Sci., 2011, 353, 335-355. 22. Onda, T.; Shibuichi, S.; Satoh, N.; Tsujii, K. Langmuir, 1996, 12, 2125-2127. 23. De Angelis, F et.al. N. Photon, 2011, 5, 682-687.

Chapter 3

Overview of Micro-Cantilever (MC) Sensors

3.1 Micro-cantilever (MC) Background

Micro-cantilevers (MCs) have been used extensively as both chemical and biological sensors for decades. The first report of a cantilever mechanical transducer was realized by

Norton in 1943.1 Shaver furthered the work in 1969 by designing a bimetallic hydrogen detector that was based off of a cantilever mechanical transducer. He was able to detect hydrogen gas concentrations around 50 ppm in the presence of nitrogen gas.2 One of the major problems with these early cantilevers was the low accuracy and sensitivity that is required for a useful sensor, which could be attributed to their larger initial size (100 mm long and 125 µm thick).3

Researchers began to realize that cantilevers made on the micro-scale would be a vast improvement in accuracy and sensitivity;3-7 however, the technology was not yet available to fabricate MCs.

With the help of atomic force microscopy (AFM) and new technological advances in microfabrication, MC probes were shown to add additional sensitivity to the apparatus. AFM used the cantilever tip to scan or raster across a surface, with the force between the tip and the surface causing a deflection of the cantilever. The measured deflection was then used to map the topography of the surface.8 Researchers began to notice that a MC’s behavior would change based off of physical changes in the environment, including humidity, temperature, pressure, and acoustic noise.9 These observations sparked the interest of the scientific community and led to the AFM cantilever eventually becoming known as a sensitive analytical tool for quantitative applications.10-13 Typical MCs are tiny plates with dimensions ranging from 0.2-1 µm thick, 100-

500 µm long, and 20-100 µm wide, with one end of the MC being affixed to a solid base. A typical cantilever can be seen in Figure 3.1.1. For AFM applications, a small sharp tip can be

Figure 3.1.1: Image of typical MC defining the thickness, width, and length.

36 added to the bottom of the cantilever for very sensitive mapping of surface topology. However,

MCs responsible for chemical and biological sensing do not typically have sharp tips.

There are several advantages to using MCs as a sensor, including small size, lower cost, real-time analysis, and easy integration with other detection methods, including Raman spectroscopy14 and chromatographic separations.15 In addition, MC platforms are very sensitivity, with reported dynamic ranges spanning three orders of magnitude.16 As stated above, transduction in MCs can be affected by physical, chemical, or biological changes, which can be measured optically and electronically.15 The shapes and geometries of MCs can be varied for any application with the use of photolithography and RIE using either silicon or silicon nitride.14

Figure 3.1.2 shows examples of several different types of MCs in various sizes and geometries.

3.2 MC Measurement Methods – Dynamic Mode

There are two common modes of MC sensing that will be discussed in this dissertation: dynamic mode and static mode. Each mode has its own applications, with the choice being dictated by the transduction mechanism and the medium in which the MC will be used. The dyanamic mode monitors the change in the resonance frequency (Rf) of the MC oscillations in order to monitor mass changes on the MC surface.13 Resonance of an object is the tendency of that object to oscillate at greater amplitudes at certain frequencies. A change in the mass on the surface of an

MC can thus be measured from monitoring the Rf of the system before and after the mass

13-14 change. Equation 3.2.1 shows the dependence of the Rf with respect to mass.

1 푅 = ×√퐾/푚∗ 3.2.1 푓 2휋

The terms are as follows: K is the MC spring constant and m* is the effective MC mass. In order to determine the effective MC mass, Equation 3.2.2 can be used.15

Figure 3.1.2: SEM images of (a) triangular MC made by the Datskos group, (b) traditional rectangular MC purchased from MikroMasch, and (c) another MC geometry made by the

Electronics Institute of Technology.

∗ 푚 = 푛푚푏 3.2.2

The terms are as follows: n is a geometric parameter, in which the value is 0.24 for a rectangular

16 cantilever, and mb is the mass of the micro-cantilever. Another method of evaluating the Rf with relation to mass change and changes in the spring constant is to use Equation 3.2.3.15

1 1 ∆푚 − = 2 3.2.3 푅푓1 푅푓0 4휋 퐾

The terms are as follows: Rf1 is the final resonance frequency, Rf0 is the initial resonance frequency, Δm is the change in mass, and K is the spring constant. For the case of a simple rectangular MC, which are used in the studies in Chapter 6, the Rf can be evaluated using

Equation 3.2.4.17

0.162 퐸푡3푊 푅 = √ 3.2.4 푓 √휌푙2

The terms are as follows: ρ is the mass density, t is the thickness of the MC, l is the length of the

MC, and w is the width of the MC. Using the dynamic mode for MC sensing has been shown to have sensitive applications, including the use of a T-shaped cantilever to detect femtogram changes in mass.18-19 By decreasing the size of the MC, as seen in the Pang et al. study,18 the sensitivity of the dynamic mode can be increased.

3.3 MC Measurement Methods – Static Mode

For the studies presented in this dissertation, the static mode MC measurement method was used. The static mode measures the deflection of the MC tip as a result of bending of the

MC. The tip deflections are caused by either stresses generated on or in the lever or external forces.3 As seen in Figure 3.3.1, a laser spot can be reflected off the tip of the MC and focused on

Figure 3.3.1: Diagram of a typical static mode experiment in which the change in position of the laser on the PSD is shown before and after the MC has bent.

40 a position sensitive detector (PSD) to monitor changes in a MC. The most important aspect to conducting a static mode measurement is to ensure there are asymmetric surfaces on the MC, resulting in an active and passive side. A successful experiment involves the deposition of a coating on the active side of the MC that shows an affinity for the analyte molecule, while the passive side is not responsive to the analyte molecule. The goal is to cause a differential change

(passive vs. active sides) in the surface stress large enough for detection.

There are three models that have been developed to explain how the surface stress will change on a MC’s active side (Figure 3.3.2). The first model involves the analyte interacting directly with a monolayer surface. Analytes can either physisorb on to the surface through van der Waals forces or through chemical bonding to the monolayer. The physisorption of the analytes can polarize the surface, inducing dipoles; however, the energy is small, so there are very little changes in the surface stress as a result. Chemisorption has a higher energy interaction with the surface, which leads to larger changes in surface stress.20-21 If there is an excess in the

Gibb’s free energy on the surface, a spontaneous adsorption processes will occur which could lead to a possible reduction in the surface stress. There are two possible outcomes after adsorption. Either the stress is compressive causing the MC to bend downward, or the MC bends upward, which is referred to as tensile stress.20-21 This means that large static deformations will occur in MCs with large initial surface free energies, or, as stated in Chapter 6, the larger the initial surface stress, the higher the potential signal change.

A second model for surface stress changes on the active side of MCs involves the interaction of analytes with thicker than monolayer coatings. The analyte can permeate the thicker coating, which can alter several different forces within the coating, including osmotic, dispersion, steric, and electrostatic forces.23 This leads to surface stress changes in the cantilever,

42 such as swelling.

A third model can be used to explain the changes in surface stress with an active side consisting of a nanostructured surface. These types of coatings are desirable in that they cause high initial surface stress due to the disorder of the coating.24 The binding of the analyte to the surface causes a large change in stress and larger MC deformation. Studies completed by the

Sepaniak group have shown the enhancement of MC response by two orders of magnitude in which the MC had a nanostructured surface.25-26 In addition, MCs with this type of stress have been shown to have improved reversibility.16

3.4 MC Readout Methods

In the case of dynamic mode measurements, a typical experiment involves the placement of the MC array on a stable surface in which a laser spot is focused on one individual MC at a time. The laser spot is reflected on to a PSD and an oscilloscope in conjunction with a lock-in amplifier is used to measure the Rf of each MC. A speaker then emits a range of frequencies in order to isolate the Rf. The change in the Rf cannot be measured easily over time, so usually the measurements are taken before and after an analyte has been exposed to the MC surface. The Rf of each MC can act as an identifier as it unique to each MC.

A typical static mode measurement uses the optical beam deflection method in which a laser spot is focused on an individual MC and reflected onto a PSD. Instead of monitoring the Rf, the bending of the MC can be monitored in real-time by tracking the movement of the laser spot across the PSD. MC tip displacements as low as 10-14 m have been reported using this optical arrangement,27 with routine measurements capable of detecting 10-10 reliably.21 This readout method has several advantages including a linear response, simplicity, and reliability; however,

43 this method is susceptible to interferences to the medium around the MCs, so measurements taken in vacuum are ideal.

3.5 Conclusion

MCs are a valuable chemical and biological sensor that can be employed in many different applications. For the purposes of this dissertation, the focus will be on HF gas sensing with PSO acting as the active layer. The MCs used will be an array commercially manufactured to ensure stable and consistent MCs. Chapter 6 will summarize a series of studies utilizing the static bending method of measurement to monitor real-time changes in the bending and surface stress of the MCs after exposure to the etching nature of the HF gas.

3.6 References

1. F.J. Norton, Gas analyzing and control apparatus, (General Electric Co.). US, 1943. 2. Shaver, P.J. Rev. Sci. Intrum.1969, 40, 901. 3. Lavrik, N.V.; Sepaniak, M.J.; Datskos, P.G. Rev. Sci. Instrum. 2004, 575A. 4. Patton, J.F.; Lavrik, N.V.; Joy, D.C., Datskos, P.G.; Smith, D.B.; Hunter, S.R.; Sepaniak, M.J. Nanotechnology. 2012, 23, doi:10.1088/0957-4484/23/46/465403. 5. Chapman, P.J.; Vogt ,F.; Dutta, P.; Datskos, P.G. ; Devault, G.L.; Sepaniak, M.J. Anal. Chem. 2007, 79, 364-370. 6. Patton, J. F.; Sepaniak, M. J.; Smith, D. B.; Datskos, P.G.; Hunter, S.R.; Sensors and Actuators A. 2010, 163, 464-470. 7. Patton, J. F.; Sepaniak, M. J.; Smith, D. B.; Datskos, P.G.; Hunter, S.R.; Sensors and Actuators A, 163, (2010),464-470 8. Alexander, S.; Hellemans, L.; Marti, O.; Schneir, J.; Elings, V.; Hansma, P. K.; et al J. Appl. Phys. 1989, 65, 1, 164-167. 9. Binnig, G.; Quate, C. F.; Gerber, C. Phys. Rev. Lett. 1986, 56, 930-933. 10. Thundat, T; Warmack, R. J.; Chen, G.Y.; Allison, D. P. Appl. Phys. Lett. 1994, 64, 2894- 2903. 11. Thundat,T.; Wachter, S.L.; Sharp, S.L.; and Warmack, R.J.; Appl. Phys. Lett. 1995, 66, 1695. 12. Barnes, J. R.; Stephenson, R. J.; Welland, M. E.; Gerber, C.; Gimzewski, J. K. Nature 1994, 372, 79 13. Itoh, T.; Suga,T.; Applied Physics Letters, 1994, 64(1), 37-39. 14. Wang, Z.; Miao, J.; Tan, C.W.; Xu, T. J. Electroceram. 2010, 24, 25-32. 15. Chapman, P.J.; Vogt, F.; Dutta, P.; Datskos, P.G.; Devault, G.L.; Sepaniak, M.J. Anal. Chem. 2007, 79, 364-370. 16. Sepaniak, M.J.; Datskos, P.G.; Lavrik, N.V.; Tipple, C. Anal. Chem. 2002, 568A-575A. 17. Vahist, S.K. J. Mater. 2007, doi:10.2240/azojono0115. 18. Johnson, B.N.; Mutharasan, R. Biosensors and Bioelectronics. 2012, 32, 1-18. 19. Pang, W.; Yan, L.; Zhang, H.; Yu, H.Y.; Kim, E.S.; Tang, W.C. Appl. Phys. Lett. 2006, 88, 243503. 20. Datskos, P.G.; Lavrik, N.V.; Sepaniak, M.J. Chemical and Biological Sensors Based on Microcantilevers. In Smart Sensors and MEMS. Yurish, S.Y.; Gomes, T.S.R., Ed.; Kluwer Academic Publishers: Netherlands, 2004, 331. 21. Chapman, P.J. “Approaches to Generating Selectivity in Microcantilever Sensors.” PhD diss., University of Tennessee, 2008. 22. Lakes, R. Science. 1987, 235, 1038. 23. Israelachvili, J. Intramolecular and Surface Forces, Academic Press, San Diego, CA, 1991. 24. R. Koch, Journal of Physics: Condensed Matter, 6 (1994) 9519. 25. Lavrik, N.V.; Tipple, C.A.; Sepaniak, M.J.; Datskos, P.G. Chem. Phys. Lett. 2001, 336, 371. 26. Tipple, C.A.; Lavrik, N.V.; Culha, M.; Headrick, J.; Datskos, P.G.; Sepaniak, M.J. Anal. Chem. 2002, 74, 3118. 27. Meyer, G.; Amer, N.M. Appl. Phys. Lett. 1990, 57, 2089.

Chapter 4

Superhydrophobic Analyte Concentration Utilizing Colloid-Pillar Array SERS Substrates

The research presented in Chapter 4 has been adapted from a research article published in

Analytical Chemistry (Wallace, R.A.; Charlton, J.J.; Kirchner, T.B.; Lavrik, N.V.; Datskos, P.G.;

Sepaniak, M.J. Superhydrophobic Analyte Concentration Utilizing Colloid-Pillar Array SERS

Substrates. Anal.Chem., 2014, 86, 23, 11819-11825.) This chapter focuses on using a superhydrophobic surface as method of pre-concentration, followed by SERS detection through the use of silver colloid.

4.1 Abstract

The ability to detect a few molecules present in a large sample is of great interest for the detection of trace components in both medicinal and environmental samples. Surface enhanced

Raman spectroscopy (SERS) is a technique that can be utilized to detect molecules at very low absolute numbers. However, detection at trace concentration levels in real samples requires properly designed delivery and detection systems. The following work involves superhydrophobic surfaces that have as a framework deterministic or stochastic silicon pillar arrays formed by lithographic or metal dewetting protocols, respectively. In order to generate the necessary plasmonic substrate for SERS detection, simple and flow stable Ag colloid was added to the functionalized pillar array system via soaking. Native pillars and pillars with hydrophobic modification are used. The pillars provide a means to concentrate analyte via superhydrophobic droplet evaporation effects. A ≥ 100-fold concentration of analyte was estimated, with a limit of detection of 2.9 × 10–12 M for mitoxantrone dihydrochloride. Additionally, analytes were delivered to the surface via a multiplex approach in order to demonstrate an ability to control droplet size and placement for scaled-up uses in real world applications. Finally, a concentration process involving transport and sequestration based on surface treatment selective wicking is demonstrated.

4.2 Introduction

As an information-rich technique, Raman spectroscopy exhibits many analytical advantages, although sensitivity is not among them. However, the small cross sections of the

Raman process are routinely overcome via enormous electromagnetic fields located near properly designed plasmonic surface enhanced Raman spectroscopy (SERS) substrates.1-7

Substrate development has been a major focus of SERS research for decades with field enhancements capable in the best of cases to detect a single molecule.8-9 SERS substrates can be characterized into two distinct categories, which include simple, random substrates, such as Ag colloid, silver island films, and polymer–metal nanocomposites,10 and engineered substrates, such as silver films-over nanosphere (AgFON) substrates11 and varying lithographic approaches.12-15 However, an area of SERS research that does not generally receive enough consideration is the development of approaches to deliver general classes of polarizable analytes to the SERS substrate for detection. The SERS substrates used in the following work includes a framework composed of deterministic and stochastic silicon pillar arrays, with the enhancement coming from the use of Ag colloid. Ag colloid can be easily prepared and can exhibit typical

SERS enhancement factors in the range of 105 to 106.16-17 However, when fortuitously aggregated, Ag colloid can reach single molecule detection levels.18 Pillar array systems are advantageous both as a novel approach to separation media and as a platform for superhydrophobic functionalization due to their microscale roughened surface.19

The use of superhydrophobic platforms have recently been gaining significant interest due to the many applications in a wide range of fields and industries from anti-icing coatings for ships20 to droplet manipulation on lab-on-a-chip substrates.21-22 It has recently been demonstrated that superhydrophobic platforms can be used to concentrate droplets of dilute analyte for SERS

48 and fluorescence detection.23 The concentration is a result of the droplet maintaining a Cassie–

Baxter state for a given amount of time, allowing the droplet to undergo evaporation. The

Cassie–Baxter state involves the water droplet keeping in contact with only the tops of the pillars, leaving air underneath the droplet. The transition to the Wenzel state in which the droplet will penetrate the pillars is a result of the competition of the energy barrier and external forces, in which the magnitudes are on the same order of magnitude.24

Superhydrophobic platforms have characteristics that include a hydrophobic surface that is roughened on both the micro- and nanoscale, along with low surface energy.25 The contact angle of a water droplet must be greater than 150° and exhibit a roll-off angle less than 10°.26

The pillars used in this study have both micro- and nanoscale roughness as a result of the fabrication process, which can be seen in previous work.27 The surface was then functionalized to generate a superhydrophobic substrate.

In this article, the goal was to develop a relatively simple procedure that could generate reproducible SERS substrates capable of detection of small, dilute samples. Previous studies involving the concentration of dilute analytes via droplets on top of superhydrophobic surfaces23 were conducted, in which low limits of detection were achieved. This work is an extension of the previous, with a focus on analytical metrics, including effective droplet size and composition, detection limit, concentration factor, dynamic range, and reproducibility. In order to evaluate the different figures of merit, mitoxantrone dihydrochloride was chosen. The analyte exhibits resonance enhancement, which allows for lower limits of detection in SERS.12 Practically, the analyte is an anthracylineantineoplastic agent that is used in the treatment of certain types of cancer, such as acutemyeloid leukemia and metastatic breast cancer.28 An additional method of concentrating dilute samples was demonstrated by wicking a dilute sample through a pillar array

49 that was functionalized to be half hydrophobic and half hydrophilic. To demonstrate high throughput and the ability to scale-up the droplet evaporation concentration method, a prototypical electro-osmotic delivery system was constructed with the ability to multiplex in a small area on the specialized surface.

4.3 Materials

Quantitative and qualitative analyses were completed using a Jobin Yvon Labram Raman

Spectrometer with a HeNe 20 mW laser at a wavelength of 632.8 nm, coupled with a charge- coupled device (Horiba Jobin Yvon, Villeneuve-d'Ascq, France) controlled by LabSpec 5 software.

The silicon based deterministic and stochastic pillar arrays were fabricated and characterized at the Center for Nanophase Materials Science (Oak Ridge National Laboratory, Oak Ridge, TN).

The mitoxantrone dihydrochloride (MIT) was obtained from Santa Cruz Biotechnology (Santa

Cruz, CA). The rhodamine 6g (R6G), aminothiophenol (ATP), and crystal violet (CV) were obtained from Sigma Aldrich (St. Louis, MO).

4.4 Preparation of Silver Colloid

The Ag colloid was prepared using the method described by Lee and Meisel.29

Approximately, 85 mg of silver nitrate was dissolved in 500 mL of HPLC grade water and heated to a boil in a 1000 mL beaker. Ten milliliters of a 1% sodium citrate solution was added dropwise to the heated solution. The solution was kept at a boil for 1 h and then taken off to cool to room temperature. Twenty-eight milliliter aliquots of the stock Ag colloid solution were then concentrated via centrifugation to a volume of 4.0 mL.

4.5 Fabrication of Deterministic Pillar Arrays

To fabricate the deterministic pillar arrays, the deterministic protocol previously reported

50 was used,19 and was described in Chapter 2. Briefly, a CAD program is used to define the pillar pattern and a Heidelberg LW, Model DWL66 laser writer (Center for Nanophase Materials

Science, Oak Ridge National Laboratory, Oak Ridge, TN) is used to create an initial chrome mask. Subsequently, a double layer of positive photoresist (lift-off resist LOR-1A overcoated by positive tone photoresist 955CM-2.1, MicroChem Corp.) was added to the top of a silicon wafer.

The pattern for the arrays was made using a Quintel Inc. contact aligner—designed to mask off the nonpillared areas which were to be etched. Using UV light, holes were formed in the positive photoresist where the pillars would be etched. Approximately 15 to 20 nm of chromium was then deposited onto the wafer to act as the etchant mask, after which the remaining photoresist is removed, leaving only areas of chromium which were not etched. A Bosch process which alternates between etching and adding a passivation layer of fluoropolymer was performed to generate pillars with a height of 20 μm (System 100 Plasma Etcher, Oxford Instruments). A 100 nm layer of silicon oxide was deposited onto the wafer surface using plasma enhanced chemical vapor deposition (PECVD) (System 100 Plasma Deposition Tool, Oxford Instruments). The wafers were then scribed and cleaved into individual 1 cm by 3 cm pillar array chips prior to phase modification.19 All of the deterministic pillar arrays were then functionalized with octadecyltrichlorosilane (Acros Organics, New Jersey) to enhance the hydrophobicity of the substrate.30

4.6 Fabrication of Stochastic Pillar Arrays

The stochastic pillar arrays were created by a metal dewetting procedure based off the work done by Lee and Kim,31 and was described in Chapter 2. Fabrication of the stochastic pillar arrays was completed by first depositing a thin (∼10 nm) Pt film (Thermonics Laboratory, VE-

240) onto a p-type silicon wafer that had 100 nm of thermally grown SiO2. Using a cold wall

51 furnace (Easy Tube 3000, First Nano, Ronkonkoma, NY), the platinum was thermally annealed by setting the radiative heat source to its maximum power (22 kW) for 8 s with a maximum temperature of 900 °C. The furnace contained a 10:1 mixture of argon and hydrogen at 735 Torr.

The dewetting of the metal film onto the surface creates a stochastic circular masking pattern, which can be used as a mask for reactive ion etching (Oxford PlasmaLab, Oxford Instruments,

U.K.) in order to make the height of the high aspect ratio silicon pillars approximately 1–2 μm.32-

33 All of the stochastic pillar arrays were then functionalized with octadecyltrichlorosilane (Acros

Organics, NJ) to enhance the hydrophobicity of the substrate.30

4.7 Silver Colloid Delivery to the Pillar Array Surface

Different methods of Ag colloid delivery were explored, with the goal of having an even layer of Ag colloid on the pillar array surface without the use of complicated instrumentation or techniques. The first method of Ag colloid delivery consisted of developing the Ag colloid in ethanol solution through the superhydrophobic pillar array system via capillary action. Scanning electron microscope (SEM) images were taken to evaluate the surface qualitatively. The SERS substrate then was soaked in an ethanolic 1 × 10–6 M ATP solution. ATP was chosen for its affinity to adsorbing to the silver surface, forming a monolayer.34

The first method of Ag colloid delivery involved developing the colloid solution through the deterministic pillar array via capillary action. The SEM images of the developed Ag colloid solution within the pillar arrays did not appear to deposit adequate Ag colloid onto the surface. A relative standard deviation (RSD) of the SERS signal in a random 0.25 mm2 area of the pillar array was then calculated to evaluate the surface quantitatively. It is important to note that the

RSD was calculated with a laser spot that was approximately 7 μm2 through a 10× microscope objective. Improvements in signal reproducibility can be achieved from a sample translation

52 technique (STT), such as circular-STT (using a mechanical chopper) and linear-STT (repetitious lateral movement of stage), reported in previous work.37-38 A RSD of the peak area at the 1078 cm–1 peak was calculated to compare with other delivery methods. This peak was chosen because of its correspondence to the carbon–sulfur bond, which is an in-plane and in-phase vibrational mode in ATP.39 A RSD of the developing method was calculated to be 249%. One reason for the lack of Ag colloid on the surface was attributed to the “clogging” of the pillar array system at the end of the pillar array where the Ag colloid solution entered (Figure 4.7.1a).

A second method of delivery involved soaking the superhydrophobic pillar array in a water- based Ag colloid solution for 24 h. In order to increase the contact of the Ag colloid solution with the superhydrophobic pillar array surface, the Ag colloid was suspended in 50% ethanol and

50% deionized water, and the pillar array was soaked in solution for 24 h. Additional SEM images were taken and compared, along with the RSD of the SERS signal in a random 0.25 mm2 area within in the pillar arrays. The SEM images of the soaking method, in which Ag colloid was suspended in water, revealed a large amount of Ag colloid being deposited inside the deterministic pillar array system; however, drying effects made the distribution of the colloid nonuniform (Figure 4.7.1b). The calculated RSD of this soaking method was 87.5%, The SEM images of the soaking method in which the Ag colloid was suspended in 50% ethanol and 50% deionized water revealed that the Ag colloid was more uniform than previously (Figure 4.7.1c), and the RSD was calculated to be 69%. From the SEM images and the calculated RSD values of the development and soaking methods, it was determined that the soaking procedure in which the solvent was 50% ethanol and 50% deionized water was the better method of Ag colloid delivery to the surface. The soaking procedure involving 50% ethanol and 50% water was repeated for the stochastic pillar arrays (Figure 4.7.1d), and the calculated RSD value was 76%.

Figure 4.7.1: SEM images of different colloidal delivery systems including (a) development of

Ag colloid in ethanol up the pillar array via capillary action, (b) soaking the deterministic pillar array in the Ag colloid solution with water as the solvent, (c) soaking the deterministic pillar array in the Ag colloid solution with 50% ethanol and 50% water as the solvent, and (d) soaking the stochastic pillar array in the Ag colloid solution with 50% ethanol and 50% water as the solvent.

In addition, previous work with a disk on pillar substrates has demonstrated that an aqua regia rinse can remove oxidized silver disks, allowing for recycling of the expensive substrate.13

Despite the costs of fabrication of the pillar framework herein, the reusability of the substrate described via chemical rinses allows for repetitive application of the hydrophobic phase and the simply prepared and stable Ag colloid.

4.8 Analyte Delivery to the Pillar Array Surface

The initial delivery of analytes involved the use of a National Scientific 25 μL pipet

(Thermo Scientific, Suwanee, GA) to manually place droplets onto the superhydrophobic surface. Varying droplet sizes were allowed to evaporate at room temperature.

Although the manual placement of the droplets functioned adequately, it is a very time and labor intensive procedure. Thus, an apparatus was designed to demonstrate the ability to multiplex via electro-osmotic flow. The experimental setup comprised of a 30 cm long standard coated fused silica capillary tube (TSP, o.d. = 365 μm, i.d. = 78 μm, Polymicro Technologies,

Phoenix, AZ) that was fixed at both ends with a 22G × 1 1/2 in. BD PrecisionGlide Needle

(Becton, Dickinson and Company, Franklin Lakes, NJ). The capillary tube was filled with the analyte solution via a syringe at one end and was placed in a container containing the same analyte solution. The other end of the capillary with the needle was fixed via two aluminum plates and adjusted to a height such that both ends of the capillary were at the same height to prevent gravity flow (Figure 4.11.1a). The positive electrode was then placed into the analyte reservoir, and the negative electrode was connected to the metal plates. A voltage of 2.5 kV was then applied across the electrodes to allow for flow from the positive to the negative. The droplets were allowed to grow in size until the desired droplet size was achieved, and the droplets were then lowered onto the superhydrophobic substrate for analyte concentration.

Another method of delivery involved the use of a deterministic pillar array that was functionalized to be hydrophobic on one end and hydrophilic on the other, with porous silicon oxide (PSO) being vapor deposited onto the arrays prior to functionalization.35-36 Concentration of the sample would occur at the “border” between the hydrophilic and hydrophobic sections by developing an aqueous solution up the hydrophilic end. To create the hydrophilic and superhydrophobic array, one end of the pillar array was dipped into honey and the other end was exposed to a C4 vapor under vacuum, overnight. The pillars were then rinsed with deionized water and toluene to remove excess honey and C4. The hydrophilic end was then dipped into a piranha solution (70% H2SO4/30% H2O2) for 15 min to remove any residual honey that may have gotten stuck in the PSO within the array. The array was then soaked for 24 h in a 50% ethanol and 50% water mixture of concentrated Ag colloid as before. A 1 × 10-6 M aqueous solution of R6G was then developed up the pillar array using a horizontal development chamber,19 and the fluorescence at the border between the hydrophobic and hydrophilic areas of the array was monitored over time using a Nikon Eclipse E600 and Q-capture software. In addition, the SERS signal of the R6G over time was monitored at the border.

4.9 Droplet Size and Composition

Three different characteristics, including droplet size, ionic strength, and solvent composition were examined to determine any effects the changes had on superhydrophobic behavior of the droplets. First, the contact angles of droplets of varying size from 1 to 10 μL were measured on both deterministic and stochastic pillar arrays to determine if the droplets were showing superhydrophobic behavior. For the range from 1 to 5 μL, the contact angles on the deterministic pillar arrays were near 160°, while in the range from 1 to 6 μL, the contact angle of the droplets on the stochastic pillar arrays were also near 160°. The droplets ranging in size from

6 to 10 μL on the deterministic pillar arrays had smaller contact angles (∼150°), while the droplets in the range from 7 to 10 μL on the stochastic pillar arrays exhibited contact angles near

155°. In order to ensure a droplet maintained its superhydrophobic behavior, droplets in the range from 1 to 5 μL were used for both the deterministic and stochastic pillar arrays.

In addition to estimating the contact angles of the droplets in varying sizes, the initial contact area of the droplet was also measured. Upon comparison of both the initial contact areas of the droplets at each size, a pattern emerges in which the droplets on the stochastic pillar arrays have less contact with the surface than the droplets on the deterministic pillars (Figure 4.9.1a).

The smaller contact area suggests that the stochastic pillars exhibit slightly greater superhydrophobic behavior, in part to a lower surface energy compared to the deterministic pillars.40-42

Ionic strength was tested to determine if it had any effect on the superhydrophobic behavior of the droplet on both the deterministic and stochastic pillar arrays. Droplets of water (5

μL) containing concentrations of sodium bromide ranging from 1 to 20 mM were manually placed on top of the deterministic and stochastic pillar arrays. However, the ionic strength did not appear to have any effect on the contact angle or contact area of the droplets.

A test involving solvent composition was then completed, which included making droplets with varying amounts of deionized water, methanol, and ethanol. The study revealed that the droplet must be almost completely composed of water in order to maintain its superhydrophobic behavior on both the deterministic and stochastic pillar arrays (Figure 4.9.1b).

The droplets on the stochastic pillars appeared to be slightly more superhydrophobic than the deterministic pillars, as suggested by the increased contact angle of the droplets observed.

4.10 Determination of the Concentration Factor, Dynamic Range, Detection Limit, and

Reproducibility

In order to estimate the concentration factor of the droplets, the initial contact area of the droplet was compared to the dried area after evaporation to determine the degree of concentration. The concentration factor was determined to be approximately 100. In addition, the

SERS spectra of varying concentrations of 5 μL MIT droplets after evaporation were obtained.

These SERS spectra were compared to pillar arrays that were exposed to the MIT in varying concentrations via developing the analyte up the array, with ethanol as the solvent. By comparing the concentrated droplet to the developed analyte, the degree of concentration can be estimated.

The peak intensity at 1318 cm-1 was used for comparison. The average concentration factor of 5

μL droplets was estimated to be about 110. The concentration factor can be seen by comparing the SERS spectra of the concentrated droplet of the 12 pM MIT to the developed 1.2 nM MIT

(Figure 4.10.1a). The spectra nearly overlap one another, and their normalized peak ratios are close to one another, indicating that the concentration of a 5 μL 10 pM MIT droplet will result in a spectrum similar to that of a developed 1 nM solution through a SERS substrate.

A calibration plot of the intensity of the peak at 1318 cm-1 vs the concentration of the

MIT was generated to determine the dynamic range of the concentration procedure. The dynamic range was determined to be from 1 × 10–7 M to 1 × 10–10 M, following the equation y = 1 × 107x

+ 3.9162, with a correlation coefficient of 0.9868 (Figure 4.10.1b). The observed limit of detection (LOD) was found to be 2.9 × 10–12 M, which is smaller than other reported LOD values for MIT using Ag colloid for SERS enhancement.43-44 The estimated number of MIT molecules observed under the laser beam at the LOD was calculated to be approximately 200, based on a laser spot size of 2.31 um2 (50× objective).

Figure 4.10.1: (a) Calibration plot of the averaged MIT peak at 1318 cm-1 vs the concentration of MIT and (b) a comparison of the SERS spectra of concentrated 12 pM MIT vs developed 1.2 nM MIT.

Evaluation of the reproducibility was completed by calculating an RSD of the highest peak intensities at 1318 cm-1 of five 5 μL droplets of MIT. A raster plot of the droplet spot (0.25 mm2) after evaporation was collected using the 50× microscope objective. The RSD values were calculated in the range of 5–8% for the five separate trials of each concentration seen in Figure

3a. The RSDs were measured using two different deterministic pillar array substrates that were soaked in the same Ag colloid solution. Three trials were conducted on one of the pillar arrays, and two trials were conducted on the other, demonstrating reproducibility in both run-to-run and substrate-to-substrate. Upon evaluation of the RSD values for concentrations within the dynamic range (10–7 M to 10–10 M), it was determined that the procedure was reproducible.

4.11 Electro-osmotic Delivery of Analyte for Multiplexing

Other than increased sample throughput, one of the most important characteristics of multiplexing is the ability to achieve reproducibility. There have been many advances in multiplexing using capillary electrophoresis that have shown reproducible results.45-46 Some issues that arise with reproducibility when using electro-osmotic flow come from variations in the injection voltage, injection time, temperature, ionic strength, and pH of the sample.45 Many of these issues were addressed upon fabrication of the small scale prototype electro-osmotic delivery system used in this work. The voltage remained constant and continuous at 2.5 kV throughout the delivery of the analytes. Additionally, the analytes were kept at room temperature at a pH near 7.0. Upon comparison of different analytes, the concentrations were kept the same in order to maintain a consistent ionic strength throughout the samples.

To demonstrate the reproducibility of the electro-osmotic delivery system that was fabricated, 5 μL of 1.0 × 10-8 M MIT were allowed to flow through four capillaries via a 2.5 kV potential. The power source was then turned off, and the four droplets were lowered and

61 delivered to the superhydrophobic deterministic pillar array surface for evaporation. The pillar array was previously soaked in concentrated Ag colloid as before. It is important to note that these steps could be robotically automated and scaled-up to produce a large number of droplets.

Upon evaporation, the four spots were found via a 9 mm × 1 mm area SERS raster across the substrate using the Jobin Yvon Labram Raman Spectrometer. Each droplet spot was then individually scanned to find the largest MIT SERS signal. The largest SERS signals from each spot were compared, and an RSD of 5.0% was calculated, which suggested the system was reproducible.

Four separate solutions consisting of MIT, R6G, CV, and ATP at a concentration of 1.0 ×

10-8 M were made and loaded into the capillaries for electro-osmotic delivery. Five microliters of each solution was allowed to pass through the capillary to form droplets at the end of the syringe needles. The droplets were allowed to evaporate on the superhydrophobic surface, and the spots were then analyzed as previously (Figure 4.11.1b-c). SERS signals of each analyte were obtained, the prototype exhibited the ability to concentrate multiple samples in a small area, which could be scaled up and automated such that a large number of samples could be concentrated onto the surface.

4.12 Concentration of Analyte and Wicking

Another method of concentration of analyte utilizes a unique, spatially defined functionalization of a pillar array surface such that half of the array is rendered hydrophilic while the other half is hydrophobic, forming a “border.” Contact angles of both the hydrophilic (20°) and hydrophobic (155°) ends of the array were measured to confirm the functionalization succeeded. The aqueous solution of R6G was then allowed to develop up the pillar array, with the solution constantly being added via the wick (soaked in 1 × 10-6 M R6G). Such methods are

Figure 4.11.1: (a) Schematic diagram of the electro-osmotic delivery system used along with the

(b) actual droplets that were spotted and allowed to evaporate on the superhydrophobic substrate.

(c) The representative rasters of the droplet spots after evaporation and the largest SERS spectrum found in each droplet spot are presented to demonstrate the ability to multiplex.

63 used to concentrate aqueous samples in reversed phase HPLC during the injection process.37-48

The fluorescence was measured over a 2 min time period to monitor the stacking effect of the analyte at the “border.” The R6G was visually identifiable through the fluorescence microscope as soon as 10 s after the solvent front began to flow into the “border” region, with the intensity growing exponentially for about 90 s after initial contact with the “border” (Figure 4.12.1a-b).

Additionally, the SERS intensity at 1507 cm-1 of the R6G was monitored over time via a raster of the same 6 spots every 10 seconds at the “border.” The signals from one spot were chosen and compared over time, with the signal appearing to reach a maximum at about 110 s (Figure

4.12.1b-c).

From the measured fluorescence and SERS intensities, mass transfer of the R6G can be seen at the “border,” demonstrating a novel method of analyte concentration. One advantage of this technique is the ability to concentrate trace amounts of analyte within a sample to a centralized location for detection. This could be especially useful for very small sample sizes typically needed in the medical field. In addition, this technique is capable of being multiplexed by selectively functionalizing specific areas of the arrays in a multiple channel configuration while masking others, thereby increasing throughput.

4.13 Conclusions

This work demonstrates the ability to use a relatively simple procedure to generate a superhydrophobic SERS substrate capable of concentrating dilute samples such that detection is possible. The combination of the simple Ag colloid preparation, along with the functionality and reusability of both the deterministic and stochastic pillar arrays is a cost-effective method of generating superhydrophobic substrates capable of SERS detection. The presented electro- osmotic delivery system is a prototype that could be scaled-up and automated for use in many

65 fields of study including medicine. The ability to multiplex is of the utmost importance for efficiency and speed in the laboratory. This system is a cost-effective alternative that allows the researcher the ability to observe multiple analytes of interest, with the potential application to environmental or medicinal samples in future work, using very little sample and substrate. In addition, a novel method of analyte concentration via wicking was demonstrated utilizing both hydrophilic and superhydrophobic regions of a pillar array.

4.14 Acknowledgements

This material is based on work supported in part by the National Science Foundation under

Grant CHE-1144947 with the University of Tennessee. A portion of this research was conducted at the Center for Nanophase Materials Sciences, which is sponsored at Oak Ridge National

Laboratory by the Scientific User Facilities Division, Office of Basic Energy Sciences, U.S.

Department of Energy. A portion of this work was supported by the U.S. Department of Energy,

SunShot Program of the Office Energy Efficiency and Renewable Energy.

4.15 References

1. McQuillan, J. A. Notes Rec. R. Soc. 2009, 63, 105−109. 2. Haynes, C. L.; Yonzon, C. R.; Zhang, Z.; Van Duyne, R. P. J. Raman Spectrosc. 2005, 36, 471−484. 3. Yan, B.; Thubagere, A.; Premasiri, W. R.; Ziegler, L. D.; Negro, L.D.; Reinhard, B. M. ACS Nano 2009, 3 (5), 1190−1202. 505 4. Litorha, M.; Haynes, C. L.; Haes, A. J.; Jensen, T. R.; Van Duyne, R. P. J. Phys. Chem. B 2001, 105 (29), 6907−6915. 5. Hatab, N. A.; Hsueh, C.; Gaddis, A. L.; Retterer, S. T.; Li, J.; Eres, G.; Zhang, Z.; Gu, B. Nano Lett. 2010, 10 (12), 4952−4955. 6. Wang, W.; Ruan, C.; Gu, B. Anal. Chim. Acta 2006, 1, 121−126. 7. Zeisel, D.; Deckert, V.; Zenobi, R.; Vo-Dinh, T. Chem. Phys. Lett. 1998, 283, 381−385. 8. Li, L.; Hutter, T.; Steiner, U.; Mahajan, S. Analyst 2013, 138, 4574−4578. 9. Botta, R.; Upender, G.; Sathyavathi, R.; Rao, D. N.; Bansal, C. Mater. Chem. Phys. 2013, 137, 699−703. 10. Giesfeldt, K. S.; Connatser, R. M.; De Jesus, M. A.; Lavrik, N.V.; Dutta, P.; Sepaniak, M. J. Appl. Spectrosc. 2003, 57 (11), 1346−1352. 11. Zhang, X.; Zhao, J.; Whitney, A. V.; Elam, J. W.; Van Duyne, R. P. J. Am. Chem. Soc. 2006, 128, 10304−10309. 12. De Jesus, M. A.; Giesfeldt, K. S.; Oran, J. M.; Abu-Hatab, N. A.;Lavrik, N. V.; Sepaniak, M. J. Appl. Spectrosc. 2005, 59 (12), 1501−1508. 13. Wells, S. M.; Polemi, A.; Lavrik, N. V.; Shuford, K. L.; Sepaniak, M. J. Chem. Commun. 2011, 47, 3814−3816. 14. Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem. B 2001, 105, 5599−5611. 15. Gopinath, A.; Boriskian, S. V.; Premasiri, W. R.; Ziegler, L.; Reinhard, B. M.; Negro, L. D. Nano Lett. 2009, 9 (11), 3922−3929.11. 16. Clou, S.; Cao, Y.; Huang, H.; Su, D.; Huang, C. J. Phys. Chem. C. 2009, 113, 9520−9525. 17. Xu, H.; Aizpurua, J.; Kall, M.; Apell, P. Phys. Rev. E 2000, 62 (3), 4318−4324. 18. Camden, J. P.; Deringer, J. A.; Wang, Y.; Masiello, D. J.; Marks, L. D.; Schatz, G. C.; Van Duyne, R. P. J. Am. Chem. Soc. 2008, 130, 12616−12617. 19. Kirchner, T. B.; Hatab, N. A.; Lavrik, N. V.; Sepaniak, M. J. Anal. Chem. 2013, 85, 11802−11808. 20. Gao, D.; Jones, A. K.; Sikka, V. K. Anti-Icing Superhydrophobic Coatings. U.S. Patent No. US20100314575 A1, December 16, 2101. 21. Draper, M. C.; Crick, C. R.; Orlickaite, V.; Turek, V. A.; Parkin, I. P.; Edel, J. B. Anal. Chem. 2013, 85, 5405−5410. 22. Mertaniemi, H.; Jokinen, V.; Sainiemi, L.; Franssila, S.; Marmur, A.; Ikkala, O.; Ras, R. H. A. Adv. Mater. 2011, 23, 2911−2914. 23. De Angelis, F.; Gentile, F.; Mecarini, F.; Das, G.; Moretti, M.; Candeloro, P.; Coluccio, M. L.; Cojoc, G.; Accardo, A.; Liberale, C.; Zaccaria, R. P.; Perozziello, G.; Tirinato, L.; Toma, A.; Cuda, G.; Singolani, R.; Di Fabrizio, E. Nat. Photonics 2011, 5, 683−688. 24. Murakami, D.; Hiroshi, J.; Takahara, A. Langmuir 2014, 30, 2061−6067. 25. Celia, E.; Darmanin, T.; Givenchy, E. T.; Amigoni, S.; Guittard, F. J. Colloid Interface Sci. 2013, 402, 1−18.

26. Kang, S.-M.; Hwang, S.; Jin, S.-H.; Choi, C.-H.; Kim, J.; Park, B. J.; Lee, D.; Lee, C.-S. Langmuir 2014, 30 (10), 2828−2834. 27. Taylor, L. C.; Lavrik, N. V.; Sepaniak, M. J. Anal. Chem. 2010, 82, 9549−9556. 28. Manikas, A. C.; Beobide, A. S.; Voyiatzis, G. A. Analyst 2009, 134, 587−592. 29. Lee, P. C.; Meisel, D. J. Phys. Chem. 1982, 86, 3391−3395. 30. Hennion, M. C.; Picard, C.; Caude, M. J. Chromatogr. 1978, 166, 21−35. 31. Lee, J.; Kim, B. Mater. Sci. Eng., A 2007, 769−773. 32. Charlton, J.J; Lavrik, N.V.; Bradshaw, J.A.; Sepaniak, M.J. ACS Appl. Mater. Interfaces 2014, 6, 17894−17901 DOI: 10.1021/am504604j. 33. Agapov, R. L.; Srijanto, B.; Fowler, C.; Briggs, D.; Lavrik, N. V.; Sepaniak, M. J. Nanotechnology 2013, 24 (50), 505302. 34. Skantarova, L.; Orinak, A.; Orinakova, R.; Lofaj, F. Nano-Micro Lett. 2012, 4 (3), 184−188. 35. Ceiler, M. F.; Kohl, P. A.; Bidsrup, S. A. J. Electrochem. Soc. 1995, 142 (6), 2067−2071. 36. Yang, P.; Liu, L.; Mo, J.; Yang, W. Semicond. Sci. Technol. 2010, 25 (4), 043001. 37. De Jesus, M. A.; Giesfeldt, K. S.; Sepaniak, M. J. Appl. Spectrosc. 2003, 57 (4), 428−438. 38. Freye, C. E.; Crane, N. A.; Kirchner, T. B.; Sepaniak, M. J. Anal. Chem. 2013, 85, 3991−3998. 39. Zheng, J.; Zhou, Y.; Li, X.; Ji, Y.; Lu, T.; Gu, R. Langmuir 2003, 19, 632−636. 40. Dupuis, A.; Yeomans, J. M. Langmuir 2005, 21 (6), 2624−2629. 41. McHale, G.; Aquil, S.; Shirtcliffe, N. J.; Newton, M. I.; Erbil, H. Y. Langmuir 2005, 21 (24), 11053−11060. 42. Jopp, J.; Grull, H.; Yersushalmi-Rozen, R. Langmuir 2004, 20 (23), 10015−10019. 43. McLaughlin, C.; MacMillan, D.; McCardle, C.; Smith, W. E. Anal. Chem. 2002, 74, 3160−3167. 44. Ackermann, K. R.; Henkel, T.; Popp. J. Chem. Phys. Chem. 2007, 8, 2665−2670. 45. Xue, G.; Pang, H.; Yeung, E. S. Anal. Chem. 1999, 71, 2642−2649. 46. He, Y.; Yeung, E. S. Electrophoresis 2003, 24, 101−108. 47. Chiaia, A. C.; Banta-Green, C.; Field, J. Environ. Sci. Technol. 2008, 42, 8841−8848. 48. Snyder, L. R.; Kirkland, J. J. Introduction to Modern Liquid Chromatography, 2nd ed.; John Wiley & Sons: New York, 1979; pp 729−731.

Chapter 5

Ultra-Thin Layer Chromatography with Integrated Silver Colloid-Based SERS Detection

The research presented in Chapter 5 has been adapted from a research article published in

Electrophoresis (Wallace, R.A.; Lavrik, N.V.; Sepaniak, M.J. Ultra-thin layer chromatography with integrated silver colloid-based SERS detection. Electrophoresis, 2017, 38, 361–367.) This chapter focuses on the use of our deterministic pillars for separations and detection utilizing silver colloid as a means to use SERS.

5.1 Abstract

Simplified lab-on-a-chip techniques are desirable for quick and efficient detection of analytes of interest in the field. The following work involves the use of deterministic pillar arrays on the micro-scale as a platform to separate compounds, and the use of Ag colloid within the arrays as a source of increased signal via surface enhanced Raman spectroscopy (SERS). One problem traditionally seen with SERS surfaces containing Ag colloid is oxidation; however, our platforms are superhydrophobic, reducing the amount of oxidation taking place on the surface of the Ag colloid. This work includes the successful separation and SERS detection of a fluorescent dye compounds (resorufin and sulforhodamine 640), fluorescent anti-tumor drugs (Adriamycin and Daunomycin), and purine and pyrimidine bases (adenine, cytosine, guanine, hypoxanthine, and thymine).

5.2 Introduction

Analytical techniques on the micro-scale for accurate and proficient sample analysis is vital to providing scientists the tools necessary to expedite results, especially for complex mixtures of samples. One such technique is the use of highly ordered, lithographic, micro-scale pillar arrays

(deterministic pillar arrays) which have been previously shown to be an effective form of planar chromatography for separations of analytes of interest.1-5 The benefits of using the silicon-based

70 deterministic pillar arrays include uniformity and reusability, along with the freedom to tailor geometries. Additionally, the deterministic pillars provide a scaled-down version of traditional thin-layer chromatography. These silicon pillars can be easily functionalized with a hydrophobic or hydrophilic phase, further demonstrating their practicability in separating complex mixtures.

When combined with a potentially sensitive detection method, these ultra-thin layer chromatography (UTLC) platforms prove to be quite valuable.

Surface enhanced Raman spectroscopy (SERS) has proven to be a sensitive detection method for trace amounts of biologically relevant compounds, with enhancement factors routinely as much as 106.6-8 SERS requires the laser excitation of localized surface plasmons on a roughened metal surface or in metal nanoparticles, causing an oscillation of the free electron density with respect to the fixed ions in a metal. The two most commonly used metals for SERS are silver and gold, due to their high electrical conductivity and their stability in ambient, electrochemical, and ultrahigh vacuum environments.8 There are many different types of both random9-11 and engineered SERS substrates12-15 that have been used with various advantages and disadvantages for each. Unlike silver nanorods or our well-modelled disk on pillar substrates,12 silver colloid is more random in their geometry and size. We chose to use silver colloid for its ease of fabrication, scalability to large quantities, and potential for single molecule detection.11 Oxidation can limit the robustness, especially for silver-based SERS substrates. The following work demonstrates the ability to limit the oxidation of the silver colloid, thereby increasing the lifetime of the detection platform.

The superhydrophobic nature of our deterministic pillar arrays after functionalization with a C18 phase provided us with a couple of advantages. One of the biggest advantages was the ability to spot our samples in precise locations with minimal spreading. Contrary to previous reports

71 involving the use of silver nanorods as the means to both separate and detect analytes using

SERS,16-17 we have the ability to change the stationary phase independent of the SERS detection scheme, enhancing the separation selectivity possibilities for our chromatographic platforms. In addition, superhydrophobic surfaces prevent water contamination on the surface18, which we show can potentially slow the oxidation process of the silver colloid used in this work.

Previous work from others has shown the ability to combine liquid chromatography (LC) with SERS as a valid method of separation and detection of analytes of interest.19-23 In addition, the Sepaniak group has previously demonstrated the ability to detect compounds after separations on thin-layer chromatography (TLC) plates and in micro-fluidic platforms, utilizing SERS.24-25

The following work involves the integration of several operations including the ability to use small samples due to the superhydrophobic nature of the functionalized pillars, separations with the use of our deterministic pillar arrays, and detection with SERS via Ag colloid addition to the pillars. This work represents a significant advance by combining high efficiency chromatography with SERS. We show the separations of several biologically relevant compounds that have observable SERS signals, which is the first demonstration of SERS-based detection using our ultra-thin layer chromatography platforms.

5.3 Materials

A Nikon Eclipse E600 fluorescent microscope (Nikon Instruments Inc., Mel-ville, NY) with Q-capture software was used for qualitative analysis, along with a Jobin Yvon Labram Raman

Spectrometer with a HeNe 9 mW laser at a wavelength of 632.8 nm, coupled with a charge-coupled device (Horiba Jobin Yvon, Ville-neuved’Ascq, France) via LabSpec 5 software. The deterministic pillar array substrates were fabricated and characterized at the Center for Nanophase Materials

Science (Oak Ridge National Laboratory, Oak Ridge, TN). The mitoxantrone dihydrochloride

(MIT) was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). The resorufin was obtained from Molecular Probes (Eugene, OR). The sulforhodamine 640 was purchased from Fisher

Scientific (Pittsburgh, PA). The Adriamycin (A1), Daunomycin (D1), adenine, cytosine, guanine, hypoxanthine, and thymine were purchased from Sigma-Aldrich (St. Louis, MO).

5.4 Fabrication of Deterministic Pillar Arrays

Fabrication of deterministic pillar arrays has been described in previous work from our group,1,26 as well as in Chapter 2. In short, a silicon wafer is overcoated with a double layer of positive photoresist, followed by exposure to UV light through a mask using a Quintel Inc. contact aligner. Chromium is then deposited onto the surface, followed by the metal lift off and etching of the pillars via a Bosch process using a System 100 Plasma etcher (Oxford Plasma Technology

Inc.). A 100 nm layer of porous silicon oxide was then deposited onto the pillar surface using room temperature plasma enhanced chemical vapor deposition (PECVD).27 The pillar dimensions were

2 µm in diameter and 20 µm in height, with the inter-pillar distance being 1 µm.The pillar arrays were made into dimensions of 1 cm by 3 cm for the fluorescent compounds and 1 mm by 3 cm for the purine and pyrimidine bases. All of the pillar arrays were then functionalized using octadecyltrichlorosilane (Acros Organics, New Jersey) in order to generate a superhydrophobic, reversed phase surface.

5.5 Preparation of the Silver Colloid-Pillar Array Surface

The Ag colloid used in the following work was prepared us-ing the same method as described by Lee and Meisel.11 Briefly, 85 mg of silver nitrate was dissolved in 500 mL of HPLC grade water and heated to a boil, while continuously stirring, in a 1000 mL beaker. Ten milliliters of a 1% sodium citrate solution was added dropwise. The beaker was taken off the heat and stirred

73 for 1 hour at room temperature. Approximately 60 milliliter aliquots of the Ag colloid solution were then concentrated via centrifugation to a volume of 4.0 mL, with half of the solution consisting of ethanol and the other half containing deionized water. The average size of the Ag colloid was about 80 nm. The pillar arrays were then soaked overnight in the Ag colloid solution to allow for sufficient deposition of the colloid into the pillar array.

5.6 Analyte Delivery to the Surface and Development

Droplets were manually placed onto the superhydrophobic surface using a National

Scientific 25 µL pipet (Thermo Scientific, Suwanee, GA). For all the fluorescent compounds, each droplet was precisely 5 µL in volume for consistency. In the case of the purine and pyrimidine bases, 1 µL droplets of each compound were spotted on top of one another after drying. After deposition and evaporation of the droplets, a SERS spectrum of the droplet spot is obtained, as well as a fluorescence image for comparison later. The pillar arrays were then exposed to the mobile phase and allowed to develop vertically for approximately 1 minute. The mobile phase used in this work consisted of 60/40% ethanol/water. A fluorescence image of the developed original spot was then obtained, followed by subsequent images of the separated fluorescent compounds. SERS data was then obtained from rasters of the areas in which the fluorescent compounds were located to confirm their identities. It is important to note that the separations carried out in this manuscript were initially separated without Ag colloid for comparison, excluding the purine and pyrimidine bases in which only SERS activity was observed.

5.7 Silver Colloid Stability

To justify the use of SERS as a detection mechanism for separations on our deterministic pillar arrays, it must first be demonstrated that the Ag colloid is stable enough to provide a

74 reasonable platform for detection. First, the Ag colloid must be shown be unaffected by the introduction of the mobile phase (60/40% ethanol/water). In order to test this, a visual inspection using SEM images before and after introduction of the mobile phase were observed (Figure 5.7.1a and b). The silver colloid appeared stable at the bases of the pillars, which is a favorable outcome for SERS detection. We presented the SEM images of the pillars in such a way to show both the geometry as well as the presence of Ag colloid in the pillars. The slight loss of colloid observed in

Figure 5.7.1b is only seen on the outside of the pillar array.

In addition to visually observing the silver colloid before and after the introduction of a mobile phase, a study was conducted using SERS signal over several large areas to determine the spread of silver colloid. Due to its affinity for silver, aminothiophenol was used as the indicator for the presence of silver throughout the pillar array. A raster of a 10,000 µm2 area with steps of

100 µm in each direction were observed over three separate pillar arrays and averaged. The relative standard deviation (RSD) of the SERS signal of the aminothiophenol was 69% before the mobile phase and 75% after. The small change in RSD indicated that the colloid appeared to stay within the pillar arrays, despite the introduction of a mobile phase. We attribute the continued presence of the colloid to the Van der Waals interactions of the colloid to the functionalized silicon dioxide surface, which several other studies have suggested.28-30

The next step in establishing the validity of this form of separation and detection, is to determine the time sensitivity of the SERS platforms. Ag colloid is highly prone to oxidation and it would be desirable to have a detection platform that is stable in the laboratory setting for a reasonable length of time. To provide a basis for how stable the Ag colloid is on a flat surface and at room temperature, Ag colloid could settle onto a microscope slide. A 5 µL droplet of 1 µM MIT was deposited on top of the Ag colloid covered microscope slide and allowed to dry. SERS spectra

Figure 5.7.1: SEM images of (a) before and (b) after the chromatographic development with the

60% ethanol and 40% water mobile phase, showing the silver colloid on the surface.

76 of the MIT were observed for the 5 microscope slides immediately after drying and agreed with previous work.26 Five different microscope slides sprayed with Ag colloid were then left on top of the lab bench for 24 hours, and SERS spectra were obtained again. This was repeated once more for a total elapsed period of 2 days. The normalized mean and RSD of the intensity of the MIT peak at 1318 cm-1 were compared (Figure 5.7.2a). After 2 days of sitting at room temperature in the lab, the MIT SERS signal was only 1% of the original signal (Figure 5.7.2b).

The same process was repeated with the superhydrophobic pillar arrays, in which Ag colloid was allowed to fall into the pillars. Five microliter droplets of 1 µM MIT were deposited and allowed to evaporate on top of the pillar arrays. SERS activity of the MIT were observed and compared initially, at 3 days, and at 6 days (Figure 5.7.2c-d). In addition, the pillar arrays were stored both in and out of vacuum to determine if there would be a difference in oxidation rate of the Ag colloid. Upon comparison, it appeared as though the SERS signal of the MIT was consistent after 6 days of being both in and out of vacuum. This presented an interesting discovery, in that the Ag colloid appeared to be “protected” from oxidation when settled at the bottom of the superhydrophobic pillar arrays. We believe that the combination of water vapor with molecular oxygen increases the likelihood of oxidation, however, our superhydrophobic surface reduces the amount of water condensate and water vapor coming into contact with the Ag colloid. Because our platforms resisted oxidation for at least 6 days, we hypothesize that the lack of water vapor plays a role in the oxidation process. This means that the pillar arrays can be prepared several days ahead of a potential separation, and still maintain good SERS activity.

5.8 Fluorescence UTLC-SERS Separations Utilizing Pillar Array Platforms

To test the viability of our UTLC-SERS platforms, a 5 µL droplet containing 1 µM resorufin and 1 µM sulforhodamine 640 was deposited onto the superhydrophobic surface to

78 concentrate and descend into the pillars. The pillar array was then lowered into a mobile phase of

60/40% ethanol/water and allowed to develop for 75 seconds. From the raster of each separated band, a chromatogram was generated based off of the SERS signal of the strongest peak for each component - ~554 cm-1 for resorufin and ~1483 cm-1 for sulforhodamine 640 (Figure 5.8.1a).

Using the Nikon Eclipse E600 fluorescent microscope, images of the original spot after development were obtained (Figure 5.8.1b), along with images of the resorufin and sulforhodamine 640 bands (Figure 5.8.1c and 5.781d). The separation of the components was then observed using the Jobin Yvon Labram Raman Spectrometer via a rastering pattern. SERS spectra of each component including the original spot were collected, with the strongest spectrum for each being presented (Figure 5.8.1e-g). The resorufin spectrum agreed with previously obtained spectra.31 It is important to note that the chromatogram peaks are not Gaussian due to the small number of data points available from the raster of each band, each having an average of about 8 to

10 data points. Because the separated bands are not Gaussian in shape the spot shapes are prone to interpretation, which is common in planar chromatography. This can be attributed to several factors, including stacking and drying effects, especially at the solvent front.1 It is also important to note that the goal of this study is to qualitatively analyze our system, however, previous work has demonstrated the quantitative ability of our pillar array platforms also including a limit of detection of 2.9 pM for MIT.26

Another set of fluorescence separations utilizing our UTLC-SERS platforms was performed using the anti-tumor drugs A1 and D1, which naturally fluoresce. The same process was repeated as before, including the deposition of a 5 µL droplet containing 10 µM A1, 10 µM

D1, and 1 µM sulforhodamine 640 onto the superhydrophobic surface. After development for 75 seconds using a 60/40% ethanol/water mobile phase, fluorescence images and SERS activity were

(c&f), and the sulforhodamine 640 band (d&g).

81 observed (Figure 5.8.2b- i). Both the A1 and the D1 SERS spectra agreed with other studies.31-34

The spectra of both A1 and D1 were found to be very similar with no distinctive differences in the vibrational bands observed; however, our UTLC platform easily separated the two compounds.

Note that the spots are narrower in the direction of flow, presumably due to sample stacking.1 In order to determine the identity of the A1 and D1 band, an additional separation was performed in which D1 was 3 times as concentrated as A1. The location of the brighter band would then indicate the identity of the band as D1, which was found to be the most retained and consistent with prior work.35

To estimate efficiency, we used H = σ2/d, with H equal to the plate height (efficiency), σ equal to one-fourth of the apparent spot size, and d equal to the distance traveled.1 Based on the analysis of the fluorescence images of three successful separations, the D1 band had an average efficiency of 4.3 µm with a RSD of 7.3%, the A1 band had an average efficiency of 4.0 µm with a

RSD of 8.4%, and the sulforhodamine 640 had an average efficiency of 5.1 µm with a RSD of

44%. All three components had good efficiencies, with A1 and D1 having acceptable RSD values.

The poor RSD of the sulforhodamine 640 band may be attributed to extreme drying effects at the solvent front. The chromatogram again was made from the SERS signal of the most prominent peak in each spectra, with about 10 to 19 data points (Figure 5.7.2a). The strongest peak for each component included ~1196 cm-1 for A1 and D1 and ~1483 cm-1 for sulforhodamine 640.

5.9 UTLC-SERS Separations with Pillar Arrays

With several successful separations using our UTLC-SERS platforms, the next step was to look into analytes that were not fluorescent. Five purine and pyrimidine bases were chosen due to their importance biologically, and their unique SERS spectra. The solution for each purine base was prepared with the pH being adjusted based off of pKa values. Adenine, thymine and cytosine

82 were each prepared to a pH of approximately 7.0. Guanine was prepared at a pH of 6.0, and hypoxanthine was prepared at a pH of 5.0.36-37 Initially, different pairs of the purine and pyrimidine bases were spotted onto the surface and separated using 60/40% ethanol/water to determine approximate retardation factors and their SERS activity. Once it was determined that each component would separate, 1 µL droplets of each purine base were deposited onto the surface in the same spot, with each having a concentration of 100 µM. Again, the mobile phase used was

60/40% ethanol/water. The SERS activity was then observed via several point by point rasters across the deterministic pillar array.

Three of the five purine and pyrimidine bases separated completely, with good resolution, while guanine and thymine appeared to co-elute (Figure 5.9.1a), and the order of elution agreed with previous work.19,37 However, distinctive spectra for each of the five purine and pyrimidine bases were observed and compared (Figure 5.9.1b-g).38-42 In order to evaluate the efficiency of the separation, we used the SERS activity from raster plots, in which the distance between each raster spot was 20 µm. Again, we used our approximate method of calculating efficiency, and the efficiencies for each purine and pyrimidine bases were less than 1 µm. Although the calculated efficiencies based on SERS and rastering are artificially low, we still had a successful separation that also provided vibrational spectra.

5.10 Conclusions

This work demonstrates that deterministic pillar arrays have the ability to separate compounds efficiently, even in the presence of silver colloid. Using both fluorescence and SERS signals, plate heights of 5 µm or less are achieved. We also presented a five component separation of biologically relevant purine and pyrimidine bases. The combination of simply made silver colloid with the deterministic pillars was shown to be a novel method of separation and detection.

(b) the original spot after development, (c) adenine, (d) hypoxanthine, (e) guanine, (f) thymine, and (g) cytosine.

This work could lead to more integrated and automated complex separation and detection systems, such as extracted biological samples for the anti-tumor drugs and DNA bases studied herein, or metabolomics compounds in a variety of sample matrices. In addition to the colloid being stable to flow, the superhydrophobic nature of our pillars provided a unique outcome in that the silver colloid appeared to resist oxidation with the pillar array, thereby extending their usability form a day to at least a week.

5.11 Acknowledgements

This material is based upon work supported by the National Science Foundation under

Grant CHE-1144947 with the University of Tennessee. A portion of this research was conducted at the Center for Nanophase Materials Sciences, which is sponsored at Oak Ridge National

Laboratory by the Scientific User Facilities Division, Office of Basic Energy Sciences, U.S.

Department of Energy.

5.12 References

1. Kirchner, T.B.; Hatab, N.A.; Lavrik, N.V.; Sepaniak, M.J. Anal. Chem. 2013, 85, 11802– 11808. 2. Chandra, D.; Yang, S. Accounts Chem. Res. 2010, 43, 1080-1091. 3. Malshe, W.D.; Eghbali, H.; Clicq, D.; Vangelooven, J.; Gardeniers, H.; Desmet, G. Anal. Chem. 2007, 79, 5915-5926. 4. Gzil, P.; Vervoort, N.; Baron, G.V.; Desmet, G. Anal. Chem. 2003, 75, 6244-6250. 5. He, B.; Regnier, F. J. Pharm. Biomed. Anal. 1998, 17, 925-932. 6. Xu, H.; Aizpurua, J.; Kall, M.; Apell, P. Phys, Rev. E. 2000, 62, 4318-4324. 7. Ciou, S.; Cao, Y.; Huang, H.; Su, D.; Huang, C. J. Phys. Chem. C. 2009, 113, 9520-9525. 8. Zhang, X.; Yonzon, C.R.; Young, M.A.; Stuart, D.A.; Van Duyne, R.P. IEE Proc.- Nanobiotechnol. 2005, 152, 195-206. 9. Schlegel, V.L.; Cotton, T.M. Anal. Chem. 1991, 63, 241-247. 10. Giesfeldt, K. S.; Connatser, R. M.; De Jesus, M. A.; Lavrik, N. 518V.; Dutta, P.; Sepaniak, M. J. Appl. Spectrosc. 2003, 57, 1346− 5191352. 11. Lee, P.C.; Meisel, D. J. Phys. Chem. 1982, 86, 3391-3395. 12. Polemi, A.; Wells, S.M.; Lavrik, N.V.; Sepaniak, M.J.; Shuford, K.L. J. Phys. Chem. 2011, 115, 13624-13629. 13. Wells, S.M.; Polemi, A.; Lavrik, N.V.; Shuford, K.L.; Sepaniak, M.J. Chem. Commun., 2011, 47, 3814-3816. 14. Zhang, X.; Zhao, J.; Whitney, A.V.; Elam, J.W.; Van Duy-ne, R.P. J. Am. Chem. Soc. 2006, 128, 10304-10309. 15. Haynes, C.L.; Van Duyne, R.P.; J. Phys. Chem. B. 2001, 105, 5599-5611. 16. Chen, J.; Abell, J.L.; Huang, Y.; Zhao, Y. P. SPIE., 2012. 17. Chen, J.; Huang, Y.; Zhao, Y. J. Mat. Chem. B. 2015, 3, 1898-1906. 18. Zhang, X.; Shi, F.; Niu, J.; Jiang, Y.; Wang, Z. J. Mater. Chem. 2007, 18, 621-633. 19. Cowcher, D.P.; Jarvis R.; Goodacre, R.; Anal. Chem. 2014, 86, 9977-9984. 20. Chen, J.; Abell, J.; Huang, Y.; Zhao, Y. Lab Chip. 2012, 12, 3096-3102. 21. Trachta, G.; Schwarze, B.; Saguller, B.; Brehm, G.; Schneider, S. J. Mol. Struct. 2004, 693, 175-185. 22. Carrillo-Carrion, C.; Aremtna, S.; Simonet, B.M.; Valcarcel, M.; Lenl, B. Anal. Chem. 2011, 83, 9391-9398. 23. Cabalin, L.M.; Ruperez, A.; Laserna, J.J. Anal. Chim. Acta. 1996, 318, 203-210. 24. Freye, C.E.; Crane, N.A.; Kirchner, T.B.; Sepaniak, M.J. Anal. Chem. 2013, 85, 3991-3998. 25. Connaster, R.M.; Riddle, L.A.; Sepaniak, M.J. J. Sep. Sci. 2004, 27, 1545-1550. 26. Wallace, R.A.; Charlton, J.J.; Kirchner, T.B.; Lavrik, N.V.; Datskos, P.G.; Sepaniak, M.J. Anal. Chem. 2014, 86, 11819-11825. 27. Charlton, J.J.; Lavrik, N.; Bradshaw, J.A.; Sepaniak, M.J. ACS Appl. Mater. Interfaces. 2014, 6, 17894-17901. 28. Matope, S.; Rabinovich, Y.I.; Van der Merwe, A.F. Colloid Surface A. 2012, 411, 87-93. 29. Min, Y.; Akbulut, M.; Kristiansen, K.; Golan, Y.; Israelch-vili, J. Nat. Mater. 2008, 7, 527- 538. 30. Luo, Y.; Zhao, R.; Pendry, J.B. P. Natl. Acad. Sci. USA. 2014, 111, 18422-18427. 31. Connaster, R.M. PhD. Dissertation, University of Tennessee, Knoxville, 2006. 32. Nabiev, I.R.; Morjani, H.; Manfait, M. Eur. Biophys. J. 1991, 19, 311-316.

33. Heywang, C.; Chazalet, S.; Masson, M.; Bolard, J. Biophys. J. 1998, 75, 2368-2381. 34. Eliasson, C.; Loren, A.; Murty, K.V.G.K.; Josefson, M.; Kall, M.; Abrahamsson, J.; Abrahamsson, K. Spectrochim. Acta. 2001, 57, 1907-1915. 35. Sepaniak, M.J. Yeung, E.S. J. Chromatogr. 1980, 190, 377-383. 36. Chowdhury, J.; Mukherjee, K.M.; Misra, T.N. J. Raman Spectrosc. 2000, 31, 427-431. 37. Sheng, R.; Ni, F.; Cotton, T.M. Anal. Chem. 1991, 63, 437-442. 38. Huang, Q.; Kaiser, K.; Benner, R. Limno. Oceanogr. 2012, 10, 608-616. 39. Cortes, S.S.; Garcia-Ramos, J.V. J. Raman Spectrosc. 1992, 23, 61-66. 40. Pergolese, B.; Bonifacio, A.; Bigotto, A. Phys. Chem. Chem. Phys. 2005, 7, 3610-3613. 41. Otto, C.; van den Tweel, T.J.J.; de Mul, F.F.M.; Greve, J.; J. Raman Spectrosc. 1986, 17, 289-298. 42. Barhoumi, A.; Zhang, D.; Tam, F.; Halas, N. J. Am. Chem. Soc. 2008, 130, 17740-17747.

Chapter 6

Evaluation of Porous Silicon Oxide on Silicon Micro-Cantilevers for Sensitive Detection of

Gaseous HF

The research presented in Chapter 6 has been adapted from a research article currently being submitted for publication. This chapter focuses on evaluating the potential of porous silicon oxide as a sensitive coating on top of micro-cantilever mechanical sensors for trace hydrogen fluoride gas sensing. Physical observations of the changes in the micro-cantilevers during exposure to the hydrogen fluoride gas are noted, along with an analytical treatment of the physical data.

6.1 Abstract

Sensitive detection of harmful chemicals in industrial applications is pertinent to safety.

In this work, we demonstrate the use of a sensitive silicon micro-cantilever (MC) system with a porous silicon oxide layer deposited on the active side of the MCs that have been mechanically manipulated to increase sensitivity. Included is the evaluation of porous silicon oxide present on different geometries of MCs and exposed to varying concentrations of hydrogen fluoride in humid air. Profilometry, in addition to the signal generated by the stress-induced PSO coating and bending of the MC, were used as methods of evaluation.

6.2 Introduction

The versatility of micro-cantilevers (MCs) allows their use in many different applications in sensing, ranging from medical applications, such as disease screening, to the detection of chemical and biological warfare agents.1 One application that has garnered interest has been the use of MCs as a trace gas sensor.2-4 MCs have the ability to detect physical, chemical, or biological changes which results in MC static bending and changes in their resonant vibrational frequency.1,3-6 The MC provides several advantages over other mass sensitive transducers, including sensitivity, low cost, small dimensions, and fast response time.4,7-11

MC sensing can be broken down into two modes of detection, static and dynamic modes

– in this work we used the static mode. In our implementation of the static mode we used an optical readout for the signal transduction and aimed to modify only one of the surfaces of the

MC.6,12 There have been many different modifications to the surface of MCs for the sensing of various chemicals and biologicals such as gold coating for biological sample detection13-16 and palladium coatings for hydrogen gas detection.17-19 Another system of interest for MCs is a silicon oxide coating for HF detection, and previous studies have used silicon oxide coated probes for HF gas detection .20-21 This novel work involves the first exploration into the use of porous silicon oxide (PSO) as a potentially sensitive surface modification for MC sensing applications, in particular for HF gas detection. Interest in the use of PSO as a surface modification includes increased surface area due to the nano-porous nature of the material, and lower cost, as compared to other surface modifications such as gold.22-23 In addition, there is an increased possibility of enhanced analyte absorption due to the nanoscale porosity of the PSO.

This work is being presented to demonstrate the potential of using PSO in combination with mechanically manipulated MCs for sensitive detection of gaseous hydrogen fluoride (HF).

At low concentration, HF gas has been shown in the past to adsorb onto silicon oxide covered surfaces.24-26 Previous work has also shown the adsorption of HF onto silicon oxide coated MCs.20 However, the following work is unique in that the PSO coating is nano-porous

(greater surface area) and the MCs are mechanically manipulated prior to the PSO coating, which results in the manufacture of MCs in a stressed state and with an asymmetrical deposition of PSO (prefedrentially on one side of the MCs). The following work involves a homemade sample cell in which a syringe pump drives dilute HF gas into the cell where the prepared MCs are present. It is important to note that generally the more initial surface stress present on a MC,

90 the higher the sensitivity.27 In most cases, the stress changes in bending due to the HF exposure to the PSO is measured via the movement of a reflected laser beam, which is tracked with a position sensitive detector (PSD). The resulting change in voltage that is recorded by the PSD is then used for comparison of varying levels of contamination and varying concentrations of HF, with the results indicating that PSO as a surface modification on the MC was reproducible and sensitive to HF gas in the range of 30 ppm to 3000 ppm.

6.3 Materials

A sample cell was made in house and covered with a polycarbonate sheet purchased from

Tap Plastics Inc (Stockton, CA). Hydrofluoric acid (48%) was purchased from Sigma-Aldrich

(St. Louis, MO). Headspace HF gas was diluted in a 1 L Restek polypropylene sample bag

(Philadelphia, PA). The “plumbing” consisted of 1/4 inch and 1/8 inch tubing connected with a three way Idex Health & Science shut-off valve made of Tefzel® (Oak Harbor, WA).

6.4 MC Preparation and PSO Deposition

All the silicon MC sets were purchased from MikroMasch (Watsonville, CA), and each set of MCs had a total of sixteen MCs. Each individual MC had the following dimensions: 400

µm long x 100 µm wide x 1 µm thick. Each MC set was mechanically manipulated using another

MC base. For the addition of a reactive layer on the surface, each MC set was exposed to room temperature plasma-enhanced chemical vapor deposition (PECVD) to deposit a 200 nm layer of porous silicon oxide on the top surface of the MCs (System 100 Plasma Deposition Tool, Oxford

Instruments).22-23 For the purposes of reproducibility, the thickness of the PSO on the MCs was observed using scanning electron microscopy (SEM) prior to HF exposure.

6.5 Controlled Exposure of Hydrogen Fluoride Gas to Porous Silicon Oxide

Hydrogen fluoride (HF) gas was collected from the headspace of a 48% solution of hydrofluoric acid and diluted into a 1 L gas sample bag. HF gas was diluted with 90% humid air obtained from the headspace of a deionized water bottle. The entire system was closed after dilution of the HF gas. As seen in Figure 6.5.1, the diluted gas sample was then drawn up through a 100 mL gas-tight syringe, and then pumped at a rate of 20 mL/min into an attached sample cell for a total of 5 minutes of exposure. During exposure of the HF gas to the PSO coated MCs, a Coherent 5 mW 635 nm class 3a laser (Santa Clara, CA) was focused on the MC tip. The optical beam bending as a result of HF gas interacting with the PSO surface was monitored with a PSD and amplified with an On-Trak OT301 precision position sensing amplifier (Irvine, CA). A Tektronix TDS 460A four channel digitizing oscilloscope interpreted the amplified signal (Beaverton, OR). The isolated signal was sent to a Stanford Research

Systems Model SR850 DSP lock-in amplifier to interpret changes in bending (Sunnyvale, CA).

In addition to monitoring static bending of the MCs during exposure, a Wyko NT9800 Optical profilometry system was used for measuring the bending of the MCs before and after HF gas exposure in order to determine any changes in cantilever shape as a result of the change in surface stress. Ellipsometry measurements were taken to evaluate changes in porosity and thickness of the PSO layer before and after HF gas exposure using a spectroscopic imaging ellipsometer (SIE, Beaglehole).

6.6 Evaluation of Shielding and Mechanically Distorting MCs and the Subsequent

Deposition of Porous Silicon Oxide

MCs used for chemical sensing in the static bending mode require that only one side of the cantilever is modified with a responsive phase that has preferential interaction with the target

93 analyte. That interaction can involve adsorption, partitioning, or bioaffinity. The asymmetric nature of the modified MC assures that the analyte induce changes in surface stress is largely on the active side and is not counterbalanced by similar stress changes on the passive side. In many cases an adsorption process serves to reduce surface stress on the active surface and, as such, benefits from a large initial surface stress. These factors contributing to analyte induced signal were enhanced in this work by mechanically manipulating the MC which served the dual functions of shielding the back (passive) side from PSO deposition and leaving the MC is a greater stressed state.

To accomplish this, a MC base was inserted underneath a MC array, which shielded the bottom and mechanically manipulated the MCs such that they bent upward slightly. The slight bending in the MC added to the surface stress and can be seen below in Figures 6.6.1 and 6.6.2.

By depositing asymmetrically, the overall change in sensitivity increases due to the presence of active and passive sides of the MCs. Additional SEM images and profilometry measurements were taken before and after PSO deposition in order to determine the changes in bending caused by the addition of the PSO (Figure 6.6.1c-e). Upon comparison of Figure 6.6.1c and d, the change in bending is evident, especially at the MC tips, where the MC appears to bend back down after PSO deposition. Further examination of these MCs with profilometry (Figure 6.6.1c) revealed a similar outcome.

The native MCs did not have enough initial surface stress and/or assymetry without the use of a shield to be capable of detecting changes in the bending for HF gas exposure below

1000 ppm. This can be attributed to a fairly even deposition of PSO on both sides of the MC, leaving equal opportunity for the HF gas to interact with both sides of the MC. Conversely, when shielded the morphologies (active vs. passive) were as seen in Figure 6.6.1a and b.

Figure 6.6.1: SEM images of the (a) top and (b) bottom of a fully distorted MC in which 200 nm of PSO was deposited. SEM images of (c) the end of a MC with a base underneath and before

200 nm of PSO has been deposited, and (d) the same cantilever after deposition of 200 nm PSO.

In addition, (e) profilometry measurements along the length of the cantilever prior to PSO deposition with the base present, after PSO deposition and before the base is removed, and after the base has been removed.

100 ppm HF. (c) Additionally, a comparison of the voltage changes when 1000 ppm HCl gas is added to 100 ppm HF with fully bent MCs.

The changes in the bending of the MCs is represented by the voltage changes detected by the PSD and recorded by the lock-in amplifier. In addition to mechanically manipulating the

MCs, the HF gas was mixed with 90% humid air, which appeared to increase the sensitivity of the MCs to the HF without interference to the signal (Figure 6.6.2b). As seen in Figure 6.6.2b, there was no noticeable change in the bending for native MCs during exposure to 100 ppm HF gas, however, there was a significant change in the bending after shielding by mechanically manipulating the MC either partially or fully.

There is generally interest in determining if contaminants in the sample would affect the ability to detect HF gas. Hydrogen chloride gas is a possible contaminant in the uranium enrichment process28 and thus a compound of interest. So, a 1000 ppm sample of hydrogen chloride gas was exposed to a fully bent MC with 200 nm of PSO on the surface. The resulting signal did not indicate any significant change in the bending (Figure 6.6.2c). Subsequently, the

1000 ppm HCl gas was mixed with 100 ppm HF gas and exposed to the fully bent MCs (Figure

6.6.2c). The resulting change in signal was compared to previous runs with only HF gas and it was determined that HCl gas had very little effect on the bending signal, indicating that our detection method could be conducted with contaminants such as water vapor and HCl gas present.

6.7 Physical Changes in Bending of MCs After Trace HF Gas Exposure

Mechanically manipulating the MCs prior to PSO deposition is shown to improve sensitivity dramatically by the mechanisms presented above; i.e., leaving the MC in a highly stressed state and asymmetrically modifying with PSO. At high concentrations HF can chemically etch away SiO2. However, at low concentrations it is expected to reversibly adsorb onto the silica surfaces.24-26 Thus it was surprising to find that the sensitivity to HF diminished

97 significantly after the initial exposure (data not shown). In order to observe the physical changes in the MC after HF exposure, optical profilometry was performed. Figure 6.7.1 shows that the bending of the MC after it is exposed and then unexposed to HF is at least partially retained.

That is to say it does not fully return to the pre-HF stress state. Possible explanations include HF induced slippage of the PSO layer along the MC surface or an altered morphology that might be reflected in porosity changes. Retention of some adsorbed HF within the deep nano-pores of the unique PSO layer cannot be ruled out as well. Ellipsometry performed on a PSO surface before and after HF did not reveal a noticeable change in the porosity. However, since the change in

PSO layer volume that would be associated with the ~4 µm difference in the positions of the MC tip (Figure 6.7.1c) would be only ~0.008%, it is likely the sensitivity of the ellipsometry was inadequate to detect such a minute change.

6.8 Analytics of Trace HF Gas Exposure to MCs with a Porous Silicon Oxide Layer

In order to establish the reliability of PSO as a suitable reactive layer for the detection of trace HF gas, a variety of HF gas concentrations were exposed to three different MC sets, each having a 200 nm layer of PSO on the active side. The concentrations ranged from 30 ppm HF to

3000 ppm HF. The slope of the initial drop in the voltage signal was when recorded for triplicate runs and and shown to be reprodiucible (pooled RSD = 8.5%). The slopes were then compared over the range of concentrations (Figure 6.8.1). Analysis of the data points indicated a linear relationship between the concentration of HF gas and the slope of the initial drop in voltage signal (R2 = 0.988). Note that the measurement at every concentration for each fully manipulated MC set was performed on a restored MC set (all PSO was removed with concentrated aqueous HF). With this outcome, PSO appears to be a viable surface modification for the detection of HF gas on MCs.

Figure 6.7.1: Optical profilometry of the MCs showing the changes in the bending before and after exposure to 100 ppm HF gas in 90% humid air for the (a)native, (b)partially distorted, and

(c)fully distorted MCs. Average signal from five MCs from the same set, with three different sets being used for each type of bending.

Figure 6.8.1: Calibration curve of the average slope of the initial drop in voltage signal observed via the PSD using varying concentrations of HF gas ranging from 30 ppm to 3000 ppm.

100

6.9 Conclusions

We demonstrated for the first time the applicability of a nano-porous silicon oxide modified MC to the sensitive detection of HF gas. It is also uniquely shown that mechanically manipulating the extended MC structures can enhance initial stress as well as the asymmetric deposition of responsive phases on opposing sides of the MC, both important to sensitivity.

Linear responses and immunity to certain contaminates was demonstrated. Due to non- reversible changes in MC bending upon HF exposure it is currently necessary to use a new or regenerated system for each measurement. Possible explanations for this behavior were presented but further studies of this surface phenomena are warranted with the eventual goal of creating a MC system that is more conveniently reusable. This work provides insight into a new surface modification used as created, or functionalize though siloxane chemistries, for a variety of MC applications.

6.10 Acknowledgements

The work performed was supported by the Laboratory Director’s Research and

Development Program of Oak Ridge National Laboratory. Part of the research was supported by the Center for Nanophase Materials Sciences, which is sponsored at Oak Ridge National

Laboratory by the Scientific User Facilities Division, Office of Basic Energy Sciences, U.S.

Department of Energy. This work was supported by the United States Department of Energy

(DOE) NA-22. Oak Ridge National Laboratory is operated for the U.S. Department of Energy by UT-Battelle under Contract No. DE-AC05-00OR22725.

101

6.11 References

1. Vahist, S.K. Int. J. Nanotechnol. 2007, 3, 1-15. 2. Barnes, J.R. et.al. Rev. Sci. Instrum. 1994, 65, 3793–3798. 3. Wachter, E.A.; Thundat, T. Rev. Sci. Instrum. 1995, 66, 3662–3667. 4. Long, Z.; Kou, L.; Sepaniak, M.J.; Hou, X. Rev. Anal. Chem. 2013, 32, 135-158. 5. Lavrik, N.V.; Sepaniak, M.J.; Datskos, P.G. Rev. Sci. Instrum. 2004, 75, 2229-2253. 6. Datskos, P.G.; Lavrik, N.V.; Sepaniak, M.J. Smart Sensors and MEMS, Kluwer Academic Publishers, Netherlands, 2005, 331–379. 7. Patton, J.F. et. al. Sens. Actuators, A. 2010, 163, 464-470. 8. Arlett, J.L.; Myers, E.B.; Roukes, M.L. Nat. Nanotechnol. 2011, 6, doi:10.1038/nnano.2011.44. 9. Wachter, E. A.; Thundat, T. Rev. Sci. Instrum. 1995, 66, 3662–3671. 10. Thundat, T.; Wachter, E. A.; Sharp, S. L.; Warmack, R. J. Appl. Phys. Lett. 1995, 66, 1695–1697. 11. Thundat, T.; Chen, G. Y.; Warmack, R. J.; Allison, D. P.; Wachter, E. A. Anal. Chem. 1995, 67, 519–521. 12. Sepaniak, M.J.; Datskos, P.G.; Lavrik, N.V.; Tipple, C. Anal. Chem. 2002, 74, 568A- 575A. 13. Ji, H.F, et. al. Chem. Commun. 2000, 457. 14. Haag, A.; Nagai, Y.; Lennox, R.B.; Grutter, P. J Tech. Instrum. 2015, doi: 10.1140/epjti/s40485-014-0011-5. 15. Sandberg, R. et. al. J. Micromech. Microeng. 2005, 15, 2249. 16. Ji, H.; Hansen, K.M.; Hu, Z.; Thundat, T. Sens. Actuators, B. 2001, 72, 233-238. 17. Patton, J.F.; Hunter, S.R.; Sepaniak, M.J.; Datskos, P.G.; Smith, D.B. Sens. Actuators, A. 2010, 163, 464-470. 18. Patton, J.F. et. al. Nanotechnology. 2012, 23, 465403. 19. Iannuzzi, D. et. al. J. Opt. Soc. Am. B: Opt. Phys. 2006, 1-4. 20. Mertens, J. et. al. Sens. Actuators, B. 2003, 99, 58-65. 21. Chen, I.; Lin, S.; Lin, T.; Du, J. Sensors (Basel). 2011, 11, 1907–1923. 22. Lincoln, D.R.; Lavrik, N.V.; Kravhcenko, I.I.; Sepaniak, M.J. Anal. Chem. 2016, 88, 8741-8748. 23. Charlton, J.J.; Lavrik, N.; Bradshaw, J.A.; Sepaniak, M.J. ACS Appl. Mater. Interfaces. 2014, 6, 17894-17901. 24. Jang, W.I. et. al. J. Micromech. Microeng. 2002, 12, 297-306. 25. Knotter, D.M. J. Am. Chem. Soc. 2000, 122, 4345-4351. 26. Verhaverbeke, S. et. al. J. Electrochem. Soc. 1994, 141, 2852-2857. 27. Lavrik, N.V.; Tipple, C.A.; Sepaniak, M.J.; Datskos, P.G. Biomed. Microdevices. 2001, 3, 35-44.

102

Chapter 7

Concluding Remarks

103

The content of this dissertation included a broad description of three different forms of chemical sensing that involved utilizing nanostructures and nanomechanics for sensitive analyte detection. The discussion began in Chapter 1 with an overview of surface enhanced Raman spectroscopy (SERS) in which the concept and contributions to the enhancement of the spectroscopic method were discussed and applied in the studies found in Chapter 4. The studies involved exploiting the functionalized superhydrophobic surface to concentrate analytes on top of the surface, followed by laser irradiation to monitor the Raman signal. Dilute samples of an anti-cancer drug, mitoxantrone dihydrochloride, was analysed using these platforms, and a limit of detection of 2.9 × 10–12 M was observed.

Additionally, Chapter 2 included a description of thin-layer chromatography and the applications that can be employed using both deterministic and stochastic pillar arrays. Chapters

4 and 5 both included the use of the fabricated ultra-thin layer chromatographic pillar arrays, with Chapter 5 including actual separations conducted on the platform. The studies outlined in

Chapter 5 provided insight into the ability of the deterministic pillar arrays to not only serve as a suitable chromatographic platform with silver colloid present in the pillars, but also allowed for the detection of analytes using SERS with good reproducibility. A separation of 5 purine and pyrimidine bases was able to be observed through SERS. As an added benefit, the silver colloid was shown to have improved stability within the pillar array, due to the superhydrophobic nature of the surface.

A third topic was introduced in Chapter 3 with the presentation of micro-cantilevers. The development and concepts behind the mechanical sensor were discussed, including the two different modes of detection, dynamic and static modes. The concepts from the static mode of detection were applied to the studies in Chapter 6, in which an asymmetric PSO layer was coated

104 on to the MCs. The change in the bending was then observed in real-time as HF has was allowed to flow across the MCs. The study was able to suggest that PSO could serve as a viable sensitive coating for MC sensing of HF gas and perhaps many other applications.

105

Vita

Ryan Wallace was born in Lima, OH on March 29, 1990 to Brian and Karen Wallace.

Most of his childhood was spent in Noblesville, IN, where he attended and graduated high school

(2008) with an Academic Honors Diploma and in the top 10% of his class. He then attended

Manchester University (formerly Manchester College) to begin his work on a B.S. in Chemistry and eventually decided to pursue a PhD in chemistry. With the help of his undergraduate research mentor, Dr. Mark Bryant, Ryan completed his senior research project and graduated from Manchester University with a B.S. in Chemistry with a minor in Biology. Ryan was then recruited by the University of Tennessee’s Chemistry graduate program, where he joined Dr.

Michael J. Sepaniak’s research group. He began working on a full research assistantship in 2015 by collaborating with Dr. Panos G. Datskos’s group at Oak Ridge National Laboratory. In his graduate career, Ryan has published two manuscripts with a third submitted for review.

106