Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Surface Engineering Assisted Directed Assembly-based Printing of
Electronic Devices
A Dissertation Presented
By
Salman Ali Abbasi
to
The Department of Mechanical and Industrial Engineering
in partial fulfillment of the requirements
for the degree of
Doctor of Philosophy
in the field of
Mechanical Engineering (Materials Science)
Northeastern University
Boston, Massachusetts
December 2019
ii
DEDICATION
To my parents, Shafquat and Noshaba, who have always been unconditional in their support,
and who strive each day to teach me what is noble in life.
iii
ABSTRACT
The potential environmental and economic benefits of avoiding the resource-intensive, subtractive
semiconductor fabrication processes have driven the development of additive manufacturing
techniques such as inkjet, screen, transfer and aerosol jet printing. Most printing techniques,
however, have poor resolution, while techniques capable of nanoscale printing suffer from low
throughput and poor morphology control over large areas. Thus, there is need to improve the
resolution, throughput and control of existing processes, and for a high-throughput, readily
integrable process capable of printing micro and nanoscale features with controllable thickness on
a variety of substrates for various device applications.
The first part of this thesis enhances the understanding of previously demonstrated capillary force
assisted and transfer-based printing methods with nanoscale resolution. The effect of geometrical
constraints on the thickness and morphology of printed features is investigated for fluidic assembly
on resist patterned substrates by varying the resist height, feature width, and withdrawal speed. A
confinement effect, which increases thickness for narrower, deeper features is identified and
explained. For transfer printing onto polymers, the structural properties needed for successfully
embedding materials into a polymer substrate from a template are investigated by varying crosslink
density and molecular weight of crosslinked and amorphous polymer films, respectively. Our
results indicate that an increase in either characteristic adversely affects the polymer’s ability to
be printed on, which points at chain mobility as a key parameter.
In the second part of the thesis, a newly devised approach called the Fast-Fluidic Assembly and
Transfer method, or FFAsT, is presented. This technique utilizes dip-coating to control the
iv
dewetting front on a chemically heterogenous surface and exploits its high differential wettability
to achieve selective printing of nanomaterials. Lithography-assisted silanized surfaces are printed
with conductive, semiconducting and insulating materials from 500 µm down to 300 nm at a high
speed. The thickness and morphology of the printed film is tailored by changing the withdrawal
speed and modulating the ink properties (surface tension, nanomaterial concentration, viscosity).
The newly developed process has also been used to print carbon nanotube field-effect transistors
on both rigid and flexible substrates.
The last part of this thesis investigates nanomaterial-based consumer products from environmental
and policy perspectives. Nanomaterials used in the manufacture of devices and other products have
a high embodied energy and a potential risk when released to the environment during the lifecycle
of the product. We address the first concern by a cumulative energy demand analysis to compare
the energy requirements to print 14-nm node transistors on a one cm2 silicon chip using directed
assembly with the top-down conventional fabrication. Results indicate that the embodied energy
of conventionally fabricated structures could exceed that of printed circuits by an order of
magnitude. Moreover, nanomaterials contribute less to the overall embodied energy than surface
preparation processes. Finally, the role of current producer responsibility landscape in the U.S. in
preventing the release of nanomaterials from nano-enabled consumer products is explored.
Relevant product stewardship state and federal laws enacted until 2017 applicable to architectural
paint, household batteries, nanopharmaceutical drugs and consumer electronics are considered.
We estimate that the current legal framework only captures ~ 10% of the ENMs used in consumer
products in the U.S. A cost-benefit study for recovering quantum dots from television displays is
also performed to assess the viability of recovery for reuse initiatives.
v
ACKNOWLEDGEMENTS
I would like to express my profound gratitude to my advisors, Prof. Jackie Isaacs and Prof. Ahmed
Busnaina for their guidance, patience and constant support throughout these years, without which,
I could not have completed this work.
I must thank my committee members, Prof. Marilyn Minus, Prof. Yung Joon Jung, and Prof.
Moneesh Upmanyu for their time and their feedback to improve this work. I thank Prof. Matthew
Eckelman and Prof. Christopher Bosso, for their valuable insights on the environmental and policy
part of my thesis. A special thanks to Prof. Randall Erb, for his support and advice, and for being
an excellent teacher.
I would like to thank Dr. Zhimin Chai whose constructive criticism helped improve my research
skills. I thank Dr. Hobin Jeong, David McKee and Dr. Sivasubramanian Somu for teaching me
microfabrication, and for their friendship. I also want to acknowledge support from some other
group members, Adnan Korkmaz, Ahmed Abdelaziz, Jukyung Lee, and Sharon Kotz.
I want to thank Serkan Erbis, Burak Sancaktar, Amir Namin, Ahmed Abbasi, Kanwar Ankush,
Saurabh and Alolika Mukhpadhyay for their friendship and support throughout the past few years.
The Friday bicycle trips with Anthony Childress and Wei-Hong Wang were a true stress relief.
I thank Prof. Delcie Durham, who pushed me to present my work at the Sustainable Nano
conference in Boston where I met my PhD advisor, and Carol Lynn Alpert, who sat next to me
during dinner and put in a good word for me to Prof. Jackie Isaacs.
I want to thank my parents for their support and encouragement, and my sisters, Fatima and
Ayesha, for always setting the bar high, and pushing me to do better.
Finally, I would like to express the deepest gratitude to my wife, Rafia, for her untiring support
and constant encouragement throughout this journey. She made me breakfast each day, taught me
mathematics for the qualifying exam, took care of our son, and made the figures in my papers look
good, all while managing her own PhD research. Her resolve, hard work, and perseverance will
always serve as an example for me and our son, Daoud.
vi
TABLE OF CONTENTS
1. Introduction ............................................................................................................................. 1
1.1. Motivation for Research ................................................................................................... 1
1.2. Theoretical Background and Literature Review .............................................................. 2
1.2.1. Important Definitions and Governing Forces ........................................................... 2
1.2.2. Inkjet, Screen and Flexographic Printing.................................................................. 5
1.2.3. Microcontact and Nanotransfer Printing (nTP) ........................................................ 6
1.2.4. Resist-patterned Fluidic Assembly ........................................................................... 6
1.2.5. Nanoscale Offset Printing Process - NanoOPS ........................................................ 7
1.3. Research Objectives and Thesis Outline .......................................................................... 8
2. Fluidic Assembly Process ...................................................................................................... 10
2.1. Assembly Mechanism .................................................................................................... 10
2.1.1. Thickness Control ................................................................................................... 13
2.1.2. Morphology Characterization and Optimization .................................................... 19
2.2. Applications of Resist-patterned Fluidic Assembly ....................................................... 24
2.2.1. Flexible Transparent Silver Grid Electrode ............................................................ 24
2.2.2. CNT-based Real-time Flexible Lactate Sensor....................................................... 36
3. Fast-Fluidic Assembly Method ............................................................................................. 47
3.1. Motivation for a Rapid Printing Process ........................................................................ 47
3.2. Literature Review of Selective Deposition on Chemically Heterogenous Surfaces ...... 48
vii
3.3. Mechanism of the Fast-Fluidic Assembly Process ........................................................ 51
3.4. Development of the Fast-Fluidic Assembly Process...................................................... 53
3.5. Characterization of the Fast-Fluidic Assembly Process ................................................. 56
3.5.1. Controlling Film Thickness .................................................................................... 56
3.5.2. Characterizing Film Morphology ........................................................................... 61
3.5.3. Electrical Properties of Printed Films ..................................................................... 63
3.6. Applications of the Fast-Fluidic Assembly Process ....................................................... 66
3.6.1. Wafer-scale Printing of CNTs for Field-effect Transistors .................................... 66
4. Directed Assembly for Transfer Printing .............................................................................. 68
4.1. NanoOPS Process Description ....................................................................................... 68
4.2. Transfer Challenge: Polymer Characterization .............................................................. 71
4.2.1. Motivation for Polymer Characterization ............................................................... 71
4.2.2. Polymer Properties Characterization ...................................................................... 73
4.2.3. Crosslinked Polymers: Effect of Crosslink Density ............................................... 77
4.2.4. Amorphous Polymers: Effect of Molecular Weight ............................................... 84
4.2.5. Summary of Characterization and Challenges ........................................................ 87
4.3. Transfer Printing through FFAsT Templates instead of Damascene Templates ........... 88
4.4. Applications of Transfer Printing Process ..................................................................... 90
4.4.1. Fully Printed, All-Carbon Flexible FETs................................................................ 90
5. Lifecycle Management of Nano-enabled Products ............................................................... 99
viii
5.1. Energy Demand for Printing Nanoscale Electronics...................................................... 99
5.1.1. CED Goal and Scope for Printed FET .................................................................. 100
5.1.2. Process and material inventory analysis ............................................................... 102
5.1.3. Computing energy demand ................................................................................... 104
5.1.4. Conclusions ........................................................................................................... 109
5.2. End-of-Life Management of Nanoenabled Products: Product Stewardship Strategies 110
5.2.1. Introduction ........................................................................................................... 110
5.2.2. Methods................................................................................................................. 112
5.2.3. Analysis................................................................................................................. 115
5.2.4. Discussion ............................................................................................................. 130
6. Summary and Future Work ................................................................................................. 137
6.1. Summary ...................................................................................................................... 137
6.2. Contributions ................................................................................................................ 138
6.3. Future Work ................................................................................................................. 139
6.3.1. Fluidic Assembly .................................................................................................. 139
6.3.2. FFAsT Process ...................................................................................................... 140
6.3.3. NanoOPS Transfer Printing .................................................................................. 140
6.3.4. End-of-Life Management of Nano-enabled Products ........................................... 141
7. Appendix A.......................................................................................................................... 158
8. Appendix B .......................................................................................................................... 163
ix
LIST OF FIGURES
Figure 1-1 - The effect of withdrawal speed on film thickness for the capillary and draining dip-
coating regimes.2 Capillary regime is active for withdrawal speeds smaller than ~ 0.1 mm/sec. .. 4
Figure 2-1 – Evaporation of liquid stranded between particles causing convective flux.6 ........... 10
Figure 2-2 – Depiction of particle and water flux established due to evaporation at the meniscus.4
....................................................................................................................................................... 11
Figure 2-3 - Substrate preparation for resist-patterned fluidic assembly. ..................................... 12
Figure 2-4 – (a) A schematic depiction of relevant parameters. The printed film thickness as a
function of resist height for a 2 µm wide feature for different withdrawal speeds (b). The ratio of
the film thickness to the resist thickness as a function of feature width for different resist heights
are shown in (b) and (c), respectively. .......................................................................................... 14
Figure 2-5 - Mechanism explaining the confinement effect leading to thicker deposition for
narrower features. Bottom shows cross-sections of 2 µm and 100 µm printed at 0.05 mm/min using
2 µm thick photoresist. .................................................................................................................. 16
Figure 2-6 – Ratio of film thickness to resist thickness plotted against the withdrawal speed for (a)
2 µm and (b) 10 µm wide features for different resist thicknesses. .............................................. 18
Figure 2-7 - Silver printed in a 2 µm wide and 1.3 µm thick photoresist feature shows a concave
profile. ........................................................................................................................................... 20
Figure 2-8 - Bilayer patterning method to functionalize the photoresist with an organosilane
compound. ..................................................................................................................................... 21
x
Figure 2-9 - (a) and (b) show the profiles obtained from printing silver on samples developed for
10 sec and 30 sec, respectively before removing the photoresist. The profile of the printed structure
after removing the photoresist is shown in (d) while (e) shows the cross-sectional profile of the
same captured using confocal microscopy. .................................................................................. 22
Figure 2-10 - (a) and (b) explain the concept of edge roughness. The isolated edge of a 350 nm
wide and 100 nm deep feature in PMMA is shown in (c), while (d) and (e) show features of similar
width printed with silver using 100 nm and 600 nm deep PMMA layer, respectively. ............... 23
Figure 2-11 - Sample preparation steps are summarized in (a)-(d). The dip-coating-based printing
process is demonstrated in (e) while the mechanism based on the difference in wettability of the
PET and the photoresist is visually explained in (f), (g) and (h). ................................................. 27
Figure 2-12 - SEM micrographs of the printed grid (1.3 µm thick photoresist, 0.5 mm/min) after
hotplate sintering are shown in (a) and (b). The same sample after flash sintering is shown in (c)
and (d) and the dark areas in (c) show silver crystals. AFM micrograph of a printed silver line
(hotplate sintered) is shown in (e) and the extracted cross-sectional profile is shown in (f). A large
area 3D-projection captured with a confocal microscope is shown in (g). ................................... 28
Figure 2-13 - Sheet resistance versus withdrawal speed for 0.5 µm (a) and 1.3 µm (b) thick
photoresists. The inset in (b) shows the I-V curves (hotplate only) showing linear behavior. The
average thickness and sheet resistance for each withdrawal speed for 0.5 µm (c) and 1.3 µm (d)
thick photoresists. ......................................................................................................................... 31
Figure 2-14 - Optical transmittance of the grid electrode printed with sheet resistance of 43 Ω/sq
(1.3 µm thick photoresist, 0.25 mm/min). Change in resistance of the electrode due to bending to
xi
different radii of curvature for a single cycle (b) and the change in resistance as a function of total
bending cycles for 2 mm radius of curvature (c). ......................................................................... 34
Figure 2-15 - SEM micrographs of the nanoscale grid printed on a glass wafer through e-beam
lithography are shown in (a), (b) and (c). The I-V characteristics of the same electrode for
difference sintering processes are shown in (d). ........................................................................... 35
Figure 2-16 - (a) The CNT directed assembly mechanism and (b) The sensor fabrication and
functionalization procedure shown schematically. ....................................................................... 38
Figure 2-17 -(a) Raman spectra of the printed CNTs before and after PPy modification. The peaks
at 936 cm-1, 983 cm-1, 1047 cm-1, 1257 cm-1and 1411 cm-1 for modified CNTs correspond to
those of electropolymerized polypyrrole films. (b) shows the effect of PPy modification on the
resistance drift with sensor immersed in 1X-phosphate buffered saline (PBS) at 0.1 V. ............. 41
Figure 2-18 - (a) Response of the sensor for lactate concentration of 1.5 mM – 20 mM and (b)
normalized response plotted against the corresponding lactate concentration. ............................ 43
Figure 2-19 - The normalized response is plotted against concentration for each day (a). Intra-
batch characterization of the response of three sensors showing minimal response variation (b).
....................................................................................................................................................... 45
Figure 3-1 – The dip-coating process shown schematically in the evaporative (a) and draining (b)
regimes.5 The shifting of stagnation point that shifts to towards the reservoir surface resulting in a
thicker entrained film is depicted in (c).2 ...................................................................................... 48
Figure 3-2 - Substrate preparation for the FFAsT process. .......................................................... 54
Figure 3-3 - Examples of features printed with the FFAsT method. ............................................ 55
xii
Figure 3-4 - Film thickness attained for different withdrawal speed for DEG, IPA and DIW (a) and
a comparison with Darhuber’s model (b). .................................................................................... 58
Figure 3-5 - Morphology and cross-sectional profiles of features printed with IPA, DIW and DEG
diluted inks. The mechanism showing transition between controlled dewtting dip-coating and
uncontrolled dewetting is also shown. .......................................................................................... 59
Figure 3-6 - Effect of particle concentration on film thickness at the withdrawal speed of 40
mm/min for DEG-based ink.......................................................................................................... 60
Figure 3-7 – 3D sideview and cross-sectional profiles of 200 µm square (left), 100 µm square, and
100 µm hollow square with 10 µm linewidth wall. ...................................................................... 61
Figure 3-8 - Cross-sectional profiles and 3D models of the same feature dried at (a) 30 ˚C (RH ~
45%), (b) 30 ˚C (RH ~ 10%) and (c) 75 ˚C (RH ~ 30%) using 20 wt% Ag ink in DEG. ........... 62
Figure 3-9 - SEM image and corresponding isolated edge (yellow outline) for a 350 nm wide
printed feature 100 mm/min.......................................................................................................... 63
Figure 3-10 - Particle size increase and sintering of silver from 200 °C to 400 °C. The
corresponding decrease in sheet resistance is shown in (f). .......................................................... 65
Figure 3-11 - Device structure (a) and the CNTs printed into channels (b). The transfer and output
characteristics are shown in (c) and (d), respectively. .................................................................. 67
Figure 4-1 – Damascene template fabrication process. ................................................................ 68
Figure 4-2 - Electrophoretic assembly process.1 ........................................................................... 69
Figure 4-3 - NanoNex tool used for transfer printing.7 ................................................................. 70
xiii
Figure 4-4 - Transfer printing results for different polymer films above their glass transition
temperatures. The damascene template before and after CNT assembly is shown in (a) and (b),
respectively. (c), (d), (e) and (f) show the results for TPU, PET, PEN and PETG. ..................... 73
Figure 4-5 - Calorimetry curves for the four TPU grades investigated in this study.................... 78
Figure 4-6 – CNT residue on templates post-transfer process for 1074A (a), 1095A (b), NPT(c)
and 1055D(d). ............................................................................................................................... 80
Figure 4-7 - (a) and (b) show storage moduli and tan delta for spincast and hot-rolled films of
1055D-TPU, respectively. The DSC curves of the same TPU grade in the form of pellets, spincast
film, and hot-rolled film are compared in (c). .............................................................................. 82
Figure 4-8 - DSC curves for the PETG samples. .......................................................................... 85
Figure 4-9 - Residue thickness as a function of transfer time for each of the PETG samples. .... 87
Figure 4-10 - Silver transfer printed to a PETG substrate (a) showing clean transfer and sharp
edges. A cross-section of the transferred electrode pair is shown in (c) while the roughness of the
electrode is shown in (d). A cross-section of the entire feature embedded into the substrate is seen
in (e). Assembled MWCNTs are shown in (f) and electropolymerized polypyrrole into 100 µm x
20 µm channels is shown in (g). ................................................................................................... 90
Figure 4-11 - Schematic depictions of substrate patterning and assembly processes used for SWNT
and MWNT are shown in (a) and (b), respectively. The assembly mechanism for SWNTs and
MWNTs is explained in (c) and (d), respectively. The assembled SWNTs on a Si wafer are shown
in (e), while (f) shows the same for MWNTs. A confocal microscope image and cross-sectional
profile of MWNTs electrophoretically assembled on a Si wafer are shown in (g). ..................... 94
xiv
Figure 4-12 - (a) Polyimide substrate preparation for transfer printing and the sequential transfer
of SWNTs and MWNTs (b). The prepared substrate with printed CNT FETs is shown in (c), and
(d) shows an optical image of an array of source-drain electrodes with SWNTs. ........................ 96
Figure 4-13 - Transfer (a) and output (b) characteristics of a CNT FET. (c) On/off ratio of 27
devices. (d) Field-effect mobility of 27 devices. .......................................................................... 98
Figure 5-1 - A typical FinFET structure (not drawn to scale – based on a similar sketch from Intel).3
..................................................................................................................................................... 102
Figure 5-2 - Process sequence for printing 14 nm node FET. ................................................... 103
Figure 5-3 - Probability density function of the profit made through recycling of quantum dots in
televisions from 2020-2030 for Scenario A (($300/television) (a) and (b) and Scenario B
($1500/television) (c) and (d) assuming 100% yield (a) and (c) and 50% yield (b) and (d). The
probability density function for Scenario A($300/television) with decreasing recovery cost is
shown in (e) while (f) compares the profit for decreasing vs fixed recovery cost for a period of 10
years. ........................................................................................................................................... 126
Figure 5-4 - Sensitivity analyses for a) Scenario A ($300/television) and b) Scenario B
($1500/television) showing the profitability of the recovery process to be most affected by the
mass of QDs in on-surface type televisions. For the higher recovery cost (Scenario B), the percent
of faulty screens affects the profit margin more profoundly than for lower recovery cost (Scenario
A). The mass of QDs in on-edge type screens affects the profitability the least. ....................... 129
Figure 6-1 - A graphical comparoson of the Fast-Fluidic Assembly process with the previously
developed processes at the CHN. The fastest drying time for the ink has been considered for this
chart............................................................................................................................................. 139
xv
LIST OF TABLES
Table 4-1 - Summary of properties of the characterized films. .................................................... 74
Table 4-2 - Mechanical properties of polymers tested for transfer. .............................................. 75
Table 4-3 - Mechanical properties of the four grades of TPU obtained from Lubrizol. ............... 77
Table 4-4 - Summary of TPU characterization for transfer printing. ........................................... 83
Table 4-5 - Properties of the four grades of PETG. ...................................................................... 84
Table 4-6 - Comparison of damascene template fabrication with the SAM functionalized template
fabrication. .................................................................................................................................... 89
Table 5-1 - Material and energy inputs for each process step used for printing a cm2 silicon chip
with FETs. ................................................................................................................................... 106
Table 5-2 - The type, and the total amount of ENMs used in select product categories worldwide
and in the U.S. in 2010. ‘Broader Category’ refers to the category from Keller et al. 2013 to
quantify the consumption of ENMs. The ‘Selected Application’ refers to each product category
investigated for this study. Selected applications are classified within the broader categories.
Assumptions made to estimate the amounts are included, with 12% share of the U.S. in the global
ENM use is assumed for all product categories. ......................................................................... 116
Table 5-3 - Total annual amount of collectable and recoverable nanomaterials for nano-enabled
applications investigated: paints and coatings, household batteries, pharmaceutical drugs, and
consumer electronics. .................................................................................................................. 131
1
1. Introduction
1.1. Motivation for Research
The aggressive scaling and densification of integrated circuits (ICs) has remarkably increased the
processing and storage capacities of modern computers. Faster computers and memory structures
enable us to address incredibly complex challenges facing our planet. The abundance and
favorable properties of silicon such as the relative ease of doping and a suitable native oxide has
played a pivotal role in advancements in electronics manufacturing. Over many years, the
semiconductor fabrication industry has developed and optimized material deposition and etching
techniques that are used with lithography to make the silicon chips used in modern-day electronics.
Semiconductor fabrication currently faces three challenges. First, semiconductor processing has
become exceedingly complex and costly. A state-of-the-art fabrication facility could cost around
$20 billion to build; an amount afforded only by a few large corporations.8 The manufacturing
costs are passed on to the consumers, which hegemonizes access to cutting-edge technology by
the privileged, and isolates the marginalized. Secondly, because IC fabrication requires energy
intensive, high-vacuum, subtractive film deposition and etching processes, the environmental cost
of semiconductor fabrication is increasing.9 Thirdly, many emerging applications require circuits
on novel substrates such as paper or textiles, which are currently incompatible with typical
semiconductor fabrication processes.
To decrease the economic and environmental burden of semiconductor manufacturing and increase
its versatility, researchers are devising solution-based processes to additively print ICs and sensors
instead of vapor deposition and etching processes.10 Characteristically, solution-based printing
processes do not require high-vacuum and temperature, which can significantly lower the
2
economic and environmental costs of manufacturing electronics. Various printing techniques such
as inkjet printing,11 gravure printing,12 flexographic printing,13 and transfer printing14 are
demonstrated. However, most of these techniques are currently limited by their resolution, which
is around 10 µm and only be used for low to mid-level electronics.15 Therefore, processes capable
of printing in the nanoscale regime comparable to that of traditional semiconductor fabrication
industry are required. Nanoscale features can be printed by coopting lithography techniques used
in semiconductor fabrication as lithography is key in defining the minimum feature size. Following
this approach, many lithography assisted printing techniques have been proposed, which will be
introduced later in this chapter. Generally, these methods face challenges pertaining to scalability,
complexity and control that are discussed in the next section.
1.2.Theoretical Background and Literature Review
1.2.1. Important Definitions and Governing Forces
The research presented herein focuses on directed assembly-based printing processes that strongly
depend on the tribological properties of solid and liquid surfaces. Some key terms and concepts,
therefore, need to be explained. A more thorough literature review is provided at the start of each
chapter discussing the subject at hand.
1) Intermolecular forces: Weak forces that occur due to coulombic interactions between
permanent or temporary charge centers of different molecules. van der Waals (vdW) and
hydrogen bonding are important examples of intermolecular forces. Intermolecular forces
are critical in most surface related phenomenon such as adhesion and interfacial energy.16
2) Surface energy and surface tension: Surface energy surface energy and surface tension is a
direct consequence of intermolecular forces. Unbalanced atomic or molecular forces at the
3
surface of a solid or liquid give rise to surface energy (for solids) or surface tension (for
liquids).16
3) Contact angle: the contact angle is the angle made by a solid and the tangent to the liquid-
gas interface and indicates the interfacial energy between the solid and the liquid.16
4) Hydrophilicity and hydrophobicity: Hydrophilicity and hydrophobicity are terms used to
indicate the wettability of a surface by water. If a surface ‘likes’ water, and the contact
angle is significantly less than 90°, the surface is said to be hydrophilic. A contact angle of
greater than 90° between a surface and water indicates non-favorable interaction and thus,
hydrophobicity.16
5) Capillary action/force: the flow of liquid in narrow channels due to intermolecular forces.
Capillary force defines the climbing of a liquid on a solid surface.
6) Dip-coating: Dip-coating is a widespread film deposition method in which a substrate is
immersed and withdrawn from a liquid at a fixed speed.2 Depending on the withdrawal
speed, dip-coating may operate in either the capillary (evaporation controlled) or the
draining (gravity/viscosity controlled) regime. In the evaporation-controlled regime, film
thickness decreases with increasing withdrawal speed, while the opposite is true for the
draining regime as shown in Figure 1-1.
The solid layer thickness deposited from a solution through dip-coating in the capillary
regime is given by:
ℎ𝑓 =𝑐𝑀𝐸
𝛼𝜌𝐿𝑈0
4
where ℎ𝑓 is the final thickness, 𝑐 is the solute concentration, 𝑀 is the solute molecular
weight, 𝐸 is the evaporation rate, 𝛼 is the film porosity, 𝜌 is the solute density, 𝐿 is the film
width, and 𝑈0 is the withdrawal speed. Thickness control in the draining regime will be
discussed in detail in Chapter 3.
7) Hydroxylation: Hydroxylation is generating -OH groups on a surface by treating it with an
oxidizing agent such as a strong acid or a plasma. Hydroxylation polarizes surfaces and
increases their surface energy. The density if hydroxylation varies with the process used
and the chemical nature of the surface.17 Hydroxylated surfaces are hydrophilic (like
water) due to hydrogen bonding between surface -OH groups and water.
8) Silanization: The reaction of -OH groups on a surface with an organosilane molecule
deposits a thin self-assembled monolayer (SAM) on it, which significantly lowers the
surface energy and makes it hydrophobic.18 This process is called silanization.
Figure 1-1 - The effect of withdrawal speed on film thickness for the capillary and draining dip-coating regimes.2
Capillary regime is active for withdrawal speeds smaller than ~ 0.1 mm/sec.
5
9) Nanomaterials: a material less than 100 nm in at least one dimension. Examples include
carbon nanotubes (CNTs), graphene, and silver nanoparticles.
10) Polymers: Polymers are a class of materials that consist of long chains of repeating units
of smaller molecules called monomers. The chemical characteristics of monomers and
nature of their bonding to each other determines the physical, chemical and mechanical
properties of polymers.19
1.2.2. Inkjet, Screen and Flexographic Printing
Inkjet printing is amongst the most commercially successful electronics printing techniques to
date.11 The mechanism is implemented without highly sophisticated hardware. A nozzle controlled
through piezoelectric action is used to eject droplets of a colloidal ink on a surface where it dries.
Patterns of conducting, semiconducting and insulating materials can be printed on a variety of
substrates. Despite being truly additive, inkjet printing is unsuitable for large-scale integration as
the reliable maximum resolution is tens of microns.
Screen printing involves a porous, patterned stencil placed on top of the substrate to be printed
followed by spreading and squeegeeing of the desired ink.20 The ink passes through the stencil and
the pattern is transferred to the underlying substrate. While screen printing is a robust printing
method, it has a poor resolution and a poor control of the thickness of the printed layer.
In flexographic or relief printing, a flexographic plate with the desired features is inked by rotating
it on a cylinder. The inked plate then makes contact with a substrate, often a flexible film loaded
and spun on a cylinder, to transfer the ink. This method is best suited for printing on flexible
substrates. While some novel alterations of flexography claim capability of printing submicron
features,15 the minimum feature size is limited to several µm.
6
1.2.3. Microcontact and Nanotransfer Printing (nTP)
Transfer printing is a printing technique that numerous studies have demonstrated as a viable
option for printing electronics in the nanoscale regime.21-22 Microcontact printing is a soft-
lithography based transfer printing technique in which ink is transferred from a
polydimetylesiloxane (PDMS) stamp to a target substrate.23 The stamp is made by pouring, curing
and removing an elastomer such as PDMS on a master mold. The method was first devised by
Whitesides et al at Harvard University.24 A similar method called nanotransfer printing (nTP)
capable of printing nanoscale features has been developed by Rogers et al.25 In nTP, photocurable
PDMS stamps patterned with nanoscale features are used to selectively transfer vapor deposited
films to a substrate of choice through favorable interfacial surface energies.26 While nTP is similar
to microcontact printing, it utilizes vapor deposition for precise and repeatable inking of the stamp
and uses a new chemistry of PDMS elastomer, which helps avoid some of the problems faced by
microcontact printing such as shrinkage and fracture of the stamp during use. Although promising,
the method still relies on high-vacuum deposition, which is undesirable for low cost
manufacturing.
1.2.4. Resist-patterned Fluidic Assembly
Fluidic assembly is a well-known particle assembly method that relies on the shape of the meniscus
of a solvent on a wettable (hydrophilic) surface. When such a surface is immersed in an appropriate
liquid, the meniscus rises and wets it. Evaporation at the ‘foot’ of the meniscus leaves particles
behind. The evaporated liquid molecules are replaced by molecules from the bulk of the
suspension, which establishes a convective flow from the bulk to the suspension that also brings
along suspended particles. If the substrate is withdrawn out of the particle suspension at a slow,
7
fixed rate during the process, particles are assembled across the entire surface. This particle
assembly method was first formally demonstrated in 1993,27 but was later modified at the Center
for High-Rate Nanomanufacturing (CHN) to assemble particles into nano and microscale features
patterned using UV or electron beam sensitive resists.28-29 However, while fluidic assembly has
been demonstrated, control over thickness and morphology remains explored. Moreover, the
method is yet to be demonstrated for functional devices on the wafer-scale.
1.2.5. Nanoscale Offset Printing Process - NanoOPS
This is an offset printing method to print nanoscale features onto plastic substrates through directed
assembly and transfer that has been developed at the CHN. A damascene template with nano or
microscale conductive features is used to electrophoretically assemble and transfer nanomaterials
onto polymer substrates. Optical or ebeam lithography is used with film deposition and etching
techniques to create the desired conductive pattern on the surface of the wafer. The area outside
the conductive pattern is silanized, which makes it hydrophobic (water contact angle > 90°). The
surface of the conductive film - typically gold - is piranha treated to render it moderately
hydrophilic (water contact angle ~ 30°). This differential wetting behavior promotes selectivity
during the electrophoretic assembly process.
The offset printing process can be divided into two distinct subprocesses: 1) the inking process
and 2) the transfer process. During inking, electric field is used to assemble nanomaterials into
conductive patterns on a damascene template that serves as the ‘stamp’. The assembled particles
are then transferred to a polymer substrate by applying pressure to bring the template and the
substrate into contact followed by applying heat. Heat softens the polymer substrate to promote
embedment of the assembled nanomaterial film into the polymer while pressure ensures conformal
8
contact during the transfer process. During embedment, softening of the polymer substrate
increases the total contact area between the nanoparticle film and the polymer substrate from the
top and the sides. Due to the increase in contact area, the adhesion force between the polymer and
the nanoparticle film also increases, and eventually overcomes the adhesion between the particle
film and the template. At this point, the template is separated from the substrate and the
nanoparticle film undergoes an adhesive failure at the nanoparticle/template interface causing the
nanomaterials to be transferred to the polymer.
Some issues facing the NanoOPS process remain unaddressed. First, it is unknown as to which
structural and physical properties of the polymer substrate affects the transfer process. Moreover,
heating the polymer substrate causes volumetric expansion, which obstructs the alignment of
subsequent layers and decreases the fidelity of the printed features. Hence, to control the transfer
process at the nanoscale, it is important to understand the underlying physics. However, very little
literature specifically on embedding-based printing is available. Most available work associates
the transfer success to the glass transition temperature of the polymer substrate and no study
systematically analyzes all variables of interest such as the structure and type of the thermoplastic,
and the effect of glass transition temperature on each type of polymer with respect to transfer
printing. To fill this gap, this thesis studies the polymer properties and physical processes relevant
to embedment-type transfer printing.
1.3.Research Objectives and Thesis Outline
Based on the discussion above, the research presented herein had four objectives:
a) investigate thickness morphology control of features printed via fluidic assembly
- understand how the resist height and width correlate to printed film thickness
9
- optimize the morphology of features printed using the fluidic assembly process
- use the fluidic assembly process to print functional devices
b) develop a scalable printing method to complement existing printing technology
- improve the scalability of the fluidic assembly process
- investigate physical parameters to control film thickness and morphology
- demonstrate functional devices using the newly developed method
c) investigate the polymer properties conducive to embedment-type transfer printing
- understand the polymer properties needed for successful transfer printing
- demonstrate use of transfer printing for device applications
d) analyze the end-of-life (EoL) of the printed devices from environmental, economic and
policy perspectives.
The first chapter explains the motivation behind the presented research and introduces a few widely
known printing technologies developed elsewhere as well as at the CHN. The second chapter
focuses on the fluidic assembly process and discusses the approach used to understand the role of
resist thickness and its surface properties in controlling the thickness and morphology of the
printed features. The third chapter introduces an innovative surface engineering assisted dip-
coating method for fast fluidic assembly and its application to rigid and flexible printed electronics.
The investigation into the role of polymer properties in the transfer printing process is described
in the fourth chapter. The fifth chapter discusses the potential energy benefits of printing
electronics at the nanoscale, and end-of-life (EoL) of nano-enabled products from the viewpoint
of existing waste management policy in the United States (U.S.). The last chapter summarizes the
major findings and outlines possible future work.
10
2. Fluidic Assembly Process
2.1.Assembly Mechanism
The fluidic assembly process, to state simply, is driven by the evaporation of liquid at the meniscus
on a solid/liquid interface. When a hydrophilic substrate (water contact angle ~ 0°) is immersed in
a particle suspension, the liquid climbs the substrate and forms a concave meniscus due to capillary
action. Evaporation at the meniscus and leaves behind the particles suspended therein. The
evaporated molecules at the meniscus are replaced by molecules from the bulk of the suspension,
thereby establishing a convective flow that also brings along particles from the suspension.
While the dynamics of an evaporating meniscus had been thoroughly investigated before,30 the
first formal investigation into the mechanism behind particle assembly via evaporation at a
meniscus was published Denkov et al in 1992.6 They hypothesized that particles were assembled
onto a surface when the thickness of the liquid meniscus incident on it became less than the particle
size. Moreover, they suggested that the evaporation of the remaining liquid stranded between the
particles increases the curvature of the local menisci, which creates a flow of liquid from the bulk
to the particles that brings the suspended particles along as depicted in Figure 2-1. The process
repeats as the meniscus evaporates and slowly recedes causing particles to assemble on the surface.
Figure 2-1 – Evaporation of liquid stranded between particles causing convective flux.6
11
While Denkov et al. had also shown that the rate and thickness of the assembly depended strongly
on the evaporation rate and surfactant concentration, a deeper study into the kinetics of the particle
assembly process was published by Dushkin et al. in 1993,31 who identified the importance of
interparticle attraction to the assembly process, and introduced a factor β to capture the difference
in the velocities of water molecules and suspended particles due to interparticle interaction (β ≈ 1
when water velocity equals particle velocity). The work published by Denkov et al. and Dushkin
et al. had demonstrated evaporation driven assembly and identified several relevant parameters.
However, their studies had two shortcomings: 1) their assembly method could only assemble
particles over small areas, and b) a robust quantitative model combining the identified parameters
was not available. This gap was addressed in 1996 by Nagayama et al. who proposed the use of
dip-coating to assemble the particles over large areas by matching the assembly rate of the particles
to the withdrawal rate of the substrate from a particle suspension.4 They had earlier provided
Figure 2-2 – Depiction of particle and water flux established due to evaporation at the meniscus.4
12
detailed insights into the capillary forces at play between particles during assembly,32 and applied
the newly developed understanding to the dip-coating-based approach depicted in Figure 2-2.
Nagayama et al.’s most significant contribution was demonstrating the formation of continuous
particle films over large areas by matching the rate of evaporation to the withdrawal speed of the
substrate. This invigorated interest in this assembly technique for various applications and the
method was extended to templated surfaces to selectively assemble particles by exploiting
wettability difference through chemical modification over large areas.33-34 Assembly on
geometrically patterned substrates using etched or lithographically patterned substrates were
demonstrated. A method to exploit the difference in wettability of a patterned photo or electron
beam sensitive resist and an underlying hydroxylated surface was developed and investigated for
nanotube assembly in nanoscale channels over large substrates at the CHN by Jung et al.28, 35-36
This process, developed at the CHN, is the focus of this study.
Consider the substrate patterning process depicted in Figure 2-3. The underlying SiO2 layer is
significantly more hydrophilic than the surrounding photoresist surface due to piranha or plasma
treatment, which generates a large number of -OH groups on the Si surface.28, 37-38 If this resist
patterned substrate is immersed in a nanoparticle suspension, the difference in the wettability of
Figure 2-3 - Substrate preparation for resist-patterned fluidic assembly.
Si/SiO2
Resist
Si/SiO2
Si/SiO2
Si/SiO2
Hydroxylation Lithography
OH OH OH OH OH OH
13
SiO2 and the photoresist will cause the shape of the menisci on both surfaces to be different, leading
to a difference in evaporation rate and hence, particle assembly on both surfaces.39-40 Dip-coating
this resist-patterned substrate at a fixed speed provides particle assembly along its entirety.
Initial research had primarily focused on assembling 2D particle arrays. At the CHN, however, the
technique was extended to 3D particle structures for device applications, for which controlling the
thickness and the morphology is critical. The next sections discuss the efforts of the authors to
understand of the factors dictating thickness and morphology of 3D structures printed using dip-
coating assisted fluidic assembly.
2.1.1. Thickness Control
Controlling film thickness is important for a printing or fabrication process. While Denkov6 and
Nagayama4 had shown that increasing particle concentration or increasing the evaporation rate (or
decreasing the withdrawal speed) typically increased particle assembly and hence layer thickness,
the extent to which the resist thickness (tR), feature width (W) and the ratio between them affect
the thickness of the printed feature (tF) is not explained by their models. This study investigates
the role of these parameters in controlling the thickness of features printed using the resist
patterned fluidic assembly process.
This study investigates the effect of resist thickness (tR) and feature width (W) on the thickness of
the printed film (tF). We assembled silver nanoparticles from an aqueous, nearly monodisperse
suspension, approximately 5 wt%, 30 nm – 50 nm (Novacentrix® JS101A diluted with deionized
water) into channels with W = 2 µm, 10 µm, 15 µm and 20 µm for tR = 0.5 µm, 1.3 µm, 1.8 µm
and 2.7 µm. Four different withdrawal speeds, 0.05 mm/min, 0.1 mm/min, 0.25 mm/min and 0.5
mm/min were used, and the assembly was performed at a relative humidity of ~ 20% inside a
14
desiccated chamber. The high concentration of silver was used to fully fill the large, microscale
features. The selectivity of particle assembly generally decreases with increasing particle
concentration.
Si/SiO2 wafers were used as substrates. Before patterning, each wafer was piranha cleaned
(2(H2SO4):1(H2O2)) and treated with O2 plasma in a reactive ion etch system (Anatech 100) for 2
min (115 W, 0.38 Torr, 15 sccm). Shipley S1805, S1813, S1818 and S1827 positive tone
photoresists, which are 0.5 µm, 1.3 µm, 1.8 µm and 2.7 µm thick (4000 RPM, 45 sec), respectively,
were used to pattern the wafers. After assembly, the resist was removed by gently rinsing followed
by immersion in acetone and isopropyl alcohol (IPA). The average feature edge-to-edge thickness
Figure 2-4 – (a) A schematic depiction of relevant parameters. The printed film thickness as a function of resist
height for a 2 µm wide feature for different withdrawal speeds (b). The ratio of the film thickness to the resist
thickness as a function of feature width for different resist heights are shown in (b) and (c), respectively.
0
0.5
1
1.5
2
2.5
3
0 0.5 1 1.5 2 2.5 3
t F(µ
m)
tR (µm)
0.05 mm/min
0.1 mm/min
0.25 mm/min
0.5 mm/min
w = 2 µm
(b)
(c) (d)
AgtR
W
tF
(a)
0
0.2
0.4
0.6
0.8
1
1.2
0 2 4 6 8 10 12 14 16 18 20 22
t F/t
R
W (µm)
0
0.2
0.4
0.6
0.8
1
1.2
0 2 4 6 8 10 12 14 16 18 20 22
t F/t
R
W (µm)
0.05 mm/min 0.25 mm/min
15
(tF) was measured using confocal microscopy (Zeiss SmartProof 5). The geometrical parameters
of interest are shown schematically in Figure 2-4(a).
As seen in Figure 2-4(b), the resist thickness (tR) dictates the film thickness (tF) almost linearly for
a 2 µm wide feature when the withdrawal speed is ~ 0.1 mm/min or less. This indicates that for
the same withdrawal speed, printed film thickness can be tuned by adjusting the resist thickness.
The relationship between resist thickness and the thickness of the printed film for the same W can
be attributed to an increased meniscus length for thicker photoresists. The sidewall of the
photoresist has a contact angle of 55˚ with water (unless it is baked upon which it increases to
nearly 80˚ but the sidewall shape and edges become rounded), which may further decrease for
particle suspensions due to surfactants and possibly polar photoresist sidewalls.41 A contact angle
of around 50˚ allows the liquid to climb the sidewall of the resist and hence the total length of the
receding meniscus inside the feature increases. Logically, the higher the sidewall, the greater the
total meniscus length will be as long as it is less than the capillary length, which is typically ~ 2
mm for most fluids on wettable surfaces.42 For example, the total length of the meniscus in a 2 µm
wide feature with a 500 nm thick photoresist is approximately 3 µm. For a 2.7 µm thick photoresist,
the length increases to 7.4 µm. The increased meniscus length increases particle assembly on the
sidewall, and hence an increase in the average thickness of the assembled nanoparticle film.
Next, it was ascertained if this thickness control technique was independent of the feature width
W. It was found that a ratio-relationship between the resist thickness and the film thickness due to
the above described effect is more pronounced for narrower features (2 µm). This is revealed if
the ratio of film thickness tF to resist thickness tR is plotted against feature width W for different
16
resist thicknesses as shown in Figure 2-4(c) for the withdrawal speed of 0.05 mm/min, which
Figure 2-5 - Mechanism explaining the confinement effect leading to thicker deposition for narrower features.
Bottom shows cross-sections of 2 µm and 100 µm printed at 0.05 mm/min using 2 µm thick photoresist.
Front View
t = 0
Side View
t = 0
Front View
t = t1
Side View
t = t1
Point A
(a)
(b)
(c)
(d)
Top View
Side ViewFront View
(e) (f)
Top View
t = 0Top View
t = t1
Average thickness: 2.147 µm Average thickness: 0.829 µm
(g) (h) W = 2 µm W = 100 µm
17
shows that the ratio tF :tR drops with increasing feature width. Therefore, it is necessary to explain
the effect of feature width (W) on the thickness of the printed film.
For this purpose, we propose an addendum along the theory proposed by Denkov et al.6 to explain
this effect. Please refer to Figure 2-5 that accompanies the following description. Imagine a
receding meniscus in a 2 µm wide feature on a hydroxylated SiO2 surface. Due to the difference
in wettability, the precursor foot of the meniscus on SiO2 is longer than that on the photoresist
sidewall. Therefore, the contact line on the resist sidewall and the underlying surface, although
connected, is spatially misaligned. The point at which the resist wall, referred to as Point A, meets
the underlying resist is of prime importance. As the meniscus on the sidewall recedes, particles
assemble at the said point reducing the width of the trench, before the meniscus on the underlying
substrate reaches Point A. When the meniscus on the substrate approaches Point A, and the
thickness of the meniscus decreases to below the thickness of the particle layer previously
assembled by the contact line on the sidewall, the increased curvature of the meniscus between the
assembled particles generates additional capillary pressure that pulls more particles to Point A that
get assembled, hence increasing the thickness of the final film. This assembly takes place on both
sidewalls, and if the assembled layer thickness is on the order of the width of the feature, the
assembled particles on both sides may bridge at the center of the feature, thereby increasing the
thickness in the center of the channel, as well. However, for significantly wider channels, the
particles getting assembled at the sidewall-substrate junction cannot form a bridge, and the layer
in the center between the sidewalls does not become significantly thicker with increasing resist
thickness. We refer to the described effect as the confinement effect. The presence of this effect is
corroborated by Figure 2-4(c) and (d), for features printed at 0.05 mm/min and 0.25 mm/min,
18
respectively. The results indicate that a deeper, high vertical aspect ratio feature can receive a
thicker nanoparticle assembly layer. Moreover, as the withdrawal speed increases, the effect is less
noticeable due to the shorter time available to the particles to reach the receding meniscus.
We have established that for features ~ 2 µm, the resist thickness can be controlled by adjusting
the height of the photoresist for withdrawal speed of 0.05 mm/min. However, it was not fully clear
if the how the relationship between the film thickness and the resist thickness may evolve as a
function of withdrawal speed. An interesting trend emerges if the ratio of the film thickness to the
resist height is plotted against the withdrawal speed for 2 µm and 10 µm wide features, as shown
in Figure 2-6(a) and (b), respectively. It is seen that tF/tR clusters tightly and approaches ‘1’ for all
resist height values when the withdrawal speed is low (≤ 0.1 mm/min). However, the film thickness
values spread over a wider range, especially for the 10 µm wide lines, as the withdrawal speed
increases. We attribute this to 0.05 mm/min being purely in the capillary regime, while the
assembly beyond 0.1 mm/min is due to the combined effect of evaporation and draining, and hence
prone to fingering instabilities.43
Figure 2-6 – Ratio of film thickness to resist thickness plotted against the withdrawal speed for (a) 2 µm and (b)
10 µm wide features for different resist thicknesses.
0
0.2
0.4
0.6
0.8
1
1.2
0 0.1 0.2 0.3 0.4 0.5
t F/t
R
Withdrawal Speed (mm/min)
0.5 um
1.3 um
1.8 um
2.7 um
0
0.2
0.4
0.6
0.8
1
1.2
0 0.1 0.2 0.3 0.4 0.5
t F/t
R
Withdrawal Speed (mm/min)
0.5 um
1.3 um
1.8 um
2.7 um
w = 2 µm w = 10 µm(a) (b)
19
We conclude this section with the following remarks on thickness control using fluidic assembly:
a) thickness is most well-controlled for features approaching the limits of optical
photolithography, which is advantageous as very large or thick features are seldom used in
modern electronics,
b) ideally, the withdrawal speed should be minimized to operate in the purely capillary
regime,
c) the ratio of tF to tR for wider features may be increased by using even higher concentration
inks albeit at a greater loss of selectivity,
d) fluidic assembly is a slow process, and the concentration of ink can change considerably if
used consecutively for a few days due to evaporation leading to inter-batch variation in
thickness.
2.1.2. Morphology Characterization and Optimization
Controlling the morphology of a feature is as important as controlling its thickness. The
morphology of printed features was examined using atomic force microscopy (AFM) and scanning
electron microscopy (SEM), which revealed that the profile of printed structures was concave with
ears at the edges as shown in Figure 2-7. This effect, what we call the rabbit ears effect, is more
pronounced at higher withdrawal speeds. We attribute the rabbit ears effect to the capillary action
of the ink on the sidewalls of the photoresist. As previously mentioned, the resist is not highly
hydrophobic, and the contact angle is approximately 55°, which cannot prevent the ink from
climbing the sidewall due to capillary action. The shape of the meniscus is an important factor that
controls the number of particles deposited on the sidewall. Controlling the shape of the meniscus
on the sidewall of the photoresist can suppress the rabbit ear effect by preventing the capillary
20
action of the ink on the sidewall. Curvature of the surface contacting a liquid affects the capillary
force. A detailed discussion of the effect can be found elsewhere, but generally, the apparent
contact angle decreases on a convex surface and because the capillary force is directly proportional
to the contact angle, the capillary force decreases, as well. While concave sidewalls are achievable
through isotropic dry etching, making convex sidewalls is not straightforward. Moreover, convex
surfaces may have other undesirable effects on the morphology.
The contact angle on the sidewall of the photoresist can also be increased by functionalizing the
photoresist with an organosilane molecule. Organosilanes form covalently bond with the -OH
groups on a hydroxylated surface. Here, a newly developed bilayer patterning process using a
negative tone photoresist and a lift-off resist (LOR) is described. Negative resists are UV-sensitive
polymers that crosslink when exposed to UV or heated and become resistant to chemical attack.
The process starts by hydroxylation of a Si/SiO2 substrate using O2 plasma treatment. A 300 nm
thick lift-off resist layer was then spincoated on the wafer (LOR 3A, 3000 RPM, 60 sec) and baked
at 150 °C for 5 min. This was followed by spincoating, baking and exposure of the N2020 negative
Figure 2-7 - Silver printed in a 2 µm wide and 1.3 µm thick photoresist feature shows a concave profile.
21
tone resist. Post-exposure, the development of the N2020 negative tone resist in AZ 300 metal-ion
free (MIF) developer was controlled precisely to remove the N2020 from the feature and leave a
~ 50 nm thick LOR layer residue. The patterned photoresist was then chemically crosslinked
through a 5 min hard-bake at 125 °C before being O2 plasma treated after which it was immediately
exposed to organosilane vapor for 3 min at 95 °C. Post-silanization, the LOR residue was removed
with highly controlled development in AZ 726 MIF developer (10 sec, gentle shaking). Although
the N2020 resist is hardbaked, it can be attacked by the developer in a prolonged exposure, hence
removing the silane functionalization. The bilayer patterning process is shown in Figure 2-8.
A 5 wt% ink was used to assemble Ag at 0.25 mm/min and the cross-section of each sample was
examined using the SEM as shown in Figure 2-9. Results indicate that minimizing the development
time after silanization to remove the residue LOR is critical to suppressing the climbing of ink
onto the photoresist walls. Confocal microscopy was also used to generate cross-sectional profiles
Figure 2-8 - Bilayer patterning method to functionalize the photoresist with an organosilane compound.
Undeveloped LOR
Post-development
Si/SiO2
LOR
Si/SiO2
Si/SiO2
OH OH OH OH OH
-ve N2020 PR
LORSi/SiO2
N2020LOR
N2020
LORSi/SiO2
N2020LOR
N2020
OH OH
OH
OH OH
LORSi/SiO2
N2020LOR
N2020
Si-R Si-R
Si-R
Si-R Si-R
LORSi/SiO2
N2020LOR
N2020
Si-R Si-R
Si-R
Si-R Si-R
1 µm
1 µm
22
of the printed silver lines and confirmed the elimination of the rabbit ear effect as seen in Figure
2-9(e), which is a significant improvement of the fluidic assembly process. The average height of
the feature was comparable to that achieved using a positive tone photoresist at the same
withdrawal speed.
Line-edge-roughness (LER), is another important characteristic of printed or evaporated features.
It is defined as the standard deviation of an edge from a perfectly straight line as depicted in Figure
2-9(a) and (b) and becomes important when characterizing nanoscale devices. Many reasons, for
example the difference in development rate of the resist and the size of the polymer molecule the
photoresist is composed of, contribute to the edge roughness in conventional lithography. A
detailed discussion on the causes of LER is available elsewhere.44
Figure 2-9 - (a) and (b) show the profiles obtained from printing silver on samples developed for 10 sec and 30 sec,
respectively before removing the photoresist. The profile of the printed structure after removing the photoresist is
shown in (d) while (e) shows the cross-sectional profile of the same captured using confocal microscopy.
10s development 30s development(a) (b)
(c)(d)
(e)
(f) (g)76 Pre-silanization
106 Post-silanization
2 µm
1 µm
2 µm
23
In the context of printed features, particle size and morphology are additional factors contributing
to the edge roughness. Here, we present some estimates of the edge-roughness and comment on
the effect of particle size and resist thickness on it. The root means square (RMS) edge roughness
is a commonly used metric and was analyzed using AnalyzeStripes plugin for ImageJ.45-46To study
the LER of features printed using fluidic assembly, first we characterized the edge roughness of a
350 nm wide feature written in PMMA using electron beam lithography, which was found to be
Figure 2-10 - (a) and (b) explain the concept of edge roughness. The isolated edge of a 350 nm wide and 100 nm deep
feature in PMMA is shown in (c), while (d) and (e) show features of similar width printed with silver using 100 nm
and 600 nm deep PMMA layer, respectively.
(a) (b)
(c) (d)(e)
350 nm350 nm 300 nm
LERRMS = 3.7 nmLERRMS = 3.6 nm LERRMS = 18.7 nm
24
less than 10 nm. A feature of similar width and the same length printed with silver had an edge
roughness of ~ 24 nm. The increase in the edge roughness is attributed to the curvature of the
spherical particles at the edge. When the thickness of the PMMA layer was changed to about 600
nm, the edge roughness rose sharply to approach 100 nm due to the particles that fall at the edge
during lift-off. The results are shown in Figure 2-9.
2.2.Applications of Resist-patterned Fluidic Assembly
2.2.1. Flexible Transparent Silver Grid Electrode
Metal grid electrodes have emerged as a flexible and cost-effective alternative to indium tin oxide
(ITO) as transparent conductive electrodes (TCEs) for optoelectronic devices such as touch
screens,47 liquid crystal displays,48 solar cells49 and light-emitting diodes (OLEDs).50 Grid TCEs
provide good conductivity, flexibility and high optical transmittance, potentially at a low cost. The
transparency and conductivity of metal grid electrodes can be tailored by tuning the grid’s line
width, thickness and pitch.51 Previously, grid electrodes have been developed using both vapor
deposition and solution-based methods using patterning techniques such as electron beam
lithography (EBL),52 photolithography53 and nanoimprint lithography.54 For some emerging
applications, decreasing the line width of grid electrodes to the limit of photolithography and
possibly below that is required. Smaller line widths provide increased transparency to visible light
without decreasing the conductivity or the density of the grid. For next generation optoelectronic
applications such as displays for virtual and augmented reality, smaller, densely packed LEDs are
needed to eliminate nuisances such as the screen-door effect,55 in which the unlit region between
pixels is visible to the user. To operate such smaller LEDs, one requirement is grid electrodes with
a line width of less than or equal to the size of the LED pixel for maximum optical transmittance.
25
Moreover, decreasing line widths may reduce shadow loss in solar panels thereby increasing their
efficiency.49 Therefore, there is strong impetus to make grid electrodes with submicron and
nanoscale line widths. While vapor deposition methods can deliver high-resolution electrodes with
excellent electrical properties, they are costly because of being subtractive and requiring high
vacuum. Moreover, vapor deposition methods are incompatible with roll-to-roll processing and
are particularly challenging to employ with large, flexible substrates needed for emerging lighting
and solar cell applications. Solution-based methods such as inkjet printing,56 ink evaporation,57
spincasting,58 doctor-blading/coating,59 UV embossing,60 femtosecond laser printing,61 vacuum
filtration,62 electrohydrodynamic jet printing,63 microcontact printing23, and transfer printing64
have thus been used to create metal grid electrodes. While most solution-based methods do not
require high temperatures/pressures., they are either subtractive or non-scalable as some methods
involve non-selectively coating the entire substrate with ink followed by the removal of excess.
Nanoparticle inks are expensive and have a high-embodied energy65, which makes subtractive
printing processes less viable for largescale application as wastage of ink for such processes is
unavoidable. The most significant technical drawback of solution-based methods, however, is the
inability to provide line widths of less than ~ 10 µm on large substrates (inch scale) required for
many emerging applications. Hence, there is need for a cost-effective and scalable solution-based
method to selectively make high-resolution structures (<< 10 µm) over large substrates. To the
best of our knowledge, the highest resolution demonstrated by solution-based processes for grid
structures over large areas (inch scale) is ~ 5 µm using a combination of electrohydrodynamic jet
and transfer printing.66 A proprietary process by Suzhou NanoGrid Ltd. claims to print grids down
to 3 µm by pushing a nanoparticle paste into trenches with a doctor blade. Our study, however,
26
demonstrates grid electrodes with even smaller line widths using a facile directed-assembly based
method to selectively print silver-based gird electrodes on lithographically patterned transparent
substrates. First, the process for printing grid electrodes with 2 µm line width and 100 µm pitch
on a 5 cm x 5 cm flexible polyethylene terephthalate (PET) substrate is described. Towards the
end of the paper, the capability of the proposed process to print a grid electrode with nanoscale
line width (300 nm) is demonstrated on glass.
The process starts by treating the PET substrate with oxygen (O2) plasma to hydroxylate its surface.
The substrate was then patterned with a grid pattern using photolithography with 2 µm line width
and 100 µm pitch. The choice of the photoresist is used to control the depth of the patterned grid.
In this study, two different photoresist thicknesses, ~0.6 µm thick (Microposit® S1805) and 1.3
µm thick (Microposit® S1813), were used to create grid patterns using UV lithography. The sample
preparation is described in more detail in the experimental section. Post-patterning, the contact
angle of water on the photoresist was 68°, which increased to nearly 80° after hard-baking the
patterned PET substrate at 115 °C on the hotplate for 2 min. After preparing the PET sample,
fluidic assembly was used with a 2 wt% aqueous silver ink as shown in Figure 2-11. Each patterned
substrate was lowered vertically into a tank containing silver ink and withdrawn at a fixed speed
using a dipcoater. The withdrawal speed was varied between 0.25 mm/min to 1 mm/min (0.25
mm/min, 0.5 mm/min, 0.75 mm/min, 1 mm/min) for samples prepared with both thin and thick
photoresists (0.6 µm and 1.3 µm) to characterize the effect of: a) the thickness of the patterned
grid, and b) the withdrawal speed, on the electrical properties of the printed transparent electrodes.
Both parameters affect the thickness and density of the printed silver lines as demonstrated later.
The dip-coating-based printing process used herein exploits the preferential wettability of the
27
plasma treated PET substrate achieved by a photoresist template. Depending on the withdrawal
speed, the dip-coating process operates in either the evaporative or the draining regime, the physics
of which are described elsewhere.2, 39 The nature of the fluid/air interface, defined by the surface
tension of the liquid and the surface energy of the solid, play an important role in the assembly
process.67-68 For slower withdrawal speeds as is the case for this study, evaporative regime is
dominant, and the thickness of the deposited/assembled film decreases with increasing withdrawal
speed. In this regime, capillary flow wicks the ink into the thin patterned channels and assembly
takes place due to the evaporation of the solvent at the solid/liquid interface. During assembly,
capillary force causes the ink meniscus to spread and readily wet the exposed, O2 plasma treated
PET surface as it is highly hydrophilic while the spreading of ink on the photoresist is suppressed
due to its low surface energy as depicted in Figure 2-11 (f), (g) and (h).
Figure 2-11 - Sample preparation steps are summarized in (a)-(d). The dip-coating-based printing process is
demonstrated in (e) while the mechanism based on the difference in wettability of the PET and the photoresist is
visually explained in (f), (g) and (h).
O2 plasma
Spincoat
photoresist
Lithography
OH-
OH- OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
Withdrawal
direction
Silver ink
Photoresist
Contact angle ~ 80 .
O2 treated PET
Contact angle ~ 0 .
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
(a)
(e)
(f)
(g)Concave
meniscus
(b)(c)
(d)
(h)
Convex
meniscus
28
This difference in the wetting behavior of the meniscus at the air/ink/substrate interface causes the
liquid in the ink to evaporate faster at the meniscus on the PET surface than the photoresist surface.
As the solvent evaporates at the meniscus, the evaporated volume is replenished by convective
flow from the bulk of the suspension that brings along suspended particles. The evaporation of the
liquid at the meniscus leaves behind the suspended silver particles assembled in patterned grid
lines.69 Because the ink does not spread readily on the photoresist surface and the evaporation rate
is slow, very few particles are deposited on the photoresist surface compared to inside the patterned
grid on the PET surface.
Figure 2-12 - SEM micrographs of the printed grid (1.3 µm thick photoresist, 0.5 mm/min) after hotplate sintering
are shown in (a) and (b). The same sample after flash sintering is shown in (c) and (d) and the dark areas in (c) show
silver crystals. AFM micrograph of a printed silver line (hotplate sintered) is shown in (e) and the extracted cross-
sectional profile is shown in (f). A large area 3D-projection captured with a confocal microscope is shown in (g).
10 µm
100 µm
2 µm
500 nm
400 nm
(a) (b)
(c) (d) 0
500
1000
0 10 20
nm
µm
(e)
(f)
(g)
29
After dip-coating, silver pads were deposited using electron beam evaporation to aid electrical
characterization after which the samples were first sintered on a hotplate at 150 °C for 20 min
before its sheet resistance was measured. A xenon lamp generated flash of light (wavelength 200-
1000 nm) was then used to sinter the samples and the sheet resistance was characterized, again.
This method called flash or photonic sintering is particularly attractive for flexible substrates as it
can heat the surface of a substrate without significantly affecting its bulk. Scanning electron
microscopy (SEM) was used to study the morphology of the printed silver grid after each sintering
process. Atomic force microscopy (AFM) and confocal microscopy were used to ascertain the
profile and the average thicknesses of the printed silver lines, respectively. The printed grid along
with the SEM, AFM, and confocal microscopy micrographs of printed silver are shown in Figure
2-12. Particles coalesce, and large silver crystals form after flash sintering as shown in Figure
2-12(c) and 7(d), which increases the conductivity of the printed electrode.
Electrical characterization was performed by measuring the sheet resistance of each sample using
a four-probe method to eliminate contact resistances, as described in literature.52 Two probes were
landed on each vapor-deposited silver pad and the current was swept from -100 µA to 100 µA
between probes 1 and 2 while measuring the voltage between probes 3 and 4. The sheet resistance
of each sample was measured before sintering, after sintering for 20 min at 150 °C on the hotplate,
and after flash sintering.
Sheet resistance is plotted as a function of withdrawal speed for the grids printed with 0.6 µm thick
and 1.3 µm thick photoresists in Figure 2-13(a) and (b), respectively. In addition to the increase in
conductivity post flash sintering as seen in Figure 2-13(b), two trends emerge. First, for both
photoresists, sheet resistance increases with increasing withdrawal speed indicating a direct
30
relationship between them. Second, for the same withdrawal speed, the thicker photoresist
provides lower sheet resistance. Because the sheet resistance only depends on the grid thickness
provided the line width is fixed, the results suggest that the grid thickness increases with a)
decreasing withdrawal speeds, and b) the thickness of the photoresist used for patterning. The
thickness dependence of the withdrawal speed in the evaporative regime for dip-coating is well-
studied.70-71 In the evaporative regime, increasing the withdrawal speed does not alter the
evaporation rate at the meniscus but shortens the time window for particles to be transported to the
the meniscus, which causes the number of particles getting deposited on the surface to decrease.
The dependence of grid thickness on the photoresist thickness has been explained in 2.1.1.
Generally, the grids printed with the thicker PR and at slower withdrawal speeds were thicker and
had lower sheet resistance values as shown in Figure 2-13(c) and (d) for 0.6 µm and 1.3 µm thick
photoresists, respectively. However, two major anomalies were noticed: 1) the change in grid
thickness is not linearly proportional to the change in sheet resistance, and 2) for comparable
average thicknesses, the sheet resistance values for the two photoresists differed greatly. This
suggests that the variation in electrical performance for different withdrawal speeds cannot solely
be attributed to different grid thicknesses. Seemingly, at faster withdrawal speeds, not only the
grid thickness decreases, but the packing density of the nanoparticles also decreases, which lessens
the effectiveness of the subsequent sintering processes. Thus, the sheet resistances, even for
comparable grid thicknesses, are smaller for the thinner photoresist for slower withdrawal speeds.
For instance, the grid electrode printed with 0.6 µm thick photoresist at 0.25 mm/min is about 300
nm thick and has a sheet resistance of the 367 Ω/sq. In contrast, the grid electrode printed with 1.3
µm thick photoresist at 1 mm/min is almost twice as thick but has a higher sheet resistance.
31
According to Kirchoff’s laws, the sheet resistance (Rs) of a grid electrode can be written as:52
𝑅𝑠 = 𝑁
𝑁 + 1
𝜌𝐿
𝑤𝑡
where N is the number of lines in the grid, ρ is the resistivity of bulk silver, w is the width of each
line, L is the pitch and t is the grid thickness. Using this relationship, theoretical sheet resistances
of the printed grids were calculated and compared with the measured values. The average grid
Figure 2-13 - Sheet resistance versus withdrawal speed for 0.5 µm (a) and 1.3 µm (b) thick photoresists. The inset in
(b) shows the I-V curves (hotplate only) showing linear behavior. The average thickness and sheet resistance for each
withdrawal speed for 0.5 µm (c) and 1.3 µm (d) thick photoresists.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
0 0.25 0.5 0.75 1 1.25
Avg. T
hick
ness [µ
m]
Sheet
Resi
stan
ce [Ω
/sq]
Withdrawal Speed [mm/min]
Thin PR (0.6 µm)
Sheet Resistance
Avg. Ag Thickness
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0
200
400
600
800
1000
1200
0 0.25 0.5 0.75 1 1.25
Avg. T
hick
ness [µ
m]
Sheet
Resi
stan
ce [Ω
/sq]
Withdrawal Speed [mm/min]
Thick PR (1.3 µm)
Sheet Resistance
Avg. Ag Thickness
0
200
400
600
800
1000
1200
0 0.25 0.5 0.75 1 1.25
Sheet
Resi
stan
ce [Ω
/sq]
Withdrawal Speed [mm/min]
Thick PR (1.3 µm)
Hotplate only
Hotplate & Flash
-100
-50
0
50
100
-100 0 100
Voltag
e (m
V)
Current (µA)
0.25 mm/min
0.5 mm/min
1 mm/min
0
5000
10000
15000
20000
0 0.25 0.5 0.75 1 1.25
Sheet
Resi
stan
ce[Ω
/sq]
Withdrawal Speed [mm/min]
Thin PR (0.6 µm)
Hotplate only
(a) (b)
(c) (d)
32
thicknesses extracted from AFM measurements and the bulk resistivity of silver (1.59 x 10-8) were
used. The theoretical sheet resistances were found to be at least an order of magnitude smaller than
the measured values. For example, the calculated sheet resistance of the grid printed using 1.3 µm
thick PR at a withdrawal speed of 0.25 mm/min is about 0.8 Ω/sq, which is about 47 times and 83
times smaller than the measured values of 43 Ω/sq and 67 Ω/sq for flash sintered and hotplate
sintered samples, respectively. To put the sheet resistance data in a more realistic context, the sheet
resistance of a vapor-deposited 400 nm silver grid was compared to the sheet resistance of a printed
sample with a similar average thickness and was found to be 7 times smaller. The higher resistance
of printed samples can be attributed to charge scattering from interparticle-boundaries, surfactant
residue, and assembly defects (discontinuous lines) especially for higher withdrawal speeds. For
lower speeds, the sheet resistance values match closely with those attained by inkjet printing,
electroplating and imprinting while providing better optical transmission due to the smaller line
width.
The optical transmittance of the grid electrode was experimentally determined using an in-house
built tool in the wavelength range of 400 – 900 nm. A tradeoff between optical transmittance and
sheet resistance exists due to obvious physical constraints. Increasing line width increases
conductivity but lowers optical transmittance as wider lines reflect more light. This is defined by
the relationship between optical transmittance and the fill factor (f) that represents the ratio of the
area covered by the grid lines to the total substrate area independent of the grid thickness.61 The f
for the designed grid with a line width of 2 µm and a pitch of 100 µm is fixed at 0.04 for all samples
and hence, measuring the transmittance of a single sample should suffice. For this reason, the
transmittance of the sample with the highest sheet resistance (43 Ω/sq, 1.3 µm thick photoresist,
33
0.25 mm/min) was measured and divided by the transmittance of a blank PET sample. The average
transmittance in the visible light range (450 – 750 nm) is above 92% and peaks at around 96%
around 700 nm as shown in Figure 2-14(a). This transmittance along with a sheet resistance of 43
Ω/sq is well-suited to most optoelectronic applications. However, the trade-off between
withdrawal speed and sheet resistance for different photoresist thicknesses can be explored to tailor
electrodes for various purposes. For instance, solar cells typically require a sheet resistance of less
than 50 Ω/sq to avoid undesired power loss from joule heating,72 for which electrodes printed at
low speeds with thicker photoresists are suitable. However, for applications such as large displays
and field effect devices where a sheet resistance of up to 1000 Ω/sq is acceptable,73 a higher pulling
speed and thinner photoresists may be employed.
A major drawback of ITO is its brittleness, and any suggested alternatives should ideally
demonstrate high flexibility with a minimal loss in conductivity. To explore this aspect, a grid
electrode printed with 1.3 µm thick photoresist at 0.25 mm/min was bent post hotplate and flash
sintering to different radii of curvature between 2 – 20 mm for a single cycle, and the change in its
resistance, R/Ro, was quantified. As shown in Figure 2-14(b), no significant change in the
resistance occurred and R/Ro remained stable around 1.0 even for very small radii. Some emerging
photoelectronic applications such as foldable screens may also require the transparent electrode to
undergo many bending cycles without a significant change in its resistance. To characterize the
suitability of the proposed electrode for such applications, the fatigue characteristics were tested
by subjecting a hotplate and flash sintered grid electrode to 1000 cycles of bending to 2 mm radius
of curvature and quantifying the ratio R/Ro every 250 cycles. Interestingly, negligible change (less
34
than 5%) in sheet resistance was seen over 1000 cycles as shown in Figure 2-14(c). The results
indicate good adhesion between the printed and sintered silver and the PET substrate.
The capability to print nanoscale features is demonstrated by patterning a ~ 0.6 µm thick PMMA
layer spincoated on glass with a grid of 300 nm line width and 5 µm pitch using electron beam
lithography. Glass was chosen as the substrate to facilitate the patterning process but the same can
be achieved on a flexible substrate. Two large pads (50 µm x 25 µm) were patterned on either side
of the grid to facilitate its electrical characterization. The withdrawal speed was set at 0.07 mm/min
as the large pads also needed to be filled alongside the nanoscale features. SEM micrographs of
Figure 2-14 - Optical transmittance of the grid electrode printed with sheet resistance of 43 Ω/sq (1.3 µm thick
photoresist, 0.25 mm/min). Change in resistance of the electrode due to bending to different radii of curvature for a
single cycle (b) and the change in resistance as a function of total bending cycles for 2 mm radius of curvature (c).
0
0.5
1
1.5
2
0 5 10 15 20
R/R
o
Radius of curvature
0
20
40
60
80
100
450 550 650 750 850
Tra
nsm
itta
nce
%
Wavelength [nm]
(a)
(b)
0
0.5
1
1.5
2
0 250 500 750 1000
R/R
o
# of bending cycles
(c)r = 2 mm
35
the printed grid with nanoscale line width are shown in Figure 2-15(a), 5(b) and 5(c). The sheet
resistance was found to be ~ 5 Ω/sq after hotplate and flash sintering, which is only 7 times higher
than the theoretical value of about 0.7 Ω/sq. The improved electrical behavior of nanoscale grid is
because fewer printing defects occur over a significantly smaller area. Moreover, slower
withdrawal speeds provide better packing of particles thereby increasing conductivity.
In summary, a versatile method for printing high-resolution grid electrodes for high resolution
display and touch display applications at room temperature at a low cost. The electrode properties
can be tailored for various applications by tuning the printing process. The presented process is
scalable as many large-sized substrates can be printed simultaneously. The presented grid electrode
Figure 2-15 - SEM micrographs of the nanoscale grid printed on a glass wafer through e-beam lithography are shown
in (a), (b) and (c). The I-V characteristics of the same electrode for difference sintering processes are shown in (d).
-1
-0.5
0
0.5
1
-100 0 100
Vo
ltag
e [m
V]
Current [µA]
PMMA (0.6 µm thick)
Hotplate
Hotplate &
Flash
8.3 Ω/sq
4.4 Ω/sq
200 nm
5 µm
400 nm
(a) (b)
(c)
(c)(d)
36
with a sheet resistance of 43 Ω/sq with an average transmittance of over 92% can potentially
replace ITO or vapor-deposited grid electrodes for most applications. The process can be further
improved by utilizing a higher quality nanoparticle ink, better sintering processes, and the use of
patterning tools specialized for flexible substrates. Overall, this additive technique provides a
promising method for industrial scale printing of metal grid electrodes for current and emerging
applications.
2.2.2. CNT-based Real-time Flexible Lactate Sensor
Continuous monitoring of lactate is of great interest for clinical as well as sports medicine and
fitness applications. Lactate is produced in the body when glucose metabolizes in the absence of
excess oxygen through a process called anaerobic glycolysis, which typically occurs to meet the
body’s energy demand during rigorous physical activity.74-75 At a certain intensity of exertion
called the lactate threshold, significant accumulation of lactate in muscles starts, and symptoms
of fatigue develop. The lactate threshold is a critical measure of the endurance capacity of athletes
and is also used to measure the adequacy of oxygen supply in muscles during a workout, so the
duration or intensity of exercise can be adjusted to avoid fatigue-related health implications. A
wearable lactate sensor can identify the onset of the lactate threshold and therefore, several such
have been proposed. The earliest sensors were impractical as they were invasive,76 bulky and
lacked the required sensitivity. Because lactate is also detectable in sweat,77 tears78 and saliva,79-80
recent sensors use noninvasive pathways81-83 while exploiting the large, functionalizable surface
areas of nanomaterials including CNTs,84-86 graphene,87-89 and various nanoparticles83, 90 to
enhance sensitivity. However, most proposed sensors are complex three electrode amperometric
systems91-92 that are bulky and power inefficient.
37
A two-electrode chemiresistor architecture is a cost-effective alternative to the three-electrode
sensing system.93-94 In chemiresistive sensors, two electrodes are connected by a chemiresistor
material whose resistance changes when exposed to the target analyte. CNTs make good
chemiresistor materials for wearable sensors due to their enormous surface area, flexibility, and
high sensitivity to chemical surroundings.95 However, three key challenges are impeding the
commercialization of CNT-based chemiresistor sensors. First is the unstable zero baseline due to
the resistance drift in CNTs. It is suggested that the drift is due to the preferential adsorption of
vapor/gases or other chemical species onto sidewall defects of CNTs.96-97 Secondly,
functionalization is needed for CNTs to specifically detect a desired analyte but single-walled
CNTs (SWCNTs) that are typically used for chemiresistive biosensors lack enough chemically
active sites for dense covalent functionalization, while non-covalent functionalization methods are
unstable and lack robustness.98 The last challenge is the high sensor-to-sensor variation in baseline
resistance, which affects the inter and intra-batch reproducibility.99-100 For minimal resistance
variation on a large substrate, a CNT film of uniform density is needed.101-102 Direct vapor-based
growth of CNT films requires high temperatures and is incompatible with most plastic substrates
and therefore, solution-based methods such as spincoating103 and dropcasting94 are preferred to
make CNT-based sensors. Dropcasting has poor control and lacks repeatability. Spincoating is
unsuitable for large substrates because the density of the CNT network changes with distance from
the center of the substrate. Moreover, spincoating requires a high-concentration CNT suspension
of which a considerable amount is wasted, which is disadvantageous as high-quality nanomaterials
are expensive.
38
We directly address each of the aforementioned challenges by: a) replacing SWCNTs with multi-
walled CNTs (MWCNTs) in a chemiresistor architecture to enable robust covalent
functionalization and chemical modification of the surface without significantly increasing the
resistance, b) electrochemically modifying the CNTs with carboxylic acid functionalized
polypyrrole to simultaneously mitigate the resistance drift and significantly increase chemically
active sites for enzymatic covalent functionalization and c) demonstrating a facile directed
assembly-based printing method to print multi-walled carbon nanotubes (MWCNTs) selectively
on flexible substrates into lithographically defined channels to attain minimal intra and inter-batch
resistance variation. Conductivity of MWCNTs is significantly less affected by the modification
of their outermost walls, which is advantageous for chemiresistor sensors.98 The sensors are
covalently functionalized for lactate detection and are shown to detect L-lactate between 1 – 20
mM, which is the range is of interest for noninvasive detection of lactate in humans for both clinical
and fitness applications.104
Figure 2-16 - (a) The CNT directed assembly mechanism and (b) The sensor fabrication and functionalization
procedure shown schematically.
Deposit Ti/Au
and PR window
PET Substrate
Spincoat
S1813 PR
Pattern with
UV
Print CNTs
Strip PR
Modify with
PPyCOOH
Immobilize Enzyme
(LOx)
Contact
angle ~ 0˚ Contact
angle ~ 80˚
(a) (b)
Deposit Ti/Au
and PR window
PET Substrate
Spincoat
S1813 PR
Pattern with
UV
Print CNTs
Strip PR
Modify with
PPyCOOH
Immobilize Enzyme
(LOx)
Contact
angle ~ 0˚ Contact
angle ~ 80˚
(a) (b)
39
To start, PET film cut into a four-inch wafer was treated with oxygen plasma to make its surface
hydrophilic (contact angle ~ 0˚). The PET substrate was then patterned with 50 µm wide, 800 µm
long and ~1.3 µm deep trenches using UV photolithography and hard-baked post patterning.
MWCNTs were then selectively printed in the patterned trenches by immersing the PET substrate
vertically in the MWCNT suspension and withdrawing at 0.05 mm/min using a dip coater placed
inside a desiccated chamber with the relative humidity below 20%. The printing mechanism is
shown schematically in Figure 2-16(a). The selectivity of the printing method is due to the
difference in surface energies of the plasma treated PET surface (contact angle ~ 0˚) and the hard-
baked photoresist (contact angle ~ 80˚). The CNT suspension readily wets the hydrophilic PET
surface and the meniscus spreads upwards, thereby increasing the rate of evaporation at the
meniscus. As the liquid evaporates, the CNTs suspended in it are assembled on the PET surface.
The evaporation rate is suppressed at the meniscus on the low surface energy photoresist, resulting
in a far sparser deposition of CNTs, which are removed during the subsequent photoresist removal
process. After assembly, the photoresist was removed, and the thickness of the printed CNT
channels was measured to be ~100 nm using atomic force microscopy (AFM). The thickness and
hence the resistance of the channel can be varied by changing the dip coating speed, humidity,
concentration of the CNT suspension, and the thickness of the photoresist used.28, 105 Post
assembly, Ti/Au (5/150 nm) electrodes with a 30 µm gap were deposited using electron beam
evaporation. After metal deposition, a third layer, a 20 µm wide and 1000 µm long window, was
patterned on top of the CNT channel using AZ nLoF 2020 photoresist to only expose CNTs and
protect the metal electrodes from the liquid test medium during measurements.
40
Next, sensor functionalization starts by electrochemically modifying the CNTs with poly(1-(2-
carboxyethyl) pyrrole) (PPyCOOH). The modification of CNTs by PPyCOOH stabilizes sensor’s
baseline by mitigating resistance drift and facilitates dense covalent functionalization by providing
additional carboxylic acid functional groups for the enzyme to attach through amine coupling,
leading to high sensitivity. Several methods for electrochemical deposition of PPy films are
described in literature.106-107 We use DC chronoamperometry to electropolymerize PPy on CNTs,
which produces islands of PPy and not a continuous film.108 Each sensor was immersed in 100
mM PPyCOOH in KoH and a 1.2 V potential was applied across the two electrodes for 120 sec
after which the sensors were gently rinsed with deionized water for 5 sec to remove the excess
solution. Raman spectra of the CNTs before and after PPy modification is shown in Figure 2-17(a).
The sensor fabrication and functionalization processes are shown schematically in Figure 2-16(b).
Interestingly, the D/G ratio decreases from 1.44 to 1.28 indicating a decrease in defects in the
CNTs. Additional peaks also emerge, namely the peaks at 936 cm-1, 983 cm-1, 1047 cm-1, 1257
cm-1and 1411 cm-1, that can be matched to the Raman spectrum of electrochemically formed
polypyrrole films.109-110 Successful modification is also confirmed by the stabilization of the sensor
baseline at 0.1 V after the PPy modification when immersed in 1X-PBS as shown in Figure
2-17(b).
41
We hypothesize that PPyCOOH electrochemically polymerizes preferentially on the defect sites
of CNTs, which reduces the available defect sites for the adsorption of vapor to adsorb and
decreases the D/G ratio. The defects on the CNTs are effectively coated by a PPy layer with
carboxylic functional groups for enzyme immobilization.
After PPy modification, the sensors were incubated in 0.5 M EDC and 0.4 M NHS in MES (pH
4.7) in an airtight container for two hours to promote amine bond formation.111 The sensors were
then rinsed with deionized water and incubated in the enzyme lactate oxidase (114 units in 250 uL
1X-PBS, stored at -20 °C) for 24 hours at room temperature and then rinsed with deionized water.
Ferrocenecarboxylic acid (FCX) was used as a redox mediator and deposited using linear sweep
voltammetry was performed by dropcasting 10 mM FCX in 1X-PBS and sweeping the voltage
between -0.6 V to 1.2 V at 20 mV/s for 40 cycles. The sensors were then dried for 30 minutes
before being rinsed with 1X-PBS priming them for testing. The sensors were tested by measuring
the resistance continuously at 0.1 V applied across the sensor channel. At the start of each
Figure 2-17 -(a) Raman spectra of the printed CNTs before and after PPy modification. The peaks at 936 cm-1,
983 cm-1, 1047 cm-1, 1257 cm-1and 1411 cm-1 for modified CNTs correspond to those of electropolymerized
polypyrrole films. (b) shows the effect of PPy modification on the resistance drift with sensor immersed in 1X-
phosphate buffered saline (PBS) at 0.1 V.
480
490
500
510
520
530
540
550
0 600 1200 1800 2400 3000 3600R
esis
tance
(Ω
)
Time (s)
After PpyModification
Before PpyModification
(b)
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
600 800 1000 1200 1400 1600 1800 2000
Inte
nsi
ty (
counts
)
Raman Shift (cm-1)
(a)
PPy modified CNTs
Unmodified CNTs
480
490
500
510
520
530
540
550
0 600 1200 1800 2400 3000 3600
Res
ista
nce
(Ω
)
Time (s)
After PpyModification
Before PpyModification
(b)
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
600 800 1000 1200 1400 1600 1800 2000
Inte
nsi
ty (
counts
)
Raman Shift (cm-1)
(a)
PPy modified CNTs
Unmodified CNTs
42
experiment, each sensor was rinsed gently with deionized water for 5 sec and dried. A 25 µL
droplet of 1X-PBS was then placed on the sensor and a known volume of lactate solution in 1X-
PBS was added to change its concentration to a pre-calculated value. To test the use stability of
the sensors, and to measure the degradation in their response over time, some sensors were tested
over 10 days. The reproducibility of sensing response was also characterized.
The resistance of a sensor is inversely proportional to the thickness of the CNT channel connecting
the two metal electrodes but the total functionalizable surface area may be independent of the
thickness. Both thickness and the surface area contribute to the sensor’s response, ∆𝑅 (%)and can
be expressed as:
∆𝑅 (%) = (𝑅𝑜 − 𝑅)
𝑅𝑜 × 100
where 𝑅𝑜 is the baseline resistance of the sensor and 𝑅 is the resistance after exposure to lactate.
For chemiresistive sensing, 𝑅𝑜 − 𝑅 depends on the total functionalized surface area of the
chemiresistor material whereas 𝑅𝑜 is a function of the material’s thickness given device geometry
is fixed. If the thickness varies independently of the functionalized surface area, the response of
each sensor will be different. Therefore, the resistances of all sensors must lie within a small range
for reproducibility. We measured the resistances of thirty CNT channels before functionalization,
of which, 25 channels (83.3%) had resistances between 1 kΩ and 5 kΩ. This small range is highly
desirable from a device manufacturing standpoint. The mean resistance for 25 channels is 2.37 kΩ
with a standard deviation of 0.93 kΩ.
43
The real-time response of the sensor in 1X-PBS is shown in Figure 2-18(a) along with the
normalized signal ∆𝑅 (%) in Figure 2-18(b).
The resistance of the CNT-channel is seen to decrease with increasing concentration of lactate. We
hypothesize that the oxidation of LOx by lactate in the presence of the ferrocene electron shuttle
provides the CNTs with additional charge carriers thereby decreasing their resistance. The sensor’s
real-time response exhibits linearity between 1.5 to 20 mM as shown in Figure 3(b) with y =
0.0019x + 0.0034 and R2 = 0.9952. For each measurement, the resistance changes instantaneously
but is allowed to stabilize for 30 - 60 sec before taking a reading. The stabilization period is
required to allow the lactate to diffuse evenly throughout the droplet and to let the unstable
convective perturbation fade out. A microfluidic channel could alleviate this issue by ensuring
quasi-static changes in lactate concentration. The specificity test is shown in Appendix A in which
the sensor does not respond to either glucose or urea, two metabolites that exist commonly in
biofluids alongside lactate. This confirms successful functionalization and high specificity of the
Figure 2-18 - (a) Response of the sensor for lactate concentration of 1.5 mM – 20 mM and (b) normalized response
plotted against the corresponding lactate concentration.
44
sensor. The results in Figure 2-18 are for the sensor functionalized and tested on the day of
functionalization.
The stability for the designed sensor was characterized by testing functionalized sensors on days
1, 2, 7 and 10 with the sensors placed at room temperature exposed to air between each test and
rinsed with 1X-PBS before and after each experiment. Change in the sensor’s response to lactate,
the stability of its baseline, and the response time were quantified. Figure 2-19(a) shows the
normalized responses plotted against respective concentrations for each day. As seen in Figure
2-19(a), the response is stronger and highly linear on Day 1, but from Day 2, the response weakens
slightly with increasing nonlinearity between 10 mM and 20 mM lactate. The difference between
the response on Day 7 and Day 10 was found to be significantly smaller than between Day 1 and
Day 2. The large reduction in response from Day 1 to Day 2 can be attributed to non-covalently
attached enzyme112 on the surface of the CNTs, which is removed during the subsequent rinsing.
Additional rinsing steps can be added post-functionalization to equilibrate the sensor and to ensure
consistency in its response. After Day 2, the response depreciates due to the gradual loss of the
mediator in the system,113 as well as decreasing enzymatic activity.114 If required the redox
mediator can be trapped in a polymer matrix but it is not needed for disposable application such
as a for a single workout. The stability of the sensor’s baseline also changed over ten days. On Day
1, the sensor’s baseline was stable at the start of measurement. On Day 7, the baseline drifted but
stabilized within 100 sec, which is less than the 300 sec it took for the baseline to stabilize on Day
10. On both Day 7 and Day 10, the baseline resistance decreased by about 1.3% before stabilizing.
This could indicate the formation of additional defect sites on the CNTs during testing. Overall,
45
the results indicate suitability for a disposable microsensor with no stringent storage requirements
that can be used for up to a week.
Lack of sensor-to-sensor reproducibility is an important issue for chemiresistor sensors and can be
attributed to variation in baseline resistances. Especially for CNT-based sensors, the uniformity
and morphology of the film is critical to achieving repeatable electrical and sensing properties.115
To demonstrate the advantage of minimal resistance variation shown previously, intra-batch
sensing response characterization was performed. Three sensors from different regions on a four-
inch wafer were functionalized and tested on the same day. The normalized response of each sensor
is plotted against the concentration in Figure 2-19(b). The response of all sensors between 2 mM
to 20 mM has negligible variation, which is a result of the minimum baseline resistance variation
achieved through directed assembly and is conducive to commercial application.
Figure 2-19 - The normalized response is plotted against concentration for each day (a). Intra-batch characterization
of the response of three sensors showing minimal response variation (b).
0
0.5
1
1.5
2
2.5
3
0 5 10 15 20
∆R
(%
)
Concentration (mM)
Sensor 1
Sensor 2
Sensor 3
(b)
0
0.5
1
1.5
2
2.5
3
0 5 10 15 20
∆R
(%
)
Concentration (mM)
Day 1
Day 2
Day 7
Day 10
(a)
0
0.5
1
1.5
2
2.5
3
0 5 10 15 20
∆R
(%
)
Concentration (mM)
Sensor 1
Sensor 2
Sensor 3
(b)
0
0.5
1
1.5
2
2.5
3
0 5 10 15 20
∆R
(%
)
Concentration (mM)
Day 1
Day 2
Day 7
Day 10
(a)
46
In summary, we present a simple and scalable method for the manufacture and functionalization
of low-cost CNT enabled sensors on flexible substrates for continuous noninvasive lactate
detection. Significantly, it is shown that the resistance drift in CNT-based chemiresistive sensors
can be mitigated by modifying the CNTs with PPyCOOH, leading to a stable baseline and an
enhanced response due to denser covalent functionalization. Although a PPy/ CNT composite has
been used for amperometric sensing,116 this is the first report of PPyCOOH modification of CNTs
to reduce resistance drift and enhance sensitivity to the best of our knowledge although the exact
structural changes through which this is achieved warrants more investigation. The sensors
demonstrate fast real-time monitoring of lactate in an ionic buffer and are active for at least ten
days when stored in air at room temperature. The sensors also show minimal intra-batch resistance
and response variation over a four-inch substrate, which demonstrates the scalability and
reproducibility of the technique. The fluidic assembly-based printing method as well as PPy
modification presented in this paper is extendable to the manufacture of sensors for detecting other
metabolites such as glucose, urea and creatinine, as well as for inorganic chemicals and gases.
47
3. Fast-Fluidic Assembly Method
3.1.Motivation for a Rapid Printing Process
In Chapter 2, we introduced an evaporation-controlled fluidic assembly process performed by
withdrawing a resist-patterned substrate out of a nanoparticle suspension using a dip-coating
platform. Capillary forces and evaporation at the solid-liquid interface cause particle assembly on
the hydrophilic areas on a substrate. For particles to assemble in continuous arrays, the rate of
evaporation must match the rate of withdrawal of the substrate out of the suspension. As described
in Chapter 2, the fluidic assembly method has been used to print carbon nanotubes (CNTs) and
silver nanoparticles on photoresist patterned hydrophilic surfaces.28, 117
A major disadvantage of the fluidic assembly method is that because it depends on evaporation, it
has a low throughput, which is undesirable for commercialization. However, depending on the
withdrawal speed, dip-coating operates in either the evaporative (capillary) or the draining regime.
Until now, fluidic assembly has only been discussed in the evaporative regime, which is active
when the withdrawal speed is slow, typically less than 1 mm/min and the thickness of the deposited
film decreases with increasing withdrawal speed. As the rate of withdrawal of the substrate exceeds
the rate of evaporation and particle assembly, the dip-coating process enters the draining regime
where instead of particles, a liquid film is entrained on the substrate whose thickness increases
with increasing withdrawal speed. This mechanism is shown schematically in Figure 3-1. Higher
withdrawal speeds, if used with a photoresist patterned substrate, will reduce the selectivity of the
printing process, which is undesirable. Moreover, photoresist templated substrates cannot be used
with solvents, which are typically needed for high-concentration nanoparticle inks.
48
To avoid these drawbacks while increasing the speed of fluidic assembly, we introduce a fast-
fluidic assembly-based printing technique on an engineered surface, which operates in the draining
regime of dip-coating. In this technique, the substrate’s surface is engineered such that the features
desired to be printed are made highly hydrophilic, whereas the rest of the surface is hydrophobic
due to the presence of a self-assembled monolayer (SAM) of a silane compound. This SAM serves
a function, to some extent, similar to that of the photoresist as described above, without the
associated disadvantages. When such a chemically heterogenous surface is withdrawn from the
liquid, the liquid slips off the low-energy hydrophobic surfaces while the hydrophilic areas can
retain ink with remarkable precision. The ink retained in the features then dries, leaving the
suspended particles assembled as features.
3.2.Literature Review of Selective Deposition on Chemically Heterogenous Surfaces
The first reports of using self-assembled monolayers to attain chemically heterogenous surfaces
were published by Whitesides et al in 1994.118 Later that year, they utilized the selective wettability
of such chemically patterned surfaces to selectively and spontaneously organize thin layers of
organic liquids through a dewetting process.119 As self-organization due to spontaneous dewetting
Figure 3-1 – The dip-coating process shown schematically in the evaporative (a) and draining (b) regimes.5 The
shifting of stagnation point that shifts to towards the reservoir surface resulting in a thicker entrained film is depicted
in (c).2
(a) (b) (c)
Increasing withdrawal speed
49
was demonstrated, interest in the underlying mechanisms grew, and in 1999, a thorough treatment
providing insights into the dynamics of the dewetting process was published by Petrov et al.120
Sharma et al. later expanded that dewetting on a chemically heterogenous surface was due to local
spatial wettability gradients, and different from spinodal dewetting that takes places on chemically
homogenous hydrophobic surfaces.121-122
Soon followed the attempts to exploit spontaneous dewetting on lithographically generated
chemically heterogenous surfaces for device applications.123 However, spontaneous dewetting is
an unstable process and controlling the dewetting front is difficult, which leads to nonuniformities
in assembly over the large areas. Considering this, work exploiting selective wettability of surfaces
for deposition of materials by more controllable methods such as spincoating and doctor blading
was demonstrated.124-125 There are two major disadvantages of spincoating: 1) the film thickness
varies from center to the edge of the substrate and 2) it is not a material conservative technique as
significant amount of material slips off the substrate. For printing, it is especially disadvantageous
as nanoparticle inks that are used for printing electronics are expensive. The doctor blading
process has limitations such as inhomogeneity over large substrates and difficulty in attaining
features heights of less than 100 nm. The doctor blading process, similar to spincoating and
spraycoating, is not a material conservative process and some ink is wasted. Dip-coating, on the
other hand, is a material conservative process, which also allows for relatively precise thickness
control. The thickness of the liquid film entrained on a chemically homogenous substrate is defined
by the Landau-Levich model.126 In 2000, the first in-depth study of dip-coating on a chemically
heterogenous substrate with a patterned SAM was published by Darhuber et al. who reported
selective dip-coating of glycerin on a chemically patterned gold substrate.127 The relationship
50
between withdrawal speed and thickness of the film entrained was numerically and experimentally
estimated. Geometrical constraints applicable to positioning of the hydrophilic and hydrophobic
patches with respect to the withdrawal direction were also investigated. It was found that the
orientation of the feature, or the width of the feature that is perpendicular to the withdrawal
direction and parallel to the substrate-liquid contact line is important to the film thickness.120 Soon
after, more work investigating the morphology of dip-coated liquid films on chemically
heterogenous surfaces was published by the same group.128
Although dip-coating in the draining regime is a promising method to control the dewetting front,
which has also been mathematically modeled, except for one study on CNTs,129 its use to assemble
nanoparticle structures for device applications has remained largely unexplored. One reason is that
dip-coating in the draining regime requires high concentration particle suspensions, which until
recently, were not commercially available. Moreover, controlling the hydrophobicity at the
nanoscale is very challenging. Due to these reasons, nanoparticle structures using dip-coating in
the draining regime on chemically heterogenous structures have not been demonstrated.
The work presented herein experimentally determines the effect of withdrawal speed, particle
concentration, and solvent properties on the thickness, morphology and electrical properties of the
printed nanoparticle structures at the micro and nanoscale. We term this process the Fast-Fluidic
Assembly and Transfer (FFAsT) process. We also demonstrate the use of the FFAsT process
towards device application through both direct and transfer printing.
51
3.3.Mechanism of the Fast-Fluidic Assembly Process
Describing the proposed process through a robust mathematical model is very challenging due to
the wide variety of inks and substrates involved. However, a simpler explanation based on
tribological properties of the components in the system is presented.
A chemically heterogenous surface with hydroxylated (hydrophilic) and silanized (hydrophobic)
‘patches’ is immersed in a liquid and withdrawn at a fixed speed. As the substrate moves up, the
ink is retained by and spreads in the hydrophilic areas, while the hydrophobic regions are de-
wetted. The underlying concept is well-understood by examining the intermolecular and
intramolecular properties of the substrate and the ink. Capillary action is promoted on the
hydroxylated surface due to the dipole-dipole interaction between the -OH groups on the substrate
(generated due to plasma treatment) and the polar components of the ink’s solvent. The type and
magnitude of this interaction dictates the contact angle between the fluid and the substrate. The
role of the contact angle between the ink and the solid substrate is critical to understanding this
process. In addition, considering the heterogenous chemical nature of the surface, it is reasonable
to use the Cassie-Baxter model to calculate the apparent contact angle, 휃𝑎𝑝𝑝, using contact angle
values on hydroxylated and silanized surfaces:
𝑐𝑜𝑠 휃𝑎𝑝𝑝 = ϕ1𝑐𝑜𝑠 휃1 + ϕ2𝑐𝑜𝑠 휃2
where ϕ1 and ϕ2 are area fractions for the hydrophobic and hydrophilic regions, while 휃1 and 휃2
are the respective contact angles.
As the substrate is withdrawn from the ink bath, the hydroxylated features retain ink while the ink
dewets from the silanized regions. For a successful print, three conditions should be met: 1) the
ink must completely and cleanly dewet the hydrophobic regions, 2) the hydrophilic regions must
52
retain ink, and 3) the ink must spread evenly as a thin film in the hydrophilic regions. All three
factors are discussed below.
The dewetting from the hydrophobic regions can be explained by the work of adhesion, Wadh. The
intermolecular forces between the silanized surface and the ink molecules are significantly lower
than the gravitational force acting on the film. The work of adhesion can be calculated using the
Young-Dupre equation:
𝑊𝑎𝑑ℎ = 𝛾𝑆𝐿(1 − 𝛾𝐿𝑉𝐶𝑜𝑠휃)
where 𝛾𝑆𝐿 and 𝛾𝐿𝑉 are interfacial energies for solid-liquid and liquid-vapor phases, respectively,
while 휃 is the solid-liquid contact angle.
A similar parameter to the work of adhesion is the work of wetting. It is the work done to eliminate
a unit area of a solid-liquid interface to make a solid-gas interface. This parameter typically
describes draining of fluids from capillaries but is also suitable for dip-coating based assembly
process described here. The work of wetting, 𝑊𝑤, can be written as:
𝑊𝑤 = 𝛾𝐿(𝐶𝑜𝑠휃)
where 𝛾𝐿 is the liquid’s surface tension and 휃 is the solid-liquid contact angle.
As an example, we calculate the work of wetting of DEG on hydrophilic and hydrophobic surfaces.
For DEG, the surface tension is ~ 45 mN/m and its contact angles on hydroxylated and silanized
surfaces are ~ 0° and 78°, respectively. Simple calculations reveal the work of wetting on the
hydroxylated surface to be significantly higher than the silanized surface indicating the relative
ease of draining ink from the latter. An important aspect is that the contact angle values considered
here were less than 90°, signifying that some affinity would exist between the liquid and the
underlying substrate. However, for water the contact angle on a hydrophobic surface exceeds 90°,
53
in which case the work of wetting would become negative thereby indicating that spontaneous
dewetting without any additional force would occur.
Spontaneous spreading of the ink in the hydroxylated regions is important for film uniformity. The
ability of an ink to provide uniform features can be predicted by calculating the spreading
coefficient, S, of the liquid on the hydrophilic and hydrophobic patches. When the spreading
coefficient is positive, spontaneous wetting of a smooth solid surface by the liquid is expected. A
negative S prevents spreading. The spreading coefficient can be calculated by using the Cooper-
Nuttall criterion:
𝑆𝐿/𝑆 ≡ 𝑊𝑆 = 𝛾𝑆 − 𝛾𝐿 − 𝛾𝑆𝐿
where 𝛾𝑆 is the solid’s surface energy, 𝛾𝐿is the surface tension of the liquid, and 𝛾𝑆𝐿is the interfacial
surface energy between the solid and the liquid.
3.4.Development of the Fast-Fluidic Assembly Process
Lithography is used to selectively functionalize surfaces as described in literature. Each substrate
used for printing was piranha cleaned (H2SO4: H2O2) and treated with oxygen plasma for 2 min
(38 mTorr, 115 W, 15 sccms). Negative tone AZ N2020 was spincoated on the wafer and patterned
using UV-lithography to create a photoresist mask. This mask protects the desired features from
the subsequent surface treatment. Post lithography, the wafer was again treated with O2 plasma for
1 min (115 W, 15 sccms). The substrate was then vapor treated with an organosilane
(1H,1H,2H,2H-Perfluorododecyltrichlorosilane) compound for 3 min under mild vacuum, which
forms a SAM causing the surface energy to decrease significantly. Post SAM deposition, the
substrates were immersed in a solvent or photoresist stripper (acetone, NMP etc) to remove the
photoresist mask. Figure 3-2 schematically shows the substrate preparation process.
54
For nanoscale patterns, a negative electron beam resist, ma-N 2410, was used to pattern a 4-inch
wafer with nanoscale patterns. The resist was spincoated on an Si wafer at 3000 RPM for 30 sec
and baked on the hotplate at 90 °C for 3 min. It was then exposed to create 350 nm wide and 50
µm long lines with a 5 µm spacing before being developed in an alkaline developer.
Figure 3-2 - Substrate preparation for the FFAsT process.
(c)
Plasma treat Spincoat
photoresist
Lithography
Withdrawal
direction
Silver ink
Silanized surface
Contact angle ~ 105 .
O2 treated surface
Contact angle ~ 0 .
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-
OH-(a)
(g) (h) (i)
(b)
Remove
photoresist
OH (hydrophilic)
Si-R (hydrophobic)
Hydrophilic
Retains ink droplets
Hydrophobic
Meniscus slips
Silanization
(d)
OH-
OH-
OH-
OH-
OH-
OH-
OH-OH-
OH-
(e)
Si-R
Si-R
Si-R
Si-R
Si-R
Si-R
Si-R
Si-R
Si-R
(f)
Si-R
Si-R
Si-R
Si-R
Si-R
Si-R
Si-R
Si-R
Si-R
55
The results presented here are for a commercially available silver nanoparticle ink. The particle
suspension consisted of silver in diethylene glycol (DEG) with a concentration of 20% silver by
weight. This ink was utilized to print microscale features. Each substrate was lowered vertically
into a bath containing the silver ink and withdrawn at a fixed speed. This withdrawal speed was
varied between 10 mm/min to 100 mm/min to gauge the effect on the printed film’s morphology
and thickness. Different concentrations of the ink were also tried to capture how ink concentration
affects the thickness of the printed features.
After the first layer is printed, different methods such as chemical, thermal or plasma treatment
may be used to remove the existing SAM and prepare the surface for the next layer. For example,
a very short exposure to Ar or O2 plasma or heating the substrate to 300 ˚C can remove the SAM.
Figure 3-3 - Examples of features printed with the FFAsT method.
56
3.5.Characterization of the Fast-Fluidic Assembly Process
3.5.1. Controlling Film Thickness
The ability to control film thickness is critical for a deposition or printing process. Moreover, the
ability to predict film thickness using a mathematical model for different process conditions is also
very useful. Liquid film thickness entrained in the draining regime of the dip-coating process for
low viscosity and reasonably high surface tension solvents such as those used for nanoparticle inks
is described by the Landau-Levich model:
ℎ0 = 𝑐(휂𝑈0)
23
(𝛾𝐿𝑉)16(𝜌𝑔)
12⁄
where c ~ 0.94 for Newtonian, non-viscous liquids, 휂 is the viscosity of the solvent, 𝑈0 is the
withdrawal speed, 𝛾𝐿𝑉 is the surface tension of the liquid, 𝜌 is solvent density and 𝑔 is the
gravitational constant. According to the Landau-Levich model, a direct but nonlinear relationship
exists between the entrained film thickness and withdrawal speed. Viscosity is also directly related
to the film thickness whereas surface tension has an inverse but very weak relationship. Landau-
Levich model applies best when the ratio of the viscous and the capillary forces in the liquid is
small (< 10-2). This ratio is defined by the capillary number as:
𝐶𝑎 = 휂𝑈/𝛾
Landau-Levich model is suited to chemically homogenous surfaces. For chemically heterogenous
surfaces consisting of patches of hydrophilic and hydrophobic areas, Darhuber et al. altered the
Landau-Levich model and showed that the film thickness on such surfaces is described by:
ℎ0 = 𝐾w (휂𝑈0
𝛾𝐿𝑉)
13
57
where K ~ 0.356 for Newtonian, non-viscous liquids, w is the length of the feature parallel to the
meniscus, 휂 is the viscosity of the solvent, 𝑈0 is the withdrawal speed, and 𝛾𝐿𝑉 is the surface
tension of the liquid. To ascertain if this model describes the proposed printing method, silver ink
was printed into 100 µm x 100 µm features patterned using a SAM mask on a Si wafer. Because
the film thickness depends strongly on the solvent properties, 20wt% silver ink was prepared by
diluting the original 40 wt% silver ink in three different 1:1 solvent combination namely DEG and
deionized water (DIW), pure DEG and IPA.
The prepared substrates were printed on using withdrawal speeds between 10 mm/min to 100
mm/min, and dried on the hotplate at 30 °C. After drying, the thickness of the printed features was
measured using confocal microscopy. To compare with Darhuber’s modified model, the thickness
of the liquid film entrained by the proposed printing method was estimated by using the thickness
of the printed film along to calculate the volume of the nanoparticle structure, and then using the
concentration of the ink to calculate the volume and the thickness of the liquid film. Porosity of
the printed film is critical to this calculation and was accounted for by dividing the density of silver
by three. This factor was calculated by measuring the thickness of a nanoparticle film before and
after sintering at 800 °C for 1 min, which reduced the feature height to about a third of its original
height. The results are summarized in Figure 3-4.
58
Solid film thickness increases with withdrawal speed for the DEG and DIW-diluted inks, while
remaining virtually the same for the ink diluted with IPA. First, the large increase in the thickness
of the DIW-based ink with increasing withdrawal speed is analyzed. The contact angle of DIW-
diluted ink on the hydrophobic region is only 32°, due to which selective dip-coating does not
occur on the substrate. A thick film of ink (obeying the Landau-Levich model) is pulled out starting
around 20 mm/min, which later dewets as liquid retracts from the hydrophobic regions and
accumulates in the hydrophilic areas. Due to this, a much thicker liquid film occupies the
hydrophilic region. The morphology of the feature printed with DIW-diluted ink shown in Figure
3-5(b). The cross-section reveals a convex ‘dome’, much thicker than that obtained by IPA and
DEG diluted inks. The reason for this is explained in Figure 3-5(d). As the substrate is withdrawn
out of the ink, it is initially completely wetted and is placed on the hotplate at 30 °C. The
temperature is not high enough to cause rapid evaporation of the solvent, and instead allows for
the water molecules to evaporate first (due to its lower boiling point), and leaving the silver
Figure 3-4 - Film thickness attained for different withdrawal speed for DEG, IPA and DIW (a) and a comparison
with Darhuber’s model (b).
0
100
200
300
400
500
600
700
800
900
0 50 100Avera
ge T
hic
kness
(nm
)
Withdrawal Speed (mm/min)
Withdrawal Speed vs Film Thickness
DEG
IPA
DIW
0
0.5
1
1.5
2
2.5
3
3.5
0 50 100 150
Liq
uid
Film
Thic
kness
(µm
)
Withdrawal Speed (mm/min)
Model vs Experimental
Darhuber's
Model
Experimental
(a) (b)
59
suspended in DEG. As the water evaporates, two phenomena take place: the ink recovers its pre-
dilution properties (the contact angle increases), and the concentration of silver in the ink increases.
The increased contact angle causes the contact line to recede and move to the closest hydrophilic
region. The dewetting dynamics of a liquid film on chemically patterned have been previously
described. Once the ink is confined to the hydrophilic regions, it starts to dry, and the silver film
is formed.
Figure 3-5 - Morphology and cross-sectional profiles of features printed with IPA, DIW and DEG diluted inks. The
mechanism showing transition between controlled dewtting dip-coating and uncontrolled dewetting is also shown.
IPA 40 mm/min DIW 40 mm/min DEG 40 mm/min
Withdrawal
Direction
Controlled Dewetting
Dipcoating
Uncontrolled Dewetting
Film rupture
Withdrawal Speed or
Viscosity or
Surface Tension
60
The effect of increasing solid weight percent or particle concentration of the ink on the final film
thickness was also characterized. For this purpose, a Si wafer was patterned with 100 µm x 100
µm pads and diced into 20 mm x 10 mm chips. A commercial 40 wt% DEG-based silver
nanoparticle ink obtained from Novacentrix was diluted to concentrations of 5 wt%, 10 wt%, 15
wt% and 20 wt% using DEG and used to print the patterned features followed by drying at 30 °C.
The height of the dried solid film was measured using confocal microscopy (Zeiss, Smartzoom 5),
and confirmed using AFM. The results are shown in Figure 3-6. It is seen that increasing particle
assembly leads to a thicker film, as there are more particles in the entrained liquid film that fill the
hydrophilic region when the solvent dries. Ink concentration, in addition to withdrawal speed,
therefore can be used to control the thickness of the printed features. A caveat is the increasing
porosity of the film at lower concentration, which would need to be catered to by adjusting the
withdrawal speed or the subsequent sintering process.
As mentioned in Darhuber’s model, the half width of the feature that makes contact with the
meniscus directly affects the feature height. This is demonstrated in Figure 3-7. The average height
Figure 3-6 - Effect of particle concentration on film thickness at the withdrawal speed of 40 mm/min for DEG-based
ink.
0
20
40
60
80
100
5 10 15 20
Ave
rage
Thic
kness
(nm
)
Ink Concentration in DEG (%)
61
of a 200 µm x 200 µm feature is approximately ~200 nm higher than that of a 100 µm x 100 µm
square box. This implies that a 100% increase in surface area and edge-width caused the height of
the feature to increase by ~35%. Most modern electronic features, however, do not have this large
a range in feature sizes. If required, surface polishing can be used to achieve planar surfaces.
3.5.2. Characterizing Film Morphology
The morphology of the printed film depends largely on two factors: 1) the kinetics of solvent
evaporation from the ink, and 2) the size and orientation of the nanoparticles. The effect of
nanoparticles on surface roughness are trivial, so most discussion will focus on the kinetics of
solvent evaporation, which are dictated by the temperature used to dry the ink after printing. A 20
wt% DEG ink was used to print 100 µm x 100 µm features on Si samples and dried on a hotplate
at 30 °C, 50 °C, 75 °C and 150 °C. The resulting morphologies were characterized using confocal
microscopy. As the drying temperature increases, the widely known coffee ring effect manifests,
Figure 3-7 – 3D sideview and cross-sectional profiles of 200 µm square (left), 100 µm square, and 100 µm hollow
square with 10 µm linewidth wall.
62
where the edges of the features are significantly higher than the center. For comparison, the
morphology and the cross-sectional profile of the same feature dried at 30 °C and 75 °C is shown
in Figure 3-8.
The difference in the profiles of the features is a manifestation of the widely known coffee ring
effect. The contact line gets pinned at the edge of the droplet and evaporation rate at the center and
the edge is different due to the curvature effect. The difference in the evaporation rate establishes
a convective flow bringing particles to the contact line and causes more particles to assemble on
the edges of the features than the center. This is a significant problem in printing of circuits,
especially for inkjet printing and has been studied before.130-131 The key takeaway, with regards to
this study, is that controlling the relative humidity in addition to the temperature is important to
achieving perfectly flat features. Slower evaporation suppresses the coffee ring formation but is
deleterious to the manufacturability of the process. Using a mixture of low boiling point and high
boiling point solvents, such as DEG and IPA, can suppress the coffee ring while increasing the
drying time of the feature. However, local instabilities increase with increasing evaporation rate,
and may cause the edge roughness to increase.
Figure 3-8 - Cross-sectional profiles and 3D models of the same feature dried at (a) 30 ˚C (RH ~ 45%), (b) 30 ˚C
(RH ~ 10%) and (c) 75 ˚C (RH ~ 30%) using 20 wt% Ag ink in DEG.
(a) 30 °C, RH% ~ 45 (c) 75 °C(b) 30 °C, RH% ~10
63
LER of the printed features was also characterized. High-magnification SEM images were
analyzed using AnalyzeStripes in ImageJ as described in Chapter 2. A 350 nm wide line’s edge
was isolated, and its RMS edge roughness was estimated as 44.9 nm, which is an order of
magnitude higher than that achieved by fluidic assembly. The major contributor to this roughness
is the shape and size of the particle as seen in Figure 3-9. Smaller, more tightly packed particles
such as for fluidic assembly can reduce the edge roughness. Moreover, the LER for microscale
features printed using the FFAsT method exceeds 200 nm primarily due instabilities at the contact
line during the drying process.
3.5.3. Electrical Properties of Printed Films
In addition to thickness and morphology, good electrical transport characteristics of the printed
layers are essential for the proposed process to be used for device fabrication. Except for CNTs
Figure 3-9 - SEM image and corresponding isolated edge (yellow outline) for a 350 nm wide printed feature 100
mm/min.
200 nm
64
and other 2D materials, sintering is typically required to improve the electrical properties of
nanoparticle films as interparticle boundaries serve as scattering sites and increase resistance.
We study the electrical behavior of the printed films from three different aspects: by studying the
kinetics of the nanoparticle sintering by measuring the resistance of a silver nanoparticle film
annealed at increasing temperatures, we correlate the resistance of printed films to their thickness
and identify the role of porosity and particle density, and lastly we estimate the contact resistance
between printed silver and vapor deposited gold to gain insight into the quality of the film.
To study the kinetics of the annealing process, chips were printed at 40 mm/min using 20 wt% Ag
ink in DEG and dried at 30 °C on the hotplate. The samples were then sintered in a rapid thermal
annealing system at 200 °C, 300 °C, 400 °C, 600 °C and 800 °C for 60 sec. For each sample, SEM
was used to capture the sintering kinetics and the sheet resistance was measured using a rough 2-
probe method. Five different pads on each sample were tested and the average is presented. As
seen in Figure 3-10(a), the initial particle size is ~ 50 nm. At 200 °C, nearly all the particles are
seen to have doubled in size approaching 100 nm. From thereon until 300 °C, some particles retain
their size, and other large particulate approximately 250 nm appear. This is a manifestation of the
classical recrystallization process. By 400 °C, the particles have nearly disappeared completely,
and have been replaced by a porous film of silver. The trend in particle growth is corroborated by
Figure 3-10(e), which shows a sharp drop in sheet resistance beyond 300 °C corresponding to the
formation of a continuous sintered film as seen in Figure 3-10(d). While the film is conductive,
65
the pores of approximately ~ 1 µm are present, which may not be ideal for device making. The
pores form as the film is three times as dense as the particle assembly.
Withdrawal speed affects film thickness, and hence the resistance. This is confirmed by Figure
3-10(f) that shows the sheet resistance for a 100 µm x 100 µm pad as a function of its withdrawal
speed. Interestingly, while the film thickness does not depend linearly on withdrawal speed, the
change in sheet resistance is considerably more linear with the withdrawal speed. The contact
resistance post annealing was also characterized using the transmission line method. Silver was
printed at 40 mm/min using a 20 wt% ink in DEG into 100 µm x 40 µm channels and dried at 30
C before being sintered at 300 C using RTP. Gold electrodes (50 nm) were then deposited using
e-beam evaporation with channel sizes 2 µm, 4 µm, 6 µm and 10 µm followed by measuring
resistance across each channel size. The data is shown in Figure 3-10(g), and the y-intercept of
Figure 3-10 - Particle size increase and sintering of silver from 200 °C to 400 °C. The corresponding decrease in
sheet resistance is shown in (f).
Unannealed 200 C 300 C 400 C(a) (b) (c) (d)
0
2
4
6
8
10
12
14
16
18
0 20 40 60 80 100 120
Sheet
Resi
stan
ce (Ω
/sq)
Pulling Speed (mm/min)
Pulling Speed vs Sheet Resistance
20 wt% Silver in DEG
0
5
10
15
20
25
30
0 200 400 600 800 1000
Sheet
Resi
stan
ce (Ω
/sq)
Sintering Temperature (°C)
Sintering Temp. vs Sheet Resistance
20 wt% Silver in DEG
300 C40 mm/min
(f)(e)
0
1
2
3
4
5
6
7
0 4 8 12
Resi
stan
ce (Ω
)
Channel Length (µm)
Channel Length vs Resistance
(g)
66
the linear fit is ~3.4 Ω, which is twice the contact resistance. The small contact resistance of 1.7 Ω
meets the criteria for most practical device applications.
3.6.Applications of the Fast-Fluidic Assembly Process
3.6.1. Wafer-scale Printing of CNTs for Field-effect Transistors
To demonstrate the utility of the proposed FFAsT process, back-gated CNT transistors were
printed on a 300 nm thermally grown SiO2 on a highly doped p-type Si wafer. The wafer was
piranha cleaned (H2SO4:H2O2) for 5 min and rinsed with deionized water. UV lithography was
used along with plasma treatment to generate a chemically heterogenous surface for printing
CNTs. A 0.025 mg/mL aqueous suspension of 99.9% semiconducting CNTs obtained from
Nanointegris was used to print CNTs selectively into hydrophilic channels with dimensions of 100
µm x 10,20,30,40 µm at 100 mm/min through dip-coating. After printing the CNTs, the wafer was
treated with Ar plasma for 2 sec using Unaxis ICP-RIE to remove the SAM and make the entire
substrate hydrophilic. The gas flow was set at 10 sccm and RF1 and RF2 were 250 W each. The
processed was carried out at ~ 6 mTorr. This process is necessary to enable spincoating of
photoresist for electrode deposition. After removing the SAM, the wafer was briefly rinsed with
acetone, IPA and DIW to remove the polymer surfactants coating the CNTs and improve device
performance. Pd electrodes were deposited using the process previously described for CNT
transistors and the devices were characterized using an Agilent 4156C system. The device
architecture and the assembled CNTs are shown in Figure 3-11.Transfer and output characteristics
of the CNT devices were measured for the gate voltage between -40 V to 40 V, and a source drain
voltage of up to -1 V. The results of a typical device are shown in Figure 3-11. A typical device
67
exhibited an on/off ratio exceeding 1000, with the ON-state current in 100’s on nA. The output
characteristics conform well with the transfer curve.
It is noteworthy that the printing of CNTs using the FFAsT method was difficult for two reasons:
1) the suspension has a very low surface tension due to a large amount of surfactants resulting in
poor selectivity in some areas, and 2) because the concentration of CNTs was very low, a very
small amount of CNTs were deposited in a significant number of channels, leading to very low
ON-current values, and non-functional devices.
Figure 3-11 - Device structure (a) and the CNTs printed into channels (b). The transfer and output characteristics are
shown in (c) and (d), respectively.
100 µm
SiO2
Highly doped Si
Ti/PdTi/PdCNTs
0
50
100
150
200
250
300
350
0 0.25 0.5 0.75 1
-Id (nA
)
-Vd (V)
Output Characteristics
40 V 30 V20 V 10 V0 V -10 V-20 V -30 V-40 V
1E-12
1E-11
1E-10
1E-9
1E-8
1E-7
1E-6
-40 -30 -20 -10 0 10 20 30 40
Log
(Id)
Vg (V)
Transfer Characteristics
-250 mV
-500 mV
-750 mV
-1 V
(a) (b)
(c) (d)
68
4. Directed Assembly for Transfer Printing
4.1.NanoOPS Process Description
The offset printing system described herein uses a damascene template to selectively assemble and
transfer nanomaterials onto a polymer film. The damascene template is fabricated using
lithography and thin film deposition as shown in Figure 4-1.
The damascene template’s surface consists of metallic features amid a layer of silanized SiO2. The
conductive features are connected by a conductive film at the back. This conductive film is used
to generate localized electric field at the metallic features by applying a voltage against a counter
electrode. This assembly method is called electrophoretic assembly and is schematically shown in
Figure 4-2.
Figure 4-1 – Damascene template fabrication process.
Si
Si
Ti/Au (5/50 nm)
Si
Ti/Au (5/50 nm)
PR
Si
Si
Ti/Au (5/150 nm)Cr (80 nm)
Ti/Au
Cr Cr
Si
SiO2Au Au
Si
SiO2
Si
Si
Ti/AuSiO2
OH OH OH OH OH
Si
SiO2
OH OH OH OH OH
RRRRR
Deposit Ti/Au
Spincoat photoresist
Lithography
Deposit Ti/Au/CR
Lift-off
Deposit Ti adhesion layer
Deposit PECVD oxide
Lift-off in chrome etchant
Piranha dip
Silanization
Silanization
69
The electric field generated between the template and the counter electrode exerts a force on the
particles in a suspension that is described by:
where E is the electric field strength and q is the charge on the particle. Net charge on a particle
results from a complex interplay between the surface chemistry of particles and the physical and
chemical properties of the liquid in the suspension. A detailed discussion about particle charge and
its significance to electrophoretic assembly is outside of the scope of this work and can be found
elsewhere.16 However, a basic mathematical description of the charge q is:
where ζ is the zeta potential, which is a function of the pH and ionic strength of the solution, κ is
the Debye length, while εr and ε0 are the permittivity of suspension and free space, respectively.
Figure 4-2 - Electrophoretic assembly process.1
Pulling
Directio
n
Surface Tension and
Hydrodynamic force
Electrophoreticaly
driven NPs
(+)
SiO2
Gold
EFDF
v
+
+
+
+
+
+
+
-
-
-
-
-
-
-
+
+
+
𝐹𝐸 = 𝑞𝐸
𝑞 = 4𝜋𝑅휀𝑟휀0(1 + 𝜅𝑅)휁
70
After assembly, the assembled nanomaterial film is transferred to a polymer film by pressing the
template and the receiving substrate together and applying heat and pressure. While different
setups can be used for this purpose, most work presented in this study was performed using a
nanoimprint tool whose schematic is shown in Figure 4-3.
During the transfer process, the polymer substrate softens, the details of which are important and
will be discussed later. Pressure embeds the assembled nanomaterial film into the softened
polymer substrate. Work of adhesion or delamination is a function of the contact surface area and
the interfacial energy. During embedment the contact area between the nanomaterial film and the
polymer substrate increases, and for a complete transfer, the following condition is met: Wadh
(polymer) > Wadh (template).
The Wadh (template) depends on the nature of the contact between the nanomaterial film and the
template. For minimal Wadh (template), the assembled layer should ideally be unsintered. If the
radius of contact between the nanoparticles and the template surface is small, and no bonding
occurs between the two during assembly, the probability of successful transfer increases. However,
Figure 4-3 - NanoNex tool used for transfer printing.7
71
if the assembled nanoparticles are partially or fully sintered at the interface with the template, the
success of transfer printing may significantly decrease. This is especially true for compatible
surfaces such as gold and silver as both are face centered cubic (FCC) with similar lattice
parameters.
The major challenge facing the transfer process is that the polymer properties that enable
successful transfer of the nanomaterial film are unknown. This challenge is addressed in the next
sections.
4.2.Transfer Challenge: Polymer Characterization
4.2.1. Motivation for Polymer Characterization
Transfer printing can be either surface-to-surface ink transfer or embedment of ink into the
receiving substrate. The former operates by promoting adhesive failure at the ink/stamp interface
and adhesive bonding to the receiving substrate. The stamp/ink and substrate/ink interfacial surface
energies are critical in the transfer process. The bulk properties of the receiving substrate are not
critical, and the process occurs at ambient temperature. In embedment-type transfer, the bulk
properties of the substrate play a significant role as the nanomaterial film is embedded into a
polymer substrate during the transfer process. The polymer substrate deforms under heat and
pressure, causing the nanomaterial layer to embed into the substrate. Embedded structures provide
improved structural integrity and better adhesion to the substrate.
While embedment-based printing similar to the NanoOPS process has been previously
demonstrated, most transfer printing techniques are surface-to-surface type such as microcontact
printing where appropriate interfacial energies are deemed critical for successfully transferring ink
from a stamp to a receiving substrate. Some studies investigating microcontact printing consider
72
the point of crack initiation, and the direction and speed of crack propagation as important to the
transfer process.132 While surface-to-surface type printing has been investigated rigorously, the
physics of embedment-type transfer printing is rarely investigated. Few studies that comment on
embedment-type printing mention two primary criteria for successful ink transfer: 1) the receiving
polymer substrate must be a thermoplastic and 2) the receiving substrate must be heated close to
its glass transition temperature.14, 133 The general requirements of favorable interfacial energies
also applies to the current process.
First, we investigated the aforementioned conditions. Four thermoplastic films namely Dupont
Melinex® 454 polyethylene terephthalate (PET), Dupont Teonex® polyethylene naphthalate
(PEN), thermoplastic polyurethane (TPU) from Delphon and polyethylene tetraphthalate glycol
(PETG) were characterized. CNTs were electrophoretically assembled onto 1 cm x 1 cm
damascene templates. All films were plasma treated (O2, 115 W, 15 sccm) prior to transfer (contact
angle with water ~ 0°). Pressure was fixed at 200 psi for all four films while the temperature was
set 5 °C above their glass transition temperatures (Tg) for PET, PEN, TPU and PETG, respectively.
As seen in Figure 4-4, the CNTs could only be successfully transferred to PETG. These results
indicated that exceeding the glass transition temperature of a thermoplastic film does not guarantee
successful embedment of the nanomaterial film and that additional factors could possibly be active
and needed to be explored.
73
4.2.2. Polymer Properties Characterization
After establishing that reaching or exceeding the glass transition temperature of the thermoplastic
does not guarantee transfer, a deeper investigation into other possible factors was required.
Because surface properties and the work of adhesion are important for transfer printing, including
embedment-type printing, it was hypothesized that surface energy and roughness could be
important to the success of embedment. Thus, AFM was used to measure the root mean square
roughness (Rq) of the TPU, PET, PEN and PETG films (NX-20 Park Systems, NCHR Si probe,
non-contact mode) and the surface energy values were determined using the Fowkes Model.16
Water and hexane were used as polar and non-polar liquids, respectively. Phoenix 150 was used
Figure 4-4 - Transfer printing results for different polymer films above their glass transition temperatures. The
damascene template before and after CNT assembly is shown in (a) and (b), respectively. (c), (d), (e) and (f) show
the results for TPU, PET, PEN and PETG.
74
for contact angle measurements. The films were treated with O2 plasma for 2 min to increase their
surface energy just before experiments. Table 4-1 summarizes the results.
Table 4-1 - Summary of properties of the characterized films.
*found from literature
The results summarized in Table 4-1 indicate that the surface properties do not corelate exclusively
with transfer success. Nanomaterials were easily transferred to the TPU film, which had a higher
surface roughness than other films. This indicates that bulk mechanical properties of the polymer
are relevant for the transfer process.
A wider analysis was performed by testing the transfer process on a variety of polymers in which
the bulk mechanical properties, which are strongly affected by the structural properties of
polymers, were also compared. CNTs were transferred to each of the polymer films slightly above
the glass transition temperatures. The success of the transfer process was characterized by
inspecting the residue on the template. The results are summarized in Table 4-2.
Polymer Tg
(˚C)
Surface Energy
(mJ/cm2)
Surface Roughness
(RMS) (nm)
Min. Transfer Temp.
(˚C)
TPU <-30* 27.43 4.26 90
PETG 76 31.26 2.77 85
PET 124 34.01 2.21 No transfer
PEN ~155 41.37 1.98 Poor transfer
75
Table 4-2 - Mechanical properties of polymers tested for transfer.
*found from literature
**found from vendor’s datasheet
Generally, the results suggest that transferring nanomaterial films onto completely amorphous
polymers is easier. It may, therefore, be important to consider the difference in mechanical
properties of crystalline and amorphous polymers as they approach the glass transition. During a
heat cycle, thermoplastic polymers may undergo the glass transition and/or the melting transition.
While these transition temperatures can generally be ascribed to both crystalline and amorphous
polymers, the effect of each type of transition on the physical and mechanical properties of a
polymer varies significantly. Highly crystalline thermoplastics with ordered structures experience
little change in their physical and mechanical properties during the glass transition but experience
a sharp melting transition. In contrast, the effect of the glass transition is significantly more
pronounced on amorphous polymers, which upon exceeding the glass transition temperature, lose
their rigidity and become highly pliable. The behavior of semi-crystalline polymers depends on
Polymer Tg (˚C) %Cryst Yield Strength
(MPa)
Molecular Weight
(g/mol)
Transfer
TPU - Crosslinked 9.7** 10,000 – 50,000* Successful
PETG 76 Amorphous 30* 10,000 – 50,000* Successful
PET 124 ~80% 97** 10,000 – 50,000* Failed
PEN ~155 ~80% 138** 20,000 – 40,000* Failed
PMMA
PC
115*
160
Amorphous
Amorphous
37-73*
63*
2,000 – 350,000*
24,000 – 60,000*
Failed
Successful
76
the degree of crystallinity and typically lies between that of highly crystalline and completely
amorphous polymers. The key element linking the structural properties of a polymer to its behavior
upon reaching the glass transition is chain mobility. Polymers consist of long entangled chains of
molecules. The chains, depending on how much free volume is available, are always in motion.
There is very little motion under the glass transition but may increase considerably once the glass
transition temperature is exceeded, especially for amorphous polymers. Many factors, such as the
degree of crystallinity, molecular weight, chain length, and degree of crosslinking affect the chain
mobility, and hence the nature of the glass transition.134
We hypothesize that the structural properties that dictate the nature of glass transition are critical
to the success of the transfer printing process and hence, the mobility of the polymer chains after
exceeding the glass transition is important. Changes in a structural property that directly affects
chain mobility, such as the degree of crystallinity, rigidity of the backbone, the molecular weight,
and the degree of crosslinking will therefore affect the quality of transfer. To confirm this
hypothesis, we investigated two structural categories of thermoplastic polymers: crosslinked
polymers and amorphous polymers. We chose these polymer types because they typically provide
good transfer, and by testing the same polymer films with reduced chain mobility and gauging its
effect on the quality of transfer, the hypothesis that chain mobility is indeed key to the transfer
process could be verified. Moreover, transfer printing onto highly crystalline films is very
challenging, and semi-crystalline films are prone to structural changes during thermal processing.
For the chosen polymers, the crosslink density in polyurethane affects the chain mobility,135 while
the molecular weight and chain length influence the chain mobility in amorphous polymers.136
77
4.2.3. Crosslinked Polymers: Effect of Crosslink Density
The degree of crosslinking in a polymer profoundly affects its properties. Nielsen first reported on
this phenomenon in 1969.137 The elastic modulus increases significantly with increasing number
of crosslinks and the Tg range broadens with increasing crosslinking.138 The effect of crosslinking
on elastic moduli and damping behavior of polymers has also been investigated. The elastic moduli
of polymers are largely unaffected under the Tg regardless of the degree of crosslinking. However,
the mechanical behavior above the Tg is severely impacted.138
For this study, thermoplastic polyurethane (TPU) was chosen as the crosslinked polymer for
further characterization. It is an inexpensive polymer widely used for flexible and deformable
electronics and shows good transfer characteristics. The aim was to test TPU films with increasing
crosslink density for transfer printing. Four different grades, 1074A, 1095A, NAT-CPT, and
1055D, of Tecothane® TPU were obtained from Lubrizol Inc. as pellets. The mechanical
properties of the four different grades (provided by the vendor) are summarized in
Table 4-3.
Table 4-3 - Mechanical properties of the four grades of TPU obtained from Lubrizol.
TPU
grade
Softening
Point (°C)
Shore Hardness
(A)
Ultimate Tensile
Strength (MPa)
Tensile Strength at
300% Elongation
(MPa)
1074A ~84.7* 75A 41.36 7.51
1095A 84.5 94A 62.05 29.64
NAT-C 67.2, 152.2 32D 62.05 37.86
1055-D 140 54D 69.18 44.81
78
*Tg is likely well below 0 °C, so the softening point was used instead measured using DSC.
Quantifying crosslink density of polymer samples is not trivial. Solid-state nuclear magnetic
resonance spectroscopy (NMR) and dynamic mechanical analysis (DMA) are typically used for
rough comparative estimates of crosslink density.139 However, due to time and resource
constraints, NMR analysis could not be performed. DSC was also performed for the four grades
of TPU the results of which are shown in Figure 4-5.
The softening point for both 1074A and 1095A is around 85 °C, indicating that crosslink densities
were not very dissimilar. 1055D, on the other hand, shows no transition up until ~150 °C. NAT-
025 is more complex, as it exhibits two glass transitions, at 67.2 °C and 152.91 °C, indicating the
use of a diluent. Generally, the trend in glass transition temperatures indicate that the TPU grades
had different crosslink densities, assuming the same chemical structure. The mechanical properties
Figure 4-5 - Calorimetry curves for the four TPU grades investigated in this study.
1074A 1095A
NAT025CPT 1055D
79
of these four TPU grades also corroborate that the four TPU grades had different degrees of
crosslinking.
All TPU grades were dissolved at 50 mg/mL in tetrahydrofuran (THF) except for 1055D, which
was dissolved in dimethylformamide (DMF). 1074A pellets dissolved easily with only stirring
required, whereas the pellets for 1095A and NAT-CPT swelled in THF requiring heat for complete
dissolution. 1055D did not dissolve in THF but continued to swell for several days, and therefore
was dissolved in DMF at the same concentration instead. Here, by dissolution, spreading of the
polymer chains is implied. Each solution was spincoated on glass and baked at 50 °C for 30 min
(1055D was baked at 150 °C for 1 hr to remove DMF) before CNTs assembled on damascene
templates were transfer printed onto each sample. The pressure was fixed at 200 psi and the applied
temperature was increased from 70 °C until either complete transfer was achieved, or the polymer
film melted. The efficiency of the transfer process was characterized by observing the residue on
the template post-transfer. ImageJ was used to convert bright-field images of the template post-
transfer into binary micrographs, which were processed for area covered by CNTs. The template
with the least amount of residue was used as baseline to normalize the results.
The results in Figure 4-6 suggest that chain mobility is important to the quality of the transfer
process as CNTs were fully transferred to polymer grades with smaller crosslink density, 1074A
and 1095A. On the other hand, the grade with the highest crosslink density, 1055D, was least
conducive to transfer and highly variable and patchy transfer was obtained. Even when the
temperature was increased to 120 °C, only ~30% of the CNTs were transferred in the case of
1055D compared to 1074A. Beyond this temperature, the film showed visible signs of melting and
degradation.
80
To strengthen our findings, a commercially available hot-rolled 1055Dfilm was obtained from
Polyzen Ltd and the results were compared with the spincast 1055D film. CNTs were fully
transferred to the hot-rolled 1055D film at 120 °C and 200 psi with good fidelity. This provided a
unique opportunity to gain insights into the effect of the processing route on the suitability for
transfer printing. The physical and mechanical properties of the two films were compared using
DSC and DMA.
First, DSC was performed for both films from 30 °C to 200 °C at 10 °C/min. A stark difference
was observed between the two films: the calorimetry curve for the hot-rolled film was smooth with
a small but sharp melting peak observed at ~180 °C, indicating a relatively ordered and structurally
uniform film. In contrast, the calorimetry curve for the spincast film was significantly less smooth,
Figure 4-6 – CNT residue on templates post-transfer process for 1074A (a), 1095A (b), NPT(c) and 1055D(d).
81
with ‘melting’ occurring over a much wider range. These findings were related to the mechanical
properties of the film by DMA. To prepare a thick film from pellets, 1055D was dissolved in DMF
at 100 mg/mL and spincoated on a Si wafer. The carrier substrate was then baked at 150 °C for ~
1 hr before being peeled off.
82
For DMA, each film, approximately 10 µm thick, was cut into a 20 mm x 4 mm strip. The
frequency was set at 1 Hz with a maximum strain value of 1%. A ramp to 200 °C from room
Figure 4-7 - (a) and (b) show storage moduli and tan delta for spincast and hot-rolled films of 1055D-TPU,
respectively. The DSC curves of the same TPU grade in the form of pellets, spincast film, and hot-rolled film are
compared in (c).
0
0.1
0.2
0.3
0.4
0.5
0.6
1E+5
1E+6
1E+7
1E+8
1E+9
0 50 100 150 200 250
Tan
delta
Sto
rage
Mo
dulu
s (P
a)
Temperature (˚C)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1E+5
1E+6
1E+7
1E+8
1E+9
1E+10
0 50 100 150 200 250
Tan
delta
Sto
rage
Mo
dulu
s (P
a)
Temperature (˚C)
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0 50 100 150 200 250
Heat
flo
w (W
)
Temperature (˚C)
Pellet 1055D
Spincast
Hot-rolled 1055D
(a)
(b)
(c)
Spincast 1055D
Hot-rolled 1055D
83
temperature was performed at 10 °C/min. The apparent inconsistency in transferring nanomaterials
films to the spincast and the hot-rolled 1055D films can be explained by the DMA and DSC results
shown in Figure 4-7. The tan delta or the damping factor of the spincast film, which indicates the
order in the structure, becomes noisier around 120 °C indicating rapid softening of the film as seen
in Figure 4-7(a). This coincides with the wide ‘pseudo’ melting transition for the spincast film
seen in Figure 4-7(c). The tan delta of the hot-rolled film, shown in Figure 4-7(b), is significantly
smoother, and shows signs of softening at a considerably higher temperature than the spincast film
i.e. ~160 °C, which matches with the small but sharp melting transition on the calorimetry curve
in Figure 4-7(c). This indicates that rolling induces a higher degree of order in the cross-linked
TPU film than the structure attained by spincasting. The structural variation in the spincast film
prevents its softening in a small temperature window, hence causing patchy and incomplete
transfer of the nanomaterial film. This is a strong indication that the thermal history and the
processing induced structure, in addition to the chemistry of the polymer, are critical to transferring
nanomaterials. The results of the TPU characterization are summarized in Table 4-4.
Table 4-4 - Summary of TPU characterization for transfer printing.
TPU grade 1074A 1095A NPT-CAT 1055D Spincast 1055D Hot-rolled
Crosslink
density
Low Medium - High High
Softening point ~80 ~90 62.05 120 160
Transfer
quality @
Softening Point
High High Poor Poor (patchy) Medium
Min. transfer
temp. ˚C
75 85 N/A N/A 110
84
4.2.4. Amorphous Polymers: Effect of Molecular Weight
Polymer chains are macromolecules consisting of many smaller molecules also called monomers.
The molecular weight of a polymer is the sum of the atomic weights of molecules in a single
chain.140 It is a key parameter that profoundly affects the physical and mechanical properties of a
polymer. Especially for amorphous polymers, the mechanical behavior post glass transition
depends strongly on the polymer’s molecular weight.141 Longer chains cause more entanglement
and form temporary, physical crosslinks, thereby decreasing chain mobility.141 As a result, the
glass transition typically rises with increasing molecular weight.
In the context of this study, the goal was to assess the effect of increasing molecular weight on the
success of the transfer process. To this end, two different PETG films were obtained from Eastman
Inc and the molecular weights were estimated through their intrinsic viscosity values provided by
the vendor. Another commercial PETG film was also obtained and included in the analysis for
which the viscosity was estimated through the method described in Appendix B. The glass
transition temperature of each film was obtained from DSC. The transition temperature for each
film along with the intrinsic viscosity values are summarized in Table 4-5.
Table 4-5 - Properties of the four grades of PETG.
PETG
type
Tg (°C) Intrinsic
Viscosity
Yield Stress (MPa) Ultimate Tensile Strength
(MPa)
14471 74.5 0.75 53 26
5011 75.1 0.6 - 25
Wafer 75.35 0.86* - -
*estimated using the method described in the Appendix B.
85
Prior to transfer, the Tg was determined using DSC as shown in Figure 4-8. The Tg was found to
increased slightly with increasing intrinsic viscosity. CNTs were assembled onto 1 cm x 1 cm
chips (Nanolab 1mg/mL, 3.5 V, 180 sec) and transferred onto the four PETG samples at 5 °C
above their respective Tg at 200 psi for 60 sec. The success of transfer was evaluated similar to the
TPU samples. CNTs were transferred to all the PETG samples with little to no difference. It was,
therefore, considered that the time for which the transfer process is performed may be important
in this context. This is because once the glass transition temperature is reached, the samples with
a smaller molecular weight may be embedded the same amount in a shorter time than the polymers
with a higher molecular weight. For this purpose, assessing the depth of indentation as a function
of time was required.
An experiment was designed to test the embedment time into each of the polymers. The goal was
to vary the time of transfer printing, while measuring the thickness of the film on the template after
Figure 4-8 - DSC curves for the PETG samples.
14471 5011
PETG Wafer
14471
PETG Wafer
5011
86
each step. This approach can provide insights into the polymer’s molecular weight dependent
viscoelastic behavior during the transfer process.
CNTs are often intertwined, and their residue on the template cannot provide an accurate estimate
of the thickness of the film embedded into the polymer. Instead, approximately ~ 120 nm thick
layer of silver nanoparticles approximately 50 nm in diameter was assembled on 1 cm x 1 cm
chips. Each of the PETG films was treated with O2 plasma for 2 min (0.38 Torr, 15 sccms, 115 W)
prior to the transfer process. The transfer process was performed at 80 °C and 200 psi for each of
the films considering their closely clustered glass transition temperatures. The height of the film
was measured after transferring silver for 5 sec, 10 sec, 15 sec, 30 sec, and 60 sec. For each sample,
the change in the height after transfer was measured using AFM line scans to estimate the thickness
of the silver layer embedded in the PETG substrate. Three measurements of line scan were taken
from each template, and the highest asperity was used as the residue thickness. For all PETG
samples, a 60 sec transfer time left no residue on the template. Figure 4-9 shows the change in
residue thickness with time for each of the PETG grades. The y-intercept is the thickness before
any transfer takes place.
The results cannot be considered highly conclusive due to doubt in the absolute molecular weight
of the films used. Moreover, the PETG wafer sample was about 35% thicker than the other two
films the effect of which cannot be ruled out. However, available results suggest that the chain
mobility does play a role in the transfer process. The polymer sample with the highest intrinsic
viscosity, i.e. the PETG wafer, and hence the highest molecular weight, takes the longest to get
embedded with the nanomaterial film upon exceeding the glass transition temperature. This is
87
additional evidence in support of chain mobility being the key determinant of the success of the
transfer printing process.
4.2.5. Summary of Characterization and Challenges
In summary, two different classes of polymers, crosslinked and amorphous thermoplastics, were
characterized for the transfer process while varying their crosslink density and molecular weight,
respectively. Evidence supports the hypothesis that chain mobility beyond reaching the glass
transition temperature is key to successfully transferring nanomaterial film to polymer substrates.
Perhaps the toughest challenge faced during the study was acquiring a polymer of desired structure
with known characteristics. Often, only mechanical and thermal properties of polymers are shared
by the vendors, and important chemical and structural information needs to be indirectly estimated.
For instance, estimating the molecular weight with accuracy requires access to a gel permeation
chromatography (GPC) tool, which is not easily accessible. Similarly, solid-state NMR can be
Figure 4-9 - Residue thickness as a function of transfer time for each of the PETG samples.
0
20
40
60
80
100
120
140
0 10 20 30 40 50 60 70
Resi
due T
hic
kness
(nm
)
Transfer Process Time (s)
5011
14471
PETG Wafer
88
used to characterize the crosslink densities. Moreover, the mechanical properties quoted by the
vendors are often at or close to the room temperature, and any proprietary diluents or surface
coatings that affect its properties are unknown.
4.3.Transfer Printing through FFAsT Templates instead of Damascene Templates
The method for template fabrication introduced in Section 3.1. is complex and costly, as it requires
seven distinct vapor deposition or etching processes. Moreover, it may take a few days to fabricate
the template, which significantly hinders high volume manufacturing. An alternative to the
damascene template is a SAM patterned template fabricated using a highly doped wafer fabricated
using the technique described in Chapter 3.
The high doping makes the Si wafer a degenerate semiconductor, or a semi-metal. A SAM layer
can be a few nm thick,142 and acts like a dielectric by prohibiting electron transfer and reducing
the electric field intensity over the hydrophobic regions thereby selective deposition in the
patterned hydrophilic features. Moreover, due to the possibility of highly selective deposition on
the template, it may be used independent of electrophoresis for transfer printing. The steps for
fabrication of the damascene template are compared in Table 4-6. The total number of fabrication
steps are reduced from ten to four, and high-vacuum processes are eliminated.
89
Table 4-6 - Comparison of damascene template fabrication with the SAM functionalized template fabrication.
Step# Damascene Template Fabrication Process SAM Functionalized Template Process
1 Ti/Au deposition (high-vacuum) Surface functionalization
2 Lithography Lithography
3 Ti/Au/Cr deposition (high-vacuum) SAM deposition
4
5
6
7
8
9
10
Lift-off
Ti Deposition (high-vacuum)
SiO2 deposition (high-vacuum)
Cr lift-off
Surface functionalization
SAM deposition
Piranha treatment
Lift-off
To demonstrate the use of the SAM patterned template, we transfer silver and MWCNTs onto
PETG an TPU, respectively. The use of a SAM patterned template has also been shown to
electropolymerize polypyrrole selectively on a 4-inch doped wafer and transferred to a PETG
substrate. The same can be used to electroplate different materials such as copper, silver and nickel.
Silver was assembled on a functionalized Si wafer using a 20 wt% Ag ink in DEG at 40 mm/min.
MWCNTs were assembled into 100 µm x 100 µm source-drain electrodes on a highly doped wafer
by applying 50 µA using a counter electrode while immersed in a ~ 0.5 wt% CNT suspension in
90
NMP. For polypyrrole, a highly doped wafer was immersed in the pyrrole monomer solution and
3V were applied for 300 sec. The results are shown in Figure 4-10.
4.4.Applications of Transfer Printing Process
4.4.1. Fully Printed, All-Carbon Flexible FETs
Owing to the emergence of Internet-of-Things (IoT), the demand for low-cost, flexible electronics
for displays, wearable sensors, and smart accessories is increasing.143 Avoiding high-vacuum
vapor deposition and high-temperature processes is key to lowering the cost of such devices.144
Therefore, solution-processable materials such as organic molecules,145 polymers,146 carbon
nanotubes,147 and graphene148 have been used as semiconducting and conducting materials for
flexible devices. However, organic materials degrade if exposed to oxygen,149 while low electronic
Figure 4-10 - Silver transfer printed to a PETG substrate (a) showing clean transfer and sharp edges. A cross-section of
the transferred electrode pair is shown in (c) while the roughness of the electrode is shown in (d). A cross-section of the
entire feature embedded into the substrate is seen in (e). Assembled MWCNTs are shown in (f) and electropolymerized
polypyrrole into 100 µm x 20 µm channels is shown in (g).
91
and thermal conductivities of polymers are disadvantageous for device performance.150 Graphene
presents a variety of complex handling and processing issues. Among these materials, CNTs are
especially promising due to their low cost, superior electrical properties and high mechanical
flexibility.151 Moreover, CNTs can be semiconducting or conducting depending on their structure
and chirality, and thus can be used exclusively to make flexible and low-cost semiconductor
devices. Researchers are, therefore, pursuing development of all-carbon devices that can be highly
flexible while being low-cost.152Several methods such as vapor-deposition,153-154 spraycoating,155
aerosol jet printing,156 and inkjet printing157 have been used independently or combined with
transfer printing to fabricate all-CNT FETs. Vapor-deposition requires high vacuum/temperature
and offers poor control on chirality of CNTs, leading to large inter-device variation158 Inkjet and
aerosol jet printing, on the other hand, provide poor resolution and thickness control that inversely
affect device performance.
Directed assembly is a printing technique that enables precise placement of nanoelements using
surface, capillary, magnetic, or electrophoretic forces.159 In contrast to the above-mentioned
printing methods, directed assembly can be used to precisely and selectively deposit CNT films of
desired thicknesses and morphology. Moreover, the resolution achievable by directed assembly
far exceeds that of commonplace printing methods, with printing in nanoscale features
demonstrated.105, 160 In this paper, we combine the advantages of directed assembly techniques
with a facile transfer printing process to make high-performance, flexible all-carbon FETs. Single-
walled carbon nanotubes (SWNTs) are used as the semiconducting channel, whereas gate and
source-drain electrodes consist of metallic multi-walled carbon nanotubes (MWNTs). SU-8, a
92
carbon epoxy polymer often used as a negative tone photoresist, has been cross-linked and used as
the gate dielectric.
First, SWNTs were assembled on a Si wafer using a previously demonstrated capillary force
assisted printing method.105, 117 UV-lithography was used to pattern 100 µm wide lines with a 500
µm spacing along the length of an oxygen (O2) plasma treated (2 min, 15 sccm, 0.38 Torr, 115 W)
4-inch Si wafer. To assemble SWNTs in the patterned lines, the wafer was immersed in a 0.025
mg/mL SWNT suspension (Nanointegris 99.9% semiconducting) and withdrawn at 0.25 mm/min
using a dip-coater. This printing process exploits the wettability difference between the
hydroxylated Si substrate and the surrounding photoresist template. During assembly, capillary
force causes the ink meniscus to spread and readily wet the exposed, O2 plasma treated Si surface
as it is highly hydrophilic while the spreading of ink on the photoresist is suppressed due to its low
surface energy. This difference in the wetting behavior of the meniscus at the air/ink/substrate
interface causes the solvent in the ink to evaporate faster at the meniscus on the Si surface than the
photoresist surface.39 As the solvent evaporates at the meniscus, the evaporated volume is
replenished by convective flow from the bulk of the suspension that brings along suspended
CNTs.6 The liquid at the meniscus evaporates and leaves behind the suspended CNTs assembled
in the patterned lines.69 Because the ink does not spread readily on the photoresist surface and the
evaporation rate is slow, significantly fewer CNTs are deposited on the photoresist surface.
Substrate preparation and mechanism for the SWNT assembly process are schematically shown in
Figure 1(a) and (c), respectively. SWNTs assembled on the Si wafer are shown in Figure 1(e).
Next, MWNTs were assembled into a source-drain electrode pattern using electrophoresis on a
chemically patterned highly doped p-type Si wafer. The wafer was O2 plasma treated to make it
93
hydrophilic and patterned using AZ N2020 to mask the source-drain electrode pattern. Post-
lithography, the substrate was again treated with O2 plasma, and exposed to 1H,1H,2H,2H-
Perfluorododecyltrichlorosilane (PETS) vapor at 95 °C under a mild vacuum for 3 min. The
hydroxyl groups on the exposed Si surface covalently bond with the organosilane molecule to form
an approximately ~3 nm thick self-assembled monolayer (SAM). This process is called
silanization and is widely used to lower the surface energy of substrates and render them
hydrophobic.118 The source-drain electrode features underlying the photoresist mask remain
highly hydrophilic (water contact angle ~ 0°) while the rest of the surface is hydrophobic (water
contact angle ~ 105°). After silanization, the photoresist mask was stripped to reveal the chemically
heterogenous surface primed for MWNT assembly.
The MWNTs were electrophoretically assembled using a 0.5 wt% NMP MWCNT suspension. The
chemically patterned highly doped Si wafer served as anode and a Si wafer with a vapor-deposited
gold film was used as the counter electrode (cathode). Both substrates were vertically lowered into
a container with NMP MWCNT suspension ~ 5mm apart. A 45 µA current that generated a
potential of approximately 8 V was applied across the two substrates for 5 min. The CNTs in the
suspension carry a net-negative charge due to which they are attracted to the positively charged
chemically patterned substrate. The silane SAM serves two distinct purposes in this assembly
process: 1) it is an insulating layer that hinders the electric field outside of the hydrophilic
features161 and 2) it prevents the area outside the source-drain electrodes to retain any MWNTs
when the substrate is withdrawn due to its hydrophobicity. To the best of our knowledge, this is
the first demonstration of electrophoretic assembly on a chemically patterned highly doped Si
wafer using a SAM as an insulator without a vapor-deposited conductive layer. This Si wafer can
94
be used repeatedly, which is a significant advantage of this technique. The thickness of the MWNT
layer is approximately 100 nm as measured by confocal microscopy as shown in Figure 4-11(g).
The substrate preparation for MWNT assembly and the assembly mechanism are shown
schematically in Figure 4-11 (b) and (d), respectively. An SEM micrograph showing the assembled
MWNTs on the Si wafer is shown in Figure 4-11 (f).
The substrate to transfer print the assembled SWNTs and MWNTs was prepared next. A polyimide
(Kapton® HN) film was treated with O2 plasma and a ~ 2 µm thick layer of MWNTs was dip-
coated on it as the gate electrode using a 3 wt% MWNT suspension at ~600 mm/min followed by
drying at 40 °C on a hotplate. The edge-to-edge resistance of the dip-coated CNT layer across a 2
Figure 4-11 - Schematic depictions of substrate patterning and assembly processes used for SWNT and MWNT are
shown in (a) and (b), respectively. The assembly mechanism for SWNTs and MWNTs is explained in (c) and (d),
respectively. The assembled SWNTs on a Si wafer are shown in (e), while (f) shows the same for MWNTs. A confocal
microscope image and cross-sectional profile of MWNTs assembled on a Si wafer are shown in (g).
45uAOH-OH-
Dip-coatingStrip photoresist
Silanization & resist removal
Assembled CNTsLithography
OH-
OH-
OH-
OH-
Hydrophobic SAM
(b)
Electrophoretic assembly
OH-OH-
OH-OH-
(e)
(f)
(a)
45uA
(c) (d)
200 nm
250 µm
300 µm
200 nm
Concave
meniscus
Nearly
convex
meniscus
(g)
95
cm x 2 cm area was approximately10 kΩ. A thin piece of copper tape was then placed on the edge
of the sample to facilitate gate biasing during FET characterization. Next, a 300 nm thick SU-8
(2000.5) film was spincoated atop the MWNT layer at 3000 RPM for 60 sec followed by a soft
bake at 115 °C for 1 min. The sample was then UV exposed (Quintel 4000 mask aligner) for 60
sec and hard-baked at 200 °C for 15 min on a hotplate to cross-link the SU-8 film and improve its
dielectric properties. A solution of thermoplastic polyurethane (100 mg/mL in
dimethylformamide) was spincoated on the hard-baked SU-8 layer at 4000 RPM for 60 sec. The
TPU film serves as both a secondary gate dielectric and the receiving layer for transfer printing of
the channel and source-drain layers.
Next, the SWNTs and MWNTs were sequentially transferred onto the prepared polyimide
substrate using a nanoimprint lithography tool (Nanonex 2000). Each substrate was pressed against
the PU film on the polyimide followed by applying heat and pressure. During the transfer process,
the TPU film softens and the pressure embeds the CNTs in the TPU film.145 Embedded structures
provide excellent flexibility due to structural support by the surrounding polymer film. For both
SWNT and MWNT layers, transfer process was performed at 90 °C for 2 min at 200 psi. After
transferring both layers, bottom-gated top-contact CNT transistors were ready for characterization.
The substrate preparation process for transferring SWNT and MWNT layers is described
schematically in Figure 4-12(a) while the transfer process is shown in Figure 4-12(b).
96
Electrical characterization was performed using an Agilent 4156C semiconductor analyzer and a
Summit Cascade 12000 probe station with DCP-HTR 19 µm probe tips. Both transfer and output
characteristics of the CNT FETs were measured. 30 devices were randomly tested across three
different samples out of which 27 devices were functional while 3 devices were discontinuous,
(yield ~ 90%). For transfer characteristics, the gate voltage (Vg) was swept between 10 V to -40
V at a step of 80 mV. The source-drain voltage (Vds) was varied between -0.25 V to -1 V. The
transfer and output characteristics are shown in Figure 4-13(a) and (b), respectively.
As evident in the output characteristics, the devices do not saturate, and therefore, the field-effect
mobility was estimated in the linear regime using the following relation:162
Figure 4-12 - (a) Polyimide substrate preparation for transfer printing and the sequential transfer of SWNTs and
MWNTs (b). The prepared substrate with printed CNT FETs is shown in (c), and (d) shows an optical image of an
array of source-drain electrodes with SWNTs.
(a)
Polyimide
Spin-coat SU-8
Spin-coat PU
(b)MWNT
suspension
Dip-coat MWNT
(c)
(d)
Transferred SWNT lines aligned across the
MWNT source-drain electrodes under a microscope
MWNT transfer
SWNT transfer
Transferred SWNTs
97
𝜇𝑙𝑖𝑛 = −𝐿
𝑊𝐶𝑜𝑥𝑉𝑑𝑠
𝛿𝐼𝑑𝑠
𝛿𝑉𝑔
where 𝜇𝑙𝑖𝑛 is the mobility, L is the channel length, W is the channel width, Cox is the dielectric
capacitance, Vds is the source-drain voltage, and 𝛿𝐼𝑑𝑠
𝛿𝑉𝑔 is the slope of the transfer curve in the linear
regime. The combined capacitance of the SU-8/TPU dielectric layer was calculated to be ~1.48
nF. The mobility in the linear regime was calculated using the transfer curve shown in Figure 3(a)
to be 6.39 cm2 V-1 s-1 with an on-off ratio of ~105. The on/off ratio and the mobility of the
characterized devices are summarized in Figure 3(c) and (d), respectively. Over 70% of the devices
have an on/off ratio over 104, while mobility as high as 8.84 cm2 V-1 s-1 was measured. Both values
compare favorably with other all-CNT transistors.154, 163
In conclusion, we present a robust yet simple manufacturing pathway to fabricate all-carbon FETs
on flexible substrates. No high-vacuum or high-temperature process has been used, and directed
assembly assisted transfer printing is employed to attain high-performance devices. The template
used for transfer printing of MWNTs may be used for other materials and can be used repeatedly,
which is a major advantage from a manufacturing viewpoint. The assembly methods introduced
herein can also be used to fabricate CNTs rapidly on rigid substrates.
98
Figure 4-13 - Transfer (a) and output (b) characteristics of a CNT FET. (c) On/off ratio of 27 devices. (d) Field-
effect mobility of 27 devices.
0
1
2
3
4
5
6
7
8
Num
ber
of devi
ces
On/off ratio
1E-12
1E-11
1E-10
1E-9
1E-8
1E-7
1E-6
1E-5
-40-30-20-10010
I DS(A
)
VGS(V)
-250mV
-500mV
-750mV
-1000mV
VDS
0E+0
2E-7
4E-7
6E-7
8E-7
1E-6
1E-6
1E-6
0 0.2 0.4 0.6 0.8 1
-ID
S(A
)
-VDS(V)
10V 0V
-10V -20V
-30V -40V
0
2
4
6
8
10
12
0 1 2 3 4 5 6 7 8
Num
ber
of devi
ces
Field-effect mobility(cm2V -1s -1)
(a) (b)
(c) (d)
Kapton
MWNTs
SU-8
PU
MW MWSW L = 25 µmW = 100 µm
99
5. Lifecycle Management of Nano-enabled Products
5.1.Energy Demand for Printing Nanoscale Electronics
Semiconductor fabrication processes are highly resource intensive as they need high temperature
and vacuum, consume significant amounts of water and energy, and produce large quantities of
solvent waste. The resource intensiveness of the semiconductor fabrication industry may be
reduced by adopting additive manufacturing techniques, which are touted as significantly less
resource intensive than conventional fabrication processes albeit with very little evidence. In
contrast to the conventional top-down layer-by-layer deposition and etching processes, bottom-up
manufacturing techniques use nanomaterial-based inks to print circuits at or close to the room
temperature/pressure. However, the use of nanomaterials gives rise to two concerns: a)
nanomaterials have significantly higher embodied energies than their ‘non-nano’ forms,164-167 and
b) the true toxicity of nanomaterials is uncharacterized if exposures occur.168-170 The threat of
toxicity cannot be adequately quantified currently as many relevant characterization factors are
unknown. However, performing a cumulative energy demand (CED) assessment of the available
electronics printing methods is feasible and can help quantify any gains made by printing
electronics instead of fabricating them. CED is a popular method used to quantify the total energy
used by all processes and materials needed to manufacture a product or provide a service.171 The
assessment strategy usually follows a similar path as an environmental life cycle assessment. At
the start of the assessment, the boundary and the scope of the analysis is defined followed by
quantifying all the energy and material inputs to the process chain up to the final product or service
along with the inherent uncertainties. The cumulative energy demand is then calculated, and
recommendations can be made based on the interpretation of the results.
100
The goal of this study was to conduct a CED assessment of the printing of electronics. For this
analysis to be of wider relevance, it was necessary to analyze the printing process for a product
that is fundamental to consumer and industrial electronics. A field effect transistor (FET) is the
most significant circuit component used in modern electronics. An FET controls current between
two electrodes connected by a semiconducting channel via electric field applied through a third,
electrically insulated electrode. Metal-oxide semiconductor FET, also known as MOSFET,
supplements the ubiquitous complementary metal-oxide semiconductor (CMOS) manufacturing
process, which is used to manufacture logic gates and amplifiers for integrated circuits.172 Among
the many types of MOSFETs, the two major designs are the planar FET design and the trigate FET
design. The planar FET structure was predominantly used until the late 2000s. In 2011, Intel Corp.
introduced the 22 nm node trigate FinFET structure, which offered significant increase in the
performance and packing density of transistors. 173 It was further scaled to the 14 nm node structure
in 2015, which represents the state of the art in commercially available microprocessors today.174
Therefore, to make this study technologically relevant, it was deemed suitable that the energy cost
be estimated for printing the 14 nm node size. It is noteworthy that the 22 nm and 14 nm are not
representative of any dimensions, but simply refer to the design rules followed for a certain process
technology. A detailed technical discussion on the different FET designs/structures is beyond the
scope of this paper and can be found elsewhere.175
5.1.1. CED Goal and Scope for Printed FET
This paper estimates the cumulative energy demand of printing 14 nm node field effect transistors
on a one cm2 area. This functional unit is based on a previous study with which a useful comparison
may be drawn as explained in the discussion section.9 The boundary of the CED analysis was set
101
from the acquisition of raw materials to the printing of the final product, i.e., a single layer of 14
nm node FinFETs printed on one cm2 of silicon. The energy consumed during use and by any end-
of-life processes is not included in this analysis.
While many printing processes such as inkjet, flexographic and gravure have been demonstrated,
only two methods have demonstrated the capability to print nanoscale features with reasonable
scalability. 25, 35, 176-177 Our previous work has demonstrated methods for printing metals and
semiconductors at the nanoscale. 21, 35, 177 The first method is transfer printing, in which stamps
with nanoscale features are inked and made to come into a conformal contact with a desired
substrate to transfer the ink to the receiving substrate. Inking is achieved through solution-based
processes such as directed assembly. Because the surface and mechanical properties of polymers
are typically easier to manipulate than rigid substrates, transfer printing is more suitable for
printing on flexible substrates than rigid substrates such as silicon. The stamps are also prone to
damage and issues such as feature fidelity, post-transfer ink residue on the stamp, and a need for
regular cleaning of the features make the process complicated. The second printing method capable
of printing nanoscale features is directed assembly of nanoparticles using fluidic and/or interfacial
forces. This mostly additive process better suits this assessment. However, because many printing
methods employ lithography or surface preparation techniques, the results obtained in this study
are widely relevant. The throughput of the printing process is not considered in this study.
102
5.1.2. Process and material inventory analysis
A 14 nm node FinFET structure is shown in Figure 5-1. To calculate the mass of the materials
required to print this FinFET structure, and subsequently estimate its embodied energy, the
dimensions of the FET structure are needed. The fin and gate dimensions for 14 nm node are
estimated from the available literature and SEM micrographs.3, 178-179 Although the fin tapers from
an 18 nm wide base to a 7 nm wide top, this study assumes a constant width of 18 nm. Each fin is
42 nm in height. Although the fin length is not explicitly available, SEM micrographs were used
to estimate it to be ~90 nm. The gate, sitting perpendicular to the fin, is 30 nm long and 80 nm
wide. Its exact height is unavailable, but because it must be taller than the fin (42 nm high), it is
assumed to be 60 nm in height. The channel is assumed to be one-third (30 nm) of the fin length,
and just as high and wide as the rest of the fin. A high-κ dielectric that insulates the gate from the
channel is assumed to be 10 nm thick and 18 nm wide.
The next step was to define the materials used to print the trigate FET. In a conventionally
fabricated structure, the source and drain electrodes typically consist of highly doped silicon and
the semiconducting channel (region connecting the source and drain) is lightly doped silicon. A
thin layer of a high-κ dielectric that is deposited on top of the channel insulates it from the metallic
Figure 5-1 - A typical FinFET structure (not drawn to scale – based on a similar sketch from Intel).3
103
gate electrode. For printing a MOSFET, either the same materials as used in fabricated structures
can be chosen, or a novel structure employing different nanomaterials and predicted to be a high-
performance replacement for the conventional FET can be analysed. It is difficult to use silicon
nanoparticles for printing, because silicon nanoparticles in commercially available inks are
typically covered by a native oxide, which could cause technical difficulties. Thus, a novel, more
realistic structure is assumed, with pure semiconducting carbon nanotubes (CNTs) as the channel
material while all electrodes, including source, drain and gate, are assumed to be printed with silver
nanoparticles. A sub-10 nm CNT FET has been previously demonstrated in a study not focused on
printing.180 Although processing challenges currently impede their commercialization, CNTs have
been demonstrated to provide near-ballistic transport among other significant advantages over
conventional silicon based FETs.181 Alumina nanoparticles are assumed to print the high-κ
dielectric layer. Silicon is assumed as the underlying substrate. While it should be noted that a
14 nm node FET has not been realized using the printing process described in this study, the
feasibility of the chosen materials as well as the described process for printing nanoscale features
is previously demonstrated.
In the next stage of the assessment, all the process steps and their associated material and energy
inputs are inventoried. For the proposed printing method, a template-based directed assembly
Figure 5-2 - Process sequence for printing 14 nm node FET.
104
process exploiting selective wettability of two adjacent surfaces can be used in combination with
dip-coating to selectively assemble nanoparticles or nanotubes into lithographically defined and
chemically activated patterns.182 Chemical activation of the surface is a pre-requisite for most
printing techniques to promote ink adhesion and controlled wetting. Such activation can be
achieved through a combination of plasma and chemical treatment methods. After surface
functionalization and patterning, the substrate can be dipped into a tank containing the ink of the
desired material and withdrawn at a fixed speed. In addition to nanoparticle structures, crystalline
films of organic materials can be printed this method.5 Repeating the assembly process through
templating can be used to print each material or structure. Lift-off using organic solvents is
required after printing each layer. To improve their electrical characteristics, the printed
nanomaterials, except for single-walled carbon nanotubes (SWNTs), need to be sintered to melt
the particles and cause them to resemble a thin film. Many sintering methods such as laser, plasma
and microwave assisted sintering are previously demonstrated for this purpose.183 However, in this
study the energy consumed by the sintering step is derived for photonic curing or flash sintering.184
This method uses a powerful flash of light to cause particle sintering. The sequence for printing a
trigate FinFET is depicted in Figure 5-2. Theoretically, repeating this sequence with dielectric
layers in-between (that can be either spin-casted or evaporated) provides a pathway to printing
multi-layer circuits. Printing blanket coats of nanoparticles with high uniformity, although
challenging, may also be a possibility.
5.1.3. Computing energy demand
The mass of each material required to print a single FET was calculated using the dimensions
stated in the previous sub-section. The number of FETs on a one cm2 was calculated using the fact
105
that E7-8890 Xeon® processor has 7.2 billion 14 nm node trigate FETs on a 456 mm2 area.185
Proportionally, the number comes out to be about 1.58 billion FETs/cm2. The embodied energies
(MJ/kg) of silver nanoparticles and SWNTs were taken from literature.65 The embodied energy of
alumina (Al2O3) particles was not available so the value for zinc oxide (ZnO), another commonly
used meta-oxide nanoparticle, was used instead. Because of the high performance required from
the device, it was assumed that only the purest nanomaterials can be used, and thus, the upper
bound of the embodied energy values were used for the assessment. The embodied energy of the
silicon wafer was taken from literature.186 Material and energy consumption estimates for substrate
cleaning and photolithography are available.187-188 The energy requirements for dip-coating and
surface functionalization were estimated from lab-scale experiments. The material and energy
requirements for surface functionalization can vary widely depending on the approach taken.
However, this assessment assumed an oxygen plasma treatment. For simplicity and lack of
concrete data, the same sintering conditions were assumed for each nanomaterial used and the
energy was estimated for a Novacentrix® 1300 flash sintering tool. For all non-nano materials the
embodied energy values were computed using the USEI database with the Cumulative Energy
Demand method available in the SimaPro® environment. Table 5-1 summarizes the material and
energy inputs of the processes required for printing FETs on a one cm2 area.
106
Table 5-1 - Material and energy inputs for each process step used for printing a cm2 silicon chip with FETs.
Process step Materials used Primary
equipment
used
Process
energy (MJ)
Material
energy (MJ)
Total
energy/cm2
(MJ)
Cleaning (x 1/layer),
surface functionalization
(x 4/layer) O2 gas, H2SO4, H2O2 Plasma asher 0.2 x 4 =
0.8*
0.0001
(O2)**, 0.001
(H2SO4)**,
0.02
(H2O2)**
0.8
Photolithography (x
4/layer)
Photoresist,
developer
Photomask
aligner - - 0.51
Printing channel Pure semiconducting
SWNTs
Dipcoater 0.08* 2.67E-06** 0.08
Printing source and drain Silver nanoparticles Dipcoater 0.08* 1.12E-09** 0.08
Printing high-K
dielectric
Alumina
nanoparticles
Dipcoater 0.08* 1.67E-06** 0.08
Printing gate electrode Silver nanoparticles Dipcoater 0.08* 1.12E-09** 0.08
Lift-off
(x 4/layer)
Acetone, Isopropyl
alcohol
- 0.22 0.22*
Photonic sintering Silicon wafer Flash lamp 0.125*
Silicon substrate 0.35**
Total embodied energy for a single layer 2.48
*Value estimated purely from labscale experiments.
**Value taken directly from literature or computed using USEI database.
The total embodied energy of a single printed layer of MOSFETs on one cm2 of silicon is
approximately 2.48 MJ as shown in Table 5-1. For this number to provide any meaningful insights,
it is important to compare it with the embodied energy of FETs fabricated using the conventional
CMOS fabrication process. The 14 nm node process is proprietary and cannot be analyzed for
cumulative energy demand in a similar fashion. However, a widely cited study published in 2002
estimated the embodied energy of a conventionally fabricated 32 MB DRAM, which was
fabricated on a one cm2 silicon chip.9 The study does not mention the number or type of transistors,
but it can serve for a useful proxy with some limitations.
Several layers of MOSFETs and capacitors are interconnected and programmed to make electronic
components such as microprocessors and dynamic random-access memory (DRAM) elements.
The size of the DRAM depends on the packing density and the number of layers of FETs on the
107
chip. While 32 MB DRAM was commonplace in 2002, there is no commercially available DRAM
of a similar capacity made with the 14 nm node process. In comparison to the 2.48 MJ of energy
consumed for printing in the current study, the energy consumption for fabricating a 32 MB
DRAM was estimated at around 35 MJ, which is approximately an order of magnitude greater than
that for printing of a similar sized chip with FETs using directed-assembly of silver nanoparticles,
alumina nanoparticles, and carbon nanotubes. The study referred to herein also included a yield of
82% and 75% for the process and silicon wafer area used, respectively. Incorporating the same for
the process used in this study raises the energy consumption to above 3 MJ – still much smaller
compared to 35 MJ. The compared values do not include the energy and material consumption for
the assembly of the final chip. Again, it should be noted that this is not considered as a direct
comparison, as semiconductor fabrication processes have evolved considerably over the past 16
years. The packing density of transistors has also increased by approximately 50% since 2002.
Moreover, the DRAM architectures have significantly changed over the years with 3D DRAM
replacing older, planar architecture for most applications. It is also significant that the energy
estimate in our study is only for printing a layer of FETs, whereas a DRAM structure has equal
numbers of FETs and capacitors, which could imply the use of additional deposition and etching
processes. However, with all the caveats to considering it as a meaningful comparison, the
difference in numbers meets intuition: by avoiding vapor deposition processes, the energy use can
be significantly reduced. Inventory data for comparison of printed electronics with those fabricated
using the current 14 nm node proprietary process would provide a clearer analysis of the reduction
in cumulative energy demand. Many hidden energy and material costs are expected to be identified
when the printing process is analyzed for widescale commercialization.
108
Another interesting feature in Table 5-1 is the nearly negligible contribution of the nanomaterials
used in the process towards the total embodied energy of the printed FETs. The largest
contributions are by cleaning and surface functionalization, photolithography, and particle
sintering processes. Surface functionalization and preparation techniques are difficult to avoid
when printing. However, by devising methods to avoid photolithography, the energy requirement
can further be reduced. Noticeably, the energy required to produce nanoparticle inks most likely
exceeds the embodied energy of nanoparticles themselves. Filtration, purification,
ultracentrifugation and ultrasonication processes may be required to stabilize the inks. However,
even increasing the embodied energy of the nanomaterials by an order of magnitude does not alter
the overall results of the analysis and the conclusions remain largely unchanged.
Sintering of nanoparticles depends on many different factors and the energy consumption may
vary widely with nanoparticle material type, size, shape, functionalization, and packing density.
Metals typically sinter at lower temperatures, but ceramics such as alumina and wide band-gap
semiconductors that are increasingly used in electronics are difficult to sinter without high
temperatures. Thus, improving the analysis requires deeper characterization of the sintering
process and the energy consumed for sintering of different materials and structures. This exercise
is of broad interest as nearly all printing technologies utilizing nanomaterials will require sintering.
For some materials, however, printing of layers through nanomaterials is not possible due to strict
requirements for morphology and crystal structure. Such layers are typically deposited using
epitaxial processes. Therefore, a hybrid approach focused at minimizing vapour deposition and
etching processes must be pursued and analysed for energy consumption.
The present analysis does not take into regard the possible environmental and ecological
implications of the possible release of nanomaterials from the products they are contained in.
109
Sintering processes reduce the surface area of nanomaterials significantly, causing them to
resemble thin films, and hence diminishing the threat of toxicity. However, for materials such as
carbon nanotubes and quantum dots, sintering is not required. An environmental assessment for
any devices printed with such nanomaterials is also needed as more characterization factors
become available.
Unless the throughput of printing techniques is comparable to the conventional fabrication
processes, the functional unit chosen in this study cannot provide a holistic assessment. Dip-
coating-based printing, for example, can be a slow process based on the material and feature size
being printed. For such a case, non-process related energy consumption such as for ventilation and
lighting contribute to the overall energy footprint and need to be included in the comparative
analysis, which is possible by collecting more data for the printing process and choosing a more
appropriate functional unit. This could potentially highlight the trade-offs between avoiding direct
high-energy related processes and the in-direct energy cost to run the entire facility for longer
durations.
5.1.4. Conclusions
To conclude, this study offers a preliminary analysis of the energy gains achievable by adapting
additive manufacturing practices for printing electronics. These are the major conclusions of the
study:
1. Printing of circuit boards instead of using fabrication methods could potentially reduce
energy consumption by approximately an order of magnitude.
2. Embodied energy of nanomaterials is not a major contributor to the total energy
consumption for printed electronics.
110
3. Comparison of the estimated energy consumption to print trigate FETs with that of the
cutting-edge conventional fabrication processes to produce the same structure is vital and
must be performed for an accurate picture of the situation.
In summary, semiconductor fabrication industry uses processes that have substantial
environmental and monetary costs. The results of this preliminary study suggest that adapting
additive manufacturing approaches can potentially reduce the resulting environmental impact and
associated manufacturing costs, which could potentially enable marginalized communities and
peoples to access cutting edge technologies.
5.2.End-of-Life Management of Nanoenabled Products: Product Stewardship Strategies
5.2.1. Introduction
There is growing need to develop effective stewardship programs as consumer products become
more complex, and increasingly employ a diverse class of novel materials, including engineered
nanomaterials, or ENMs. These materials are nano-sized in at least one dimension, and offer
excellent physical, mechanical, chemical and electrical properties; because of which, they are
becoming ubiquitous across the consumer product spectrum. Approximately 300,000 metric tons
(mt) of ENMs are used annually in a variety of products, including cosmetics, electronics,
medicine, automotive, energy, environment, and coatings and paints 189-190. With increased use of
ENMs in various applications, concerns arise regarding the possible negative implications of their
largescale use in consumer products 191 for two reasons: 1) short and long-term toxicity of ENMs
is not well understood, and 2) significantly more energy is required to produce nano-enabled
products (NEPs) due to energy requirements for very high purity ENMs, which increases the cost
and the carbon footprint of their manufacture. While the Organisation for Economic Co-operation
and Development (OECD) has published guidelines on safety testing and assessment of
111
nanomaterials, not all engineered nanomaterials have been certified in accordance with the
guidelines.
The benefits of ENMs can be fully exploited without apprehensions if exposures are avoided.
Product stewardship strategies present opportunities to address these challenges, thereby allowing
potential to reduce the environmental and energy cost of the use of ENMs in consumer products.
One approach would involve material specific regulatory oversight, but this is currently
improbable due to the lack of consensus on the toxicity of and possible health risks posed by
nanomaterials. Another approach, outlined in this work, is to probe the capacity of the current
regulatory landscape to guide end-of-life best practices and the feasibility for stewardship of nano-
enabled products. Although probabilistic estimates of release of ENMs and the risks it poses to the
environment during production and use of consumer products are available,169, 192-193 the possible
effect of the current regulatory landscape on impeding the release of ENMs has not been explored
for the U.S. The following questions serve as the objectives for this study:
a) What percentage of ENMs can be prevented from meeting an uncertain fate through the
existing product stewardship and EPR infrastructure?
b) Is it economically feasible to utilize existing recycling infrastructures to recover
nanomaterials from the products collected through EPR?
c) What current strategies or future improvements in the legal infrastructure can be most
conducive towards the goal of reducing the impact of NEPs?
By addressing these questions, recommendations can be made for improving the effectiveness of
product stewardship and EPR practices in the U.S. for emerging technologies using ENMs and
NEPs.
112
5.2.2. Methods
A two-part approach was used to conduct this study. The first part focused on tracking the flow of
ENMs via the products that contained them and analyzing whether pertinent laws may impede
ENM release, and possibly aid their recovery and reuse. In the second part, the economic feasibility
of recovering nanomaterials from end-of-life (EOL) products was explored by using a hypothetical
case study of recovering quantum dots (QDs) from the television screens and probing the current
capacity of the collection and recycling infrastructure in the U.S. to handle novel materials. The
possible challenges to increasing this capacity were also identified.
Two distinct terms that are repeatedly used throughout this paper are ‘collectable’ and
‘recoverable’. By collectable, it is implied that the material may be prevented from release to the
environment by the collection of the product it is contained in, via regulation or incentives enacted
specifically for that product type. It is not necessary for a collectable material to be recyclable/
recoverable. For a nanomaterial to be considered recoverable, it is assumed that in addition to
being collectable with existing U.S. regulation, its physical and chemical properties should not
have changed significantly during use – allowing for its use in secondary applications without
much conditioning. All collected ENMs that are not recoverable are assumed to be disposed
appropriately. Hence, all recoverable materials are also collectable, but not vice versa.
If risks are established, some ENMs could receive special treatment if their release to the
environment carried unreasonably high risk. However, because of the current knowledge gaps
regarding specific ENM risks, only economic benefit is expected to drive their recovery. Lead
automobile batteries, for example, have high collection and recycling rates because they hold
known economic value 194 as well as dangerous health effects. On the other hand, collecting and
recycling millions of consumer lightbulbs that contain mercury has required greater efforts despite
113
concerns about their cumulative environmental and health risks. Considering this precedent, it is
expected that even if the risks of nanomaterials were better understood, the lack of clear economic
incentive may hinder collection efforts. Analysis of potential recovery processes and pathways to
make collection and recovery of ENMs economically beneficial is critical to safer NEP adoption.
Over 15% of ENMs produced each year are used in personal care products such as cosmetics,
detergents, and sunscreens. 195 ENMs from these products predominantly end up in wastewater
treatment plants, from where they could be potentially filtered, recovered and reused. However,
considering the heterogeneity of the mixtures obtained from filters in waste water treatment plants,
this idea is very difficult to implement. In addition, during the time spent in waste water,
nanomaterials contact a wide range of chemical species that may alter their surfaces irreversibly.
Significant amount of ENMs are also used in another dispersive application, architectural paint.196
Collection of nanoparticles from used architectural paint is not possible. Therefore, based on the
dataset used by Keller et al., we estimate that about 35% - 40% of ENMs, regardless of the
stewardship strategy adopted, cannot be easily recovered or prevented from meeting uncertain
fates. This is recognized by some manufacturers, who are modifying ENMs using chemical
pathways to render them innocuous at the end-of-life.
Several methods of recovering nanomaterials from heterogeneous mixtures have been reported in
the literature, including ultracentrifugation,197 precipitation,198 polymer induced liquid-liquid
extraction through microemulsions,199-200 and complex formations.201 These methods were
developed to lower the impact of NEPs, although only one study201 attempted to quantify the
economic or environmental gains achievable through their methods. This is because their results
were reported from lab-scale experiments.
114
Global ENM production and use have been estimated in several studies.169, 190, 202 These estimates
vary significantly, which indicates the difficulty in assessing the fluid state of nanotechnology. In
addition to the production data, the use data is also elusive. For the current investigation, data for
maximum production and use numbers of ENMs is assumed using estimates by Future Markets
Inc. and cited by Keller et al. 189, 203. A very small amount (<1%) of ENMs is consumed by research
activities; therefore, for simplicity 100% of ENMs produced were assumed to be used in household
and industrial applications.
Because NEPs remain a significant potential pathway of ENM release 204-205, the first step was to
identify and categorize the consumer products in the U.S. relevant to the objectives of this study.
Product categories were selected based on the following criteria: a) existing federal or state
stewardship/EPR laws in the U.S. that apply to the relevant product category and b) a product
category that contributes significantly to overall global consumption of ENMs. Both criteria had
to be mutually fulfilled for a product category to be included in the analysis. Hence four product
categories were included for analysis: consumer electronics, paints and coatings, batteries, and
pharmaceutical drugs. These categories capture over 35% of the total annual global production of
ENMs 189.
For the case study of QD recovery from used televisions, the current and forecasted market trends
for televisions containing QDs are considered. Thereafter, based on the sales data, we estimate the
number of QD-based televisions available for recycling over a 10-year period. The cost of
recovering QDs is assumed, and the profit made through recovery is based on the market price of
QDs adjusted over time owing to improvements in technology. The probability of net profit is then
assessed using a Monte-Carlo simulation.
115
5.2.3. Analysis
A 2012 report by Future Markets Inc. and cited by Keller et al. 189 estimated the maximum annual
global production of nanomaterials in 2010 to be around 318,200 mt. In a separate study, Keller et
al. used the Inequality-adjusted Human Development (IHDI) to estimate that the U.S. has about
12% share of the nano-enabled consumer products produced globally. If 12% of the products are
assumed to contain 12% of the ENMs produced globally, the total mass of ENMs utilized in the
U.S. can be calculated to be 38,184 mt. We also assumed this amount is uniformly distributed
among all states. Next, the legislative database of the Product Stewardship Institute 206 was utilized
to enlist the relevant U.S. stewardship laws. The applicability of these stewardship and EPR laws
was assessed for each of the product categories, and the extent to which the current regulatory
framework could aid in impeding the release or recovery of ENMs from nano-enabled products
was ascertained. Finally, the challenges to make the recovery process feasible and more effective
were identified and are discussed for each product category.
For the four product applications (paints and coatings, household batteries, pharmaceutical drugs,
and consumer electronics) we assessed the feasibility of collection and recyclability of the major
ENMs used in each application. The mass of ENMs assumed is shown in Table 5-2. Other
assumptions in Table 1 are described in more detail in the following sections. Because electronics
utilize a significantly wider variety of nanomaterials, and because the efforts for their recovery
have been the strongest, a subcategory of electronic products was analyzed in greater depth. A
rudimentary cost analysis for the recovery of quantum dots (QDs), which are increasingly used in
various applications, was also performed to assess the profitability of operating a recycling facility
for the products that contain them.
116
Table 5-2 - The type, and the total amount of ENMs used in select product categories worldwide and in the U.S. in 2010. ‘Broader Category’ refers to the category
from Keller et al. 2013 to quantify the consumption of ENMs. The ‘Selected Application’ refers to each product category investigated for this study. Selected
applications are classified within the broader categories. Assumptions made to estimate the amounts are included, with 12% share of the U.S. in the global ENM
use is assumed for all product categories.
Broader
Category
(from Keller et
al. 2013)
Global
Consumption
in Broader
Category
(mt) 189
Selected
Application within
Broader Category
(this study)
Global
Consumption for
the Selected
Application (mt)
(this study)
Calculated Mass of
ENMs in the
Selected
Application in the
U.S. (mt)
(this study using
Keller 2013
estimate of 12%
share of U.S. in
global consumption
of NEPs.)
Major
ENMs
Used in
Selected
Application
States with
Collection Laws as
of 2017 for Selected
Application
Population of the
states as % of U.S.
population in 2017
Coating,
Paints &
Pigments
80,500
In ‘Coatings,
Paints &
Pigments”
Architectural Paint 20,100
(25% of global
consumption for
‘Paints and
Coatings’
2,412
TiO2, SiO2 CA, CO, CT, ME,
OR, MN, RI, VT,
DC
23%
Energy and
Environment
43,700
in ‘Energy
and
Environment’
Household Batteries 10,925
(25% of global
consumption for
‘Energy and
Environment’
1,311 CNTs,
SiO2, Al2O3
CA, FL, NJ, NY,
ME, MD, MI, IA,
VT
32%
Medical 13,400
in ‘Medical’
Nanopharmaceutical
Drugs
6,700
(50% of ENMs
used in ‘Medical’
field.)
804 ZnO,
Al2O3,
Fe/Fe
oxides
CeO2, Ag
Several CA counties
and small cities.
Two counties in WI,
and one in IL.
6%
Electronics
and Optics
48,700
in
‘Electronics
and Optics’
Consumer
Electronics
(televisions, laptop
and desktop
computers, displays)
24,350
(50% of the
global
consumption for
‘Electronics and
Optics’)
2,922 (components)
966 (coatings)
Fe, SiO2,
Al2O3,
ZnO,
Ag/Au,
Quantum
dots
CA, NJ, NY, NC,
SC, RI, VA, VT,
TX, CT, HI, IL, IN,
ME, MD, MI, MN,
MO, OK, OR, PA,
WA, WV, WI, UT,
DC
65%
117
Paints and coatings (architectural paint)
Nanomaterials can add scratch, corrosion and stain resistance to architectural paints as well as to
coatings for textiles, electronics, and automobiles.196, 207-209 Silica and titania are the most
extensively used ENMs in paints and coatings. Although coatings, paints and pigments contribute
the most to the overall ENM production,189 they are primarily used in the manufacture of consumer
products, and not used by consumers directly. The only applications of ENM-based coatings
fulfilling the criteria for this study are in architectural paint and consumer electronics.
Architectural paint is the most common application of nano-enabled coatings.210-211 Therefore, at
least 25% of the ENMs produced annually for use in paints and coatings i.e., 20,100 mt, were
assumed to be used in architectural paint. It was also assumed that consumers have equal
preference for nano and non nano-enabled paint. The electronics industry is also an important
consumer of nano-enabled coatings212 and was assumed to have a 10% share i.e. 966 mt, in ENMs
consumption for coatings and paints in the U.S. (The recovery and reuse potential of this share is
analyzed in the consumer electronics section and is not included here.)
In 2017, 9 states with 23% of the U.S. population had enacted laws regarding takeback and safe
disposal of leftover architectural paint as shown in Table 5-2. According to the American Coatings
Association, approximately 10%, or 65 million gallons of architectural paint remains unused each
year213 and assuming that 23% of that paint could be collected (based on the population percent of
the state with relevant laws), it is estimated that about 55 mt of ENMs, primarily silica, titania,
alumina and zinc oxide, could be collected by existing takeback programs to prevent undesired
dispersion. Recovery of titania, silica, and alumina nanoparticles from unused paint is unlikely for
secondary uses given their low cost and high production volumes. However, unused paint, if
118
suitably stores prior to collection could be resold at a reduced price. It is projected that 55 mt of
ENMs could be collected through existing takeback programs but none is recoverable.
Household Batteries
Nanomaterials can reduce the weight and increase the specific energy of rechargeable batteries
used in portable electronics.214 While few batteries currently use nanomaterials (mostly carbon
nanotubes (CNTs)), their use is projected to increase.215 The exact mass of ENMs used for battery
applications is unknown. 25% of the ENMs used for energy and environment were assumed to be
dedicated to battery manufacture. This estimate is because ENMs have a high material-to-product
weight ratio for batteries than other product types. For example, a laptop battery weighing 100
grams, could contain up to 6 grams of CNTs amounting to around 6 wt.%,216 which is significantly
larger than some of the other popular applications such as CNT composites.217
Household batteries include both rechargeable and non-rechargeable batteries. Stewardship laws
in most states, except for Vermont, address rechargeable batteries only. Collection of batteries is
required by at least 9 states in the U.S., which encompass about 32% of the total U.S. population.
The collection and recycling of nano-enabled batteries has been previously explored.218 Most
nanomaterials used in the manufacture of batteries, such as multi-walled-carbon nanotubes
(MWCNTs), are inexpensive and are in an agglomerated state inside the batteries.219 Moreover,
their physical properties degrade with use over the battery’s lifetime, which hinders recovery for
secondary reuse. Lithium and cobalt are also used in batteries, but because they are currently
inexpensive, recovery of these materials has been unable to drive the collection of used batteries.
The situation regarding lithium and cobalt could change as their biggest reserves are in politically
unstable countries (Afghanistan and the Congo, respectively). If not recovered, MWCNTs would
need to be destroyed by heating to at least 660˚C, 220 which would require energy, thus increasing
119
the cost of disposal. The collection of small batteries may be financially less attractive, because
the manufacturer or the recycler may be held responsible for the proper disposal of nanotubes, or
other ENMs. In summary, while slightly more than 400 mt of ENMs contained in batteries appear
to be collectable, only a very small amount, probably less than 1% (4 mt), can be expected to be
recovered/recoverable.
Nanopharmaceutical Drugs
Nanomedicine can be broadly categorized into two main fields, nanodiagnostics and
nanopharmaceuticals.221 Detailed descriptions of these terms are available elsewhere,222-223 but in
short, nanodiagnostics and nanopharmaceuticals involve improving disease diagnoses techniques
and drug-based treatment techniques, respectively. While nanodiagnostic applications are mostly
developed for hospital/medical facility-based use and are not a part of this study,
nanopharmaceuticals (as drugs) are primarily aimed for household use by consumers. A BCC
Market Research Overview221 estimated the global market for nanopharmaceuticals in 2016 to be
15 times larger than nanodiagnostics. Based on this insight, we assume that the majority of ENMs
consumed globally in the medical field, i.e., 13,400 mt, are used in nanopharmaceuticals, and thus
in drugs. Therefore, as a conservative estimate based on the 15:1 ratio by the BCC Market
Research, 50% of ENMs used (6,700 mt of 13,400 mt) in medical applications globally were
assumed to be used by household consumers outside of medical facilities as part of prescription
and over-the-counter drugs. As mentioned earlier, the U.S. uses about 12% of the nano-enabled
products manufactured globally.203 Therefore, the mass of ENMs used in drugs in the U.S. is
assumed to be 12% of 6,700 mt, or 804 mt per year.
About 50% of the medicine dispensed to consumers in the U.S. each year remains unused and is
disposed of incorrectly.224 Assuming ENMs are homogenously distributed among drugs,
120
approximately 402 mt of ENMs (50% of 804 mt) can be released to the environment from incorrect
disposal of unused medicine in the U.S.
Several counties, mostly in California, New York, Illinois, and Washington have enacted
regulation pertaining to stewardship efforts based on collection of unused medicines from
consumers. Massachusetts adopted a state-wide substance control law that requires drug
manufacturers to either setup stewardship programs by themselves, or to enroll in a state-run
stewardship program. The population of the states and localities covered by pharmaceutical related
EPR laws represents about 6% of the total population of the U.S. Hence, a maximum of 15 mt of
ENMs (6% of 402 mt) is collectable. Because of low heterogeneity of drug mixtures, ENMs could
possibly be recovered chemically. However, due to increasing awareness regarding drug wastage,
and research suggesting that most medicines do not lose efficacy despite passing their expiration
dates,225 efforts to resell or donate unused medicines are gaining pace in the U.S., with
organizations such as SafeNetRx226 providing unused collected medicine on need basis, either free
or at a fraction of the original cost. Hence, about 6 mt of nanoparticles used in medicines could be
prevented from release through collection by the current laws, and potentially reused under some
circumstances, but none could be viably recovered.
Consumer Electronics
Nanomaterials are finding wide-ranging use in consumer electronics. Nanoparticles have a very
large surface area and the nano-shape and size may be advantageous for some electronic
applications. Nanoparticles can be suspended in suitable solvents to form stable inks that can be
used to print electronics rather than employing the conventional layer-by-layer deposition and
etching-based fabrication processes. Metallic nanoparticles are projected to be widely used as
interconnects in electronics.227 Silicon, zinc oxide, and carbon-based nanomaterials are being
121
widely explored for use in transistors.228 Iron-platinum or cobalt nanoparticles may be used for
producing very large and stable magnetic storage,229 and transparent electrodes with silver
nanoparticles have shown considerable potential to replace indium tin oxide (ITO) as front contact
for displays.230
When used for printing electronics, the nanomaterials may undergo sintering treatments such as
heating, laser sintering, or flash curing, which significantly alter the physical and chemical
properties of nanomaterials and cause them to resemble thin films. This is desirable for reducing
toxicity to which consumers might be subjected, since the potential hazard results from the small
sizes and large aspect ratios of nanomaterials and not bulk materials or thin films. However,
recovery of materials in their nano-form becomes impossible and the embodied energy due to a
very large surface area is lost. On the other hand, for some applications such as optoelectronics,
sensors and coolants, it is preferred that nanoparticles maintain their shape and other physical
properties during use. Only for these nanomaterials could recovery be possible and economically
or environmentally beneficial.
Electronics in this study are defined along the lines of the current legislative policy in the U.S.,
which varies by state, but usually includes desktop and laptop computers and their peripherals,
LCD and CRT monitors, and televisions. Based on the ubiquity of these electronic devices, it was
assumed that at least 50% of ENMs used globally per annum in electronics and optics applications,
or 24,350 mt, are utilized in these electronic devices. Cell phones, which also contribute to ENM
use in electronics, are not covered under the e-waste recycling laws, except for California, New
York and Illinois. Thus, cell phones are not included in this analysis. Based on the assumption that
12% of these nano-enabled electronics will be used in the U.S. (Keller and Lazareva 2014), it was
122
estimated that 2,922 mt of ENMs are used in electronics (desktop and laptop computers and their
peripherals, LCD and CRT monitors, and televisions) sold in the U.S. each year.
In the U.S. in 2017, 26 states with 65% of the U.S. population have enacted laws placing varying
degrees of responsibility on producers of consumer electronics. Nano-enabled products were
assumed to be evenly distributed in the U.S. Therefore, a maximum of 65% of nano-enabled
consumer electronic devices containing 1,900 mt of ENMs, in theory, could be collected at product
end-of-life. A much smaller amount, possibly less than 10% of 1900 mt, is expected to be in the
recoverable form as explained before. As stated previously, ENMs are increasingly used to make
protective coatings for electronics, with 10%, or 966 mt of ENMs assumed to be used for this
purpose in the U.S. With the assumption that all of the 65% of products with nano-enabled coatings
can be collected, about 628 mt of ENMs could be prevented from release albeit, the chances of
recovery are slim.
Consumer Electronics: The Case of Quantum Dots
Insulating nanoparticles such as silica or titania, used to improve cooling in circuits, are
inexpensive to produce,231 and carry a low-risk of human or ecological toxicity. Consequently,
even when possible, their recovery may not be of the highest priority. However, gold nanoparticles
and semiconducting nanomaterials such as quantum dots are more expensive, which may make
their recovery economically beneficial. An interesting application is the increasing use of quantum
dots in high-end displays, which are anticipated to be utilized in televisions, cell phones and
laptops.232 Quantum dots (QDs) carry a considerable ecotoxicity potential, 233 which could
promote their recovery. Ideally, under inert conditions, the physical properties of QDs should not
change much over the lifecycle of a typical television, which makes for a promising case study.
123
Analysis was performed for recovering quantum dots from television displays using a cost-benefit
approach. There are several possible end-of-life scenarios for QD-containing televisions. Fully
functional televisions can be simply refurbished and sold again. If there are any malfunctioning
parts, the television is disassembled, and the screen can be used as replacement for other
televisions, with the remaining parts scavenged for valuables. However, if the screen itself exhibits
defects, depending on the type of the screen, it can either be shredded, or scavenged for recovery
of expensive materials such as quantum dots. These materials can then be purified, characterized
for functionality, and again used in the manufacture of products. The scenarios for reuse, recycling
and recovery and recommended practices are described in the R2 Standard by Sustainable
Electronics Recycling International.234 The cost-benefit analysis in this study was performed for
the recovery scenario. Because exact data were not available, it was assumed that between 40% to
60% of the QD based televisions collected will have some malfunction in their screens. This range
was modelled using a uniform distribution.
A quantum dot is a semiconducting nanoparticle with a size-tunable bandgap, typically between 1
nm -10 nm in diameter. Depending on its size and shape, it can be excited optically or electrically
to produce light of various colors/wavelengths.235 Initially, optically induced luminescence of
quantum dots was exploited, leading to the development and commercialization of ‘on-edge’
quantum dot-based displays. Recently, interest has turned to developing large-sized electrically
driven, or ‘on-surface’ displays. In on-surface device architectures, a thin film of several
monolayers of individual QDs is stacked between the two charge carrying layers. In on-edge
architectures, a tube containing quantum dots is placed in front of LEDs to excite the QDs. The
details of these devices are available elsewhere.236
124
Available estimates suggest that ‘on-edge’ and ‘on-surface’ screens (42 in – 60 in) contain between
75-150 mg and 1,970 - 4,040 mg of cadmium selenide (CdSe) quantum dots, respectively.236 The
mass estimate for the on-edge screen seems reasonable and was used for this analysis using a
uniform distribution. However, the mass estimate of QDs for on-surface type screens seems much
larger than technically feasible. Therefore, to calculate the mass of QDs in on-surface type screens,
a close-packed film geometry was used (30-90 nm in thickness of CdSe QDs) where the maximum
possible mass range was calculated to be 70-477 mg (based on screen size) and modeled using a
uniform distribution.
For the analysis, it was assumed that in 2015, only 10% of QD-based televisions were ‘on-surface’,
with the other 90% being ‘on-edge’ type devices. A transition of 10% per annum was assumed
from on-edge to on-surface technology, resulting in all QD-based televisions with on-surface
technology by 2025 (with calculations for a 5% transition rate shown in Figure S4 in the
supplementary information). The mass of QDs per television was based on this assumption, with
uniform distribution used for the mass of QDs contained in each type of architecture. The number
of televisions available for recycling was calculated using global sales of QD televisions in 2015
and projected over the next 10 years using available forecasts.237-238 Almost half of the televisions
sold in 2025 are calculated to be QD-based. In accordance with the assumption for the use of nano-
enabled products in the U.S., 12% of QD-based televisions were assumed to be sold in the U.S.
each year.203 This number was assumed constant for the duration of the analysis. The sale price of
the recovered quantum dots was modelled using a triangular distribution, with 25% and 75% of
the market price in 2017 serving as the lower and the upper bounds, respectively. With time, the
number of ‘on-surface’ devices will increase, but the cost of producing quantum dots from primary
processes is also expected to decrease, leading to a decrease in the selling price of QDs post-
125
recovery. This sale price was assumed to reduce by 10% each year, due to forecast improvements
in the technology and increase production capacity,239 which will decrease the cost of the primary
material. The competitive differential advantage of virgin/primary QDs over secondary material is
ignored for this analysis. Based on discussion with Electronic Recyclers International Inc.
(Kocabas, Shim et al. 2006), the current television recycling/recovery capacity of the U.S. was
assumed sufficient for the recovery of QDs over the next 10 years, and hence no initial investment
was included in the analysis. Assuming a 5-year lifetime of a QD-based television, the analysis
was performed for the period of 2020-2030.
For this study, no specific recovery process was used to provide estimates for material and energy
costs; instead, approximate figures inferred through thermodynamic arguments (e.g., comparing
surface area energies) and available numbers for bulk material recovery processes are used. No
specific processes for recovering or recycling quantum dots from television displays have been
reported. Some estimates suggest the cost of recycling a television between $25 and $35.240
However, considering the sensitivity of quantum dots to chemical exposure, more sophisticated
processes to prevent exposure and preserve their physical properties would be required. However,
QD recovery from televisions could cost between 10 times to 100 times more. Because the cost of
recovery is a critical parameter, rather than using a distribution to include its effect on profit, three
126
different scenarios (A, B and C) were explored with the recovery cost fixed at $300, $1500 and
$3000 per television. The effect of other inputs on the expected profit was calculated.
Figure 5-3 - Probability density function of the profit made through recycling of quantum dots in televisions from
2020-2030 for Scenario A (($300/television) (a) and (b) and Scenario B ($1500/television) (c) and (d) assuming
100% yield (a) and (c) and 50% yield (b) and (d). The probability density function for Scenario A($300/television)
with decreasing recovery cost is shown in (e) while (f) compares the profit for decreasing vs fixed recovery cost for
a period of 10 years.
0%
20%
40%
60%
80%
100%
0
200
400
600
800
-7 3 14 25
Cu
mu
lative P
rob
ability
Fre
qu
en
cy
Billions (USD)
a) Scenario A
(100% yield)
0%
20%
40%
60%
80%
100%
0
200
400
600
800
-55 -42 -30 -17
Cu
mu
lative P
rob
ability
Fre
qu
en
cy
Billions (USD)
0%10%20%30%40%50%60%70%80%90%100%
0
200
400
600
800
-2 9 19 30
Cu
mu
lative P
rob
ability
Fre
qu
en
cy
Billions (USD)
0%10%20%30%40%50%60%70%80%90%100%
0
200
400
600
800
-10 -5 -1 3 7
Cu
mu
lative P
rob
ability
Fre
qu
en
cy
Billions (USD)
0%10%20%30%40%50%60%70%80%90%100%
0
200
400
600
800
-57 -50 -43 -37 -30
Cu
mu
lative P
rob
ability
Fre
qu
en
cy
Billions (USD)
0
200
400
600
800
1000
1200
2020 2022 2024 2026 2028 2030
Pro
fit
in M
illi
ons
(USD
)
Year
Decreasing recovery cost
Fixed recovery cost
f) Scenario A (100% yield)
b) Scenario A
(50% yield)
c) ScenarioB
(100% yield)
e) Scenario A (100% yield)
Decreasing recovery cost
d) ScenarioB
(50% yield)
Profit Loss
127
The results for the scenarios A and B are presented in Figure 5-3. The results of Scenario C
followed the exact trend as Scenario B (i.e., no chance of making a profit). Monte Carlo simulation
with 10,000 runs was used to determine the probability of breaking even over the 2020-2030, 10-
year period.
In Figure 5-3(a) and (c), the analyses assume a 100% efficiency of recovery (yield) of quantum
dots from televisions, without any loss of QDs during the lifetime of televisions, or during the
recovery processes. However, QDs may be photobleached if exposed to oxygen, which reduces
their luminescence. Therefore, scenarios assuming a 50% yield of the recovery processes are
shown in Figure 5-3(b) and (d). These results indicate that the profitability of recovery is severely
reduced at the lower yield threshold. Moreover, while the probability of earning a profit with a
reduced yield is non-zero for scenario A (with costs at $300/ television), there is no chance of
earning a profit for scenario B (when recycling costs reach $1500) according to this analysis. A
separate analysis was performed to find the crossover point for the breakeven recovery cost of
QDs per television. It was found that profitability – even assuming a 100% yield – is eliminated
as the cost of recovery per television approaches $1000.
It is important to comment on the middling results for the Scenario A ($300/television). Under the
stated assumptions, the recovery venture has a fair chance of success. However, yearly trends
reveal that for a fixed recovery cost, the profit grows for the first few years, but then decreases
steadily – implying that over a longer period the chances of making a profit would be significantly
lower. This is because while the sale price of the recovered QDs is assumed to decrease each year
(due to improved technology and increased production capacity of primary QDs), the cost of
recovering QDs from televisions is fixed. To investigate the possible effect of decrease in the
recovery cost, a separate analysis was performed for Scenario A wherein the recovery cost was
128
also assumed to decrease by 10% annually, reaching $94/television at the end of the 10-year
period, which is still three times as much as the cost of recycling a conventional television in 2017.
The probability distribution and the density function for depreciating recovery cost are shown in
Figure 5-3(e). Comparison of Figure 5-3(a) and 1(e) shows that the profitability of the venture
markedly increases with depreciating recovery cost. A comparison of the yearly trends shown in
Figure 5-3(f) suggests that for annually decreasing recovery cost, the profit is expected to increase
almost linearly before plateauing. These trends emphasize the significance of developing and
improving recovery processes for QDs to keep recovery ventures beneficial over long term. In
future, due to increased production, quantum dots will become considerably less expensive and
will find widespread use in consumer electronics. This could potentially present a challenge to the
viability of any recovery efforts. The health and environmental risks, however, will remain the
same. Therefore, it is imperative to the adoption of quantum dots, and ENMs generally, to develop
processes for their recovery and reuse.
Sensitivity analyses were performed for the parameters defined by distributions, which were 1)
QD mass on surface type (uniform distribution), 2) QD mass on edge type (uniform distribution),
3) sale price of recovered QDs per gram (triangular distribution), and 4) the percentage of faulty
screens collected each year (uniform distribution). The analyses were performed for all scenarios
using the median baseline for each distribution, and the results are shown as tornado plots in Figure
5-4. For Scenario A, where the cost of recycling QDs is assumed fixed at $300 per television, the
expected profit is most sensitive to the mass of QDs in on-surface type televisions. The sale price
per gram at which the recovered QDs are sold is the second largest contributor to uncertainty in
profits. The percentage of faulty screens collected each year as well as the mass of QDs in on-edge
type screens are the smallest contributors to the uncertainty. For Scenario B, with the recovery cost
129
assumed as $1500 per television, again, the expected profit is most sensitive to the mass of QDs
in on-surface type televisions. In contrast to Scenario A, the percentage of faulty screens processed
has a much more pronounced effect on the expected profit as shown in Figure 5-4 (b). This
indicates that beyond a certain recovery cost, any increase in the number of screens processed each
year negatively affects the profitability.
To verify this observation, a linear correlation analysis using the Pearson method revealed that the
percentage of faulty screens has a negative correlation with the expected profit. (See Figure S2 in
supplementary information). Again, as previously explained, this is because in the model it is
assumed that the selling price of the recovered QDs depreciates with time, while the cost of
recovery per television is assumed to be static.
To summarize, the estimate for the mass of QDs per television is most critical to accurately assess
the expected profits. The results also indicate the significance of the transition timeline from on-
edge to on-surface type technology. This transition will provide an increased mass of QDs for
Figure 5-4 - Sensitivity analyses for a) Scenario A ($300/television) and b) Scenario B ($1500/television) showing
the profitability of the recovery process to be most affected by the mass of QDs in on-surface type televisions. For
the higher recovery cost (Scenario B), the percent of faulty screens affects the profit margin more profoundly than
for lower recovery cost (Scenario A). The mass of QDs in on-edge type screens affects the profitability the least.
130
recovery, which will potentially offset the effect of the decline of the sale price of the recovered
material. With slower transition rates, it is anticipated that the probability of profit is significantly
reduced. Lastly, as expected, the sale price of the secondary QDs, which depends directly on the
price of the primary QDs, is important to make this venture profitable. When the recovery
processes are less expensive than primary QD production processes, the recovery system can be
financially viable. It is noteworthy that EPR strategies are typically employed when EOL
management is not financially feasible. Any chances of QD recovery, therefore, are an additional
benefit especially given their potential toxicity.
5.2.4. Discussion
The feasibility of collection and recyclability of four nano-enabled product applications (paints
and coatings, household batteries, pharmaceutical drugs, and consumer electronics with a specific
focus on QDs) is investigated. As stated earlier, it is assumed that 12% of the ENMs produced
globally – about 38,000 mt across four product categories - are estimated to be used in the U.S.
each year.203 The findings of this study suggest that maximum implementation of the current
regulations through collection could prevent approximately 10% of the nanomaterials used in
consumer products from entering uncertain pathways of release. The situation for reuse and/or
recovery is somewhat bleaker. As Table 5-3 shows, architectural paint, if collected and stored
appropriately, can provide a pathway to reuse without requiring much processing. Nanomaterials
contained in household batteries and pharmaceutical drugs are least likely to be recovered and
reused, with less than 1% recoverable. For consumer electronics, while their collection is globally
pursued, most ENMs contained in electronic products are not recoverable, although could still be
reused through secondary usage of the product they are contained in. If recycling procedures were
131
implemented, 1% – 2% of ENMs contained in electronics could be recovered for use in similar or
different applications.
It is evident from Table 5-3 that if the intent is to prevent release and dispersion to avoid unintended
consequences, there is considerable potential under the current EPR laws, which can be expanded
by merely extending the existing legal infrastructure over a wider population. However, it is also
clear that collection and recovery of nano-enabled devices could face financial challenges as the
potential for reuse in secondary applications is very low.
Table 5-3 - Total annual amount of collectable and recoverable nanomaterials for nano-enabled applications
investigated: paints and coatings, household batteries, pharmaceutical drugs, and consumer electronics. Architectural Paint Household Batteries Nanopharmaceutical Drugs Consumer Electronics
ENM use in the U.S. 2,415 mt 1,311 mt 80 mt 2,922 mt (components)
966 mt (coatings)
Population covered by laws 23 % 32 % 6 % 65 %
Collectable 56 mt 419 mt 1.5 mt 1,900 mt (components)
628 mt (coatings)
Recoverable for reuse 56 mt (~100%) <1% 0.8 mt (<1%) <1%
There is, however, untapped potential for recovery and reuse for many other nano-enabled
products that are not widely covered by the current EPR laws. For example, in most states, cell
phones are not covered by e-waste recycling laws. California, however, has enacted a separate law,
called the Cell Phone Recycling Act that requires retailers to setup and implement an effective
takeback/recycling system before selling their products to consumers.241 Manufacturers of cell
phones play no direct role in this process. Cell phones are anticipated to be one of the central
technologies to make use of nanomaterials for their casing, electronics and sensing systems.
Enactment of laws similar to California’s could lead to greater prevention of ENM release and
potentially increased recycle/reuse. The automotive recycling industry also presents interesting
132
opportunities. Although there are no mandates regarding automotive recycling in the U.S., over
95% of EOL vehicles are recycled. It is one of the most efficient recycling sector in the U.S. Nano-
reinforced materials are widely used in automobile manufacturing. Exterior surfaces of
automobiles are coated with nano-based coatings, and ENM-reinforced materials are used in
chassis and tires of automobiles. If the automotive recycling industry is equipped with the
technology to recycle nanomaterials, more ENMs could be prevented from release and possibly
recovered for reuse. It should be noted that EPR – or recycling by the manufacturer – is not the
only mode by which the nanomaterials maybe recycled or reused. Nanomaterials used in largescale
industrial applications such as catalysis are recovered and reconditioned by their primary users and
therefore, are not released to the environment. A formalization of the recovery processes currently
used in the industry may help account for other ENMs.
Several major challenges exist that could hinder the collection, recovery and reuse of nanoparticles
from consumer products, even if it were monetarily beneficial. The first is the capacity of the
current facilities to deal with the collected waste. For example, despite poor collection rates for
electronics, existing recycling facilities have difficulty in dealing with the amount of waste
available to them due to high labor and energy costs, as well an increasing number of regulations
on processing e-waste. Therefore, some recyclers find it easier to export a significant portion of
the collected waste. Around 7 million mt of e-waste were generated in the U.S. in 2014, 15% (1
million mt) of which was collected.242 Separately, it was also estimated that 8.5% of the e-waste
collected in the U.S. was exported, most of it to developing countries where waste management
practices may be problematic.243 The ironic result of stiffer U.S. e-waste laws could be higher
incidents of global release for ENMs, and a greater chance of potentially toxic exposure to
unauthorized recyclers. However, the current EPR laws increasingly require environmentally
133
sound management (ESM), the chances of greater export activity especially to unauthorized
recyclers with the expansion of electronics recycling are low.
Dealing with nano-waste with the goal of recovering ENMs for reuse is challenging. Even if
elaborate third party and public-sector collection and recycling programs were to exist, processing
would be difficult because recyclers would need product details of where exactly, and in what form
a certain nanomaterial has been used. Recycling practices in this scenario may even exacerbate an
existent problem; for example, shredding of nano-reinforced polymers may release ENMs to air.
Informative EPR, through which producers provide relevant information about their products to
aid recycling, could additionally be used to aid recovery or safe disposal of ENMs. This would
require for producers to share more detailed technical information regarding their products. An
industry-compatible sharing method can be a federally managed stewardship database through
which information about nano-enabled devices embedded into radio-frequency identification
(RFID) tags would be made accessible to authorized recyclers. These tags would inform recyclers
regarding the exact use, form, quantity as well as standard procedures for the recovery of ENMs
contained within a device. The feasibility of RFID tags for aiding stewardship has been previously
explored,244 but the scope could be extended to explore the ramifications of including more
information regarding the design and disassembly of the product. Several EU countries such as
Denmark and France have taken some initiative in this direction by adopting ‘nano registers’,
which contain products with the type and form of the ENMs they contain. Their success, however,
has been marginal due to difficulties faced in collecting information from overseas manufacturers.
This could be a significant challenge o any informative EPR efforts adopted in the U.S. Electronics
undergo frequent design changes due to which recycling systems may also need to be upgraded
frequently. Implementation of physical and financial EPR is difficult due to the wide range of
134
NEPs marketed by companies of different sizes, some of which may not be able to establish and
sustain take-back programs especially when not legally required to do so. Liability EPR is unlikely
as any long-term damage caused by the release of ENMs may not be currently measurable.
It is clear from this study that the existing e-waste recycling strategies in the U.S. are not
technologically capable of making the recovery and reuse of nanomaterials financially beneficial.
For example, during and before the 1970s, the car recycling industry in the U.S. was nearly non-
existent. The introduction of new shredder technology saw a turnover of the industry over the next
decades, with the car recycling industry becoming the 16th largest industry in the U.S. in 2017. A
similar technological leap – aimed at the recovery of nanomaterials from consumer products or
redesign of ENMs to accommodate their recycling – would help recover nanomaterials from waste
and provide a financial impetus to the recovery of consumer products in general.
Enactment of regulations for nano-enabled products is currently unlikely in the U.S. Initial
research regarding nanomaterial toxicity is presently inconclusive, and therefore, any potential
long or short-term toxicity risk is unestablished. In such a scenario, enforcement of even a weak
precautionary law is difficult for two reasons: 1) enforcing laws consumes resources and
dedicating public funds for a perceived or unestablished threat is difficult, and 2) engaging any
stakeholder, especially producers with a financial motive, is unlikely without offering clear
empirical evidence both of risks and of the monetary benefit of any stewardship strategies.
Stewardship strategies are based on the evaluation of risk. First, the need for such practices must
be established based on ENMs posing unreasonable risk, and second, the feasibility of any
suggested practice should be demonstrated. If a nanomaterial is shown to carry an unreasonably
high risk, either its stewardship will need to be made mandatory, or its use in consumer products
would have to be discouraged.
135
Identifying and quantifying environmental benefits can incentivize the recovery of relatively
inexpensive particles through policy drivers such as carbon tax or maximum disposal limits.
Determination of the exact tipping point where a carbon tax could dictate recovery with certainty,
is a challenge. Some injunctions of the California Electronic Waste Management Act such as the
collection of Electronic Waste Recycling Fee may have consequences in the context of recovering
meta/nanomaterials from consumer waste, which may not have come into focus. Moreover,
informative EPR focused legislative efforts like the proposed Right-to-Repair bills245 requiring
manufacturers to share enough information with third party recyclers to enable them to more
effectively recycle materials contained in their products may have a significant impact. An in-
depth study investigating the implications of these policies using other techniques is needed.
The stated objectives of the study were to assess the effectiveness of the current EPR landscape in
the U.S. in mitigating the release of ENMs from certain categories of nano-enabled products, to
investigate if any ENM recovery efforts at the end-of-life could be economically viable, and to
identify potential measures to aid the capture and recovery of ENMs from nano-enabled products.
Rough estimates show that the release of a maximum of about 10% of ENMs used in the U.S. may
be avoided through implementation of the current collection, recycling and Extended Producer
Responsibility (EPR) strategies for certain product categories. However, in most cases, end-of-life
nano-enabled products hold little additional value for the consumer or recyclers. Because there are
no regulations specifically pertaining to the use of ENMs in consumer products, these NEPs will
end up in landfills and/or incinerators, where they may persist.246 Secondly, based on the scenario
analyses conducted in this study, the chances of making profit through recovery and reuse of
nanomaterials from consumer products are currently slim. The chances could increase as more
nano-enabled products enter consumer markets. Significant amounts of ENMs are used in
136
potentially reusable/recyclable products such as cell phones, televisions, and computers. Provided
the right technical and legal infrastructure, ENMs in these products could potentially be recovered.
The extent to which the current technical infrastructure may be feasible for recovery and reuse of
ENMs for secondary applications is unclear, but it is critical that better ENM recovery methods
are developed. Lastly, informative EPR strategies present an effective method to reduce the
environmental and energy cost of the use of ENMs in consumer products, but, save in instances
where recoverable materials hold clear market value, likely would require government-backed
enforcement and economic incentives, both for producers and consumers, if they are to work as
envisioned.
137
6. Summary and Future Work
6.1.Summary
Fluidic assembly is a promising method, the understanding of which has been advanced by this
study. The relationship between the resist thickness and the printed film thickness has been
identified and explained based on the spatially misaligned contact line inside the printed feature.
Confinement, which is the ratio between a feature’s height and width, and the resulting total
meniscus length have been introduced as a measure to predict the thickness of the film for purely
capillary regime. A new bilayer patterning process to improve the morphology and increase the
selectivity of fluidic assembly has been developed. Evidence is presented to suggest that the rabbit
ears effect stems from the climbing of the resist sidewall by the particle suspension and can be
suppressed through precise patterning and silanization of the resist sidewall to obtain flatter printed
structures.
For the NanoOPS process, significant insights into the transfer step have been gained. The role of
chain mobility and uniformity of structure of the polymer substrate in successfully transferring
nanomaterial films has been explored. Factors that decrease chain mobility, such as higher
crosslink density, high molecular weight are less conducive to the printing process. Moreover, the
uniformity of structure resulting from the processing profoundly affects the transfer process.
A newly developed directed assembly-based printing process utilizes the dip-coating platform with
surface engineering to selectively print on a variety of rigid surfaces including Si, SiO2, sapphire,
glass, quartz, etc is also presented. The process has been shown to print conductive,
semiconducting and insulating materials from 100 µm down to sub 100 nm scale at a high speed
100 mm/min or faster. We have shown that the thickness and the morphology of the printed film
138
can be tailored by controlling the withdrawal speed and modulating solvent properties (surface
tension, concentration of nanomaterials, viscosity, volatility).
6.2.Contributions
This thesis makes contributions in four different areas.
1) We expand on the knowledge of fluidic assembly on a physically patterned surface through
dip-coating by providing insights into the confinement effect leading to increased particle
assembly. We present an experimental approach for controlling the thickness of the printed
structures and develop a novel bilayer patterning process to pattern a surface with a
functionalized photoresist.
2) We advance the understanding of the embedment-type transfer printing process by
investigating the polymer characteristics that are conducive to transfer printing. We report
that polymer chain mobility is key to the transfer process, and that a tradeoff typically exists
between transfer yield and pattern fidelity.
3) We develop and demonstrate a rapid directed assembly process to print different types of
materials for device-making in the micro and nanoscale regimes. The method increases by
the printing throughput for the fluidic assembly, and with slight modifications, can be used
to increase the manufacturability and decrease the environmental and economic cost of the
NanoOPS transfer printing process.
4) We compare the energy costs of printing and fabricating nanoscale devices and show that
despite large uncertainties, printing of devices can reduce the energy demand by almost an
order of magnitude. The fundamental physics and device performance challenges have not
been considered. We also show that wider application of the current EPR laws in the US
can significantly lower the risks of release of nanomaterials to the environment.
139
6.3.Future Work
6.3.1. Fluidic Assembly
a) A numerical model of the proposed confinement theory is needed to further verify the
particle assembly mechanism in chemically heterogenous channels and geometrically
patterned substrates.
b) The relationship developed between the photoresist and printed film thickness applies
when the photoresist is not hydrophobic. However, for the newly proposed bilayer
patterning process for resist silanization, the model may need to be experimentally revised
to predict film thickness.
Figure 6-1 - A graphical comparoson of the Fast-Fluidic Assembly process with the previously developed processes
at the CHN. The fastest drying time for the ink has been considered for this chart.
Thro
ugh
put
(m2/s
)
Demonstrated Resolution (µm)
0.1 1 10 100
1E-5
1E-4
1E-3
1E-2
1E-1
Electrophoretic Assembly
Convective Assembly
Fast-Fluidic Assembly (this work) Any Substrate
Conductive
Substrate
Electrofluidic
Assembly
Fluidic Assembly
140
6.3.2. FFAsT Process
a) While the effect of surface tension has been explicitly explored, the effect of the ink
viscosity on the selectivity, morphology and thickness of the printed structures has not been
investigated and presents an opportunity for further research.
b) Modulation of ink properties to increase the drying speed while preserving clean edges and
uniform structures should be investigated.
c) Hydroxylation of the surface is required for putting SAMs, which may not be ideal for
some device structures. Alternative methods to decrease surface energy, both chemical and
physical (such as generation of nanoscale roughness) for nitrides and metals, are needed to
increase the versatility of the FFAsT printing process.
6.3.3. NanoOPS Transfer Printing
a) Chain mobility, which aids the transfer process, is also responsible for the volumetric
expansion that reduces pattern fidelity and affects alignment between multiple layers.
Some evidence suggests that the thickness of the polymer, and the fact whether it’s a
freestanding film or spincast on a rigid substrate, affects the volumetric expansion. Thinner
polymer layers, especially when spincoated on rigid substrates, minimize the volumetric
expansion, but an in-depth systematic study to investigate this observation is required.
b) When multiple layers are printed on the same substrate, the areas with overlap are prone to
damage (unless they are polymers or CNTs). Investigating the amount of stress different
types of particle structures can take before they are irreversibly damaged is required.
c) The functionalized Si template method is cost-effective and straightforward but needs to
be extended to the nanoscale regime to truly replace the damascene template.
141
6.3.4. End-of-Life Management of Nano-enabled Products
a) Methods to aid green design of nano-enabled products keeping in view their entire lifecycle
and possible exposures are needed. For this purpose, developing a well encompassing
attribute to be used in a multi-attribute utility method is needed. Exploring the change in
risk quotient (RQ) of a material upon its release to the environment should be explored as
a possible parameter.
References
1. Jeong, H. Development of Directed-Assembly Based Printing Process for Electronics,
Sensing, and Material Application. Northeastern University, 2017.
2. Brinker, C. J., Dip coating. In Chemical Solution Deposition of Functional Oxide Thin
Films, Springer: 2013; pp 233-261.
3. Rueckes, T.; Kim, K.; Joselevich, E.; Tseng, G. Y.; Cheung, C.-L.; Lieber, C. M., Carbon
Nanotube-Based Nonvolatile Random Access Memory for Molecular Computing. Science 2000,
289 (5476), 94.
4. Dimitrov, A. S.; Nagayama, K., Continuous Convective Assembling of Fine Particles
into Two-Dimensional Arrays on Solid Surfaces. Langmuir 1996, 12 (5), 1303-1311.
5. Chai, Z.; Abbasi, S. A.; Busnaina, A. A., Scalable Directed Assembly of Highly
Crystalline 2,7-Dioctyl[1]benzothieno[3,2-b][1]benzothiophene (C8-BTBT) Films. ACS Applied
Materials & Interfaces 2018, 10 (21), 18123-18130.
6. Denkov, N.; Velev, O.; Kralchevski, P.; Ivanov, I.; Yoshimura, H.; Nagayama, K.,
Mechanism of formation of two-dimensional crystals from latex particles on substrates.
Langmuir 1992, 8 (12), 3183-3190.
7. Hines, D. R.; Ballarotto, V. W.; Williams, E. D.; Shao, Y.; Solin, S. A., Transfer printing
methods for the fabrication of flexible organic electronics. Journal of Applied Physics 2007, 101
(2), 024503.
8. Heck, S.; Kaza, S.; Pinner, D., Creating value in the semiconductor industry. McKinsey &
Company 2011.
9. Williams, E. D.; Ayres, R. U.; Heller, M., The 1.7 Kilogram Microchip: Energy and
Material Use in the Production of Semiconductor Devices. Environmental Science & Technology
2002, 36 (24), 5504-5510.
10. Perelaer, J.; Smith, P. J.; Mager, D.; Soltman, D.; Volkman, S. K.; Subramanian, V.;
Korvink, J. G.; Schubert, U. S., Printed electronics: the challenges involved in printing devices,
interconnects, and contacts based on inorganic materials (Book review). Journal of Materials
Chemistry 2010, 20 (39), 8446.
11. Singh, M.; Haverinen, H. M.; Dhagat, P.; Jabbour, G. E., Inkjet Printing—Process and Its
Applications. Advanced Materials 2010, 22 (6), 673-685.
12. Szentgyörgyvölgyi, R., 12 - Gravure Printing. In Printing on Polymers, Izdebska, J.;
Thomas, S., Eds. William Andrew Publishing: 2016; pp 199-215.
13. Izdebska, J., 11 - Flexographic Printing. In Printing on Polymers, Izdebska, J.; Thomas,
S., Eds. William Andrew Publishing: 2016; pp 179-197.
14. Carlson, A.; Bowen, A. M.; Huang, Y.; Nuzzo, R. G.; Rogers, J. A., Transfer Printing
Techniques for Materials Assembly and Micro/Nanodevice Fabrication. Advanced Materials
2012, 24 (39), 5284-5318.
15. Kim, S.; Sojoudi, H.; Zhao, H.; Mariappan, D.; McKinley, G. H.; Gleason, K. K.; Hart,
A. J., Ultrathin high-resolution flexographic printing using nanoporous stamps. Science
Advances 2016, 2 (12), e1601660.
16. Berg, J. C., An introduction to interfaces & colloids : the bridge to nanoscience.
Singapore
143
Hackensack, N.J. : World Scientific: Singapore
Hackensack, N.J., 2010.
17. Armistead, C. G.; Tyler, A. J.; Hambleton, F. H.; Mitchell, S. A.; Hockey, J. A., Surface
hydroxylation of silica. The Journal of Physical Chemistry 1969, 73 (11), 3947-3953.
18. Jung, G.-Y.; Li, Z.; Wu, W.; Chen, Y.; Olynick, D. L.; Wang, S.-Y.; Tong, W. M.;
Williams, R. S., Vapor-Phase Self-Assembled Monolayer for Improved Mold Release in
Nanoimprint Lithography. Langmuir 2005, 21 (4), 1158-1161.
19. Hiemenz, P. C., Polymer chemistry : the basic concepts. New York : M. Dekker: New
York, 1984.
20. Yao, W.; Tian, Q.; Liu, J.; Wu, Z.; Cui, S.; Ding, J.; Dai, Z.; Wu, W., Large-scale
synthesis and screen printing of upconversion hexagonal-phase NaYF4:Yb3+,Tm3+/Er3+/Eu3+
plates for security applications. Journal of Materials Chemistry C 2016, 4 (26), 6327-6335.
21. Cho, H.; Somu, S.; Lee, J. Y.; Jeong, H.; Busnaina, A., High‐Rate Nanoscale Offset
Printing Process Using Directed Assembly and Transfer of Nanomaterials. Advanced Materials
2015, 27 (10), 1759-1766.
22. Zaumseil, J.; Meitl, M. A.; Hsu, J. W. P.; Acharya, B. R.; Baldwin, K. W.; Loo, Y.-L.;
Rogers, J. A., Three-Dimensional and Multilayer Nanostructures Formed by Nanotransfer
Printing. Nano Letters 2003, 3 (9), 1223-1227.
23. Xin, Z.; Liu, Y.; Li, X.; Liu, S.; Fang, Y.; Deng, Y.; Bao, C.; Li, L., Conductive grid
patterns prepared by microcontact printing silver nanoparticles ink. Materials Research Express
2017, 4 (1), 015021.
24. Kumar, A.; Whitesides, G. M., Features of gold having micrometer to centimeter
dimensions can be formed through a combination of stamping with an elastomeric stamp and an
alkanethiol ‘‘ink’’ followed by chemical etching. Applied Physics Letters 1993, 63 (14), 2002-
2004.
25. Loo, Y.-L.; Willett, R. L.; Baldwin, K. W.; Rogers, J. A., Interfacial Chemistries for
Nanoscale Transfer Printing. Journal of the American Chemical Society 2002, 124 (26), 7654-
7655.
26. Choi, K. M.; Rogers, J. A., A Photocurable Poly(dimethylsiloxane) Chemistry Designed
for Soft Lithographic Molding and Printing in the Nanometer Regime. Journal of the American
Chemical Society 2003, 125 (14), 4060-4061.
27. Dimitrov, Continuous Convective Assembling of Fine Particles into Two-Dimensional
Arrays on Solid Surfaces. Langmuir : 1996, 12 (5), 1303-1311.
28. Jaber-Ansari, L.; Hahm, M.; Kim, T.; Somu, S.; Busnaina, A.; Jung, Y., Large scale
highly organized single-walled carbon nanotube networks for electrical devices. Materials
Science & Processing 2009, 96 (2), 373-377.
29. Xiong, X.; Jaberansari, L.; Busnaina, A.; Jung, Y. J.; Somu, S.; Upmanyu, M. Highly
Organized Single-Walled Carbon Nanotube Networks and Method of Making Using Template
Guided Fluidic Assembly. US2010183844 (A1), 2010/07/22/, 2010.
30. Sujanani, M.; Wayner, P. C., TRANSPORT PROCESSES AND INTERFACIAL
PHENOMENA IN AN EVAPORATING MENISCUS. Chem. Eng. Commun. 1992, 118, 89-
110.
144
31. Dushkin, C. D.; Yoshimura, H.; Nagayama, K., Nucleation and growth of two-
dimensional colloidal crystals. Chemical Physics Letters 1993, 204 (5), 455-460.
32. Kralchevsky, P. A.; Nagayama, K., Capillary forces between colloidal particles.
Langmuir 1994, 10 (1), 23-36.
33. Park, S.; Xia, Y., Assembly of mesoscale particles over large areas and its application in
fabricating tunable optical filters. Langmuir 1999, 15 (1), 266-273.
34. Fustin, C.-A.; Glasser, G.; Spiess, H. W.; Jonas, U., Parameters influencing the templated
growth of colloidal crystals on chemically patterned surfaces. Langmuir : the ACS journal of
surfaces and colloids 2004, 20 (21), 9114.
35. Xiong, X.; Makaram, P.; Busnaina, A.; Bakhtari, K.; Somu, S.; McGruer, N.; Park, J.,
Large scale directed assembly of nanoparticles using nanotrench templates. Applied Physics
Letters 2006, 89 (19).
36. Li, B.; Jung, H. Y.; Wang, H.; Kim, Y. L.; Kim, T.; Hahm, M. G.; Busnaina, A.;
Upmanyu, M.; Jung, Y. J., Ultrathin SWNT Films with Tunable, Anisotropic Transport
Properties. Advanced Functional Materials 2011, 21 (10), 1810-1815.
37. d’Agostino, R.; Flamm, D. L., Plasma etching of Si and SiO2 in SF6–O2 mixtures.
Journal of Applied Physics 1981, 52 (1), 162-167.
38. Gräf, D.; Grundner, M.; Schulz, R.; Mühlhoff, L., Oxidation of HF‐treated Si wafer
surfaces in air. Journal of Applied Physics 1990, 68 (10), 5155-5161.
39. Born, P.; Blum, S.; Munoz, A.; Kraus, T., Role of the Meniscus Shape in Large-Area
Convective Particle Assembly. Langmuir 2011, 27 (14), 8621-8633.
40. Malaquin, L.; Kraus, T.; Schmid, H.; Delamarche, E.; Wolf, H., Controlled Particle
Placement through Convective and Capillary Assembly. Langmuir 2007, 23 (23), 11513-11521.
41. Su, G.; Guo, Q.; Palmer, R. E., Colloidal Lines and Strings. Langmuir 2003, 19 (23),
9669-9671.
42. Berthier, J., Chapter 3 - The Physics of Droplets**This chapter was written with the
collaboration of Kenneth A. Brakke (Mathematics Department, Susquehanna University,
Selinsgrove, PA). In Micro-Drops and Digital Microfluidics (Second Edition), Berthier, J., Ed.
William Andrew Publishing: 2013; pp 75-160.
43. Troian, S. M.; Wu, X. L.; Safran, S. A., Fingering instability in thin wetting films.
Physical Review Letters 1989, 62 (13), 1496-1499.
44. Mack, C. A., Field guide to optical lithography. SPIE Press Bellingham, WA: 2006; Vol.
6.
45. Bickford, J. R. Analyze_Stripes Documentation.
https://imagejdocu.tudor.lu/doku.php?id=macro:analyze_stripes.
46. Villarrubia, J. S. In Issues in line edge and linewidth roughness metrology, AIP
Conference Proceedings, AIP: 2005; pp 386-393.
47. Frey, M. H.; Zu, L.; Hagermoser, E. S., Touch screen sensor. Google Patents: 2012.
48. Liu, Y.; Shen, S.; Hu, J.; Chen, L., Embedded Ag mesh electrodes for polymer dispersed
liquid crystal devices on flexible substrate. Optics express 2016, 24 (22), 25774-25784.
49. Li, Y.; Mao, L.; Gao, Y.; Zhang, P.; Li, C.; Ma, C.; Tu, Y.; Cui, Z.; Chen, L., ITO-free
photovoltaic cell utilizing a high-resolution silver grid current collecting layer.(Report). Solar
Energy Materials and Solar Cells 2013, 113, 85.
145
50. Park, J.; Lee, J.; Noh, Y.-Y., Optical and thermal properties of large-area OLED lightings
with metallic grids. Organic Electronics 2012, 13 (1), 184-194.
51. Misra, S., New report finds that more than 165,000 mobile health apps now available,
takes close look at characteristics & use: iMedicalApps; 2015 [updated September 17, 2015.
URL: http://www. imedicalapps. com/2015/09/ims-health-apps-report/. Accessed 2015, 8 (20),
16.
52. van de Groep, J.; Spinelli, P.; Polman, A., Transparent conducting silver nanowire
networks. Nano letters 2012, 12 (6), 3138.
53. Ghosh, D. S.; Chen, T. L., High figure-of-merit ultrathin metal transparent electrodes
incorporating a conductive grid. Applied Physics Letters 2010, 96 (4).
54. Choi, H.-J.; Choo, S.; Jung, P.-H.; Shin, J.-H.; Kim, Y.-D.; Lee, H., Uniformly embedded
silver nanomesh as highly bendable transparent conducting electrode. Nanotechnology 2015, 26
(5), 055305.
55. Rajesh Desai, P.; Nikhil Desai, P.; Deepak Ajmera, K.; Mehta, K., A Review Paper on
Oculus Rift-A Virtual Reality Headset. International Journal of Engineering Trends and
Technology 2014, 13 (4), 175-179.
56. Kahng, Y. H.; Kim, M.-K.; Lee, J.-H.; Kim, Y. J.; Kim, N.; Park, D.-W.; Lee, K., Highly
conductive flexible transparent electrodes fabricated by combining graphene films and inkjet-
printed silver grids. Solar Energy Materials and Solar Cells 2014, 124 (C), 86-91.
57. Park, J. H.; Lee, D. Y.; Kim, Y.-H.; Kim, J. K.; Lee, J. H.; Park, J. H.; Lee, T.-W.; Cho,
J. H., Flexible and Transparent Metallic Grid Electrodes Prepared by Evaporative Assembly.
ACS Applied Materials & Interfaces 2014, 6 (15), 12380-12387.
58. Huang, Y.; Bai, X.; Zhou, M.; Liao, S.; Yu, Z.; Wang, Y.; Wu, H., Large-Scale Spinning
of Silver Nanofibers as Flexible and Reliable Conductors. Nano Letters 2016, 16 (9), 5846-5851.
59. Hu, L.; Kim, H. S.; Lee, J.-Y.; Peumans, P.; Cui, Y., Scalable coating and properties of
transparent, flexible, silver nanowire electrodes. ACS nano 2010, 4 (5), 2955.
60. Dong Jin, K.; Mikyung, L.; Seung, S. L.; Jin-Ha, K., High aspect ratio silver grid
transparent electrodes using UV embossing process. AIP Advances 2017, 7 (10), 105218-
105218-7.
61. Zhao, Y.-Y.; Zheng, M.-L.; Dong, X.-Z.; Jin, F.; Liu, J.; Ren, X.-L.; Duan, X.-M.; Zhao,
Z.-S., Tailored silver grid as transparent electrodes directly written by femtosecond laser.
Applied Physics Letters 2016, 108 (22).
62. De, S.; Higgins, T. M.; Lyons, P. E.; Doherty, E. M.; Nirmalraj, P. N.; Blau, W. J.;
Boland, J. J.; Coleman, J. N., Silver Nanowire Networks as Flexible, Transparent, Conducting
Films: Extremely High DC to Optical Conductivity Ratios. ACS Nano 2009, 3 (7), 1767-1774.
63. Jang, Y.; Kim, J.; Byun, D., Invisible metal-grid transparent electrode prepared by
electrohydrodynamic (ehd) jet printing. Journal of Physics D: Applied Physics 2013, 46 (15),
155103.
64. Hjerrild, N. E.; Neo, D. C. J.; Kasdi, A.; Assender, H. E.; Warner, J. H.; Watt, A. A. R.,
Transfer Printed Silver Nanowire Transparent Conductors for PbS–ZnO Heterojunction
Quantum Dot Solar Cells. ACS Applied Materials & Interfaces 2015, 7 (12), 6417-6421.
146
65. Falinski, M. M.; Plata, D. L.; Chopra, S. S.; Theis, T. L.; Gilbertson, L. M.; Zimmerman,
J. B., A framework for sustainable nanomaterial selection and design based on performance,
hazard, and economic considerations. Nature nanotechnology 2018, 13 (8), 708.
66. Yang, C.; Kim, J., Embedding of inkjet-printed Ag-grid/ITO hybrid transparent electrode
into a plastic substrate for flexible electronic devices. Published by the Indian Academy of
Sciences 2018, 41 (1), 1-5.
67. Syms, R. R. A.; Yeatman, E. M.; Bright, V. M.; Whitesides, G. M., Surface tension-
powered self-assembly of microstructures - the state-of-the-art. Journal of
Microelectromechanical Systems 2003, 12 (4), 387-417.
68. Ng, J. M. K.; Fuerstman, M. J.; Grzybowski, B. A.; Stone, H. A.; Whitesides, G. M.,
Self-Assembly of Gears at a Fluid/Air Interface. Journal of the American Chemical Society
2003, 125 (26), 7948-7958.
69. Ni, S.; Isa, L.; Wolf, H., Capillary assembly as a tool for the heterogeneous integration of
micro- and nanoscale objects. Soft Matter 2018, 14 (16), 2978-2995.
70. Roland, S.; Gamys, C. G.; Grosrenaud, J.; Boissé, S.; Pellerin, C.; Prud’homme, R. E.;
Bazuin, C. G., Solvent Influence on Thickness, Composition, and Morphology Variation with
Dip-Coating Rate in Supramolecular PS-b-P4VP Thin Films. Macromolecules 2015, 48 (14),
4823-4834.
71. Cisneros-Zevallos, L.; Krochta, J., Dependence of coating thickness on viscosity of
coating solution applied to fruits and vegetables by dipping method. J. Food Sci. 2003, 68 (2),
503-510.
72. Willeke, G.; Weber, E. R., Advances in Photovoltaics. Part 3. Amsterdam : Academic
Press: 2014.
73. Piegari, A.; Flory, F. o., Optical thin films and coatings : from materials to applications.
Philadelphia, PA : Woodhead Pub.: 2013.
74. Scott, C., Misconceptions about Aerobic and Anaerobic Energy Expenditure. Journal of
the International Society of Sports Nutrition 2005, 2 (2), 32.
75. Scott, B. R.; Goods, P. S. R.; Slattery, K. M., High-Intensity Exercise in Hypoxia: Is
Increased Reliance on Anaerobic Metabolism Important? Frontiers in Physiology 2016, 7.
76. Meyerhoff, C.; Bischof, F.; Mennel, F. J.; Sternberg, F.; Bican, J.; Pfeiffer, E. F., On line
continuous monitoring of blood lactate in men by a wearable device based upon an enzymatic
amperometric lactate sensor. Biosensors and Bioelectronics 1993, 8 (9), 409-414.
77. Sakharov, D. A.; Shkurnikov, M. U.; Vagin, M. Y.; Yashina, E. I.; Karyakin, A. A.;
Tonevitsky, A. G., Relationship between Lactate Concentrations in Active Muscle Sweat and
Whole Blood. Bulletin of Experimental Biology and Medicine 2010, 150 (1), 83-85.
78. Haeringen, N.; Glasius, E., Collection method dependant concentrations of some
metabolites in human tear fluid, with special reference to glucose in hyperglycaemic conditions.
Albrecht von Graefes Archiv für klinische und experimentelle Ophthalmologie 1977, 202 (1), 1-
7.
79. Tékus, E.; Kaj, M.; Szabó, E.; Szénási, N. L.; Kerepesi, I.; Figler, M.; Gábriel, R.;
Wilhelm, M., Comparison of blood and saliva lactate level after maximum intensity exercise.
Acta biologica Hungarica 2012, 63 Suppl 1 (1), 89.
147
80. Zhang, W.; Du, Y.; Wang, M. L., Noninvasive glucose monitoring using saliva nano-
biosensor. Sensing and Bio-Sensing Research 2015, 4, 23-29.
81. Schabmueller, C. G. J.; Loppow, D.; Piechotta, G.; Schütze, B.; Albers, J.; Hintsche, R.,
Micromachined sensor for lactate monitoring in saliva. Biosensors and Bioelectronics 2006, 21
(9), 1770-1776.
82. Thomas, N.; Lähdesmäki, I.; Parviz, B. A., A contact lens with an integrated lactate
sensor. Sensors and Actuators B: Chemical 2012, 162 (1), 128-134.
83. Abrar, M. A.; Dong, Y.; Lee, P. K.; Kim, W. S., Bendable Electro-chemical Lactate
Sensor Printed with Silver Nano-particles. Scientific reports 2016, 6 (1), 30565.
84. Jia, W.; Bandodkar, A. J.; Valdés-Ramírez, G.; Windmiller, J. R.; Yang, Z.; Ramírez, J.;
Chan, G.; Wang, J., Electrochemical tattoo biosensors for real-time noninvasive lactate
monitoring in human perspiration. Analytical chemistry 2013, 85 (14), 6553.
85. Tsai, Y.-C.; Chen, S.-Y.; Liaw, H.-W., Immobilization of lactate dehydrogenase within
multiwalled carbon nanotube-chitosan nanocomposite for application to lactate biosensors.
Sensors and Actuators B: Chemical 2007, 125 (2), 474-481.
86. Pereira, A. C.; Aguiar, M. R.; Kisner, A.; Macedo, D. V.; Kubota, L. T., Amperometric
biosensor for lactate based on lactate dehydrogenase and Meldola Blue coimmobilized on multi-
wall carbon-nanotube. Sensors and Actuators B: Chemical 2007, 124 (1), 269-276.
87. Labroo, P.; Cui, Y., Flexible graphene bio-nanosensor for lactate. Biosensors and
Bioelectronics 2013, 41, 852-856.
88. Shao, Y.; Wang, J.; Wu, H.; Liu, J.; Aksay, I. A.; Lin, Y., Graphene Based
Electrochemical Sensors and Biosensors: A Review. Electroanalysis, 22:1027-1036 2010, 22
(10).
89. Chen, Q.; Sun, T.; Song, X.; Ran, Q.; Yu, C.; Yang, J.; Feng, H.; Yu, L.; Wei, D.,
Flexible electrochemical biosensors based on graphene nanowalls for the real-time measurement
of lactate. Nanotechnology 2017, 28 (31), 315501.
90. Nesakumar, N.; Sethuraman, S.; Krishnan, U. M.; Rayappan, J. B. B., Fabrication of
lactate biosensor based on lactate dehydrogenase immobilized on cerium oxide nanoparticles.
Journal of Colloid and Interface Science 2013, 410, 158-164.
91. Rathee, K.; Dhull, V.; Dhull, R.; Singh, S., Biosensors based on electrochemical lactate
detection: A comprehensive review. Biochemistry and Biophysics Reports 2016, 5 (C), 35-54.
92. Hickey, D. P.; Reid, R. C.; Milton, R. D.; Minteer, S. D., A self-powered amperometric
lactate biosensor based on lactate oxidase immobilized in dimethylferrocene-modified LPEI.
Biosensors and Bioelectronics 2016, 77, 26-31.
93. Pandey, S.; Nanda, K. K., Au Nanocomposite Based Chemiresistive Ammonia Sensor for
Health Monitoring. ACS Sensors 2016, 1 (1), 55-62.
94. Soylemez, S.; Yoon, B.; Toppare, L.; Swager, T. M., Quaternized Polymer–Single-
Walled Carbon Nanotube Scaffolds for a Chemiresistive Glucose Sensor. ACS Sensors 2017, 2
(8), 1123-1127.
95. Ruixian, T.; Yongji, S.; Zhongyu, H.; Liangming, W., Carbon Nanotube-Based
Chemiresistive Sensors. Sensors 2017, 17 (4), 882.
96. Li, A.; Bogdanovich, A. E.; Bradford, P. D., Aligned carbon nanotube sheet
piezoresistive strain sensors. Smart Materials and Structures 2015, 24 (9), 095004.
148
97. Wilson, H.; Ripp, S.; Prisbrey, L.; Brown, M. A.; Sharf, T.; Myles, D. J. T.; Blank, K. G.;
Minot, E. D., Electrical Monitoring of sp 3 Defect Formation in Individual Carbon Nanotubes.
The Journal of Physical Chemistry C 2016, 120 (3), 1971-1976.
98. Wang, F.; Swager, T. M., Diverse Chemiresistors Based upon Covalently Modified
Multiwalled Carbon Nanotubes. Journal of the American Chemical Society 2011, 133 (29),
11181-11193.
99. Mockus, L.; Peterson, J. J.; Lainez, J. M.; Reklaitis, G. V., Batch-to-Batch Variation: A
Key Component for Modeling Chemical Manufacturing Processes. Organic Process Research
& Development 2015, 19 (8), 908-914.
100. Russell, L. J.; Rawson, K. M., The commercialisation of sensor technology in clinical
chemistry: an outline of the potential difficulties. Biosensors 1986, 2 (5), 301-318.
101. Fu, L.; Yu, A. M., Carbon nanotube based thin films: fabrication, characterization and
applications. Reviews on Advanced Materials Science 2014, 36 (1), 40-61.
102. Torrey, J. D.; Kirschling, T. L.; Greenlee, L. F., Processing and Characterization of
Nanoparticle Coatings for Quartz Crystal Microbalance Measurements. Journal of Research of
the National Institute of Standards and Technology 2015, 120, 1-10.
103. Ngo, Y. H.; Brothers, M.; Martin, J. A.; Grigsby, C. C.; Fullerton, K.; Naik, R. R.; Kim,
S. S., Chemically Enhanced Polymer-Coated Carbon Nanotube Electronic Gas Sensor for
Isopropyl Alcohol Detection. ACS Omega 2018, 3 (6), 6230-6236.
104. Green, J. M.; Pritchett, R. C.; Crews, T. R.; McLester, J. R.; Tucker, D. C., Sweat lactate
response between males with high and low aerobic fitness. European Journal of Applied
Physiology 2004, 91 (1), 1-6.
105. Xiong, X.; Jaberansari, L.; Hahm Myung, G.; Busnaina, A.; Jung Yung, J., Building
Highly Organized Single‐Walled‐Carbon‐Nanotube Networks Using Template‐Guided Fluidic
Assembly. Small 2007, 3 (12), 2006-2010.
106. Jérôme, C.; Geskin, V.; Lazzaroni, R.; Brédas, J. L.; Thibaut, A.; Calberg, C.; Bodart, I.;
Mertens, M.; Martinot, L.; Rodrigue, D.; Riga, J.; Jérôme, R., Full-Electrochemical Preparation
of Conducting/Insulating Binary Polymer Films. Chemistry of Materials 2001, 13 (5), 1656-
1664.
107. Sadki, S.; Schottland, P.; Brodie, N.; Sabouraud, G., The mechanisms of pyrrole
electropolymerization. Chemical Society Reviews 2000, 29 (5), 283-293.
108. Lee, J.; Jeong, H.; Lassarote Lavall, R.; Busnaina, A.; Kim, Y.; Jung, Y. J.; Lee, H.,
Polypyrrole Films with Micro/Nanosphere Shapes for Electrodes of High-Performance
Supercapacitors. ACS applied materials & interfaces 2017, 9 (38), 33203.
109. Abdullah, H. S., Electrochemical polymerization and Raman study of polypyrrole and
polyaniline thin films. International Journal of Physical Sciences 2012, 7 (38), 5468-5476.
110. Nguyen Thi Le, H.; Bernard, M. C.; Garcia-Renaud, B.; Deslouis, C., Raman
spectroscopy analysis of polypyrrole films as protective coatings on iron. Synthetic Metals 2004,
140 (2), 287-293.
111. Valeur, E.; Bradley, M., Amide bond formation: beyond the myth of coupling reagents.
Chemical Society Reviews 2009, 38 (2), 606-631.
149
112. Pagán, M.; Suazo, D.; del Toro, N.; Griebenow, K., A comparative study of different
protein immobilization methods for the construction of an efficient nano-structured lactate
oxidase-SWCNT-biosensor. Biosensors and Bioelectronics 2015, 64, 138-146.
113. Gaia, R.; Angela, S.; Sergio, B.; Gavinella, L.; Giordano, M.; Grazia, G.; Susanna, N.;
Paola, B.; Maria Ilaria, D.; Vito, F.; Roberto, M.; Pier Andrea, S., Analytical Problems in
Exposing Amperometric Enzyme Biosensors to Biological Fluids. Sensors 2016, 16 (6), 780.
114. Sonawane, A.; Manickam, P.; Bhansali, S., Stability of Enzymatic Biosensors for
Wearable Applications. IEEE Reviews in Biomedical Engineering 2017, 10, 174-186.
115. Lyons, P. E.; De, S.; Blighe, F.; Nicolosi, V.; Pereira, L. F. C.; Ferreira, M. S.; Coleman,
J. N., The relationship between network morphology and conductivity in nanotube films. Journal
of Applied Physics 2008, 104 (4), 044302.
116. Meshram, B. H.; Mahore, R. P.; Virutkar, P. D.; Kondawar, S. B., Polypyrrole/Carbon
Nanotubes/Lactate Oxidase Nanobiocomposite Film based Modified Stainless Steel Electrode
Lactate biosensor. Procedia Materials Science 2015, 10, 176-185.
117. Abbasi, S. A.; Chai, Z.; Busnaina, A., Scalable Printing of High-Resolution Flexible
Transparent Grid Electrodes Using Directed Assembly of Silver Nanoparticles. Advanced
Materials Interfaces 2019, 0 (0), 1900898.
118. Kumar, A.; Biebuyck, H. A.; Whitesides, G. M., Patterning Self-Assembled Monolayers:
Applications in Materials Science. Langmuir 1994, 10 (5), 1498-1511.
119. Biebuyck, H. A.; Whitesides, G. M., Self-Organization of Organic Liquids on Patterned
Self-Assembled Monolayers of Alkanethiolates on Gold. Langmuir 1994, 10 (8), 2790-2793.
120. Petrov, J. G.; Ralston, J.; Hayes, R. A., Dewetting Dynamics on Heterogeneous Surfaces.
A Molecular−Kinetic Treatment. Langmuir 1999, 15 (9), 3365-3373.
121. Konnur; Kargupta; Sharma, Instability and morphology of thin liquid films on chemically
heterogeneous substrates. Physical review letters 2000, 84 (5), 931.
122. Kargupta, K.; Konnur, R.; Sharma, A., Instability and Pattern Formation in Thin Liquid
Films on Chemically Heterogeneous Substrates. Langmuir 2000, 16 (26), 10243-10253.
123. Wang, J. Z.; Zheng, Z. H.; Li, H. W.; Huck, W. T. S.; Sirringhaus, H., Dewetting of
conducting polymer inkjet droplets on patterned surfaces. Nature Materials 2004, 3 (3), 171.
124. Lee, K.-H.; Choi, B.-Y.; Park, J.-W.; Kang, S.-J.; Kim, S.-M.; Kim, D.-Y.; Jung, G.-Y.,
Solution processable micron- to nanoscale conducting polymer patterning utilizing selective
surface energy engineering. Organic Electronics 2010, 11 (5), 748-754.
125. Lee, K.-H.; Kim, S.-M.; Jeong, H.; Jung, G.-Y., Spontaneous nanoscale polymer solution
patterning using solvent evaporation driven double-dewetting edge lithography. Soft Matter
2012, 8 (2), 465-471.
126. Landau, L.; Levich, B., Dragging of a Liquid by a Moving Plate. In Dynamics of Curved
Fronts, Pelcé, P., Ed. Academic Press: San Diego, 1988; pp 141-153.
127. Darhuber, A. A.; Troian, S. M.; Davis, J. M.; Miller, S. M.; Wagner, S., Selective dip-
coating of chemically micropatterned surfaces. Journal of Applied Physics 2000, 88 (9), 5119-
5126.
128. Darhuber, A. A.; Troian, S. M.; Miller, S. M.; Wagner, S., Morphology of liquid
microstructures on chemically patterned surfaces. Journal of Applied Physics 2000, 87 (11),
7768-7775.
150
129. Ko, H.; Peleshanko, S.; Tsukruk, V. V., Combing and Bending of Carbon Nanotube
Arrays with Confined Microfluidic Flow on Patterned Surfaces. The Journal of Physical
Chemistry B 2004, 108 (14), 4385-4393.
130. de Gans, B.-J.; Schubert, U. S., Inkjet Printing of Well-Defined Polymer Dots and
Arrays. Langmuir 2004, 20 (18), 7789-7793.
131. Soltman, D.; Subramanian, V., Inkjet-Printed Line Morphologies and Temperature
Control of the Coffee Ring Effect. Langmuir 2008, 24 (5), 2224-2231.
132. Tucker, M. B.; Hines, D. R.; Li, T., A quality map of transfer printing. Journal of Applied
Physics 2009, 106 (10).
133. Brenckle, M. A.; Kaplan, D. L.; Omenetto, F. G., Direct Transfer Printing of Water
Hydrolyzable Metals onto Silk Fibroin Substrates through Thermal-Reflow-Based Adhesion.
Advanced Materials Interfaces 2016, 3 (15), 1600094.
134. Liu, D.; Guo, J.; Zhang, J.-H., Chain mobility and film softness mediated protein
antifouling at the solid–liquid interface. Journal of Materials Chemistry B 2016, 4 (36), 6134-
6142.
135. Kojio, K.; Furukawa, M.; Matsumura, S.; Motokucho, S.; Osajima, T.; Yoshinaga, K.,
The effect of cross-linking density and dangling chains on surface molecular mobility of network
polyurethanes. Polymer Chemistry 2012, 3 (8), 2287-2292.
136. Domanskyi, S.; Gentekos, D. T.; Privman, V.; Fors, B. P., Predictive design of polymer
molecular weight distributions in anionic polymerization. Polymer Chemistry 2019.
137. Nielsen, L. E., Cross-Linking–Effect on Physical Properties of Polymers. Journal of
Macromolecular Science, Part C 1969, 3 (1), 69-103.
138. Chiou, B. S.; Schoen, P. E., Effects of crosslinking on thermal and mechanical properties
of polyurethanes. Journal of Applied Polymer Science 2002, 83 (1), 212-223.
139. Saleesung, T.; Reichert, D.; Saalwächter, K.; Sirisinha, C., Correlation of crosslink
densities using solid state NMR and conventional techniques in peroxide-crosslinked EPDM
rubber. Polymer 2015, 56, 309-317.
140. McKeen, L., 2 - Introduction to Plastics and Polymers. In The Effect of Sterilization on
Plastics and Elastomers (Third Edition), McKeen, L., Ed. William Andrew Publishing: Boston,
2012; pp 41-56.
141. Gradin, P.; Howgate, P. G.; Seldén, R.; Brown, R. A., 16 - Dynamic-mechanical
Properties. In Comprehensive Polymer Science and Supplements, Allen, G.; Bevington, J. C.,
Eds. Pergamon: Amsterdam, 1989; pp 533-569.
142. Tillman, N.; Ulman, A.; Schildkraut, J. S.; Penner, T. L., Incorporation of phenoxy
groups in self-assembled monolayers of trichlorosilane derivatives. Effects on film thickness,
wettability, and molecular orientation. Journal of the American Chemical Society 1988, 110 (18),
6136-6144.
143. Zhan, Y.; Mei, Y.; Zheng, L., Materials capability and device performance in flexible
electronics for the Internet of Things. J. Mater. Chem. C 2014, 2 (7), 1220-1232.
144. Perelaer, J.; Smith, P. J.; Mager, D.; Soltman, D.; Volkman, S. K.; Subramanian, V.;
Korvink, J. G.; Schubert, U. S., Printed electronics: the challenges involved in printing devices,
interconnects, and contacts based on inorganic materials. Journal of Materials Chemistry 2010,
20 (39), 8446-8453.
151
145. Chai, Z.; Jeong, H.; Abbasi, S. A.; Busnaina, A. A., Fabrication of organic field effect
transistors using directed assembled and transfer printed carbon nanotube source/drain
electrodes. Applied Physics Letters 2019, 114 (10).
146. Fan, X.; Nie, W.; Tsai, H.; Wang, N.; Huang, H.; Cheng, Y.; Wen, R.; Ma, L.; Yan, F.;
Xia, Y., PEDOT:PSS for Flexible and Stretchable Electronics: Modifications, Strategies, and
Applications. Advanced Science 2019, 6 (19), 1900813.
147. Hecht, D. S.; Hu, L.; Irvin, G., Emerging Transparent Electrodes Based on Thin Films of
Carbon Nanotubes, Graphene, and Metallic Nanostructures. Weinheim, 2011; Vol. 23, pp 1482-
1513.
148. Fisichella, G.; Lo Verso, S.; Di Marco, S.; Vinciguerra, V.; Schilirò, E.; Di Franco, S.; Lo
Nigro, R.; Roccaforte, F.; Zurutuza, A.; Centeno, A.; Ravesi, S.; Giannazzo, F., Advances in the
fabrication of graphene transistors on flexible substrates. Beilstein J Nanotechnol 2017, 8, 467-
474.
149. Scholz, S.; Meerheim, R.; Walzer, K.; Leo, K., Chemical degradation mechanisms of
organic semiconductor devices. Proc SPIE 2008, 6999.
150. Le, T.-H.; Kim, Y.; Yoon, H., Electrical and Electrochemical Properties of Conducting
Polymers. Polymers (Basel) 2017, 9 (4), 150.
151. Cai, L.; Wang, C., Carbon Nanotube Flexible and Stretchable Electronics. Nanoscale Res
Lett 2015, 10 (1), 1013-1013.
152. Sun, D. M.; Liu, C.; Ren, W. C.; Cheng, H. M., All‐Carbon Thin‐Film Transistors as a
Step Towards Flexible and Transparent Electronics. Advanced Electronic Materials 2016, 2 (11),
n/a-n/a.
153. Aikawa, S.; Einarsson, E.; Thurakitseree, T.; Chiashi, S.; Nishikawa, E.; Maruyama, S.,
Deformable transparent all-carbon-nanotube transistors. Applied Physics Letters 2012, 100 (6),
063502.
154. Tseng, S. H.; Tai, N. H., Fabrication of a transparent and flexible thin film transistor
based on single-walled carbon nanotubes using the direct transfer method. Applied Physics
Letters 2009, 95 (20).
155. Lipomi, D. J.; Vosgueritchian, M.; Tee, B. C. K.; Hellstrom, S. L.; Lee, J. A.; Fox, C. H.;
Bao, Z., Skin-like pressure and strain sensors based on transparent elastic films of carbon
nanotubes. Nature nanotechnology 2011, 6 (12), 788.
156. Cao, C.; Andrews, J. B.; Franklin, A. D., Completely Printed, Flexible, Stable, and
Hysteresis-Free Carbon Nanotube Thin-Film Transistors via Aerosol Jet Printing. Advanced
Electronic Materials 2017, 3 (5), 1700057.
157. Dongil, L.; Jinsu, Y.; Juhee, L.; Byung-Hyun, L.; Myeong-Lok, S.; Hagyoul, B.; Seung-
Bae, J.; Hyejeong, S.; Sung Gap, I.; Sung-Jin, C.; Yang-Kyu, C., Logic circuits composed of
flexible carbon nanotube thin-film transistor and ultra-thin polymer gate dielectric. Scientific
Reports 2016, 6 (1).
158. Rao, R.; Pint, C. L.; Islam, A. E.; Weatherup, R. S.; Hofmann, S.; Meshot, E. R.; Wu, F.;
Zhou, C.; Dee, N.; Amama, P. B.; Carpena-Nuñez, J.; Shi, W.; Plata, D. L.; Penev, E. S.;
Yakobson, B. I.; Balbuena, P. B.; Bichara, C.; Futaba, D. N.; Noda, S.; Shin, H.; Kim, K. S.;
Simard, B.; Mirri, F.; Pasquali, M.; Fornasiero, F.; Kauppinen, E. I.; Arnold, M.; Cola, B. A.;
Nikolaev, P.; Arepalli, S.; Cheng, H.-M.; Zakharov, D. N.; Stach, E. A.; Zhang, J.; Wei, F.;
152
Terrones, M.; Geohegan, D. B.; Maruyama, B.; Maruyama, S.; Li, Y.; Adams, W. W.; Hart, A.
J., Carbon Nanotubes and Related Nanomaterials: Critical Advances and Challenges for
Synthesis toward Mainstream Commercial Applications. ACS Nano 2018, 12 (12), 11756-11784.
159. Zhang, S.; Pelligra, C. I.; Feng, X.; Osuji, C. O., Directed Assembly of Hybrid
Nanomaterials and Nanocomposites. Advanced Materials 2018, 30 (18), 1705794.
160. Somu, S.; Wang, H.; Kim, Y.; Jaberansari, L.; Hahm, M. G.; Li, B.; Kim, T.; Xiong, X.;
Jung, Y. J.; Upmanyu, M.; Busnaina, A., Topological Transitions in Carbon Nanotube Networks
via Nanoscale Confinement. ACS Nano 2010, 4 (7), 4142-4148.
161. Yilmaz, C.; Sirman, A.; Halder, A.; Busnaina, A., High-Rate Assembly of Nanomaterials
on Insulating Surfaces Using Electro-Fluidic Directed Assembly. ACS Nano 2017, 11 (8), 7679-
7689.
162. Gomes, H., Organic Field-Effect Transistors. 2016; pp 147-197.
163. Chortos, A.; Koleilat, G. I.; Pfattner, R.; Kong, D.; Lin, P.; Nur, R.; Lei, T.; Wang, H.;
Liu, N.; Lai, Y.-C.; Kim, M.-G.; Chung, J. W.; Lee, S.; Bao, Z., Mechanically Durable and
Highly Stretchable Transistors Employing Carbon Nanotube Semiconductor and Electrodes.
Advanced Materials 2016, 28 (22), 4441-4448.
164. Healy, M. L.; Dahlben, L. J.; Isaacs, J. A., Environmental Assessment of Single‐Walled
Carbon Nanotube Processes. Journal of Industrial Ecology 2008, 12 (3), 376-393.
165. Savage, N. F.; ProQuest, Nanotechnology applications for clean water. Norwich, NY :
William Andrew: Norwich, NY, 2009.
166. Abbasi, S., Exergetic Life Cycle Assessment of Electrospun Polyvinylidene Fluoride
Nanofibers. Durham, D.; Thomas, S.; Volinsky, A., Eds. ProQuest Dissertations Publishing:
2014.
167. Osterwalder, N.; Capello, C.; Hungerbühler, K.; Stark, W., Energy Consumption During
Nanoparticle Production: How Economic is Dry Synthesis? An Interdisciplinary Forum for
Nanoscale Science and Technology 2006, 8 (1), 1-9.
168. Syberg, K.; Hansen, S. F., Environmental risk assessment of chemicals and
nanomaterials--The best foundation for regulatory decision-making? Sci Total Environ 2016,
541, 784-94.
169. Gottschalk, F.; Sonderer, T.; Scholz, R. W.; Nowack, B., Possibilities and limitations of
modeling environmental exposure to engineered nanomaterials by probabilistic material flow
analysis. Environ Toxicol Chem 2010, 29 (5), 1036-48.
170. Strand, R.; Kjlberg, K., Regulating Nanoparticles: the Problem of Uncertainty. European
Journal of Law and Technology 2011, 2 (3).
171. Huijbregts, M. A. J.; Hellweg, S.; Frischknecht, R.; Hendriks, H. W. M.; Hungerbühler,
K.; Hendriks, A. J., Cumulative energy demand as predictor for the environmental burden of
commodity production. Environmental science & technology 2010, 44 (6), 2189.
172. Pimbley, J. M., Advanced CMOS process technology. San Diego : Academic Press: San
Diego, 1989.
173. Bohr, M.; Mistry, K., Intel’s revolutionary 22 nm transistor technology. Intel website
2011.
174. Bohr, M. In 14 nm process technology: Opening new horizons, 2014.
153
175. Subramanian, V.; Parvais, B.; Borremans, J.; Mercha, A.; Linten, D.; Wambacq, P.; Loo,
J.; Dehan, M.; Gustin, C.; Collaert, N.; Kubicek, S.; Lander, R.; Hooker, J.; Cubaynes, F.;
Donnay, S.; Jurczak, M.; Groeseneken, G.; Sansen, W.; Decoutere, S., Planar Bulk MOSFET
Versus FinFETs: An Analog/RF Perspective. Electron Devices, IEEE Transactions on 2006, 53
(12), 3071-3079.
176. Nag, A.; Zia, A. I.; Babu, A.; Mukhopadhyay, S. C., Printed electronics: Present and
future opportunities. 2015; Vol. 2016-, pp 380-389.
177. Kim, Y. L.; Li, B.; An, X.; Hahm, M. G.; Chen, L.; Washington, M.; Ajayan, P. M.;
Nayak, S. K.; Busnaina, A.; Kar, S.; Jung, Y. J., Highly aligned scalable platinum-decorated
single-wall carbon nanotube arrays for nanoscale electrical interconnects. ACS nano 2009, 3 (9),
2818.
178. Robert, D. C., Emerging Applications for High K Materials in VLSI Technology.
Materials 2014, 7 (4), 2913-2944.
179. Dorofeev, A. D. A. Intel's 22-nm process gives MOSFET switch a facelift.
https://www.embedded.com/print/4395587.
180. Franklin, A. D.; Luisier, M.; Han, S.-J.; Tulevski, G.; Breslin, C. M.; Gignac, L.;
Lundstrom, M. S.; Haensch, W., Sub-10 nm carbon nanotube transistor. Nano letters 2012, 12
(2), 758.
181. Javey, A.; Guo, J.; Paulsson, M.; Wang, Q.; Mann, D.; Lundstrom, M.; Dai, H., High-
field quasiballistic transport in short carbon nanotubes. Physical review letters 2004, 92 (10),
106804.
182. Xiong, X.; Jaberansari, L.; Hahm, M. G.; Busnaina, A.; Jung, Y. J., Building Highly
Organized Single‐Walled‐Carbon‐Nanotube Networks Using Template‐Guided Fluidic
Assembly. Small 2007, 3 (12), 2006-2010.
183. Fang, Z. Z., Sintering of advanced materials : fundamentals and processes. Oxford
Philadelphia, PA : Woodhead Pub.: Oxford
Philadelphia, PA, 2010.
184. Galagan, Y.; Coenen, E. W. C.; Abbel, R.; van Lammeren, T. J.; Sabik, S.; Barink, M.;
Meinders, E. R.; Andriessen, R.; Blom, P. W. M., Photonic sintering of inkjet printed current
collecting grids for organic solar cell applications. Organic Electronics 2013, 14 (1), 38-46.
185. Hennessy, J. L., Computer architecture a quantitative approach. 5th ed. ed.; Patterson, D.
A.; Asanovic, K.; ebrary, I., Eds. Amsterdam
Boston : Morgan Kaufmann/Elsevier: Amsterdam
Boston, 2012.
186. Díaz, A. C., Energy Life Cycle Assessment (LCA) of silicon-based photovoltaic
technologies and the influence of where it is manufactured and installed. University of Bacelona
2014.
187. Espinosa, N.; García-Valverde, R.; Krebs, F. C., Life-cycle analysis of product integrated
polymer solar cells. Energy & Environmental Science 2011, 4 (5), 1547.
188. Dahlben, L. J.; Eckelman, M. J.; Hakimian, A.; Somu, S.; Isaacs, J. A., Environmental
life cycle assessment of a carbon nanotube-enabled semiconductor device. Environmental
science & technology 2013, 47 (15), 8471.
154
189. Keller, A. A., Global life cycle releases of engineered nanomaterials. J Nanopart Res
2013.
190. Wigger, H.; Hackmann, S.; Zimmermann, T.; Koser, J.; Thoming, J.; von Gleich, A.,
Influences of use activities and waste management on environmental releases of engineered
nanomaterials. Sci Total Environ 2015, 535, 160-71.
191. Ray, P. C.; Yu, H.; Fu, P. P., Toxicity and environmental risks of nanomaterials:
challenges and future needs. J Environ Sci Health C Environ Carcinog Ecotoxicol Rev 2009, 27
(1), 1-35.
192. Gottschalk, F.; Kost, E.; Nowack, B., Engineered nanomaterials in water and soils: A risk
quantification based on probabilistic exposure and effect modeling. Environmental Toxicology
and Chemistry 2013, 32 (6), 1278-1287.
193. Bernd, G.; Fred, K.; Barry, P.; Ralf, K.; Michael, S.; Henning, W.; Arnim Von, G.; Fadri,
G., Risks, Release and Concentrations of Engineered Nanomaterial in the Environment.
Scientific Reports 2018, 8 (1), 1-18.
194. Nash, J.; Bosso, C., Extended Producer Responsibility in the United States. Journal of
Industrial Ecology 2013, 17 (2), 175-185.
195. Keller, A. A., Release of engineered nanomaterials from personal care products
throughout their life cycle. Journal of Nanoparticle Research 2014.
196. Khanna, A. S., Nanotechnology in High Performance Paint Coatings. Asian J. Exp. Sci.
2008, 21 (2), 25-32.
197. Cölfen, H., Feature Article: Analytical Ultracentrifugation of Nanoparticles. Polymer
News 2004, 29 (4), 101-116.
198. Salabat, A.; Eastoe, J.; Mutch, K. J.; Tabor, R. F., Tuning aggregation of microemulsion
droplets and silica nanoparticles using solvent mixtures. J Colloid Interface Sci 2008, 318 (2),
244-51.
199. Hollamby, M. J.; Eastoe, J.; Chemelli, A.; Glatter, O.; Rogers, S.; Heenan, R. K.; Grillo,
I., Separation and purification of nanoparticles in a single step. Langmuir 2010, 26 (10), 6989-
94.
200. Myakonkaya, O.; Guibert, C.; Eastoe, J.; Grillo, I., Recovery of nanoparticles made easy.
Langmuir 2010, 26 (6), 3794-7.
201. Pati, P.; McGinnis, S.; Vikesland, P. J., Waste not want not: life cycle implications of
gold recovery and recycling from nanowaste. Environ. Sci.: Nano 2016, 3 (5), 1133-1143.
202. Aitken, R. J.; Chaudhry, M. Q.; Boxall, A. B.; Hull, M., Manufacture and use of
nanomaterials: current status in the UK and global trends. Occup Med (Lond) 2006, 56 (5), 300-
6.
203. Keller, A. A.; Lazareva, A., Predicted Releases of Engineered Nanomaterials: From
Global to Regional to Local. Environmental Science & Technology Letters 2014, 1 (1), 65-70.
204. Mackevica, A.; Hansen, S. F.; Olsson, M. E., Release of nanomaterials from consumer
products and implications for consumer exposure assessment. 2016.
205. Mackevica, A.; Olsson, M.; Hansen, S., Silver nanoparticle release from commercially
available plastic food containers into food simulants. An Interdisciplinary Forum for Nanoscale
Science and Technology 2016, 18 (1), 1-11.
206. PSI https://www.productstewardship.us.
155
207. Dhoke, S. K.; Mangal Sinha, T. J.; Khanna, A. S., Effect of nano-Al2O3 particles on the
corrosion behavior of alkyd based waterborne coatings. Journal of Coatings Technology and
Research 2008, 6 (3), 353-368.
208. Yilmaz, C.; Sirman, A.; Halder, A.; Busnaina, A., High-Rate Assembly of Nanomaterials
on Insulating Surfaces Using Electro-Fluidic Directed Assembly. Acs Nano 2017, 11, 7679-
7689.
209. Shahzad, F., Electromagnetic interference shielding with 2D transition metal carbides
(MXenes). Science 2016, 353 (6304).
210. Wolfgang Dubbert, K. S., Doris Völker, Petra Apel, Use of nanomaterials in coatings.
Federal Environment Agency: Dessau-Roßlau, Germany, 2014.
211. Boostani, H.; Modirrousta, S., Review of Nanocoatings for Building Application.
Procedia Engineering 2016, 145, 1541-1548.
212. Fernando, R. H., Nanocomposite and Nanostructured Coatings: Recent Advancements. In
Nanotechnology Applications in Coatings, American Chemical Society: 2009; Vol. 1008, pp 2-
21.
213. PaintCare, A. ACA and PaintCare®: Steering a Post-Consumer Paint Solution.
https://www.paint.org/publications-resources/issue-backgrounder/aca-paintcare/.
214. Shah, A.; Ates, M.; Kotz, S.; Seo, J.; Abraham, K.; Somu, S.; Busnaina, A., A Layered
Carbon Nanotube Architecture for High Power Lithium Ion Batteries. J. Electrochem. Soc. 2014,
161 (6), A989-A995.
215. Sun, Y.; Liu, N.; Cui, Y., Promises and challenges of nanomaterials for lithium-based
rechargeable batteries. Nature Energy 2016, 1 (7), 16071.
216. Hakimian, A., Economic and Environmental Impact Analysis of MWCNT Lithium
Nickel Manganese Cobalt Oxide Batteries Manufacturing Considering Worker Safety. Isaacs, J.
A.; Abraham, K.; Cullinane, T.; Kamarthi, S., Eds. ProQuest Dissertations Publishing: 2015.
217. Bakshi, S. R.; Agarwal, A., An analysis of the factors affecting strengthening in carbon
nanotube reinforced aluminum composites. Carbon 2011, 49 (2), 533-544.
218. Nash, J.; Hagemann, K.; Bosso, C. In Stewardship of nano-enabled batteries, 2012 IEEE
International Symposium on Sustainable Systems and Technology (ISSST), 16-18 May 2012;
2012; pp 1-1.
219. Yan, J.; Liu, X.; Wang, X.; Li, B., Long-life, high-efficiency lithium/sulfur batteries from
sulfurized carbon nanotube cathodes. Journal of Materials Chemistry A 2015, 3 (18), 10127-
10133.
220. Mahajan, A.; Kingon, A.; Kukovecz, Á.; Konya, Z.; Vilarinho, P. M., Studies on the
thermal decomposition of multiwall carbon nanotubes under different atmospheres. Materials
Letters 2013, 90, 165-168.
221. Evers, P. Nanotechnology in Medical Applications: The Global Market; BCC Research
LLC: 2017.
222. Baptista Pedro, V., Nanodiagnostics: leaving the research lab to enter the clinics? In
Diagnosis, 2014; Vol. 1, p 305.
223. Weissig, V.; Pettinger, T. K.; Murdock, N., Nanopharmaceuticals (part 1): products on
the market. Int J Nanomedicine 2014, 9, 4357-73.
156
224. Law, A. V.; Sakharkar, P.; Zargarzadeh, A.; Tai, B. W.; Hess, K.; Hata, M.; Mireles, R.;
Ha, C.; Park, T. J., Taking stock of medication wastage: Unused medications in US households.
Res Social Adm Pharm 2015, 11 (4), 571-8.
225. Cantrell, L.; Suchard, J. R.; Wu, A.; Gerona, R. R., Stability of active ingredients in long-
expired prescription medications. Archives of Internal Medicine 2012, 172 (21), 1685-1687.
226. Jingqi, L.; Qing, Z.; Yehai, Y.; Sai, L.; Longqing, C., Fabrication of Carbon Nanotube
Field-Effect Transistors by Fluidic Alignment Technique. Nanotechnology, IEEE Transactions
on 2007, 6 (4), 481-484.
227. Sadie, J. A.; Subramanian, V., Three-Dimensional Inkjet-Printed Interconnects using
Functional Metallic Nanoparticle Inks. Advanced Functional Materials 2014, 24 (43), 6834-
6842.
228. Liang, Y. N.; Lok, B. K.; Wang, L.; Feng, C.; Lu, A. C. W.; Mei, T.; Hu, X., Effects of
the morphology of inkjet printed zinc oxide (ZnO) on thin film transistor performance and
seeded ZnO nanorod growth. Thin Solid Films 2013, 544, 509-514.
229. Bautin, V. A.; Seferyan, A. G.; Nesmeyanov, M. S.; Usov, N. A., Magnetic properties of
polycrystalline cobalt nanoparticles. AIP Advances 2017, 7 (4), 045103.
230. Kahng, Y. H.; Kim, M.-K.; Lee, J.-H.; Kim, Y. J.; Kim, N.; Park, D.-W.; Lee, K., Highly
conductive flexible transparent electrodes fabricated by combining graphene films and inkjet-
printed silver grids. Solar Energy Materials and Solar Cells 2014, 124, 86-91.
231. Tang, L.; Cheng, J., Nonporous Silica Nanoparticles for Nanomedicine Application.
Nano Today 2013, 8 (3), 290-312.
232. Kong, Y. L.; Tamargo, I. A.; Kim, H.; Johnson, B. N.; Gupta, M. K.; Koh, T.-W.; Chin,
H.-A.; Steingart, D. A.; Rand, B. P.; McAlpine, M. C., 3D Printed Quantum Dot Light-Emitting
Diodes. Nano Letters 2014, 14 (12), 7017-7023.
233. Hardman, R., A Toxicologic Review of Quantum Dots: Toxicity Depends on
Physicochemical and Environmental Factors. Environmental Health Perspectives 2006, 114 (2),
165-172.
234. (SERI), S. E. R. I., The Responsible Recycling (“R2”) Standard for Electronics
Recyclers. 2013.
235. Shirasaki, Y.; Supran, G. J.; Bawendi, M. G.; Bulovic, V., Emergence of colloidal
quantum-dot light-emitting technologies. Nature Photonics 2012, 7 (1), 13-23.
236. Chopra, S. S.; Theis, T. L., Comparative cradle-to-gate energy assessment of indium
phosphide and cadmium selenide quantum dot displays. Environ. Sci.: Nano 2017, 4 (1), 244-
254.
237. Franklin, A. D.; Lin, A.; Wong, H. S. P.; Zhihong Chen, H. S. P., Current Scaling in
Aligned Carbon Nanotube Array Transistors With Local Bottom Gating. Electron Device
Letters, IEEE 2010, 31 (7), 644-646.
238. Statista, Wide color gamut display shipment area worldwide from 2015 to 2018 (in
million square meters). Statista - The Statistics Portal. : 2018.
239. PRNewswire Quantum Materials Increasing Quantum Dot Production Capacity to Two
Metric Tons by Q2 2015 to Meet 4K LCD TV and Display Demand.
https://www.prnewswire.com/news-releases/quantum-materials-increasing-quantum-dot-
157
production-capacity-to-two-metric-tons-by-q2-2015-to-meet-4k-lcd-tv-and-display-demand-
300016260.html.
240. Bishop, L., In Best Buy’s Recycling Program Is Changing. Here’s How And Why., 2016;
Vol. 2018.
241. Kehoe, A. M. P. a., CELL PHONE RECYCLING ACT OF 2004. CALIFORNIA, T. S.
O., Ed. 2004.
242. Baldé, C. P., Wang, F., Kuehr, R., Huisman, J The global e-waste monitor – 2014; IAS –
SCYCLE, United Nations University: Bonn, Germany, 2015.
243. Duan, H., Reed Miller, T., Gregory, J. og Kirchain, R. Quantitative characterization of
domestic and transboundary flows of used electronics - analysis of Generation, Collection and
Export in the United States; Massachusetts Institute of Technology for StEP: 2013.
244. O'Connell, M.; Hickey, S.; Besiou, M.; Fitzpatrick, C.; Van Wassenhove, L. N.,
Feasibility of Using Radio Frequency Identification to Facilitate Individual Producer
Responsibility for Waste Electrical and Electronic Equipment. Journal of Industrial Ecology
2013, 17 (2), 213-223.
245. Li, X.; Zhang, L.; Wang, X.; Shimoyama, I.; Sun, X.; Seo, W.-S.; Dai, H., Langmuir-
blodgett assembly of densely aligned single-walled carbon nanotubes from bulk materials.
Journal of the American Chemical Society 2007, 129 (16), 4890-4891.
246. Walser, T.; Limbach, L. K.; Brogioli, R.; Erismann, E.; Flamigni, L.; Hattendorf, B.;
Juchli, M.; Krumeich, F.; Ludwig, C.; Prikopsky, K.; Rossier, M.; Saner, D.; Sigg, A.; Hellweg,
S.; Günther, D.; Stark, W. J., Persistence of engineered nanoparticles in a municipal solid-waste
incineration plant. Nature Nanotechnology 2012, 7, 520.
158
7. Appendix A
Supplementary Information Scalable Printing of High-Resolution Flexible Transparent Grid
Electrodes using Directed Assembly of Silver Nanoparticles
Figure S1. The method to prepare the grid electrodes for electrical characterization and the
measurement setup.
25 mm
25 mm
Deposit Ag with
shadow mask
Cut two perfect
squares
Use four-probe
method
V
I
Apply current
Measure voltage
159
Figure S2. (a) A flattened confocal micrograph for the grid printed at 0.25 mm/min using 1.3 um thick
photoresist along with the extracted cross-sectional profile (b). A 3D model generated using the same image
is shown in (c).
160
Supporting Information: Drift-free Polypyrrole Modified Flexible Carbon Nanotube Sensor for
Continuous Lactate Monitoring
Materials and Reagents
A 0.1g/L aqueous suspension of carboxylic functionalized multi-walled carbon nanotubes
(MWCNTs) with an average diameter of 25 nm and and length of 1 µm was obtained from Brewer
Science. Melinex® 454 polyethylene terephthalate (PET) film was obtained from Dupont. Lactate
oxidase from Pediococcus sp., pyrrole-2-carboxylic acid (99%), N-(3-Dimethylaminopropyl)-N′-
ethylcarbodiimide hydrochloride (EDC) (98%), N-hydroxysuccinimide (NHS) – (98%), sodium
L-lactate > 99% (NT), phosphate buffered saline (1X-PBS, and ferrocenecarboxylic acid (FCX)
(99%), were obtained from Sigma Aldrich. 2-(N-morpholino) ethanesulfonic acid (MES) packs
were obtained from Fischer Scientific and dissolved in deionized water to make 0.1 M MES with
pH 4.7.
Experimental Details
PET was treated with oxygen plasma at 100 W (15 sccms O2, 0.38 Torr) for 2 minutes using
Anatech100 asher before UV lithography. For the first layer, Shipley S1813 was used (4000 RPM,
45s) and baked for 60 C at 115 °C. Quintell 4000 was used for photolithography. Microchem
AZ726 developer was used for 45 sec to develop the photoresist. The PET wafer was baked again
at 115 ˚C for 2 min after development. Contact angle measurements were performed using a
Phoenix tool.
MWCNT suspension was probe sonicated (QSonica Q700) for 30 minutes while immersed in an
ice bath before assembly. After assembly, photoresist was removed by immersing the substrate in
acetone and isopropyl alcohol for 3 min each, respectively. The thickness of the assembled CNT
161
channel was measured using Park Systems NX-20 atomic force microscope. The electrodes were
patterned across the CNT channel using AZ nLoF 2020 negative tone photoresist (4000 RPM, 45
sec) followed by the deposition of Ti/Au (5/150 nm) using electron beam evaporation. A Horiba
Multiline Raman Spectrophotometer was used for Raman spectroscopy. A cascade Summit 12000
station along with Agilent 4156C was used for electrical characterization.
Figure S1(a) shows the 90 sensors fabricated on a four-inch PET substrate while 1(b) and 1(c)
show the exposed CNT channel and the printed CNTs, respectively.
Exposed CNT
channel
500 nm
5 µm
(a) (b)
(c)
162
Figure S2. The selectivity of the sensor towards lactate is confirmed by testing in urea and glucose
(c). The sensor was tested on the same day it was functionalized.
Setup for extrapolating viscosity.
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
0 100 200 300R
esi
stan
ce (Ω
)
Time (s)
10 mM Urea
8 mM Lactate
60 µM Glucose
163
8. Appendix B
Each PETG film was dissolved in DMF at 100 mg/mL by stirring for 6 hrs on a hotplate at 100
˚C. A 1 mL droplet of each polymer was passed through a thin, marked capillary tube as shown in
the image below and the time between the markings 0 and 9 was measured. Five readings were
taken for each polymer and the time was averaged and plotted against the known viscosities to
extrapolate the viscosity of the unknown sample (PETG wafer).