Monodisperse Gold Nanoparticles

0 downloads 12 Views 8MB Size Report
of Chemical Engineering, IISc, especially Prof. Sanjeev Kumar for ... In a similar vein, the technique of drop-casting colloidal solution is extended for tuning the ...
Monodisperse Gold Nanoparticles: Synthesis, Self-assembly and Fabrication of Floating Gate Memory Devices

A thesis submitted for the degree of Doctor of Philosophy in the Faculty of Engineering by

M. Girish

Department of Chemical Engineering Indian Institute of Science Bangalore 560012 (India) November 2013

Declaration

I certify that the thesis was written by me and the Departmental Guidelines were adhered to in preparation of the thesis. In writing the thesis,

1. Experimental data collected obtained by me have been presented without any bias, modifications, or alterations, and can be obtained by others using the information provided in the report. 2. I have not copied material from published/unpublished sources (reports, text books, papers, web sites etc.). I am also aware of the ethical issues involved in writing as per the Departmental Guidelines. 3. Where material from any source was used, the source was given due credit by citing it in the text of the report and giving its details in the section on references, and 4. Where material from any source was copied, it was put in quotation marks, and the source was given due credit by citing it in the text of the report and giving its details in the section on references.

Girish

Acknowledgements I take this opportunity to thank many people who made this thesis possible. First of all, I would like to thank my advisor, Dr. S. Venugopal for the countless hours of discussion and encouragement, he has provided during all stages of this thesis. I benefited greatly from his ideas and insightful conversations during the progress of this project. My sincere thanks are also due to him for obtaining TEM images, for patiently correcting my thesis and for having trust in me when things were not as expected. Next, I would like to thank Prof. Navakanta Bhat for numerous meetings even in his busy schedule. His encouragement during the meetings helped me in venturing out in an entirely new field of electronics. I would also like to thank all the Faculty members of Department of Chemical Engineering, IISc, especially Prof. Sanjeev Kumar for various discussions and group meetings during the course of this project and Prof. K Ganapathy Ayappa (DCC review chair) for monitoring my progress every year. I am highly indebted to Sankar who introduced me to the world of synthesis and assembly of nanoparticles along with training me in most of the nanoscale characterization equipments. I am grateful to all lab mates for maintaining a pleasant atmosphere in the lab. None of the device fabrication would have been possible without the help of CENSE facility technologists in training me to attain fast independent usership. The technologists include, Jayshree, Radha, Raghu and Dr. Savitha for RCA cleaning and dry oxidation, Sabiha for RF sputtering, Pavithra and Reshma for thermal evaporator, Suman for ALD, Soumya for e-beam, Vamsi and Smitha for PECVD, Pavan for ellipsometry, Suma for FIB, Manikant, Sterin, Deepak and Santosh for electrical characterization and finally Bhaskar for optical profilometer. I would also like to thank Mr. Narayan Sharma for assisting me in ellipsometric measurements during understanding of self-assembly of nanoparticles into 2D arrays at the air-water interface. Following people for sophisticated characterizations are also acknowledged, Ravi for Raman spectroscopy, Madukar from RRI for SAXS, Kamalesh from IIT Madras for MALDI measurements, Yashoda for TGA and Tripathi for Photoluminescence spectra. NMR research centre and Institute Surface Science Facility at IISc are acknowledged for NMR and XPS measurements respectively. Special thanks to Sindhuja, Asad and Revathi for their timely help during device fabrication and characterization. Special thanks are also due to my friends here in particular, Sankar, Sivaram, Girija Madam, Vinu, Joshi, Mahendra, Jayshree, Mohanraj, Samrat, Vijay, Iniyan, Pearl, Sharmila, Anand, Sushant, Shravan, Parveen, Deepali, Abhinandan, Ganesh, Arunodai, Sahoo, Aravind, Shivanaresh, Shinde, Ayush, Surya, Sandeep, Srikanth, Bala, Suman, Sainath and all for making my stay at IISc, a very special one. Last but not least, I am ever indebted to my parents and my sister’s family without whom this work would not have been possible.

……..Girish

Abstract The emergence of novel electronic, optical and magnetic properties in ordered twodimensional (2D) nanoparticle ensembles, due to collective dipolar interactions of surface plasmons or excitons or magnetic moments have motivated intense research efforts into fabricating functional nanostructure assemblies. Such functional assemblies (i.e., highly-integrated and addressable) have great potential in terms of device performance and cost benefits. Presently, there is a paradigm shift from lithography based top-down approaches to bottom-up approaches that use selfassembly to engineer addressable architectures from nanoscale building blocks. The objective of this dissertation was to develop appropriate processing tools that can overcome the common challenges faced in fabricating floating gate memory devices using self-assembled 2D metal nanoparticle arrays as charge storage nodes. The salient challenges being to synthesize monodisperse nanoparticles, develop large scale guided self-assembly processes and to integrate with Complementary Metal Oxide Semiconductor (CMOS) memory device fabrication processes, thereby, meeting the targets of International Technology Roadmap for Semiconductors (ITRS) – 2017, for non-volatile memory devices. In the first part of the thesis, a simple and robust process for the formation of waferscale, ordered arrays using dodecanethiol capped gold nanoparticles is reported. Next, the results of ellipsometric measurements to analyze the effect of excess ligand on the self-assembly of dodecanethiol coated gold nanoparticles at the air-water interface are discussed. In a similar vein, the technique of drop-casting colloidal solution is extended for tuning the interparticle spacing in the sub-20 nm regime, by altering the ligand length, through thiol-functionalized polystyrene molecules of different molecular weights. The results of characterization, using the complementary techniques of Atomic Force Microscopy (AFM) and Field-Emission Scanning Electron Microscopy (FESEM), of nanoparticle arrays formed by polystyrene thiol (average molecular weight 20,000 g/mol) grafted gold nanoparticles (7 nm diameter) on three different substrates and also using different solvents is then reported. The substrate interactions were found to affect the interparticle spacing in arrays, changing from 20 nm on silicon to 10 nm on a water surface; whereas, the height of the

resultant thin film was found to be independent of substrate used and to correlate only with the hydrodynamic diameter of the polymer grafted nanoparticle in solution. Also, the mechanical properties of the nanoparticle thin films were found to be significantly altered by such compression of the polymer ligands. Based on the experimental data, the interparticle spacing and packing structure in these 2D arrays, were found to be controlled by the substrate, through modulation of the disjoining pressure in the evaporating thin film (van der Waals interaction); and by the solvent used for drop casting, through modulation of the hydrodynamic diameter. This is the first report on the ability to vary interparticle spacing of metal nanoparticle arrays by tuning substrate interactions alone, while maintaining the same ligand structure. A process to fabricate arrays with square packing based on convective shearing at a liquid surface induced by miscibility of colloidal solution with the substrate is proposed. This obviates the need for complex ligands with spatially directed molecular binding properties. Fabrication of 3D aggregates of polymer-nanoparticle composite by manipulating solvent-ligand interactions is also presented. In flash memory devices, charges are stored in a floating gate separated by a tunneling oxide layer from the channel, and the tunneling oxide thickness is scaled down to minimize power consumption. However, reduction in tunneling oxide thickness has reached a stage where data loss can occur due to random defects in the oxide. Using metal nanoparticles as charge-trapping nodes will minimize the data loss and enhance reliability by compartmentalizing the charge storage. In the second part of the thesis, a scalable and CMOS compatible process for fabricating next-generation, non-volatile, flash memory devices using the self-assembled 2D arrays of gold nanoparticles as charge storage nodes were developed. The salient features of the fabricated devices include: (a) reproducible threshold voltage shifts measured from devices spread over cm2 area, (b) excellent retention (>10 years) and endurance characteristics (>10000 Program/Erase cycles). The removal of ligands coating the metal nanoparticles using mild RF plasma etching was found, based on FESEM characterization as well as electrical measurements, to be critical in maintaining both the ordering of the nanoparticles and charge storage capacity. Results of Electrostatic Force Microscope (EFM) measurements are presented, corroborating the need for ligand removal in obtaining reproducible memory characteristics and reducing vertical charge leakage.

The effect of interparticle spacing on the memory characteristics of the devices was also studied. Interestingly, the arrays with interparticle spacing of the order of nanoparticle diameter (7 nm) gave rise to the largest memory window, in comparison with arrays with smaller (2 nm) or larger interparticle spacing (20 nm). The effect of interparticle spacing and ligand removal on memory characteristics was found to be independent of different top-oxide deposition processes employed in device fabrication, namely, Radio-frequency magnetron sputtering (RF sputtering), Atomic Layer Deposition (ALD) and electron-beam evaporation. In the final part of the thesis, a facile method for transforming polydisperse citrate capped gold nanoparticles into monodisperse gold nanoparticles through the addition of excess polyethylene glycol (PEG) molecules is presented. A systematic study was conducted in order to understand the role of excess ligand (PEG) in enabling size focusing. The size focusing behavior due to PEG coating of nanoparticles was found to be different for different metals. Unlike the digestive ripening process, the presence of PEG was found to be critical, while the thiol functionalization was not needed. Remarkably, the amount of adsorbed carboxylate-PEG mixture was found to play a key role in this process. The stability of the ordered nanoparticle films under vacuum was also reported. The experimental results of particle ripening draw an analogy with the well-established Pechini process for synthesizing metal oxide nanostructures. The ability to directly self-assemble nanoparticles from the aqueous phase in conjunction with the ability to transfer these arrays to any desired substrate using microcontact printing can foster the development of applications ranging from flexible electronics to sensors. Also, this approach in conjunction with roll-to-roll processing approaches such as doctor-blade casting or convective assembly can aid in realizing the goal of large scale nanostructure fabrication without the utilization of organic solvents.

i

TABLE OF CONTENTS

Chapter 1

Introduction ........................................................................ 1

1.1

Introduction to Nanoscience and Nanotechnolgy ........................................... 1

1.2

Fabrication of nanostructures: Top-down vs. Bottom-up ............................... 3

1.3

Nanoparticle assembly: Scale-up and particle arrangement ........................... 5

1.4

Nanoparticle arrays for next-generation, floating gate memory devices ........ 7

1.5

Scope and structure ......................................................................................... 9

Chapter 2

Experimental Methods ..................................................... 11

2.1

Introduction ................................................................................................... 11

2.2

Synthesis of gold nanoparticles ..................................................................... 12

2.3

Fabrication of 2D nanoparticle array ............................................................ 13

2.4

Nanoscale characterization methods ............................................................. 14

2.4.1

Scanning Electron Microscopy (SEM) ...................................................................... 14

2.4.2

Transmission Electron Microscopy (TEM) ............................................................... 18

2.4.3

Focused Ion Beam (FIB)............................................................................................ 18

2.4.4

Atomic Force Microscopy (AFM) ............................................................................. 19

2.4.5

Dynamic Light Scattering (DLS) ............................................................................... 22

2.5

Spectroscopic characterization ...................................................................... 25

2.5.1

UV-Visible spectroscopy (UV-Vis) ........................................................................... 25

2.5.2

X-ray Photoelectron Spectroscopy (XPS).................................................................. 25

2.5.3

Nuclear Magnetic Resonance (NMR) ........................................................................ 27

2.5.4

Small-Angle X-ray Scattering (SAXS) ...................................................................... 28

2.5.5

Raman spectroscopy .................................................................................................. 28

2.5.6

Ellipsometry............................................................................................................... 29

2.5.7

Photoluminescence spectroscopy............................................................................... 34

2.6

Sample cleaning methods .............................................................................. 34

2.6.1

Reactive Ion Etching (RIE) ........................................................................................ 34

2.6.2

Radio Corporation of America (RCA) cleaning ........................................................ 35

2.7

Deposition techniques ................................................................................... 36

2.7.1

Radio-frequency magnetron sputtering (RF sputtering) ............................................ 36

2.7.2

Atomic Layer Deposition (ALD) ............................................................................... 36

2.7.3

Thermal evaporator .................................................................................................... 39

2.7.4

Electron-beam (e-beam) evaporator .......................................................................... 39

ii 2.8

Electrical characterization ............................................................................. 40

2.8.1

2.9

Capacitance-Voltage (CV) measurement .................................................................. 40

Summary ....................................................................................................... 42

Chapter 3 Self-assembly of ligand protected nanoparticles: Scaleup and particle arrangement in 2D arrays .......................................... 43 3.1

Introduction ................................................................................................... 43

3.2

Large-scale self-assembly of 2D nanoparticle arrays - Literature review .... 45

3.2.1

Drop-casting colloidal solutions with excess ligand on a solid substrate .................. 45

3.2.2

Drop-casting colloidal solution on curved or flat water substrate ............................. 45

3.2.3

Doctor-blade/roll-to-roll processing of 2D assembly of particles ............................. 46

3.3

Large-scale self-assembly of gold nanoparticles .......................................... 48

3.3.1

Effect of excess ligand on self-assembly of DDT capped gold nanoparticles .......... 50

3.3.2

Ellipsometric study of formation of 2D nanoparticle arrays ..................................... 54

3.3.2.1 3.3.3

3.4

Modelling of optical constants ................................................................................. 57 Large scale arrays of thiol-terminated polystyrene capped gold nanoparticles ......... 60

Tunable interparticle spacing in nanoparticle arrays..................................... 66

3.4.1

Substrate effects ........................................................................................................ 67

3.4.2

Solvent effect ............................................................................................................ 68

3.4.3

Mechanical properties of the nanoparticle array ....................................................... 73

3.4.4

Mechanism of self-assembly of nanoparticle arrays ................................................. 74

3.4.4.1

3.5

Capillary immersion forces - Background ............................................................... 77

Square packing in nanoparticle arrays .......................................................... 81

3.5.1

Marangoni instability using surface tension gradient (10 nm spacing) ..................... 81

3.5.2

Marangoni instability using temperature gradient (2 nm spacing) ............................ 86

3.6

Formation of nanopouches ............................................................................ 90

3.7

Surface Plasmon Resonance (SPR) in IR region .......................................... 95

3.8

Summary ....................................................................................................... 95

Chapter 4 Fabrication of floating-gate memory devices using gold nanoparticle arrays ................................................................................ 97 4.1

Introduction to memory devices .................................................................... 97

4.2

Working principle of flash memory – Tunneling mechanism ...................... 99

4.3

Limitations of flash memory device............................................................ 102

4.3.1

High design cost ...................................................................................................... 102

4.3.2

Capacitive coupling between floating gates ............................................................ 103

iii 4.3.3

Device reliability: Retention and Endurance ........................................................... 103

4.4

Future of flash memory devices .................................................................. 105

4.5

Device fabrication using RF sputtering ....................................................... 109

4.5.1

Process optimization ................................................................................................ 111

4.5.2

Electrical characterization........................................................................................ 118

4.5.3

Effect of interparticle spacing on memory characteristics ....................................... 123

4.5.4

Effect of plasma treatment on vertical leakage ........................................................ 132

4.6

Device fabrication using atomic layer deposition ....................................... 134

4.6.1

Process optimization ................................................................................................ 134

4.6.2

Quality of Al2O3 deposited at low temperature........................................................ 141

4.6.3

Effect of presence of ligand and interparticle spacing on memory characteristics .. 141

4.6.4 Effect of tunneling oxide thickness and control oxide thickness on memory characteristics .............................................................................................................................. 144 4.6.5

Reduction in accumulation capacitance ................................................................... 144

4.6.6

Effect of plasma treatment on device characteristics ............................................... 148

4.6.7 Possible causes for observed decrease in accumulation capacitance - metal in a dielectric or area screening .......................................................................................................... 148 4.6.8

Leakage characteristics in Al2O3 ............................................................................. 152

4.6.9

Effect of interparticle spacing .................................................................................. 157

4.7

Device fabrication using e-beam evaporation ............................................. 163

4.8

Summary ..................................................................................................... 163

Chapter 5 5.1

PEG capped monodisperse gold nanoparticles ........... 164

Introduction ................................................................................................. 164

5.2 Effect of addition of thiol-functionalized PEG molecules on citrate-stabilized metal nanoparticles ................................................................................................. 165 5.2.1

Gold ......................................................................................................................... 165

5.2.2

Palladium ................................................................................................................. 168

5.2.3

Silver ........................................................................................................................ 169

5.2.4

Platinum ................................................................................................................... 177

5.3

Effect of vacuum on PEG coated gold nanoparticles .................................. 177

5.4

Evolution of particle size distribution in chloroform/water mixture........... 186

5.5

Role of excess PEG-thiol molecules on ripening of gold nanoparticles ..... 188

5.6

Removal of excess PEG molecules from ripened nanoparticles ................. 188

5.7

Effect of molecular weight of PEG-thiol .................................................... 199

5.8

Absence of gold thiolates in PEG-thiol coated nanoparticles ..................... 200

iv 5.9

DLS photon counts ...................................................................................... 206

5.10 PEG-citrate complexation ........................................................................... 207 5.11 Role of PEG backbone ................................................................................ 213 5.12 Role of bound carboxylate ion .................................................................... 214 5.12.1

Tannic acid method ................................................................................................. 214

5.12.2

Ascorbic acid method .............................................................................................. 214

5.12.3

Ethylene glycol protocol ......................................................................................... 214

5.12.4

Sodium acrylate protocol ........................................................................................ 219

5.13 Phase-space of ripening ............................................................................... 219 5.14 Pechini process ............................................................................................ 220 5.15 Pseudocrown ethers ..................................................................................... 231 5.16 Interaction between ligand and metal.......................................................... 234 5.17 Direct aqueous self-assembly of nanoparticle ............................................. 235 5.18 Thermal stability of PEG capped gold nanoparticles .................................. 235 5.19 Summary ..................................................................................................... 239

Chapter 6 6.1

Summary and Scope for Future Work ........................ 240

Summary ..................................................................................................... 240

6.1.1

Large-scale self-assembly of nanoparticles ............................................................. 241

6.1.2 Scalable processes for fabricating non-volatile memory devices using self-assembled gold nanoparticles as charge storage nodes ................................................................................ 242 6.1.3

6.2

PEG capped monodisperse gold nanoparticles ....................................................... 244

Thesis contributions .................................................................................... 245

6.2.1

Peer-reviewed articles ............................................................................................. 245

6.2.2

Conference presentations ........................................................................................ 245

6.3

Scope for future work .................................................................................. 246

6.3.1

Large-scale self-assembly of nanoparticles ............................................................. 246

6.3.2

Fabrication of flash memory devices ...................................................................... 247

6.3.3

Aqueous-phase synthesis of monodisperse metal nanoparticles ............................. 247

Chapter 7

References ....................................................................... 248

Appendix A: Self-assembled FePt nanoparticles for bit patterned magnetic media……………………………..……...…………………………..………273 Appendix B: Calculations……………..……………………………..……..282

v

LIST OF FIGURES Figure 1.1: Effect of platinum particle size on selectivity for furfural decarbonylation to furan and furfuryl alcohol (Adapted with permission from Pushkarev et al.6. Copyright (2012) American Chemical Society). ........................................................... 2 Figure 1.2: Demonstration of polyethylene glycol (PEG) tethered Gold-Cadmium telluride nanoparticles as nanoscale thermometer (adapted with permission from Lee et al.7 Copyright (2005) John Wiley and Sons). Plot of time variation of (a) photoluminescent intensity (E) and (b) corresponding temperature variation, as a function of time, for composite nanoparticles. .............................................................. 2 Figure 1.3: Illustration of Top-down vs. Bottom-up approach using the real life examples of Chariot temple in Hampi18 and Egyptian pyramid19 respectively. Today, transistors20 are fabricated using top-down approach, which faces both fundamental and economic challenges in further scaling of devices. An attractive alternative way for fabricating nanostructures in sub-20 nm regime is bottom-up approach, which involve brick-by-by assembly of molecules or particles21............................................. 4 Figure 1.4: Illustration of the outline of this thesis ........................................................ 6 Figure 2.1: Schematic representation of monolayer formation and transfer printing developed by Santhanam and Andres64. Key fabrication steps include (a) adjusting the curvature of water surface using Teflon cell, (b) drop-casting concentrated (~ 1014 particles/mL) gold nanoparticles dispersed in organic solvent on curved water surface, (c) Langmuir-Schaefer transfer of nanoparticle array from water surface to PDMS pad, and (d) Microcontact printing of gold nanoparticle array and transfer to desired substrate. ...................................................................................................................... 14 Figure 2.2: (a) Schematic representation of position of different detectors in the Gemini column of ULTRA 55 and (b) schematic representation of functioning of filtering grid in separating secondary and backscattered electrons (reproduced from Zeiss69 with permission) .............................................................................................. 15 Figure 2.3: Example of image processing using Clemex Vision PE software, a) binary image, b) defining process frame c) the thresholded image along with an outline of the

vi threshold, (d) the original image with the outline based on thresholding and (e) histogram of circular diameter estimated within the process frame. ........................... 17 Figure 2.4: Representative FESEM images taken during sample preparation for crosssection imaging using Focused Ion Beam (FIB). The figures represent (a) platinum deposition on substrate, (b) silicon etching near the deposited platinum, (c) etching of platinum to aid omniprobe welding, (d) omniprobe welding to the lamella, (e) lamella removal, (f) transfer welding of lamella attached to the omniprobe onto a TEM grid using platinum deposition, (g) etching of omniprobe tip and (h) thinning down of lamella section to a thickness of 100 nm, so as to make it electron transparent. ......... 21 Figure 2.5: Illustration of estimation of (a) linear background (L(E)) and (b) Shirley background (S(E)) for titanium 2p XPS spectrum (reproduced from reference76). ..... 27 Figure 2.6: Illustration of estimation of background spectra using IWTA algorithm (reproduced from Galloway et al.78 with permission). (a) Superposition of original signal and the 7th level approximation after single iteration. (b) Superposition of new approximation curve and the chopped original curve at the approximation values in a. (c) Superposition of original signal and the successive approximation curves after multiple iterations, as stated in a and b. ....................................................................... 29 Figure 2.7: Schematic representation of light incident on the sample surface and getting reflected, adapted from reference81. ................................................................. 30 Figure 2.8: Schematic representation of ellipsometric wave in linear coordinate. ...... 32 Figure 2.9: Schematic representation of incident light getting reflected and transmitted on a three layer structure, with substrate, thin film and air as layers81. ....................... 32 Figure 2.10: Schematic representation of RF magnetron sputtering process used for depositing dielectric on top of gold nanoparticle array. The thickness and the quality of the film sputtered is found to depend on four parameters, namely, Argon pressure, applied voltage between target and substrate, substrate to target distance and deposition time. ............................................................................................................ 37 Figure 2.11: (a) Digital photograph of Anelva RF magnetron sputtering unit in Centre for Nano Science and Engineering (CeNSE), IISc. Digital photographs of (b) substrate holder and (c) target, in Anelva RF sputtering unit. ..................................... 37

vii Figure 2.12: Schematic representation of the first cycle of Atomic Layer Deposition (ALD) process of Al2O3, comprising of four steps namely, (i) exposure of aluminum precursor to silicon surface, (ii) reaction of TMA with OH and methane formation (iii) monolayer adsorption of TMA and (iv) formation of new hydroxyl group and oxygen bridges; adapted from reference85. ............................................................................... 38 Figure 2.13: Schematic representation of layer by layer growth of Al2O3 using ALD process, adapted from reference85. ............................................................................... 38 Figure 2.14: (a) Illustration of three states of an ideal p-MOS capacitor depending on the gate voltage, namely, (i) accumulation (negative gate voltage), (ii) depletion (close to zero gate voltage) and (iii) inversion (positive gate voltage). (b) Schematic representation of Capacitance-Voltage (CV) characteristics of a p-MOS capacitor, as the gate voltage is swept from inversion to accumulation. .......................................... 41 Figure 3.1: (a) Transmission electron micrograph of monolayer of dodecanethiol capped gold nanoparticles produced by drop-casting gold colloid solution (containing excess dodecanethiol) dispersed in toluene onto silicon nitride substrate. The top inset represents schematic of dodecanethiol capped gold nanoparticles. The bottom inset represents fast Fourier transform of the image highlighting good order. (b) Optical photograph of compact monolayer of gold nanoparticles formed on silicon nitride substrate. (Reproduced with permission from Bigioni et al.28 Copyright (2006) Nature Publishing Group). ....................................................................................................... 47 Figure 3.2: (a) Optical photograph of gold nanoparticle array floating on water surface and inking the PDMS stamp pad with nanoparticle array self-assembled at the curved air-water interface (Reproduced with permission from Santhanam and Andres64. Copyright

(2003)

American

Chemical

Society).

(b)

Optical

image

of

nanoparticle/poly (methylmethacrylate) film transferred to glass; the arrow represents the free-standing film (Reproduced with permission from Pang et al.94. Copyright (2008) American Chemical Society).(c) Schematic representation of monolayer formation by drop-casting dodecanethiol capped gold nanoparticles dispersed in hexane onto toluene surface (Reproduced with permission from Martin et al.96. Copyright (2010) American Chemical Society and (d) Schematic representation of array formation wherein the assembly at the air-toluene interface is transported onto

viii an air-water interface (Reproduced with permission from Eah31. Copyright (2011) Royal Society of Chemistry). ....................................................................................... 47 Figure 3.3: Schematic representation of large scale assembly of nanoparticles through doctor-blade casting (Reproduced with permission from Bodnarchuk et al.95 Copyright (2010) American Chemical Society). ......................................................... 48 Figure

3.4:

(a)

Digital

photograph

of

gold

nanoparticles

dispersed

in

chloroform/hexane mixture on curved water surface in rectangular Teflon cell (2x10 cm). The outlined region, dotted square box used to highlight the presence of monolayer gold nanoparticle film on the water surface while the arrow indicates the particle lost to the Teflon walls. (b-c) Representative FESEM image of sub-monolayer gold nanoparticle array formed using the rectangular Teflon cell at different magnifications. Increasing the concentration should yield monolayer gold nanoparticle array......................................................................................................... 49 Figure 3.5: Representative FESEM images showing polymer/solvent dewetting occurring while floating thiol-terminated polystyrene (molecular weight: 3000) capped gold nanoparticles on the water surface at (a) low nanoparticle concentration (< 1014 particles/ml) and (b) high particle concentrations (> 1014 particles/ml). At low particle concentration, parts of arrays due to residual stresses, portion of arrays flip on top of neighbouring domains as shown in the outlined region in a (dotted circle) while at high concentrations, multilayer is formed with larger holes (bright regions in b, corresponds to holes). .................................................................................................. 49 Figure 3.6: (a) Digital photograph of dodecanethiol capped gold nanoparticle array formed by drop casting gold nanoparticles dispersed in toluene without excess dodecanethiol molecules. The outlined regions (two circles) in a are used to highlight the occurrence of multiple nucleation events during monolayer formation.(b) Dualmagnification FESEM image of dodecanethiol capped gold nanoparticle array formed by drop casting gold nanoparticles dispersed in toluene without excess dodecanethiol molecules. .................................................................................................................... 51 Figure 3.7: (a) Digital photograph of dodecanethiol capped gold nanoparticle array formed by drop casting gold nanoparticles dispersed in toluene with 5 µL excess dodecanethiol molecules. It can be seen that the entire petridish is filled with gold

ix nanoparticle array. (b) Dual-magnification FESEM image of dodecanethiol capped gold nanoparticle array formed by drop casting gold nanoparticles dispersed in toluene with 5 µL excess dodecanethiol molecules. .................................................... 52 Figure 3.8: Representative FESEM image of monolayer gold nanoparticle array formed by spreading toluene-hexane mixture, with SPAN 20 (sorbitan monolaurate) as surfactant on the water surface in a petridish at (a) low and (b) high magnifications. ...................................................................................................................................... 53 Figure 3.9: Digital photograph of general ellipsometer set-up. Light emitted from the source, first passes through the polarizer (P), and the compensator (C) before it is reflected at the sample (S). The polarizer converts the unpolarized light into linearly polarized light. After reflection the light again passes a linear polarizer denoted the analyzer (A) before it reaches the detector. ................................................................. 55 Figure 3.10: Optical photograph during self-assembly of dodecanethiol gold nanoparticle array without addition of excess dodecanethiol to the colloidal solution at different times after drop-casting of colloidal solution (a and b). The initial patterns represent island formation, and even after 90 minutes, arrays with microscopic holes are formed. The scale bar represents 50 µm. The regions marked 1 represent the array. ...................................................................................................................................... 56 Figure 3.11: Optical photograph during self-assembly of dodecanethiol gold nanoparticle array with the addition of excess dodecanethiol to the colloidal solution at different times after drop-cast of colloidal solution (a and b). The initial patterns represent labyrinthine-like structures, enabling large scale array formation. The scale bar in a represents 50 µm, and the same magnification is used for obtaining Fig. b. The regions marked 1 represent the array. The two concentric boxes (white line in a) represent the region of interest (ROI), where the ellipsometric data is measured. ...... 56 Figure 3.12: Representative screen shots of EP4 model fit of ellipsometric data obtained after 15 minutes of drop-cast of dodecanethiol capped gold nanoparticles on the water surface containing, (a) without excess thiol and (b) with excess thiol to the colloidal solution.......................................................................................................... 59

x Figure 3.13: Evolution of film thickness obtained using ellipsometry during selfassembly of dodecanethiol capped gold nanoparticles for both with and without addition of excess dodecanethiol. The exact thickness was found to depend on the model; however, the absolute difference of initial film thickness was observed in both cases. The points represent the mean of experimental data and the error bars represent 95 % confidence interval (n=4).................................................................................... 59 Figure 3.14: Representative FESEM image of thiol-terminated polystyrene (molecular weight: 20000) capped gold nanoparticles dispersed in toluene (with 5 µL dodecanethiol), floated on the water surface. The distance between particles is approximately 2 nm, corresponding to dodecanethiol chain length and not thiolterminated polystyrene (molecular weight: 20000). This suggests that small molecule (DDT) had replaced long chain molecule (PSSH) within the time scale of minutes. Nevertheless, it can be seen that polymer/solvent dewetting is minimized. ................ 63 Figure 3.15: Dual-magnification FESEM image of thiol-terminated polystyrene (molecular weight: 20000) capped gold nanoparticles dispersed in toluene floated on the water surface modified with 5 µL of dodecanethiol (Scale bar: 2 µm). The right side image represents the zoomed portion of the small box on the left side. .............. 64 Figure 3.16: (a) FESEM image of thiol-terminated polystyrene (molecular weight: 20000) capped gold nanoparticles dispersed in toluene drop-cast on Si/SiO2 substrate (Scale bar: 20 µm). The bright regions represent bare silicon substrate while the dark regions correspond to array. It can be clearly seen that polymer/solvent dewetting is completely negated. (b) Dual-magnification FESEM image of thiol-terminated polystyrene (molecular weight: 20000) capped gold nanoparticles dispersed in toluene drop casted on Si/SiO2 substrate (Scale bar: 1 µm). The right side image represents the zoomed portion of the small box on the left side. Most of the places, it can be seen on the right hand side of the image b, multilayer is formed. ....................................... 64 Figure 3.17: (a) Low magnification FESEM image of thiol-terminated polystyrene (molecular weight: 20000) capped gold nanoparticles dispersed in toluene drop-casted on 10 minutes ozone treated Si/SiO2 substrate (Scale bar: 100 µm). The two regions marked 1 and 2 represent array and bare silicon substrate respectively. The outlined region (circle) is attributed to the blind spot of the Inlens detector used for detecting

xi secondary electrons. It can be clearly seen that polymer/solvent dewetting is completely negated. (b) High magnification FESEM image of thiol-terminated polystyrene (molecular weight: 20000) capped gold nanoparticles dispersed in toluene drop-cast on 10 minutes ozone treated Si/SiO2 substrate (Scale bar: 200 nm). ........... 65 Figure 3.18: (a) Digital photograph of drop-cast of thiol-terminated polystyrene capped gold nanoparticles dispersed in toluene on flat water surface in a petridish. (b) Low magnification FESEM image of the self-assembled 2D arrays formed by dropcasting gold nanoparticles dispersed in toluene onto a silicon substrate. The outlined region (dotted circle) in b is the blind spot of the in-lens detector used for detecting secondary electrons. ..................................................................................................... 66 Figure 3.19: Representative FESEM images of self-assembled 2D arrays of thiolfunctionalized polystyrene grafted gold nanoparticles (a–f) and their corresponding interparticle spacing distributions (g). These 2D arrays were formed by drop-casting gold nanoparticles from: a,b) toluene solution onto a UVO treated silicon substrate (with native oxide), c,d) toluene solution onto freshly cleaved graphite, e,f) toluene solution onto water. ...................................................................................................... 69 Figure 3.20: AFM cross-section profiles across arrays formed by drop-casting gold nanoparticles from: a) toluene solution onto a UVO treated silicon substrate (with native oxide), b) toluene solution onto freshly cleaved graphite, c) toluene solution onto water. (d) Schematic representation of AFM cross-section profile across nanoparticle array. AFM histogram map of the thickness profile of the array over the entire area for colloidal solution drop-cast onto a (e) silicon substrate (with native oxide), (f) freshly cleaved graphite and (g) water. The respective thicknesses are 21.7 nm, 19.3 nm and 23.7 nm, which are in agreement with the AFM line thickness values of 23.0 nm, 21.0 nm and 23.0 nm. (h) DLS size histogram of thiol-terminated polystyrene capped gold nanoparticles dispersed in toluene. ...................................... 70 Figure 3.21: Representative FESEM images of self-assembled 2D arrays of thiolfunctionalized polystyrene grafted gold nanoparticles (a–f) and their corresponding interparticle spacing distributions (g). These 2D arrays were formed by drop-casting gold nanoparticles from: a,b) tetrahydrofuran solution onto silicon (with native oxide), c,d) 50% (v/v) chloroform/cyclohexane solution onto water, and e,f) toluene solution

xii onto a water surface that was a priori modified with a thin film of thiol-functionalized polystyrene. .................................................................................................................. 71 Figure 3.22: AFM cross-section profiles across arrays formed by drop-casting gold nanoparticles

from

a)

THF

solution

onto

silicon,

b)

50%

(v/v)

chloroform/cyclohexane solution onto water, and c) toluene solution onto a water surface that was a priori modified with a thin film of thiol-functionalized polystyrene. AFM histogram map of the thickness profile of the array over the entire area for colloidal solution drop-cast using d) THF solution onto silicon, e) 50% (v/v) chloroform/cyclohexane solution onto water, and f) toluene solution onto a water surface that was a priori modified with a thin film of thiol-functionalized polystyrene The respective thicknesses are 14.0 nm, 27.9 nm and 20.4 nm, which are in agreement with the AFM line thickness values of 13.4 nm, 29.1 nm and 22.0 nm. DLS size histogram of thiol-terminated polystyrene capped gold nanoparticles dispersed in (g) THF and (h) 50% (v/v) chloroform/cyclohexane. ....................................................... 72 Figure 3.23: Illustration summarizing the substrate effect on interparticle spacing in 2D arrays formed by drop-casting a colloidal solution of thiol-functionalized polystyrene grafted gold nanoparticles. ....................................................................... 72 Figure 3.24: Representative force-displacement curves measured on top of nanoparticle films formed by drop casting a thiol-functionalized polystyrene grafted gold nanoparticles in toluene onto: a) silicon (with native oxide), b) freshly cleaved graphite, c) water, d) water surface that was a priori modified with a thin film of thiol-functionalized polystyrene (average molecular weight: 3000 g/mol), (e) silicon substrate and (f) ‘silicon’ region presumed to be polystyrene thiol coated, due to transfer printing from the water surface modified a priori with polystyrene thiol molecules. .................................................................................................................... 75 Figure 3.25: Representative AFM phase scans, from the edges of self-assembled 2D arrays formed by drop-casting gold nanoparticles from (a) toluene solution onto a water surface that was a priori modified with a thin film of thiol-functionalized polystyrene, (b) tetrahydrofuran solution onto silicon (with native oxide). The phase profiles along the length of the corresponding straight lines are shown as overlays.

xiii Negligible phase difference in (a) clearly suggests the transfer of polystyrene thiol molecules in addition to array from the water surface. ................................................ 76 Figure 3.26: (a) Digital photograph recorded during the self-assembly of gold nanoparticles on the water surface, with partial evaporation of the solvent. The region between the dotted and solid circle represent the region of partially dried film. (b) FESEM image of the nanoparticle thin film formed by transfer printing prior to complete evaporation of solvent from a colloidal solution spread on water. The dashed curve marks the putative evaporating front. The inset shows a magnified view of the arrangement of particles around a hole. (c) Optical photograph illustrating “coffee stain effect”, image reproduced from reference136. ......................................... 76 Figure 3.27: Schematic representation of attractive capillary immersion forces during self-assembly of thiol-terminated polystyrene capped gold nanoparticles. ................. 78 Figure3.28: Schematic representation of attractive or repulsive capillary immersion and floatation forces depending on the meniscus slope and contact line (adapted from Kralchevsky et al.137). .................................................................................................. 78 Figure 3.29: Plot of capillary interaction energy, ΔW and particle size, R, at fixed interparticle distance of L = 2R, for capillary immersion and floatation forces (Reproduced with permission from Kralchevsky et al.137. Copyright (1994) IOP Publishing). .................................................................................................................. 79 Figure 3.30: Representative FESEM image of nanoparticles forming a 2D square configuration obtained by using a 50% (v/v) solution of ethanol and water as the substrate at different magnifications; (a) low magnification image highlighting the formation of convective cells (bright regions correspond to nanoparticle multilayers with hexagonal order) and (b) high magnification image demonstrating the formation of a square lattice; the inset shows Fourier transform of the main image. .................. 83 Figure 3.31: Representative FESEM image of nanoparticles forming a disordered configuration by using a 1: 2 solution of ethanol and water as the substrate. ............. 83 Figure 3.32: (a) Fluorescent microscopic images of self-assembled hexagonal and striped patterns of polystyrene nanoparticles obtained by controlling the evaporation of solvent and inducing Marangoni flow of ethanol into water. (b) Schematic

xiv illustration of the nanoparticle self-assembly process highlighting the process of water condensation and generation of surface tension gradient, thereby driving the nanoparticles towards the receding contact line (Reproduced with permission from Cai and Newby143. Copyright (2010) Springer). .......................................................... 85 Figure

3.33:

Schematic

illustration

poly(methylmethacrylate)-b-polystyrene

of

poly(ethyleneoxide)-b-

(PEO-b-PMMA-b-PS)

ABC

triblock

copolymer film solvent annealing with controlled humidity. (b) Scanning Force Micrograph of square domains of 117 nm thick PEO-b-PMMA-b-PS films after annealing for 16 h under saturated benzene vapor at relative humidity of ~ 80-90 %. The top inset shows the high magnification image of the square pattern while the bottom inset represents the corresponding Fourier transform (Reproduced with permission from Tang et al.144. Copyright (2008) American Chemical Society). ....... 85 Figure 3.34: Representative FESEM image of nanoparticles forming a 2D square configuration obtained by spin coating gold nanoparticles on Si/SiO2 substrate. ....... 87 Figure 3.35: Digital photograph during self-assembly of dodecanethiol capped gold nanoparticles at the air water interface. The colloidal solution contains excess thiol and maintained at ~ 15 °C before drop casting on the water surface. The water surface was maintained at room temperature (24 °C). ............................................................. 87 Figure3.36: Representative FESEM images obtained by transfer printing the arrays fabricated by self-assembly of dodecanethiol capped gold nanoparticles at the air water interface, shown in Fig. 3.35. (a) Low magnification image highlighting the circular regions (dotted circles) and regions marked 1 and 2 corresponding to array and bare silicon substrate respectively. (b-d) High magnification images obtained by zooming in circular regions shown in a. The dotted rectangles represent domains of square lattice with 2.2 nm interparticle spacing. (e) High magnification image obtained by zooming in region marked 1 in a, illustrating hexagonal packing of particles. ....................................................................................................................... 88 Figure 3.37: (a) Optical photograph of the nanoparticle assembly created through Marangoni instability with circular patches in between the large scale monolayer. (b) Representative line scan across the circular region (marked dark line in a), indicating a thickness of ~ 14 nm, corresponding to a bilayer thickness. .................................... 89

xv Figure 3.38: (a-b) Representative FESEM images taken at different magnifications, of the polystyrene thiol-gold nanoparticle nanostructures fabricated through drop-casting polystyrene thiol capped gold nanoparticles (molecular weight: 20000 g/mol) dispersed in a solvent mixture comprising equal volumes of tetrahydrofuran and water on Si substrate. The images in b were taken using two detectors, namely, angle selective backscattered electron (left) and secondary electron (right) detector. High contrast in backscattered electrons is primarily due to the variation in atomic mass while the contrast in secondary electrons is due to surface features. ........................... 91 Figure 3.39: (a-b) Representative FESEM images taken at different magnifications, of the polystyrene thiol-gold nanoparticle nanostructures fabricated through drop-casting polystyrene thiol capped gold nanoparticles (molecular weight: 3000 g/mol) dispersed in tetrahydrofuran on Si substrate. The images in b were taken using two detectors, namely, angle selective backscattered electron (left) and secondary electron (right) detector. ........................................................................................................................ 92 Figure 3.40:(a-b) Representative FESEM images taken at different magnifications, of the polystyrene thiol-gold nanoparticle nanostructures fabricated through drop-casting polystyrene thiol capped gold nanoparticles (molecular weight: 20000 g/mol) dispersed in toluene on the water surface spread with dodecane layer. The images in b were taken using two detectors, namely, angle selective backscattered electron (left) and secondary electron (right) detector........................................................................ 93 Figure 3.41: UV-Visible spectra of polystyrene thiol capped gold nanoparticle in THF mixed with varying amounts of water. Remarkably, at higher water contents in the colloidal solution, nanoparticle-polystyrene assemblies have SPR in the IR region. .. 94 Figure 3.42: (a) Representative FESEM image of the polystyrene thiol-gold nanoparticle nanostructures fabricated through drop-casting polystyrene thiol capped gold nanoparticles (molecular weight: 3000 g/mol) dispersed in THF and water (with 20% volume).The images were taken using two detectors, namely, angle selective backscattered Representative

electron FESEM

(left)

and

image

of

secondary electron the

polystyrene

(right) thiol-gold

detector.

(b)

nanoparticle

nanostructures fabricated through drop-casting polystyrene thiol capped gold

xvi nanoparticles (molecular weight: 3000 g/mol) dispersed in THF and water (with 50% volume). ....................................................................................................................... 94 Figure 4.1: (a) Estimate of share of non-volatile memory market in comparison with a total market (adapted from Electronics, CA publications154). (b) The evolution of flash memory market in comparison with DRAM market over the last 6 years. For the first time in 2012, the flash memory business has exceeded DRAM market, adapted from IC Insights153)............................................................................................................... 98 Figure 4.2: Schematic representation of analogous behavior of flow valve and Field Effect Transistor (FET). The extent of opening of flow valve determines the liquid flow rate from inlet to outlet, analogous to the gate voltage determining the extent of conduction channel formation between source and drain. ........................................... 98 Figure 4.3: (a) Schematic representation of field-effect floating gate transistor.(b) Each field effect transistor (FET) is connected using bit lines, word lines and source lines for program, erase and/or read operation (adapted from reference157). (c) Schematic representation of trapping/detrapping of electrons by controlling the gate voltage. Depending on the presence or absence of electrons in a floating gate, memory states are assigned as 0 or 1. ...................................................................................... 100 Figure 4.4: Schematic energy-level representation of Fowler-Nordheim tunneling in a MOS capacitor156. ...................................................................................................... 101 Figure 4.5: Schematic representation of definition of “Pitch”. “Pitch” refers to the centre-centre distance between two FETs.................................................................. 101 Figure 4.6: (a) Estimation of increase of design cost with a decrease in feature size, adapted from reference158. (b) Comparison of design cost with required revenue as the feature size decreases, for successful commercialization, adapted from reference159. .................................................................................................................................... 101 Figure 4.7: Flash technology scaling for the last two decades, adapted from Intel164. .................................................................................................................................... 104 Figure 4.8: Plot showing requirement of floating gate height for the effective suppression of gate coupling based on design rule (adapted from Kim et al.161). ..... 104

xvii Figure 4.9: (a) Schematic representation of energy band diagram of MOS capacitor (p-Si).(b) Schematic representation of energy band diagram of floating gate MOS capacitor.(c) Schematic representation of energy band diagram for (i) thick tunneling oxide highlighting negligible electron tunneling probability from floating gate to substrate. Electron trapping is necessary for reliable memory storage device. (ii) thin tunneling oxide highlighting higher electron tunneling probability. This will result in data loss. (iii) Thin metallic film sandwiched between substrate and control oxide. Metallic film offers deep potential well, thereby minimizing electron tunneling probability in comparison to semiconductor. This will aid in scaling down the tunneling oxide thickness aggressively, even down to ~ 3 nm, with higher data reliability. ................................................................................................................... 106 Figure 4.10: Schematic representation of charge retention characteristics in the presence of defects for (a) conventional floating gate and (b) nanoparticle floating gate memory. In conventional memory, a small leakage would cause all the stored electrons (data) to tunnel back to the substrate, while a few defects in a nanoparticle memory would not lead to complete loss of data due to charge compartmentalization. .................................................................................................................................... 106 Figure 4.11: (a) Schematic representation of formation of nanoparticles using annealing process. (b) Scanning electron micrograph of platinum nanoparticles obtained by annealing a thin film at 950 °C for 10 s; indicating large polydispersity (inset). Reproduced with permission from Lee et al.60. Copyright (2005) Springer. 107 Figure 4.12: (a) Representative FESEM image of gold nanoparticle array formed by spin coating (Reproduced with permission from Leuw et al177. Copyright (2010) John Wiley and Sons). (b) Representative FESEM image of gold nanoparticle array after annealing the sample at 500 °C (Reproduced with permission from Leuw et al177. Copyright (2010) John Wiley and Sons). .................................................................. 107 Figure 4.13: Plot showing the band gap vs. dielectric constant for typical dielectric materials (Reproduced with permission from Robertson192. Copyright (2000) AIP Publishing). ................................................................................................................ 110 Figure 4.14: Schematic representation of process flow sheet for fabrication of nanoparticle floating-gate memory device. Microcontact printing of gold nanoparticle

xviii array on 10 nm SiO2/silicon wafer was done followed by gadolinium oxide deposition using RF sputtering. Aluminium contacts were deposited for device testing............ 110 Figure 4.15: Representative FESEM image of dodecanethiol capped gold nanoparticle array embedded in gadolinium oxide, which was deposited under standard conditions; particle coalescence and loss of hexagonal ordering is observed. ............................. 112 Figure 4.16: Schematic of the RF magnetron sputtering process for gadolinium oxide deposition, indicating the various parameters that were optimized in this study. ..... 112 Figure 4.17: Representative high magnification cross-sectional FESEM images of the MOS capacitor with gold nanoparticle arrays as charge storage nodes, obtained using (a) angle selective backscattered electron (AsB) and (b) secondary electron (SE) detector. Contrast in backscattered electron imaging is predominantly due to the difference in atomic weight (Z) while secondary electron imaging depends on surface features. Here, the sample contains both gold (Z = 197) and gadolinium oxide (Z = 363.49/5=72), hence gold nanoparticles are seen clearly in AsB and not using SE detector. The edge of the sample appears as a very faint line (indicated by the arrow labelled gadolinium oxide) in (a), while it appears very bright in (b) due to the predominance of edge effects in SE imaging. A 2-D profile plot of the grayscale values across the MOS capacitor (inset in a) indicates the thickness of array to be ~ 7 nm, based on full width at half maximum. (c) Low magnification cross-sectional FESEM image of the MOS capacitor, obtained using angle selective backscattered electron detector indicating large scale uniformity. The dense packing of the array precludes visualization of individual nanoparticles. These samples were obtained by cleaving the silicon substrate. .................................................................................... 115 Figure 4.18: (a) Schematic illustrating the expected cross-section of the device before and after RF sputter deposition, indicating that most of the ligand molecules have been removed from the ligand layer after gadolinium deposition. (b) AFM crosssectional height profiles across the edge of the nanoparticle array before and after gadolinium oxide deposition by RF sputtering. A decrease in the difference of height from 8.2 nm to 6.6 nm indicates that most of the ligands have been removed during the RF sputtering process. .......................................................................................... 116

xix Figure 4.19: Representative FESEM image of MOS capacitor fabricated with a 30 nm thick gadolinium oxide layer, indicating poor lateral resolution, which hinders the visualization of the in-plane ordering of the nanoparticle array. ............................... 116 Figure 4.20: (a) Capacitance-Voltage (CV) curve of a MOS capacitor with the gold nanoparticle array as floating gate, obtained by bi-directional sweep from inversion to accumulation. The appearance of a hysteresis loop indicates that gold nanoparticles act as charge storage nodes. The capacitor area is 7.07 x 10-4 cm2, and the measurement frequency used was 1 MHz. The inset shows an optical photograph of several MOS capacitor devices. The white dots are the aluminium top contacts. (b) Capacitance-voltage characteristics of Si/SiO2/Gd2O3 (without any gold nanoparticle). The control samples do not show any hysteresis with bidirectional voltage sweeps, indicating the absence of any interface traps.(c) Histogram of flatband voltage shifts measured from 14 devices using P/E voltages of ±7 V. The distribution of flatband voltage shifts is 0.66 ± 0.05 V. (d) Variation of flatband voltage shifts as a function of P/E voltages. The points represent the average of flatband voltage shifts and the error bars represent 95% confidence interval (n≥5). .......................................................... 120 Figure 4.21:(a) Representative charge retention characteristics of MOS capacitors physically spread across ~ cm2 area, indicating large scale uniformity of nanoparticle array. P/E was carried out at a voltage of ± 7 V, while reading voltage was set at -0.8 V. (b) Schematic energy band diagram depicting flatband condition of the floating gate structure comprising of gold nanoparticles sandwiched between silicon dioxide and gadolinium oxide. (c) EFM phase change profiles obtained at a reading voltage of 3 V and at a height of 80 nm from the surface. Charge was injected initially, by applying -6 V from 0 to 120 s, in contact mode at the spot marked “X” in AFM topography scan (top left inset). EFM phase change image (after 150 s) is shown in the top right inset, with a line marking the cross-section along which phase change profiles were measured. (d) Comparison of normalized charge stored obtained using capacitance and EFM measurements. (e) Endurance characteristic of a device indicating a stable memory window even after 104 P/E cycles (by biasing at ± 7 V). .................................................................................................................................... 121

xx Figure 4.22: (a) Au 4f core level XPS spectrum from a MOS capacitor with embedded nanoparticle array indicating the presence of gold in metallic form (84.1 and 87.7 eV). XPS spectra of (b) Gadolinium 3d and 4d levels, and (c) Oxygen 1s level indicate the predominant presence of hydroxide (532.4 eV), possibly as gadolinium hydroxide (Gd(OH)3) and is attributed to the hygroscopic nature of rare earth oxides. XPS spectra of (d) Carbon 1s and (e) Sulphur 2p core levels for MOS capacitors fabricated with and without (control) gold nanoparticle arrays embedded in gadolinium oxide. For MOS capacitors with the embedded nanoparticle array, the presence of C=O (288.7 eV) and trace amount of sulphur (162.7 eV) indicates that some residual ligands are trapped inside. C=O bonds are expected to have formed during the annealing steps by decomposition of the trapped ligands. In these figures, the symbols represent measured data points, the black lines represent deconvoluted peaks, and the red lines represent the overall fit. The numbers in the panels represent the peak values of the fitted curves. .................................................................................................... 122 Figure 4.23: Representative FESEM image of close-packed 2D array of thiolterminated polystyrene capped gold nanoparticles (nominally 7 nm sized gold cores) with (a) 4.4 ± 0.6 nm interparticle spacing, corresponding to a number density of 7.7x1011 particles/cm2 and (b)13.3 ± 2.3 nm interparticle spacing, corresponding to a number density of 2.4x1011 particles/cm2, on Si/SiO2 prior to gadolinium oxide deposition. The insets show the corresponding histograms of interparticle spacing in these arrays................................................................................................................. 124 Figure 4.24: (a) Representative plan-view FESEM images of samples obtained after deposition of a 15 nm thick gadolinium oxide layer onto a 2D array of thiolterminated polystyrene capped gold nanoparticles (nominally 7 nm sized gold cores) with 4.4 nm interparticle spacing, followed by thermal annealing at 500 °C. The left hand side image was obtained using angle selective backscattered electrons, which provide information on the atomic number of the material, and the right hand side image was obtained using secondary electrons, which provide information on the surface topography. Scale bar represents 20 nm. (b)Representative plan-view FESEM images (using secondary electron detector) of samples obtained after deposition of a 15 nm thick gadolinium oxide layer onto 2D array of thiol-terminated polystyrene

xxi capped gold nanoparticles (nominally 7 nm sized gold cores) with 13 nm interparticle spacing, followed by thermal annealing at 500 °C. Scale bar represents100 nm. ..... 125 Figure 4.25: Capacitance-voltage characteristics of MOS capacitors fabricated using arrays of gold nanoparticles capped with thiol-terminated polystyrene of molecular weight (a) 3000 g/mol and (b) 20000 g/mol, embedded in gadolinium oxide on Si/SiO2. The samples do not exhibit any hysteresis despite the presence of gold nanoparticles that can serve as charge storage nodes. This could be possible only if thiol-terminated polystyrene molecules can be converted to carbonaceous material during annealing......................................................................................................... 127 Figure 4.26: (a) Representative EFM phase profile of MOS capacitors fabricated using arrays of gold nanoparticles capped with thiol-terminated polystyrene of molecular weight 3000 g/mol embedded in gadolinium oxide on Si/SiO2. The units of x and y axis correspond to µm. Charge was injected initially by applying -6 V from 2 minutes while the reading voltage was fixed at 3 V at a height of 70 nm from the surface. The color coded scale bar represents the magnitude of phase change observed during EFM imaging. The EFM measurements are carried out in a standard mode (see schematic). (b) Schematic representation of modified EFM measurement with an intermediate step of contacting the tip with sample before reading. (c) EFM phase profile of modified procedure for imaging MOS capacitor incorporating gold nanoparticle arrays. Negligible presence of charge stored after reading indicates leakage paths for electrons to detrap through the top-oxide. ..................................... 127 Figure 4.27: Schematic representation of modified process flow sheet for nanoparticle floating-gate memory device fabrication, with an additional step of RF plasma etching to fabricate bare nanoparticle array (plasma treated array) under optimized conditions. All other steps remain identical. ................................................................................ 128 Figure 4.28: Representative EFM phase profiles of charge injection on oxygen plasma treated Si/SiO2 substrate after (a) 1 min and (b) 5 min. Plasma conditions: power - 20 W, flow rate - 50 sccm, and time - 2 minutes. Charge was injected initially by applying -8 V from 1 minute while the reading voltage was fixed at 3 V at a height of 70 nm from the surface. The color coded scale bar represents the magnitude of phase change observed during EFM imaging. The images indicate negligible charge storage

xxii after 5 minutes. (c) Representative EFM phase profile of charge injection on oxygen plasma treated Si/SiO2 substrate, followed by forming gas annealing at 450 °C for 30 minutes. Retention of phase change in the programmed region (white region) indicates elimination of interface trap sites in the silicon dioxide film. .................... 128 Figure 4.29: Representative plan-view FESEM images of samples obtained after deposition of a 15 nm thick gadolinium oxide layer, followed by thermal annealing at 500 °C, onto plasma treated gold nanoparticle array with (a) 4.4 nm interparticle spacing and (b) 13.3 nm interparticle spacing. The polystyrene ligands were removed by RF plasma etching under optimized conditions. The left hand side image (in a) was obtained using angle selective backscattered electrons, which provide information on the atomic number of the material, and the right hand side image was obtained using secondary electrons, which provide information on the surface topography. ........... 129 Figure 4.30: (a) Capacitance–voltage curve of a MOS capacitor obtained by bidirectional sweep from inversion to accumulation with plasma treated gold nanoparticle array with 4.4 nm interparticle spacing as floating gate. (b) Histogram of flatband voltage shifts measured from 8 different devices. (c) Capacitance–voltage curve of a MOS capacitor obtained by bi-directional sweep from inversion to accumulation with plasma treated gold nanoparticle array with 13.3 nm interparticle spacing as floating gate. (d) Capacitance–voltage curve of a MOS capacitor obtained by bi-directional sweep from inversion to accumulation with plasma treated gold nanoparticle array with 2 nm interparticle spacing as floating gate. ......................... 131 Figure 4.31: Comparison of EFM phase profiles of standard and modified procedure for imaging MOS capacitors incorporating untreated and plasma treated gold nanoparticle arrays. The units of x and y axis correspond to µm. Charge was injected initially by applying -6 V from 2 minutes while the reading voltage was fixed at 3 V at a height of 70 nm from the surface. The color coded scale bar represents the magnitude of phase change observed during EFM imaging. ..................................... 133 Figure 4.32: Representative charge retention characteristics of MOS capacitors fabricated with 2D arrays with different interparticle spacing of 2 nm (Au-DDT), 4.4 nm (Au-PSSH (3K)) and 13.3 nm (Au-PSSH (20K)). P/E was carried out at a voltage of ± 7 V, while reading voltage was set at -0.8 V for DDT capped array and -1.5 V for

xxiii polystyrene thiol capped gold nanoparticle arrays. The hole decay was found to be significant for arrays with larger interparticle spacing. ............................................. 133 Figure 4.33: Representative FESEM images of dodecanethiol capped gold nanoparticle arrays obtained after deposition of 18 nm of Al2O3 using ALD process at 250 °C at (a) low and (b) high magnification. Note the images were taken after annealing the sample in a forming gas atmosphere at 450 °C for 20 minutes. .......... 135 Figure 4.34: Representative FESEM image of thiol-terminated polystyrene capped gold nanoparticle arrays (molecular weight: 20000 g/mol) obtained after deposition of 18 nm of Al2O3 using ALD process at 250 °C, (a) without ligand removal and (b) with ligand removal using oxygen plasma treatment, followed by annealing of the sample in a forming gas atmosphere at 450 °C for 20 minutes. ............................................ 137 Figure 4.35: Representative FESEM images of dodecanethiol capped gold nanoparticle arrays obtained after deposition of Al2O3 using ALD process at 150 °C, (a) without ligand removal and (b) with ligand removal using hydrogen plasma treatment followed by annealing of the sample in a forming gas atmosphere at 450 °C for 20 minutes. ........................................................................................................... 138 Figure 4.36: Representative FESEM images dodecanethiol capped gold nanoparticle arrays obtained after deposition of Al2O3 using ALD process at 135 °C, without ligand removal (a) and with ligand removal using hydrogen plasma treatment (b). Note, the images were taken after annealing of the sample in forming gas atmosphere at 450 °C for 20 minutes. Representative FESEM image of thiol-terminated polystyrene capped gold nanoparticle arrays (molecular weight: 20000 g/mol) obtained after deposition of Al2O3 using ALD process at 135 °C, without ligand removal (c) and with ligand removal using oxygen plasma treatment (d). Note, the images were taken after annealing of the sample in forming gas atmosphere at 450 °C for 20 minutes. (e) High magnification cross-sectional FESEM image of the MOS capacitor with arrays fabricated using dodecanethiol capped gold nanoparticles as charge storage nodes. The dense packing of the array precludes visualization of individual nanoparticles. Platinum is present due to FIB sample preparation. (f)High magnification cross-sectional FESEM image of the MOS capacitor obtained using (i) backscattered electron and (ii) secondary electron detector. Contrast in backscattered

xxiv electron imaging is predominantly due to the difference in atomic weight while contrast in secondary electron imaging depends on surface features. ....................... 139 Figure 4.37: (a) Capacitance-Voltage (CV) curve of a MOS capacitor without the gold nanoparticle array as floating gate obtained by depositing Al2O3 using ALD process at 135°C, obtained by bi-directional sweep from inversion to accumulation. The appearance of a hysteresis loop indicates the presence of interface traps in significant numbers. (b)Capacitance-Voltage (CV) curve of a MOS capacitor with the 2D arrays using dodecanethiol capped gold nanoparticles as floating gate obtained by depositing Al2O3 using ALD process at 135°C, obtained by bi-directional sweep from inversion to accumulation. The appearance of a hysteresis loop indicates a lot of interface traps. .................................................................................................................................... 140 Figure 4.38: Capacitance-Voltage (CV) curve of a MOS capacitor without the gold nanoparticle array as floating gate obtained by depositing Al2O3 using ALD process at 375°C, obtained by bi-directional sweep from inversion to accumulation. The appearance of a no hysteresis loop indicates high quality oxide without any trap charges. ...................................................................................................................... 140 Figure 4.39: (a) Schematic of the device structure of MOS capacitor with Si/SiO2/Al2O3. (b) Capacitance-Voltage (CV) curve of a MOS capacitor without the gold nanoparticle array as floating gate obtained by depositing Al2O3 using ALD process at 135°C followed by immediately annealing the sample at 400°C under nitrogen ambient for 20 minutes inside the ALD reactor. The absence of a hysteresis loop indicates the formation of high quality oxide without any trap charges. ........... 140 Figure 4.40:Capacitance-Voltage (CV) curve of a MOS capacitor fabricated using bare gold nanoparticle arrays with (a) 2 nm interparticle spacing (DDT) and (b) 20 nm interparticle spacing (PSSH, 20000 g/mol) as floating gate. The devices were fabricated by depositing Al2O3 using ALD process at 135°C, followed by nitrogen annealing at 400°C for 20 minutes. The appearance of a hysteresis loop for devices fabricated using plasma treated nanoparticle array indicates the storage of electrons inside gold nanoparticles while the untreated array sample (c) 2 nm interparticle spacing (DDT capped gold nanoparticles) and (d) 20 nm interparticle spacing (PSSH

xxv capped gold nanoparticles) shows continues trapping/detrapping of charges due to the possible formation of carbonaceous product during annealing. ................................ 143 Figure 4.41: Capacitance-Voltage (CV) curve of a MOS capacitor with the 2D bare gold nanoparticle arrays with (a) 2 nm interparticle spacing (DDT capped gold nanoparticles) and (b) 20 nm interparticle spacing (PSSH capped gold nanoparticles) as floating gate. The devices were fabricated by depositing Al2O3 using ALD process at 135°C, followed by nitrogen annealing at 400°C for 20 minutes. The appearance of a hysteresis loop for devices fabricated using plasma treated nanoparticle array sample indicates the storage of electrons inside gold nanoparticles while the devices fabricated using untreated array with (c) 2 nm interparticle spacing and (d) 20 nm interparticle spacing shows continues trapping/detrapping of charges due to the possible formation of carbonaceous product during annealing. (e) CV curve of MOS capacitor without gold nanoparticle array. (f) Schematic of the device structure of MOS capacitor. .......................................................................................................... 145 Figure 4.42: Capacitance-Voltage (CV) curve of a MOS capacitor without the gold nanoparticle array subjected to the RF-plasma treatment as necessary to fabricate bare nanoparticle arrays at (a) 2 nm and (b) 20 nm spacing as floating gate. The absence of hysteresis loop and no significant reduction in capacitance at accumulation indicate plasma treatment does not cause substrate damage, similar to CV curve of a MOS capacitor without gold nanoparticle array, not subjected to any plasma treatment (c). (d) CV curve of a MOS capacitor fabricated with bare arrays with 20 nm spacing as floating gate shows significant reduction in accumulation capacitance. This experiment clearly highlights the mild plasma treatment does not cause substrate damage and the hysteresis and decrease in accumulation capacitance arise due to the presence of bare nanoparticle arrays.(e) Schematic of the device structure of MOS capacitor. .................................................................................................................... 150 Figure 4.43: Current-Voltage (IV) characteristic of MOS capacitor fabricated with (i) SiO2 and Al2O3 (MOS blank), (ii) untreated and (iii) bare 2D arrays of nanoparticle with 20 nm spacing. This experiment suggests mild plasma treatment used for fabricating bare arrays does not affect the leakage current. ....................................... 151

xxvi Figure 4.44: (a-b) Representative FESEM images of gold porous film of thickness 5.6 nm at different magnifications. (c) Representative FESEM image of Al2O3 deposited on gold porous film, followed by annealing the sample at 450 °C for 20 minutes, under forming gas atmosphere. .................................................................................. 153 Figure 4.45: Current voltage characteristic of MIM structure (with gold porous filmAl2O3-Aluminium architecture), showing high leakage currents (~ µA) in addition to very low breakdown voltage of ~ 1V. Note the substrate was annealed at 450 °C for 20 minutes in forming gas environment. ................................................................... 154 Figure 4.46: Current voltage characteristic of MOS structure (with Aluminium-SiAl2O3-Aluminium) with ALD process carried out at 375 °C. Even in this structure, very high leakage current and low breakdown voltage was observed. ...................... 154 Figure 4.47: (a) Capacitance-Voltage (CV) curve of a MOS capacitor with pure tunnelling oxide structure. The MOS capacitor shows no significant hysteresis, indicating negligible traps in the oxide layer. (b)Capacitance-Voltage (CV) curve of a MOS capacitor with tunnelling oxide, SiO2 and control oxide, Al2O3 structure. The MOS capacitor shows no significant hysteresis, indicating negligible traps in the oxide layer. Capacitance-Voltage (CV) curve of a MOS capacitor with the 2D(c) untreated array and (d) bare gold nanoparticle array (with 2 nm spacing) as floating gate. The devices were fabricated by depositing Al2O3 using ALD process at 135°C, followed by nitrogen annealing at 400°C for 20 minutes. The sample was subsequently annealed under forming gas environment at 850 °C for 20 minutes. The hysteresis loop is observed for devices fabricated using bare nanoparticle array (0.9 V) and not for untreated array. This can be attributed to the possible conversion of entrapped organic ligand into carbonaceous product during annealing. CapacitanceVoltage (CV) curve of a MOS capacitor with the 2D (e) untreated array and (f) bare gold nanoparticle array (with 20 nm spacing) as floating gate. The sample was subsequently annealed under forming gas environment at 850 °C for 20 minutes. The hysteresis loop is observed for devices fabricated using bare nanoparticle array (0.9 V) and not for untreated array. This can be attributed to the possible conversion of entrapped organic ligand into carbonaceous product during annealing. .................... 155

xxvii Figure 4.48: Representative FESEM images of (a) untreated and (b) bare nanoparticle array with 2 nm spacing (DDT capped gold nanoparticles), after ALD deposition of Al2O3 followed by forming gas annealing at 850°C for 20 minutes. Representative FESEM images of (c) untreated and (d) bare nanoparticles with 20 nm spacing (using thiol-terminated polystyrene capped gold nanoparticles of molecular weight: 20000 g/mol), after ALD deposition of Al2O3 followed by forming gas annealing at 850°C for 20 minutes. In all the figures, left and right images were obtained using backscattered electron and secondary electron detector respectively. ....................... 156 Figure 4.49: Representative FESEM image of 2D arrays of gold nanoparticles capped with (a) dodecanethiol, (b) thiol-terminated polystyrene (molecular weight: 7000 g/mol), (c) thiol-terminated polystyrene (molecular weight: 10000 g/mol) and (d) thiol-terminated polystyrene (molecular weight: 20000 g/mol). The respective histograms of interparticle spacing are presented in Fig. e-h. ................................... 158 Figure 4.50: Representative FESEM image of bare gold nanoparticle array with interparticle spacing of (a) 2.2 nm (dodecanethiol capped gold nanoparticle array), (b) 7.9 nm (thiol-terminated polystyrene (molecular weight: 7000 g/mol) capped gold nanoparticle array), (c) 8.9 nm (thiol-terminated polystyrene (molecular weight: 10000 g/mol) capped gold nanoparticle array), and (d) 19.7 nm (thiol-terminated polystyrene (molecular weight: 20000 g/mol) capped gold nanoparticle array); after plasma treatment under optimized conditions. .......................................................... 159 Figure 4.51: Capacitance-Voltage (CV) curve of a MOS capacitor with the bare nanoparticle array with interparticle spacing of (a) 2.2 nm (dodecanethiol capped gold nanoparticle array), (b) 7.9 nm (thiol-terminated polystyrene (molecular weight: 7000 g/mol) capped gold nanoparticle array), (c) 8.9 nm (thiol-terminated polystyrene (molecular weight: 10000 g/mol) capped gold nanoparticle array), and (d) 19.7 nm (thiol-terminated polystyrene

(molecular

weight:

20000 g/mol) capped gold

nanoparticle array); after plasma treatment under optimized conditions. The devices were fabricated by depositing Al2O3 using ALD process at 135°C, followed by nitrogen annealing at 400°C for 20 minutes. The sample was subsequently annealed under forming gas environment at 850 °C for 20 minutes. The largest memory window was obtained for arrays prepared using thiol-terminated polystyrene

xxviii molecules of molecular weight of 10000 g/mol, an intermittent spacing. The memory window followed the trend of 8.9 nm (5.2 V)> 7.9 nm (4.5 V)> 2.2 nm (2.9 V)> 19.7 nm (0.8 V) interparticle spacing. Values in the bracket represent the memory window. .................................................................................................................................... 160 Figure 4.52:Capacitance-Voltage (CV) curve of a MOS capacitor with the interparticle spacing of (a) 2.2 nm (dodecanethiol capped gold nanoparticle array), (b) 7.9 nm (thiol-terminated polystyrene (molecular weight: 7000 g/mol) capped gold nanoparticle array), (c) 8.9 nm (thiol-terminated polystyrene (molecular weight: 10000 g/mol) capped gold nanoparticle array), and (d) 19.7 nm (thiol-terminated polystyrene (molecular weight: 20000 g/mol) capped gold nanoparticle array); after plasma treatment under optimized conditions. The devices were fabricated by depositing Al2O3 using e-beam process at room temperature. The sample was subsequently annealed under forming gas environment at 450 °C for 20 minutes. The largest memory window was obtained for arrays prepared using thiol-terminated polystyrene molecules of molecular weight of 10000 g/mol, an intermittent spacing, similar to earlier ALD process. The memory window followed the trend of 8.9 nm (1.5 V)> 7.9 nm (0.6 V)~ 2.2 nm (0.6 V)~ 19.7 nm (0.6 V) interparticle spacing. Values in the bracket represent the memory window. ............................................... 161 Figure 5.1: Representative FESEM image of (a) citrate capped gold nanoparticles (b) thiol-functionalized polyethylene glycol capped gold nanoparticles and their respective size histograms (c and d). The molar ratio of PEG-thiol to gold is 10:1 and the PEG-thiol capped gold nanoparticles were drop-cast on silicon substrate after ageing the colloidal solution for ~16 hours at 25 °C. ................................................ 166 Figure 5.2: Thermogravimetric analysis (TGA) of citrate capped and PEG-thiol capped gold nanoparticles. ......................................................................................... 167 Figure 5.3: Representative DLS size histograms of (a) citrate capped and (b) PEGthiol capped gold nanoparticles. The molar ratio of PEG-thiol to gold is 10:1 and the PEG-thiol capped gold nanoparticles were analyzed after ageing the colloidal solution for ~16 hours at 25 °C. ............................................................................................... 167 Figure 5.4: UV-Vis spectrum of a palladium chloride solution that was boiled with citrate solution. ........................................................................................................... 170

xxix Figure 5.5: Digital photographs taken during the process of synthesizing citratecapped palladium nanoparticles, (a) at the start of the reaction and (b) after boiling for 1 hour. ........................................................................................................................ 170 Figure 5.6: Representative FESEM image of (a) citrate capped palladium nanoparticles and (b) aged PEG-thiol coated gold nanoparticles after 9 days. (c) UVVis spectra of palladium nanoparticles before and after addition of PEG-thiol. (d) Digital photograph of citrate capped palladium nanoparticles before (left) and after addition of PEG-thiol (right). 5µL of PEG-thiol was added to 5 mL of colloidal solution and the PEG-thiol capped gold nanoparticles were drop-cast on silicon substrate after ageing the colloidal solution for 9 days at 74 °C. .............................. 171 Figure 5.7: X-ray photoemission spectra of PEG-thiol capped palladium nanoparticles, after ripening....................................................................................... 171 Figure 5.8: Representative FESEM image of aged PEG-thiol capped palladium nanoparticles after 15 days from particle ripening. ................................................... 172 Figure 5.9: Representative FESEM images of citrate capped silver nanoparticles, after ageing at different magnifications. ............................................................................ 173 Figure 5.10: Representative FESEM image of PEG-thiol capped silver nanoparticles, after ageing at 25 °C for 1 day. .................................................................................. 174 Figure 5.11: UV-Vis spectra of (a) citrate capped silver nanoparticles and (b) aged PEG-thiol capped silver nanoparticles after 1 day at 25 °C. ..................................... 174 Figure 5.12: Photoluminescence spectra of citrate capped silver nanoparticles and PEG-thiol capped silver nanoparticles, excited at 320 nm. Upon continued excitation, the intensity of the PEG-thiol capped sample decreases, which is attributed to photobleaching. The dotted lines represent the repeated measurement of the corresponding sample without altering the parameters. ............................................ 176 Figure 5.13: Representative FESEM image of citrate capped platinum nanoparticles .................................................................................................................................... 176 Figure 5.14: Representative FESEM image of PEG-thiol coated platinum nanoparticles, after ageing for 12 days at different magnifications. .......................... 179

xxx Figure 5.15: X-ray photoemission spectra of PEG-thiol capped platinum nanoparticles, after ageing for 12 days ...................................................................... 179 Figure 5.16: (a) Representative TEM image of PEG-thiol capped gold nanoparticles and (b) corresponding size histogram. The molar ratio of PEG-thiol to gold was 10 and the PEG-thiol capped gold nanoparticles were drop-cast on TEM grid after ageing the colloidal solution for ~1 day at 25 °C. ................................................................. 179 Figure 5.17: Representative TEM image during imaging thiol-functionalized PEG capped gold nanoparticles, highlighting thick organic film in the shape of open-ended wrench. ....................................................................................................................... 180 Figure 5.18: (a) Representative TEM image of thiol-functionalized PEG capped gold nanoparticles near organic film and (b) respective size histogram. ........................... 180 Figure 5.19: TEM images of thiol-functionalized PEG capped gold nanoparticles, highlighting low polydispersity, obtained at different locations. Images show particles with two contrasts, suggesting possible presence of organic film. The particles inside the region marked with a dotted oval (b) were found to be more spherical than the particles found away from this region ....................................................................... 181 Figure 5.20: Representative TEM image of thiol-functionalized PEG capped gold nanoparticles, highlighting particles assembled in ordered and disordered fashion, next to each other. ...................................................................................................... 182 Figure 5.21: Representative TEM images of PEG-thiol-functionalized gold nanoparticles obtained at (a) low and (b) high magnification. These sub-2 nm clusters were found to increase in size during the course of imaging. .................................... 182 Figure 5.22: Representative FESEM images of thiol-functionalized PEG capped gold nanoparticles at different magnifications before (a-c) and after treating the sample in TEM chamber for 1 hour, without switching on the electron beam (d-f). The regions marked 1 and 2 (in a) represent nanoparticle array and substrate respectively. Clearly, the higher vacuum in the TEM chamber has affected the size distribution of the particles. ..................................................................................................................... 183 Figure 5.23: Representative FESEM images by drop-cast of either (a) concentrated PEG-thiol capped gold nanoparticle solution (~ 20 times the original concentration) or

xxxi (b) freeze-drying the standard colloidal solution was found to result in formation of supracrystals, a 3D assembly of PEG-thiol coated gold nanoparticles. The dotted rectangles (in a) represent Moiré patterns due to two ordered pattern, namely, ordered array and SEM scan line. The regions marked 1 and 2 represent monolayer array (shown in d) and supracrystal (shown in c). .............................................................. 184 Figure 5.24: (a-f) Representative FESEM images of thiol-functionalized PEG capped gold nanoparticles at different magnifications after leaving the sample imaged in Fig. 5.23c inside the FESEM chamber for 1 day. The dotted rectangles highlight regions of different contrast due to the possible partial removal of PEG molecules under vacuum. Clearly, the transformation of supracrystals into linear assemblies, as well as changes in size distribution can only be attributed to the effect of vacuum on the organic film. ............................................................................................................... 185 Figure 5.25: Representative FESEM images of (a) PEG-thiol coated gold nanoparticles, (b) PEG-thiol coated gold nanoparticles (5 mL) after addition of small amounts of chloroform (0.2 mL), (c) boiled PEG-thiol coated gold nanoparticles dispersed in water and chloroform, and (d) the colloidal solution after being cooled back to room temperature. ......................................................................................... 187 Figure 5.26: Representative FESEM images of PEG-thiol coated gold nanoparticles with the ratio of PEG-thiol to Au, (a) 0.15, (b) 1, (c) 10, and (d-f) their respective size histograms. PEG-thiol coated gold nanoparticles were drop-cast on silicon substrate, after ageing the colloidal solution for 1 day at 25 °C. ............................................... 190 Figure 5.27: Representative low magnification FESEM images of PEG-thiol coated gold nanoparticles with the molar ratio of PEG-thiol to Au of, (a) 0.15, (b) 1, (c) 10, highlighting Moiré patterns. The dotted rectangles (in b) represent regions of Moiré pattern. Moiré patterns can appear due to the interference between two regular patterns. Here, SEM scan lines act as reference pattern, while the ordered nanoparticle array forms the second pattern. PEG-thiol capped gold nanoparticles were drop-cast on silicon substrate after ageing the colloidal solution for 1 day at 25 °C. ............... 192 Figure 5.28: Representative low-magnification FESEM image of PEG-thiol coated gold nanoparticles mixed with excess hydrogen peroxide and sodium citrate and

xxxii boiled for 2 minutes. The region marked 1, 2 and 3 correspond to circular domains of 2D nanoparticle array, organic film and bare silicon respectively. ........................... 193 Figure 5.29: (a) Representative FESEM images of PEG-thiol coated gold nanoparticles mixed with excess hydrogen peroxide and sodium citrate after boiling for 2 minutes, showing regions of (b) highly disordered nanoparticle array (region marked 2 in a), (c and d) highly ordered nanoparticle array. The left hand side and right hand side images in a and b were obtained using backscattered electron and secondary electron detector respectively. (c) High magnification FESEM image of circular regions reveal highly ordered array with their corresponding size histogram (d). .............................................................................................................................. 194 Figure 5.30: Representative FESEM image of PEG-thiol coated gold nanoparticles mixed with excess hydrogen peroxide and sodium citrate after boiling for 2 minutes. .................................................................................................................................... 194 Figure 5.31: Representative high-magnification FESEM images, obtained using both secondary electron and backscattered electron detector, of boiled PEG-thiol coated gold nanoparticles mixed with excess hydrogen peroxide and sodium citrate at different locations; (a) region of organic PEG film (as region marked by the oval in secondary electron detector was not translated onto the backscattered electron image) and (b) region of gold-rich complex (as regions marked with circles in secondary electron image also appear as very bright spots in backscattered electron image). ... 195 Figure 5.32:(a) Representative FESEM image of thiol-functionalized PEG capped gold nanoparticles boiled with a mixture of hydrogen peroxide and sodium citrate after boiling for 20 minutes, suggesting fusion of nanoparticles due to removal of ligands and the (b) representative DLS size histogram. ............................................ 196 Figure 5.33: (a) Representative FESEM images of (a) PEG-thiol coated gold nanoparticles derived from 5 nm citrate capped gold nanoparticles62 and (b) PEG-thiol coated gold nanoparticles derived from ~ 40 nm citrate capped gold nanoparticles (Frens method68). ....................................................................................................... 196 Figure 5.34: Representative FESEM image of ~ 40 nm citrate capped gold nanoparticles after changes in pH through the addition of sodium citrate (such that the

xxxiii final molar ratio of citrate to gold reached 5.2), highlighting ripening process through the formation of small sized particles using two detectors, namely, backscattered electron (left) and secondary electron (right) detectors. ............................................ 198 Figure 5.35: Representative FESEM image of citrate capped gold nanoparticles after immediate addition of PEG-thiol molecules of molecular weight 5000 g/mol; highlighting possible ripening process through the formation of etching of larger particles (dotted rectangle). ........................................................................................ 198 Figure 5.36: Representative FESEM images of ripened gold nanoparticles using PEGthiol molecules having molecular weight of (a-b) 1000 g/mol, and (c-d) 5000 g/mol. The insets in b and d represent respective size histograms. DLS measurements of PEG-thiol coated gold nanoparticles with and without boiling for different molecular weights, (e) 1000 g/mol, and (f) 5000 g/mol. The molar ratio of PEG-thiol to gold was maintained 10 for both the cases and the sample was boiled for 10 minutes before analysis. ...................................................................................................................... 201 Figure 5.37: (a) Representative FESEM images of addition of PEG-thiol 5000 molecules to ripened particles using PEG-thiol 356. (b) Representative FESEM images of addition of PEG-thiol 356 molecules to ripened particles using PEG-thiol 5000. The insets represent respective size histograms............................................... 202 Figure 5.38: Representative FESEM images of ripened gold nanoparticles using PEGthiol molecules of molecular weight 20000 g/mol at (a) low and (b) high magnification. The molar ratio of PEG to gold is 10 and the sample was aged for 1 day at 25 °C. The regions marked 1 and 2 (in a) represent close-packed and non-close packed arrays respectively. ........................................................................................ 203 Figure 5.39: Low m/z range, MALDI spectra of PEG-thiol capped gold nanoparticles ionized both negatively and positively, at different ratios of PEG-thiol to gold of 0.15 and 10 plotted with different grid lines (a-f). The molecular weight of Sinapic acid (matrix), PEG-thiol, Au1-PEG-thiol is 224, 356 and 553 respectively. It can be seen that peaks coincide well with the Sinapic acid and not with gold thiolates or any complex. ..................................................................................................................... 204

xxxiv Figure 5.40: High m/z range, MALDI spectra of PEG-thiol capped gold nanoparticles ionized both negatively and positively, at different ratios of PEG-thiol to gold of 0.15 and 10. ........................................................................................................................ 205 Figure 5.41: (a) Time-dependent average DLS photon count intensity after addition of PEG-thiol (PEG/Au: 10) and aged the sample at 74 °C. Extraction of time constant (~ 4.1 hours) based on the time-dependent average DLS photon count intensity at 74 °C. There is a sudden drop in DLS photon count intensity from 40 to 30 kcps after the addition of PEG-thiol to citrate capped gold nanoparticles. (b) Time-dependent average DLS photon count intensity with respect to temperature, after addition of PEG-thiol (PEG/Au: 10). It can also be that there is a significant batch to batch variation. .................................................................................................................... 205 Figure 5.42: (a) Time-dependent average DLS photon count intensity with respect to temperature for the same batch of citrate capped gold nanoparticles, after addition of PEG-thiol (PEG/Au: 10). It can also be seen that the photon count intensity of citrate capped gold nanoparticles at 25 °C does not increase much while at 74 °C, the increase is attributed to the fusion of particles........................................................... 208 Figure 5.43: Small-Angle X-ray Scattering (SAXS) spectra of citrate capped and PEG-thiol capped gold nanoparticles. As the intensity at low q values remains the same, the amount of gold contained in the two samples can be considered equal. ... 208 Figure 5.44: 1H Nuclear Magnetic Resonance (NMR) of PEG-thiol dispersed in D2O. The numbers embedded with the NMR data correspond to the respective numbers marked in PEG molecular structure (as inset).The characteristic peak values at 2.7 (SH, labelled 1), 2.9 (-S-CH2, labelled 2), 3.3 (-OCH3, labelled 3), 3.5 and 3.6 (hydrogen from ethylene glycol chain, labelled 4 and 5) ppm. ................................. 210 Figure 5.45: (a) 1H Nuclear Magnetic Resonance (NMR) of citrate capped gold nanoparticles dispersed in D2O. The number embedded with the NMR data corresponds to the respective numbers marked in PEG molecular structure (as inset). (b) 1H Nuclear Magnetic Resonance (NMR) of PEG-thiol capped gold nanoparticles dispersed in D2O. The inset represents the full spectrum scan. The peak at 2.7 ppm suggests that there is a significant presence of bound citrate present in the PEG capped gold nanoparticle. .......................................................................................... 211

xxxv Figure 5.46: Raman spectra of PEG-thiol, sodium citrate and PEG-thiol-sodium citrate solutions in water. The sample concentrations correspond to the respective amounts in PEG-thiol capped gold nanoparticles. Raman spectra of citrate capped gold nanoparticles, with the characteristic carboxylate peaks at 1558 and 1604 cm-1; thereby suggesting the presence of only bridging bidentate complex. Raman spectra of PEG-thiol coated gold nanoparticle indicating negligible presence of Au-S peak ~ 300 cm-1. ........................................................................................................................... 212 Figure 5.47: Representative low magnification FESEM image of PEG capped gold nanoparticles (without any thiol moiety). This was formed after addition of PEG molecules of molecular weight 300 g/mol without any thiol functionalization to citrate capped gold nanoparticles and aged for 1 day. The molar ratio of PEG to gold is 10. The regions marked with solid circles represent Moiré patterns while the dotted rectangle represents image distortion due to the possible presence of organic film. . 215 Figure 5.48: (a-c) Representative high magnification FESEM image of PEG capped gold nanoparticles (without any thiol moiety) obtained at different locations and magnifications. This was formed after addition of PEG molecules of molecular weight 300 g/mol without any thiol functionalization to citrate capped gold nanoparticles and after being aged for 1 day. The molar ratio of PEG to gold is 10. (d) Size histogram of PEG capped gold nanoparticles. ................................................................................ 216 Figure 5.49: Representative FESEM image of (a) polydispersed tannic acid capped gold nanoparticles (13.8 ± 4.3 nm), (b) after addition of PEG-thiol and ageing for 16 hours (14.5 ± 2.1 nm) and (c) after post-factorial addition of sodium citrate to one day aged tannic acid-PEG-thiol capped gold nanoparticles (d) after one week (10.4 ± 4.9 nm) ............................................................................................................................. 217 Figure 5.50: (a) Representative FESEM image of ascorbic acid capped gold nanoflowers. (b) Representative FESEM image of gold nanoflowers after addition of PEG-thiol and ageing the sample for 1 day at 25 °C. ................................................ 218 Figure 5.51: Representative FESEM image of citrate capped gold nanoparticles synthesized with ethylene glycol as solvent (instead of water) before (a) and after addition of PEG-thiol (b,c). Fig d represent the FESEM image of gold nanoparticles synthesized with ethylene glycol as both reducing agent and stabilizing agent without

xxxvi citrate while e represents gold nanoparticles after capping with PEG-thiol molecules. The molar ratio of PEG to gold was maintained at 10 and the sample was aged at 25 °C for 1 day. ............................................................................................................... 221 Figure 5.52: Representative FESEM images of sodium acrylate capped gold nanoparticles (a) before and (b) after addition of PEG-thiol molecules and aged for 1 day. Representative FESEM image of citrate capped gold nanoparticles with citrate to gold nanoparticles (c) before and (d) after addition of PEG-thiol and aged for 2 days. Clearly excess citrate prevents ripening of particles. The molar ratio of PEG to gold is 10................................................................................................................................ 222 Figure 5.53: Phase space of PEG-thiol, gold, sodium citrate highlighting the regions of particle ripening. The black closed circles are the ones at which particle ripening occurred while the red closed circles are the points correspond to conditions of no evident particle ripening. Couple of experiments marked as points with closed half red and black circles were carried to identify the limits of boundary lines for particle ripening. There were only indications of particle ripening. ...................................... 223 Figure 5.54: Representative FESEM images of gold nanoparticles synthesized using modified sodium acrylate protocol, (a) before and (b-d) after addition of PEG-thiol at different magnifications. The region marked 1 (in c) represent due to the possible presence of organic film. The molar ratio of PEG to gold is 10 and the sample was aged for ~ 18 hours at 25 °C. ..................................................................................... 224 Figure 5.55: (a) Representative FESEM image of gold nanoparticles synthesized using Pechini process with the addition of PEG-thiol molecules (molecular weight: 356 g/mol) a priori. (b) UV-Vis spectra of as-synthesized gold nanoparticles (as shown in a) indicate suppression of the characteristic gold SPR peak at 520 nm, indicating sub-3 nm particles. .................................................................................... 225 Figure 5.56:Representative FESEM image of gold nanoparticles synthesized using citrate reverse protocol, with the a priori addition of PEG molecules (molecular weight: 300 g/mol); without ageing (a-b), with ageing for 12 hours at 25 °C (c-f), and with ageing for 1 day at 4 °C (g-h). The molar ratio of PEG to gold is 10. The regions marked 1 and 2 in c represent array and bare substrate respectively. ........................ 227

xxxvii Figure 5.57: Representative FESEM images of gold nanoparticles formed using the standard Pechini process with increasing the gold concentration 10 times (a-b) and decreasing the gold concentration 10 times (c-d) when compared to standard concentration. The molar ratio of PEG to gold is 1 and 100 respectively. ................ 228 Figure 5.58: (a-b) Representative FESEM image of gold nanoparticles synthesized using acetone dicarboxylic acid and PEG molecules (molecular weight: 300 g/mol). Both nanoparticles and nanoplates were observed. (c-d) Representative FESEM images of gold nanoparticles synthesized using acetone dicarboxylic acid alone. The concentration of nanoplates increased without the addition of PEG molecules. (e) Representative FESEM images of gold nanoparticles synthesized using acetone dicarboxylic acid and PEG molecules (with final pH modified to 6.5 using sodium hydroxide solution a priori). The molar ratio of PEG to gold was maintained at 10. .................................................................................................................................... 230 Figure 5.59: (a-b) Representative FESEM images of boiled aqueous boiled mixture of PEG and sodium citrate drop-casted on silicon substrate. ......................................... 232 Figure 5.60: (a) Schematic of formation of pseudocrown ether of poly(ethylene glycol) diacrylate (PEGDA)/metal complexes (adapted from reference255). (b) Schematic of formation of 3D complex network of citrate and PEG, through chelation of metal nanoparticle or metal ions. ........................................................................... 232 Figure 5.61: (a-d) Representative FESEM images of silver nanoparticles and nanorods prepared through standard Pechini process. (e) UV-Vis spectra of as-synthesized silver nanocolloid; shoulder at ~280 nm indicates the presence of silver nanorods in solution. (f) Representative FESEM image of silver nanorods prepared through Pechini process, without the addition of PEG molecules. The molar ratio of PEG to silver is 10. ................................................................................................................. 234 Figure 5.62: (a) Schematic representation of the cross-linked hydrogel (Reproduced with permission from Shapiro261. Copyright (2011) Elsevier). (b) TEM image of the freeze-dried manno/xylene gel indicating highly polydisperse mesh size (Reproduced with permission from Sakurai et al.262. Copyright (2003) American Chemical Society). ..................................................................................................................... 236

xxxviii Figure 5.63: Representative FESEM images gold nanoparticles drop-casted on hydrogen plasma treated PDMS substrate and transfer printed to the silicon substrate. .................................................................................................................................... 237 Figure 5.64: Digital photograph demonstrating reversible transformation of red to blue colour of the PEG-thiol coated gold nanoparticle solution, with temperature. .. 237 Figure 5.65: (a-c) Representative FESEM images after thermal cycling of PEG-thiol coated gold nanoparticles derived from tannic acid capped gold nanoparticles. The formation of nanopouches can be attributed to the possible side reaction of tannic acid with PEG molecules264............................................................................................... 238

xxxix

LIST OF TABLES

Table 2-1: Nomenclature of various symbols used in ellipsometric section ............... 33 Table 3-1: Physical and chemical properties of commonly used solvents and ligands. ...................................................................................................................................... 62 Table 3-2: List of symbols and the respective nomenclature, used in this section. ..... 78 Table 4-1: Work function of different metals that have been used as floating gates. 107 Table 4-2: Effect of RF sputtering process conditions on gold nanoparticle array. The sample was annealed at 500°C for 30 min under forming gas atmosphere ............... 113 Table 4-3: Effect of presence of ligand and nanoparticle density on the value of accumulation capacitance .......................................................................................... 146 Table 4-4: Effect of presence of ligand on accumulation capacitance. Devices fabricated with PECVD also showed reduction in capacitance at accumulation. ..... 146 Table 4-5: Comparison of experimental and predicted accumulation capacitance values due to the either change in effective dielectric constant (metal in a dielectric) or effective area screening (based on nanoparticle density). ......................................... 151 Table 4-6: Summary of effect of interparticle spacing in nanoparticle arrays and process parameters on memory window of the floating gate memory devices presented in this chapter............................................................................................. 162

1

Chapter 1 Introduction 1.1 Introduction to Nanoscience and Nanotechnolgy The transformation of scientific knowledge into functional end products has been the principle objective of engineering research. Major breakthroughs in various technologies have been accomplished due to the invention of new materials. The traditional approach of design of new functional products is to control the physicochemical properties by choosing appropriate materials1. Currently, there is a paradigm shift and widespread optimism, “thanks to the emergence of advanced nanoscale characterization tools in the last couple of decades”, in controlling and manipulating the morphology of nanostructures to achieve the desired functionality2. Nanoscience refers to the study of the fundamental principles of molecules and structures with at least one dimension between 1 to 100 nanometers, while nanotechnology refers to the application of these structures to form useful devices3. Nanotechnology involves the capability to synthesize, characterize and control such artificial structures. Nanoparticles are one of the fundamental building blocks of nanotechnology having all three dimensions in the sub-100 nm regime4. Nanoparticles exhibit behavior which is intermediate between that of a bulk solid and a molecular system. The scientific study of metal nanoparticles has generated a lot of interest from the time of Michael Faraday’s pioneering experiments in 18575. Apart from the size of nanoparticles, control over shape as well as composition, such as core-shell or bimetallic nanoparticles, also aid in manipulating the physico-chemical properties. As an example, selectivity in the furfural decarbonylation and hydrogenation reactions can be modulated through platinum nanoparticle size and shape6. Smaller sized particles (< 2 nm) predominantly give rise to furan as a product of decarbonylation, while larger sized particles (5 – 7 nm) yield both furan and furfuryl alcohol (Fig. 1.1). Further, the octahedral particles (6.2 nm) selectively yield furfuryl alcohol while cube shaped particles (6.8 nm) result in an equal amount of furan and furfuryl alcohol6.

2

Figure 1.1: Effect of platinum particle size on selectivity for furfural decarbonylation to furan and furfuryl alcohol (Adapted with permission from Pushkarev et al.6. Copyright (2012) American Chemical Society).

a

b

Figure 1.2: Demonstration of polyethylene glycol (PEG) tethered Gold-Cadmium telluride nanoparticles as nanoscale thermometer (adapted with permission from Lee et al.7 Copyright (2005) John Wiley and Sons). Plot of time variation of (a) photoluminescent intensity (E) and (b) corresponding temperature variation, as a function of time, for composite nanoparticles.

3 Another example of the application of hybrid nanostructures is the molecular scale thermometer developed by interlinking gold and cadmium telluride nanoparticles using polyethyleneglycol (PEG) molecules (Fig. 1.2)7. The underlying mechanism in nanoscale thermometer involves plasmon resonance and exciton-plasmon interaction7. Nanoparticles of noble metals namely gold, silver and copper have attracted the largest share of scientific interest and, in particular, gold nanoparticles have been studied extensively in the literature for various technological applications. The physico-chemical properties of gold nanoparticles differ from their bulk counterparts4. For example, gold nanoparticles exhibit remarkable catalytic activity, a lowering of their melting point temperature, surface plasmon resonance in the visible range and single electron charging behavior8. In addition to their fascinating individual properties, metal nanoparticle ensembles have opened up a plethora of applications ranging from electronics to biotechnology due to the emergence of dipolar interactions of surface plasmons or excitons or magnetic moments9-12. Nanoparticle ensembles have been used as substrates in Surface Enhanced Raman Scattering (SERS)12, templates for chemiresistive sensor12,13, molecular scale thermometer7, molecular ruler14, and strain gauges15. Furthermore, nanoparticle arrays can be used as templates for growing ultra-high density nanowires with the potential for enhancing the efficiency of solar cells16.

1.2 Fabrication of nanostructures: Top-down vs. Bottom-up In 1959, Richard Feynman discussed, in his seminal talk entitled “There is plenty of room at the bottom”, the possibility of manipulating and controlling things at the nanoscale17. In principle, there are two approaches for fabrication of any structure, namely, top-down and bottom up (Fig. 1.3)2. Top-down approaches involve the removal of material from the bulk in a pre-defined fashion so as to achieve the desired functionality. Alternatively, the bottom-up approaches involves brick-by-brick assembly to obtain the desired target. Examples of top-down approach include Ellora temple in India, world’s largest monolith structure, constructed by sculpting approximately 20000 tons of rock and Chariot temple in Hampi, India; while the Taj

4

Figure 1.3: Illustration of Top-down vs. Bottom-up approach using the real life examples of Chariot temple in Hampi18 and Egyptian pyramid19 respectively. Today, transistors20 are fabricated using topdown approach, which faces both fundamental and economic challenges in further scaling of devices. An attractive alternative way for fabricating nanostructures in sub-20 nm regime is bottom-up approach, which involve brick-by-by assembly of molecules or particles21.

Mahal, one of the new seven wonders of the world and Egyptian pyramids can be treated as being built through bottom-up approach. Until now, conventional top-down approach of optical lithography has been able to scale down to meet the ever growing demand of miniaturization in the field of microelectronics1. However, with the feature sizes approaching sub-30 nm regime, further scaling is hampered due to both economic and fundamental limitations. On the other hand, bottom-up approaches, i.e. construction of nanostructures using building blocks such as atoms or molecules are deemed to be promising candidates for controlling and manipulating things at the nanoscale due to its ease and cost-effectiveness. Bottom-up approaches include directwrite dip-pen nanolithography22, protein based assembly23, molecular assembler24 etc. Even though, aforementioned molecular level assembly processes are scientifically fascinating, they find limited use in practical applications due to their slow nature1. For example, atom by atom construction of 1 mm3 silicon will take at least 10 years or to make 1 mole of some compound at a rate 109 s would take at least 19 million years1. Alternatively, replacement of atoms/molecules with nanoparticles as a building block in the self-assembly process can circumvent the time limitation in the nanostructure fabrication1. In simple terms, self-assembly is a process of transformation from an initially disordered state into a final ordered state. Typically, the self-assembly process is close to equilibrium in which local interactions between

5 particles or building blocks drive the final structure. Various forces acting at different length scales such as van der Waals, electrostatic, magnetic, hydrodynamic, biological or depletion interactions etc. can be exploited for particle assembly in a desired fashion25. There are extremely few systems, wherein a priori the self-assembled structures are predicted based on inter-particle interactions. Many interactions that can lead to self-assembled structures are yet to be explored in a substantial manner. Further, the translation of the promise of self-assembly processes having technological applications is hampered severely due to issues related to both scalability and precise particle arrangement in functional nanoparticle based architectures26,27.

1.3 Nanoparticle

assembly:

Scale-up

and

particle

arrangement To date, several approaches have been proposed and implemented for self-assembly of nanoparticles into two-dimensional (2D) arrays at different length scales. But many of these approaches cannot be integrated with large scale device fabrication28-32, as they are limited in terms of substrate compatibility33, or limited to a particular size range1. Research efforts to scale up the existing batch scale self-assembly process are yet to begin in a serious way. In addition, until now the self-assembly processes have been restricted to colloidal particles dispersed in organic solvent32. So far, only a few reports have attempted self-assembly of particles directly from aqueous medium, and have met with limited success34. Any demonstration of scalable self-assembly approach directly from aqueous medium will have beneficial cost ramifications in terms of both economical as well as environmental cost in the field of flexible nanoelectronics, sensors etc. Thus, the challenge is two-fold: first to design scalable approach for self-assembly of nanoparticle arrays with precise particle arrangement and second to extend self-assembly approach so as to assemble particles directly from aqueous medium, instead of from an organic medium.

6

Figure 1.4: Illustration of the outline of this thesis

7

1.4 Nanoparticle arrays for next-generation, floating gate memory devices Over the last two decades, major advancement in the field of nanoscience and nanotechnology can be attributed to the significant efforts made in nanoelectronics, i.e. the drive towards miniaturization for making devices smaller, cheaper and faster11. Single electron transistors35, ultra-high density memory devices36, point-of-care diagnostics37, and electronic nose38 are few applications of nanotechnology. An integration of low-cost, inkjet printed gold nanoparticle sensor chip with simple microfluidic immunoarray has helped in the detection of two cancer biomarker proteins in 5 µL samples in 8 minutes39. The nanoassemblies enabled to have high sensitivity and ultralow detection. Assay time of 8 minutes is shown to have a clinically relevant detection limit of 5 pg per mL39. Assemblies of nanoparticles have also attracted interest in various applications such as light-emitting diodes (LEDs), photo-luminescence (PL) devices, biosensors etc., as it enhances the luminescence of fluorophores by coupling with localized surface plasmon resonances resulting from metallic nanostructures40. The efficiency of these devices critically depends on the distance between the metal surface and the fluorophore as the interaction between these molecules decay exponentially with distance. Organic light-emitting diodes (OLEDs) have shown potential for lighting source and display applications due to their high brightness, low power consumption, and superior color features41. Even though, 100% internal quantum efficiency have been achieved through phosphorescent emitting materials, the light extraction efficiency still remains poor (~20%) due to issues arising from total internal reflection, and the surface plasmon losses at the interface. Various approaches have been employed to reduce the losses. Significant ones include: (a) surface microstructure fabrication42, (b) use of photonic crystals43, (c) corrugated cathode44, (d) optical microcavity structures45 etc. Of late, utilization of localized surface plasmon resonance of metallic nanostructures in OLEDs has gained prominence44,46. Another similar area where light harvesting is an issue is the Organic Photovoltaics (OPVs), which has tremendous potential for becoming a next-generation renewable energy source. As mentioned earlier, one of the major drawbacks of these devices is

8 the insufficient photon absorption of the photoactive layer, i.e., efficient harvesting of sun-light is very poor. Nanoparticle assemblies have also been exploited in these applications to utilize localized surface plasmon resonance (LSPR) to enhance the light absorption efficiency47,48. Memory operations have been achieved through various mechanisms such as charge storage, electromechanical, phase transition, and magnetic storage49. Hybrid organic/inorganic nanocomposites (as organic bistable devices) are one class of nonvolatile memory device which is gaining prominence due to its mechanical (flexible) characteristics50. A typical structure of organic bistable devices incorporating nanocomposite consists of composite organic molecules, metal or semiconductor nanoparticle layer sandwiched between two metal electrodes. The electrical properties of these devices are characterized through current-voltage (IV) measurements50. The applied voltage across the device is varied in a cycle from negative bias to positive bias and then back to negative bias. The hysteresis or the memory characteristics is determined through two distinct states, high current (ON) and low current (OFF) positions. The voltage at which the OFF-ON transition occurs is termed as turn-on voltage, which is analogous to writing process in a conventional memory cell. Recently, n-terminal memory devices (two-terminal, three-terminal) using charge storage have gained lot of interest in fabrication of next-generation memory devices49. A two-terminal memory device with metal or semiconductor nanoparticles, nanowires, grapheme nanoribbons acting as one of the active materials has been demonstrated as resistive switch2,51-55. These resistive switches have been coupled with light-emitting diodes, which can potentially be used in electronic books54,56. Three-terminal memory devices utilizing the charge storage nodes are extensively studied57. Conventional floating gate flash memory devices are three-terminal metaloxide-semiconductor transistors49. They are the present technology driver in the semiconductor industry, which has doubled the data density every 18 months (Moore’s law) due to several advancements made in lithography based processes58. In floating gate/flash memory devices, charges are stored in a floating gate, separated by a tunneling oxide layer from the channel. The tunneling oxide thickness is scaled down to ensure high gate coupling ratio and to minimize power consumption. Based on International Technology Roadmap (ITRS - 2011), the 2017 target of Effective

9 Oxide Thickness (EOT) needs be less than 10 nm for fabricating next-generation nonvolatile memory devices59. At such small length scales, a few random defects in the oxide are unavoidable, which results in unacceptable device endurance and retention characteristics. Hence, alternative charge-trapping elements such as nanoparticles, organic materials, Silicon-Oxide-Nitride-Oxide-Silicon (SONOS) etc. have been proposed instead of conventional thin p-Si, which can minimize the data loss and enhance reliability through compartmentalization of charge storage58,60. Selfassembled 2D metal nanoparticle arrays can act as charge storage nodes in the floating gate memory devices. The salient challenges being to synthesize monodisperse nanoparticles, develop large scale guided self-assembly processes and to integrate with Complementary Metal Oxide Semiconductor (CMOS) memory device fabrication processes, thereby, meeting the targets of ITRS 2017 for nonvolatile memory devices.

1.5 Scope and structure The primary scope of this thesis is to integrate bottom-up self-assembly process with CMOS device fabrication steps, which will aid in fabricating next generation nonvolatile memory devices. In order to achieve this, first, a scalable bottom-up approach to fabricate wafer scale 2D arrays with tunable particle arrangement was developed. Second, the scalable and CMOS compatible process for fabricating next-generation, non-volatile, floating gate memory devices using the as-prepared self-assembled 2D arrays of gold nanoparticles as charge storage nodes were developed. Third, the possibility of synthesizing and assembling monodisperse nanoparticles from aqueous medium was also investigated. This thesis documents the results of research carried out on the above mentioned topics. The outline of thesis is illustrated in Fig. 1.4. The thesis is organized as follows. Chapter 2 describes the methods used for metal nanoparticle synthesis using the classical Turkevich-citrate61,62 and tannic acid protocol63. Next, detailed protocols to fabricate high-density, uniform monolayer of particles and transfer printing to any desired substrate is presented29,64. The chapter concludes with an overview of all the

10 nanoscale characterization methods used in the work, in addition to the procedure followed for sample preparation and data analysis. Chapter 3 presents the development of a scalable process for assembly of particles into ordered 2D arrays with precise particle positioning. It involves a detailed investigation using ellipsometry of the underlying mechanisms involved in the wafer scale self-assembly of gold nanoparticles. Then, systematic studies exploiting the effects of substrate and Marangoni instabilities to tune particle arrangement in 2D nanoparticle arrays is presented. Chapter 4 presents details of the development of CMOS compatible, scalable processes for fabricating floating gate memory devices using self-assembled 2D nanoparticle arrays as charge storage nodes. The device performance characteristics were studied using complementary measurement techniques, namely, capacitancevoltage and electrostatic force microscopy measurements, and are discussed in detail in this chapter. The effect of interparticle spacing within arrays on memory characteristics and device performance is also presented. Chapter 5 presents a simple process for self-assembly of gold nanoparticles directly from aqueous medium through PEG capping of gold nanoparticles. Addition of PEG molecules to citrate capped gold nanoparticles was found to form highly monodisperse particles, due to PEG-carboxylate interactions. The synthesis of monodisperse nanoparticles and their large scale, highly ordered assembly form the basis for integration of bottom-up approach with the well-established top-down approaches for commercial applications, such as floating gate memory devices. Finally, chapter 6 presents a summary of thesis contributions, and the scope for future work. The appendix contain results of an exploratory work on synthesis and assembly of iron-platinum nanoparticles for ultra-high density magnetic data storage applications65.

11

Chapter 2 Experimental Methods 2.1 Introduction The delayed emergence of nanotechnology vis-à-vis biotechnology as an independent research field is due to a lag in the invention of appropriate characterization tools. Even though, gold colloids were synthesized in a systematic manner way back in 18575, the microscopic characterization of particles was feasible only after many decades, beginning with the seminal work on the development of an electron microscope by Ernst Ruska and Max Knoll in 193166. Although, the understanding of forces or interactions between particles has been well established for decades, development of instruments to quantify such nanoscale interactions wasfound to be an extremely daunting task. For example, for measuring adhesion force between two substrates, the force or the distance has to be measured in nN or nm respectively. Hence, the noise level in the system has to be of the order of piconewtons or picometers. This makes the construction of characterization instruments extremely difficult and even today many advances are still being made to improve the performance of the existing instruments. With the development of newer technologies like piezo-controllers, cantilevers etc. Atomic Force Microscopy (AFM) has become feasible now. The advent of suitable characterization techniques such as electron microscopy, atomic force microscopy, ellipsometry etc., has led to the quantification and understanding of hitherto unexplored areas in nanotechnology. Dynamic Light Scattering (DLS), Transmission Electron Microscopy (TEM), Atomic Force Microscopy (AFM), UV-Visible Spectroscopy are few of the commonly employed characterization tools for determining particle size and its distribution, each having its own advantages and disadvantages. Hence, an accepted method of characterization is to compare the results from different characterization techniques to get a unified picture. Nanoparticles and their assembly form fundamental building blocks of nanoscience and nanotechnology4. In this chapter, first, experimental protocols used to synthesize and assemble monodisperse metal nanoparticles into ordered 2D arrays are presented. Next, the details of characterization instruments and procedures used in this study to characterize the nanomaterials are presented.

12

2.2 Synthesis of gold nanoparticles Aqueous gold colloids of size 7 ± 0.7 nm were synthesized at room temperature using tannic acid as both reducing and stabilizing agent, based on the method developed earlier in our group63. Typically, 30 mL of 0.64 mM aqueous chloroauric acid, at a pH of 3.2, was added dropwise to 45 mL of 0.9 mM aqueous tannic acid, maintained at pH 7.0. The pH of the reaction mixture was always maintained above 6.4 by the addition of requisite amounts of 1 % (w/v) KOH solution intermittently. The pH of the solution is critical, and the amount of buffering agent needed to be added was found to vary between different batches of tannic acid. Particles of different size were obtained by scaling the amount of chloroauric acid added. Dodecanethiol (DDT) capping of gold nanoparticles was accomplished by mixing 5 mL ethanol solution (containing 5 µL dodecanethiol) with 5 mL of aqueous gold colloid, maintained at a pH of 4. The pH of the solution is very critical because lower pH (less than 4) can result in aggregation of particles, while at higher pH (greater than 5) precipitating particles by centrifugation will be difficult. Further, the amount of thiol to be added is also critical. Addition of thiol greater than 5 µL used to result in loss of particles in the supernatant. After adding DDT, the solution was left undisturbed for 4 hours, and then centrifuged at 3500 rpm for 30 minutes. The precipitate was then washed with 5 mL ethanol twice to remove free dodecanethiol molecules. The final precipitate was dispersed in 0.4 mL of toluene (5 nm particles) or 1:3 (volume) chloroformtoluene mixture (7 nm particles), prior to array formation. However, larger size particles could be suspended in pure toluene by increasing the ligand length. The ability to disperse larger particles arises either due to the enhanced screening of interparticle van der Waals attraction by using a higher dielectric constant medium (i.e. adding chloroform) or by increasing the magnitude of repulsive interactions due to an increase in the chain length of the ligands. Polymer grafting to gold nanoparticle was achieved by mixing 7 mL acetone solution (containing 0.02% (w/v) thiolterminated polystyrene, molecular weight: 20000 g/mol, Polymer Source Inc.) with 5 mL of aqueous gold colloid67. [The disulfide bonds need to be broken before use of aged polymer batches. For this, 0.2 mL of hydrazine hydrate is added to polymeric solution and aged for 10 minutes before adding citrate capped gold colloid. Citrate capped gold colloid is obtained by heating a mixture of 0.4 mL of 1 % (w/v) sodium

13 citrate solution and 10 mL of tannic acid capped gold nanoparticle. As the solution becomes warm (~ 60 °C), 3 mL of hydrogen peroxide solution (6 % sol.) is added to degrade the tannic acid. Also, care needs to be taken to ensure hydrazine hydrate used is not from a very old batch].The solution was left undisturbed overnight, and then centrifuged at 3500 rpm for 30 minutes. The precipitate was then washed with 5 mL acetone twice to remove excess thiol-terminated polystyrene molecules. The final precipitate was dispersed in the desired organic solvent prior to array formation. Aqueous gold nanoparticles were also synthesized using well-established classical citrate/Turkevich protocol61. The overall molar ratio (MR) of sodium citrate to chloroauric acid was adjusted to 5.2 by adding 1 mL of trisodium citrate (34 mM concentration, 1% w/v solution) to boiling chloroauric acid solution. The total reaction volume was maintained at 25 mL, and the concentration of chloroauric acid in the reaction mixture was 0.254 mM. Experiments were also carried out by reversing the order of reagent addition to obtain monodisperse nanoparticles, based on the report of Sivaraman et al62. Also, few experiments were carried out at different molar ratios of sodium citrate to chloroauric acid, based on the report of Frens, to change particle size68.

2.3 Fabrication of 2D nanoparticle array Close-packed nanoparticle films were formed by controlling the curvature of water surface using Teflon cell designed by Santhanam et al. as shown in Fig. 2.129. Briefly, thin circular hollow disc machined in Teflon of 2 mm thickness (5 cm in diameter witha2cm diameter circular hole machined in its center) is placed on an open stand in a petridish. Leveler is used to ensure the Teflon disc is level. Deionized water is added until the water surface inside the hole assumes a convex upward curvature. Nanoparticle films were formed by drop-casting 0.4 mL (particle concentration ~ 1014/mL) of the colloidal solution on the water surface and allowing the organic solvent to evaporate (few minutes), as shown in Fig. 2.1. A hollow thick glass cylinder is placed to eliminate undesirable air currents present in the fume hood. Nanoparticle films were transferred to the desired substrate using poly (dimethylsiloxane) (PDMS) stamps64. PDMS stamps were prepared using Sylgard 184

14

a

b

c

d

Figure 2.1: Schematic representation of monolayer formation and transfer printing developed by Santhanam and Andres64. Key fabrication steps include (a) adjusting the curvature of water surface using Teflon cell, (b) drop-casting concentrated (~ 1014 particles/mL) gold nanoparticles dispersed in organic solvent on curved water surface, (c) Langmuir-Schaefer transfer of nanoparticle array from water surface to PDMS pad, and (d) Microcontact printing of gold nanoparticle array and transfer to desired substrate.

(Dow Corning). A silicon wafer with native oxide (stuck to a glass slide) is placed in a plastic weighing cup and covered with a 10:1 by weight mixture of PDMS oligomer and catalyst. The polymerization process is allowed to proceed at 80 °C for 4 h. The pads are peeled off the silicon chip and cut to the desired shape. The unreacted monomers were removed by sequential immersion in fresh hexane and fresh ethanol (at least 3 times for 10 minutes each). The pads are blown dry in a stream of nitrogen before use.

2.4 Nanoscale characterization methods 2.4.1 Scanning Electron Microscopy (SEM) Scanning Electron Microscope (SEM) is an imaging technique which provides topographical and elemental information at magnifications ranging from 10 x to 1000 kx with high depth of field. The topography present in images based on the secondary/backscattered electrons generated due to the impingement of high energy electron beam on specimen makes this an invaluable tool. Field-Emission Scanning Electron Microscopic (FESEM) images were obtained using ULTRA-55, Zeiss NTS

15

a

b

Figure 2.2: (a) Schematic representation of position of different detectors in the Gemini column of ULTRA 55 and (b) schematic representation of functioning of filtering grid in separating secondary and backscattered electrons (reproduced from Zeiss69 with permission)

Gmbh, at an operating voltage of 10 to 15 kV. ULTRA 55 FESEM is based on the ZEISS GEMINI® FESEM column with beam booster which aids in efficient detection of Secondary Electrons (SEs) and Backscattered Electrons (BSEs). The subsequent discussion in this section is taken from the report and presentation given by Zeiss69 to Department of Chemical Engineering, IISc. The column comprises of three detection systems fully integrated in the column: (i) high efficiency In-lens SE detector for high contrast topography; (ii) in-column EsB (Energy selective Backscattered electron) detector for low kV ultra high resolution material contrast; and (iii) integrated annular AsB (Angle selective Backscattered electron) detector for compositional and crystal orientation imaging, as shown in Fig. 2.2. In addition, there is also a lateral SE2 (Evarhort-Thornley or secondary electron 2) detector for surface topography imaging. Secondary electrons (SEs) having energy of less than 50 eV emerges from the specimen surface (due to the loss of energy, inelastic scattering) while the backscattered electrons (BSEs)are generated below the surface (no significant loss of energy, elastic scattering) from a larger scattering volume than the SEs. The low energy, SEs generated at the impact point of the primary electron beam is intercepted by the weak electrical field at the sample surface. They are then accelerated to a high energy by the field of the electrostatic lens and focused on the annular In-lens detector inside the beam booster located above the objective lens.

16 While the In-lens detector provides the best high resolution information, a lateral SE detector (Evarhort-Thornley)

in the specimen chamber provides optimum

topographical information. Signals from both the detectors can be mixed to deliver optimum image quality. This image mixing from two different detectors can be processed in real time in the same screen. For high resolution imaging, elastically scattered BSEs have to be detected. These high angle BSEs which form typically in a cone of 15° angle to the primary beam are attracted by the electrical field of the column and injected back into the column. Both Rutherford and Mott scattering take place as high energy beam impinges on the specimen. Mott scattering is similar to Rutherford scattering, but electrons are used instead of Alpha particles as they are much smaller (around 4 orders of magnitude). This enables them to penetrate the atomic nucleus, giving valuable insight into the nuclear structure. Rutherford scattered electrons are detected using BSE which is generated using multiple scattering processes; giving rise to high Z (atomic number) contrast. Mott scattered electrons are in principle single scattered electrons. This results in channeling contrast and resolution. They are detected using AsB detector. The method of separating and detecting the backscattered electrons is called: Energy and angle selective Backscattered detection, hence the name EsB and AsB detectors respectively. The detection efficiency of the EsB is around 85% of the high angle backscattered electrons. Only a fraction of the secondary electrons are deflected into the hole of the In-lens SE detector giving detection efficiency of over 90%. High angle EsB detector has small acceptance angle while the AsB detector has a large acceptance angle. As all the detectors are in line of optic axis, this alleviates alignment requirements during detector change. The images presented in this study were obtained using Secondary Electron (SE) through inlens mode, unless stated otherwise. Few images were also obtained using Angle Selective Backscattered electron (AsB) detector, which is sensitive to the atomic number (Z) of the material. Samples were prepared by dropcasting the colloidal solution (~ 5 µL) on silicon substrate, unless stated otherwise. Multiple images taken at different locations were used for image analysis and determining the size distribution. The edge to edge interparticle spacing and particle size distribution were obtained using customized code written in IgorPRO software by

17

b

a

c

e

d

Figure 2.3: Example of image processing using Clemex Vision PE software, a) binary image, b) defining process frame c) the thresholded image along with an outline of the threshold, (d) the original image with the outline based on thresholding and (e) histogram of circular diameter estimated within the process frame.

Ms.

Shruti

Seshadri

(code

can

be

accessed

at

http://chemeng.iisc.ernet.in/venu/temanalysis.pdf).Clemex Vision PE software was also used for obtaining particle size distribution. After setting the scale bar, the image was converted to binary format using grey threshold (Fig. 2.3a). This process was done manually. A process frame is then defined, wherein the features inside the process frame are only analyzed. In the example shown, a square box (with red color) represents the process frame (Fig. 2.3b, c). After this, ‘separate’ command is used to isolate particles. It is a morphological filter that separates features by distance analysis. Finally, the original image can be superimposed with the thresholded image with outlines alone to judge the level of thresholding (Fig. 2.3d). For particle size and distribution, the circular diameter is used. The equivalent circular diameter is calculated from the particle area (using pixels), as shown in Fig. 2.3e. For a particular sample, different images were obtained at different locations on the sample. The particle size and distribution of all images from a particular sample at an appropriate magnification is averaged. Thus, the standard deviation reported in this study is ‘pooled standard deviation’.

18

2.4.2 Transmission Electron Microscopy (TEM) Transmission Electron Microscopy (TEM) is a high resolution microscopic technique similar to optical microscope wherein light is replaced by electrons. The resolution of optical microscope is limited by the wavelength of light. Thus, use of electrons, which has much lower wavelength, enables at least 3 orders of magnitude better resolution than optical microscopes. In TEM, electrons are directed towards sample, which travel through vacuum from the top of a microscope. Electromagnetic lenses are used to focus the electrons into a collimated beam. Some electrons are transmitted through the sample depending on the specimen under analysis. The transmitted electrons fall on the fluorescent screen located at the bottom of the microscope and form the image. TEM images were obtained using Tecnai F30 operating at a voltage of 200kV. Samples for TEM were prepared by drop casting few drops (~ 0.1 mL) of the sample on a carbon grid that was placed on a lint-free tissue paper. Few samples were also prepared by drop-casting on silicon nitride membrane. Clemex Vision PE software was used to obtain the particle size and distribution from the TEM images. The particle size distribution analysis is similar to the analysis of the images obtained using FESEM.

2.4.3 Focused Ion Beam (FIB) Focused Ion Beam (FIB) is a technique used to deposit or etch materials from the sample. It resembles SEM in design, except for the difference that there are two columns, one an electron column similar to SEM chamber for image analysis, and the other an ion column which has focused ion beam either for sputtering, or imaging or etching. In this study, Helios NanoLab 600i, ultra-high resolution, dual beam was used for imaging and TEM cross-section sample preparation. The TEM sample preparation involves the following steps: (a) platinum deposition (Fig. 2.4a), (b) silicon etching (Fig. 2.4b,c), (c) welding omniprobe with platinum deposited on silicon (Fig. 2.4d), (d) silicon etching to remove lamella from the sample (Fig. 2.4e), (e) transfer welding the lamella attached to the omniprobe on to a TEM grid using platinum deposition (Fig. 2.4f), (f) etching the omniprobe from the lamella (Fig.

19 2.4g), and (g) thinning down the lamellar section to a thickness of 100 nm (Fig. 2.4h), so as to make it electron transparent. In addition to conventional secondary electron and backscattered electron detectors, this equipment has Scanning Transmission Electron Microscopic (STEM) detector, which enables to have resolution comparable to TEM with the advantage of imaging at lower vacuum and voltages when compared to TEM.

2.4.4 Atomic Force Microscopy (AFM) Atomic Force Microscopy (AFM) belongs to the family of scanning probe microscopy in which forces are used for constructing the topological information of the sample. First, a sharp tip is brought near the sample to be investigated. Next, the tip is scanned along the surface with either constant force or constant height mode. A feedback loop alters the height of the tip (or force) to ensure constant force (or height). The change in displacement of tip with respect to the drive amplitude is used for obtaining the topographical information of the sample. AFM characterization was performed usingMFP-3D (Asylum Research) in a clean room environment, maintained at 21°C and 45 % RH. A set of silicon nitride tips (Olympus, OMCL-AC240 TS, nominal spring constant: 20-40 N/m, resonant frequency: 340 kHz) were used for imaging and force spectroscopy. The heights of the nanoparticle films were obtained by averaging across several cross-sections from images spread across the sample (typically several mm’s across), and the standard deviations of the measured values (n>100 cross-sections) were approximately 1 nm. The film thickness was also estimated using histogram maps, obtained over entire area. The force-displacement curves reported were repeatable across several areas, and were further verified by intermittently checking the reproducibility of force-displacement curves on bare silicon substrate. The adhesion forces reported were based on the calculated spring constant obtained using Thermal-Noise method, which utilizes equi-partition theorem. It is based on the fact that for an ideal spring, the thermal energy (kBT; kB: Boltzmann constant and T: Temperature) can be equated to the potential energy of the spring (k /2; k: spring constant of the cantilever and : averaged thermal noise). However, due to the presence of non-ideality in the system, the above equation is

20

a

b

c

d

e

f

21

g

h

Figure 2.4: Representative FESEM images taken during sample preparation for cross-section imaging using Focused Ion Beam (FIB). The figures represent (a) platinum deposition on substrate, (b) silicon etching near the deposited platinum, (c) etching of platinum to aid omniprobe welding, (d) omniprobe welding to the lamella, (e) lamella removal, (f) transfer welding of lamella attached to the omniprobe onto a TEM grid using platinum deposition, (g) etching of omniprobe tip and (h) thinning down of lamella section to a thickness of 100 nm, so as to make it electron transparent.

modified further to account for the effects of cantilever tilt, aspect ratio etc70. The measurement of spring constant involves two steps: (i) estimation of sensitivity of cantilever based on the slope in the contact region (from force spectroscopy), and (ii) estimation of resonant frequency of the cantilever using thermal tuning. Electrostatic Force Microscopy (EFM) characterization, a variant of AFM, was performed using an Asylum Research-MFP3D instrument maintained at 21°C and45% RH, to study the charge storage characteristics of MOS capacitor at nanoscale. A set of platinum coated silicon tips (Olympus, OMCL-AC-240 TM, spring constant: 2 N/m, resonant frequency: 70 kHz) were used for imaging and EFM measurements. First, AFM topography scan was performed on the substrate using tapping mode. Then, charge was injected at -6 V for 2 min by placing the tip in contact with the sample (programming). Immediately afterwards, the AFM tip was taken to a height of80 nm above the surface while applying a tip voltage of 3 V (reading) and scanning by retracing the AFM topography recorded earlier. Localized negative/positive charges result in attractive/repulsive forces, which can be seen as changes in phase lag i.e. the phase lag between applied signal and tip oscillation. Scanning 80 nm above the surface helps in eliminating van der Waals interactions and capturing

22 only long-range electrostatic effects. The programming and reading voltages are kept constant unless stated otherwise.

Operating instructions for EFM operation in MFP-3D 1. In the main panel, select CONTACT MODE 2. From the Programming pull down menu, open up the CROSSPOINT PANEL 3. Click on the check box that says "no auto change XPT" 4. In the CHIP parameter box, select OUTC. 5. Click on the WRITE CROSSPOINT SWITCH button. 5. Use the "pick-a-point" feature and then "go there". You may want toselect "show tip location" so that you can confirm the tip position and engage 6. Type in the following command at the command line.... printtd_WriteValue("C%Output", 10)[this will apply a 10V bias to the tip] To turn it off, simply type the following at the command line: printtd_WriteValue("C%Output", 0)[this will apply a 0V bias to the tip]

2.4.5 Dynamic Light Scattering (DLS) Dynamic light scattering (DLS) is used to measure the size and distribution of particles especially in the nanometer range. DLS is also denoted as Photon Correlation Spectroscopy (PCS) and Quasi-Elastic Light Scattering (QELS). DLS is applicable only to measure particles which are small enough to be suspended in a liquid and undergo Brownian motion. For example, microemulsions, liposomes, latexes, nanoparticles etc., are generally characterized through DLS. In DLS, the focused beam from a laser source impinges upon particles that are undergoing random Brownian motion. As a result, the scattered light fluctuates with time at a given scattering angle which is collected by an avalanche photodiode (APD). This section is adapted from reference71. At any given instant, the total scattered light intensity at a given scattering angle depends on the position, shape and size of the particles within

23 the scattering volume. The total scattered intensity detected by the detector must be equal to the sum of scattered intensity by all particles (independently scattering particles). If the concentration of the sample is too high, then the intensity detected by the detector is less than the sum of the individuals (multiple scattering), because of the interaction between particles. This situation should be avoided by appropriate dilution of the sample. These temporal fluctuations of the intensity are correlated using a digital auto-correlator to obtain diffusion coefficient Dt. The hydrodynamic diameter dHis then obtained with the help of Stokes-Einstein equation,

Dt 

k BT 3d H

(2.1)

wherekB is Boltzmann constant and η is the viscosity of the sample at temperature T. The second order auto-correlation function (ACF) is given by,

C ( )  n( )n(t   ) 

(2.2)

where n(t ) is the number of photons counted over a sampling interval time t and n(t   ) is the number of photons counted over





centered at

but delayed in time by

 . Then, Seigart relation is used, which is given by



C ( )  B 1  f g ( )

2



(2.3)

For a monodisperse sample,

g ( )  exp(  )

(2.4)

  Dt q 2

(2.5)

and

where  , the line width of the frequency broadened distribution of the scattered light. The symbol, q, refers to the amplitude of the scattering wave vector, which depends on the refractive index of the liquid, wavelength of the laser and the scattering angle. In the case of dispersions being polydisperse (particles with a distribution of sizes rather than a single size), there is a distribution of diffusion coefficients. The particles of a particle size class contribute their own exponential to the auto-correlating

24 function (ACF). Thus, the correlation function is now a sum of exponentials. There are many algorithms like ‘method of cumulants’, Non-Negative Least Squares (NNLS), continuous (CONTIN)etc., to interpret the correlation function, as a sum of exponentials. In the cumulants method, the logarithm of the ACF is expanded as a power series with time. The coefficients are the cumulants. The first cumulant is equal to the reciprocal of the average relaxation time. The second cumulant is a measure of variance. The ratio of second cumulant to the square of the first cumulant is referred to as ‘poly’. If the second and higher order cumulants are zero, the ACF is a single exponential. Alternatively, an inverse Laplace transform of the data is used to obtain the distribution function. Different software packages like Non-Negative Least Squares (NNLS), CONTIN can be used to get an estimate of the particle size distribution. Here, the data reported is based on NNLS fit. Dynamic Light Scattering (DLS) instrument [Model BI-200SM] from Brookhaven Co. with BI-9000AT correlator system is used for obtaining autocorrelation function. Sample cells were cleaned with 2 % Micron-90 solution (Cole-Parmer) at 60 °C, followed by de-ionized water and ethanol. The samples were used without any filtration. Multiple scattering and interparticle aggregation effects were eliminated by using appropriate concentration of the sample. Interpretation of DLS results depends on an understanding of the principles of Brownian motion and light scattering. Hence, to characterize the sample under true Brownian motion, concentration of the sample was decreased until the intensity detected by the detector decreases with dilution (to preclude multiple scattering). The results with baseline difference of less than 1 % were only considered for analysis. The NNLS intensity plot was used to analyze data because, when the intensity plot is converted to a number density plot, based on Rayleigh scattering, the higher modes generally vanish. Hence, to obtain results without any bias, intensity plot (actual measured value) was used and the peak height was used in selecting the dominant mode, keeping in mind the effect of size on scattered intensity. A Gaussian curve was fitted using Origin software to the dominant mode in the intensity weighted particle size distribution obtained using DLS. The software gives mean (μ) and linewidth (L). From linewidth, standard deviation (σ) was calculated using the relation σ = 0.849 L/2. The percentage polydispersity is obtained as σ*100/μ. Few DLS measurements involving temperature dependence

25 were carried out using ZetaPALS instrument from Brookhaven Ltd. Zeta potential values were estimated using phase analysis light scattering. The reported values represent the mean of at least 6 measurements.

2.5 Spectroscopic characterization 2.5.1 UV-Visible spectroscopy (UV-Vis) Even though, Michael Faraday was the first to recognize that the red color of colloidal gold was related to the size of the particles, Mie was the first to explain this phenomenon theoretically, in 1908, by solving Maxwell’s equation for absorption and scattering of electromagnetic radiation by spherical particles72. The peak observed in absorption of light by metallic nanoparticles is attributed to the coherent oscillation of the conduction band electrons induced by an electromagnetic field. These collective oscillations are known as surface plasmons. This effect is treated as small particle effect, as this cannot be seen either in atoms or bulk. This occurs due to physical confinement of electrons in the nanoparticle. In recent years, probing of particle shape, size and its distribution using Surface Plasmon Resonance (SPR) has gained a lot of interest. In literature, the variation of peak position with particle size has been characterized for many systems73. The wavelength at which SPR occurs can be used for characterization of nanoparticles. In the case of gold, the SPR occurs in the visible regime, which can be detected by UV-Visible spectroscopy. In this study, UV-Visible spectrometer from Shimadzu Scientific Instruments (Model: UV-2100) was employed for sample characterization. The sample was scanned over a wavelength ranging from 200 nm to 900 nm with a resolution of ± 1 nm. The solvent used for dispersing the colloid was used for background subtraction.

2.5.2 X-ray Photoelectron Spectroscopy (XPS) X-ray Photoelectron Spectroscopy (XPS) is a quantitative spectroscopic technique to obtain elemental composition as well as the oxidation state of the elements present in the sample. It uses monochromatic x-rays (typically soft x-rays; energy: 200-2000 eV) to excite electrons from the core levels of the sample into vacuum.

26 Consider material A whose core electron is ejected by the impact of monochromatic x-rays, 𝐴 + ℎ𝜐 → 𝐴+ + 𝑒 −

(2.6)

whereh is Planck’s constant and 𝜐 is the frequency of incident beam. By applying energy conservation, 𝐸(𝐴) + ℎ𝜐 = 𝐸(𝐴+ ) + 𝐸(𝑒 − )

(2.7)

𝐾𝑖𝑛𝑒𝑡𝑖𝑐 𝑒𝑛𝑒𝑟𝑔𝑦 = ℎ𝜐 + [𝐸(𝐴) − 𝐸(𝐴+ )] = ℎ𝜐 + 𝐵𝑖𝑛𝑑𝑖𝑛𝑔 𝑒𝑛𝑒𝑟𝑔𝑦

(2.8)

The kinetic energy distribution of the emitted photoelectrons is then analyzed for determining the composition and the oxidation state of the sample. In this study, XPS spectra were recorded using a Thermo ScientificMultilab-2000 instrument with Al Kα (energy of 1486.6 eV) as the X-ray radiation source. XPS spectra were fitted using Fityk software74. The peak positions were calibrated with respect to the C1s peak (284.8 eV). Typically, the XPS spectrum of the sample is a combination of contributions from the spectrometer and energy loss processes within the sample, in addition to the electronic spectra from the material under investigation. The x-rays can penetrate deep into the surface, beyond the surface layers responsible for photoelectric and Auger peaks in XPS spectrum. This can result in electrons undergoing inelastic collisions within the sample, thus altering the electrons detected in the system. These energy loss processes result in background count in XPS spectra which needs to be subtracted before making quantitative analysis. One of the widely used background subtraction method is the Shirley background method, which states that the background at any binding energy is proportional to the area of the peak above the background at higher kinetic energies75. For example, the linear background value L(E) at any point E is estimated using the formulae, [𝐸 −𝐸]

[𝐸−𝐸1 ]

2 −𝐸1

2 −𝐸1 ]

𝐿(𝐸 ) = 𝐼1 [𝐸 2

+ 𝐼2 [𝐸 ]

(2.9)

where, E1and E2 are distinct energies and I1and I2 are intensity values chosen such that background values merge with the spectral bins at E1and E2. An example is

27

a

b

Figure 2.5: Illustration of estimation of (a) linear background (L(E)) and (b) Shirley background (S(E)) for titanium 2p XPS spectrum (reproduced from reference76).

demonstrated for titanium doublet pair (Fig. 2.5a)76. On the contrary, estimation of Shirley background value S(E) at point E involves an iterative procedure to account for the data variation. Shirley background is estimated using,

𝑆(𝐸 ) = 𝐼2 + [𝐼1 − 𝐼2 ]

𝐴2 (𝐸) 𝐴1 (𝐸)+𝐴2 (𝐸)

(2.10)

where A1andA2 represent the integrated areas as depicted in Fig. 2.5b, provided the value of S(E)is known. With an initial estimate of areas, the background value is estimated and the values are iteratively refined to obtain a better estimate of areas and hence S(E).Fityk software74 can automatically calculate and subtract the Shirley background.

2.5.3 Nuclear Magnetic Resonance (NMR) NMR spectroscopy is a technique used to determine physical or chemical properties of the molecules present in the system. It depends on the phenomenon of nuclear magnetic resonance, wherein chemical shift of resonant frequencies of the nuclei are extracted to determine the structure of the molecule. In this study, 1H-NMR spectra for the polymer were recorded on a Bruker NMR spectrometer at 400 MHz using water as solvent (facility available at NMR Research Centre, IISc).

28

2.5.4 Small-Angle X-ray Scattering (SAXS) Small-Angle X-ray Scattering (SAXS) is a scattering technique similar to Dynamic Light Scattering (DLS) with the difference that x-ray is used as a source. Also, the scattered x-rays are recorded at very low angles (less than 10°), which is sensitive to both shape and size of the material involved. In this study, SAXS data were obtained using Hecus S-3 Micro system (available at Raman Research Institute, Bangalore; Courtesy: Prof. Raghunathan), with a 1D Positional Sensitive Detector (PSD). The sample temperature was maintained at 30 °C and accumulation time used was 5101 s.

2.5.5 Raman spectroscopy Raman spectroscopy is a spectroscopic technique used to obtain the chemical composition and structure of the sample. It involves the polarization of electrons in a molecular bond using monochromatic light source, resulting in loss of energy which is termed as either inelastic scattering or Raman effect. It is characteristic of the energy of a probed molecular bond. In this study, Raman spectra were recorded using LabRAM HR (UV) system with the 514 nm laser (Courtesy: Prof. Umapathy, IISc).The baseline correction in Raman spectra was done using COBRA software77. The program uses Iterative Wavelet Transform Algorithm (IWTA) to obtain the background spectrum. IWTA is based on the Discrete Wavelet Transform (DWT), which is a modified version of Continuous Wavelet Transform (CWT)78. The signal transformation is another form of representing the signal, so that further mathematical operations can be carried out. The wavelet transform provides a time-frequency representation of the signal. In CWT, 𝑋(𝑡), the signal to be analyzed is transformed 𝑡−𝜏

using basis function Ψ (

𝑠

)through,

𝑋(𝜏, 𝑠) =

1 √|𝑠|

𝑡−𝜏

∫ 𝑥(𝑡). Ψ (

𝑠

) 𝑑𝑡

(2.11)

where𝜏 is the translational parameter, 𝑠 is the scale parameter (inverse of frequency). Scaling function helps in either compressing or expanding the signal. A large scale(low frequency) expands the signal so as to detect intensity information over a

29

Figure 2.6: Illustration of estimation of background spectra using IWTA algorithm (reproduced from Galloway et al.78 with permission). (a) Superposition of original signal and the 7 th level approximation after single iteration. (b) Superposition of new approximation curve and the chopped original curve at the approximation values in a. (c) Superposition of original signal and the successive approximation curves after multiple iterations, as stated in a and b.

small range while small scales (high frequency) provide a global picture. In essence, the CWT is obtained by changing the scale of the window under analysis, shifting the window in time, multiplying the signal and integrating over all times. The DWT helps in fast computation of wavelet transform using sub-band coding enabling time-scale representation of digital signal. The signal to be analyzed is passed through filters with different cutoff frequencies at different scales79,80. Figure 2.6a shows the original Raman spectra and the 7th level approximation curve after single iteration. In Fig. 2.6b, all the points above the approximation spectrum is set equal to approximation spectrum itself and then the new background spectra is obtained by applying DWT to this modified signal. This process of chopping the spectra and redefining the background spectra using DWT is repeated until the convergence criterion set by the user is met (Fig. 2.6 c).

2.5.6 Ellipsometry Ellipsometry is a non-destructive optical technique to determine film thickness and optical properties based on the characterization of light reflection from samples. This is obtained based on the change in polarization of light before and after reflection from the film surface. The name “ellipsometry” originate due to the fact that the shape of the polarized light becomes elliptical upon reflection. When a linearly polarized light wave is incident on the surface at some angle, the plane of incidence is defined as a plane perpendicular to the surface and contains a vector in the direction of wave

30

Figure 2.7: Schematic representation of light incident on the sample surface and getting reflected, adapted from reference81.

propagation81. This vector is termed as wavevector, kin. The light wave consists of electric field, E and magnetic field, B. The two components of E are respectively parallel and perpendicular to the plane of incidence, and the vectors are termed as π (or p-) and σ (or s-) polarized light respectively. Upon incidence on the surface, the linearly polarized light transforms into elliptically polarized light wave (Fig. 2.7). By using Jones formalism for the ratios of reflected electric fields to incident electric fields81, the elliptic polarized light wave can be rewritten in terms of amplitude and phase difference (Fig. 2.8). Eiy Eix Aiy exp iiy  Aix exp iix  Xi   X o Eoy Eox Aoy exp  joy  Aox exp  jox 

(2.12)

For a linearly polarized light wave, Aix = Aiy; and 𝜙𝑖 = 0. This leads to,





X i Aox  expi ox  oy   tan expi  X o Aoy

(2.13)

Thus, Δ represents the phase difference between reflected π (or p-) and σ (or s-) polarized light and ψ represents the angle determined from the amplitude ratio between π (or p-) and σ (or s-) polarized light. These are the two parameters which are determined from the ellipsometer. Alternatively, when light is incident upon a three layer structure (with substrate, thin film and air), some part of it gets reflected, and

31 some gets transmitted depending on the refractive indices of the layers, as shown in Fig. 2.9. The amplitude reflected coefficient can be expressed as the sum of all the reflected waves as depicted in Fig. 2.9, r012  r01  t01t10r12 exp i 2   t01t10r10r122 exp i 4   ...

(2.14)

The nomenclature of all the symbols used in this section is presented in Table 2-1.

As the infinite series y = a + ar + ar2 + … reduces to a/(1-r) for r < 1, the reflected coefficient can be rewritten as,

r012  r01 

t01t10r12 exp i 2  1  r10r12 exp i 2 

(2.15)

Similarly, the amplitude transmission coefficient can be expressed as, t012 

t01t10 exp  i  1  r10r12 exp  i 2  

(2.16)

Knowing individual reflection and transmission coefficients, the reflection coefficient can be expressed in terms of quadratic expression in x, which is defined as,

x  exp i 2 

(2.17)

Note, earlier the same reflection coefficient is expressed in terms of Δ and ψ, which can be measured experimentally. The above expression can be rewritten by employing the definition of the wave vector,  2d  ln( x)  i 2 n cos     

(2.18)

Rewriting the equation, the thickness of the film can be expressed as, d

i ln( x ) 4n cos 

(2.19)

32

Figure 2.8: Schematic representation of ellipsometric wave in linear coordinate.

Figure 2.9: Schematic representation of incident light getting reflected and transmitted on a three layer structure, with substrate, thin film and air as layers81.

33

Table 2-1: Nomenclature of various symbols used in ellipsometric section

Symbol

Nomenclature

E

Electric field

X

Jones vector (describes the polarization of light)

n

Refractive index of medium

r

Reflectance

t

Transmittance

β

Phase change due to light propagation through thin film

λ

Wavelength of incident light

A

Amplitude of light wave

𝝓

Phase difference between p- and s- polarized light

ψ

Amplitude ratio between p- and s- polarized light

θ

Incident angle of light wave

34 The measured Δ and ψ (contained in x) depends strongly on the film thickness and optical properties of the material. Ellipsometry involves measurement of Δ and ψ by either varying the incident light wavelength at a particular angle of incidence or varying the angle of incidence at particular wavelength. Variable Angle Spectroscopic Ellipsometer (VASE, J.A. Wollam Co. Inc) was used for obtaining ellipsometric data by varying incident light wavelength at fixed angle of incidence while Accurion (EP3 SW) was used for obtaining ellipsometric data by varying angle of incidence at fixed incident light wavelength. The data from the ellipsometer were modeled by building a film structure with in-built optical constants (details are given later in the sec. 3.3.2.1). In addition, ellipsometric data at the air-water interface was obtained by combining nulling ellipsometry with microscopy.

2.5.7 Photoluminescence spectroscopy Photoluminescence spectroscopy is a spectroscopic technique to study the electronic structure of materials. It involves the photo-excitation of the sample and the measurement of wavelength dependence of the intensity of resulting inelastic photoemission. In this study, photoluminescence spectra were obtained using Witec alpha SNOM setup. All excitations were performed with the blue line 488 nm of an Argon ion laser, while the transmitted light was detected either using a photomultiplier tube or a charge coupled device detector.

2.6 Sample cleaning methods 2.6.1 Reactive Ion Etching (RIE) Reactive Ion Etching (RIE) is an etching technique, commonly used in microfabrication, to etch materials from the surface using high-energy plasma. Plasma is often referred as the fourth state of matter. Plasma is a collection of electrons, ions, neutrals, free radicals and photons. The ionization of neutral molecules helps in sustaining the plasma in the steady state. Plasmas can be generated by a variety of discharge techniques, such as direct current discharges (DC), low-frequency discharges (kHz range), radio-frequency discharges (RF, MHz range) and microwave

35 discharges (GHz range)82. Here, RF based plasma is used for sample treatment. In addition to etching applications, plasma has been conventionally used to surface treat the samples such as silicon wafer, PDMS pads etc. In reactive ion etching, the conventional etching involving reactions between the polymer substrate and neutral species (in the plasma) can be accelerated through ion bombardment. In the present study, RIE from Milman Thin Film Systems Private Ltd. was used to remove the ligands from the ligand coated nanoparticles based on the optimum conditions (to retain hexagonal ordering in the 2D nanoparticle array) reported12. Overall, it is better to use mild RF power (low ion energies) so as to ensure both the ligand removal and minimize the ion bombardment. Typically, for removing dodecanethiol ligand, hydrogen plasma was used for 20 s at a gas flow rate of 50 sccm, while for removing polystyrene thiol ligands of molecular weight (3000 and 20000 g/gmol), oxygen plasma was used for 50 s and 4 min at a gas flow rate of 100 sccm respectively. Furthermore, surface pre-treatment of silicon wafer and PDMS samples were also carried out using UV-Ozone treatment for 2 minutes using a commercial UVO cleaner (Jelight-42).

2.6.2 Radio Corporation of America (RCA) cleaning Radio Corporation of America (RCA) has standardized wafer cleaning steps for removing metallic and organic contaminants83. Typically, the cleaning has four steps, (i) RCA 1 cleaning – dipping silicon wafer in the mixture containing 5:1:1 :: H2O:H2O2:NH4OH at 75 °C for 10 min for the removal of organic contaminants; (ii) dilute HF dip - removal of native silicon dioxide layer by dipping the silicon wafer in 10:1:: H20:HF solution for 20 s; (iii) RCA 2 cleaning – dipping silicon wafer in the mixture containing 6:1:1 :: H2O:H2O2:HCl at 75 °C for 10 min for the removal of metallic contaminants; (iv) dilute HF dip - again removal of thin silicon dioxide layer, which may have formed during RCA cleaning, using dilute HF solution (as in step ii). The cleaning is performed just before loading the silicon wafer for dry oxidation. The storage of the samplewas found to form poor quality native oxide (based on ellipsometric measurements) and leads to a loss in electrical quality.

36

2.7 Deposition techniques 2.7.1 Radio-frequency magnetron sputtering (RF sputtering) Radio-frequency magnetron sputtering (RF sputtering) is one of the physical vapor deposition techniques used to deposit thin films of both metals and dielectric materials. This involves deposition of the ejected target material (source) onto the substrate using high energy plasma discharge. The thickness and the quality of the film sputtered depend mainly on two parameters, namely, plasma power and substrate to target distance (Fig. 2.10). Magnetron helps in focusing the sputtered material onto the substrate. Sputtering can be done in two modes, namely, Direct Current (DC) or Radio Frequency (RF). For conducting samples, both DC and RF modes can be employed, while for non-conducting samples RF mode is employed. This is because use of DC mode leads to building up of electrons on the dielectric material and the plasma might extinguish. In this study, RF magnetron sputtering (ANELVA SPF-332 H) is employed. Digital photograph of the RF sputtering unit is presented in Fig. 2.11. The base pressure was maintained at 2 x10-5 mbar, unless stated otherwise and argon gas was used to generate plasma.

2.7.2 Atomic Layer Deposition (ALD) Atomic layer deposition involves layer-by-layer deposition process for depositing thin films of metals, oxides, nitrides and sulfides. The deposition process involves four steps: (i) exposure of substrate to metal precursor such as trimethyl aluminium (TMA), (ii) nitrogen purge to remove unadsorbed species from the substrate, (iii) exposure of the second precursor such as water, and (iv) nitrogen purge again to remove unadsorbed or unreacted species from the substrate84. For example, during the first cycle, TMA adsorbs on the silicon (OH) substrate, which is then followed by water exposure that converts adsorbed TMA to alumina (Fig. 2.12). Al(CH3)3 (g) + SiOH (s) SiO-Al(CH3)3(s) + CH4 2 H2O (g) + SiO-Al(CH3)3(s) SiO-Al(OH)2 (s) + 2 CH4

(2.20) (2.21)

37

Figure 2.10: Schematic representation of RF magnetron sputtering process used for depositing dielectric on top of gold nanoparticle array. The thickness and the quality of the film sputtered is found to depend on four parameters, namely, Argon pressure, applied voltage between target and substrate, substrate to target distance and deposition time.

a

b

c

Figure 2.11: (a) Digital photograph of Anelva RF magnetron sputtering unit in Centre for Nano Science and Engineering (CeNSE), IISc. Digital photographs of (b) substrate holder and (c) target, in Anelva RF sputtering unit.

38 i

ii

iii

iv

Figure 2.12: Schematic representation of the first cycle of Atomic Layer Deposition (ALD) process of Al2O3, comprising of four steps namely, (i) exposure of aluminum precursor to silicon surface, (ii) reaction of TMA with OH and methane formation (iii) monolayer adsorption of TMA and (iv) formation of new hydroxyl group and oxygen bridges; adapted from reference85.

Figure 2.13: Schematic representation of layer by layer growth of Al 2O3 using ALD process, adapted from reference85.

39 In the subsequent cycles, the deposited alumina acts as substrate and the cycle continues till the desired thickness is obtained (Fig. 2.13). Al(CH3)3 (g) + AlOH (s) Al-Al(CH3)3(s) + CH4

(2.22)

2 H2O (g) + AlO-Al(CH3)3(s) AlO-Al(OH)2 (s) + 2 CH4

(2.23)

The pulse times for all the steps were set at 0.2 ms and the growth temperature at 250 °C, unless otherwise specified. The quality and thickness of thin film depend critically on the amount of nucleation events i.e., adsorption probability of initial precursor on to the substrate84. Also, the temperature of the growth has been reported to significantly affect the quality of thin film86.

2.7.3 Thermal evaporator Evaporation is a simple and fast method for the deposition of thin films onto a substrate. Depending on the source of evaporation, the process is classified as thermal or e-beam or flash evaporation. The source material, for example, aluminium is evaporated in vacuum and condensed onto the substrate. In this study, thermal evaporator (HindHivac 15F6) was employed for making aluminium metal contacts. As in RF sputtering, parameters such as substrate-target distance, vacuum conditions etc., dictate the quality of thin films deposited in thermal evaporation. The chamber pressure was maintained at 5 x 10-5 mbar during aluminum evaporation. The thickness of the film is monitored using quartz crystal monitor. For device top-contact, shadow evaporation was carried out by using commercially available 300 µm shadow mask.

2.7.4 Electron-beam (e-beam) evaporator In electron-beam (e-beam) evaporator, electron beam is used as the source of evaporation. In this study, e-beam evaporator (Tecport, Symphony Precision) is used for depositing alumina, control oxide for fabricating memory devices. The chamber was maintained at room temperature. The base pressure was maintained at 4.3 x 10-

40 6

mbar with the deposition rate of 1Å/s. The oxygen flow was maintained at 1 sccm

during oxide deposition. The target to substrate distance was maintained at 52 cm.

2.8 Electrical characterization 2.8.1 Capacitance-Voltage (CV) measurement Capacitance-Voltage (CV) measurements aid in quantifying the charges stored in the Metal-Oxide-Semiconductor (MOS) capacitor. A MOS capacitor is essentially composed of an oxide that is sandwiched between a semiconductor and a metal gate. The capacitance of the MOS capacitor depends on the applied gate voltage, as depicted in Fig. 2.14; one identifies three regions, namely, inversion, depletion and accumulation depending on the gate voltage applied. For p-type MOS capacitor, accumulation represents the accumulation of holes (majority charge carriers) at the surface, depletion represents the lack of any major carriers at the surface and inversion represents the accumulation of opposite/minority charge carriers (invert) at the surface. The threshold voltage is the voltage which separates depletion from inversion. The flatband voltage represents the demarcation between accumulation and depletion. Typically, CV curves were recorded by applying bidirectional sweeps from inversion to accumulation and then back to inversion. To avoid sudden stress in the device, voltage bidirectional sweeps were sequentially increased from 1 V to 10 V (or maximum voltage) with the step size of 1 V. The capacitance also depends on the measurement frequency. At very low frequencies (referred to as quasi-static conditions), the holes are generated at a faster rate in the surface layer, resulting in sweeping away of the charge carriers to the Si/SiO2 interface. At high frequencies, the generation rate is not fast enough to allow the formation of a hole charge density at the Si/SiO2 interface, and hence leads to a lower capacitance values. Hence, for all measurements in the study, the frequency was set at 1 MHz (high frequency). Also, the hold and delay times were set at 100 ms to ensure steady state conditions. Capacitance-voltage characteristics were recorded using HP4284A capacitance meter. In addition, capacitance time measurements were performed after applying a bias voltage equal to flat band voltage. The details of the Program/Erase (P/E) voltages are mentioned later in the sec. 4.5.2.

41

a

b

Figure 2.14: (a) Illustration of three states of an ideal p-MOS capacitor depending on the gate voltage, namely, (i) accumulation (negative gate voltage), (ii) depletion (close to zero gate voltage) and (iii) inversion (positive gate voltage). (b) Schematic representation of Capacitance-Voltage (CV) characteristics of a p-MOS capacitor, as the gate voltage is swept from inversion to accumulation.

42

2.9 Summary The main aim of this chapter was to summarize the merits and demerits of various advanced nanoscale characterization tools used in this study for investigating physical and chemical properties of the fabricated/synthesized nanostructures. Experimental protocols for synthesizing and assembling nanoparticles into 2D arrays was also presented. These form the basis for the work reported in this thesis.

43

Chapter 3 Self-assembly

of

ligand

protected nanoparticles: Scale-up and particle arrangement in 2D arrays 3.1 Introduction The emergence of novel electronic, optical and magnetic properties in ordered twodimensional (2D) nanoparticle ensembles, due to collective dipolar interactions of surface plasmons or excitons or magnetic moments, and the appeal of cost-effective nanostructure fabrication have motivated intense research efforts into fabricating selfassembled nanostructures10. Several applications of nanoparticle arrays, viz. as catalysts for templated nanowire growth for dye sensitized solar cells87, as catalyst layers in fuel cells88,89, as electrical transducers of chemical or biological binding events90 etc., require control over interparticle spacing in the 10-50 nm regime. In addition, few applications such as magnetic recording media demand nanostructures in a square pattern, as opposed to the ubiquitous hexagonal pattern91. Thus, there is a demand for a scalable approach to control particle arrangement in 2D nanoparticle arrays. Nanoparticle ensembles can be obtained through two approaches, namely, topdown and bottom-up1. For nanostructure fabrication in the sub 50 nm regime, topdown lithography based approaches are hampered due to both economic and fundamental limitations while the bottom-up self-assembly based processes are hindered due to two serious problems, namely, (a) scalability and (b) control over particle arrangement in arrays. Of the several bottom up approaches proposed so far in the literature, many such as dip-pen lithography22, molecular assembler24 etc., are time consuming; in contrast, self-assembly of ligand protected nanoparticles12 offers an attractive avenue for cost-effective large scale fabrication of nanostructures. To date, several approaches have been proposed for fabricating large scale 2D arrays using self-assembly of ligand protected nanoparticles. Significant ones include: (i) self-assembly on a solid substrate by drop-casting28, (ii) self-assembly at fluid interfaces such as air-water64 or air-toluene interface31 in conjunction with

44 microcontact printing which enables transfer of arrays to any desired substrate and (iii) convective assembly of particles on solid substrate26. Self-assembly on a solid substrate is driven by wetting properties and limited by issues related to substrate compatibility33 while the self-assembly at fluid-fluid interfaces has been demonstrated only over small domains (few cm scale). Convective based assembly has so far been demonstrated only for larger size particles1. Moreover, a detailed understanding of the process involved is not generally available, and scale up of these self-assembly processes is yet to begin in any serious way. Self-assembly of ligand protected nanoparticles29 or the use of diblock copolymer micelles92 as templates are the most preferred routes for fabricating ordered 2D arrayswith precise particle arrangement. Self-assembly of ligand protected nanoparticles is a simple technique, wherein the interparticle spacing is typically tuned by changing ligand molecules; Self-assembly of diblock copolymer templates offers the possibility of tuning nanoparticle array packing and spacing using different substrates. However, ligands based on alkane chains that are most commonly used only vary the interparticle spacing in the sub-5 nm range, whereas the diblock copolymer method is restricted to substrates on which the polymer can form micelles with well-defined shape and, typically, the spacing is controllable only above 25 nm. Grafting polymeric ligands directly onto nanoparticles using functionalized endgroups is an attractive avenue for controlling the order and spacing of nanoparticle arrays in the 5–50 nm range67. In this chapter, first a simple process for fabricating wafer scale arrays in few minutes by adding excess surfactant is presented. The role of excess ligand and its implications during solvent drying and array fabrication is studied using complementary measurement techniques namely, SEM and ellipsometry, i.e., SEM provides topography of the film after drying while the ellipsometric data provides details on film formation during drying. Then, the study was extended to arrays fabricated using gold nanoparticles capped with thiol-terminated polystyrene of different molecular weights67. Large scale arrays were found to form by simple dropcasting technique by controlling the effects of solvent dewetting. Next, a systematic study involving the effect of substrate on the interparticle spacing in monolayer arrays formed by self-assembly of polymer grafted nanoparticles is discussed. Finally, a

45 process to fabricate arrays with square packing based on convective shearing at a liquid surface induced by miscibility of colloidal solution with the substrate is proposed. Fabrication of 3D aggregates of polymer-nanoparticle composite by manipulating solvent-ligand interactions is also discussed.

3.2 Large-scale self-assembly of 2D nanoparticle arrays Literature review

Significant efforts in making large scale array fabrication using ligand protected gold nanoparticles include:

3.2.1 Drop-casting colloidal solutions with excess ligand on a solid substrate Jaeger’s group has developed a simple drop-coating technique to assemble particles over cm scale on a smooth solid substrate (Fig. 3.1)28.They demonstrated their concept by drop-casting 5 nm dodecanethiol capped gold nanoparticles with excess dodecanethiol, on silicon nitride surface28,30. The success of the process depends on finding a right balance between the rate of solvent evaporation and nanoparticle diffusion. They conjectured that a decrease of the solvent evaporation rate in a controlled manner occurs with the addition of excess dodecanethiol molecules, thereby leading to large scale assembly. This technique is limited to the use of ultrasmooth solid substrates for spreading colloidal solutions.

3.2.2 Drop-casting colloidal solution on curved or flat water substrate Santhanam and coworkers have developed a simple and scalable technique by evaporating nanoparticles suspended in organic solvent on a curved water surface(Fig.

46 3.2a)29. This process was inspired by the work of Nagayama and coworkers, wherein a curved mercury surface was used to fabricate single domain of highly ordered protein crystals93. This technique in conjunction with microcontact printing enables the transfer of arrays from the water surface onto any desired substrate, overcoming the major limitation of the process of assembling particles directly on a solid substrate64. Many variations to this approach have been reported in the literature. The significant ones include: (i) Pang et al. formed free-standing films by evaporating gold nanoparticles in toluene with poly methylmethacrylate (PMMA) on flat water surface (Fig. 3.2b)94, (ii) Eah has extended the approach of particle drying from airtoluene interface to air-water interface to increase the 2D superlattice domain area (Fig. 3.2c,d)31, and (iii) Wen and Majitech have reported a large scale assembly technique by controlling the evaporation of the nanoparticles in a binary solvent mixture of toluene and hexane32. They have also reported that in the case of dodecanethiol coated gold nanoparticles, use of excess ligand is necessary to form large scale arrays, unlike the case of iron oxide nanoparticles32.

3.2.3 Doctor-blade/roll-to-roll processing of 2D assembly of particles A potential industrial technique for roll-to-roll assembly of nanoparticles is the technique of doctor-blade casting. Recently, a large-scale monolayer of 11 nm sized Wustite/Cobalt core-shell particles was fabricated. The ordering was uniform with the exception of few defects and irregularities (Fig. 3.3)95. Along similar line is the work of Malaquin and coworkers, which involves convective assembly of 100 nm gold nanoparticles or 500 nm latex particles on plasma treated PDMS substrate by controlling evaporation rate indirectly by manipulating temperature, relative humidity etc26. Self-assembly of particles on PDMS substrate enables direct transfer of arrays to any desired substrate. However, this comes with the cost of substrate compatibility, and as the PDMS is incompatible with organic solvents, this technique is limited to self-assembly of aqueous based colloids. Thus, it can be realized that there is no single robust technique for large-scale self-assembly of nanoparticles in the size range of 2-20 nm. More importantly, the process needs to be compatible with different

47

a

b

Figure 3.1: (a) Transmission electron micrograph of monolayer of dodecanethiol capped gold nanoparticles produced by drop-casting gold colloid solution (containing excess dodecanethiol) dispersed in toluene onto silicon nitride substrate. The top inset represents schematic of dodecanethiol capped gold nanoparticles. The bottom inset represents fast Fourier transform of the image highlighting good order. (b) Optical photograph of compact monolayer of gold nanoparticles formed on silicon nitride substrate. (Reproduced with permission from Bigioni et al.28 Copyright (2006) Nature Publishing Group).

a

c

b

d

Figure 3.2: (a) Optical photograph of gold nanoparticle array floating on water surface and inking the PDMS stamp pad with nanoparticle array self-assembled at the curved air-water interface (Reproduced with permission from Santhanam and Andres64. Copyright (2003) American Chemical Society). (b) Optical image of nanoparticle/poly (methylmethacrylate) film transferred to glass; the arrow represents the free-standing film (Reproduced with permission from Pang et al.94. Copyright (2008) American Chemical Society).(c) Schematic representation of monolayer formation by drop-casting dodecanethiol capped gold nanoparticles dispersed in hexane onto toluene surface (Reproduced with permission from Martin et al.96. Copyright (2010) American Chemical Society and (d) Schematic representation of array formation wherein the assembly at the air-toluene interface is transported onto an air-water interface (Reproduced with permission from Eah31. Copyright (2011) Royal Society of Chemistry).

48

Figure 3.3: Schematic representation of large scale assembly of nanoparticles through doctor-blade casting (Reproduced with permission from Bodnarchuk et al.95 Copyright (2010) American Chemical Society).

substrates. It is imperative to understand the mechanisms that dictate self-assembly process before venturing out into scale-up. The directed self-assembly scheme of alkanethiol encapsulated gold nanoparticles on a water surface is attractive due to its capability of transferring the arrays that are floating on the water surface to any desired substrates (eg. flexible or rigid) using soft-lithographic techniques. In the following section, a process to fabricate large scale arrays of dodecanethiol coated gold nanoparticles and thiol-terminated polystyrene capped gold nanoparticles will be discussed.

3.3 Large-scale self-assembly of gold nanoparticles In the study of Santhanam et al.29, it has been highlighted that the curvature of water surface is critical for obtaining large scale monolayer devoid of macroscopic holes. As an exploratory work, a rectangular cell of dimensions 2 x 10 cm is fabricated thereby ensuring the same water curvature as reported earlier (Fig. 3.4a). Monolayer of dodecanethiol capped gold nanoparticles was fabricated over large area in the rectangular cell (see Fig. 3.4b,c). However, this process is inefficient as more than 90 % of particle were lost to the Teflon walls. Further, non-uniform, local nanoscale ordering of nanoparticles was observed. Furthermore, in the Teflon cell approach, the array is pinned to the wall, thereby resulting in residual stress in the film. The stresses in the film are more evident when the capping ligand is changed from dodecanethiol to thiol-terminated polystyrene of molecular weight 3000 g/mol (Fig. 3.5a). Also,

49

a

b

c

Figure 3.4: (a) Digital photograph of gold nanoparticles dispersed in chloroform/hexane mixture on curved water surface in rectangular Teflon cell (2x10 cm). The outlined region, dotted square box used to highlight the presence of monolayer gold nanoparticle film on the water surface while the arrow indicates the particle lost to the Teflon walls. (b-c) Representative FESEM image of sub-monolayer gold nanoparticle array formed using the rectangular Teflon cell at different magnifications. Increasing the concentration should yield monolayer gold nanoparticle array.

a

b

Figure 3.5: Representative FESEM images showing polymer/solvent dewetting occurring while floating thiol-terminated polystyrene (molecular weight: 3000) capped gold nanoparticles on the water surface at (a) low nanoparticle concentration (< 10 14 particles/ml) and (b) high particle concentrations (> 1014 particles/ml). At low particle concentration, parts of arrays due to residual stresses, portion of arrays flip on top of neighbouring domains as shown in the outlined region in a (dotted circle) while at high concentrations, multilayer is formed with larger holes (bright regions in b, corresponds to holes).

50 unlike dodecanethiol capped gold nanoparticles, centimeter-scale gold nanoparticle array capped with thiol-terminated polystyrene could be obtained only by using a narrow range of concentration by using a circular Teflon cell approach. Typically, either finger-like pattern having inter-connected nanoparticle array or large circular holes surrounded by nanoparticle arrays, as shown in Fig. 3.5b, were formed. These observations suggested that stresses in the film must be avoided to form large scale array.

3.3.1 Effect of excess ligand on self-assembly of DDT capped gold nanoparticles The Teflon cell approach is not easily scalable to the wafer-scale due to losses associated with adsorption of nanoparticles on the walls, while the array formation by drop casting on a solid substrate is controlled by the topography and wetting properties of the system, and is limited by issues related to material compatibility33. In order to overcome the residual stress in the film, 0.4 mL of 5 nm dodecanethiol capped gold nanoparticles dispersed in toluene was drop-cast on free water surface in a petridish (Tissue culture, 900030A). The colloidal solution does not spread uniformly leading to multiple lens-like formations, which affected uniformity over large scale (Fig. 3.6a). This non-uniform spreading led to the formation of monolayer/multilayer of gold nanoparticle arrays as shown in Fig. 3.6b. The key to any successful self-assembly approach is to allow the colloidal solution to spread uniformly. The use of excess ligands to form ordered arrays have been reported for various systems97, with the actual mechanism yet to be identified. Some proposed explanations are: (i) the role of excess ligand acting as high boiling point solvent98, (ii) the alteration of contact line motion of solvent and dynamics of dewetting99, (iv) excess ligands form a monolayer on the subphase which enables rearrangement of nanoparticles, in addition to reducing the attractive forces between the particles97. Herein, a hybrid approach comprising of utilizing excess ligands to facilitate the spreading of colloidal solution on water surface was employed. The nanoparticle array

51

a

b

Figure 3.6: (a) Digital photograph of dodecanethiol capped gold nanoparticle array formed by drop casting gold nanoparticles dispersed in toluene without excess dodecanethiol molecules. The outlined regions (two circles) in a are used to highlight the occurrence of multiple nucleation events during monolayer formation.(b) Dual-magnification FESEM image of dodecanethiol capped gold nanoparticle array formed by drop casting gold nanoparticles dispersed in toluene without excess dodecanethiol molecules.

52

a

b

Figure 3.7: (a) Digital photograph of dodecanethiol capped gold nanoparticle array formed by drop casting gold nanoparticles dispersed in toluene with 5 µL excess dodecanethiol molecules. It can be seen that the entire petridish is filled with gold nanoparticle array. (b) Dual-magnification FESEM image of dodecanethiol capped gold nanoparticle array formed by drop casting gold nanoparticles dispersed in toluene with 5 µL excess dodecanethiol molecules.

53

a

b

Figure 3.8: Representative FESEM image of monolayer gold nanoparticle array formed by spreading toluene-hexane mixture, with SPAN 20 (sorbitan monolaurate) as surfactant on the water surface in a petridish at (a) low and (b) high magnifications.

54 was formed by drop casting 0.4 mL of 5 nm gold nanoparticles dispersed in toluene (containing 5 µL dodecanethiol) on the water surface (9 cm diameter, petridish), as shown in Fig. 3.7a. Almost, the entire surface of water was covered with a nanoparticle array within a few minutes, resulting in large area close-packed monolayer (Fig. 3.7b).To understand the role of the nature of the surfactants, several surfactants like SPAN 20, TWEEN 20, TRITON-X 100 were also tried. The water soluble surfactants, namely, Tween 20 and Triton-X 100, engulfed the nanoparticles into the water phase while the SPAN 20 resulted in large area monolayer (Fig. 3.8).

3.3.2 Ellipsometric study of formation of 2D nanoparticle arrays Nanoparticle self-assembly is a process wherein the interaction between nanoparticles results in final structure. As a result of evaporation induced dewetting, various structures such as isolated islands, labyrinthine structures, fractal structures, circular rings etc. have also been observed100-110. Thus, it becomes imperative to understand the differences in the self-assembly of dodecanethiol capped gold nanoparticles with and without addition of excess surfactant. Both modeling and experimental approaches have been employed to investigate the self-assembly of nanoparticles into 2D arrays. Rabani and co-workers have employed coarse-grained lattice models to explain how solvent fluctuations during the evaporation affect the final nanostructure, with many of their predictions yet to be corroborated101. In-situ studies using tools such as AFM, SAXS, optical microscopy have also been used to understand the nucleation and growth behavior during self-assembly of 2D nanoparticle arrays111. Very recently, Park and co-workers have used in-situ TEM measurements, to understand the drying mediated self-assembly of platinum nanoparticles in real time111. Based on their observation, they propose an important role of capillary forces developed in the evaporating solvent front, in addition to lateral immersion forces. The aforementioned techniques used for investigating self-assembly are limited to nanoscale or at the most microscopic level in their spatial resolution. Recently, an alternate measurement technique, ellipsometry (Fig. 3.9), has generated a lot of attention for in-situ characterization of polymer nanostructures, as it provides

55

Figure 3.9: Digital photograph of general ellipsometer set-up. Light emitted from the source, first passes through the polarizer (P), and the compensator (C) before it is reflected at the sample (S). The polarizer converts the unpolarized light into linearly polarized light. After reflection the light again passes a linear polarizer denoted the analyzer (A) before it reaches the detector.

information both at macroscopic and microscopic levels. In the following, the use of ellipsometry to analyze the effect of excess surfactant on the self-assembly of dodecanethiol coated gold nanoparticles at the air-water interface is presented. Dodecanethiol capped gold nanoparticles dispersed in toluene were drop-cast on the water surface, and ellipsometric data were collected at various incident angles during the evaporation of solvent.The optical images obtained during the process of array formation as the solvent evaporates for the case without excess thiol is presented in Fig. 3.10 and with excess thiol is presented in Fig. 3.11. Without the addition of excess surfactant, the initial pattern is similar to the islands of particles, which eventually aggregate to form large scale monolayer arrays with multiple voids. Also, multiple nucleation events can be observed visually during self-assembly of nanoparticles. On the contrary, self-assembly of dodecanethiol capped gold nanoparticles containing excess dodecanethiol resulted in labyrinthine structure, which eventually form large scale array without any voids. This is surprising because labyrinthine structures have been observed in simulations after complete evaporation of solvent112.

56

2 min

90 min

1

1

a

b

Figure 3.10: Optical photograph during self-assembly of dodecanethiol gold nanoparticle array without addition of excess dodecanethiol to the colloidal solution at different times after drop-casting of colloidal solution (a and b). The initial patterns represent island formation, and even after 90 minutes, arrays with microscopic holes are formed. The scale bar represents 50 µm. The regions marked 1 represent the array.

1 min

68 min

1 1

a

b

Figure 3.11: Optical photograph during self-assembly of dodecanethiol gold nanoparticle array with the addition of excess dodecanethiol to the colloidal solution at different times after drop-cast of colloidal solution (a and b). The initial patterns represent labyrinthine-like structures, enabling large scale array formation. The scale bar in a represents 50 µm, and the same magnification is used for obtaining Fig. b. The regions marked 1 represent the array. The two concentric boxes (white line in a) represent the region of interest (ROI), where the ellipsometric data is measured.

57 3.3.2.1 Modelling of optical constants

The data from ellipsometry were modeled based on air-film-substrate three layer structures with water and gold film as substrate and layer stack respectively (Model 1). As long as the films are uniform, with known n and k (refractive and extinction index) values, finding the root of the quadratic equation as explained in the experimental section of ellipsometry in the previous chapter (see sec. 2.5.6) becomes trivial. However, “the gold film layer”, considered in this study is a highly inhomogeneous medium whose refractive indices cannot be determined directly. Alternatively, to treat the problem of macroscopically inhomogeneous medium, i.e., a medium in which bulk properties such as dielectric function, elastic modulus, conductivity etc. is varying in space, an effective medium approximation is employed113,114. For example, in a metal-dielectric composite, consisting of collection of metallic and dielectric grains in a particular arrangement, the effective function is obtained by utilizing the component fractions. The derivation of effective dielectric constant involves solving equations for the electric field and dipole moment per unit volume and then volume-averaging. Depending on the direction of electric field, i.e., parallel or perpendicular, the averaging leads to optical equivalent of capacitors in parallel or in series,

  f a  a  f b b 1   f a  a  fb  b

(3.1) (3.2)

where, ε represents dielectric constant of the medium and f represents volume fraction115. The corresponding subscripts a and b represent respective properties of the two phases a and b, in the system. These expressions represent very crude approximations, because it represents two extreme cases, where either the local fields are parallel to the surface (no screening of charges) or perpendicular to the surface (maximum screening of charges). Modifications to this equation are accomplished by inclusion of a screening parameter, s, which is given by (s = 1/z–1, 0 108 s).Electrostatic Force Microscopy (EFM) measurements also show a similar variation of the phase change with time (Fig. 4.21c,d). The phase change, measured by EFM is known to be proportional to the amount of stored charge194,195 which is estimated to be 2.3 x 10-7 C/cm2 (~ 1 e-/nanoparticle) in this case. Details of the estimation of charge storage based on EFM measurements are presented in Appendix B.2. The facts that the charge retained remains almost constant after the initial decay and that the cross-sectional profiles of phase-change show the stored charges to be bound laterally exclude a charge loss mechanism involving lateral charge dispersion. A similar trend in retention behavior has been attributed to a reduction in tunneling probability caused by lowering of the in-built electric field with charge loss196; it is believed that this reasoning can also explain the trends seen in Fig. 4.21a and Fig. 4.21c,d. Significantly, endurance characterization tests show that the memory window persists even after 104 P/E cycles (Fig. 4.21e). The rapid shift in flatband voltages during the first 100 cycles is attributed to the generation of stress induced defects in the oxide stack179. Altogether, these results confirm that it is indeed the gold nanoparticles that act as charge storage nodes in these devices. Furthermore, the reproducibility of the electrical characteristics across devices attests to the uniformity of the array of nanoparticles at the macroscopic scale, which signifies that the proposed process can be easily adopted for production of future flash memory devices. XPS spectrum of gold-4f region (Fig. 4.22a) shows the presence of gold in pure metallic form (4f7/2 line at 84.1 eV). The absence of peaks corresponding to gold silicide formation (~ 85 eV), which degrades semiconductor device performance, highlights the compatibility of the proposed bottom-up process with CMOS technology, in contrast to approaches based on ion bombardment and annealing197. XPS spectrum of gadolinium 3d and 4d levels (Fig. 4.22b) confirm the presence of gadolinium in +2 charged state, while the spectrum of oxygen 1s level (Fig. 4.22c) shows the predominant presence of gadolinium hydroxide (532.4 eV) in the stack, which is attributed to the hygroscopic nature of rare earth oxides198. XPS scan of carbon 1s region (Fig. 4.22d) indicates the presence of C=O moieties (288.7 eV)199, possibly formed by the decomposition of the ligands during sputtering. The presence of sulphur peaks at 162.7 eV (Fig. 4.22e) also corroborates the earlier AFM based finding that some portions of the ligands are incorporated into the oxide stack.

120

a

c

b

d

Figure 4.20: (a) Capacitance-Voltage (CV) curve of a MOS capacitor with the gold nanoparticle array as floating gate, obtained by bi-directional sweep from inversion to accumulation. The appearance of a hysteresis loop indicates that gold nanoparticles act as charge storage nodes. The capacitor area is 7.07 x 10-4 cm2, and the measurement frequency used was 1 MHz. The inset shows an optical photograph of several MOS capacitor devices. The white dots are the aluminium top contacts. (b) Capacitance-voltage characteristics of Si/SiO2/Gd2O3 (without any gold nanoparticle). The control samples do not show any hysteresis with bidirectional voltage sweeps, indicating the absence of any interface traps.(c) Histogram of flatband voltage shifts measured from 14 devices using P/E voltages of ±7 V. The distribution of flatband voltage shifts is 0.66 ± 0.05 V. (d) Variation of flatband voltage shifts as a function of P/E voltages. The points represent the average of flatband voltage shifts and the error bars represent 95% confidence interval (n≥5).

121

a

b

d

c

e

Figure 4.21:(a) Representative charge retention characteristics of MOS capacitors physically spread across ~ cm2 area, indicating large scale uniformity of nanoparticle array. P/E was carried out at a voltage of ± 7 V, while reading voltage was set at -0.8 V. (b) Schematic energy band diagram depicting flatband condition of the floating gate structure comprising of gold nanoparticles sandwiched between silicon dioxide and gadolinium oxide. (c) EFM phase change profiles obtained at a reading voltage of 3 V and at a height of 80 nm from the surface. Charge was injected initially, by applying -6 V from 0 to 120 s, in contact mode at the spot marked “X” in AFM topography scan (top left inset). EFM phase change image (after 150 s) is shown in the top right inset, with a line marking the cross-section along which phase change profiles were measured. (d) Comparison of normalized charge stored obtained using capacitance and EFM measurements. (e) Endurance characteristic of a device indicating a stable memory window even after 104 P/E cycles (by biasing at ± 7 V).

122

a

b

c

d

e

Figure 4.22: (a) Au 4f core level XPS spectrum from a MOS capacitor with embedded nanoparticle array indicating the presence of gold in metallic form (84.1 and 87.7 eV). XPS spectra of (b) Gadolinium 3d and 4d levels, and (c) Oxygen 1s level indicate the predominant presence of hydroxide (532.4 eV), possibly as gadolinium hydroxide (Gd(OH)3) and is attributed to the hygroscopic nature of rare earth oxides. XPS spectra of (d) Carbon 1s and (e) Sulphur 2p core levels for MOS capacitors fabricated with and without (control) gold nanoparticle arrays embedded in gadolinium oxide. For MOS capacitors with the embedded nanoparticle array, the presence of C=O (288.7 eV) and trace amount of sulphur (162.7 eV) indicates that some residual ligands are trapped inside. C=O bonds are expected to have formed during the annealing steps by decomposition of the trapped ligands. In these figures, the symbols represent measured data points, the black lines represent deconvoluted peaks, and the red lines represent the overall fit. The numbers in the panels represent the peak values of the fitted curves.

123

4.5.3 Effect of interparticle spacing on memory characteristics Control over nanoparticle ordering and spacing is necessary, as these can critically affect the performance of floating gate devices167,200. Intuitively, smaller interparticle spacing would be considered best for increasing the storage density, while larger interparticle spacing leads to faster programming, as well as improved retention characteristics167. To understand the impact of interparticle spacing at the nanoscale, the Metal-Oxide-Semiconductor (MOS) capacitors having gold nanoparticle arrays with larger (4.4 and 13.3 nm) interparticle spacings, as the floating gate were fabricated. This was accomplished by simply replacing the dodecanethiol ligand used earlier by thiol-terminated polystyrene ligands with molecular weights of 3000 and 20000 g/mol; thereby, also highlighting the versatility and ease of the developed process. Representative plan-view FESEM image of a self-assembled 2D array of thiol-terminated polystyrene capped gold nanoparticles on top of a Si/SiO2 substrate is shown in Fig. 4.23. Gold nanoparticles (nominally 7 nm size) form highly-ordered hexagonal arrays with interparticle spacing of 4.4 ± 0.6 nm (Fig. 4.23a) and 13.3 ± 2.3 nm (Fig. 4.23b) for thiol-terminated polystyrene of molecular weight 3000 and 20000 g/mol respectively. Representative plan-view FESEM images of samples obtained after deposition of a 15 nm thick gadolinium oxide layer onto 2D arrays of thiolterminated polystyrene capped gold nanoparticles followed by thermal annealing at 500 °C are shown in Fig. 4.24a and Fig. 4.24b respectively. At the optimal conditions of RF sputtering found earlier, the ordering of these polystyrene-coated arrays was disturbed, but there was no evidence of particle coalescence. Electrical characterization of devices fabricated using arrays of ligand coated gold nanoparticles (both with 4.4 nm spacing and 13.3 nm spacing) did not show any hysteresis in CV plots (Fig. 4.25a,b), despite the presence of gold nanoparticles that can serve as charge storage nodes. The loss of memory window is attributed to the possible formation of conductive carbonaceous products. Such carbonaceous products are likely formed, due to the decomposition, during the annealing steps, of polystyrene thiol ligands entrapped in the oxide stack. This aspect was further investigated using EFM studies of oxide stacks formed. In EFM measurements, the3K sample exhibited a larger phase change than the sample containing DDT coated

124

a

b

Figure 4.23: Representative FESEM image of close-packed 2D array of thiol-terminated polystyrene capped gold nanoparticles (nominally 7 nm sized gold cores) with (a) 4.4 ± 0.6 nm interparticle spacing, corresponding to a number density of 7.7x1011 particles/cm2 and (b)13.3 ± 2.3 nm interparticle spacing, corresponding to a number density of 2.4x1011 particles/cm2, on Si/SiO2 prior to gadolinium oxide deposition. The insets show the corresponding histograms of interparticle spacing in these arrays.

125

a

b

Figure 4.24: (a) Representative plan-view FESEM images of samples obtained after deposition of a 15 nm thick gadolinium oxide layer onto a 2D array of thiol-terminated polystyrene capped gold nanoparticles (nominally 7 nm sized gold cores) with 4.4 nm interparticle spacing, followed by thermal annealing at 500 °C. The left hand side image was obtained using angle selective backscattered electrons, which provide information on the atomic number of the material, and the right hand side image was obtained using secondary electrons, which provide information on the surface topography. Scale bar represents 20 nm. (b)Representative plan-view FESEM images (using secondary electron detector) of samples obtained after deposition of a 15 nm thick gadolinium oxide layer onto 2D array of thiol-terminated polystyrene capped gold nanoparticles (nominally 7 nm sized gold cores) with 13 nm interparticle spacing, followed by thermal annealing at 500 °C. Scale bar represents100 nm.

126 array, indicating larger storage capacity, which is attributed to the possibility of polystyrene molecules also acting as charge storage sites (Fig. 4.26a)201. This is unlike the CV measurements in Fig. 4.25a. This difference in observing contrasting charge storage characteristics is attributed to the differences in the two measurement techniques. In standard EFM measurement (4.26a), the sample is charged by placing the tip and sample in contact, while the reading of the stored charges, with the tip biased to the opposite polarity, takes place out of contact. However, in CV measurements the probe is always in contact with the sample and thus provides a leakage path to ground for the stored charges. To replicate this situation in EFM experiments, an intermediate step was introduced, wherein the tip remains in contact with the sample for some time after charging, prior to the reading step (Fig. 4.26b).With the intermediate step of reading in contact, the sample showed negligible charge storage characteristics (Fig. 4.26c). These results are in accord with the hypothesis that a conductive pathway is formed in the gadolinium oxide that completely compromises its dielectric properties. The most likely cause of this is the formation of carbonaceous products by ligand decomposition. In order to circumvent the issue of presence of ligand, bare gold nanoparticle arrays were fabricated by using mild RF oxygen plasma etching, based on an earlier report12. A schematic of the modified process flow for silicon dioxide/gold nanoparticle array/gadolinium oxide stack formation is shown in Fig.4.27. All the device fabrication steps remain identical except for an additional step of ligand removal using RF plasma etching. Preliminary EFM characterization showed that oxygen plasma treatment on bare Si/SiO2 substrate results in deterioration of the electrical properties of the thin silicon dioxide layer (Fig. 4.28a,b). Thus, to reduce the amount of interface traps, the sample was subjected to forming gas annealing at 450 °C for 30 minutes. EFM measurements indicated that the charge retention properties were restored to its original condition (Fig. 4.28c). Thus, as a precaution to reduce the amount of interface traps, an additional step of forming gas annealing at 450 °C for 30 minutes was carried out after array deposition and RF oxygen plasma treatment. This processing step was not implemented in the case of the ligand coated nanoparticle array, as the nanoparticles coalesced upon annealing without top-oxide deposition. Again, this poor thermal stability is attributed

127

a

b

Figure 4.25: Capacitance-voltage characteristics of MOS capacitors fabricated using arrays of gold nanoparticles capped with thiol-terminated polystyrene of molecular weight (a) 3000 g/mol and (b) 20000 g/mol, embedded in gadolinium oxide on Si/SiO2. The samples do not exhibit any hysteresis despite the presence of gold nanoparticles that can serve as charge storage nodes. This could be possible only if thiol-terminated polystyrene molecules can be converted to carbonaceous material during annealing.

a

b

c

Figure 4.26: (a) Representative EFM phase profile of MOS capacitors fabricated using arrays of gold nanoparticles capped with thiol-terminated polystyrene of molecular weight 3000 g/mol embedded in gadolinium oxide on Si/SiO2. The units of x and y axis correspond to µm. Charge was injected initially by applying -6 V from 2 minutes while the reading voltage was fixed at 3 V at a height of 70 nm from the surface. The color coded scale bar represents the magnitude of phase change observed during EFM imaging. The EFM measurements are carried out in a standard mode (see schematic). (b) Schematic representation of modified EFM measurement with an intermediate step of contacting the tip with sample before reading. (c) EFM phase profile of modified procedure for imaging MOS capacitor incorporating gold nanoparticle arrays. Negligible presence of charge stored after reading indicates leakage paths for electrons to detrap through the top-oxide.

128

Figure 4.27: Schematic representation of modified process flow sheet for nanoparticle floating-gate memory device fabrication, with an additional step of RF plasma etching to fabricate bare nanoparticle array (plasma treated array) under optimized conditions. All other steps remain identical.

a

b

1 min

5 min

5 min

c

Figure 4.28: Representative EFM phase profiles of charge injection on oxygen plasma treated Si/SiO 2 substrate after (a) 1 min and (b) 5 min. Plasma conditions: power - 20 W, flow rate - 50 sccm, and time - 2 minutes. Charge was injected initially by applying -8 V from 1 minute while the reading voltage was fixed at 3 V at a height of 70 nm from the surface. The color coded scale bar represents the magnitude of phase change observed during EFM imaging. The images indicate negligible charge storage after 5 minutes. (c) Representative EFM phase profile of charge injection on oxygen plasma treated Si/SiO2 substrate, followed by forming gas annealing at 450 °C for 30 minutes. Retention of phase change in the programmed region (white region) indicates elimination of interface trap sites in the silicon dioxide film.

129

a

b

Figure 4.29: Representative plan-view FESEM images of samples obtained after deposition of a 15 nm thick gadolinium oxide layer, followed by thermal annealing at 500 °C, onto plasma treated gold nanoparticle array with (a) 4.4 nm interparticle spacing and (b) 13.3 nm interparticle spacing. The polystyrene ligands were removed by RF plasma etching under optimized conditions. The left hand side image (in a) was obtained using angle selective backscattered electrons, which provide information on the atomic number of the material, and the right hand side image was obtained using secondary electrons, which provide information on the surface topography.

130 to the presence of ligands119. The removal of polystyrene ligands by RF oxygen of all the steps involved in device fabrication (Fig. 4.29a); this is in contrast to the earlier report on dodecanethiol coated nanoparticle arrays. This can be attributed to the ligand removal during exposure of dodecanethiol coated nanoparticle array to argon plasma during RF sputtering. In the case of the longer PSSH ligands, the ligands may not be removed during gadolinium oxide deposition process and so, are entrapped within the oxide stack. Therefore, the loss of nanoparticle ordering observed in Fig. 4.24a is attributed to the presence of ligands. A reproducible flatband voltage shift of 1.26 ± 0.17 V (n = 8) and 0.53 ± 0.06 (n=3) was obtained for the devices fabricated after incorporating the ligand removal step in the process flow for arrays with 4.4 nm and 13.3 nm interparticle spacing (Fig. 4.30a-c). Interestingly, the memory window obtained for devices with 4.4 nm spacing is larger than the memory window (0.66 V) reported earlier using dodecanethiol coated gold nanoparticle arrays with 2 nm spacing, although the number density was only 7.7x1011 particles/cm2 in the present case, as opposed to the number density of 1.2x1012 particles/cm2. It is to be noted that in the case of devices using dodecanethiol capped gold nanoparticles, ligand was not removed completely. To determine the role of ligand on memory characteristics, devices were made with bare nanoparticle arrays having 2 nm spacing. This was obtained by pre-treating the array with mild hydrogen plasma under optimized conditions as reported12. The memory window is 0.7 V (Fig. 4.30d), which is equal to the earlier value of 0.66 ± 0.05 V obtained using ligand coated array. This indicates that the ligand has no significant effect on memory window. Hou et al. had modeled the interaction of 3D electrostatic fields in an array of nanoparticles and found that lower particle densities lead to faster P/E times167. However, the P/E times used in these measurements were found to result in saturation of flatband voltage shifts in all the cases, and so, changes in the rate of charge tunneling during P/E cannot account for the observed differences. At this juncture, it is surmised that with the increased number density of gold nanoparticles in the case of dodecanethiol sample, the limiting case of a continuous gold layer is approached, and therefore, it is more susceptible to defects in the oxide stack. As in the previous section, gadolinium oxide acts as the tunneling oxide for the devices reported herein. Because of the low barrier height and

131

a

c

b

d

Figure 4.30: (a) Capacitance–voltage curve of a MOS capacitor obtained by bi-directional sweep from inversion to accumulation with plasma treated gold nanoparticle array with 4.4 nm interparticle spacing as floating gate. (b) Histogram of flatband voltage shifts measured from 8 different devices. (c) Capacitance–voltage curve of a MOS capacitor obtained by bi-directional sweep from inversion to accumulation with plasma treated gold nanoparticle array with 13.3 nm interparticle spacing as floating gate. (d) Capacitance–voltage curve of a MOS capacitor obtained by bi-directional sweep from inversion to accumulation with plasma treated gold nanoparticle array with 2 nm interparticle spacing as floating gate.

132 higher defect assisted tunneling compared to silicon dioxide, this results in clockwise hysteresis loops. With further increase in interparticle spacing to 13 nm, the memory window decreased to 0.53 V. This is expected because with an increase in interparticle spacing, the density of charge storage elements decreases.

4.5.4 Effect of plasma treatment on vertical leakage Earlier it was hypothesized that the formation of conductive pathways by decomposition of entrapped ligands resulted in the loss of integrity of the oxide stack, specifically that the gadolinium oxide’s insulating properties were degraded. To verify this, EFM measurements were carried out in two different configurations on the two plasma-treated samples, using standard and modified EFM techniques, as explained in Fig. 4.26. A comparison of EFM phase maps for the four scenarios involving the two samples measured by the standard and modified EFM procedures (for arrays with 4.4 nm spacing) is shown in Fig. 4.31. Clearly, without the intermediate step, both ‘with ligand (untreated)’ as well as ‘without ligand (plasma-treated)’ samples show charge storage capabilities. Moreover, the “with ligand” sample exhibits a larger phase change, indicating larger storage capacity, which is attributed to the possibility of polystyrene molecules also acting as charge storage sites. However, with the intermediate step of reading in contact, only the device fabricated using gold nanoparticle arrays without ligands retains the charge. These results are in accord with the hypothesis that a conductive pathway is formed in the gadolinium oxide that compromises its dielectric properties. The most likely cause of this is the formation of carbonaceous products by ligand decomposition. On the contrary, characterization of retention in these devices showed that hole decay was significant, unlike devices fabricated using dodecanethiol capped nanoparticle arrays (Fig. 4.32). The quality of the RF sputtered oxide needs to be enhanced to meet standard industry requirements. The defect assisted clockwise tunneling can only be avoided by deposition of highquality, top-oxide. Other alternative strategies namely, atomic-layer deposition and ebeam evaporation were attempted to ascertain the scope of integration of these processes with bottom-up self-assembly process. These results are discussed in the subsequent sections.

133

Figure 4.31: Comparison of EFM phase profiles of standard and modified procedure for imaging MOS capacitors incorporating untreated and plasma treated gold nanoparticle arrays. The units of x and y axis correspond to µm. Charge was injected initially by applying -6 V from 2 minutes while the reading voltage was fixed at 3 V at a height of 70 nm from the surface. The color coded scale bar represents the magnitude of phase change observed during EFM imaging.

Figure 4.32: Representative charge retention characteristics of MOS capacitors fabricated with 2D arrays with different interparticle spacing of 2 nm (Au-DDT), 4.4 nm (Au-PSSH (3K)) and 13.3 nm (Au-PSSH (20K)). P/E was carried out at a voltage of ± 7 V, while reading voltage was set at -0.8 V for DDT capped array and -1.5 V for polystyrene thiol capped gold nanoparticle arrays. The hole decay was found to be significant for arrays with larger interparticle spacing.

134

4.6 Device fabrication using atomic layer deposition Based on the 2011 International Technology Roadmap for Semiconductors (ITRS) non-volatile memory technology projection table59, for 2017, the target of Effective Oxide Thickness (EOT) for tunneling and control oxide has been set at 5 nm and 10 nm respectively. This demands a precise control of not only tunneling oxide quality, but also control oxide. Atomic Layer Deposition (ALD) is an attractive avenue for fabricating next generation flash memory devices as it can control both oxide thickness and its quality, precisely at sub-nm level. Alumina (Al2O3) is the most extensively studied dielectric fabricated using ALD process. Since alumina has high dielectric constant (~ 8) when compared to silicon dioxide (~ 4), it is also one of the potential candidates for fabricating next-generation memory devices. In this section, investigations on the optimization of ALD set-up for Al2O3 deposition on nanoparticle array and the characterization of memory devices fabricated will be discussed. The effects of interparticle spacing on memory characteristics will also be presented.

4.6.1 Process optimization Conventionally, ALD process has been carried out at higher temperatures, typically, greater than 250 °C84. There are few reports of low temperature ALD deposition of alumina; where in the quality of oxide is comparable to the high-temperature grown oxide by modifying the process flow86. With the huge demand for flexible electronics, it becomes necessary and imperative to optimize deposition process on nanoparticle array at low temperature so as to have wide applicability. First, the ALD deposition process was carried out at standard condition of 250 °C in Beneq-250 reactor, with alternate pulse cycles of trimethylaluminium (TMA) and water vapor, with intermittent nitrogen purging. The pulse times were all set at 0.2 ms. At 250 °C, alumina deposition on top of arrays fabricated using dodecanethiol capped gold nanoparticles (for untreated treated sample), led to loss of ordering and aggregation (Fig. 4.33a,b); FESEM images were obtained after annealing in forming gas atmosphere at 450 °C for 20 min. This is attributed to the melting of ligand in arrays fabricated with dodecanethiol capped gold nanoparticles, which initiates around

135

a

b

Figure 4.33: Representative FESEM images of dodecanethiol capped gold nanoparticle arrays obtained after deposition of 18 nm of Al2O3 using ALD process at 250 °C at (a) low and (b) high magnification. Note the images were taken after annealing the sample in a forming gas atmosphere at 450 °C for 20 minutes.

136 200°C as reported12. More severe effect of ALD deposition on untreated array can be seen in arrays formed with thiol-terminated polystyrene capped gold nanoparticles (Fig. 4.34a). An enhancement in mobility of nanoparticle upon formation of molten polymer is attributed to be the cause for the loss of ordering by particle coalescence12. On the contrary, “bare” nanoparticle arrays with 20 nm (PSSH [MW: 20000 g/mol] capped gold nanoparticles) spacing was found to maintain particle ordering (Fig. 4.34b). The arrays with 20 nm spacing were prepared by drop-cast of colloidal solution on Si/SiO2 substrate. To mitigate the effect of temperature during deposition, the ALD process was next carried out at 150 °C, with all other conditions remaining same. The samples fabricated using dodecanethiol capped gold nanoparticles was found to maintain size and ordering for both untreated sample (Fig. 4.35a) and plasma treated sample (Fig. 4.35b). Note the FESEM images were obtained after annealing the samples at 450 °C for 20 minutes in forming gas atmosphere.

Based on the work of Sivaraman and Santhanam12, it can be presumed that the size of the particle do not play a major role during melting of monolayer protected nanoparticles, and that the major cause for disordering arises due to the melting of ligand. Korgel et al. have reported a pre-melting transition temperature of 140 °C, for dodecanethiol capped silver nanoparticles of size 7 nm202. So as a next step, the ALD deposition process was carried at 135 °C, so as to avoid disturbing the ordering of array. In these samples, both order and size were maintained for both untreated (Fig. 4.36a,c) and plasma treated samples (Fig. 4.36b,d), for samples fabricated with 2 and 20 nm spacing. The FESEM images were obtained after forming gas annealing of samples at 450 °C for 20 minutes. FESEM of cross-section sample (prepared using FIB), shows embedding of gold nanoparticle array between Al2O3 and SiO2 (Fig. 4.36e,f). This process optimization enabled us to investigate the effect of ligand, interparticle spacing on memory characteristics in addition to device performance.

137

a

b

Figure 4.34: Representative FESEM image of thiol-terminated polystyrene capped gold nanoparticle arrays (molecular weight: 20000 g/mol) obtained after deposition of 18 nm of Al 2O3 using ALD process at 250 °C, (a) without ligand removal and (b) with ligand removal using oxygen plasma treatment, followed by annealing of the sample in a forming gas atmosphere at 450 °C for 20 minutes.

138

a

b

Figure 4.35: Representative FESEM images of dodecanethiol capped gold nanoparticle arrays obtained after deposition of Al2O3 using ALD process at 150 °C, (a) without ligand removal and (b) with ligand removal using hydrogen plasma treatment followed by annealing of the sample in a forming gas atmosphere at 450 °C for 20 minutes.

139

a

b

c

d

e f

Figure 4.36: Representative FESEM images dodecanethiol capped gold nanoparticle arrays obtained after deposition of Al2O3 using ALD process at 135 °C, without ligand removal (a) and with ligand removal using hydrogen plasma treatment (b). Note, the images were taken after annealing of the sample in forming gas atmosphere at 450 °C for 20 minutes. Representative FESEM image of thiolterminated polystyrene capped gold nanoparticle arrays (molecular weight: 20000 g/mol) obtained after deposition of Al2O3 using ALD process at 135 °C, without ligand removal (c) and with ligand removal using oxygen plasma treatment (d). Note, the images were taken after annealing of the sample in forming gas atmosphere at 450 °C for 20 minutes. (e) High magnification cross-sectional FESEM image of the MOS capacitor with arrays fabricated using dodecanethiol capped gold nanoparticles as charge storage nodes. The dense packing of the array precludes visualization of individual nanoparticles. Platinum is present due to FIB sample preparation. (f)High magnification cross-sectional FESEM image of the MOS capacitor obtained using (i) backscattered electron and (ii) secondary electron detector. Contrast in backscattered electron imaging is predominantly due to the difference in atomic weight while contrast in secondary electron imaging depends on surface features.

140

a

b

Figure 4.37: (a) Capacitance-Voltage (CV) curve of a MOS capacitor without the gold nanoparticle array as floating gate obtained by depositing Al2O3 using ALD process at 135°C, obtained by bidirectional sweep from inversion to accumulation. The appearance of a hysteresis loop indicates the presence of interface traps in significant numbers. (b)Capacitance-Voltage (CV) curve of a MOS capacitor with the 2D arrays using dodecanethiol capped gold nanoparticles as floating gate obtained by depositing Al2O3 using ALD process at 135°C, obtained by bi-directional sweep from inversion to accumulation. The appearance of a hysteresis loop indicates a lot of interface traps.

Figure 4.38: Capacitance-Voltage (CV) curve of a MOS capacitor without the gold nanoparticle array as floating gate obtained by depositing Al2O3 using ALD process at 375°C, obtained by bi-directional sweep from inversion to accumulation. The appearance of a no hysteresis loop indicates high quality oxide without any trap charges.

a

b

Figure 4.39: (a) Schematic of the device structure of MOS capacitor with Si/SiO2/Al2O3. (b) Capacitance-Voltage (CV) curve of a MOS capacitor without the gold nanoparticle array as floating gate obtained by depositing Al2O3 using ALD process at 135°C followed by immediately annealing the sample at 400°C under nitrogen ambient for 20 minutes inside the ALD reactor. The absence of a hysteresis loop indicates the formation of high quality oxide without any trap charges.

141

4.6.2 Quality of Al2O3 deposited at low temperature Capacitance-Voltage (CV) measurements of the as-fabricated devices led to irreproducible hysteresis, suggesting the presence of interface traps in the MOS capacitors. A representative CV plot of MOS capacitor is shown in Fig. 4.37a. The MOS capacitor without gold nanoparticles (control sample) also showed hysteresis in CV plot (Fig. 4.37b) unlike the MOS capacitor with just tunneling oxide, SiO2. This suggests that the electrical quality of alumina film deposited at 135 °C and postannealed at 450 °C is poor. To investigate the effect of temperature on quality of Al2O3, a control sample was fabricated with Al2O3 grown at 375 °C. The CV plot showed no significant hysteresis (Fig. 4.38), indicating the absence of interface traps. With post-deposition annealing of the oxide not yielding desired improvements to the quality of oxide grown at 135 °C and also as nanoparticle arrays with smaller interparticle spacing (< 2 nm) do not withstand high temperature processing, a MOS capacitor was fabricated such that the Al2O3 deposition takes place at 135 °C and then the sample was immediately ramped up to 400 °C (maximum permissible temperature in ALD reactor), instead of exposing the sample to the atmosphere, and then followed by forming gas annealing. The rest of the device fabrication process remained the same as before. Now, the CV plot of the control sample did not show any discernible hysteresis (Fig. 4.39) similar to the MOS fabricated using Al2O3 deposited at 375 °C. Thus, even a brief exposure to atmosphere (even in clean room conditions) can lead to defect formation, when oxides are deposited at low temperatures.

4.6.3 Effect of presence of ligand and interparticle spacing on memory characteristics MOS capacitors were fabricated with both untreated and bare gold nanoparticle arrays using 2 nm (dodecanethiol) or 20 nm spacing (thiol-terminated polystyrene, MW: 20000 g/mol). CV plots showed significant memory window (4.4 V for 2 nm spacing and 6.2 V for 20 nm spacing, Fig. 4.40a,b) for devices fabricated with bare gold nanoparticle arrays, while it is absent for ligand-coated (unreliable charging for 2 nm spacing and < 1 V for 20 nm spacing, Fig. 4.40c,d) for a bidirectional sweep of ±10

142 V. CV plot of MOS capacitor with dodecanethiol capped gold nanoparticle array exhibited continuous injection of charge behavior (as there was no clear demarcation of inversion and accumulation regions) while the devices with thiol-terminated polystyrene capped gold nanoparticle array exhibited little hysteresis. This may be attributed to the possible presence of charge trapping sites, as in the devices fabricated using RF sputtering. The observation of significant memory window for devices with bare nanoparticle arrays is in concord with the earlier study of devices fabricated using RF sputtering of gadolinium oxide, in terms of the need for removal of ligand for observing memory behavior. Therefore, it was surmised that the argon plasma in the RF sputtering chamber would have removed the dodecanethiol (shorter chain length) in the case of untreated sample while in the case of ALD chamber, as there was no additional plasma, an additional step of pre-plasma treatment is needed to remove dodecanethiol ligands similar to the longer chain length molecule of thiolterminated polystyrene. The hysteresis was again observed as a clockwise loop, suggesting tunneling of electrons through the alumina and not silicon dioxide. In this study, the CV measurements were carried out by applying voltage from the bottom and measurement of current from the top, so as to avoid stray signals. The voltages are then reversed to obtain gate voltage as per convention. To rule out the possibility of error occurring due to this biasing configuration, a CV measurement of MOS capacitor fabricated with bare gold nanoparticle arrays (2 nm spacing) was carried out in the opposite manner, where in the voltage bias is applied at the top and the current is a measured from the bottom. There was not any significant difference in the hysteresis loop direction or in the value of capacitance. This rules out the possibility of error occurring due to the measurement configuration. This also suggests the presence of leakage paths in the control oxide, which will be discussed in the sec. 4.6.8. Also the capacitance value at accumulation decreases significantly from that of a blank sample. These indicate the possibility of leaky oxide film. This aspect will be further discussed in sec. 4.6.5 and sec.4.6.7.

143

a

c

b

d

Figure 4.40:Capacitance-Voltage (CV) curve of a MOS capacitor fabricated using bare gold nanoparticle arrays with (a) 2 nm interparticle spacing (DDT) and (b) 20 nm interparticle spacing (PSSH, 20000 g/mol) as floating gate. The devices were fabricated by depositing Al2O3 using ALD process at 135°C, followed by nitrogen annealing at 400°C for 20 minutes. The appearance of a hysteresis loop for devices fabricated using plasma treated nanoparticle array indicates the storage of electrons inside gold nanoparticles while the untreated array sample (c) 2 nm interparticle spacing (DDT capped gold nanoparticles) and (d) 20 nm interparticle spacing (PSSH capped gold nanoparticles) shows continues trapping/detrapping of charges due to the possible formation of carbonaceous product during annealing.

144

4.6.4 Effect of tunneling oxide thickness and control oxide thickness on memory characteristics The thickness of tunneling oxide, SiO2 was now reduced to 5 nm. As expected, CV plot showed an increase in the memory window with the reduction in tunneling oxide thickness at the same voltage, i.e., 14.8 V for 2 nm spacing and 1.3 V for 20 nm spacing (Fig. 4.41a,b). Again, significant memory window was only obtained for devices with bare gold nanoparticle arrays and not for untreated samples (Fig. 4.41c,d). Also, the control sample (without any gold nanoparticle arrays) did not show any discernible hysteresis in CV plot (Fig. 4.41e). However, device to device variability was high, as many devices did not yield any memory window. This can be attributed to the poor electrical quality of tunneling oxide. As the purge time of oxygen during dry oxidation was only 40 s, there could be a possibility of presence of pinhole defects which can act as leakage pathways for stored charges. Optimizing the quality of tunneling oxide of less than 5 nm thicknesses is a research problem on its own. On the other hand, increasing the control oxide thickness, led to lower memory window at the same voltages, which is expected as it requires higher energy to cross the barrier. However, in both cases, the hysteresis in CV plots was still clockwise, which is surprising because the physical thickness of control oxide is 3 times that of the thickness of tunneling oxide, and yet the electron tunneling during programming and erasing was found to occur only through thicker control oxide and not the thinner tunneling oxide.

4.6.5 Reduction in accumulation capacitance In addition to the larger memory window for devices fabricated using ALD deposition process when compared to RF sputtering, the major difference was in the value of capacitance at accumulation (Fig. 4.41). More interestingly, the capacitance value changed very little in the presence of ligand-coated gold nanoparticles (Fig. 4.41c,d) and more significantly for bare gold nanoparticle arrays (Fig. 4.41a,b). The results have been summarized in Table 4-3. A slight increase in the case of untreated sample can be explained due to the entrapment of organic ligands. This decrease in

145

a

b

c

d

e

f

Figure 4.41: Capacitance-Voltage (CV) curve of a MOS capacitor with the 2D bare gold nanoparticle arrays with (a) 2 nm interparticle spacing (DDT capped gold nanoparticles) and (b) 20 nm interparticle spacing (PSSH capped gold nanoparticles) as floating gate. The devices were fabricated by depositing Al2O3 using ALD process at 135°C, followed by nitrogen annealing at 400°C for 20 minutes. The appearance of a hysteresis loop for devices fabricated using plasma treated nanoparticle array sample indicates the storage of electrons inside gold nanoparticles while the devices fabricated using untreated array with (c) 2 nm interparticle spacing and (d) 20 nm interparticle spacing shows continues trapping/detrapping of charges due to the possible formation of carbonaceous product during annealing. (e) CV curve of MOS capacitor without gold nanoparticle array. (f) Schematic of the device structure of MOS capacitor.

146

Table 4-3: Effect of presence of ligand and nanoparticle density on the value of accumulation capacitance

Architecture

Control sample (without nanoparticle)

Spacing – 2 nm

Untreated array SiO2: 5 nm & Al2O3: 23 nm Al2O3: 11 nm & Al2O3: 20 nm

180 pF

207 pF

Plasma treated array 25 pF

93 pF

78 pF

28 pF

Spacing – 20 nm

Untreated array 219 pF

Plasma treated array 84pF

40 pF

63 pF

Table 4-4: Effect of presence of ligand on accumulation capacitance. Devices fabricated with PECVD also showed reduction in capacitance at accumulation.

Architecture

SiO2: 9 nm & Al2O3: 34 nm (Two step: 135°C and 375°C) SiN: 30 nm & SiO2: 7 nm

Control sample (without nanoparticle)

Spacing – 20 nm

Untreated array

Plasma treated array

120 pF

100pF

35 pF

90 pF

40 pF

60 pF

147 capacitance at accumulation was not observed for devices fabricated with RF sputtering process for deposition of top-oxide. In the case of RF sputtering process, the gadolinium oxide is sputtered directly from the target while in the case of ALD process, it involves various steps such as adsorption of precursor and purging of water vapor followed by chemical reaction. The presence of water vapor could have transformed the ligand into better trap site. Also, it has been reported that for deposition of alumina at low temperatures, sufficient time between purge cycles need to be provided so as to avoid chemical vapor deposition. So, it was decided to alter the purge timing of precursors based on the study of Groner et al86. The purge times of TMA, nitrogen, water, nitrogen were set at 0.2 s, 1 s, 0.4 s, and 2 s respectively so as to avoid chemical vapor deposition. In addition, the nanoparticle arrays are thermally stable (~ 450 °C, as tested in this study), once encapsulated in alumina matrix. In practice, the array should be stable up to the melting point of alumina, which is 2072 °C. However, as the device had aback contact made of aluminum, the practical temperature up to which the devices can be annealed is limited to the melting temperature of aluminium, which is 660 °C. As it is well known that the high temperature deposition of alumina is superior in terms of quality, it was decided to alter the deposition process in the following manner: the first 20 cycles (~ 2 nm) were deposited at 135 °C, while the remaining 320 cycles (~28 nm) were deposited at 375 °C. Even in this case, there was a significant reduction in the value of capacitance at accumulation for devices with bare gold nanoparticle arrays (Table 4-4). Also, the hysteresis loops were found to be clockwise, i.e., tunneling of charges still occurs through alumina and not silicon dioxide. To understand this effect further, memory devices were fabricated using a different process: control oxide consisting of 30 nm silicon nitride (SiN) and 7 nm silicon dioxide (SiO2) were deposited using plasma enhanced chemical vapor deposition at 375 °C. Even then, there was a reduction in the capacitance value at accumulation and the clockwise hysteresis loop was still observed (Table 4-4).

148

4.6.6 Effect of plasma treatment on device characteristics In the literature, there are reports of plasma damage leading to poor electrical quality of ultra-thin oxides. In order to test whether the pre-plasma treatment which was used for ligand removal to fabricate bare nanoparticle arrays might be the cause for a decrease in capacitance, a control MOS capacitor without gold nanoparticles was fabricated and subjected to the RF-plasma treatment as necessary to fabricate bare nanoparticle arrays at 2 nm and 20 nm spacing. CV plots neither show discernible hysteresis nor decrease in accumulation capacitance, for the control sample subjected to the RF-plasma treatment as necessary to fabricate bare nanoparticle arrays at2 nm (Fig. 4.42a) and 20 nm interparticle spacing (Fig. 4.42b). As shown in the previous sections, the CV plot of the MOS capacitor fabricated without gold nanoparticle (control sample) did not show any hysteresis indicating high-quality oxide without any interface traps (Fig. 4.42c) while the device fabricated with bare nanoparticle arrays with 20 nm spacing showed significant reduction in accumulation capacitance from ~ 70 pF to ~ 20 pF (Fig. 4.42d). These experiments showed that the mild plasma condition used for the ligand removal is not the cause for the poor electrical quality of deposited oxide film. Current-voltage (IV) measurements were carried out on MOS capacitors, both with and without gold nanoparticles (untreated and bare nanoparticle arrays (20 nm spacing), to obtain the leakage current density. IV measurements (Fig. 4.43) highlight MOS capacitors without gold nanoparticle and bare gold nanoparticle arrays have similar leakage current densities (less than nA). The untreated sample, on the contrary, was able to withstand electric fields beyond -25 V, which can be attributed to the entrapment of thiol-terminated polystyrene molecules.

4.6.7 Possible causes for observed decrease in accumulation capacitance - metal in a dielectric or area screening So far, aforementioned experiments confirm that reduction in accumulation capacitance is correlated to the presence of bare gold nanoparticle arrays in a dielectric. Also, the amount of decrease is dependent on the amount of metal

149 inclusions, i.e., the accumulation capacitance was found to decrease with an increase in particle density. This suggests that physical presence of metal might be a cause for the decrease in capacitance value at accumulation. Capacitance is defined as the ratio of product of dielectric constant and effective area to the distance between the plates. In this study, as the distance is identical in all the MOS capacitors (apart from ligand height), the difference in capacitance must arise either due to changes in effective dielectric constant or effective area. There are two different schools of thought for changes in effective dielectric constant namely, (i) reduction in dielectric constant, due to the metal inclusions in a dielectric114, and (ii) an increase in dielectric constant due to the possible formation of percolative pathways203. In the former case, various models like Bruggeman, Lorentz-Lorenz, Maxwell-Garnett etc. have been proposed. Details about these models were presented in the sec. 3.3.2.1. A three layer film structure is constructed, wherein the volume fraction of metal was presumed to affect only the intermittent layer and not the dielectric constant of top and the bottom layers. Lorentz-Lorenz model was used as the effective medium approximation in this study, which is represented below,

  1    1   1   1  f a  b   f a  a  2   2   2  a   b 

(4.2)

where, ε represents dielectric constant of the medium and f represents volume fraction. The corresponding subscript a and b represents respective properties of the two phases in the system. The effective dielectric constant changes from an initial value of 8 to 4.1 and 7.4 for an interparticle spacing of 2 nm and 20 nm respectively. This leads to capacitance values of 148 and 181 pF respectively, which do not match with the measured values of 25 and 84 pF respectively (Table 4-5). Estimation of capacitance based on the decrease in the area, obtained by extracting the free space in 2D nanoparticle array, leads to decrease in capacitance to values of 52 and 167, which again differs from the experimental value. Hence, the explanation of decrease in accumulation capacitance must lie elsewhere.

150

a

b

c

d

e

Figure 4.42: Capacitance-Voltage (CV) curve of a MOS capacitor without the gold nanoparticle array subjected to the RF-plasma treatment as necessary to fabricate bare nanoparticle arrays at (a) 2 nm and (b) 20 nm spacing as floating gate. The absence of hysteresis loop and no significant reduction in capacitance at accumulation indicate plasma treatment does not cause substrate damage, similar to CV curve of a MOS capacitor without gold nanoparticle array, not subjected to any plasma treatment (c). (d) CV curve of a MOS capacitor fabricated with bare arrays with 20 nm spacing as floating gate shows significant reduction in accumulation capacitance. This experiment clearly highlights the mild plasma treatment does not cause substrate damage and the hysteresis and decrease in accumulation capacitance arise due to the presence of bare nanoparticle arrays.(e) Schematic of the device structure of MOS capacitor.

151

Figure 4.43: Current-Voltage (IV) characteristic of MOS capacitor fabricated with (i) SiO 2 and Al2O3 (MOS blank), (ii) untreated and (iii) bare 2D arrays of nanoparticle with 20 nm spacing. This experiment suggests mild plasma treatment used for fabricating bare arrays does not affect the leakage current.

Table 4-5: Comparison of experimental and predicted accumulation capacitance values due to the either change in effective dielectric constant (metal in a dielectric) or effective area screening (based on nanoparticle density).

Control sample (without nanoparticle) 184.5 (180)

Spacing – 2 nm (q = 0.32) ε = 4.1 A = 0.53

148 pF(25pF) 52 pF(25 pF)

Spacing – 20 nm (q = 0.035) ε = 7.4 A = 0.95

181 pF(84 pF) 167 pF(84 pF)

152

4.6.8 Leakage characteristics in Al2O3 Until now, the direction of hysteresis loop observed in all the MOS capacitors has been clock-wise, suggesting injection of electrons through Al2O3 instead of SiO2. This is surprising because even with 3 times the thickness of tunneling oxide, the electrons were found to tunnel through the thicker oxide. For a particle of mass, m and energy, E, the tunneling probability, T of the potential barrier with length, L and height, U is given by156

 2mU  E    T  exp  2 L  h2  

(4.3)

using Wentzel–Kramers–Brillouin (WKB) approximation. For 10 nm SiO2 and 30 nm Al2O3, the tunneling probability is computed to be exp(-41x) and exp(-70x) respectively, assuming effective mass =0.2 times the mass of an electron204. This suggests electron tunneling must occur through SiO2 rather than Al2O3. However this is not observed in the experiments. Another possibility is the defect assisted tunneling through alumina. To test this hypothesis, it was decided to measure the leakage currents across the alumina deposited using ALD process. For this, the alumina was deposited on 5.6 nm sputtered gold porous film using ALD (Fig. 4.44). After fabricating MOS capacitors using this substrate, a part of the substrate is immersed in 10% KOH solution to etch the alumina. This results in electrical access to the Metal-Insulator-Metal (MIM) structure, which can be utilized to measure IV characteristic across the alumina film alone. As surmised earlier, the leakage currents were found to be of the order of µA and more interestingly in some of the devices, the breakdown voltage was found to be less than 2 V (Fig. 4.45). Note, the thickness of alumina was 34 nm. Also, a MOS capacitor was fabricated with 10 nm Al2O3 as tunneling oxide deposited at 375 °C. Surprisingly, the breakdown voltage of high-temperature grown oxide is very low when compared to literature report (Fig. 4.46). This demands a serious inspection of

153

a

b

c

Figure 4.44: (a-b) Representative FESEM images of gold porous film of thickness 5.6 nm at different magnifications. (c) Representative FESEM image of Al2O3 deposited on gold porous film, followed by annealing the sample at 450 °C for 20 minutes, under forming gas atmosphere.

the precursor materials used, and the reactor set-up, which is beyond the scope of this work. On the other hand, one way to reduce the leakage paths is to increasing the annealing temperature. As expected, after annealing the sample at 850 °C, no significant reduction in accumulation capacitance was observed for all the MOS capacitors (with and without nanoparticle), as shown in Fig. 4.47. This suggests that contact at the gold-alumina interface could have been an issue. FESEM images of the sample are shown in Fig. 4.48, which shows that the ordering is mostly maintained even after annealing at 850 °C, due to encasement in alumina. Now, this enables to investigate the important effect of interparticle spacing effect on memory characteristics.

154

Figure 4.45: Current voltage characteristic of MIM structure (with gold porous film-Al2O3-Aluminium architecture), showing high leakage currents (~ µA) in addition to very low breakdown voltage of ~ 1V. Note the substrate was annealed at 450 °C for 20 minutes in forming gas environment.

Figure 4.46: Current voltage characteristic of MOS structure (with Aluminium-Si-Al2O3-Aluminium) with ALD process carried out at 375 °C. Even in this structure, very high leakage current and low breakdown voltage was observed.

155

a

b

c

d

e

f

Figure 4.47: (a) Capacitance-Voltage (CV) curve of a MOS capacitor with pure tunnelling oxide structure. The MOS capacitor shows no significant hysteresis, indicating negligible traps in the oxide layer. (b)Capacitance-Voltage (CV) curve of a MOS capacitor with tunnelling oxide, SiO2 and control oxide, Al2O3 structure. The MOS capacitor shows no significant hysteresis, indicating negligible traps in the oxide layer. Capacitance-Voltage (CV) curve of a MOS capacitor with the 2D(c) untreated array and (d) bare gold nanoparticle array (with 2 nm spacing) as floating gate. The devices were fabricated by depositing Al2O3 using ALD process at 135°C, followed by nitrogen annealing at 400°C for 20 minutes. The sample was subsequently annealed under forming gas environment at 850 °C for 20 minutes. The hysteresis loop is observed for devices fabricated using bare nanoparticle array (0.9 V) and not for untreated array. This can be attributed to the possible conversion of entrapped organic ligand into carbonaceous product during annealing. Capacitance-Voltage (CV) curve of a MOS capacitor with the 2D (e) untreated array and (f) bare gold nanoparticle array (with 20 nm spacing) as floating gate. The sample was subsequently annealed under forming gas environment at 850 °C for 20 minutes. The hysteresis loop is observed for devices fabricated using bare nanoparticle array (0.9 V) and not for untreated array. This can be attributed to the possible conversion of entrapped organic ligand into carbonaceous product during annealing.

156

a

b

c

d

Figure 4.48: Representative FESEM images of (a) untreated and (b) bare nanoparticle array with 2 nm spacing (DDT capped gold nanoparticles), after ALD deposition of Al2O3 followed by forming gas annealing at 850°C for 20 minutes. Representative FESEM images of (c) untreated and (d) bare nanoparticles with 20 nm spacing (using thiol-terminated polystyrene capped gold nanoparticles of molecular weight: 20000 g/mol), after ALD deposition of Al2O3 followed by forming gas annealing at 850°C for 20 minutes. In all the figures, left and right images were obtained using backscattered electron and secondary electron detector respectively.

157

4.6.9 Effect of interparticle spacing Memory devices were fabricated with different interparticle spacing using dodecanethiol capped gold nanoparticle arrays (~ 2.2 ± 0.8 nm spacing) and thiolterminated polystyrene capped gold nanoparticle arrays with different molecular weights namely, 7000, 10000 and 20000 g/mol (7.9 ± 1.8 nm, 8.9 ± 1.4 nm and 19.7 ± 3.3 nm spacing respectively), as shown in Fig. 4.49. The bare nanoparticle arrays were fabricated by exposing to plasma treatment under optimized conditions based on the report12. To remove the sulphonates after hydrogen plasma treatment of dodecanethiol capped gold nanoparticle arrays, the substrate was rinsed with water followed by ethanol. FESEM images after plasma treatment are shown in Fig. 4.50ad.The ALD process was employed to deposit alumina with thickness of 34 nm. This was followed by forming gas annealing at 850 °C for 20 minutes. More interestingly, the CV characteristic of MOS capacitors showed an optimum interparticle spacing for observing larger memory window (Fig. 4.51). The memory window followed the trend of 8.9 nm (5.2V)> 7.9 nm (4.5 V)> 2.2 nm (2.9 V)> 19.7 nm (0.8V)with varying interparticle spacing. Values in the bracket represent the memory window. This trend is similar to the observation made in devices prepared using RF sputtering, reported earlier in this chapter. Lower memory window observed for the case of larger interparticle spacing can be explained based on the reduction in number density of particles. On the contrary, lower memory window with highest particle density can arise due to two situations, (i) more leakage paths through the oxide or (ii) inability to inject more electrons into the nanoparticle due to columbic repulsion amongst the neighboring particle. At this juncture, it is not possible to discriminate between these two explanations. The variation in inversion capacitance is attributed to the presence of residual back oxide that was found to be difficult to remove, due to the size of the sample, using buffered HF solution. This modification was adopted so as to anneal the sample at 850 °C, and then deposit the back contact made of aluminium.

158

a

b

c

d

e

f

g

h

Figure 4.49: Representative FESEM image of 2D arrays of gold nanoparticles capped with (a) dodecanethiol, (b) thiol-terminated polystyrene (molecular weight: 7000 g/mol), (c) thiol-terminated polystyrene (molecular weight: 10000 g/mol) and (d) thiol-terminated polystyrene (molecular weight: 20000 g/mol). The respective histograms of interparticle spacing are presented in Fig. e-h.

159

a

b

c

d

Figure 4.50: Representative FESEM image of bare gold nanoparticle array with interparticle spacing of (a) 2.2 nm (dodecanethiol capped gold nanoparticle array), (b) 7.9 nm (thiol-terminated polystyrene (molecular weight: 7000 g/mol) capped gold nanoparticle array), (c) 8.9 nm (thiol-terminated polystyrene (molecular weight: 10000 g/mol) capped gold nanoparticle array), and (d) 19.7 nm (thiolterminated polystyrene (molecular weight: 20000 g/mol) capped gold nanoparticle array); after plasma treatment under optimized conditions.

160

a

b

c

d

Figure 4.51: Capacitance-Voltage (CV) curve of a MOS capacitor with the bare nanoparticle array with interparticle spacing of (a) 2.2 nm (dodecanethiol capped gold nanoparticle array), (b) 7.9 nm (thiolterminated polystyrene (molecular weight: 7000 g/mol) capped gold nanoparticle array), (c) 8.9 nm (thiol-terminated polystyrene (molecular weight: 10000 g/mol) capped gold nanoparticle array), and (d) 19.7 nm (thiol-terminated polystyrene (molecular weight: 20000 g/mol) capped gold nanoparticle array); after plasma treatment under optimized conditions. The devices were fabricated by depositing Al2O3 using ALD process at 135°C, followed by nitrogen annealing at 400°C for 20 minutes. The sample was subsequently annealed under forming gas environment at 850 °C for 20 minutes. The largest memory window was obtained for arrays prepared using thiol-terminated polystyrene molecules of molecular weight of 10000 g/mol, an intermittent spacing. The memory window followed the trend of 8.9 nm (5.2 V)> 7.9 nm (4.5 V)> 2.2 nm (2.9 V)> 19.7 nm (0.8 V) interparticle spacing. Values in the bracket represent the memory window.

161

a

b

c

d

Figure 4.52:Capacitance-Voltage (CV) curve of a MOS capacitor with the interparticle spacing of (a) 2.2 nm (dodecanethiol capped gold nanoparticle array), (b) 7.9 nm (thiol-terminated polystyrene (molecular weight: 7000 g/mol) capped gold nanoparticle array), (c) 8.9 nm (thiol-terminated polystyrene (molecular weight: 10000 g/mol) capped gold nanoparticle array), and (d) 19.7 nm (thiolterminated polystyrene (molecular weight: 20000 g/mol) capped gold nanoparticle array); after plasma treatment under optimized conditions. The devices were fabricated by depositing Al 2O3 using ebeam process at room temperature. The sample was subsequently annealed under forming gas environment at 450 °C for 20 minutes. The largest memory window was obtained for arrays prepared using thiol-terminated polystyrene molecules of molecular weight of 10000 g/mol, an intermittent spacing, similar to earlier ALD process. The memory window followed the trend of 8.9 nm (1.5 V)> 7.9 nm (0.6 V)~ 2.2 nm (0.6 V)~ 19.7 nm (0.6 V) interparticle spacing. Values in the bracket represent the memory window.

162 Table 4-6: Summary of effect of interparticle spacing in nanoparticle arrays and process parameters on memory window of the floating gate memory devices presented in this chapter Batch

Mean interparticle

Top oxide (deposition

Memory

Number.

spacing

route)

window

Thickness

(average value) 1

2 nm

Gd2O3 (RF sputtering)

0.66 V

SiO2: 10 nm Gd2O3: 15 nm

2

4.4 nm

Gd2O3 (RF sputtering)

1.26 V

13.3 nm

Gd2O3 (RF sputtering)

0.53 V

2 nm

Al2O3 (ALD)

4.4 V

SiO2: 10.1 nm Al2O3: 18.7 nm

3

20 nm

Al2O3 (ALD)

6.2 V

2 nm

Al2O3 (ALD)

14.8 V

SiO2: 5.1 nm Al2O3: 23.4 nm

4

20 nm

Al2O3 (ALD)

1.3 V

2.2 nm

Al2O3 (ALD)

2.9 V

SiO2: 9 nm Al2O3: 34 nm

5

7.9 nm

Al2O3 (ALD)

4.5 V

8.9 nm

Al2O3 (ALD)

5.2 V

19.7 nm

Al2O3 (ALD)

0.8 V

2.2 nm

Al2O3 (e-beam)

0.6 V

SiO2: 9 nm Al2O3: 37 nm

7.9 nm

Al2O3 (e-beam)

0.6 V

8.9 nm

Al2O3 (e-beam)

1.5 V

19.7 nm

Al2O3 (e-beam)

0.6 V

163

4.7 Device fabrication using e-beam evaporation An alternative, attractive room temperature deposition process, namely, electronbeam evaporation was also used to deposit Al2O3. Arrays with four different interparticle spacing as described in the previous section for ALD deposition was employed. The thickness of the oxide deposited was 37 nm. After the deposition of alumina on top of nanoparticle arrays, the sample was annealed under forming gas environment at 450 °C for 20 minutes. The capacitance-voltage of the MOS capacitor devices are shown in Fig. 4.52. Similar to the ALD process, an optimum in spacing for obtaining larger memory window was found, i.e., 2D arrays fabricated with thiolterminated polystyrene (molecular weight: 10000 g/mol) capped gold nanoparticles yielded the largest memory window. The memory window followed the trend of 8.9 nm (1.5 V)> 7.9 nm (0.6 V) ~ 2.2 nm (0.6 V) ~ 19.7 nm (0.6 V) with varying interparticle spacing. Again, lack of high-purity source material for e-beam evaporation led to the observation of clockwise hysteresis loops.

4.8 Summary A CMOS-compatible process for the realization of a high-density (>1012 particles/cm2) floating gate memory device using self-assembled 2D arrays of the gold nanoparticle as storage nodes was developed. The fabricated MOS capacitors, using RF sputtering process for top-oxide deposition, show excellent retention (decay time constant > 108 s) and endurance (>10000 P/E cycles) characteristics as well as a narrow distribution of the memory window across devices. This optimized process can be readily adapted for fabricating multilevel flash memory devices at the sub-20 nm node, as iteasily provides nanoscale control of the particle size and interparticle spacing. Experimental results show an optimum interparticle spacing (arrays with thiol-terminated polystyrene capped gold nanoparticles of molecular weight: 10000 g/mol) for obtaining larger memory window. Table 4-6 provides a summary of the effect of various process parameters on memory window. The quality of top-oxide deposited, using atomic layer deposition and electron-beam evaporation needs to be further improved, so as to enable commercial production of next-generation devices.

164

Chapter 5 PEG capped monodisperse gold nanoparticles 5.1 Introduction The fabrication of monodisperse metal nanoparticles in colloidal form has garnered a lot of interest due to its potential applications ranging from electronics to biotechnology10,11. In particular, biological applications demand precise control of particle size due to the influence of size on both bio-distribution and clearance205-208. Also, the first step in fabrication of ordered array by self-assembly involves the synthesis of monodisperse nanoparticles – a fundamental building block in nanoscience and nanotechnology. Nanoparticle synthesis in liquid phase can be classified

into

two

approaches

namely,

aqueous61,209

and

organic

phase

synthesis210,211. Typically, aqueous phase synthesis result in particle size distribution that exhibit high polydispersity while organic phase synthesis involve high temperature processing to achieve lower polydispersity. Even today, many modifications to classical citrate protocol are being proposed to improve the monodispersity of particles synthesized62,212. On the other hand, organic based approaches involve an additional step of either size-selective precipitation or digestive ripening of colloidal solution for obtaining monodisperse nanoparticles from the assynthesized polydisperse particles213-215. Size-selective precipitation involves tedious processes such as solvent/non-solvent dispersion/redispersion of colloidal sample, while digestive ripening process involves high temperature, and long durations of refluxing an organic colloidal solution with excess ligands. Recently, digestive ripening procedure has also been applied to aqueous gold colloids using mercaptopropanediol as ligand molecules34. Until now, for nanostructure fabrication using selfassembly, either nanoparticles are synthesized in the organic phase or need to be phase-transferred from aqueous to organic media. Thus, it is of technological importance and environmental benefit to synthesize and self-assemble monodisperse nanoparticles directly from aqueous medium.

165 This chapter first presents the results for synthesizing and assembling monodisperse metal nanoparticles directly from aqueous medium based on well-established Turkevich protocol61, by the addition of excess polyethylene glycol (PEG) molecules.Next, it discusses various factors involved in observing ripening of particles.

5.2 Effect of addition of thiol-functionalized PEG molecules on citrate-stabilized metal nanoparticles

5.2.1 Gold 5 mL of freshly prepared citrate capped gold nanoparticles of size 14.0 ± 3.1 nm (µ ± σ; Fig. 5.1a,b) were mixed with 5 µL of thiol-functionalized polyethylene glycol (PEG-thiol, molecular weight: 356 g/gmol) and aged overnight (~ 16 hours)at 25 °C. The molar ratio of PEG-thiol to gold corresponds to 10. To image the particles after PEG capping, a 5 µL aliquot was drop-cast on silicon substrate, and surprisingly, the size was found to be 11.0 ± 0.7 nm (Fig. 5.1c,d). Digestive ripening of particles using excess ligands by heating, for one or two hours, the colloidal solution near the boiling point of the solvent is well established214. However, here, it was observed that focusing of particle size occurs at room temperature. Thermogravimetric analysis (TGA) data of PEG-thiol capped gold nanoparticle suggests degradation of PEG-thiol around 300 °C (Fig. 5.2). Citrate capped gold nanoparticles show 16.8 % weight loss at 600 °C. This value is higher than the reports of Manson et al.216 (~10 %) and Larson-Smith and Pazzo217 (~ 5 %). This variation in weight loss can be attributed to the possibility of excess bound water. For PEG-thiol capped gold nanoparticles, the weight loss was found to be 79.8 % at 600 °C. Also, the major weight loss was found to occur between 280 and 320 °C, correlating with the thermal degradation temperature of polymer216.

166

a

b

c

d

Figure 5.1: Representative FESEM image of (a) citrate capped gold nanoparticles (b) thiolfunctionalized polyethylene glycol capped gold nanoparticles and their respective size histograms (c and d). The molar ratio of PEG-thiol to gold is 10:1 and the PEG-thiol capped gold nanoparticles were drop-cast on silicon substrate after ageing the colloidal solution for ~16 hours at 25 °C.

167

Figure 5.2: Thermogravimetric analysis (TGA) of citrate capped and PEG-thiol capped gold nanoparticles.

a

b

Figure 5.3: Representative DLS size histograms of (a) citrate capped and (b) PEG-thiol capped gold nanoparticles. The molar ratio of PEG-thiol to gold is 10:1 and the PEG-thiol capped gold nanoparticles were analyzed after ageing the colloidal solution for ~16 hours at 25 °C.

168

Dynamic Light Scattering measurements (DLS) corroborate the FESEM result of decrease in polydispersity before and after addition of PEG-thiol (Fig. 5.3). The DLS mean hydrodynamic diameters before and after PEG-thiol addition were 17.8 ± 2.0 nm and 17.1 ± 0.2 nm. As the chain length of the PEG-thiol molecule is 2.8 nm, the gold core diameter, after addition of PEG-thiol is deduced to be 11.5 nm (=17.1 – 2x PEG-thiol chain length), which is in concord with FESEM data (Fig. 5.1c,d). To investigate whether this is a generic phenomenon or system specific, the process of addition of PEG-thiol was extended to other metal nanoparticles, namely, citrate capped- palladium, silver and platinum nanoparticles.

5.2.2 Palladium The use of concentration ratios of sodium citrate to palladium (5.2, similar to gold nanoparticle synthesis) did not result in nanoparticle formation. In fact, there was no notable color difference even after heating the solution for four hours. Based on the report62, that reverse addition of gold salt increases the reaction rate, an experiment was carried out using the reverse mode of addition, i.e., adding palladium chloride to boiling citrate solution. However, no discernible reaction occurred, and the color of the solution remained pale yellow. The UV-Vis spectrum of the solution (Fig. 5.4) after reaction showed a peak at 250 nm, which is characteristic of palladium chloride in water218. Citrate capped palladium nanoparticles were then successfully synthesized based on the earlier report of Turkevich and Kim219. Briefly, 4.125 mg of palladium chloride (PdCl2) mixed with 0.5 mL of 1 N HCl was diluted to 25 mL, followed by the addition of 50 mL of 34 mM sodium citrate solution. The solution was heated to its boiling point in a beaker, which was attached with a reflux condenser, for 4 hours (Fig. 5.5a) After 1 hour, the initial pale yellow color started to change to dark brown (Fig. 5.5b). With continued heating, the color of the solution turned to intense brown. The particles were found to be highly polydispersed, with size ranging from 5 to 30 nm (Fig. 5.6a). After additing excess PEG-thiol (PEG to Pd molar ratio of 10) and

169 heating to 74 °C (using a Peltier bath) for duration of 9 days, the size of the nanoparticle decreased to 4.7 ± 1.1 nm (Fig. 5.6b). The UV-Vis spectra showed significant spectral change immediately after the addition of PEG-thiol (Fig. 5.6c,d). The palladium nanoparticles do not show any absorption features, while after PEGthiol addition, a couple of peaks at 270 nm and 310 nm appeared immediately, which is attributed to the possible formation of palladium complexes218. X-ray photoelectron spectroscopic (XPS) analysis suggests the presence of palladium in Pd2+ form (may be palladium oxide) based on the location of binding energy peak of Pd 3d5/2 and 3d3/2 peaks at 337.8 and 342.9 eV respectively (Fig. 5.7), whereas pure palladium peak corresponds to 335.2 and 341.1 eV respectively220. However, the size and polydispersity of the nanoparticles were found to increase upon further storage, for another 15 days at room temperature, to around 39.2 ± 7.9 nm (Fig. 5.8).

5.2.3 Silver Citrate-capped silver nanoparticles were synthesized based on the report of Jin et al221. A mixture of 1 mL of 30 mM sodium citrate and 2 mL of 5 mM silver nitrate solution was made up to 98 mL using DI water. Upon boiling of this mixture, 1 mL of freshly prepared, ice-cold 50 mM NaBH4 solution was added. The reaction was carried out under argon atmosphere in a Schlenk flask. Upon ageing at 4 °C (inside refrigerator), the as-synthesized citrate capped silver nanoparticles transformed into rods, cubes etc.,in concord with an earlier report222 (Fig. 5.9). So, in this study, only freshly prepared citrate capped nanoparticles were used. With the addition of PEG-thiol to the freshly prepared citrate-capped silver nanoparticles, the color of the solution transformed from pale yellow to transparent, after being stored for 1 day at 25 °C. As there was no evidence of precipitate in the vial or any aggregated particles in FESEM image of the sample obtained by dropcasting the colloidal solution on to a silicon substrate, it was speculated that the nanoparticles were transformed into small nanoclusters (Fig. 5.10). The film-like features seen in the FESEM image is attributed to the presence of an organic layer, possibly excess PEG-thiol (Fig. 5.10).The size of these clusters could not be determined using FESEM.

170

Figure 5.4: UV-Vis spectrum of a palladium chloride solution that was boiled with citrate solution.

a

b

Figure 5.5: Digital photographs taken during the process of synthesizing citrate-capped palladium nanoparticles, (a) at the start of the reaction and (b) after boiling for 1 hour.

171 a

b

c

d a

PEG-thiol capped Pd NPs Citrate capped Pd NPs

Figure 5.6: Representative FESEM image of (a) citrate capped palladium nanoparticles and (b) aged PEG-thiol coated gold nanoparticles after 9 days. (c) UV-Vis spectra of palladium nanoparticles before and after addition of PEG-thiol. (d) Digital photograph of citrate capped palladium nanoparticles before (left) and after addition of PEG-thiol (right). 5µL of PEG-thiol was added to 5 mL of colloidal solution and the PEG-thiol capped gold nanoparticles were drop-cast on silicon substrate after ageing the colloidal solution for 9 days at 74 °C.

Figure 5.7: X-ray photoemission spectra of PEG-thiol capped palladium nanoparticles, after ripening.

172

Figure 5.8: Representative FESEM image of aged PEG-thiol capped palladium nanoparticles after 15 days from particle ripening.

173

a

b

Figure 5.9: Representative FESEM images of citrate capped silver nanoparticles, after ageing at different magnifications.

174

Figure 5.10: Representative FESEM image of PEG-thiol capped silver nanoparticles, after ageing at 25 °C for 1 day.

Figure 5.11: UV-Vis spectra of (a) citrate capped silver nanoparticles and (b) aged PEG-thiol capped silver nanoparticles after 1 day at 25 °C.

175 The UV-Visible spectra of citrate capped silver nanoparticles before and after addition of PEG-thiol molecules show a significant decrease in absorbance at 400 nm after PEG addition, which can be attributed to either a plausible change in number concentration of particles or a change in their size(Fig. 5.11). Photoluminescence spectra of PEG-thiol capped silver nanoparticles shows greater enhancement when compared to original citrate-capped silver nanoparticles (Fig. 5.12)223. The particles were excited at 320 nm and the spectra were collected from 350 nm to 600 nm. The citrate capped silver nanoparticles (>5 nm) showed very little photoluminescent activity, the response being similar to water (Fig. 5.12). The dotted lines represent the repeated measurement without altering the parameters, clearly showing that there is negligible effect of laser excitation on the citrate colloids. PEG-thiol capped silver nanoclusters, showed photoluminescent enhancement and their spectra were blueshifted, which can be attributed to the metal-ligand charge transfer absorption as reported earlier223. Note, both citrate capped silver nanoparticles and PEG-thiol capped silver nanoparticles were used as is without any dilution. However, with repeated measurement (dotted line and dashed line), photobleaching was found to occur in the case of PEG-thiol capped nanoparticles. Photobleaching is a phenomenon where in the absorption of photon degrades the polymer and causes a reduction in photoluminescent intensity with time. Use of a fresh solution of PEG-thiol capped silver nanoparticles yielded the original result (unbleached). All these experiments suggest the presence of silver nanoclusters as hypothesized using UV-Vis spectra. These observations are similar to the recently observed effect of treatment of mercapto-succinic acid (MSA) with citrate capped silver nanoparticles above 70 °C by Dhanalakshmi and co-workers224. However, in this study the etching effect was also observed at room temperature. With an increase in temperature to 74 °C, particle etching occurred within two hours. Also, similar reports of ligand based etching of silver nanoparticles have been reported using Glutathione225. In addition to various reports of thiol based etching of silver nanoparticles with different backbone, Radziuk et al., have reported a reduction in polydispersity of silver nanoparticles by the addition of PEG molecules followed by thermal treatment at 90 °C for 2 hours226. The ratio of PEG to silver used was 0.1 in their study, which is much lower than that used in our study.

176

Figure 5.12: Photoluminescence spectra of citrate capped silver nanoparticles and PEG-thiol capped silver nanoparticles, excited at 320 nm. Upon continued excitation, the intensity of the PEG-thiol capped sample decreases, which is attributed to photobleaching. The dotted lines represent the repeated measurement of the corresponding sample without altering the parameters.

Figure 5.13: Representative FESEM image of citrate capped platinum nanoparticles

177

5.2.4 Platinum Citrate capped platinum nanoparticles were synthesized based on a modification of the recipe of Harriman et al227. The standard mode of addition of citrate to platinum precursor was reversed, so as to facilitate faster reduction and lower polydispersity. This mode of addition was adopted based on the report of Sivaraman et al. 62 25 mL of water was mixed with 1 mL of sodium citrate solution (1% w/v) and allowed to boil for few minutes. After that, 0.33 mL of H2PtCl6solution (1% w/v) was injected into the boiling solution. The amount was chosen such that the molar ratio of platinum to citrate is equal to that used for the case of gold, i.e., 5.2. The color of the solution changed from yellow to light brown in 30 minutes. The reaction was allowed to continue for another 3 hours. The FESEM image shows clustering of particles (Fig. 5.13), which were difficult to resolve individually. 5 µL of PEG-thiol was added to 5 mL of freshly prepared citrate stabilized platinum nanoparticles. The sample was kept in a water bath maintained at 74 °C using Peltier stage. In contrast with the earlier systems, the size of citrate capped platinum nanoparticles was found to increase after PEG-thiol addition (molar ratio of PEG-thiol to platinum was 10) and ageing for 12 days (~ 6.3 ± 1.2 nm, Fig. 5.14). X-ray photoelectron spectroscopic (XPS) analysis confirms the presence of platinum in metallic form based on the location of binding energy peaks of Pt 4f5/2 and 4f7/2 peak at 71.2 and 74.6 eV, respectively 228(Fig. 5.15). The presence of platinum in oxide form with peaks at 72.6 and 76.3 eV was also observed. This is most likely due to oxidation of the atoms at the surface of the nanoparticles. All these experiments suggest that addition of PEG-thiol molecules to citrate capped metal nanoparticles results in changes to the particle size distribution, the extent of which depends on the nature of the metal. This point will be revisited in a later section (see sec.5.16) of this chapter. To further understand the phenomena involved, a detailed study of the factors responsible for the reshaping of gold nanoparticles was then carried out.

5.3 Effect of vacuum on PEG coated gold nanoparticles When the PEG-thiol coated gold nanoparticle samples were imaged in TEM, a significant fraction of the particles were odd shaped and polydispersed, similar to as-

178 synthesized citrate capped gold nanoparticles with a size of 17.6 ± 2.7 nm (Fig. 5.16). A region of thick organic film in the shape of open-ended wrench can be seen in Figure 5.17. Efforts to image at higher magnifications in this region were not successful, due to charging effects. Furthermore, domains of monodisperse particles (size: 15.9 ± 1.4 nm) that appear to embedded in an organic film were sometimes observed (Fig. 5.18). Further, the particles (region marked with an oval) were found to be more spherical than the particles found away from this region (Fig. 5.19). In fact, the contrast between the particles and the substrate was found to be less in the region marked with an oval, as compared to regions away from this. Mixed regions of particles assembled in both ordered and disordered manner were also observed (Fig. 5.20). Also, the particle size is significantly different to the sizes obtained using FESEM, and this discrepancy cannot be simply attributed to the higher resolutions associated with TEM, as the FESEM also has the required resolving power at the 10 nm lengthscale. In addition, sub-2 nm clusters were also observed (Fig. 5.21). These nanoclusters were observed at very low magnifications, so it can be surmised that these may not be due to the impingement of electron beam (Fig. 5.21a). In fact, the sizes of these clusters increased during imaging at higher magnifications similar to the recent report of impact of electron beam on particle size229 (Fig. 5.21b). The change in nanoparticle size, shape and polydispersity observed in TEM measurements suggests two possibilities: (a) higher vacuum in the TEM chamber (~10-9mbar)when compared to FESEM (~10-6 mbar) leading to ligand desorption, (b) bombardment with higher electron energies (~ 200 kV) as compared with FESEM (~ 10 kV) leading to particle re-shaping. To discern between these two factors, it was decided to carry out a couple of experiments: (i) to place the sample in TEM chamber for one hour without opening the column valve that connects the electron gun to the chamber (chamber vacuum: ~10-9 mbar) followed by imaging it in FESEM (chamber vacuum: ~10-6 mbar; lower by 3 orders of magnitude) and (ii) to place the sample in FESEM for prolonged periods of time (chamber vacuum: ~10-7 mbar, which is achieved after few hours). The colloidal solution was drop-cast on silicon nitride TEM substrate to mimic the nature of the silicon substrate used in FESEM imaging, as well as to facilitate handling in both TEM and FESEM.

179

a

b

Figure 5.14: Representative FESEM image of PEG-thiol coated platinum nanoparticles, after ageing for 12 days at different magnifications.

Figure 5.15: X-ray photoemission spectra of PEG-thiol capped platinum nanoparticles, after ageing for 12 days

a b

Figure 5.16: (a) Representative TEM image of PEG-thiol capped gold nanoparticles and (b) corresponding size histogram. The molar ratio of PEG-thiol to gold was 10 and the PEG-thiol capped gold nanoparticles were drop-cast on TEM grid after ageing the colloidal solution for ~1 day at 25 °C.

180

Figure 5.17: Representative TEM image during imaging thiol-functionalized PEG capped gold nanoparticles, highlighting thick organic film in the shape of open-ended wrench.

a b

Figure 5.18: (a) Representative TEM image of thiol-functionalized PEG capped gold nanoparticles near organic film and (b) respective size histogram.

181

a

b

Figure 5.19: TEM images of thiol-functionalized PEG capped gold nanoparticles, highlighting low polydispersity, obtained at different locations. Images show particles with two contrasts, suggesting possible presence of organic film. The particles inside the region marked with a dotted oval (b) were found to be more spherical than the particles found away from this region

182

Figure 5.20: Representative TEM image of thiol-functionalized PEG capped gold nanoparticles, highlighting particles assembled in ordered and disordered fashion, next to each other.

a

b

Figure 5.21: Representative TEM images of PEG-thiol-functionalized gold nanoparticles obtained at (a) low and (b) high magnification. These sub-2 nm clusters were found to increase in size during the course of imaging.

183

a

b

c

d

e

f

Figure 5.22: Representative FESEM images of thiol-functionalized PEG capped gold nanoparticles at different magnifications before (a-c) and after treating the sample in TEM chamber for 1 hour, without switching on the electron beam (d-f). The regions marked 1 and 2 (in a) represent nanoparticle array and substrate respectively. Clearly, the higher vacuum in the TEM chamber has affected the size distribution of the particles.

184 b c

a

c

d

Figure 5.23: Representative FESEM images by drop-cast of either (a) concentrated PEG-thiol capped gold nanoparticle solution (~ 20 times the original concentration) or (b) freeze-drying the standard colloidal solution was found to result in formation of supracrystals, a 3D assembly of PEG-thiol coated gold nanoparticles. The dotted rectangles (in a) represent Moiré patterns due to two ordered pattern, namely, ordered array and SEM scan line. The regions marked 1 and 2 represent monolayer array (shown in d) and supracrystal (shown in c).

185

a d

b d

c d

d d

e d

f

Figure 5.24: (a-f) Representative FESEM images of thiol-functionalized PEG capped gold nanoparticles at different magnifications after leaving the sample imaged in Fig. 5.23c inside the FESEM chamber for 1 day. The dotted rectangles highlight regions of different contrast due to the possible partial removal of PEG molecules under vacuum. Clearly, the transformation of supracrystals into linear assemblies, as well as changes in size distribution can only be attributed to the effect of vacuum on the organic film.

186 Before the sample was placed in high vacuum chamber in TEM, the sample was imaged in FESEM which highlights highly, ordered array of monodispersed nanoparticles (Fig. 5.22a-c). Typically, the particles were found in patches of highly ordered arrays, which appear to behave like an organic film, in that their spacing and ordering were altered upon prolonged e-beam exposure as if a thin piece of cotton cloth was being pulled apart in tension (also seen in Fig. 5.23d).Notably, the particles exhibit a monodisperse size distribution. After placing in the high vacuum chamber of a TEM for 1 hour without opening the column valve that connects electron gun to the chamber, the sample was again analyzed in FESEM. The images clearly indicate both loss of size and ordering after being placed in high vacuum chamber (Fig. 5.22df).Interestingly, drop-casting either concentrated PEG-thiol capped gold nanoparticles (~ 20 times the original concentration) or freeze-drying the colloidal solution was found to result in formation of supracrystals, a 3D assembly of PEG-thiol coated gold nanoparticles (Fig. 5.23a-c). When this sample was placed in FESEM chamber for prolonged periods of time (~ 1 day, without e-beam exposure), transformation of supracrystals into linear assemblies of particles was observed along with a broadening of particle size distribution (Fig. 5.24). Cheng and Cao230, in a recent simulation study, have theoretically found that beyond a critical concentration of nanoparticles in solution, PEG grafted gold nanoparticles aggregate into nanowires.

5.4 Evolution

of

particle

size

distribution

in

chloroform/water mixture PEG molecules have good solubility in chloroform, and soa small aliquotof chloroform (0.2 mL) was added to ripened, aqueous PEG-thiol coated gold nanoparticles (5 mL). Upon heating, the color of the colloid changed from red to blue to light brown. Upon subsequent cooling, the color of the solution reverted back to red. FESEM images indicate the formation of rods and other odd-shaped particles at the higher temperature (the colloidal solution was drop-cast when the color of the solution was blue), and the size change was reversible upon cooling (Fig. 5.25). This clearly suggests that PEG grafted gold nanoparticle can reshape and resize depending on the solvent and temperature conditions.

187

a d

b d

c d

d d

Figure 5.25: Representative FESEM images of (a) PEG-thiol coated gold nanoparticles, (b) PEG-thiol coated gold nanoparticles (5 mL) after addition of small amounts of chloroform (0.2 mL), (c) boiled PEG-thiol coated gold nanoparticles dispersed in water and chloroform, and (d) the colloidal solution after being cooled back to room temperature.

188

5.5 Role of excess PEG-thiol molecules on ripening of gold nanoparticles To understand further the role of free PEG-thiol molecules, freshly prepared citrate capped gold nanoparticles (14 ± 3.1 nm) were mixed with different amounts of PEGthiol molecules, namely, PEG/Au ratio of 0.15 (which is equivalent to surface atoms of gold present in the solution, assuming a uniform size of 14 nm), 1 and 10. After ageing the sample for 1 day at 25 °C, the mean sizes were found to be 13.3 ± 2.1 nm, 13.0 ± 0.8 nm, and 11.0 ± 0.7 nm for PEG/Au ratio of 0.15, 1 and 10 respectively (Fig. 5.26a-f). Low magnification FESEM images at different PEG/Au ratios clearly show that Moiré patterns appear for arrays with higher amounts of PEG-thiol, i.e., for the case of 1 (Fig. 5.27). Moiré patterns appear due to the interference between two regular patterns. In this study, SEM scan lines act as one of the reference pattern, while the ordered nanoparticle array forms the second pattern. Such Moiré patterns have been reported for dodecanethiol capped gold nanoparticles31. Similar Moiré patterns were also reported for block copolymer monolayer structure231.

5.6 Removal of excess PEG molecules from ripened nanoparticles In order to further understand the role of excess free PEG in solution in enhancing the monodispersity of PEG-thiol grafted gold nanoparticles, monodisperse PEG-thiol coated gold nanoparticles (molar ratio of PEG to gold of 10)were boiled with small aliquots of hydrogen peroxide (2 mL peroxide was added to 5 mL of colloid), as peroxides are known to degrade PEG molecules232. Excess sodium citrate (2 mL of 1 % (w/v) solution) was also added to the solution to ensure the presence of adequate stabilizers. After a couple of minutes of boiling, an aliquot from the solution was drop-cast on silicon substrate. Circular domains (marked 1) of 2D nanoparticle arrays were found in a layer of organic film (marked 2); the region marked 3 corresponds to silicon (Fig. 5.28). Such contrast in FESEM images of molecular films have been ascribed to the differing electron affinities of the materials involved233. A higher

189 magnification FESEM image of the regions marked as 1 and 2 in the Fig. 5.29 a and b revealed highly ordered and disordered nanoparticle arrays respectively. The mean size of the particles in the ordered region was found to be 8.6 ± 1.7 nm (Fig. 5.29c,d), which is about 2 nm lower than that of the mean size of the particles, before adding peroxide. In the circular domains two characteristic features were observed, namely, (1) white bright spots (marked with a circle) and (2) ripple like features (marked with an oval), as shown in Fig. 5.30. To understand those features further, highmagnification images were taken using both secondary electron and backscattered electron detector. The angle selective backscattered (AsB) electron image and secondary electron (SE) images shed light on a couple of issues, (a) presence of organic film around the gold nanoparticles, as ripple in SE image was not translated in AsB image (Fig. 5.31a) and (b) possible presence of gold as some gold-rich complex as the bright spots encircled in Fig. 5.31b are visible in AsB as such. These results indicate that, given the absence of any noticeable deposition of metallic gold on the containers, the gold atoms have been redistributed onto other nanoparticles or form complexes upon oxidation of PEG/organic film by peroxide. After the solution was boiled for 20 minutes, SEM and DLS results (Fig. 5.32) indicate that the particles have fused and that they are electrostatically stabilized (zeta potential changed from 0.2 ± 2 mV to -46 ± 3 mV). These results show that the PEG molecules have been oxidized by peroxide boiling for 20 minutes and that the citrate molecules are adsorbed onto the particle surfaces. Further, the primary particles have clustered into small aggregates. DLS size histogram also corroborate these findings, as the size distribution shows two modes in the 10-100 nm size range that are attributed to the primary particles and their agglomerated clusters respectively. Alternatively, instead of boiling ripened PEG-thiol coated gold nanoparticles with hydrogen peroxide, excess PEG molecules were removed by high speed centrifugation or membrane filtration. Excessive washing resulted in irreversible aggregation of particles. These observations are consistent with the recent reports of Shon et al.234,235, wherein when the excess PEG molecules were removed from thesolution, the stability of nanoparticles decreased and also there were signs of aggregation. Interestingly, these complexes are labile enough to desorb under ambient

190 a d

b d

c d

d d

e d

f

Figure 5.26: Representative FESEM images of PEG-thiol coated gold nanoparticles with the ratio of PEG-thiol to Au, (a) 0.15, (b) 1, (c) 10, and (d-f) their respective size histograms. PEG-thiol coated gold nanoparticles were drop-cast on silicon substrate, after ageing the colloidal solution for 1 day at 25 °C.

191 a d

b d

192

c d

Figure 5.27: Representative low magnification FESEM images of PEG-thiol coated gold nanoparticles with the molar ratio of PEG-thiol to Au of, (a) 0.15, (b) 1, (c) 10, highlighting Moiré patterns. The dotted rectangles (in b) represent regions of Moiré pattern. Moiré patterns can appear due to the interference between two regular patterns. Here, SEM scan lines act as reference pattern, while the ordered nanoparticle array forms the second pattern. PEG-thiol capped gold nanoparticles were dropcast on silicon substrate after ageing the colloidal solution for 1 day at 25 °C.

193

Figure 5.28: Representative low-magnification FESEM image of PEG-thiol coated gold nanoparticles mixed with excess hydrogen peroxide and sodium citrate and boiled for 2 minutes. The region marked 1, 2 and 3 correspond to circular domains of 2D nanoparticle array, organic film and bare silicon respectively.

194 a d

b d

c d

d d

Figure 5.29: (a) Representative FESEM images of PEG-thiol coated gold nanoparticles mixed with excess hydrogen peroxide and sodium citrate after boiling for 2 minutes, showing regions of (b) highly disordered nanoparticle array (region marked 2 in a), (c and d) highly ordered nanoparticle array. The left hand side and right hand side images in a and b were obtained using backscattered electron and secondary electron detector respectively. (c) High magnification FESEM image of circular regions reveal highly ordered array with their corresponding size histogram (d).

Figure 5.30: Representative FESEM image of PEG-thiol coated gold nanoparticles mixed with excess hydrogen peroxide and sodium citrate after boiling for 2 minutes.

195 a d

b d

Figure 5.31: Representative high-magnification FESEM images, obtained using both secondary electron and backscattered electron detector, of boiled PEG-thiol coated gold nanoparticles mixed with excess hydrogen peroxide and sodium citrate at different locations; (a) region of organic PEG film (as region marked by the oval in secondary electron detector was not translated onto the backscattered electron image) and (b) region of gold-rich complex (as regions marked with circles in secondary electron image also appear as very bright spots in backscattered electron image).

196 b d

a d

Figure 5.32:(a) Representative FESEM image of thiol-functionalized PEG capped gold nanoparticles boiled with a mixture of hydrogen peroxide and sodium citrate after boiling for 20 minutes, suggesting fusion of nanoparticles due to removal of ligands and the (b) representative DLS size histogram.

a d

b d

Figure 5.33: (a) Representative FESEM images of (a) PEG-thiol coated gold nanoparticles derived from 5 nm citrate capped gold nanoparticles62 and (b) PEG-thiol coated gold nanoparticles derived from ~ 40 nm citrate capped gold nanoparticles (Frens method 68).

197 conditions from a substrate after several months. On the contrary, ageing of colloidal solution for several months did not result in any changes to particle size distribution, as the results of DLS measurements were identical to the initial value of 17.1 ± 0.2 nm. All these results suggest that excess ligand molecules/complexes are in equilibrium with the adsorbed molecules/complexes on the surface of gold nanoparticles, and as far as this equilibrium is maintained, particles exhibit very narrow size distributions. Very recently, Tantakitti and co-workers, have reported nanoscale clustering of carbohydrate thiols in mixed self-assembled monolayers on gold236. They have also reported that during storage, sugar and oligo(ethylene glycol) thiols can laterally diffuse, facilitating intermolecular interactions between the components leading to nanoscale clustering. Further, AFM measurements suggest gel like structure. Hence, it can be surmised that, excess PEG molecules are essential even when the particles are drop-cast on the substrate. With the removal of PEG molecules either through vacuum or ageing in the ambient atmosphere, the particles restructure and reshape. However, similar rapid ripening process was neither observed starting from smaller nanoparticles62 (< 5nm) synthesized through citrate (Fig. 5.33a) nor starting from larger-size particles, synthesized using Frens protocol68 (Fig. 5.33b), despite waiting for reasonable time intervals (4 days) or using higher temperatures (~ 100 °C for 10 minutes). As the pH of the colloidal solution containing larger-sized particles was acidic and also the amount of sodium citrate added was minimal, it could have affected the ripening process. So, additional sodium citrate was added to this solution to make the amount equivalent to the standard Turkevich protocol. Evidence for ripening process was seen through the formation of small sized particles (Fig. 5.34), although complete ripening was not observed even after a matter of weeks. It is surmised that enough steric repulsion might not be in place to overcome the van der Waals attraction between particles, as the FESEM images reveal flocculation of particles unlike uniform 2D assembly observed in smaller size particles earlier. So, the use of longer chain PEG-thiol ligand molecules, viz. 1000 and 5000 g/mol molecular weight molecules was attempted to provide appropriate steric repulsion.

198

Figure 5.34: Representative FESEM image of ~ 40 nm citrate capped gold nanoparticles after changes in pH through the addition of sodium citrate (such that the final molar ratio of citrate to gold reached 5.2), highlighting ripening process through the formation of small sized particles using two detectors, namely, backscattered electron (left) and secondary electron (right) detectors.

Figure 5.35: Representative FESEM image of citrate capped gold nanoparticles after immediate addition of PEG-thiol molecules of molecular weight 5000 g/mol; highlighting possible ripening process through the formation of etching of larger particles (dotted rectangle).

199

5.7 Effect of molecular weight of PEG-thiol In view of varying the final size of ripened particles from 11 nm, longer chain length PEG-thiol molecules (molecular weight: 1000 and 5000 g/mol) were employed. The molar ratio of PEG to gold was maintained at 10. Standard citrate colloid of mean size 14 nm was employed for this study. Even after two days, significant reduction in particle size was not observed. Nevertheless, with PEG 5000, drop-casting of colloidal solution, immediately after PEG-thiol addition showed formation of extremely small particles of sizes less than 5 nm (Fig. 5.35). This suggests that particle ripening should be indeed possible. With appropriate activation, size focusing of particles could occur as observed for lower molecular PEG-thiol (356 g/gmol). Experimental observations show particle sizes of ~ 10.7 ± 2.3 nm and 14.1 ± 2.4 nm nm for PEG-thiol molecules of molecular weight 1000 and 5000 g/mol respectively (Fig. 5.36a-d). DLS measurements confirm changes in both the particle size and reduction in polydispersity with boiling (Fig. 5.36e,f). However, the extent of monodispersity is poorer when compared to particle ripened using PEG-thiol molecules of molecular weight 356g/mol. At this juncture, it is speculated, distribution in molecular weight of PEG-thiol used could be the reason for observed variation in polydispersity using different molecular weight PEG molecules. Surprisingly, zeta potential measurements showed –14 ± 3.3 mV and –16.8 ± 2.0 mV for particles capped with PEG-thiol 1000 and 5000 molecules, thereby suggesting that some of the citrate molecules are still adsorbed and retain their charge even after addition of the PEG-thiol molecules. To understand the effect of ligand length, experiments were carried out where in PEG-thiol molecules of molecular weight 5000 (molar ratio of PEG-thiol to gold: 10) was added to ripened gold nanoparticles capped with PEG-thiol molecules of molecular weight 356 and vice-versa. Addition of PEGthiol 5000 molecules to highly monodispersed PEG-thiol 356 molecules capped gold nanoparticles of size (11.0 ± 0.7 nm) resulted in loss of monodispersity and also an increase in mean size to 13.2 ± 1.9 nm (Fig. 5.37a), while the addition of PEG-thiol 356 molecules to particles capped with PEG-thiol 5000 capped gold nanoparticles resulted in a reduction of mean size to 11.9 ± 1.6 nm (Fig. 5.37b). On the contrary, with further increase in chain length (molecular weight: 20000 g/mol) for PEG-thiol

200 coated gold nanoparticles (molar ratio of PEG-thiol to gold: 10), there were clear indications of linear assembly of nanoparticles (Fig. 5.38b), unlike the preference of hexagonal close-packed assembly observed earlier, for lower chain length PEG ligands. The size distribution was narrower as compared to the original citratestabilized colloidal solution, but was not as monodisperse as those particles obtained using PEG 356.

5.8 Absence

of

gold

thiolates

in

PEG-thiol

coated

nanoparticles Based on the results of peroxide boiling experiment in section 5.6, it can be surmised that size focusing of particles might occur through etching of larger particles into smaller particles via the formation of soluble gold complexes. To test the hypothesis of presence of gold complexes, Matrix Assisted Laser Desorption Ionization (MALDI) measurements were carried out. MALDI spectra showed no characteristic peaks corresponding to gold thiolates for PEG-thiol coated gold nanoparticles at two different ratios of PEG/Au, 0.15 and 10 (Fig. 5.39). Negative and positive ion MALDI spectra were collected for both the samples for m/z values ranging from 50 to 100000. For the m/z values ranging from 50 to 1000, there were 4 multiplet peaks occurring with a periodicity of 224 (Fig. 5.39a,b). The molecular weight of PEG-thiol is 356 g/gmol, while that of single gold atom attached to PEG-thiol corresponds to 553 (=197+356). So the peaks correspond to neither pure PEG-thiol nor gold-PEG-thiol complex but rather to the molecular mass of 224, which corresponds to Sinapic acid, the matrix material (Fig. 5.39c-f). Furthermore, there were no significant peaks observed for m/z values ranging from 1000 to 100000 in both the samples (Fig. 5.40). This suggests the absence of gold complexes or gold nanoclusters less than 2 nm size237 in the aged sample. Therefore, the bright films (i.e., electron rich domains) seen in Fig. 5.31b could have occurred during the process of drying on a substrate, as part of the sample preparation of precursor due to the effect of peroxide. Although there is no evidence for presence of gold complex in the aged solution, the possibility of their formation immediately after PEG-thiol addition cannot be ruled out.

201 a d

b d

c d

d d

e d

f

Figure 5.36: Representative FESEM images of ripened gold nanoparticles using PEG-thiol molecules having molecular weight of (a-b) 1000 g/mol, and (c-d) 5000 g/mol. The insets in b and d represent respective size histograms. DLS measurements of PEG-thiol coated gold nanoparticles with and without boiling for different molecular weights, (e) 1000 g/mol, and (f) 5000 g/mol. The molar ratio of PEG-thiol to gold was maintained 10 for both the cases and the sample was boiled for 10 minutes before analysis.

202

a d

b d

Figure 5.37: (a) Representative FESEM images of addition of PEG-thiol 5000 molecules to ripened particles using PEG-thiol 356. (b) Representative FESEM images of addition of PEG-thiol 356 molecules to ripened particles using PEG-thiol 5000. The insets represent respective size histograms.

203

a d

b d

Figure 5.38: Representative FESEM images of ripened gold nanoparticles using PEG-thiol molecules of molecular weight 20000 g/mol at (a) low and (b) high magnification. The molar ratio of PEG to gold is 10 and the sample was aged for 1 day at 25 °C. The regions marked 1 and 2 (in a) represent closepacked and non-close packed arrays respectively.

204

a d

b d

c d

d d

e d

f

Figure 5.39: Low m/z range, MALDI spectra of PEG-thiol capped gold nanoparticles ionized both negatively and positively, at different ratios of PEG-thiol to gold of 0.15 and 10 plotted with different grid lines (a-f). The molecular weight of Sinapic acid (matrix), PEG-thiol, Au1-PEG-thiol is 224, 356 and 553 respectively. It can be seen that peaks coincide well with the Sinapic acid and not with gold thiolates or any complex.

205

a d

b d

c d

d d

Figure 5.40: High m/z range, MALDI spectra of PEG-thiol capped gold nanoparticles ionized both negatively and positively, at different ratios of PEG-thiol to gold of 0.15 and 10.

a d

b d

Figure 5.41: (a) Time-dependent average DLS photon count intensity after addition of PEG-thiol (PEG/Au: 10) and aged the sample at 74 °C. Extraction of time constant (~ 4.1 hours) based on the time-dependent average DLS photon count intensity at 74 °C. There is a sudden drop in DLS photon count intensity from 40 to 30 kcps after the addition of PEG-thiol to citrate capped gold nanoparticles. (b) Time-dependent average DLS photon count intensity with respect to temperature, after addition of PEG-thiol (PEG/Au: 10). It can also be that there is a significant batch to batch variation.

206

5.9 DLS photon counts In addition to particle size and polydispersity, Dynamic Light Scattering measurements can provide vital information on the number density of particles, provided the scattered signal is only from the single scattering events. The average photon count intensity (I) is dependent on both number density (n) and particle size (d), as any slight variation in either of these parameters can significantly affect the photon counts as

I α d6

(5.1)

In the current study, there was a sudden drop in DLS count rate, immediately after the addition of the excess PEG-thiol molecules to the citrate capped gold nanoparticles (PEG/gold: 10) as shown in Fig. 5.41a. Remarkably, the DLS count rate was also found to be dependent on temperature. With an increase in temperature from 25 °C to 74 °C, the ageing time for observing monodispersity decreased from ~16 hours to 4.1 hours, based on the time dependence DLS photon counts (obtained by fitting a time constant to the time-dependence of DLS photon count at 74 °C) as shown in Fig. 5.41a. However, there was significant batch to batch variation in photon counts of citrate capped gold nanoparticles, i.e., the initial photon count of citrate capped gold nanoparticles varied from 30 to 45 kcps between batch to batch as seen from the values at time = 0 in Fig. 5.41b. To overcome the issue of the batch to batch variation, PEG capping experiments were conducted from the same batch of citrate capped gold nanoparticles (as the ageing of citrate capped gold nanoparticles for 1 day did not affect their size or photon counts), as shown in Fig. 5.42. Still a significant jump was observed from 23.6 kcps to 42.6 kcps and 56.9 kcps after addition of PEG-thiol molecules (molar ratio of PEG to gold corresponds to 10) at temperatures of 25 °C and 74 °C respectively. Neither drift in the equipment at 25 °C nor at 74 °C can be attributed to this sudden jump in photon counts. Also, at 74 °C, there was a steady increase of photon counts to 37 kcps after 5 hours in pure citrate capped gold nanoparticles, which can be attributed to the slight fusion of particles as determined from changes in particle size distribution (Fig. 5.42). As the mean hydrodynamic size remains constant before and after PEG-thiol addition, only an increase in number density of particles (leading to multiple scattering events) can explain the sudden drop

207 of photon counts. The increase in number density of particles can be explained based on the argument that the gold nanoparticles capped with different chain length have different photon count rate, even though the gold core sizes are comparable (molecular weight: 356, 1000 and 5000 g/mol yielded 10.8, 39.8 and 43.2 kcps). This suggests that while considering the intensity, the hydrodynamic diameter must be taken into account not just the core diameter, even though scattering from pure polymers is negligible in comparison to metal atoms. Further, the values are obtained after diluting the sample with water to ensure that multiple scattering effects are avoided and not for time dependence measurements. As SAXS intensity at low q values before and after addition of PEG-thiol molecules remains the same, the amount of gold atoms contained within the sample is deemed to be equal (Fig. 5.43). The SAXS equipment does not have the resolution to go to q values closer to zero, and so parameters such as number density, mean size and polydispersity cannot be estimated with precision and so, are not reported here.

5.10 PEG-citrate complexation Zeta potential measurements showed a decreasing trend from -54, -35, to -0.2 mV as the PEG-thiol/Au molar ratio changed from 0, 0.15 to 10 respectively. This suggests that at PEG-thiol to gold ratio of 10, the particles are sterically stabilized, unlike the other two cases which are electrostatically stabilized. To understand this aspect further, Nuclear Magnetic Resonance (NMR) measurements were carried out. 1H NMR of free PEG-thiol in solution shows characteristic peak values at 2.7 (-SH, labelled 1),2.9 (-S-CH2, labelled 2), 3.3 (-OCH3, labelled 3), 3.5 and 3.6 (hydrogen from ethylene glycol chain, labelled 4 and 5) ppm respectively (Fig. 5.44). 1H NMR of citrate capped gold nanoparticles shows a characteristic peak at 2.6 ppm corresponding to –CH2 protons of citrate (Fig. 5.45a)238. 1H NMR of PEG-thiol capped gold nanoparticles shows all the peaks corresponding to PEG-thiol molecules (Fig. 5.45b). Surprisingly, there is an additional peak attributed to bound citrate that is also present at 2.6 ppm. This is counterintuitive because zeta potential after addition of PEG-thiol molecules was found to be only -0.2 mV, so a complete replacement of

208

Figure 5.42: (a) Time-dependent average DLS photon count intensity with respect to temperature for the same batch of citrate capped gold nanoparticles, after addition of PEG-thiol (PEG/Au: 10). It can also be seen that the photon count intensity of citrate capped gold nanoparticles at 25 °C does not increase much while at 74 °C, the increase is attributed to the fusion of particles.

Figure 5.43: Small-Angle X-ray Scattering (SAXS) spectra of citrate capped and PEG-thiol capped gold nanoparticles. As the intensity at low q values remains the same, the amount of gold contained in the two samples can be considered equal.

209 citrate ions with PEG molecules was expected. However, 1H NMR data clearly shows this is not the case. To understand this aspect further, Raman measurements were carried out in solution. The signals arising from the solution are weak as the amount of molecules present in the sample is equivalent to the amounts present in the final colloid. The Raman spectra of pure PEG-thiol in water did not show any characteristic peak (Fig. 5.46). Typically, the carboxylate peaks of citrate appear between 1380 to 1600 cm-1 239. More specifically, the symmetric and asymmetric stretching modes of carboxylate anion appear in the range of 1414-1425 cm-1 and 1560-1580 cm-1 respectively. The intensity ratio and the wavenumber difference between symmetric and asymmetric modes can be used to determine the structure of adsorbed carboxylate ion. The wavenumber difference between the ionic carboxylate symmetric and asymmetric stretching is 145-164 cm-1. If the wavenumber difference between stretching modes is higher than the ionic value, then the presence of citrate as unidentate complex is indicated, otherwise a bidentate complex is expected to be formed. From the intensity ratio between different stretching modes, one can estimate relative amounts of bidentate and unidentate complexes present in the sample. The Raman spectra of pure sodium citrate in water showed the presence of peaks at 1400, 1440, 1558 and 1566 cm-1, as shown in Fig. 5.46. This suggests the presence of carboxylate anion in ionic state, as the wavenumber difference between the symmetric and asymmetric modes is around 158-166 cm-1. After the addition of PEG-thiol to sodium citrate in water, the difference between the symmetric and asymmetric modes decreased to 85 cm-1, suggesting the presence of carboxylate as bidentate complex (as shown in Fig. 5.46). The Raman spectra of citrate capped gold nanoparticles showed peaks at 1558 and 1604 cm-1, suggesting only asymmetric stretching mode of carboxylate ion (Fig. 5.46). Also, as the intensity ratio between I(1604) to I(1558) is greater than 1, only bridging bidentate complexes of carboxylate ions are formed. This observation is in contrast to the observation reported earlier in the literature, where in the 60 nm citrate capped gold nanoparticles gave rise to Raman signals that showed the presence of unidentate and ionic complexes, in addition to bidentate complexes239,240. The observation of only bidentate complex in the present study can be attributed to the reduced size of the nanoparticle (~ 15 nm), which has higher radius of curvature, thereby favoring the formation of bidentate carboxylate complex..

210

Figure 5.44: 1H Nuclear Magnetic Resonance (NMR) of PEG-thiol dispersed in D2O. The numbers embedded with the NMR data correspond to the respective numbers marked in PEG molecular structure (as inset).The characteristic peak values at 2.7 (-SH, labelled 1), 2.9 (-S-CH2, labelled 2), 3.3 (-OCH3, labelled 3), 3.5 and 3.6 (hydrogen from ethylene glycol chain, labelled 4 and 5) ppm.

211

a d

b d

Figure 5.45: (a) 1H Nuclear Magnetic Resonance (NMR) of citrate capped gold nanoparticles dispersed in D2O. The number embedded with the NMR data corresponds to the respective numbers marked in PEG molecular structure (as inset). (b) 1H Nuclear Magnetic Resonance (NMR) of PEG-thiol capped gold nanoparticles dispersed in D2O. The inset represents the full spectrum scan. The peak at 2.7 ppm suggests that there is a significant presence of bound citrate present in the PEG capped gold nanoparticle.

Intensity (a.u.)

212

Figure 5.46: Raman spectra of PEG-thiol, sodium citrate and PEG-thiol-sodium citrate solutions in water. The sample concentrations correspond to the respective amounts in PEG-thiol capped gold nanoparticles. Raman spectra of citrate capped gold nanoparticles, with the characteristic carboxylate peaks at 1558 and 1604 cm-1; thereby suggesting the presence of only bridging bidentate complex. Raman spectra of PEG-thiol coated gold nanoparticle indicating negligible presence of Au-S peak ~ 300 cm-1.

213 Raman spectra of PEG-thiol capped gold nanoparticle showed the presence of bound citrate (1560 cm-1) and a weak signal corresponding to Au-S at ~ 300 cm-1. This result is in concord with NMR measurements (Fig. 5.45b) that citrate is still bound to the gold surface and that the gold-thiol bonds are not present in significant numbers.

5.11 Role of PEG backbone Raman measurements and 1H NMR spectra of PEG-thiol capped gold nanoparticles suggest that the presence of PEG backbone is critical, while thiol group is not critical for the observed size changes. This is in contrast to all earlier reports on “digestive ripening” of gold nanoparticles that were shown to occur through the formation of gold-thiolate complexes213. To test this hypothesis, an additional experiment was performed where PEG molecules of molecular weight 300 g/mol without any thiol functionalization were added to a solution of freshly prepared citrate capped gold nanoparticles and aged for 1 day. The molar ratio of PEG to gold was maintained at 10. A low magnification FESEM image of the PEG capped gold nanoparticles shows Moiré patterns similar to the PEG-thiol capped gold nanoparticles (Fig. 5.47). After focusing on the particles at high magnification (200kx) and then obtaining image at low magnification (19 kx) shows image distortion due to the possible presence of organic/PEG film (Fig. 5.47). The FESEM images obtained at high magnification clearly show enhanced monodispersity of gold nanoparticles (Fig. 5.48).The size of the particle is 10.3 ± 1.4 nm, which is similar to the value obtained using PEG-thiol. These results suggest that, after the addition of excess PEG molecules, the gold nanoparticles are covered with citrate molecules, which are, in turn, crowned with PEG molecules. Such crowning by PEG molecules has been attributed to the interactions of lone pairs of electrons present on the oxygen atoms of PEG molecules223. It is also believed that by varying the molecular weight of PEG molecules without thiol group should enable to alter the particle size. As the polydispersity is higher in the case of higher molecular weight (5000 g/mol) PEG molecules, this will not enable to improve the polydispersity of initial citrate capped gold nanoparticles. Hence it was decided to carry out the experiments only with lower molecular weight (300 g/mol) PEG molecules.

214

5.12 Role of bound carboxylate ion As NMR and Raman measurements showed the presence of bound citrate, the role of sodium citrate in particle ripening was investigated next. The gold nanoparticles with and without bound citrate, of similar size range were synthesized using different approaches. The results of these experiments are presented next.

5.12.1 Tannic acid method Tannic acid capped gold nanoparticles of mean size 13.8 ± 4.3 nm were prepared as reported earlier63 (Fig. 5.49a). After adding PEG-thiol molecules (molar ratio of PEG to gold: 10:1), there was no evidence of change in particle size distribution after ageing the sample for 16 hours. The mean size was found to be 14.5 ± 2.1 nm (Fig. 49b). Also, addition of premixed solutions of PEG-thiol and sodium citrate to tannic acid capped gold nanoparticles did not show evidences of size focusing, thereby indicating the importance of adsorbed citrate ions. The mean size was found to be 10.4 ± 4.9 nm (Fig. 5.49c,d).

5.12.2 Ascorbic acid method Flower like nanostructures of varying size from 10 to 40 nm were prepared based on the report of Boca and coworkers (Fig. 5.50a)241, using ascorbic acid as the reducing agent. Adding excess PEG-thiol molecules (molar ratio of PEG to gold: 10:1) did not result in significant changes in particle size distribution (Fig. 5.50b).

5.12.3 Ethylene glycol protocol Standard Turkevich protocol was modified by replacing water with ethylene glycol as solvent (Fig. 5.51a). After addition of excess PEG-thiol molecules (molar ratio of PEG/Au: 10:1), and aging for 1 day, particle ripening was clearly observed (Fig. 51b,c). Next, as ethylene glycol can act as both reducing and stabilizing agent, the nanoparticle synthesis was carried out without the addition of citrate (Fig. 5.51d).

215

Figure 5.47: Representative low magnification FESEM image of PEG capped gold nanoparticles (without any thiol moiety). This was formed after addition of PEG molecules of molecular weight 300 g/mol without any thiol functionalization to citrate capped gold nanoparticles and aged for 1 day. The molar ratio of PEG to gold is 10. The regions marked with solid circles represent Moiré patterns while the dotted rectangle represents image distortion due to the possible presence of organic film.

216

a d

c d

b d

d d

Figure 5.48: (a-c) Representative high magnification FESEM image of PEG capped gold nanoparticles (without any thiol moiety) obtained at different locations and magnifications. This was formed after addition of PEG molecules of molecular weight 300 g/mol without any thiol functionalization to citrate capped gold nanoparticles and after being aged for 1 day. The molar ratio of PEG to gold is 10. (d) Size histogram of PEG capped gold nanoparticles.

217

a d

b d

c d

d d

Figure 5.49: Representative FESEM image of (a) polydispersed tannic acid capped gold nanoparticles (13.8 ± 4.3 nm), (b) after addition of PEG-thiol and ageing for 16 hours (14.5 ± 2.1 nm) and (c) after post-factorial addition of sodium citrate to one day aged tannic acid-PEG-thiol capped gold nanoparticles (d) after one week (10.4 ± 4.9 nm)

218

a d

b d

Figure 5.50: (a) Representative FESEM image of ascorbic acid capped gold nanoflowers. (b) Representative FESEM image of gold nanoflowers after addition of PEG-thiol and ageing the sample for 1 day at 25 °C.

219 When PEG-thiol was added to this solution, there was no evidence of ripening even after waiting for prolonged durations (2 days) (Fig. 5.51e). This experiment again highlights the role of adsorbed citrate in particle ripening.

5.12.4 Sodium acrylate protocol Gold nanoparticles of size 10.0 ± 2.3 nm were synthesized using sodium acrylate as reducing and stabilizing agents based on an earlier report242 (Fig. 5.52a).Upon addition of excess PEG-thiol molecules (molar ratio of PEG/Au: 10:1), there was no significant evidence of size focusing as the mean particle size was found to be 13.4 ± 3.7 nm (Fig. 5.52b). This is surprising because both sodium acrylate and sodium citrate have the same functional unit, carboxylate ion. The only notable difference between the Turkevich protocol and the sodium acrylate protocol is the amount of the free carboxylate ions. So, the Turkevich protocol was also modified by increasing the molar ratio of sodium citrate to gold from 5 to 20 (mean size of particle remains same, i.e., 16.0 ± 3.4 nm, as shown in Fig. 5.52c). In the Turkevich protocol with excess citrate, there was no evidence of ripening even after waiting for prolonged periods of time (2 days), the mean size was found to be 16.8 ± 4.1 nm (Fig. 5.52d).

5.13 Phase-space of ripening A phase-space of ripening was constructed in concentration domain (citrate, gold and PEG concentration) (Fig. 5.53). The black closed circles represent instances of particle ripening, while the red closed circles correspond to instances where ripening didn’t occur. To demarcate the ‘zone of ripening’, additional experiments were carried out and are marked as points with closed half red and black circles. There were signs of particle ripening, but the process did not proceed to completion. Based on this phase-space of particle ripening, it becomes evident that the amount of citrate, PEG and gold is critical for particle ripening to be observed. To test this further, a modification of sodium acrylate protocol was carried out such that the amount of free carboxylate ions in the colloidal sample remains the same as standard Turkevich protocol where ripening was observed (Fig. 5.54a). As surmised, enhancement of

220 monodispersity was observed in the modified sodium acrylate protocol, as shown in Fig. 5.54b-d with the addition of PEG-thiol molecules (molar ratio of PEG to gold is 10) and after being aged for 18 hours at 25 °C. This experiment clearly highlights the role of excess carboxylate ion in the colloidal solution and its implications on particle ripening.

5.14 Pechini process All the aforementioned experiments suggest that presence of carboxylate ions and PEG molecules is critical for observing particle ripening. These observations are analogous to one of the widely-used process for synthesizing multi-component oxide powders, polymerizable complex method, which is also known as Pechini process243-246. It involves the formation of polymeric resin by gelation of the reaction mixture consisting of desired metal ions, chelating agent (citric acid, EDTA) and polyol (ethylene glycol). The metal ions are trapped inside the organic matrix. Control of particle size and shape of metal oxide nanoparticles can be achieved, albeit moderately, through calcination of polymeric resin at higher temperature (~ 300500°C). Kim and co-workers have synthesized monodispersed barium ferrite nanoparticles (4.7 ± 0.8 nm) without calcination, through the formation of metalcitrate complexes247. Until now, the Pechini process has been widely used to synthesize metal oxide nanostructures only248. In addition, in the Pechini process, the molar ratios of citric acid and ethylene glycol, has been reported to affect the chain size of the polymer and its cross-linking degree while the amount of metal ions determine the polymerization rate249,250. Our experiments, in fact, do suggest particle ripening only in the small zone of the ratio of citrate, PEG and gold. Particle ripening process through addition of PEG-thiol was neither observed for smaller particles (< 5nm) nor for larger size particles (>30 nm). It is pertinent to note that these particles were obtained by varying the amount of citrate. Furthermore, the order of addition of precursors, ethylene glycol and citric acid has been found to significantly affect the formation of barium titanium citrate gel using the Pechini process249. Similarly, it has been highlighted earlier that post-factorial addition of citrate does not result in particle ripening.

221 a d

b d

c d

d d

e d

Figure 5.51: Representative FESEM image of citrate capped gold nanoparticles synthesized with ethylene glycol as solvent (instead of water) before (a) and after addition of PEG-thiol (b,c). Fig d represent the FESEM image of gold nanoparticles synthesized with ethylene glycol as both reducing agent and stabilizing agent without citrate while e represents gold nanoparticles after capping with PEG-thiol molecules. The molar ratio of PEG to gold was maintained at 10 and the sample was aged at 25 °C for 1 day.

222

a d

b d

c d

d d

Figure 5.52: Representative FESEM images of sodium acrylate capped gold nanoparticles (a) before and (b) after addition of PEG-thiol molecules and aged for 1 day. Representative FESEM image of citrate capped gold nanoparticles with citrate to gold nanoparticles (c) before and (d) after addition of PEG-thiol and aged for 2 days. Clearly excess citrate prevents ripening of particles. The molar ratio of PEG to gold is 10.

223

Figure 5.53: Phase space of PEG-thiol, gold, sodium citrate highlighting the regions of particle ripening. The black closed circles are the ones at which particle ripening occurred while the red closed circles are the points correspond to conditions of no evident particle ripening. Couple of experiments marked as points with closed half red and black circles were carried to identify the limits of boundary lines for particle ripening. There were only indications of particle ripening.

224 a d

b d

c d

d d

Figure 5.54: Representative FESEM images of gold nanoparticles synthesized using modified sodium acrylate protocol, (a) before and (b-d) after addition of PEG-thiol at different magnifications. The region marked 1 (in c) represent due to the possible presence of organic film. The molar ratio of PEG to gold is 10 and the sample was aged for ~ 18 hours at 25 °C.

225 a d

b d

Figure 5.55: (a) Representative FESEM image of gold nanoparticles synthesized using Pechini process with the addition of PEG-thiol molecules (molecular weight: 356 g/mol) a priori. (b) UV-Vis spectra of as-synthesized gold nanoparticles (as shown in a) indicate suppression of the characteristic gold SPR peak at 520 nm, indicating sub-3 nm particles.

In order to test whether all these process are thermodynamically controlled, an additional experiment was conducted where in a mixture of PEG-thiol and sodium citrate (24 mL of water + 1 mL of 1 % sodium citrate + 25 µL PEG-thiol, molecular weight 356 g/mol) was taken to boiling followed by the addition of gold chloride solution (0.25 mL of 1 % HAuCl4). The color of the solution remained colorless even after 30 minutes of continued boiling. FESEM image of solution showed the presence of sub-3 nm particles (Fig. 5.55a). UV-Vis spectra also corroborated the presence of sub-3 nm particles as the characteristic Surface Plasmon Resonance (SPR) peak of gold nanoparticle at 520 nm was not prominent. This suggests that thiol group prevents growth of particle. To overcome the issue of thiol binding, PEG molecules without thiol moiety were added instead of PEG-thiol molecules. In this case, the color of the solution turned light blue almost immediately, and turned to wine red after 8 minutes. This is interesting because in the experiment carried out without the addition of PEG molecules, the color of the colloidal solution changed to wine red in less than 2 minutes62. More interestingly, the particles are highly polydisperse (Fig. 5.56a-b). However, with ageing of the colloidal solution at 25 °C for 12 hours, the particles become monodisperse with mean size of 12.1 ± 1.8 nm (Fig. 5.56c-f). The same sample aged at 4°C (kept in a refrigerator) does not undergo similar transformation (Fig. 5.56g-h), clearly highlighting the role of temperature.

226 Halving the gold concentration yielded particles of size 9.0 ± 1.2 nm, while doubling the gold concentration yielded 10.7 ± 1.8 nm. At this juncture, it is difficult to explain the trend of particle size with gold concentration. There could be interplay of both pH and variation of the initial number of nuclei which can cause this variation. To understand the pH effect, two more experiments were carried out: (1) the gold concentration was increased to 10 times and (2) reduced 10 times of the standard concentration, keeping all other precursor concentration same. With increased gold concentration to 10 times, the FESEM images showed presence of hexagonal plates in addition to large size particles (Fig. 5.57a-b). On the contrary, lowering the gold concentration by 10 times, the size of the particles increased to ~ 30 nm, and the spread of the size distribution was high (Fig. 5.57c-d). With ageing of the colloidal solutions at 25°C, the size of the particles did not alter in both cases. All these experiments suggest that optimum amounts of PEG-citrate-gold (“sweet zone”) are critical to observe size focusing. In parallel, to understand the role of temperature, an additional experiment was conducted where in a mixture of PEG and sodium citrate (24 mL of water + 1 mL of 1 % sodium citrate + 25 µL PEG, molecular weight 300 g/mol) was taken to boiling and cooled to room temperature. After 10 minutes, the gold chloride solution (0.25 mL of 1 % HAuCl4) was added to the above mixture maintained at room temperature and stirred for 30 minutes. The color of the solution remained colorless for few hours and a faint blue color started to appear after 4 hours. The solution color did not change even after ageing the solution for 1 week. It is well known that acetone dicarboxylic acid is an intermediate in the citrate protocol62,251,252 that is responsible for autocatalytically reducing gold ions to metallic form. So, the aforementioned experiment was modified such that instead of sodium citrate, acetone dicarboxylic acid was used to initiate the reduction at room temperature, i.e., gold chloride solution (0.25 mL of 1 % HAuCl4) was added to a mixture of PEG and sodium citrate (24 mL of water + 0.5 mL of 1 % sodium citrate + 25 µL PEG-thiol, molecular weight 300 g/mol). The color of the solution turned purple-red in a couple of minutes. The FESEM images showed presence of spherical size particles (~ 30-200 nm) in addition to nanoplates (Fig. 5.58a-b). To understand

227 a d

b d

c d

d d

e d

f

g d

h d

Figure 5.56:Representative FESEM image of gold nanoparticles synthesized using citrate reverse protocol, with the a priori addition of PEG molecules (molecular weight: 300 g/mol); without ageing (a-b), with ageing for 12 hours at 25 °C (c-f), and with ageing for 1 day at 4 °C (g-h). The molar ratio of PEG to gold is 10. The regions marked 1 and 2 in c represent array and bare substrate respectively.

228

a d

b d

c d

d d

Figure 5.57: Representative FESEM images of gold nanoparticles formed using the standard Pechini process with increasing the gold concentration 10 times (a-b) and decreasing the gold concentration 10 times (c-d) when compared to standard concentration. The molar ratio of PEG to gold is 1 and 100 respectively.

229 the effect of reverse mode of addition (addition of gold precursor to reducing agent; the standard mode of addition is addition of reducing agent to gold precursor), the above experiment was repeated without any PEG molecules. Interestingly, the concentration of nanoplates increased in the colloidal sample (Fig. 5.58c-d). Earlier, it was shown that pH of the reaction mixture can induce nanoplate formation (Fig. 5.57a-b), and it was found here that the final colloid solution pH was 4.5. So to overcome the pH effect, the experiment at room temperature using acetone dicarboxylic acid was modified with a priori addition of requisite amounts of sodium hydroxide solution, so as to maintain final pH of 6.5. As surmised, only nanoparticles were found (Fig. 5.58e). The particles were found to be more than 30 nm and the color of the colloidal solution remained purple even after 1 week. Very recently, Muhammed and co-workers have observed size focusing of polydisperse silver nanoparticles by using reducing agent and excess ligands made of poly (ethylene glycol) appended with lipoic acid at one end and a reactive (COOH/NH2) or inert (-OCH3) functional group at the other end253. The authors have hypothesized that the free thiols in solution can etch the nanoparticles, progressively leading to nanoclusters. In parallel, Shiers and co-workers have also reported thiolate driven etching of silver nanoparticles254. The authors have reported tiopronin (C5H9NO3S, carboxylic acid and thiol containing biomolecule) as ligand which has been found to form Ag(I)-tiopronin polymer chains arranged into layers. Also, Shiers et al. have reported that these layers form into 2D polymeric nanosheets which can engulf

polydisperse

nanoparticles

and

transform

them

into

monodisperse

nanoparticles254. In the current study, it is surmised similar phenomenon could be occurring where in PEG and sodium citrate form a complex, which can transform polydisperse gold nanoparticles into monodisperse nanoparticles. To study the feasibility of the formation of an organic complex between PEG and citrate, an additional experiment was conducted wherein aqueous solution of PEG and sodium citrate was heated to boiling for 30 minutes, cooled to room temperature and then drop-cast on silicon substrate. Polymeric nanostructures were observed (Fig. 5.59a-b). Electron beam was found to damage these nanostructures, while focusing in FESEM. All these evidences suggest particle ripening due to PEG-citrate complexation.

230

a d

b d

c d

d d

e d

Figure 5.58: (a-b) Representative FESEM image of gold nanoparticles synthesized using acetone dicarboxylic acid and PEG molecules (molecular weight: 300 g/mol). Both nanoparticles and nanoplates were observed. (c-d) Representative FESEM images of gold nanoparticles synthesized using acetone dicarboxylic acid alone. The concentration of nanoplates increased without the addition of PEG molecules. (e) Representative FESEM images of gold nanoparticles synthesized using acetone dicarboxylic acid and PEG molecules (with final pH modified to 6.5 using sodium hydroxide solution a priori). The molar ratio of PEG to gold was maintained at 10.

231

5.15 Pseudocrown ethers Network of polymeric pseduocrown ethers (cyclic polyethers with the repeating unit of (-CH2-CH2-O-)n, n > 2) can be formed by photopolymerization of poly(ethylene glycol) diacrylate (PEGDA)/metal complexes255,256. Crown ethers are also formed by linking PEGDA monomers with metal ions through the ether oxygen complexation resulting in bridging acrylate groups. This eventually increases the possibility of cyclopolymerization which can result in a cross-linked network with crown ether structures255. The process, suggested by Elliot and co-workers255,256, is schematically illustrated in Fig. 5.60a. The mesh size of crown ethers is found to be controlled through molecular weight of polymers, cross-linking agent etc255-259. Also, ethylene glycol-citrate complexes through chelation of metal-carboxylate-ethylene oxide (in the PEG backbone) have been reported to form chain networks246. It is surmised that the formation of thin sheets or complex 3D networks of PEG-citrategold to be the plausible structure which can aid in particle ripening as shown in Fig. 5.60b. However, the exact details of this size transformation process could not be unveiled. To test the generality of the “Pechini-type” process for forming monodisperse metal nanoparticles, the process was carried out using silver salt instead of gold salt. Surprisingly, in addition to spherical particles, significant fraction of silver nanorods was observed (Fig. 5.61a-c). More interestingly, the size of some of the nanorods exceeded a length of 2 µm (Fig. 5.61d). UV-Vis spectrum shows a small hump near 280 nm (Fig. 5.61e), which is characteristic of silver nanorods260. To verify whether the formation of nanorod is due to the reverse addition of precursor solution or presence of PEG molecules, an additional experiment was conducted wherein the process was repeated without the addition of PEG molecules. Silver nanorods were observed in this case as well, but not with similar density or aspect ratio (Fig. 5.61f). UV-Vis spectra also corroborated the FESEM result (Fig. 5.61e).

232 b d

a d

Figure 5.59: (a-b) Representative FESEM images of boiled aqueous boiled mixture of PEG and sodium citrate drop-casted on silicon substrate.

a d

b d

Figure 5.60: (a) Schematic of formation of pseudocrown ether of poly(ethylene glycol) diacrylate (PEGDA)/metal complexes (adapted from reference255). (b) Schematic of formation of 3D complex network of citrate and PEG, through chelation of metal nanoparticle or metal ions.

233 a

b d

c d

d d

e d

234 f

Figure 5.61: (a-d) Representative FESEM images of silver nanoparticles and nanorods prepared through standard Pechini process. (e) UV-Vis spectra of as-synthesized silver nanocolloid; shoulder at ~280 nm indicates the presence of silver nanorods in solution. (f) Representative FESEM image of silver nanorods prepared through Pechini process, without the addition of PEG molecules. The molar ratio of PEG to silver is 10.

5.16 Interaction between ligand and metal In literature, a well established method for size reduction of nanoparticle is the digestive ripening process, which involves thiol or ligand based etching214. The current study cannot be deemed as digestive ripening process because it was shown that thiol moiety is not essential for observing ripening behavior. The digestive ripening process is also characterized by two other observations; namely, ripening occurs close to the boiling point of solvent, and final size of the particle is the same irrespective of the initial nanoparticle size and amount of excess ligand. Hence, the explanation for current ligand based etching must be different. Ostwald ripening process, the standard ripening mechanism, involves the growth of larger particles at the expense of smaller particles in order to minimize the surface energy. Thus, it cannot explain the data of different size focusing behavior of different metals with the same ligand. Formation of hydrogel-like network261 (wherein the mesh size is controlled the concentration of cross-linking agent as shown in Fig. 5.62a), cannot explain the observed monodispersity of particles, because the size of these meshes are highly polydisperse262 when compared to the size of the particles (Fig. 5.62b). Kuo et

235 al. have shown the final size of the gold nanoparticle can vary during ligand exchange reactions depending on the solvent and the ligand263. Their hypothesis is based on the Gibbs-Thompson effect, which involves the radius of the nanoparticle and the chemical potential through surface free energy per unit area at the adsorbate-gold 𝑎𝑑𝑠 interface (𝐺𝑝𝑒𝑟 𝑙𝑖𝑔𝑎𝑛𝑑 ). They have shown that the final size of the particle (r) is given

by

𝑟=

−2𝑙 𝑎𝑑𝑠 𝑘+𝑚𝐺𝑝𝑒𝑟𝑙𝑖𝑔𝑎𝑛𝑑

(5.2)

where k, l and m are size-independent parameters used in simplifying the model equation reported as in Kuo et al263. The above equation suggests that the final size of the particle will be larger for weaker metal-ligand interaction. So, it can be surmised that based on the current study, the interactions follow, Ag-cit-PEG > Pt-cit-PEG >Pd-cit-PEG > Au-cit-PEG.

5.17 Direct aqueous self-assembly of nanoparticle As a proof of concept, 5 µL of PEG-thiol coated gold nanoparticle was drop-cast on hydrogen plasma treated PDMS slab and transferred onto Si wafer (Fig. 5.63). Very large scale arrays were formed and this represents a promising route for aqueousbased self-assembly of gold nanoparticle arrays. Transferring arrays to any desired substrates using microcontact printing will further foster the development of applications ranging from flexible electronics to sensors.

5.18 Thermal stability of PEG capped gold nanoparticles When the PEG-thiol coated gold nanoparticles (derived from citrate, molar ratio of PEG to gold is 10) were heated to boiling, the color of the colloidal solution turned blue after heating. More remarkably, when the colloidal solution was cooled to room temperature, the color of the solution reverted to red color (Fig. 5.64). This color transition can be exploited for miniaturized sensing applications. More interestingly,

236

a d

b d

Figure 5.62: (a) Schematic representation of the cross-linked hydrogel (Reproduced with permission from Shapiro261. Copyright (2011) Elsevier). (b) TEM image of the freeze-dried manno/xylene gel indicating highly polydisperse mesh size (Reproduced with permission from Sakurai et al.262. Copyright (2003) American Chemical Society).

237 a d

b d

Figure 5.63: Representative FESEM images gold nanoparticles drop-casted on hydrogen plasma treated PDMS substrate and transfer printed to the silicon substrate.

Figure 5.64: Digital photograph demonstrating reversible transformation of red to blue colour of the PEG-thiol coated gold nanoparticle solution, with temperature.

238 a d

b d

c d

Figure 5.65: (a-c) Representative FESEM images after thermal cycling of PEG-thiol coated gold nanoparticles derived from tannic acid capped gold nanoparticles. The formation of nanopouches can be attributed to the possible side reaction of tannic acid with PEG molecules 264.

239 this phenomena was observed even after multiple heating-cooling cycles (7 cycles were conducted). Couple of reasons can be envisaged for the color transformation due to the thermal cycling, (1) change in particle size and shape and (2) dipolar coupling due to the reduction in interparticle spacing between the particles. As discussed in Section 5.4, it is believed the color transition is to do more with morphological change of particles in this case. On the contrary, when the thermal cycling experiments were repeated with PEG-thiol coated gold nanoparticles derived from tannic acid molecules (monodisperse particles of size 7 nm63, and molar ratio of PEG to gold is 10), nanopouches were observed (Fig. 5.65a-c). This can be attributed to the possible reaction of tannic acid with PEG molecules264.

5.19 Summary A simple process for generating monodispersed metal nanoparticles in aqueous phase was developed. A systematic study was conducted in order to understand the role of excess ligand in enabling size focusing. PEG-carboxylate interaction is found to be critical for observing particle ripening. Also, the presence of thiol moiety was not critical for the phenomena to be observed. Finally, as a proof of concept, PEG-thiol coated gold nanoparticles were drop-cast on PDMS, which enables transfer of arrays to any desired substrate.

240

Chapter 6 Summary

and

Scope

for

Future Work 6.1 Summary The primary scope of this thesis is to integrate the bottom-up self-assembly process with Complementary Metal Oxide Semiconductor (CMOS) device fabrication steps, which will aid in fabricating next-generation non-volatile memory devices. The aforementioned problem has been broken into 3 sub-tasks, (i) large-scale selfassembly of nanoparticles, (ii) floating gate memory device fabrication using the asprepared self-assembled 2D arrays of gold nanoparticles as charge storage nodes and (iii) development of process to synthesize monodisperse PEG capped gold nanoparticles so as to assemble particles directly from aqueous medium. This thesis reports new ways of synthesizing monodisperse gold nanoparticles, the development and optimization of their self-assembly to integrate in fabrication of floating-gate memory device. A new process for transformation of polydispersed citrate capped gold nanoparticles into monodisperse nanoparticles through the addition of PEG molecules were discussed through various characterization techniques such as NMR, MALDI, Raman spectroscopy etc., and also extension of the approach to various metallic systems. The synthesized nanoparticles were then utilized to form wafer-scale 2D array of gold nanoparticles. It was also shown that the inter-particle spacing in these 2D arrays can be tuned through polymer grafting. The effects of solvent and the nature of the substrate were also discussed. Also, a new approach to tune packing geometry by modulating the non-expensive shear forces was also presented. These arrays were then utilized to fabricate floating gate memory device with different inter-particle spacing. The process developed enabled to have devices with excellent retention (decay time constant ~ 10 years) and endurance (>10000 program/erase cycles) characteristics. Also, it was experimentally shown that there is an inter-particle spacing which yields highest memory window, which is counter-intuitive. In the following sections, detailed summaries of the results of our work on the three main tasks are presented.

241

6.1.1 Large-scale self-assembly of nanoparticles With the thrust in miniaturization of electronic components, it becomes imperative to fabricate nanostructures with precise particle positioning. As the conventional lithographic processes progress towards a brick wall due to both fundamental and economic limitations, an alternative, cost-effective, bottom-up approach becomes a promising avenue for fabrication of ultra-high density nanostructures1. However, the translation of this promise is hampered due to two main reasons: (i) inability to form large scale ordered nanostructures and (ii) ineffectiveness in controlling particle arrangement in nanoparticle arrays. In this regard, a simple, wafer scale (~ 9 cm diameter) process for fabrication of monolayer of dodecanethiol capped gold nanoparticles was developed. Unlike the previous approaches26,28, the current self-assembly process occurs at the air-water interface, which in conjunction with microcontact printing enables transfer of arrays to any desired substrate, flexible or rigid64. The key to large scale fabrication of arrays lies in spreading of colloidal solution at the water sub-phase with the aid of excess ligand. The kinetics of self-assembly of nanoparticles arrays with and without excess ligand was investigated using ellipsometry. An enhancement in evaporation rate of solvent and film spreading for the colloidal solution containing excess thiol was observed, when compared to pure colloidal solution. The optical micrographs suggest labyrinthine like or island like structures depending on the presence or absence of excess surfactant.For larger interparticle spacing in arrays, thiol-terminated polystyrene capped gold nanoparticles was used. By using toluene as solvent, large scale arrays were formed by drop-casting on silicon substrate by suppressing polymer/solvent dewetting. Until now, the interparticle spacing in arrays was tuned by changing the ligand length (through molecular weight)67. Interestingly, in this study, it was shown that the spacing in arrays can be tuned from 10 to 20 nm by modulating the van der Waals interaction (disjoining pressure) between the substrate and the nanoparticle film. Also, the spacing can be altered by controlling the hydrodynamic diameter of the colloids by controlling the solvent properties. A balance of steric repulsive forces and attractive capillary immersion forces was found to control the assembly of nanoparticles.

242 Square packing in arrays have been realized using DNA based templates146. Herein, it was shown that the arrays with square packing can be formed by exploiting the Bernard-Marangoni instability arising due to surface tension gradient142. The Marangoni convective cells can arise either due to gradient in concentration or temperature. First, the square packing arrays with spacing of 10 nm was formed using thiol-terminated polystyrene capped gold nanoparticles with ethanol and water as subphase. Small domains of arrays having square packing were formed by using spin casting at high speeds (~8500 rpm). Next, it was shown that Marangoni cells can be created using temperature gradient. This results in square arrays with an interparticle spacing of 2 nm. In addition to tuning interparticle spacing and packing in arrays, by using a combination of good solvent and bad solvent such as tetrahydrofuran and water or dodecane as subphase, it was shown that surface plasmon resonance band of the resultant nanostructures can be tuned over wide range of wavelengths. More interestingly, with higher water content, the surface plasmon resonance of thiolterminated polystyrene capped gold nanoparticles can be shifted in the infra-red region, which can be exploited in biomedical applications265.

6.1.2 Scalable processes for fabricating non-volatile memory devices using self-assembled gold nanoparticles as charge storage nodes The exponential increase in the field of semiconductor nanoelectronics has led to the unimaginable rise of portable electronics155. Currently, flash memory is considered as the technology driver of the semiconductor industry, relying on floating gate transistor based mechanism for memory characteristics. The successful endeavor of the semiconductor industry in doubling the memory density every 18 months has relied on the ability to continually reduce feature sizes. Several advancements in fabrication of flash memory devices have been made with the intent of making devices “smaller, cheaper and faster”. Unfortunately, with the feature size reaching sub-25 nm, further device scaling is hampered mainly due to two challenges: (1) significant cell-to-cell interference as the spacing between the floating gates has now become comparable to

243 memory cell size, and (2) non-scalability of tunneling oxide thickness due to unreliable charge retention after several program/erase cycles58. Integration of bottom-up, self-assembly approach for fabrication of nanostrutures with the CMOS process, can help in realization of next-generation, non-volatile memory devices58. Unfortunately, this promise has been severely hampered due to the present incompetent approaches to fabricate thermally stable array to withstand CMOS device processing steps resulting in poor device performance266. In this regard, scalable processesing steps were developed to deposit control oxide without disturbing the order of nanoparticle array. Remarkably, after oxide deposition the devices can withstand temperature as high as 850 °C. In this work, three deposition technqiues namely, RF magnetron sputtering, Atomic Layer Deposition and Electron-beam deposition were employed to deposit control oxide. Some salient features of the fabricated devices include: (a) reproducible threshold voltage of devices spread across cm2 area, (b) excellent retention (>10 years) and endurance characteristics (>10000 P/Ecycles) in some cases, and (c) enhanced thermal stability of the ordered nanoparticle array (>500 °C). For arrays with larger interparticle spacing (>2 nm), pre-plasma treatment for removal of ligand was found to be critical for observing memory characteristics. The presence of ligand was surmised to form carbonaceous deposit which can result in poor memory characterstics. Interestingly, there is an optimum in interparticle spacing in arrays for observing larger memory window. This trend was independent of the type of deposition technqiues used. Unfortunately, in all of the devices it was observed that tunneling of electrons is from the top-oxide instead of conventional silicon substrate. This is attributed to the inability to deposit device-quality oxide using the available equipments. It is believed that with the optimization of processes for depositing of high-quality oxide, the goal of next-generation, memory devices using self-assembled 2D arrays of nanoparticles as charge storage nodes can be realized using the processes developed as part of this thesis.

244

6.1.3 PEG capped monodisperse gold nanoparticles Classical Turkevich-citrate protocol61 is one of the widely-sought after method for synthesizing metal nanoparticles. Even today, many modifications to this approach are being proposed to synthesize monodisperse nanoparticles62. Until now, selfassembly of nanoparticles has been realized mainly through colloidal solution dispersed in organic solvents. Hence, it becomes imperative based on environmental considerations, to develop processes for synthesizing and assembling monodisperse nanoparticles directly from aqueous medium. In this regard, it was shown that addition of excess PEG molecules to citrate capped gold nanoparticles can enhance monodispersity of particles. PEG-carboxylate interaction was found to be critical for observing particle ripening. Unlike digestive ripening213, Ostwald ripening etc., the process is found to depend on parameters such as initial particle size, PEG molecular weight, order in which the reactants are added, presence of bound citrate etc. This suggests that process occurs only when there is a complexation of PEG and bound citrate molecules in optimum amounts. The details of this process were found to be in line with the well-studied Pechini process243 for synthesizing metal oxide nanostructure films. To determine the generality of the ripening process through the addition of PEG molecules, it was extended to other systems such as palladium, platinum and silver. Interestingly, the size focusing was found to depend on nature of material. Finally, as a proof of concept, self-assembly of PEG-thiol coated gold nanoparticles were demonstrated on plasma treated PDMS substrate. This in conjunction with approaches such as doctor-blade casting95 or convective assembly26 can aid in realizing the dream of roll-to-roll processing of nanoparticle array fabrication. In addition, this can also foster the development of fields such as flexible electronics and sensors as the arrays can now be transferred to any desired substrate.

245

6.2 Thesis contributions 6.2.1 Peer-reviewed articles 1. G. Muralidharan, N. Bhat and V. Santhanam, “Scalable processes for fabrication of non-volatile memory devices using self-assembled 2D array of gold nanoparticles as charge storage nodes”, Nanoscale, 3, (2011), 4575-4579. 2. G. Muralidharan, S. Sivaraman and V. Santhanam, “Effect of substrate on particle arrangement in arrays formed by self-assembly of polymer grafted nanoparticles”, Nanoscale, 3, (2011), 2138-2141. 3. G. Muralidharan, N. Bhat, V. Santhanam “Ultra-high density floating gate devices using self-assembled 2D arrays of gold nanoparticles”, In Emerging Electronics (ICEE), 2012 International Conference on (pp.1-4). IEEE; doi: 10.1109/ICEmElec.2012.6636241. 4. G. Muralidharan and V. Santhanam, “Enhancement of monodispersity and ordering in self-assembly of aqueous gold nanoparticles through PEG-adsorbed carboxylate interaction”, under preparation.

6.2.2 Conference presentations 1. G. Muralidharan*, N. Bhat, V. Santhanam “High density memory devices using self-assembled gold nanoparticle arrays as floating gates”, American Institute of Chemical Engineers – AIChE Annual meeting, Minneapolis, USA, October 2011. (ORAL) 2. G. Muralidharan*, S. Sivaraman, V. Santhanam “Self-assembly of polymer grafted nanoparticles with tunable spacing and packing”, International Conference on Materials for Advanced Technologies – ICMAT, Singapore, June 2011. (POSTER) 3. G. Muralidharan*, N. Bhat, V. Santhanam “Self-assembled 2D arrays of gold nanoparticle for non-volatile memory applications”, “Indo-US workshop on Nanoparticle Assembly: From Fundamentals to Applications”, New Delhi, December 2011. (POSTER)

246 4. G. Muralidharan*, N. Sharma, V. Santhanam “Large-scale self-assembly of dodecanethiol capped gold nanoparticles at the air-water interface”, “4th International Conference on Advanced Nano Materials”, Chennai, October 2012. (POSTER) 5. G. Muralidharan*, N. Bhat, V. Santhanam “Ultra-high density floating gate devices using self-assembled 2D arrays of gold nanoparticles”, “IEEE Conference on Emerging Electronics”, Mumbai, December 2012. (POSTER)

6.3 Scope for future work As discussed in Section 6.1, this thesis aimed at integrating the bottom-up selfassembly process with CMOS device fabrication steps, so as to fabricate nextgeneration floating-gate memory devices using gold nanoparticles as charge storage nodes. The contributions of this thesis have been summarized in Section 6.1. However, further research work is needed to enable industrial adaptation of these results and these are elaborated herein

6.3.1 Large-scale self-assembly of nanoparticles 1. Wafer scale, high-density 2D arrays of dodecanethiol capped gold nanoparticles were self-assembled at the air-water interface by addition of excess ligand to the colloidal solution. Ellipsomtery was used to understand the mechanism of selfassembly of nanoparticles. As the thickness of the film reached sub 100 nm in less than 6 minutes, which happens to be the first data point, new ways need to be devised to slow down the dynamics of self-assembly process. One way to do this is to carry out the self-assembly process at lower temperatures such as 10 °C. 2. Local domains of square lattices of nanoparticle arrays were demonstrated by exploiting convective Marangoni instability. Alternative strategies to create shear stresses in the particles over longer periods of time need to be devised, which can potentially aid in large scale square lattice formation without the need for complicated molecular ligand designs.

247

6.3.2 Fabrication of flash memory devices 1. In all of devices fabricated, the injection of electrons was through the top-oxide, due to defect assisted tunneling. To circumvent this, process to fabricate highquality, control oxide needs to be implemented. 2. Field-effect transistor (FET) can be devised so that electrons can then be tunneled through silicon substrate instead of control oxide. 3. The effect of particle size on memory characteristics can also be investigated.

6.3.3 Aqueous-phase synthesis of monodisperse metal nanoparticles 1. By the addition of excess PEG-thiol molecules to citrate capped gold nanoparticles, monodispersity of nanoparticles can be enhanced significantly. The presence of PEG-carboxylate complex (in optimum amounts) is required for observing particle ripening. The experimental results are in concord with the wellestablished Pechini process for fabricating metal oxide nanostructures. These results open up the avenue for large scale fabrication of nanostructures directly from aqueous medium. PEG coated nanoparticles can be used to directly assemble the particles on PDMS substrate, which can aid in array fabrication using roll-toroll process. 2. The interaction of PEG, citrate and gold needs to be investigated through appropriate modeling/simulation techniques to generalize the process.

248

Chapter 7 References 1.

Kraus T. Assembly and printing of micro and nano objects Cuvillier Verlag; Dissertation, Eidgenössische Technische Hochschule ETH Zürich, Nr. 17334, 2007.

2.

Lu W, Lieber CM, Nanoelectronics from the bottom up, Nature Materials, 2007, 6(11), pp.841.

3.

Ratner M, Ratner D. Nanotechnology: A gentle introduction to the next big idea: Pearson Publications; 2005.

4.

Schmid G. Nanoparticles: From theory to application: Wiley-VCH Verlag GmbH and Co.; 2004.

5.

Faraday M, The bakerian lecture: Experimental relations of gold (and other metals) to light, Philosophical Transactions of the Royal Society of London 1857, 147, pp.145.

6.

Pushkarev VV, Musselwhite N, An K, Alayoglu S, Somorjai GA, High structural sensitivity of vapor-phase furfural decarbonylation hydrogenation reaction network as a function of size and shape of pt nanoparticles, Nano Letters, 2012, 12(10), pp.5195.

7.

Lee J, Govorov AO, Kotov NA, Nanoparticle assemblies with molecular springs: A nanoscale thermometer, Angewandte Chemie, 2005, 117(45), pp.7605.

8.

Andres RP, Bein T, Dorogi M, Feng S, Henderson JI, Kubiak CP, Mahoney W, Osifchin RG, Reifenberger R, "Coulomb staircase" at room temperature in a self-assembled molecular nanostructure, Science, 1996, 272(5266), pp.1323.

9.

Zabet-Khosousi A, Dhirani AA, Charge transport in nanoparticle assemblies, Chemical Reviews, 2008, 108(10), pp.4072.

10.

Nie Z, Petukhova A, Kumacheva E, Properties and emerging applications of self-assembled structures made from inorganic nanoparticles, Nature Nanotechnology, 2010, 5(1), pp.15.

11.

Homberger M, Simon U, On the application potential of gold nanoparticles in nanoelectronics and biomedicine, Philosophical Transactions of the Royal

249 Society A: Mathematical, Physical and Engineering Sciences, 2010, 368(1915), pp.1405. 12.

Sivaraman SK, Santhanam V, Realization of thermally durable close-packed 2d gold nanoparticle arrays using self-assembly and plasma etching, Nanotechnology, 2012, 23(25), pp.255603.

13.

Steinecker WH, Kim SK, Bohrer FI, Farina L, Kurdak Ç, Zellers ET, Electron-beam patterned monolayer-protected gold nanoparticle interface layers on a chemiresistor vapor sensor array, IEEE Sensors Journal, 2011, 11(2), pp.469.

14.

Sönnichsen C, Reinhard BM, Liphardt J, Alivisatos AP, A molecular ruler based on plasmon coupling of single gold and silver nanoparticles, Nature Biotechnology, 2005, 23(6), pp.741.

15.

Herrmann J, Müller KH, Reda T, Baxter GR, Raguse B, De Groot GJJB, Chai R, Roberts M, Wieczorek L, Nanoparticle films as sensitive strain gauges, Applied Physics Letters, 2007, 91(18).

16.

Law M, Greene LE, Johnson JC, Saykally R, Yang P, Nanowire dyesensitized solar cells, Nature Materials, 2005, 4(6), pp.455.

17.

Feynman R, There's plenty of room at the bottom, Journal of Microelectromechanical Systems, 1992, 1, pp.60.

18.

http://getaway2india.files.wordpress.com/2011/07/hampi-chariot2.jpg, accessed on 24 December 2012.

19.

http://en.wikipedia.org/wiki/File:All_Gizah_Pyramids.jpg, accessed on 23 December 2012.

20.

http://www.brightsideofnews.com/print/2011/5/6/intel-manufacturing-eventin-munich-22nm-breakthrough-analyzed.aspx, accessed on 23 December 2012.

21.

http://www.geekosystem.com/ibm-12-atom-storage/,

accessed

on

23

December 2012. 22.

Piner R, Zhu, J, Xu, F, Hong, S, Mirkin, C, "Dip-pen" nanolithography, Science, 1999, 283, pp.661.

250 23.

Whitesides G, Mathias, JP, Seto, CT, Molecular self-assembly and nanochemistry: A chemical strategy for the synthesis of nanostructures, Science, 1991, 254, pp.1312.

24.

Eric Drexler K. Engines of creation: The coming era of nanotechnology Doubleday; 1986.

25.

Bishop KJM, Wilmer CE, Soh S, Grzybowski BA, Nanoscale forces and their uses in self-assembly, Small, 2009, 5(14), pp.1600.

26.

Malaquin L, Kraus T, Schmid H, Delamarche E, Wolf H, Controlled particle placement through convective and capillary assembly, Langmuir, 2007, 23(23), pp.11513.

27.

Kraus T, Malaquin L, Schmid H, Riess W, Spencer ND, Wolf H, Nanoparticle printing with single-particle resolution, Nature Nanotechnology, 2007, 2(9), pp.570.

28.

Bigioni TP, Lin XM, Nguyen TT, Corwin EI, Witten TA, Jaeger HM, Kinetically driven self assembly of highly ordered nanoparticle monolayers, Nature Materials, 2006, 5(4), pp.265.

29.

Santhanam V, Liu J, Agarwal R, Andres RP, Self-assembly of uniform monolayer arrays of nanoparticles, Langmuir, 2003, 19(19), pp.7881.

30.

Desireddy A, Joshi CP, Sestak M, Little S, Kumar S, Podraza NJ, Marsillac S, Collins RW, Bigioni TP, Wafer-scale self-assembled plasmonic thin films, Thin Solid Films, 2011, 519(18), pp.6077.

31.

Eah SK, A very large two-dimensional superlattice domain of monodisperse gold nanoparticles by self-assembly, Journal of Materials Chemistry, 2011, 21(42), pp.16866.

32.

Wen T, Majetich SA, Ultra-large-area self-assembled monolayers of nanoparticles, ACS Nano, 2011, 5(11), pp.8868.

33.

Muralidharan G, Sivaraman SK, Santhanam V, Effect of substrate on particle arrangement in arrays formed by self-assembly of polymer grafted nanoparticles, Nanoscale, 2011, 3(5), pp.2138.

34.

Stoeva SI, Smetana AB, Sorensen CM, Klabunde KJ, Gram-scale synthesis of aqueous gold colloids stabilized by various ligands, Journal of Colloid And Interface Science, 2007, 309(1), pp.94.

251 35.

Klein DL, Roth R, Lim AKL, Alivisatos AP, McEuen PL, A single-electron transistor made from a cadmium selenide nanocrystal, Nature, 1997, 389(6652), pp.699.

36.

Ouyang J, Chu CW, Szmanda CR, Ma L, Yang Y, Programmable polymer thin film and non-volatile memory device, Nature Materials, 2004, 3(12), pp.918.

37.

Medley CD, Smith JE, Tang Z, Wu Y, Bamrungsap S, Tan W, Gold nanoparticle-based colorimetric assay for the direct detection of cancerous cells, Analytical Chemistry, 2008, 80(4), pp.1067.

38.

Baik JM, Zielke M, Kim MH, Turner KL, Wodtke AM, Moskovits M, Tinoxide-nanowire-based electronic nose using heterogeneous catalysis as a functionalization strategy, ACS Nano, 2010, 4(6), pp.3117.

39.

Krause CE, Otieno BA, Latus A, Faria RC, Patel V, Gutkind JS, Rusling JF, Rapid microfluidic immunoassays of cancer biomarker proteins using disposable inkjet‐printed gold nanoparticle arrays, ChemistryOpen, 2013.

40.

Fujiki A, Uemura T, Zettsu N, Akai-Kasaya M, Saito A, Kuwahara Y, Enhanced fluorescence by surface plasmon coupling of au nanoparticles in an organic electroluminescence diode, Applied Physics Letters, 2010, 96, pp.043307.

41.

Xiao Y, Yang J, Cheng P, Zhu J, Xu Z, Deng Y, Lee S, Li Y, Tang J, Surface plasmon-enhanced electroluminescence in organic light-emitting diodes incorporating au nanoparticles, Applied Physics Letters, 2012, 100, pp.013308.

42.

Sun Y, Forrest SR, Enhanced light out-coupling of organic light-emitting devices using embedded low-index grids, Nature photonics, 2008, 2(8), pp.483.

43.

Do YR, Kim Y-C, Song Y-W, Lee Y-H, Enhanced light extraction efficiency from organic light emitting diodes by insertion of a two-dimensional photonic crystal structure, Journal of Applied Physics, 2004, 96(12), pp.7629.

44.

Koo WH, Jeong SM, Araoka F, Ishikawa K, Nishimura S, Toyooka T, Takezoe H, Light extraction from organic light-emitting diodes enhanced by spontaneously formed buckles, Nature photonics, 2010, 4(4), pp.222.

252 45.

Li Y, Tang J, Xie Z, Hung L, Lau S, An efficient organic light-emitting diode with silver electrodes, Chemical Physics Letters, 2004, 386(1), pp.128.

46.

Kwon MK, Kim JY, Kim BH, Park IK, Cho CY, Byeon CC, Park SJ, Surface‐plasmon‐enhanced light‐emitting diodes, Advanced Materials, 2008, 20(7), pp.1253.

47.

Chen M-C, Yang Y-L, Chen S-W, Li J-H, Aklilu M, Tai Y, Self-assembled monolayer immobilized gold nanoparticles for plasmonic effects in small molecule organic photovoltaic, ACS Applied Materials & Interfaces, 2013, 5(3), pp.511.

48.

Nishijima Y, Rosa L, Juodkazis S, Surface plasmon resonances in periodic and random patterns of gold nano-disks for broadband light harvesting, Optics Express, 2012, 20(10), pp.11466.

49.

Ouyang J, Application of nanomaterials in two-terminal resistive-switching memory devices, Nano reviews, 2010, 1.

50.

Kim TW, Yang Y, Li F, Kwan WL, Electrical memory devices based on inorganic/organic nanocomposites, NPG Asia Materials, 2012, 4(6), pp.e18.

51.

Yang Y, Ouyang J, Ma L, Tseng RH, Chu CW, Electrical switching and bistability in organic/polymeric thin films and memory devices, Advanced Functional Materials, 2006, 16(8), pp.1001.

52.

Scott JC, Bozano LD, Nonvolatile memory elements based on organic materials, Advanced Materials, 2007, 19(11), pp.1452.

53.

Burr GW, Kurdi BN, Scott JC, Lam CH, Gopalakrishnan K, Shenoy RS, Overview of candidate device technologies for storage-class memory, IBM Journal of Research and Development, 2008, 52(4.5), pp.449.

54.

Ma L, Liu J, Pyo S, Yang Y, Organic bistable light-emitting devices, Applied Physics Letters, 2002, 80(3), pp.362.

55.

Li Y, Sinitskii A, Tour JM, Electronic two-terminal bistable graphitic memories, Nature Materials, 2008, 7(12), pp.966.

56.

Tseng RJ, Ouyang J, Chu C-W, Huang J, Yang Y, Nanoparticle-induced negative differential resistance and memory effect in polymer bistable lightemitting device, Applied Physics Letters, 2006, 88(12), pp.123506.

253 57.

Tiwari S, Rana F, Hanafi H, Hartstein A, Crabbé EF, Chan K, A silicon nanocrystals based memory, Applied Physics Letters, 1996, 68(10), pp.1377.

58.

Lee JS, Recent progress in gold nanoparticle-based non-volatile memory devices, Gold Bulletin, 2010, 43(3), pp.189.

59.

http://www.itrs.net/Links/2011ITRS/Home2011.htm, accessed on 15 June 2012.

60.

Lee C, Meteer J, Narayanan V, Kan EC, Self-assembly of metal nanocrystals on ultrathin oxide for nonvolatile memory applications, Journal of Electronic Materials, 2005, 34(1), pp.1.

61.

Turkevich J, Stevenson PC, Hillier J, A study of the nucleation and growth processes in the synthesis of colloidal gold, Discussions of the Faraday Society, 1951, 11, pp.55.

62.

Sivaraman SK, Kumar S, Santhanam V, Monodisperse sub-10 nm gold nanoparticles by reversing the order of addition in turkevich method - the role of chloroauric acid, Journal of Colloid And Interface Science, 2011, 361(2), pp.543.

63.

Sivaraman SK, Kumar S, Santhanam V, Room-temperature synthesis of gold nanoparticles - size-control by slow addition, Gold Bulletin, 2010, 43(4), pp.275.

64.

Santhanam V, Andres RP, Microcontact printing of uniform nanoparticle arrays, Nano Letters, 2004, 4(1), pp.41.

65.

Sun S, Recent advances in chemical synthesis, self-assembly, and applications of fept nanoparticles, Advanced Materials, 2006, 18(4), pp.393.

66.

Ruska

E,

Knoll

M,

Die

magnetische

sammelspule

für

schnelle

elektronenstrahlen, The magnetic concentrating coil for fast electron beams, Z. techn. Physik, 1931, 12, pp.389. 67.

Yockell-Lelièvre H, Desbiens J, Ritcey AM, Two-dimensional selforganization of polystyrene-capped gold nanoparticles, Langmuir, 2007, 23(5), pp.2843.

68.

Frens G, Controlled nucleation for the regulation of the particle size in monodisperse gold suspensions, Nature, 1973, 241, pp.20.

69.

Personal communication. Zeiss NTS GmbH, 2008.

254 70.

http://www.physics.uwo.ca/~hutter/calibration/afmcal.html, accessed on 20 June 2012.

71.

Berne BJ, Pecora R. Dynamic light scattering: With applications to chemistry, biology, and physics: Dover Publications; 2000.

72.

Mie G, Contribution to the optics of turbid media, especially colloidal metallic suspensions, Translated into English from Annales de Physique, 1908, 25(3), pp.377.

73.

Gaikwad AV, Verschuren P, Eiser E, Rothenberg G, A simple method for measuring the size of metal nanoclusters in solution, Journal of Physical Chemistry B, 2006, 110(35), pp.17437.

74.

http://fityk.nieto.pl, accessed on 20 June 2011.

75.

Shirley DA, High-resolution x-ray photoemission spectrum of the valence bands of gold, Physical Review B, 1972, 5(12), pp.4709.

76.

http://www.casaxps.com/help_manual/manual_updates/,

accessed

on

23

December 2012. 77.

http://www.victoria.ac.nz/raman/publis/codes/cobra.aspx, accessed on 12 May 2012.

78.

Galloway C, Ru ECL, Etchegoin P, An iterative algorithm for background removal in spectroscopy by wavelet transforms, Applied Spectroscopy, 2009, 63(12), pp.1370.

79.

http://users.rowan.edu/~polikar/WAVELETS/WTpart4.html. accessed on 23 December 2012.

80.

http://etd.lib.fsu.edu/theses/available/etd-1124200385039/unrestricted/09_ds_chapter2.pdf. accessed on 23 December 2012.

81.

http://homes.nano.aau.dk/kp/Ellipsometry/main.pdf, accessed on 23 December 2012.

82.

Olde Riekerink MB. Structural and chemical modification of polymer surfaces by gas plasma etching, Universiteit Twente; PhD Thesis, 2001.

83.

Kern W, Puotinen DA, Cleaning solutions based on hydrogen for use in silicon semiconductor technology, R.C.A. Review, 1970, 31(2), pp.187.

84.

George SM, Atomic layer deposition: An overview, Polymer, 2010, 1550, pp.125.

255 85.

http://www.sigmaaldrich.com/technical-documents/articles/materialmatters/the-savannah-ald-system.html, accessed on 11 June 2012.

86.

Groner M, Fabreguette F, Elam J, George S, Low-temperature al2o3 atomic layer deposition, Chemistry of Materials, 2004, 16(4), pp.639.

87.

Gonzalez-Valls I, Lira-Cantu M, Vertically-aligned nanostructures of zno for excitonic solar cells: A review, Energy and Environmental Science, 2009, 2(1), pp.19.

88.

Watanabe M, Sei H, Stonehart P, The influence of platinum crystallite size on the electroreduction of oxygen, Journal of Electroanalytical Chemistry, 1989, 261(2), pp.375.

89.

Lux KW, Rodriguez KJ, Template synthesis of arrays of nano fuel cells, Nano Letters, 2006, 6(2), pp.288.

90.

Wang L, Shi X, Kariuki NN, Schadt M, Wang GR, Rendeng Q, Choi J, Luo J, Lu S, Zhong CJ, Array of molecularly mediated thin film assemblies of nanoparticles: Correlation of vapor sensing with interparticle spatial properties, Journal of the American Chemical Society, 2007, 129(7), pp.2161.

91.

Ruiz R, Dobisz E, Albrecht TR, Rectangular patterns using block copolymer directed assembly for high bit aspect ratio patterned media, ACS Nano, 2011, 5(1), pp.79.

92.

Pietsch T, Gindy N, Mahltig B, Fahmi A, Fabrication of functional nanoobjects via self-assembly of nanostructured hybrid materials, Journal of Polymer Science, Part B: Polymer Physics, 2010, 48(14), pp.1642.

93.

Dushkin C, Yoshimura H, Nagayama K, Nucleation and growth of twodimensional colloidal crystals, Chemical Physics Letters, 1993, 204(5), pp.455.

94.

Pang J, Xiong S, Jaeckel F, Sun Z, Dunphy D, Brinker CJ, Free-standing, patternable nanoparticle/polymer monolayer arrays formed by evaporation induced self-assembly at a fluid interface, Journal of the American Chemical Society, 2008, 130(11), pp.3284.

95.

Bodnarchuk M, Kovalenko, M, Pichler, S, Fritz-Popovski, G, Hesser, G, Heiss, W, Large-area ordered superlattices from magnetic wustite/cobalt

256 ferrite core/shell nanocrystals by doctor blade casting, ACS Nano, 2010, 4, pp.423. 96.

Martin MN, Basham JI, Chando P, Eah SK, Charged gold nanoparticles in non-polar solvents: 10-min synthesis and 2d self-assembly, Langmuir, 2010, 26(10), pp.7410.

97.

Lau CY, Duan H, Wang F, He CB, Low HY, Yang JKW, Enhanced ordering in gold nanoparticles self-assembly through excess free ligands, Langmuir, 2011, 27(7), pp.3355.

98.

Lu C, Chen Z, O'Brien S, Optimized conditions for the self-organization of cdse-au and cdse-cdse binary nanoparticle superlattices, Chemistry of Materials, 2008, 20(11), pp.3594.

99.

Pauliac-Vaujour E, Moriarty P, Meniscus-mediated organization of colloidal nanoparticles, Journal of Physical Chemistry C, 2007, 111(44), pp.16255.

100.

Ge G, Brus L, Evidence for spinodal phase separation in two-dimensional nanocrystal self-assembly, Journal of Physical Chemistry B, 2000, 104(41), pp.9574.

101.

Rabani E, Reichman DR, Geissler PL, Brus LE, Drying-mediated selfassembly of nanoparticles, Nature, 2003, 426(6964), pp.271.

102.

Blunt MO, Martin CP, Ahola-Tuomi M, Pauliac-Vaujour E, Sharp P, Nativo P, Brust M, Moriarty PJ, Coerced mechanical coarsening of nanoparticle assemblies, Nature Nanotechnology, 2007, 2(3), pp.167.

103.

Maillard M, Motte L, Pileni MP, Rings and hexagons made of nanocrystals, Advanced Materials, 2001, 13(3), pp.200.

104.

Stowell C, Korgel BA, Self-assembled honeycomb networks of gold nanocrystals, Nano Letters, 2001, 1(11), pp.595.

105.

Moriarty P, Taylor MDR, Brust M, Nanostructured cellular networks, Physical Review Letters, 2002, 89(24), pp.2483031.

106.

Martin CP, Blunt MO, Pauliac-Vaujour E, Stannard A, Moriarty P, Vancea I, Thiele U, Controlling pattern formation in nanoparticle assemblies via directed solvent dewetting, Physical Review Letters, 2007, 99(11).

257 107.

Stannard A, Martin CP, Pauliac-Vaujour E, Moriarty P, Thiele U, Dual-scale pattern formation in nanoparticle assemblies, Journal of Physical Chemistry C, 2008, 112(39), pp.15195.

108.

Ohara PC, Heath JR, Gelbart WM, Self-assembly of submicrometer rings of particles from solutions of nanoparticles, Angewandte Chemie (International Edition in English), 1997, 36(10), pp.1078.

109.

Govor LV, Reiter G, Parisi J, Bauer GH, Self-assembled nanoparticle deposits formed at the contact line of evaporating micrometer-size droplets, Physical Review E - Statistical, Nonlinear, and Soft Matter Physics, 2004, 69(6 1), pp.061609.

110.

Pauliac-Vaujour E, Stannard A, Martin CP, Blunt MO, Notingher I, Moriarty PJ, Vancea I, Thiele U, Fingering instabilities in dewetting nanofluids, Physical Review Letters, 2008, 100(17), pp.176102.

111.

Park J, Zheng H, Lee WC, Geissler PL, Rabani E, Alivisatos AP, Direct observation of nanoparticle superlattice formation by using liquid cell transmission electron microscopy, ACS Nano, 2012, 6(3), pp.2078.

112.

Stannard A, Dewetting-mediated pattern formation in nanoparticle assemblies, Journal of Physics Condensed Matter, 2011, 23(8), pp.083001.

113.

Smith G, Dielectric constants for mixed media, Journal of Physics D: Applied Physics, 2001, 10(4), pp.L39.

114.

Sancho-Parramon J, Janicki V, Zorc H, On the dielectric function tuning of random metal-dielectric nanocomposites for metamaterial applications, Optics Express, 2010, 18(26), pp.26915.

115.

Nagy N, Deák A, Hórvölgyi Z, Fried M, Agod A, Bársony I, Ellipsometry of silica nanoparticulate langmuir-blodgett films for the verification of the validity of effective medium approximations, Langmuir, 2006, 22(20), pp.8416.

116.

Johnson PB, Christy R, Optical constants of the noble metals, Physical Review B, 1972, 6(12), pp.4370.

117.

Love JC, Wolfe DB, Haasch R, Chabinyc ML, Paul KE, Whitesides GM, Nuzzo RG, Formation and structure of self-assembled monolayers of

258 alkanethiolates on palladium, Journal of the American Chemical Society, 2003, 125(9), pp.2597. 118.

Lin XM, Jaeger HM, Sorensen CM, Klabunde KJ, Formation of long-rangeordered nanocrystal superlattices on silicon nitride substrates, Journal of Physical Chemistry B, 2001, 105(17), pp.3353.

119.

Sivaraman SK. Metal nanoparticle engineering: Development of a process ‘toolkit’

for

synthesis

and

self-assembly

into

functional

nanoscale

architectures Indian Institute of Science; Dissertation, July 2012. 120.

Yim TJ, Choi H, Zhang X, Placement of alkanethiol-capped au nanoparticles using organic solvents, Journal of Colloid And Interface Science, 2010, 346(1), pp.17.

121.

Sherrington DC, Preparation, structure and morphology of polymer supports, Chemical Communications, 1998, (21), pp.2275.

122.

Mukherjee R, Das S, Das A, Sharma SK, Raychaudhuri AK, Sharma A, Stability and dewetting of metal nanoparticle filled thin polymer films: Control of instability length scale and dynamics, ACS Nano, 2010, 4(7), pp.3709.

123.

http://murov.info/webercises.htm, accessed on 25 March 2013.

124.

Tarasov V. Control of the self-assembly of alkanethiol-coated gold nanoparticles in the solid state, Massachusetts Institute of Technology; 2008.

125.

http://www.chemicalbook.com/ChemicalProductProperty, accessed on 25 March 2013.

126.

Petri D, Characterization of spin-coated polymer films, Journal of the Brazilian Chemical Society, 2002, 13(5), pp.695.

127.

Klecha E, Arfaoui I, Richardi J, Ingert D, Pileni MP, 2d silver nanocrystal ordering modulated by various substrates and revealed using oxygen plasma treatment, Physical Chemistry Chemical Physics, 2011, 13(7), pp.2953.

128.

Bansmann J, Kielbassa S, Hoster H, Weigl F, Boyen HG, Wiedwald U, Ziemann P, Behm RJ, Controlling the interparticle spacing of au-salt loaded micelles and au nanoparticles on flat surfaces, Langmuir, 2007, 23(20), pp.10150.

259 129.

Laskar IR, Watanabe S, Hada M, Yoshida H, Li J, Iyoda T, Tuning surface interactions to control shape and array behavior of diblock copolymer micelles on a silicon substrate, Surface Science, 2009, 603(4), pp.625.

130.

Yockell-Lelièvre H, Gingras D, Vallée R, Ritcey AM, Coupling of localized surface plasmon resonance in self-organized polystyrene-capped gold nanoparticle films, Journal of Physical Chemistry C, 2009, 113(51), pp.21293.

131.

Yerushalmi-Rozen R, Klein J, Dewetting of solvated polymer brushes by thin liquid films, Journal of Physics Condensed Matter, 1997, 9(37), pp.7753.

132.

Dussaud AD, Troian SM, Dynamics of spontaneous spreading with evaporation on a deep fluid layer, Physics of Fluids, 1998, 10, pp.23.

133.

Kaliappan SK, Cappella B, Temperature dependent elastic–plastic behaviour of polystyrene studied using afm force–distance curves, Polymer, 2005, 46(25), pp.11416.

134.

Nyamjav D, Ivanisevic A, Structural and adhesion properties of surfaces functionalized with polyelectrolytes and polystyrene particles, Talanta, 2005, 67(3), pp.503.

135.

Martin JI, Wang ZG, Schick M, Effects of polymer brush self-assembly on spreading and thin film stability, Langmuir, 1996, 12(20), pp.4950.

136.

Deegan RD, Bakajin O, Dupont TF, Huber G, Nagel SR, Witten TA, Contact line deposits in an evaporating drop, Physical Review E - Statistical Physics, Plasmas, Fluids, and Related Interdisciplinary Topics, 2000, 62(1 B), pp.756.

137.

Kralchevsky PA, Denkov ND, Paunov VN, Velev OD, Ivanov IB, Yoshimura H, Nagayama K, Formation of two-dimensional colloid crystals in liquid films under the action of capillary forces, Journal of Physics: Condensed Matter, 1994, 6(23A), pp.A395.

138.

Paunov VN, Kralchevsky PA, Denkov ND, Nagayama K, Lateral capillary forces between floating submillimeter particles, Journal of Colloid And Interface Science, 1993, 157(1), pp.100.

139.

de Gennes PG, Polymers at an interface; a simplified view, Advances in Colloid and Interface Science, 1987, 27(3-4), pp.189.

140.

Israelachvili JN. Intermolecular and surface forces. Second edition ed1992.

260 141.

Chen Y, Fu J, Ng KC, Tang Y, Cheng W, Free-standing polymer–nanoparticle superlattice sheets self-assembled at the air–liquid interface, Crystal Growth & Design, 2011, 11(11), pp.4742.

142.

Bassou N, Rharbi Y, Role of bénard-marangoni instabilities during solvent evaporation in polymer surface corrugations, Langmuir, 2009, 25(1), pp.624.

143.

Cai Y, Zhang Newby BM, Polymeric microstructure arrays consequence of marangoni flow-induced water droplets, Applied Physics A: Materials Science and Processing, 2010, 100(4), pp.1221.

144.

Tang C, Bang J, Stein GE, Fredrickson GH, Hawker CJ, Kramer EJ, Sprung M, Wang J, Square packing and structural arrangement of abc triblock copolymer spheres in thin films, Macromolecules, 2008, 41(12), pp.4328.

145.

Sun CH, Min WL, Linn NC, Jiang P, Jiang B, Large-scale assembly of periodic nanostructures with metastable square lattices, Journal of Vacuum Science and Technology B: Microelectronics and Nanometer Structures, 2009, 27(3), pp.1043.

146.

Zhang J, Liu Y, Ke Y, Yan H, Periodic square-like gold nanoparticle arrays templated by self-assembled 2d DNA nanogrids on a surface, Nano Letters, 2006, 6(2), pp.248.

147.

Ji Q, Acharya S, Hill JP, Richards GJ, Ariga K, Multi-dimensional control of surfactant-guided assemblies of quantum gold particles, Advanced Materials, 2008, 20(21), pp.4027.

148.

Nie Z, Fava D, Kumacheva E, Zou S, Walker GC, Rubinstein M, Selfassembly of metal-polymer analogues of amphiphilic triblock copolymers, Nature Materials, 2007, 6(8), pp.609.

149.

Niikura K, Iyo N, Higuchi T, Nishio T, Jinnai H, Fujitani N, Ijiro K, Gold nanoparticles coated with semi-fluorinated oligo (ethylene glycol) produce sub-100 nm nanoparticle vesicles without templates, Journal of the American Chemical Society, 2012, 134(18), pp.7632.

150.

Chen J, Saeki F, Benjamin J, Cang H, Cobb MJ, Li Z-Y, Au L, Zhang H, Kimmey MB, Li X, Gold nanocages: Bioconjugation and their potential use as optical imaging contrast agents, Nano Letters, 2005, 5(3), pp.473.

261 151.

Hu M, Chen J, Li Z-Y, Au L, Hartland GV, Li X, Marquez M, Xia Y, Gold nanostructures: Engineering their plasmonic properties for biomedical applications, Chemical Society Reviews, 2006, 35(11), pp.1084.

152.

Hou TH, Lee J, Shaw JT, Kan EC, Flash memory scaling: From material selection to performance improvement, In MRS Proceedings, Cambridge University Press, 2008, 1071, pp.3.

153.

http://www.eetimes.com/electronics-news/4234608/Flash-revenue-expectedto-overtake-DRAM-in-2012, accessed on 25 July 2012.

154.

http://www.electronics.ca/publications/products/Flash-Memory-Market-to2020.html, accessed on 25 June 2012.

155.

Lai SK, Flash memories: Successes and challenges, IBM Journal of Research and Development, 2008, 52(4-5), pp.529.

156.

Paolo Pavan LL, Andrea Marmiroli. Floating gate devices: Operation and compact modeling: Kluwer Academic Publishers; 2004.

157.

http://www.hitequest.com/Kiss/Flash_terms.htm, accessed on 20 June 2012.

158.

http://www.rambus.com/assets/documents/110331_gsa_holt.pdf, accessed on 20 June 2012.

159.

http://pftle.net/?cat=47, accessed on 20 June 2012.

160.

http://www.itrs.net/Links/2011Winter/12_Lithography.pdf, accessed on 20 June 2012.

161.

Kim K, Technology for sub-50 nm dram and nand flash manufacturing, Electron Devices Meeting, IEDM Technical Digest, IEEE International, 2005, pp.323.

162.

Lu CY, Hsieh KY, Liu R, Future challenges of flash memory technologies, Microelectronic Engineering, 2009, 86(3), pp.283.

163.

Kim SJ, Lee JS, Flexible organic transistor memory devices, Nano Letters, 2010, 10(8), pp.2884.

164.

http://nanoparticles.org/pdf/BanerjeeS.pdf, accessed on 20 June 2012.

165.

Derhacobian N, Hollmer SC, Gilbert N, Kozicki MN, Power and energy perspectives of nonvolatile memory technologies, Proceedings of the IEEE, 2010, 98(2), pp.283.

262 166.

Pavel AA, Khan MA, Kirawanich P, Islam NE, Impact of high-κ dielectric and metal nanoparticles in simultaneous enhancement of programming speed and retention time of nano-flash memory, Solid-State Electronics, 2008, 52(10), pp.1536.

167.

Hou TH, Lee C, Narayanan V, Ganguly U, Kan EC, Design optimization of metal nanocrystal memory - part i: Nanocrystal array engineering, IEEE Transactions on Electron Devices, 2006, 53(12), pp.3095.

168.

Chang JJ, Nonvolatile semiconductor memory devices, Proceedings of the IEEE, 1976, 64(7), pp.1039.

169.

Laibowitz RB, Stiles PJ, Charge storage on small metal particles, Applied Physics Letters, 1971, 18(7), pp.267.

170.

Frohman-Bentchkowsky D, Lenzlinger M, Charge transport and storage in metal-nitride-oxide-silicon (mnos) structures, Journal of Applied Physics, 1969, 40(8), pp.3307.

171.

Fogle AD, Darling D, Blish Ii RC, Daszko E, Flash memory under cosmic and alpha irradiation, IEEE Transactions on Device and Materials Reliability, 2004, 4(3), pp.371.

172.

Cester A, Wrachien N, Gasperin A, Paccagnella A, Portoghese R, Gerardi C, Radiation tolerance of nanocrystal-based flash memory arrays against heavy ion irradiation, IEEE Transactions on Nuclear Science, 2007, 54(6), pp.2196.

173.

Wrachien N, Cester A, Portoghese R, Gerardi C, Investigation of proton and x-ray irradiation effects on nanocrystal and floating gate memory cell arrays, IEEE Transactions on Nuclear Science, 2008, 55(6), pp.3000.

174.

Choi S, Kim SS, Chang M, Hwang H, Jeon S, Kim C, Highly thermally stable tin nanocrystals as charge trapping sites for nonvolatile memory device applications, Applied Physics Letters, 2005, 86(12), pp.1.

175.

Duguay S, Grob JJ, Slaoui A, Le Gall Y, Amann-Liess M, Structural and electrical properties of ge nanocrystals embedded in sio2 by ion implantation and annealing, Journal of Applied Physics, 2005, 97(10), pp.1.

176.

Liu WL, Lee PF, Dai JY, Wang J, Chan HLW, Choy CL, Song ZT, Feng SL, Self-organized ge nanocrystals embedded in hfalo fabricated by pulsed-laser

263 deposition and application to floating gate memory, Applied Physics Letters, 2005, 86(1), pp.013110. 177.

Leuw CC, Chen ST, Liu FK, Spin-coating-derived gold-nanoparticle memory, Journal of the American Ceramic Society, 2010, 93(10), pp.3142.

178.

Guan W, Long S, Liu M, Li Z, Hu Y, Liu Q, Fabrication and charging characteristics of mos capacitor structure with metal nanocrystals embedded in gate oxide, Journal of Physics D: Applied Physics, 2007, 40(9), pp.2754.

179.

Ferdousi F, Sarkar J, Tang S, Shahrjerdi D, Akyol T, Tutuc E, Banerjee SK, Protein-assembled nanocrystal-based vertical flash memory devices with al 2o 3 integration, Journal of Electronic Materials, 2009, 38(3), pp.438.

180.

Miura A, Uraoka Y, Fuyuki T, Yoshii S, Yamashita I, Floating nanodot gate memory fabrication with biomineralized nanodot as charge storage node, Journal of Applied Physics, 2008, 103(7).

181.

Hong AJ, Liu CC, Wang Y, Kim J, Xiu F, Ji S, Zou J, Nealey PF, Wang RL, Metal nanodot memory by self-assembled block copolymer lift-off, Nano Letters, 2010, 10(1), pp.224.

182.

Lee JW, Ryu SW, Shin DO, Kim BH, Kim SO, Choi YK, Geometric effects of nanocrystals in nonvolatile memory using block copolymer nanotemplate, Solid-State Electronics, 2009, 53(6), pp.640.

183.

Lee JS, Kim YM, Kwon JH, Sim JS, Shin H, Sohn BH, Jia Q, Multilevel data storage memory devices based on the controlled capacitive coupling of trapped electrons, Advanced Materials, 2011, 23(18), pp.2064.

184.

Gupta RK, Kusuma DY, Lee P, Srinivasan M, Covalent assembly of gold nanoparticles for nonvolatile memory applications, ACS Applied Materials & Interfaces, 2011, 3(12), pp.4619.

185.

Gupta RK, Krishnamoorthy S, Kusuma DY, Lee PS, Srinivasan M, Enhancing charge-storage capacity of non-volatile memory devices using templatedirected assembly of gold nanoparticles, Nanoscale, 2012, 4(7), pp.2296.

186.

Lee JS, Cho J, Lee C, Kim I, Park J, Kim YM, Shin H, Lee J, Caruso F, Layer-by-layer assembled charge-trap memory devices with adjustable electronic properties, Nature Nanotechnology, 2007, 2(12), pp.790.

264 187.

Paul S, Pearson C, Molloy A, Cousins MA, Green M, Kolliopoulou S, Dimitrakis P, Normand P, Tsoukalas D, Petty MC, Langmuir-blodgett film deposition of metallic nanoparticles and their application to electronic memory structures, Nano Letters, 2003, 3(4), pp.533.

188.

Kolliopoulou S, Dimitrakis P, Normand P, Zhang HL, Cant N, Evans SD, Paul S, Pearson C, Molloy A, Petty MC, Tsoukalas D, Integration of organic insulator and self-assembled gold nanoparticles on si mosfet for novel nonvolatile memory cells, Microelectronic Engineering, 2004, 73-74, pp.725.

189.

Ristau R, Tiruvalam R, Clasen PL, Gorskowski EP, Harmer MP, Kiely CJ, Hussain I, Brust M, Electron microscopy studies of the thermal stability of gold nanoparticle arrays, Gold Bulletin, 2009, 42(2), pp.133.

190.

Dufourcq J, Bodnar S, Gay G, Lafond D, Mur P, Molas G, Nieto JP, Vandroux L, Jodin L, Gustavo F, Baron T, High density platinum nanocrystals for non-volatile memory applications, Applied Physics Letters, 2008, 92(7), pp.073102.

191.

Han ST, Zhou Y, Xu ZX, Huang LB, Yang XB, Roy V, Microcontact printing of ultrahigh density gold nanoparticle monolayer for flexible flash memories, Advanced Materials, 2012, 24(26), pp.3556.

192.

Robertson J, Band offsets of wide-band-gap oxides and implications for future electronic devices, Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures, 2000, 18(3), pp.1785.

193.

Pu J, Kim SJ, Kim YS, Cho BJ, Evaluation of gadolinium oxide as a blocking layer of charge-trap flash memory cell, Electrochemical and Solid-State Letters, 2008, 11(9), pp.H252.

194.

Schaadt DM, Yu ET, Sankar S, Berkowitz AE, Charge storage in co nanoclusters embedded in sio2 by scanning force microscopy, Applied Physics Letters, 1999, 74(3), pp.472.

195.

Guillemot C, Budau P, Chevrier J, Marchi F, Comin F, Alandi C, Bertin F, Buffet N, Wyon C, Mur P, Imaging of stored charges in si quantum dots by tapping and electrostatic force microscopy, Europhysics Letters, 2002, 59(4), pp.566.

265 196.

Huang S, Banerjee S, Tung RT, Oda S, Electron trapping, storing, and emission in nanocrystalline si dots by capacitance-voltage and conductancevoltage measurements, Journal of Applied Physics, 2003, 93(1), pp.576.

197.

Beyer V, Eicchorn F, von Borany J, Mucklich A, Muller T, Ion beam synthesis and charge storage behavior of au nanocrystals in thin sio2 layers, Journal of Applied Physics, 2008, 104(1), pp.024512.

198.

Cheng X, Xu D, Song Z, He D, Yu Y, Zhao Q, Shen D, Characterization of gadolinium oxide film by pulse laser deposition, Applied Surface Science, 2009, 256(3), pp.921.

199.

Artyushkova K, Levendosky S, Atanassov P, Fulghum J, Xps structural studies of nano-composite non-platinum electrocatalysts for polymer electrolyte fuel cells, Topics in Catalysis, 2007, 46(3-4), pp.263.

200.

Shaw J, Hou, TH, Raza, H, Kan EC, Statistical metrology of metal nanocrystal memories with 3-d finte-element analysis, IEEE Transactions on Electron Devices, 2009, pp.1729.

201.

Prime D, Paul S, First contact-charging of gold nanoparticles by electrostatic force microscopy, Applied Physics Letters, 2010, 96(4).

202.

Korgel BA, Zaccheroni N, Fitzmaurice D, Melting transition of a quantum dot solid: Collective interactions influence the thermally-induced order-disorder transition of a silver nanocrystal superlattice, Journal of the American Chemical Society, 1999, 121(14), pp.3533.

203.

Xu J, Wong C, Effect of the polymer matrices on the dielectric behavior of a percolative high-k polymer composite for embedded capacitor applications, Journal of Electronic Materials, 2006, 35(5), pp.1087.

204.

Meng S, Basceri C, Busch B, Derderian G, Sandhu G, Leakage mechanisms and dielectric properties of tin-based metal-insulator-metal capacitors, Applied Physics Letters, 2003, 83(21), pp.4429.

205.

Baptista P, Pereira E, Eaton P, Doria G, Miranda A, Gomes I, Quaresma P, Franco R, Gold nanoparticles for the development of clinical diagnosis methods, Analytical and Bioanalytical Chemistry, 2008, 391(3), pp.943.

206.

Cho WS, Cho M, Jeong J, Choi M, Cho HY, Han BS, Kim SH, Kim HO, Lim YT, Chung BH, Acute toxicity and pharmacokinetics of 13 nm-sized peg-

266 coated gold nanoparticles, Toxicology and Applied Pharmacology, 2009, 236(1), pp.16. 207.

Larsen EKU, Nielsen T, Wittenborn T, Birkedal H, Vorup-Jensen T, Jakobsen MH, Ostergaard L, Horsman MR, Besenbacher F, Howard KA, Kjems J, Sizedependent accumulation of pegylated silane-coated magnetic iron oxide nanoparticles in murine tumors, ACS Nano, 2009, 3(7), pp.1947.

208.

Lipka J, Semmler-Behnke M, Sperling RA, Wenk A, Takenaka S, Schleh C, Kissel T, Parak WJ, Kreyling WG, Biodistribution of peg-modified gold nanoparticles following intratracheal instillation and intravenous injection, Biomaterials, 2010, 31(25), pp.6574.

209.

Slot JW, Geuze HJ, A new method of preparing gold probes for multiplelabeling cytochemistry, European Journal of Cell Biology, 1985, 38(1), pp.87.

210.

Brust M, Walker M, Bethell D, Schiffrin DJ, Whyman R, Synthesis of thiolderivatised gold nanoparticles in a two-phase liquid-liquid system, Journal of the Chemical Society, Chemical Communications, 1994, (7), pp.801.

211.

Park J, An K, Hwang Y, Park JEG, Noh HJ, Kim JY, Park JH, Hwang NM, Hyeon T, Ultra-large-scale syntheses of monodisperse nanocrystals, Nature Materials, 2004, 3(12), pp.891.

212.

Ojea-Jiménez I, Romero FM, Bastús NG, Puntes V, Small gold nanoparticles synthesized with sodium citrate and heavy water: Insights into the reaction mechanism, Journal of Physical Chemistry C, 2010, 114(4), pp.1800.

213.

Sidhaye DS, Prasad BLV, Many manifestations of digestive ripening: Monodispersity, superlattices and nanomachining, New Journal of Chemistry, 2011, 35(4), pp.755.

214.

Prasad BLV, Sorensen CM, Klabunde KJ, Gold nanoparticle superlattices, Chemical Society Reviews, 2008, 37(9), pp.1871.

215.

Prasad BLV, Stoeva SI, Sorensen CM, Klabunde KJ, Digestive-ripening agents for gold nanoparticles: Alternatives to thiols, Chemistry of Materials, 2003, 15(4), pp.935.

216.

Manson J, Kumar D, Meenan BJ, Dixon D, Polyethylene glycol functionalized gold nanoparticles: The influence of capping density on stability in various media, Gold Bulletin, 2011, 44(2), pp.99.

267 217.

Larson-Smith K, Pozzo DC, Competitive adsorption of thiolated polyethylene glycol and alkane-thiols on gold nanoparticles and its effect on cluster formation, Langmuir, 2012, 28(37), pp.13157.

218.

Boily

JF,

Seward

TM,

Palladium

(ii)

chloride

complexation:

Spectrophotometric investigation in aqueous solutions from 5 to 125°c and theoretical insight into pd-cl and pd-oh2 interactions, Geochimica et cosmochimica acta, 2005, 69(15), pp.3773. 219.

Turkevich J, Kim G, Palladium: Preparation and catalytic properties of particles of uniform size, Science, 1970, 169(3948), pp.873.

220.

Gaspar A, Dieguez L, Dispersion stability and methylcyclopentane hydrogenolysis in pd/al2o3 catalysts, Applied Catalysis A: General, 2000, 201(2), pp.241.

221.

Jin R, Cao YC, Hao E, Métraux GS, Schatz GC, Mirkin CA, Controlling anisotropic nanoparticle growth through plasmon excitation, Nature, 2003, 425(6957), pp.487.

222.

Tang B, Xu S, An J, Zhao B, Xu W, Photoinduced shape conversion and reconstruction of silver nanoprisms, Journal of Physical Chemistry C, 2009, 113(17), pp.7025.

223.

Gao J, Fu J, Lin C, Lin J, Han Y, Yu X, Pan C, Formation and photoluminescence of silver nanoparticles stabilized by a two-armed polymer with a crown ether core, Langmuir, 2004, 20(22), pp.9775.

224.

Dhanalakshmi L, Udayabhaskararao T, Pradeep T, Conversion of double layer charge-stabilized [email protected] colloids to thiol passivated luminescent quantum clusters, Chemical Communications, 2012, 48(6), pp.859.

225.

Yuan X, Luo Z, Zhang Q, Zhang X, Zheng Y, Lee JY, Xie J, Synthesis of highly fluorescent metal (ag, au, pt, and cu) nanoclusters by electrostatically induced reversible phase transfer, ACS Nano, 2011, 5(11), pp.8800.

226.

Radziuk D, Skirtach A, Sukhorukov G, Shchukin D, Möhwald H, Stabilization of silver nanoparticles by polyelectrolytes and polyethylene glycol, Macromolecular Rapid Communications, 2007, 28(7), pp.848.

268 227.

Harriman A, Millward GR, Neta P, Richoux MC, Thomas JM, Interfacial electron-transfer reactions between platinum colloids and reducing radicals in aqueous solution, Journal of Physical Chemistry, 1988, 92(5), pp.1286.

228.

Sen F, Gökagaç G, Different sized platinum nanoparticles supported on carbon: An xps study on these methanol oxidation catalysts, The Journal of Physical Chemistry C, 2007, 111(15), pp.5715.

229.

Sanyal U, Datta R, Jagirdar BR, Colloidal calcium nanoparticles: Digestive ripening in the presence of a capping agent and coalescence of particles under an electron beam, RSC Advances, 2012, 2(1), pp.259.

230.

Cheng L, Cao D, Aggregation of polymer-grafted nanoparticles in good solvents: A hierarchical modeling method, Journal of Chemical Physics, 2011, 135(12).

231.

Hexemer A, Stein GE, Kramer EJ, Magonov S, Block copolymer monolayer structure measured with scanning force microscopy moiré patterns, Macromolecules, 2005, 38(16), pp.7083.

232.

Dietrick McGinnis B, Dean Adams V, Joe Middlebrooks E, Degradation of ethylene glycol using fenton's reagent and uv, Chemosphere, 2001, 45(1), pp.101.

233.

Saito N, Wu Y, Hayashi K, Sugimura H, Takai O, Principle in imaging contrast in scanning electron microscopy for binary microstructures composed of organosilane self-assembled monolayers, The Journal of Physical Chemistry B, 2003, 107(3), pp.664.

234.

Shon HK, Son M, Park KM, Rhee CK, Song NW, Park HM, Moon DW, Lee TG, Tof-sims imaging and spectroscopic analyses of peg-conjugated aunps, Surface and Interface Analysis, 2011, 43(1-2), pp.628.

235.

Shon HK, Kim SJ, Park HM, Moon DW, Song NW, Lee TG, Characterization of peg-conjugated gold nanoparticles using a statistical analysis on time-offlight sims images, Surface and Interface Analysis, 2013, 45(1), pp.225.

236.

Tantakitti F, Burk-Rafel J, Cheng F, Egnatchik R, Owen T, Hoffman M, Weiss DN, Ratner DM, Nanoscale clustering of carbohydrate thiols in mixed self-assembled monolayers on gold, Langmuir, 2012, 28(17), pp.6950.

269 237.

Martin MN, Li D, Dass A, Eah SK, Ultrafast, 2-min synthesis of monolayerprotected gold nanoclusters (d< 2 nm), Nanoscale, 2012, 4(14), pp.4091.

238.

Majzik A, Fülöp L, Csapó E, Bogár F, Martinek T, Penke B, Bíró G, Dékány I, Functionalization of gold nanoparticles with amino acid, β-amyloid peptides and fragment, Colloids and Surfaces B: Biointerfaces, 2010, 81(1), pp.235.

239.

Mak JSW, Rutledge SA, Abu-Ghazalah RM, Eftekhari F, Irizar J, Tam NCM, Zheng G, AS H, Recent developments in optofluidic-assisted raman spectroscopy, Progress in Quantum Electronics, 2013, 37(1), pp.1.

240.

Mak SWJ. Raman characterization of colloidal nanoparticles using hollowcore photonic crystal fibers, University of Toronto; Dissertation, 2011.

241.

Boca S, Rugina D, Pintea A, Barbu-Tudoran L, Astilean S, Flower-shaped gold nanoparticles: Synthesis, characterization and their application as sersactive tags inside living cells, Nanotechnology, 2010, 22(5), pp.055702.

242.

Hussain I, Brust M, Papworth AJ, Cooper AI, Preparation of acrylatestabilized gold and silver hydrosols and gold-polymer composite films, Langmuir, 2003, 19(11), pp.4831.

243.

Pechini MP. Method of preparing lead and alkaline earth titanates and niobates and coating method using the same to form a capacitor: US Patent 3,330,697; 1967.

244.

Warren SC, Perkins MR, Adams AM, Kamperman M, Burns AA, Arora H, Herz E, Suteewong T, Sai H, Li Z, Werner J, Song J, Werner-Zwanziger U, Zwanziger JW, Grätzel M, Disalvo FJ, Wiesner U, A silica sol-gel design strategy for nanostructured metallic materials, Nature Materials, 2012, 11(5), pp.460.

245.

Kakihana M, Yoshimura M, Synthesis and characteristics of complex multicomponent oxides prepared by polymer complex method, Bulletin of the Chemical Society of Japan, 1999, 72(7), pp.1427.

246.

Wilkinson G, Gillard RD, McCleverty JA. Comprehensive coordination chemistry: Ligands: Pergamon Press; 1987.

247.

Kim SG, Wang WN, Iwaki T, Yabuki A, Okuyama K, Low-temperature crystallization of barium ferrite nanoparticles by a sodium citrate-aided synthetic process, Journal of Physical Chemistry C, 2007, 111(28), pp.10175.

270 248.

Cushing BL, Kolesnichenko VL, O'Connor CJ, Recent advances in the liquidphase syntheses of inorganic nanoparticles, Chemical Reviews, 2004, 104(9), pp.3893.

249.

Fang TT, Wu MS, Tsai JD, 13c nmr study of the solution chemistry of barium titanium citrate gels prepared using the pechini process, Journal of the American Ceramic Society, 2002, 85(12), pp.2984.

250.

Fang TT, Tsay JD, Effect of ph on the chemistry of the barium titanium citrate gel and its thermal decomposition behavior, Journal of the American Ceramic Society, 2001, 84(11), pp.2475.

251.

Ojea-Jiménez I, Bastús NG, Puntes V, Influence of the sequence of the reagents addition in the citrate-mediated synthesis of gold nanoparticles, The Journal of Physical Chemistry C, 2011, 115(32), pp.15752.

252.

Kumar S, Gandhi KS, Kumar R, Modeling of formation of gold nanoparticles by citrate method, Industrial and Engineering Chemistry Research, 2007, 46(10), pp.3128.

253.

Muhammed MAH, Aldeek F, Palui G, Trapiella-Alfonso L, Mattoussi H, Growth of in situ functionalized luminescent silver nanoclusters by direct reduction and size focusing, ACS Nano, 2012, 6(10), pp.8950.

254.

Shiers MJ, Leech R, Carmalt CJ, Parkin IP, Kenyon AJ, Self-assembled ultrahigh aspect ratio silver nanochains, Advanced Materials, 2012, 24(38), pp.5227.

255.

Elliott BJ, Willis WB, Bowman CN, Polymerization kinetics of pseudocrown ether network formation for facilitated transport membranes, Macromolecules, 1999, 32(10), pp.3201.

256.

Elliott BJ, Willis WB, Bowman CN, Pseudo-crown ethers as fixed site carriers in facilitated transport membranes, Journal of Membrane Science, 2000, 168(1), pp.109.

257.

Park IH, Johnson Jr CS, Gabriel DA, Probe diffusion in polyacrylamide gels as observed by means of holographic relaxation methods: Search for a universal equation, Macromolecules, 1990, 23(5), pp.1548.

271 258.

Bryant SJ, Anseth KS, Hydrogel properties influence ecm production by chondrocytes photoencapsulated in poly (ethylene glycol) hydrogels, Journal of biomedical materials research, 2001, 59(1), pp.63.

259.

Bryant SJ, Chowdhury TT, Lee DA, Bader DL, Anseth KS, Crosslinking density

influences

chondrocyte

metabolism

in

dynamically

loaded

photocrosslinked poly (ethylene glycol) hydrogels, Annals of biomedical engineering, 2004, 32(3), pp.407. 260.

Sun Y, Yin Y, Mayers BT, Herricks T, Xia Y, Uniform silver nanowires synthesis by reducing agno3 with ethylene glycol in the presence of seeds and poly (vinyl pyrrolidone), Chemistry of Materials, 2002, 14(11), pp.4736.

261.

Shapiro YE, Structure and dynamics of hydrogels and organogels: An nmr spectroscopy approach, Progress in Polymer Science, 2011, 36(9), pp.1184.

262.

Sakurai K, Jeong Y, Koumoto K, Friggeri A, Gronwald O, Sakurai S, Okamoto S, Inoue K, Shinkai S, Supramolecular structure of a sugar-appended organogelator explored with synchrotron x-ray small-angle scattering, Langmuir, 2003, 19(20), pp.8211.

263.

Kuo CT, Yu JY, Huang MJ, Chen C, On the size evolution of goldmonolayer-protected clusters by ligand place-exchange reactions: The effect of headgroup− gold interactions, Langmuir, 2010, 26(9), pp.6149.

264.

Gupta R, Bradoo S, Saxena R, Rapid purification of extracellular tannase using polyethylene glycol–tannic acid complex, Letters in applied microbiology, 1997, 24(4), pp.253.

265.

Sperling RA, Rivera Gil P, Zhang F, Zanella M, Parak WJ, Biological applications of gold nanoparticles, Chemical Society Reviews, 2008, 37(9), pp.1896.

266.

Koliopoulou S, Dimitrakis P, Goustouridis D, Normand P, Pearson C, Petty MC, Radamson H, Tsoukalas D, Metal nano-floating gate memory devices fabricated at low temperature, Microelectronic Engineering, 2006, 83(4-9 SPEC. ISS.), pp.1563.

272