Unconventional Approaches to Micro- and Nanofabrication

for Electronic and Optical Applications

A dissertation presented

by

Darren John Lipomi

to

The Department of Chemistry and Chemical Biology

In partial fulfillment of the requirements

for the degree of

Doctor of Philosophy

in the subject of

Chemistry

Harvard University

Cambridge, Massachuestts

June, 2010

© 2010 – Darren John Lipomi

All rights reserved.

iii

Dissertation Advisor: Submitted by:

Professor George M. Whitesides Darren John Lipomi

Unconventional Approaches to Micro- and Nanofabrication for

Electronic and Optical Applications

Abstract

New applications in electronics and optics require methods of forming micro- and

nanostructures in ways that are applicable to different classes of materials and substrates.

These methods should also be simple, inexpensive, and accessible to the greatest number

of users possible. This dissertation explores several unconventional methods of forming

micro- and nanostructures for electronic and optical applications. Chapter 1 provides an

overview of the most prominent method of nanofabrication in this dissertation,

nanoskiving, which is the use of mechanical sectioning of encapsulated thin films to

generate nanostructures. Several ancillary techniques are used to prepare substrates for

nanoskiving: shadow evaporation, chemical synthesis, electrodeposition, soft lithographic

molding, spin-coating, and rolling. Chapter 2, Appendix I, II, and IV, survey the

materials and methods of nanoskiving. Appendices III and VI describe combinations of

nanoskiving with soft lithographic molding to generate and replicate two-dimensional

structures for electronic and optical applications. Chapter 3 and Appendix V detail

contact and non-contact methods to manipulate the structures produced by nanoskiving

and transfer them to optical fibers and other optical structures. Chapter 4 and Appendix

VII describe two additional forms of unconventional fabrication: shadow evaporation,

iv

and fabrication using a commercial nanoindentation system. Proof-of-principle

applications demonstrated using nanoskiving include chemical sensors (Chapter 1,

Appendices I and V), substrates for electrodeposition (Appendix III), organic

photodetectors (Appendix I), plasmonic waveguides (Appendix IV), and near-IR filters

(Appendix VI); applications for nanoindentation include substrates for surface-enhanced

Raman scattering (Appendix VII); and applications of shadow evaporation include field-

effect transistors, logic gates, and other electronic structures (Chapter 4).

v

Dedicated to Dina

vi

Acknowledgements

Over the last five years, I have had the opportunity to work beside almost one

hundred fellow members of the Whitesides Group. While I learned at least one thing

from nearly all of them, several individuals deserve specific mention. Paul Bracher has

sat in the desk behind mine nearly all of the last five years. His unique perspective and

advice have been invaluable, and I wish him all the best at Caltech and beyond. Ryan

Chiechi and Emily Weiss were my earliest mentors in the lab; they introduced me to

organic electronics, and sparked my interest in nanotechnology for the energy challenge.

Later, Michael Dickey taught me all of the basics of micro- and nanofabrication, while

Ben Wiley introduced me to plasmonics, and, through his example, equipped me with the

confidence to approach potential collaborators early and often. I give significant credit to

Qiaobing Xu for laying the groundwork for nanoskiving. While I didn’t learn the

technique from him directly, I hope my contributions were worthy of his.

I would not have graduated on time without the skilled hands of Ramses

Martinez, who played a crucial role in three of my final projects. Similarly, I thank

Mikhail Kats, Phil Kim, Sung Hoon Kang, and Parag Deotare—colleagues from other

labs with whom I’ve had many stimulating discussions and valuable collaborations. I also

acknowledge the friendship and collaboration of labmates Bill Reus, Filip Ilievski, Ludo

Cademartiri, Rob Rioux, and Jinlong Gong. I’m proud to have remained close to Tom

Baker and Meg Thurlow, fellow CCB students who started with me in 2005. Thanks also

to my GPC colleagues—Jason Beiger, Jessica Wu, Lizzy Hulme, Leslie Vogt, and Lu

Wang. I wish you all the best in your future scientific (and non-scientific) adventures.

vii

I thank my faculty collaborators Federico Capasso, Marko Lončar, Venky

Narayanamurti, Mara Prentiss, and Joanna Aizenberg for willingly volunteering your

time, resources, and students to our shared projects. I offer special thanks to Joanna,

Hongkun Park, and Tobias Ritter for serving on my GAC and defense committees, and to

Tony Shaw for helping me, and many of my colleagues, navigate the trials of graduate

school. Thanks also to the Whitesides office staff, T.J. Martin, Melissa LeGrand, Tracie

Smart, Bob Holt, Launa Johnston, Terri Howard, Elisa Lenssen, and Sandy Rosen.

I’m eternally grateful to my early mentors in chemistry: Tom Dowd from Hilton

High School, who never allowed us to fall into the abyss of mediocrity, and Jim Panek

and John Straub from Boston University, with whom I first learned the art of scientific

research, and who opened my eyes to the intersection of chemistry, physics, biology, and

human experience. I would not have made it this far without all of you.

George Whitesides. We met at the Beckman Scholars Symposium in July 2003.

Even though I considered myself an organic synthetic chemist by the time I applied for

graduate school, the possibility of joining the Whitesides Lab as a “wildcard” option

cemented my decision to come to Harvard, and I have no regrets. His tireless pursuit of

truly innovative ideas, experimental rigor, and transparent prose has profoundly

influenced both my sentiments and my professionalism. I intend to pay it forward.

I would lastly, and most importantly, like to thank my family. My mother and

father, Rosalind and Mariano, and sisters, Andrea and Deena, have supported me in every

career aspiration I’ve had in life: first inventor, then doctor, then classical musician, then

scientist, and finally back to inventor. Finally, I thank my wonderful wife Dina, to whom

I dedicate this work, and my life.

viii

Table of Contents

Chapter 1. Use of Thin Sectioning (Nanoskiving) to Fabricate Nanostructures for

Electronic and Optical Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Abstract. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Why Nano?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

Nanofabrication. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3

Soft Lithography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Ultramicrotomy and Nanoskiving. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Microtomy and Microscopy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6

The Ultramicrotome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7

The Process of Sectioning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Nanoskiving. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10

Diamond Knives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Equipment and Materials: Minimum Requirements. . . . . . . . . . . . . . . . . 14

Scope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Electronic Applications of Nanoskiving. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

Fabrication of Nanoelectrodes for Electrodeposition. . . . . . . . . . . . . . . . 16

Fabrication of Addressable Nanowires Separated by a Nanogap. . . . . . . 18

Fabrication of Chemoresistive Nanowires of Palladium. . . . . . . . . . . . . .19

Fabrication of Chemoresistive Conjugated Polymer Nanowires. . . . . . . .20

ix

Arranging Nanowires of Different Types Using Magnetic Mooring. . . . 21

Fabrication of an Ordered Bulk Heterojunction

of Conjugated Polymers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .24

Optical Applications of Nanoskiving. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

Fabrication of Gold Nanowires and Size-Dependent

Plasmon Resonance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27

Fabrication of Single-Crystalline Gold Nanowires for

Plasmonic Waveguiding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Fabrication of 2D Arrays of Nanostructures. . . . . . . . . . . . . . . . . . . . . . . 31

Plasmonic Properties of Two-Dimensional Arrays of Nanostructures. . . 36

Integration of Plasmonic Arrays with Optical Fibers. . . . . . . . . . . . . . . . 38

Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40

Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .42

Chapter 2. Survey of Materials for Nanoskiving and Influence of the Cutting

Process on the Nanostructures Produced. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .52

Abstract. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .53

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Experimental Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

Results and Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .66

x

Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .86

Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .88

Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .90

Chapter 3. Patterning the Cleaved Facets of Optical Fibers with Metallic

Nanostructures Using Nanoskiving. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .95

Abstract. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .96

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

Experimental Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

Results and Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .101

Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .108

Acknowledgements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .109

Chapter 4. Transistors Formed from a Single Lithography Step Using Information

Encoded in Topography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .112

Abstract. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .113

Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

TEMIL. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

Experimental Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

Results and Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .125

xi

Advantages and Limitations of TEMIL. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .130

Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .132

Experimental. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

Advantages and Disadvantages of Shadow Evaporation. . . . . . . . . . . . . . . .141

Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

References and Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .146

Appendices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

Appendix I. Laterally Ordered Bulk Heterojunction of Conjugated Polymers:

Nanoskiving a Jelly Roll. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

Appendix II. Fabrication of Conjugated Polymer Nanowires by Edge Lithography. . .160

Appendix III. Electrically Addressable Parallel Nanowires with 30 nm

Spacing from Micromolding and Nanoskiving. . . . . . . . . . . . . . . . . . . . . . . . . . 171

Appendix IV. Fabrication of Surface Plasmon Resonators by Nanoskiving Single-

Crystalline Gold Microplates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .186

Appendix V. Integrated Fabrication and Magnetic Positioning of Metallic and

Polymeric Nanowires Embedded in Thin Epoxy Slabs. . . . . . . . . . . . . . . . . . . .196

xii

Appendix VI. Fabrication and Replication of Arrays of Single- or Multicomponent

Nanostructures by Replica Molding and Mechanical Sectioning. . . . . . . . . . . . 210

Appendix VII. Micro- and Nanopatterning of Inorganic and Polymeric Substrates by

Indentation Lithography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .238

Chapter 1

Use of Thin Sectioning (Nanoskiving) to Fabricate Nanostructures for Electronic

and Optical Applications

Darren J. Lipomi and George M. Whitesides

Department of Chemistry and Chemical Biology, Harvard University

12 Oxford St., Cambridge, MA, U.S.A. 02138

2

Abstract

This chapter reviews nanoskiving—a simple and inexpensive method of

nanofabrication. Nanoskiving requires three steps: i) deposition of a metallic,

semiconducting, ceramic, or polymeric thin film onto an epoxy substrate (which may be

topographically patterned); ii) embedding this film in epoxy, to form an epoxy block,

with the film as an inclusion; and, the key step, iii) sectioning the epoxy block into slabs

with an ultramicrotome. These epoxy slabs, which can be 30 nm – 10 µm thick, contain

nanostructures whose lateral dimensions are equal to the thicknesses of the embedded

thin films, and thus can be as thin as 10 nm. When combined with soft lithographic

molding and other processes, nanoskiving can produce patterns of structures that can be

transferred to almost any substrate, and that would be difficult or impossible to generate

by other procedures. Optical applications of structures produced by this method include

surface plasmon resonators, plasmonic waveguides, and frequency-selective surfaces.

Electronic applications include nanoelectrodes for electrochemistry, chemoresistive

nanowires, and heterostructures of organic semiconductors that exhibit a photovoltaic

effect.

3

1. Introduction

1.1. Why Nano?

Many of the most important phenomena in nature—the binding of proteins and

ligands, the interaction of light with matter, and the mechanism of charge transport in

materials—occur on the scale of 1 – 100 nm. Processes that occur over this range of

sizes, which begins with large molecules and ends with objects that are resolved with

conventional microscopes, are the purview of the field known as nanoscience.

Nanostructured materials display properties not found in bulk materials. These properties

comprise the effects size confinement, including size-dependent band gaps in quantum

dots,1 localized surface plasmon resonances in sub-wavelength particles,2 exceptional

strength and ballistic transport of electrons in carbon nanotubes,3 and consequences of the

fact that these structures are “all—or mostly—surface”.4

Beyond discovery-based scientific inquiry in these areas, there are also

opportunities for technological development. Nanostructured materials have already

pushed the electronics industry toward faster, cheaper, and more efficient devices,5 are

making inroads into medicine,6 and could contribute in a significant way to sensing,

communication, and computation based on nanophotonics.7,8 Developing methods of

generating and patterning nanostructures in ways that are reproducible, scalable,

inexpensive, general with respect to materials, and as widely accessible to as many users

as possible, is thus an important motivation for the sciences of materials chemistry and

nanofabrication.

1.2. Nanofabrication

4

Nanofabrication refers to the generation of patterns whose individual elements

have at least one lateral dimension between approximately 1 nm and 100 nm.9

Nanofabrication, along with microfabrication before it, has been a key enabler of modern

science and technology, and has underpinned essentially all electronics since the

invention of the integrated circuit in 1958.5 Nanofabrication has two principal steps:

mastering (e.g., forming master structures such as amplitude and phase masks for

photolithography) and replication. Mastering encodes new nanoscale information in a

form from which it can be replicated. In semiconductor manufacturing, the principal tool

of mastering is electron-beam lithography (EBL), which creates patterns in a photomask.

Mastering is a time-intensive process, and may require twenty hours to produce a single

mask.10 Replication of this pattern takes the form of photolithography, in which light

passes through the photomask and creates an image on a wafer coated with a film of

light-sensitive polymer called photoresist. Modern exposure tools generate around 100

copies/min.10 After chemical processing, the surface of the material comprising the wafer

can be modified in the areas of the film unprotected by photoresist. Iteration of these

processes generates the devices and connections on a chip.

An empirical trend—Moore’s Law—shows that the number of transistors per

microprocessor has doubled approximately every 18 months, with concomitant decreases

in cost, power consumption, and increases in processing speed and storage capacity for

memory devices.11,12 This trend has become a self-fulfilling prophecy, which has

motivated the development of new steppers for projection photolithography,13 chemistry

for photoresists,14 and other technologies.5 The state-of-the-art in photolithography

produces an average half-pitch in memory devices of 32 nm using 193 nm light combined

5

with immersion optics,15 phase-shifting masks,16 and multiple exposures.17 Next-

generation lithographic tools, including extreme ultraviolet lithography (EUVL),18

maskless lithography (ML2, which would use thousands of electron beams to replicate

patterns without the need for a physical master),19 and step-and-flash imprint lithography

(SFIL)20 are expected to drive the average half-pitch down to 16 nm by 2019, according

to the International Technical Roadmap for Semiconductors.10

Semiconductor devices are manufactured using the most sophisticated processes

ever employed for commercial products. The scale of investment in these tools is so high

(and the precision is so impeccable), that it does not make sense to compete with them for

their designed purpose—manufacturing multilayered semiconductor devices on planar,

rigid substrates. There are at least five reasons to explore “unconventional” methods of

fabrication. (1) Cost: photolithographic steppers are prohibitively expensive, particularly

for universities. (2) Accessibility: scanning-beam lithographic and photolithographic

tools are usually found in a cleanroom, whose construction, operation, and maintenance

impose a significant financial burden on an institution. (3) Incompatibility: organics,

biologics, and other nontraditional materials often cannot be patterned directly using

conventional tools, nor can they by processed using the same equipment, and in the same

cleanroom, that is used for electronics. (4) Form factors: conventional tools are

incompatible with non-planar,21 mechanically compliant,22 or very small (< 100 µm)

substrates.23 (5) Overkill: there are a large number of potential applications of

nanotechnology—in biology, optics, chemistry, devices for the conversion and storage of

energy, and other areas—that are significantly more tolerant of defects than are

semiconductor devices, and whose requirements can be satisfied using simpler tools.

6

1.3. Soft Lithography

Soft lithography24 is a set of techniques whose key step is the transfer of patterns

using an elastomeric stamp or mold, which is made from poly(dimethylsiloxane)

(PDMS), perfluorpolyethers,6 or other polymers.25 There are three general modes of soft

lithography: i) molding (replica molding,21,26,27 solvent-assisted micromolding,28 and

micromolding in capillaries29); ii) printing (microcontact printing,30-33 charge printing,34

and nanotransfer printing35); and iii) near-field optical lithography (in two or three

dimensions).36-38 The key steps of all forms of soft lithography rely on physical contact,

which is not subject to the diffraction of light, or scattering of beams of charged

particles.9 There is a fourth mode for the generation and replication of patterns with

nanoscale features that is often, but not always, combined with soft lithography: thin

mechanical sectioning using an ultramicrotome, or “nanoskiving”.39

2. Ultramicrotomy and Nanoskiving

2.1. Microtomy and Microscopy

Sectioning with a microtome has been a tool of microscopists since John Hill

described the first instrument in 1770. This manually operated device could produce

sections of timber as thin as 25 µm, for analysis with a light microscope.40 Use of the

microtome was restricted, for the most part, to biology, until the invention of the

transmission electron microscope (TEM) in the 1930s.41 Transmission of electrons

through a specimen required a device capable of producing sections with thicknesses <

100 nm. This device became known as the ultramicrotome. Ultramicrotomy enabled

7

microstructural analysis not only of biological specimens, but of inorganic materials as

well. It is a complimentary technique to ion thinning and electropolishing for the

preparation of hard materials for TEM (it is the primary method of the preparation of

polymeric samples, however).42

The history of microtomy can be found in several books and reviews.

Bracegirdle’s book, A History of Microtechnique, describes the development of

microtomy between 1770 and 1910.43 The review by Pease and Porter provides an

account of the co-development of electron microscopy and ultramicrotomy,41 while that

of Malis and Steele is the most complete review of ultramicrotomy, in the context of

inorganic materials science, through 1990.42 The book by Goldstein et al. covers all

aspects of embedding and sectioning hard and soft materials, including histological

samples.44

2.2. The Ultramicrotome

Figure 1.1a shows a modern ultramicrotome. Its components include a

stereomicroscope, a movable stage that holds the knife, and a sample chuck (1.1b)

attached to a movable arm, that holds the epoxy block. The movable arm controls the fine

positioning of the block and can advance in steps as small as 1 nm toward a single-

crystalline diamond knife (1.1c). The mechanism of fine control involves a stepper motor

connected to a spindle, and a lever that transforms micrometer-length displacements of

the spindle into nanometer-length displacements of the arm.39 The arm and the epoxy

block advance toward the knife in an elliptical path when viewed from the side, as drawn.

The speed of cutting is 0.1 – 10 mm/s, and produces sections at a rate of 0.5 – 2 Hz. The

8

Figure 1.1. Photographs and schematic drawings of the tools of ultramicrotomy and

nanoskiving. (a) A photograph of a Leica UC6 ultramicrotome. The user aligns the block

face to the diamond knife using a stereomicroscope. (b) A side view of the sample chuck

and knife holder as the epoxy block impinges upon the knife. (c) A top view of the

single-crystalline diamond blade and the water-filled trough. (d) A “ribbon” of epoxy

slabs floating on the surface of water. The green color indicates that the slabs are

approximately 250 nm – 300 nm thick. (d) A schematic drawing of the “perfect loop”

tool, with which the user lifts the sections off the surface of the water in a thin film of

water and transfers them to a substrate. (d) A schematic drawing of the sectioning

process. Reproduced in part with permission from ref. 39. Copyright 2008, American

Chemical Society.

9

nascent epoxy slabs slide onto the surface of a water bath in the form of individual slabs

or ribbons of connected slabs (1.1d). Assembly of the slabs into ribbons depends on the

shape of the manually trimmed facet of the block, which determines the extent to which

the slabs can aggregate. The best method of transferring the sections from the water bath

to a substrate is to use the “perfect loop” tool (Electron Microscopy Sciences), which

suspends the slab in a film of water, from which it can be transferred to essentially any

substrate (1.1e).

2.3. The Process of Sectioning

Figure 1.1f is a schematic drawing of the sectioning process. Sectioning involves

a complicated interplay of processes: compression of the sample during the initiation of

cutting, and of the slab thereafter; tension perpendicular to the plane of sectioning;

generation of new surfaces; bending, as the slab reorients from vertical to horizontal;

shearing stress (greatest in materials with low flexibility); friction of the slabs on the

knife; and generation of heat.45 The embedding medium should have two properties, i) a

relatively high value of elastic modulus (~3 GPa, materials that are too compliant deflect

from the knife edge, rather than cleave), and ii) a high yield stress after which the

material undergoes plastic deformation (~70 MPa, otherwise the slab will deform upon

sectioning).45 Crosslinked epoxy resins fill most of these criteria at ambient temperatures,

though it is possible to section softer materials using cryogenic temperatures. Our

laboratory has achieved good results with Araldite 50246 and Epo-Fix (Electron

Microscopy Sciences),47 which are thermally curable, and excellent results with UVO-

10

114 (Epotek), which is UV-curable. The surfaces of the epoxy slabs are smooth, with

values of roughness (rms) of ~0.5 nm.48

The sectioning process is similar to methods of mechanical cleavage in

metalworking. Both microtomed metallic specimens and micromachined chips undergo

compression in the direction of cutting and exhibit shear lamellae perpendicular to that

direction.42 The extent to which the propagation of the crack extends ahead of the edge of

the knife depends on the amount of energy the sample can absorb by plastic flow prior to

fracture. The orientation of cleavage planes in crystalline samples also determines the

extent of fragmentation upon sectioning brittle materials, as Antonovsky observed in

samples of alumina.49

2.3. Nanoskiving

There are two modes of combining nanoskiving with replica molding and thin-

film deposition (Figure 1.2a). The first mode is cutting perpendicular to a topographically

patterned or planar thin film. This operation produces a structure whose geometry is the

cross sectional profile of the original molded structure, a grating, as shown (1.2b).39 A

process of dry etching removes the epoxy matrix and leaves behind a nanostructure in the

shape of a periodic square wave. The second mode is cutting parallel to a topographically

patterned film. This operation produces parallel nanowires that correspond to the

sidewalls of the topographic features of the grating (1.2c).

To determine the applicability of different materials to nanoskiving, we

performed a survey of thin films, deposited using different methods—evaporation,

sputter-coating, electroless deposition, deposition in an electrochemical cell, spin-coating,

11

Figure 1.2. Two general strategies to generate nanostructures by sectioning thin films

coated on a polymer substrate bearing relief features. Soft lithographic procedures

generate a PDMS mold, which templates the formation of an inverse replica in epoxy.

Physical vapor deposition produces a metallic film (e.g., Au) on the replica. (Sputtering

can be used to coat the epoxy conformally, while evaporation can be used to coat only the

desired sidewalls, by line-of-sight deposition.) Additional epoxy embeds the entire

structure to form a block. Sectioning the block perpendicular to the topographic features

produces a nanowire with the form of a cross-sectional profile of the grating, while

sectioning parallel generates parallel nanowires, which come from the metalized

sidewalls of the embedded structure. Reproduced with permission from ref. 39. Copyright

2008, American Chemical Society.

12

and solution-phase synthesis and subsequent deposition. The four major conclusions are:

i) for evaporated, elemental films, soft and compliant materials tend to remain intact upon

sectioning (softer than platinum, or those with bulk values of hardness < 500 MPa), while

hard and stiff materials tend to fragment (those harder than nickel); ii) platinum and

nickel are on the borderline between soft and hard, for which the extent of fragmentation

depends on the method of deposition, and the morphology of the film; iii) the extent of

fragmentation is higher when the orientation of the film is parallel to the direction of

cutting than when the film is perpendicular to it; and iv) the speed of cutting has no effect

on the frequency of defects, from 0.1 mm/s to 10 mm/s (which is consistent with Jesior’s

observation that the cutting speed also has no effect on compression50). We have

successfully formed nanostructures of aluminum, copper, silver, gold, lead, bismuth,

palladium, platinum, nickel, germanium, silicon dioxide, all conducting and

semiconducting polymers tested, and films of lead sulfide nanocrystals (Figures 1.3a –

1.3n).

2.6. Diamond Knives

The knife is the most important part of the microtome. Our laboratory uses a 35°

diamond knife, 1.8 – 2.4 mm in length, whose edge has a radius of curvature of 3 – 6

nm.51 The cost of a knife is $2,000 – $3,000. Knives must be re-sharpened every 6 – 12

months; this service is about half the cost of a new knife. Damage to the knife takes the

form of chipping (rather than homogeneous “dulling”). Chips in the knife cause scoring

of the epoxy slabs in the direction of cutting. Most scores have a width of 50 – 300 nm.

The most rapid deterioration of the quality of a knife we have observed occurred when

13

Figure 1.3. Examples of representative spans of nanowires formed by obtaining sections

of the metallic, polymeric, and semiconducting thin films. Each nanowire is physically

continuous over > 100 µm.

14

sectioning thick films (~500 nm) of hard materials (e.g. Ti) and micron-scale ceramic

objects (e.g. optical fibers). Significant chipping of the knife also occurs when hard

inorganic dust particles become inadvertently embedded in the epoxy blocks.

2.7. Equipment and Materials: Minimum Requirements

Almost all of the equipment needed for nanoskiving can be found in shared

facilities at most universities. The most important item needed to do nanoskiving is an

ultramicrotome, which is standard equipment in laboratories of electron microscopy. New

ultramicrotomes are approximately $60,000. We generally deposit metallic films by

evaportion or sputter-coating. Each method has useful characteristics. Electron-beam or

thermal evaporation produces a collimated beam of atoms, which can coat selectively the

sidewalls of photolithographic features using line-of-sight deposition. The advantage of

sputter-coating is that it can coat the sidewalls of topographic features conformally. Spin-

coating produces films of conjugated polymers, with uniform thickness, on planar

substrates. We have also formed thin films using electroless deposition, growth in an

electrochemical cell, and plasma-enhanced chemical vapor deposition (PECVD). All of

these methods produce films that are either polycrystalline or amorphous that, when

sectioned, produce nanostructures of roughly the same morphology (but likely retrograde

due to compression and shear) as the film from which it was cut. Shape-selective

synthesis of microparticles of a variety of materials can produce single-crystalline

precursors for nanoskiving.52,53 These structures, in turn, produce single-crystalline

nanostructures after sectioning.47

15

2.8. Scope

This review is organized by optical and electronic applications of structures

produced by nanoskiving. Optical applications include linear and two-dimensional arrays

of metallic nanostructures for applications based on localized surface plasmon resonances

(LSPRs) for frequency-selective surfaces, high-quality, single-crystalline gold nanowires

for plasmonic waveguiding, and methods of stacking and arranging these structures with

each other and with pre-deposited structures (e.g., optical waveguides). Electronic

applications include metallic nanowires as nanoelectrodes for electrochemistry, and

chemoresistive nanowires of conjugated polymers and palladium. As an optoelectronic

application, we have also fabricated a heterostructure of conjugated polymers, which

exhibits a photovoltaic effect when placed between electrodes with asymmetric work

functions.

3. Electronic Applications of Nanoskiving

3.1. Introduction

Conventional methods of nanofabrication are already well-suited to some

nanoelectronic applications. Modern microelectronic devices have had nanoscale

dimensions for about a decade, though the interesting effects of size reduction (e.g.,

tunneling through leaky gate dielectrics) are usually treated as something to be

suppressed.54 True nanoelectronics is in an exploratory phase. It combines new materials

and structures—carbon nanotubes,3 graphene,55 and semiconducting, metallic, or

dielectric nanowires56—that are addressed using conventional lithography. Some of the

most exciting new directions include integrating nanoelectronics with photonics on the

16

same chip.57 This section highlights five applications of nanoskiving to which

conventional fabrication is not easily applied. The structures and applications are i)

fabrication of electrically isolated, patterned electrodes for electrochemistry;58 ii)

individually addressable, parallel nanowires separated by a nanogap for nanoelectrodes;59

iii) parallel nanowires of metals60 and polymers61 with high pitch for chemical sensing;

iv) junctions of nanowires positioned using magnetic interactions for different purposes;60

and v) a heterostructures of conjugated polymers for photodetectors.48

3.2. Fabrication of Nanoelectrodes for Electrodeposition

One of the challenges in using patterned nanoelectrodes for electrochemistry is

connecting the nanostructures to an electrometer in a way that blocks the conductive

substrate from the solution, so that only the patterned nanostructures are exposed to the

electrochemical solution. In the first report of nanoskiving, Xu et al. fabricated an array

of metallic nanowires embedded in an epoxy slab and placed it on a substrate bearing a

gold film. The gold film was in contact with the underside of the metallic nanostructures.

The epoxy slab covered the conductive substrate, so that only the surfaces of the gold

nanowires were exposed to the solution. Figure 1.4a is a schematic drawing of array of

gold nanowires, which functioned as the working electrode in the electrochemical cell.

Figures 1.4b and 1.4c show these nanowires, 2 µm long × 50 nm wide, before and after

electrodeposition of additional gold. This experiment was the first demonstration of the

electrical continuity of nanowires fabricated by nanoskiving.

17

Figure 1.4. (a) Apparatus used to electrodeposit gold on nanowire electrodes shown in

(b). An image of the nanowires after electrodeposition is shown in (c). Reproduced in

part with permission from ref. 39. Copyright 2008, American Chemical Society. (d) A

schematic drawing of three parallel, addressable gold nanowires. (e) Parallel gold

nanowires separated by a 30 nm gap. (f) The same gold nanowires after electrochemical

deposition of polyaniline. (g) Current-voltage characteristics across the gap between

nanowires with and without polyaniline. Reproduced with permission from ref. 59.

Copyright 2008, American Chemical Society. (h) A group of five parallel nanowires of

palladium spanning a 10-µm gap between Au electrodes. (i) The electrical response of the

nanowires in their native state, when exposed to a stream of hydrogen gas, and a control

experiment in which the nanowire was exposed to a stream of compressed air.

18

3.3. Fabrication of Addressable Nanowires Separated by a Nanogap

Parallel nanowires that are separated by a nanoscale gap in the lateral dimension

are useful for a number of applications, in sensing,62,63 as electrodes for

dielectrophoresis64,65 and electrochemistry,66 in molecular electronics,67,68 and as a

platform for interrogating phenomena that occur over the nanoscale in charge transport35

or plasmonics.69 There are few methods of generating such nanowire electrodes.

Electron-beam lithography and FIB lithography and milling are two such methods, but, in

addition to the usual drawbacks of high cost, inconvenience, and low throughput, it is

difficult to produce nanoscale gaps over large lengths. It is nearly impossible in an

academic laboratory to use photolithography to contact structures that are separated by

fewer than 100 nm. A simple method to produce transferrable electrodes could be a tool

of significance in discovery-driven nanoelectronic research.

We fabricated transferrable, parallel electrodes bearing a nanogap using a

combination of micromolding in capillaries, physical vapor deposition, and nanoskiving.

Figure 1.4d is a schematic drawing of three parallel nanowires that are separated by gaps

< 100 nm in the parallel region and > 10 µm in the diverging, addressable region.59 We

demonstrated the electrical continuity of these nanowires by electrodepositing the

conducting polymer, PPy, in the nanogap. We addressed the nanowires individually using

low-resolution (10 µm) photolithography using a printed transparency mask in contact

mode. Both nanowires together served as the working electrode. Figures 1.4e and 1.4f

show two parallel nanowires, separated by a gap of 30 nm, before and after

electrodeposition. Figure 1.4g shows the current-voltage (I-V) characteristics of the gap

between the junction with and without the conductive polymer.59

19

3.4. Fabrication of Chemoresistive Nanowires of Palladium

Nanowires are well-suited for chemical sensing, because they have a high ratio of

area to volume; this feature permits rapid diffusion of an analyte into and out of a wire

(or adsorption/desorption from its surface),70,71 and rates of response and recovery that

are superior to those of devices based on thin films or fibrous networks. Palladium is an

attractive material for nanoelectronic devices because of its resistance to oxidation,

reproducible loss in conductivity (chemoresistivity) upon absorption of H2, and is soft

enough to be sectioned with the ultramicrotome without fragmentation. Penner and

coworkers have fabricated and studied palladium nanowires and their characteristics as

hydrogen gas sensors.72 These nanowires can be prepared by templated

electrodeposition,63 or by step-edge decoration of highly oriented pyrolytic graphite, or

other templates.73 We have fabricated palladium nanowires with rectangular cross

sections and high pitch using a combination of iterative template stripping,74 followed by

nanoskiving.60 Figure 1.4h is a SEM image of the 10-µm span of five palladium

nanowires between gold electrodes (w = 60 nm, h = 80 nm, each).

To characterize the nanowires electrically, we tested them for function as sensors

for hydrogen gas. We began by evaporating two Au contact pads on the nanowires

through a stencil mask, which defined a span of the nanowires of 10 µm. We then etched

the epoxy matrix with an air plasma to free the sidewalls of the nanowires. Figure 1.4i

shows three plots of current density vs. applied voltage. The first, “native”, represents the

conductivity of the nanowires in the ambient atmosphere of the laboratory. The second,

“H2”, shows the lower conductivity of the nanowire when exposed to a stream of

20

hydrogren gas. The third, “ctrl”, is a control experiment, in which we exposed the

nanowires to a stream of air. The control experiment and the native conductivity of the

nanowires yielded identical electrical characteristics. From the value of current at 10 mV

and the dimensions of the nanowires, we calculated a conductivity of 2.4 x 104 Ω-1cm-1.

For comparison, the conductivity of bulk palladium is 9.5 x 104 Ω-1cm-1.

3.5. Fabrication of Chemoresistive Conjugated Polymer Nanowires

Organic semiconductors are a class of materials whose properties such as

reversibly oxidation/reduction and modifiable conductivity by electrical gating render

them attractive materials for chemical and biological sensors.75 Incorporation of

molecular recognition elements into semiconducting, conjugated polymer nanowires is

relatively straightforward by synthesis, while modifications of carbon nanotubes and

inorganic nanowires require functionalization of the surfaces, carried out post-

fabrication.76 Other possible uses for conjugated polymer nanowires are as tools for

studying one-dimensional charge transport,77 or as field-effect transistors,78 actuators79 or

interconnects.80

There is not yet a truly general technique for the fabrication of conjugated

polymer nanowires. Examples of methods that satisfy some of the criteria of cost,

accessibility, and generality to different materials, are electrodeposition in templates,63

dip-pen nanolithography,81 nanoimprint lithography,82 and electrospinning.83

Electrospinning is particularly versatile. Craighead and coworkers have used scanned

electrospinning84 to deposit single nanowires of polyaniline85 and poly(3-

hexylthiophene)78 on a rotating substrate, while Xia and coworkers have developed an

21

approach to deposit uniaxial collections of nanofibers of a range of inorganic and organic

materials.86,87

Using a procedure that involves stacking spin-coated films of conjugated

polymers, followed by nanoskiving, it was possible to obtain nanowires with rectangular

cross sections individually, in bundles, or in parallel with high pitch.61 We began by spin-

coating two conjugated polymer, poly(benizimidazobenzophenantrholine ladder) (BBL)

and poly(2-methoxy-5-(2’-ethylhexyloxy)-1,4-phenylenevinylene) (MEH-PPV)

alternately on the same substrate, such that fifty 100-nm layers of BBL were separated by

fifty 100-nm layers of MEH-PPV. Release of this free-standing, 10-µm-thick film, and

subsequent sectioning, provided a cross section which bore 100-nm-wide strips of the

two conjugated polymers. Etching the MEH-PPV with an air plasma left behind parallel

BBL nanowires (Figures 1.5a and 1.5b), and dissolving BBL with methanesulfonic acid

left behind MEH-PPV nanowires (Figures 1.5c and 1.5d). Figure 1.5e is a plot of current

density vs. voltage (J-V) of a group of MEH-PPV nanowires, of the type shown in Figure

1.5d, when exposed to I2 vapor.

3.6. Arranging Nanowires of Different Types Using Magnetic Mooring

One of the central challenges in promoting discoveries of nanoscience into

technological applications is the ability to manipulate and position nanostructures on a

surface. We refer specifically to nanostructures fabricated by bottom-up methods, such as

solution-phase synthesis52 or vapor-liquid-solid growth56 (structures fabricated by top-

down procedures are usually formed exactly where they are desired). The usual

procedure to interrogate a nanostructure is to deposit structures randomly, and then to

22

Figure 1.5. (a) BBL nanowires before (bottom right) and after (top left) etching the

sacrificial polymer, MEH-PPV, with an air plasma. The unetched region was protected

with a conformal slab of PDMS. (b) A group of 50 parallel BBL nanowires with 200 nm

pitch. (c) A single MEH-PPV nanowire. (d) A group of fifty MEH-PPV nanowires. (e) A

J-V plot of the nanowires in (d) in the presence of I2 vapor. (f) Three nanowires in an

electrically continuous junction: a poly(3-hexylthiophene) (P3HT) nanowire spanning a

gap between gold nanowires. (g) A I-V plot of the junction shown in (f) in the presence of

I2 vapor. Reproduced with permission from refs. 61 (a – e) and 60 (f, g). Copyright 2008

and 2009, American Chemical Society. (h) A heterojunction of conjugated polymers

sandwiched between two electrodes with asymmetric work functions. (i) Upon irradiation

with white light, the junction produces a photovoltaic effect. Reproduced with permission

from ref. 48. Copyright 2008, Wiley-VCH Verlag GmbH & Co. KGaA.

23

select a serendipitously positioned structure, for example, a nanowire spanning two

electrodes or sitting on an optical waveguide in the proper orientation. The more elements

a system has—e.g., nanowires, waveguides, electrodes, and quantum dots—the lower the

probability that random assembly can generate a desired geometry. This section focuses

on one-dimensional structures3 but the processes we describe would be applicable to

other structures as well.

There are several methods of aligning nanowires in groups and individually.

Methods to align nanowires in group include shear alignment of nanowires suspended in

fluids,88-92 including wafer-scale alignment in bubble-blown films;93 brushing

suspensions of nanowires over a lithographically patterned substrate to create highly

aligned regions of nanowires on exposed areas;94 and alignment of nanowires in a

Langmuir-Blodgett trough.95 Methods of positioning single nanowires include optical

tweezing96 and opto-electronic tweezing;97 methods of manipulation by direct contact

with scanning probe tips98 and micromanipulators;99 and electrophoretic alignment of

over pre-patterned electrodes.100

We used a procedure that combined nanoskiving with non-contact, magnetic

manipulation of the polymeric slabs containing particles of nickel as well as nanowires of

metals and polymers. This process used the slabs as physical tethers to connect the

nanowires of interest to the sacrificial nickel particles (in the form of strips and powder).

We transferred these slabs to a substrate, along with ~5 µL of water. The slabs floated on

the pool of water and were thus mobile under the influence of an external permanent

magnet attached to a micromanipulator. As the pool of water evaporated, the slabs, along

24

with the nanostructures they contained, adhered to the substrate, with an average

deviation from the intended position of 16 µm.

To show that it was possible to form electrically continuous junctions between

nanowires of different types, we placed a single nanowire of poly(3-hexylthiophene)

(P3HT) across two parallel gold nanowires. This geometry could be useful in measuring

nanoscale charge transport in optoelectronic polymers, and in the fabrication of chemical

sensors101 or field-effect transistors based on single nanowires.78 Poly(3-hexylthiophene)

undergoes an insulator-to-metal transition upon exposure to I2.102 We deposited two

parallel Au nanowires, which were embedded in the same epoxy slab. Separately, we

fabricated a P3HT nanowire (100 nm × 100 nm cross section), co-embedded with nickel

powder in epoxy, positioned it to span the 50-µm gap between Au nanowires (Figure

1.5f). Figure 1.5g is an I-V plot of the nanowire when exposed to I2 (“doped”) and in the

absence of the I2 (“undoped”). It should also be possible to use this technique for four-

terminal measurements, which would allow decoupling of the contact resistance from the

true resistance of a nanowire.70

3.7. Fabrication of an Ordered Bulk Heterojunction of Conjugated Polymers

Nanoskiving is one of a few techniques of nanoscale patterning whose features

can be made of different materials and whose components touch in the lateral dimension.

Forming densely packed features that are touching in the lateral dimension has significant

potential to address a long-standing problem in organic photovoltaic cells, which require

two organic semiconductors (e-donor and e-acceptor) to form an intimate heterojunction

on the scale of approximately 10 nm.103 A persistent challenge in fabricating these

25

devices is that the distance an exciton can travel before it decays (the exciton diffusion

length, or LD) is about 10 times shorter than the thickness of material required for

efficient absorption of photons (100 to 200 nm). The architecture that satisfies the

requirements of both LD and the thickness for optimal absorption of light is known as the

ordered bulk heterojunction.104 It has a cross section of p-type and n-type phases that is

intermixed on the length scale of LD and is 100 to 200 nm thick.

We used nanoskiving to fabricate an ordered bulk heterojunction of two

conjugated polymers. The process had three steps: i) spin-coating a composite film with

100 alternating layers of BBL (e-acceptor) and MEH-PPV (e-donor); ii) rolling this

multilayer film into a cylinder (a “jelly roll”); and iii) nanoskiving the jelly roll (Figure

1.5h).48 The cross-section of a slab of the jelly roll has an interdigitated arrangement of

the two polymers. The thickness of the slab is determined by the ultramicrotome and the

spacing between the two materials is determined by spin-coating. We placed a slab of this

structure between two electrodes with asymmetric work functions (Figure 1.5h), tin-

doped indium oxide (ITO) and eutectic gallium-indium (EGaIn), the heterostructures

exhibited a photovoltaic response under white light (1.5i). Selective excitation of BBL

with red light confirmed that the photovoltaic effect was the result of photoinduced

charge transfer between BBL and MEH-PPV. Although the power conversion efficiency

of these devices were low (< 0.1%), we believe that this approach to fabricating

donor/acceptor heterojunctions could be useful in photophysical studies, and might

ultimately suggest new approaches to OPV devices.48

4. Optical Applications of Nanoskiving

26

4.1. Introduction

Metallic nanostructures with well-defined geometries are the building blocks of

the branch of optics known as plasmonics.105,106 A surface plasmon is a quantum of

oscillation of charge at a metal-dielectric interface, driven by electromagnetic radiation.

Localized surface plasmon resonances can be excited in nanoparticles, whose dimensions

are much smaller than the wavelength of excitation. The energy of the LSPR is a function

of the size and shape of the particle, and its dielectric environment.105 Applications of

plasmonic nanoparticles include optical filters;107,108 substrates for optical detection of

chemical and biological analytes using LSPRs109 or surface-enhanced Raman scattering

(SERS);110-112 substrates for enhanced luminosity;113 materials to augment absorption in

thin-film photovoltaic devices;114 metamaterials8,115 with negative magnetic

permeabilities116 and refractive indices;117 and materials for perfect lenses,118 and

invisibility cloaking.119

The most sophisticated arrays of plasmonic structures are fabricated using

EBL,120 FIB,117 or direct laser writing.121 There are also a number of chemical, soft

lithographic, and other unconventional approaches to producing plasmonic materials.106

Solution-phase synthesis can produce single-crystalline metallic structures of different

shapes and materials.52,122 Nanosphere lithography, pioneered by Van Duyne and

coworkers, uses self-assembled spheres as a stencil mask, in which the void spaces

between the spheres direct the deposition of metal on the substrate by evaporation.123,124

Rogers, Odom, Nuzzo, and coworkers have used soft lithographic techniques, such as

patterning photoresists with conformal phase-shifting masks,125 as well as soft

nanoimprint lithography,126 to form arrays of nanoholes in metallic films127 and

27

pyramidal shells.128 This section describes the use of nanoskiving to generate

nanostructures for a variety of optical applications.

4.2. Fabrication of Gold Nanowires and Size-Dependent Surface Plasmon Resonance

The combination of patterning or molding, thin-film deposition, and sectioning

can control each dimension of the structures produced by nanoskiving. Nanoskiving can

also produce large numbers of identical structures in a single array.39 Plasmonic

applications—e.g., sensors based on changes in the frequency of LSPRs, optical

polarizers, filters—require uniform absorption across arrays of particles. Monodisperse

particles satisfy this requirement, while groups of polydisperse particles absorb broadly.2

Figure 1.6a summarizes the method of fabrication used to form collinear arrays of

identical nanowires of gold.129 The key result of the process is that each dimension is

determined precisely: the length (x) by photolithography, the width (y) by the thickness of

the evaporated film of gold, and the height (z) by the set thickness of the ultramicrotome.

The cross sections were as thin as 10 nm × 30 nm; all nanowires were 2 µm long.

Illumination of groups of these nanowires excites plasmon resonances along their

transverse (y) axes. In order to test the optical homogeneity of the nanowires, Xu et al.

collected the spectra of four nanowires individually. The nanowires exhibited

overlapping resonances, which implied that they were geometrically monodisperse.

Figure 1.6b shows the darkfield images of these nanowires and the colors they scattered

as a function of height (z). Scattered light was passed through a polarizer perpendicular to

the long axes of the nanowires. There was a red shift in the peak of the scattered intensity

with increasing height (Figure 1.6c). This observation was consistent with finite-

28

Figure 1.6. (a) Summary of the procedure used to fabricate gold nanowires with

controlled dimensions. (b) Dark-field image of nanowires showing scattered light. The

energies at which the nanowires resonate is a function of the cross sectional dimensions

of the nanowires. (c) Spectra of the samples shown in (d). Reproduced with permission

from ref. 129. Copyright 2006, Wiley-VCH Verlag GmbH & Co. KGaA.

29

difference time-domain (FDTD) simulations. The ability to tune the size, shape, and

composition of metallic structures is a useful capability of nanoskiving for optical

applications.129

4.3. Fabrication of Single-Crystalline Gold Nanowires for Plasmonic Waveguiding

Nanophotonic devices, including photonic integrated circuits, require

waveguiding of optical energy in sub-wavelength dimensions.7,130 Patterned metal strips

or can guide light using SPPs, but efforts to produce efficient plasmonic waveguides from

these structures have been hindered by the fact that the surfaces of polycrystalline

evaporated films are too rough to support propagation of the lengths needed.131 Recently,

Ditlbacher and coworkers,132 and others,133 have shown that high quality, single-

crystalline silver nanowires can confine the energies of incident photons to propagating

surface plasmon polaritons (SPPs), which travel along the longitudinal axes of the

nanowires (in Section 3.2, LSPRs were excited along the transverse axes). For example,

microfabricated strips of silver exhibited propagation lengths of 2.5 µm, where single

crystalline silver nanowires have propagation lengths of 10 µm, due to the low loss of the

smooth surface.132 Gold nanowires might be superior to silver nanowires, because gold

does not oxidize under ambient conditions, but studies of SPPs along gold nanowires had

not been performed, in part because the synthesis of silver nanowires was well

established.134

Using a procedure that combined chemical synthesis of microplates of gold135 and

nanoskiving (Figure 1.7a), we were able to produce collinear arrays of high quality,

single crystalline nanowires.47 (Figure 1.7b shows a dark-field optical image and 1.7c

30

Figure 1.7. (a) Schematic representation of the procedure used to deposit, embed, and

section gold microplates into nanowires. (b) Dark-field and (c) SEM images of a group of

colinear single-crystalline nanowires. (d) Schematic drawing of the orientation of the

nanowire on a prism with respect to the wave vector (k) of the impinging white light. (e)

Spectra of scattered light from both the input and output tips of the nanowires. The

minima of the input and the maxima of the output intensities correspond to the

wavelengths of maximum coupling into the nanowire (Fabry-Perot resonance). (f)

Scattering spectra of a polycrystalline nanowire fabricated by photolithography,

evaporation, and nanoskiving. (g) Two nanowires placed perpendicular to each other. (h)

An optical micrograph of a nanowire placed on a microfabricated waveguide. The inset is

an SEM close-up.

31

shows and SEM image.) In order to determine if these nanowires could be used to

confine and guide light using SPPs, we mounted a nanowire on a prism and illuminated it

with unpolarized white light under total internal reflection. We oriented the nanowire

parallel to the evanescent wave generated by at the surface of the prism (Figure 1.7d). We

observed light scattering from the ends of the nanowire. Figure 1.7e shows the spectra of

scattered light from the input and output tips of the nanowire. The minima of the input

spectrum and the maxima of the output spectrum corresponded to wavelengths at which

maximum coupling of the light into the nanowire occurred due to constructive

interference of the SPP modes reflecting back and forth between the two tips. These

wavelengths correspond to those that reproduced themselves after a full round trip. The

light did not scatter from the center of the nanowire because the wave vector of the

surface plasmon is higher than that of light in air.47

Using magnetic mooring (Section 3.6), it was possible to position and orient

nanowires on top of each other (Figure 1.7f) or on top of topographic features, such as

microfabricated polymeric waveguides (Figure 1.7g), by floating the slabs containing

these nanowires in a pool of water and positioning them with an external permanent

magnet controlled by a micromanipulator.60 We positioned these nanowires with an

average center-to-center deviation of 16 µm, and orientational deviation of 3°.60 This

process could be used to generate more complex arrangements of elements in order to

produce multicomponent photonic devices comprising, for example, photonic and

plasmonic waveguides,130 semiconductor nanowires,57 and single-photon emitters.7

4.4. Fabrication of 2D Arrays of Nanostructures

32

Sections 3.2 and 3.3 focused on one-dimensional nanostructures. Applications

such as optical filters,107,108 substrates for surface-enhanced Raman spectroscopy,8,110 and

metamaterials115,117,120,124 required two-dimensional arrays of nanostructures. Using a

procedure that combined molding an epoxy substrate by soft lithography, thin-film

deposition, embedding, and sectioning parallel to the plane of topography, we were able

to produce two-dimensional arrays of nanostructures using nanoskiving. Figure 1.8 shows

an example of this process. First, we molded an array of epoxy nanoposts by soft

lithography. This array was coated conformally with gold by sputter coating, then coated

by polypyrrole (PPy) using electrochemical growth, then coated a second time with gold.

These procedures produced an array of four-layered, coaxial nanoposts with radial

symmetry. When embedded in epoxy and sectioned into slabs, the slabs contained

radially symmetric discs of epoxy, gold, PPy, and gold. These arrays could be transferred

to essentially any substrate. An optional step was to etch the organic components using

an air plasma. Etching left behind arrays of free-standing, concentric rings of gold.

Figures 1.9a – 1.9j show a series of structures fabricated by this and related procedures.

There are at least five important aspects of the structures produced that cannot be

replicated easily, if at all, with other techniques: i) the linewidths of the structures are

determined by the thickness of the thin film, not the dimensions of the original

topographic master; ii) the height of the structures can be tuned over a large range (80 nm

– 2 µm demonstrated in Figure 1.9d and 1.9e), simply by changing the set thickness on

the ultramicrotome; iii) the structures can be composed of two or more materials in the

same plane, without the need for multiple steps of patterning; iv) the components can be

in physical contact in the lateral dimension; and v) many slabs may be obtained from a

33

Figure 1.8. Summary of the procedure used to fabricate concentric rings by thin film

deposition and thin sectioning of high-aspect-ratio nanoposts. Sputter-coating produced a

film of Au on an array of epoxy nanoposts (step 1). This film served as the working

electrode for the conformal electrodeposition of PPy (step 2). A second sputter-coating

provided a nanopost array with a core-shell-core-shell composition (step 3). Embedding

this structure in additional epoxy formed a block (step 4). Sectioning this block with the

ultramicrotome yielded an epoxy slab containing the nanostructures (step 5). This

structure could be transferred from the water bath on which the nanostructures float to

any substrate (not shown). Treatment with an air plasma simultaneously etched the epoxy

matrix and the PPy in between the Au rings (step 6).

34

Figure 1.9. Scanning electron micrographs of two-dimensional arrays of nanostructures.

(a) An array of nanoposts coated with gold. (b) Gold nanorings. (c) Double rings of gold

separated by a layer of polypyrrole. (d) Double rings of gold after etching the organic

components with an air plasma. (e) Coaxial cylinders of gold obtained by cutting a 2-µm-

thick slab of the sample like that from which (d) was derived. (f) Counterfacing,

concentric rings of gold. The array contains a mixture of the two structures shown in the

inset. (g) Concentric cylinders of gold. The cylinders are segmented because the

sidewalls of the silicon master were scalloped due to the Bosch process of deep reactive-

ion etching. (h) Counterfacing crescents of silver and silicon. (i) Three-layer crescents of

gold on the inside, silicon dioxide in the middle, and palladium on the outside. (j) Rings

of lead sulfide nanocrystals obtained by sectioning an array of epoxy nanoposts drop-cast

with a solution of the crystals in hexanes.

35

Figure 1.9 (Continued)

36

single embedded structure (we have produced as many as 60 consecutive cross sections,

100 nm thick, from a single embedded array of 8-µm, gold-coated epoxy nanoposts).

4.5. Plasmonic Properties of Two-Dimensional Arrays of Nanostructures

In Section 4.2, we described visible LSPRs in along the short axes of gold

nanowires. It is also possible to excite LSPRs around the perimeters of rings, split-rings,

and L-shaped structures.136 Figure 1.10a shows a spectrum in the near-to-mid IR of two

different samples: a 2D array of rings with d = 330 nm (dotted line), and an array of

concentric rings where the inner ring had d = 330 nm, and the outer ring had d = 730 nm.

The single ring produced one resonance around 2.5 µm. The double ring produced two

resonances (solid curve), where the resonance due to the inner ring was higher in energy

than the single ring of the same dimensions, because of coupling between the two rings.

Our experimental observations were consistent with FDTD simulations (Figure 1.10b).

These structures behaved as short-wavelength IR frequency-selective surfaces.

The arrays of rings were isotropic, and thus the resonances were not dependent on

polarization. In order to test the ability to produce anisotropic structures for polarization-

dependent applications, Xu et al. produced the L-shaped structures in Figure 1.10c by a

combination of molding and line-of-sight deposition of gold.136 Figure 1.10d shows the

resonance of the array in response to unpolarized light. It consists of two distinct modes.

The mode at 8.4 µm was due to an oscillation in a line that connects the two termini of

the L, and was excited by linearly polarized light that was parallel to that line (Figure

1.10e). The mode at 4.8 µm was excited by light polarized perpendicular to the line that

connects the two ends of the L (Figure 1.10f). This polarization bisected the structure

37

Figure 1.10. Infrared spectra of two-dimensional arrays of metallic nanostructures. (a)

Comparison of spectra between single rings (dotted line) and double rings (solid line). (b)

Finite-difference time-domain simulations of the structures measured in (a). (c) Dark-

field image of an array of L-shaped structures. The inset is an SEM of an individual

structure. (d) Mid-IR transmission spectrum of the array excited using unpolarized light.

(e, f) Transmission spectra of the structures when the polarization is oriented (e) parallel

and (f) perpendicular to the line connecting the termini of the L-shaped structures.

and produced two orthogonal regions that oscillated in phase. These observations were

again consistent with FDTD simulations.136

38

4.6. Integration of Plasmonic Arrays with Optical Fibers

The thin slab of epoxy in which the structures produced by nanoskiving are

embedded provides a visible handle to transfer arrays to substrates.46 There is, thus, a

major challenge in optics to which nanoskiving seems particularly (perhaps uniquely)

well-suited—modifying the cleaved facets of optical fibers with arrays of nanostructures.

The ability to control the emission from fibers using filters or polarizers, or the

fabrication of sensors for in situ, label-free detection of chemical or biological analytes

using either SERS137 or LSPRs,138 are possible applications of modified optical fibers.139

Attachment of plasmonic arrays to the cleaved facets of fibers is not straightforward by

conventional means, however. Photolithographic patterning of the facets of fibers would

require deposition, exposure and development of photoresist on a small area (d ~ 100

µm).23 Examples of unconventional methods to integrate plasmonic elements with optical

fibers include anisotropic chemical etching, to form arrays of sharp cones,140 and

transferring gold structures fabricated by EBL from a surface to which gold adhered

weakly.111

Figure 1.11a summarizes the procedure we used to mount arrays of metallic

nanostructures on the facets of fibers. We were able to do so by submerging the floating

slabs by pressing down, from above, with the tip of a fiber. This action submerged the

slab, which became attached to the tip of the fiber as we withdrew it from the water-filled

trough of the ultramicrotome. After allowing the water on the tip of the fiber to

evaporate, exposure to an air plasma using a bench-top plasma cleaner (1 torr, 100 W, 10

min) left behind the nanostructures only on the facet of the fiber (Figures 1.11b and

1.11c).

39

Figure 1.11. (a) Schematic illustration of the procedure used to transfer arrays of

plasmonic nanostructures to the cleaved facet of an optical fiber. (b) A SEM image of a

facet of an optical fiber bearing a square array of gold crescents. (c) A facet bearing a

grating of parallel gold nanowires.

40

5. Conclusions

Nanoskiving is simple and inexpensive method of generating nanostructures for

applications in optics and electronics, though the optical applications are presently more

developed than are the electronic ones. Nanoskiving is an unconventional approach to

patterning structures that introduces “cutting” as a step of replicating patterns, which is

analogous to “printing” and “molding” in soft lithography, and “exposure” in

photolithography. While the most complex structures produced by nanoskiving require a

topographically patterned template, the nanoscale dimensions do not come from this

template. Rather, they correspond to the thicknesses of the evaporated, sputter-coated,

spin-coated, or chemically deposited thin films. In this sense, nanoskiving serves

simultaneously as a technique of mastering and replication of new nanoscale information.

This characteristic has no analogue among other techniques of nanofabrication.

Nanoskiving has a low barrier to entry, in terms of initial capital investments and

the learning curve. The first several steps of all of the procedures discussed in this

chapter—e.g., soft lithographic molding and thin-film deposition—are well established in

the literature and widely practiced.141 A user can start generating high-quality

nanostructures after a half-day session of training and fewer than ten hours of practice.

Additional experience increases the speed with which samples can be trimmed, and

sectioned, but the quality of the nanostructures obtained depends most strongly on the

preparation of blocks—molding, deposition, and embedding—rather than on the process

of sectioning itself.

As with any technique, nanoskiving has its disadvantages. It is limited to

generating non-crossing line segments. The technique works best for polymers and

41

metals softer than platinum; extensive fragmentation of brittle materials—hard metals,

crystalline oxides, and some amorphous semiconductors—limits the generality of

nanoskiving (though the number of materials for which nanoskiving does work makes it

nevertheless very general). Nanoskiving is still subject to all of the deleterious artifacts of

ultramicrotomy, the most important of which are scoring and compression. Scoring can

be avoided by working in an environment uncontaminated with hard, microscopic

particles of dust. Compression causes two deleterious effects. First, it distorts square

arrays of nanostructures into rectangular arrays (8.5% compression for UVO-114).

Second, it imposes compressive stress on embedded films that lie parallel to the direction

of cutting, and these segments are prone to fragmentation. The use of oscillating knives

and other methods50,142 will mitigate these deleterious effects as nanoskiving develops.143

There are several future directions of nanoskiving. In optics, the two most salient

are i) the ability to fabricate structures of multiple materials and ii) the integration of

arrays of metallic nanoparticles with optical fibers and other components. The fabrication

of three-dimensional metamaterials is also an area of potential significance for

nanoskiving.117,120 Stacking and laminating structures could be a route toward 3D

materials with different geometries and compositions within or between layers.46,60 Other

areas in which nanoskiving has potential are nanoelectrochemistry,66 patterning nanoscale

magnetic particles for digital storage,144 membranes for size or shape-selective

diffusion,145 devices for energy conversion and storage,146 and patterning functional

surfaces for biology.147

The ability to produce consecutive cross sections—quasi copies—of structures

suggests that thin sectioning could be useful in manufacturing. The most significant

42

impediment to transforming nanoskiving from a technique for research to one of

manufacturing is replacing the manual steps (aligning the embedded structures with the

knife edge and collecting the sections from the water-filled trough) with automated ones.

A recent technological development—reel-to-reel lathing ultramicrotomy—stands out as

potentially useful for high-throughput and large-area nanoskiving.148 Nanoskiving might,

ultimately, suggest new ways of nanomanufacturing by cutting.

6. Acknowledgements

This research was supported by the National Science Foundation under award

PHY-0646094. The authors used the shared facilities supported by the NSF under

MRSEC (DMR-0213805 and DMR-0820484). This work was performed in part using the

facilities of the Center for Nanoscale Systems (CNS), a member of the National

Nanotechnology Infrastructure Network (NNIN), which is supported by the National

Science Foundation under NSF award no. ECS-0335765. CNS is part of the Faculty of

Arts and Sciences at Harvard University. D.J.L. acknowledges a Graduate Fellowship

from the American Chemical Society, Division of Organic Chemistry, sponsored by

Novartis.

References

(1) Murray, C. B.; Kagan, C. R.; Bawendi, M. G. Annu. Rev. Mater. Sci. 2000, 30, 545-610.

(2) Mulvaney, P. MRS Bull. 2001, 26, 1009-1014. (3) Avouris, P. Phys. Today 2009, 62, 34-40. (4) Whitesides, G. M.; Lipomi, D. J. Faraday Disc. 2009, 143, 373-384.

43

(5) Pease, R. F.; Chou, S. Y. Proc. IEEE 2008, 96, 248-270.

(6) Gratton, S. E. A.; Williams, S. S.; Napier, M. E.; Pohlhaus, P. D.; Zhou, Z. L.; Wiles, K. B.; Maynor, B. W.; Shen, C.; Olafsen, T.; Samulski, E. T.; Desimone, J. M. Acc. Chem. Res. 2008, 41, 1685-1695.

(7) Akimov, A. V.; Mukherjee, A.; Yu, C. L.; Chang, D. E.; Zibrov, A. S.;

Hemmer, P. R.; Park, H.; Lukin, M. D. Nature 2007, 450, 402-406.

(8) Cubukcu, E.; Yu, N. F.; Smythe, E. J.; Diehl, L.; Crozier, K. B.; Capasso, F. IEEE J. Sel. Top. Quantum Electron. 2008, 14, 1448-1461.

(9) Gates, B. D.; Xu, Q. B.; Stewart, M.; Ryan, D.; Willson, C. G.;

Whitesides, G. M. Chem. Rev. 2005, 105, 1171-1196. (10) Willson, C. G.; Roman, B. J. ACS Nano 2008, 2, 1323-1328.

(11) Cavin, R. K.; Zhirnov, V. V.; Herr, D. J. C.; Avila, A.; Hutchby, J. J. Nanopart. Res. 2006, 8, 841-858.

(12) Lundstrom, M. Science 2003, 299, 210-211. (13) Gower, M. Microlith. World 2004, 13, 16-17. (14) Ito, H. J. Photopolym. Sci. Technol. 2008, 21, 475-491.

(15) Lopez-Gejo, J.; Kunjappu, J. T.; Zhou, J.; Smith, B. W.; Zimmerman, P.; Conley, W.; Turro, N. J. Chem. Mater. 2007, 19, 3641-3647.

(16) Ronse, K. C. R. Phys. 2006, 7, 844-857. (17) Mack, C. A. IEEE Spectrum 2008, 45, 40-45.

(18) Bratton, D.; Yang, D.; Dai, J. Y.; Ober, C. K. Polym. Adv. Technol. 2006, 17, 94-103.

(19) Berglund, C. N.; Leachman, R. C. IEEE Trans. Semicond. Manuf. 2010,

23, 39-52. (20) Willson, C. G. J. Photopolym. Sci. Technol. 2009, 22, 147-153.

(21) Xia, Y. N.; Kim, E.; Zhao, X. M.; Rogers, J. A.; Prentiss, M.; Whitesides, G. M. Science 1996, 273, 347-349.

(22) Kim, D. H.; Rogers, J. A. Adv. Mater. 2008, 20, 4887-4892.

44

(23) Smythe, E. J.; Dickey, M. D.; Whitesides, G. M.; Capasso, F. ACS Nano 2009, 3, 59-65.

(24) Xia, Y. N.; Whitesides, G. M. Angew. Chem., Int. Ed. 1998, 37, 551-575. (25) Goh, C.; Coakley, K. M.; McGehee, M. D. Nano Lett. 2005, 5, 1545-1549.

(26) Xu, Q. B.; Mayers, B. T.; Lahav, M.; Vezenov, D. V.; Whitesides, G. M. J. Am. Chem. Soc 2005, 127, 854-855.

(27) Xia, Y. N.; McClelland, J. J.; Gupta, R.; Qin, D.; Zhao, X. M.; Sohn, L.

L.; Celotta, R. J.; Whitesides, G. M. Adv. Mater. 1997, 9, 147-149.

(28) Kim, E.; Xia, Y. N.; Zhao, X. M.; Whitesides, G. M. Adv. Mater. 1997, 9, 651-654.

(29) Jeon, N. L.; Choi, I. S.; Xu, B.; Whitesides, G. M. Adv. Mater. 1999, 11,

946-950.

(30) Kane, R. S.; Takayama, S.; Ostuni, E.; Ingber, D. E.; Whitesides, G. M. Biomater. 1999, 20, 2363-2376.

(31) Xia, Y. N.; Whitesides, G. M. Langmuir 1997, 13, 2059-2067.

(32) Biebuyck, H. A.; Larsen, N. B.; Delamarche, E.; Michel, B. IBM J. Res. Dev. 1997, 41, 159-170.

(33) Xia, Y. N.; Qin, D.; Whitesides, G. M. Adv. Mater. 1996, 8, 1015-1017.

(34) Cao, T. B.; Xu, Q. B.; Winkleman, A.; Whitesides, G. M. Small 2005, 1, 1191-1195.

(35) Xue, M. Q.; Yang, Y. H.; Cao, T. B. Adv. Mater. 2008, 20, 596-600.

(36) Shir, D. J.; Jeon, S.; Liao, H.; Highland, M.; Cahill, D. G.; Su, M. F.; El-

Kady, I. F.; Christodoulou, C. G.; Bogart, G. R.; Hamza, A. V.; Rogers, J. A. J. Phys. Chem. B 2007, 111, 12945-12958.

(37) Maria, J.; Jeon, S.; Rogers, J. A. J. Photochem. Photobiol., A 2004, 166,

149-154.

(38) Rogers, J. A.; Paul, K. E.; Jackman, R. J.; Whitesides, G. M. Appl. Phys. Lett. 1997, 70, 2658-2660.

(39) Xu, Q. B.; Rioux, R. M.; Dickey, M. D.; Whitesides, G. M. Acc. Chem.

Res. 2008, 41, 1566-1577.

45

(40) Hill, J. The Construction of Timber; Imperial Academy: London, 1770. (41) Pease, D. C.; Porter, K. R. J. Cell Biol. 1981, 91, S287-S292. (42) Malis, T. F.; Steele, D. Mater. Res. Soc. Symp. Proc. 1990, 199, 3-50.

(43) Bracegirdle, B. M. A History of Microtechnique; Cornell University Press: Ithaca, NY, 1978.

(44) Goldstein, J. N., D.; Joy, D.; Lyman, C.; Echlin, P.; Lifshin, E.; Sawyer,

L.; Michael, J. Scanning Electron Microscopy and X-Ray Analysis; 3 ed.; Springer, 2003.

(45) Acetarin, J. D.; Carlemalm, E.; Kellenberger, E.; Villiger, W. J. Electron

Microsc. Tech. 1987, 6, 63-79. (46) Xu, Q.; Rioux, R. M.; Whitesides, G. M. ACS Nano 2007, 1, 215-227.

(47) Wiley, B. J.; Lipomi, D. I.; Bao, J. M.; Capasso, F.; Whitesides, G. M. Nano Lett. 2008, 8, 3023-3028.

(48) Lipomi, D. J.; Chiechi, R. C.; Reus, W. F.; Whitesides, G. M. Adv. Funct.

Mater. 2008, 18, 3469-3477. (49) Antonovsky, A. Microsc. Res. Tech. 1995, 31, 300-307. (50) Jesior, J. C. J. Ultrastruct. Mol. Struct. Res. 1986, 95, 210-217.

(51) Matzelle, T. R.; Gnaegi, H.; Ricker, A.; Reichelt, R. J. Microsc. Oxford 2003, 209, 113-117.

(52) Xia, Y.; Xiong, Y. J.; Lim, B.; Skrabalak, S. E. Angew. Chem., Int. Ed.

2009, 48, 60-103.

(53) Burda, C.; Chen, X. B.; Narayanan, R.; El-Sayed, M. A. Chem. Rev. 2005, 105, 1025-1102.

(54) Robertson, J. Rep. Prog. Phys. 2006, 69, 327-396. (55) Geim, A. K.; Novoselov, K. S. Nat. Mater. 2007, 6, 183-191.

(56) Xia, Y. N.; Yang, P. D.; Sun, Y. G.; Wu, Y. Y.; Mayers, B.; Gates, B.; Yin, Y. D.; Kim, F.; Yan, Y. Q. Adv. Mater. 2003, 15, 353-389.

46

(57) Falk, A. L.; Koppens, F. H. L.; Yu, C. L.; Kang, K.; Snapp, N. D.; Akimov, A. V.; Jo, M. H.; Lukin, M. D.; Park, H. Nat. Phys. 2009, 5, 475-479.

(58) Xu, Q. B.; Gates, B. D.; Whitesides, G. M. J. Am. Chem. Soc. 2004, 126,

1332-1333.

(59) Dickey, M. D.; Lipomi, D. J.; Bracher, P. J.; Whitesides, G. M. Nano Lett. 2008, 8, 4568-4573.

(60) Lipomi, D. J.; Ilievski, F.; Wiley, B. J.; Deotare, P. B.; Loncar, M.;

Whitesides, G. M. ACS Nano 2009, 3, 3315-3325.

(61) Lipomi, D. J.; Chiechi, R. C.; Dickey, M. D.; Whitesides, G. M. Nano Lett. 2008, 8, 2100-2105.

(62) Ramanathan, K.; Bangar, M. A.; Yun, M. H.; Chen, W. F.; Mulchandani,

A.; Myung, N. V. Nano Lett. 2004, 4, 1237-1239.

(63) Yun, M. H.; Myung, N. V.; Vasquez, R. P.; Lee, C. S.; Menke, E.; Penner, R. M. Nano Lett. 2004, 4, 419-422.

(64) Bezryadin, A.; Dekker, C.; Schmid, G. Appl. Phys. Lett. 1997, 71, 1273-

1275.

(65) Holzel, R.; Calander, N.; Chiragwandi, Z.; Willander, M.; Bier, F. F. Phys. Rev. Lett. 2005, 95.

(66) Murray, R. W. Chem. Rev. 2008, 108, 2688-2720. (67) Tao, N. J. Nat. Nanotechnol. 2006, 1, 173-181.

(68) Weiss, E. A.; Kriebel, J. K.; Rampi, M. A.; Whitesides, G. M. Philos. Trans. R. Soc. London, Ser. A 2007, 365, 1509-1537.

(69) Lassiter, J. B.; Aizpurua, J.; Hernandez, L. I.; Brandl, D. W.; Romero, I.;

Lal, S.; Hafner, J. H.; Nordlander, P.; Halas, N. J. Nano Lett. 2008, 8, 1212-1218.

(70) Liu, H. Q.; Kameoka, J.; Czaplewski, D. A.; Craighead, H. G. Nano Lett.

2004, 4, 671-675.

(71) McQuade, D. T.; Pullen, A. E.; Swager, T. M. Chem. Rev. 2000, 100, 2537-2574.

(72) Walter, E. C.; Favier, F.; Penner, R. M. Anal. Chem. 2002, 74, 1546-1553.

47

(73) Walter, E. C.; Murray, B. J.; Favier, F.; Kaltenpoth, G.; Grunze, M.;

Penner, R. M. J. Phys. Chem. B 2002, 106, 11407-11411.

(74) Weiss, E. A.; Kaufman, G. K.; Kriebel, J. K.; Li, Z.; Schalek, R.; Whitesides, G. M. Langmuir 2007, 23, 9686-9694.

(75) Roberts, M. E.; Sokolov, A. N.; Bao, Z. N. J. Mater. Chem. 2009, 19, 3351-3363.

(76) Ramanathan, K.; Bangar, M. A.; Yun, M.; Chen, W.; Myung, N. V.;

Mulchandani, A. J. Am. Chem. Soc. 2005, 127, 496-497.

(77) Duvail, J. L.; Retho, P.; Fernandez, V.; Louarn, G.; Molinie, P.; Chauvet, O. J. Phys. Chem. B 2004, 108, 18552-18556.

(78) Liu, H. Q.; Reccius, C. H.; Craighead, H. G. Appl. Phys. Lett. 2005, 87. (79) Smela, E. Adv. Mater. 2003, 15, 481-494.

(80) Samitsu, S.; Shimomura, T.; Ito, K.; Fujimori, M.; Heike, S.; Hashizume, T. Appl. Phys. Lett. 2005, 86.

(81) Salaita, K.; Wang, Y. H.; Mirkin, C. A. Nat. Nanotechnol. 2007, 2, 145-

155.

(82) Dong, B.; Lu, N.; Zelsmann, M.; Kehagias, N.; Fuchs, H.; Torres, C. M. S.; Chi, L. F. Adv. Funct. Mater. 2006, 16, 1937-1942.

(83) Greiner, A.; Wendorff, J. H. Angew. Chem., Int. Ed. 2007, 46, 5670-5703.

(84) Kameoka, J.; Czaplewski, D.; Liu, H. Q.; Craighead, H. G. J. Mater. Chem. 2004, 14, 1503-1505.

(85) Liu, H. Q.; Kameoka, J.; Czaplewski, D. A.; Craighead, H. G. Nano Lett.

2004, 4, 671-675.

(86) Li, G.; Shrotriya, V.; Huang, J. S.; Yao, Y.; Moriarty, T.; Emery, K.; Yang, Y. Nat. Mater. 2005, 4, 864-868.

(87) McCann, J. T.; Chen, J. I. L.; Li, D.; Ye, Z. G.; Xia, Y. A. Chem. Phys.

Lett. 2006, 424, 162-166. (88) Stover, C. A.; Koch, D. L.; Cohen, C. J. Fluid Mech. 1992, 238, 277-296.

48

(89) Zhong, Z. H.; Wang, D. L.; Cui, Y.; Bockrath, M. W.; Lieber, C. M. Science 2003, 302, 1377-1379.

(90) Cui, Y.; Lieber, C. M. Science 2001, 291, 851-853.

(91) Huang, Y.; Duan, X. F.; Wei, Q. Q.; Lieber, C. M. Science 2001, 291, 630-633.

(92) Messer, B.; Song, J. H.; Yang, P. D. J. Am. Chem. Soc. 2000, 122, 10232-

10233. (93) Yu, G. H.; Cao, A. Y.; Lieber, C. M. Nat. Nanotechnol. 2007, 2, 372-377.

(94) Fan, Z. Y.; Ho, J. C.; Jacobson, Z. A.; Yerushalmi, R.; Alley, R. L.; Razavi, H.; Javey, A. Nano Lett. 2008, 8, 20-25.

(95) Jin, S.; Whang, D. M.; McAlpine, M. C.; Friedman, R. S.; Wu, Y.; Lieber,

C. M. Nano Lett. 2004, 4, 915-919.

(96) Pauzauskie, P. J.; Radenovic, A.; Trepagnier, E.; Shroff, H.; Yang, P. D.; Liphardt, J. Nat. Mater. 2006, 5, 97-101.

(97) Jamshidi, A.; Pauzauskie, P. J.; Schuck, P. J.; Ohta, A. T.; Chiou, P. Y.;

Chou, J.; Yang, P. D.; Wu, M. C. Nat. Photonics 2008, 2, 85-89.

(98) Postma, H. W. C.; Sellmeijer, A.; Dekker, C. Adv. Mater. 2000, 12, 1299-1302.

(99) Sirbuly, D. J.; Law, M.; Pauzauskie, P.; Yan, H. Q.; Maslov, A. V.;

Knutsen, K.; Ning, C. Z.; Saykally, R. J.; Yang, P. D. Proc. Natl. Acad. Sci. USA 2005, 102, 7800-7805.

(100) Smith, P. A.; Nordquist, C. D.; Jackson, T. N.; Mayer, T. S.; Martin, B.

R.; Mbindyo, J.; Mallouk, T. E. Appl. Phys. Lett. 2000, 77, 1399-1401.

(101) Xue, M. Q.; Zhang, Y.; Yang, Y. L.; Cao, T. B. Adv. Mater. 2008, 20, 2145-2150.

(102) McCullough, R. D. Adv. Mater. 1998, 10, 93-116.

(103) Thompson, B. C.; Frechet, J. M. J. Angew. Chem., Int. Ed. 2008, 47, 58-77.

(104) Coakley, K. M.; McGehee, M. D. Chem. Mater. 2004, 16, 4533-4542. (105) Maier, S. A.; Atwater, H. A. J. Appl. Phys. 2005, 98.

49

(106) Stewart, M. E.; Anderton, C. R.; Thompson, L. B.; Maria, J.; Gray, S. K.;

Rogers, J. A.; Nuzzo, R. G. Chem. Rev. 2008, 108, 494-521. (107) Love, J. C.; Paul, K. E.; Whitesides, G. M. Adv. Mater. 2001, 13, 604-607.

(108) Wu, D. M.; Fang, N.; Sun, C.; Zhang, X.; Padilla, W. J.; Basov, D. N.; Smith, D. R.; Schultz, S. Appl. Phys. Lett. 2003, 83, 201-203.

(109) Sharma, A. K.; Jha, R.; Gupta, B. D. IEEE Sens. J. 2007, 7, 1118-1129.

(110) Kneipp, K.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S. J. Phys. Condens. Matter 2002, 14, R597-R624.

(111) Smythe, E. J.; Dickey, M. D.; Bao, J. M.; Whitesides, G. M.; Capasso, F.

Nano Lett. 2009, 9, 1132-1138.

(112) Liu, X. F.; Sun, C. H.; Linn, N. C.; Jiang, B.; Jiang, P. J. Phys. Chem. C 2009, 113, 14804-14811.

(113) Fort, E.; Gresillon, S. J. Phys. D, Appl. Phys. 2008, 41.

(114) Peumans, P.; Bulovic, V.; Forrest, S. R. Appl. Phys. Lett. 2000, 76, 2650-2652.

(115) Klar, T. A.; Kildishev, A. V.; Drachev, V. P.; Shalaev, V. M. IEEE J. Sel.

Top. Quantum Electron. 2006, 12, 1106-1115.

(116) Pendry, J. B.; Holden, A. J.; Robbins, D. J.; Stewart, W. J. IEEE Trans. Microwave Theory Tech. 1999, 47, 2075-2084.

(117) Valentine, J.; Zhang, S.; Zentgraf, T.; Ulin-Avila, E.; Genov, D. A.;

Bartal, G.; Zhang, X. Nature 2008, 455, 376-U32. (118) Pendry, J. B. Phys. Rev. Lett. 2000, 85, 3966-3969. (119) Pendry, J. B.; Schurig, D.; Smith, D. R. Science 2006, 312, 1780-1782.

(120) Liu, N.; Guo, H. C.; Fu, L. W.; Kaiser, S.; Schweizer, H.; Giessen, H. Nat. Mater. 2008, 7, 31-37.

(121) Gansel, J. K.; Thiel, M.; Rill, M. S.; Decker, M.; Bade, K.; Saile, V.; von

Freymann, G.; Linden, S.; Wegener, M. Science 2009, 325, 1513-1515.

(122) Ahmadi, A.; Ghadarghadr, S.; Mosallaei, H. Opt. Express 2010, 18, 123-133.

50

(123) Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem. B 2001, 105, 5599-5611.

(124) Gwinner, M. C.; Koroknay, E.; Fu, L. W.; Patoka, P.; Kandulski, W.; Giersig, M.; Giessen, H. Small 2009, 5, 400-406.

(125) Paul, K. E.; Zhu, C.; Love, J. C.; Whitesides, G. M. Appl. Opt. 2001, 40,

4557-4561.

(126) Stewart, M. E.; Mack, N. H.; Malyarchuk, V.; Soares, J.; Lee, T. W.; Gray, S. K.; Nuzzo, R. G.; Rogers, J. A. Proc. Natl. Acad. Sci. USA 2006, 103, 17143-17148.

(127) Henzie, J.; Barton, J. E.; Stender, C. L.; Odom, T. W. Acc. Chem. Res.

2006, 39, 249-257.

(128) Lee, J.; Hasan, W.; Stender, C. L.; Odom, T. W. Acc. Chem. Res. 2008, 41, 1762-1771.

(129) Xu, Q. B.; Bao, J. M.; Capasso, F.; Whitesides, G. M. Angew. Chem., Int.

Ed. 2006, 45, 3631-3635.

(130) Pyayt, A. L.; Wiley, B.; Xia, Y. N.; Chen, A.; Dalton, L. Nat. Nanotechnol. 2008, 3, 660-665.

(131) Nagpal, P.; Lindquist, N. C.; Oh, S. H.; Norris, D. J. Science 2009, 325,

594-597.

(132) Ditlbacher, H.; Hohenau, A.; Wagner, D.; Kreibig, U.; Rogers, M.; Hofer, F.; Aussenegg, F. R.; Krenn, J. R. Phys. Rev. Lett. 2005, 95.

(133) Allione, M.; Temnov, V. V.; Fedutik, Y.; Woggon, U.; Artemyev, M. V.

Nano Letters 2008, 8, 31-35.

(134) Graff, A.; Wagner, D.; Ditlbacher, H.; Kreibig, U. Eur. Phys. J. D 2005, 34, 263-269.

(135) Kan, C. X.; Zhu, X. G.; Wang, G. H. J. Phys. Chem. B 2006, 110, 4651-

4656.

(136) Xu, Q. B.; Bao, J. M.; Rioux, R. M.; Perez-Castillejos, R.; Capasso, F.; Whitesides, G. M. Nano Lett. 2007, 7, 2800-2805.

(137) Lucotti, A.; Zerbi, G. Sens. Actuators, B 2007, 121, 356-364.

51

(138) Sharma, A. K.; Jha, R.; Gupta, B. D. Ieee Sensors Journal 2007, 7, 1118-1129.

(139) Leung, A.; Shankar, P. M.; Mutharasan, R. Sens. Actuators, B 2007, 125,

688-703.

(140) Guieu, V.; Talaga, D.; Servant, L.; Sojic, N.; Lagugne-Labarthet, F. J. Phys. Chem. C 2009, 113, 874-881.

(141) Qin, D.; Xia, Y. N.; Whitesides, G. M. Nat. Protoc. 2010, 5, 491-502. (142) Jesior, J. C. J. Ultrastruct. Res. 1985, 90, 135-144. (143) Studer, D.; Gnaegi, H. J. Microsc. Oxford 2000, 197, 94-100.

(144) Shi, J.; Gider, S.; Babcock, K.; Awschalom, D. D. Science 1996, 271, 937-941.

(145) McKeown, N. B.; Budd, P. M. Chem. Soc. Rev. 2006, 35, 675-683. (146) Shah, K.; Shin, W. C.; Besser, R. S. Sens. Actuators, B 2004, 97, 157-167.

(147) Weibel, D. B.; DiLuzio, W. R.; Whitesides, G. M. Nat. Rev. Microbiol. 2007, 5, 209-218.

(148) Kasthuri, N.; Hayworth, K.; Tapia, J. C.; Schalek, R.; Nundy, S.;

Lichtman, J. W. Soc. Neurosci. Abstr. 2009.

Chapter 2

Survey of Materials for Nanoskiving and Influence of the Cutting Process on the

Nanostructures Produced

Darren J. Lipomi,1 Ramses V. Martinez,1 Robert M. Rioux,2 Ludovico Cademartiri,1

William F. Reus,1 and George M. Whitesides1

1Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street,

Cambridge, MA, 02138

2Department of Chemical Engineering, 158 Fenske Laboratory, Pennsylvania State

University, University Park, PA 16802-4400

53

Abstract

This paper examines the factors that influence the quality of nanostructures

fabricated by sectioning thin films with an ultramicrotome (“nanoskiving”). It surveys

different materials (metals, ceramics, semiconductors, and conjugated polymers),

deposition techniques, (evaporation, sputter deposition, electroless deposition, chemical-

vapor deposition, solution-phase synthesis, and spin-coating), and geometries (nanowires

or two-dimensional arrays of rings and crescents). It then correlates the extent of

fragmentation of the nanostructures with the composition of the thin films, the methods

used to deposit them, and the parameters used for sectioning. There are four major

conclusions. i) Films of soft and compliant metals (those that have bulk values of

hardness less than or equal to those of palladium, or ≤ 500 MPa) tend to remain intact

upon sectioning, while hard and stiff metals (those that have values of hardness greater

than or equal to those of platinum, or ≥ 500 MPa) tend to fragment. ii) All conjugated

polymers tested form intact nanostructures. iii) The extent of fragmentation is lowest

when the direction of cutting is perpendicular to the exposed edge of the embedded film.

iv) The speed of cutting—from 0.1 mm/s to 8 mm/s—has no effect on the frequency of

defects. Defects generated during sectioning include scoring from defects in the knife,

delamination of the film from the matrix, and compression of the matrix. The materials

tested were: aluminum, titanium, nickel, copper, palladium, silver, platinum, gold, lead,

bismuth, germanium, silicon dioxide (SiO2), alumina (Al2O3), tin-doped indium oxide

(ITO), lead sulfide nanocrystals, the semiconducting polymers MEH-PPV, P3HT, and

BBL, and the conductive polymer PEDOT:PSS.

54

Introduction

This paper surveys materials for use in nanoskiving—a process of fabrication

whose key step is sectioning planar or topographically patterned thin films with an

ultramicrotome (Figure 2.1),1,2 and correlates their composition and method of deposition

with the qualities of structures produced (e.g., morphology and extent of fragmentation).

We examined two types of structures—long nanowires and two-dimensional (2D) arrays

of circular or semicircular particles—formed by sectioning thin films for defects due to

intrinsic mechanical properties of the materials (e.g., brittleness). We also explored the

origin of defects due to artifacts of the sectioning process, and conclude that compression

of the matrix, scoring due to chips in the edge of the knife, and delamination of the film

from the matrix are the most important. Nanoskiving is useful for generating

nanostructures for applications in electronics3 and optics,4,5 and has the potential to be

used for manufacturing certain types of structures (e.g., nanowires and 2D arrays of

nanostructures). The information in this paper will be useful to those concerned with

designing processes for fabrication based on nanoskiving.

Background

Unconventional Nanofabrication. “Nanofabrication” generates structures with

sizes of ≤ 100 nm in at least one lateral dimension.6,7 Most commercial nanofabrication

takes place in the semiconductor industry, where the high-precision tools of electron-

beam writing and photolithography dominate, and are likely to continue to do so for the

foreseeable future.8,9 Many new applications for nanostructures, particularly in

photonics,10 chemical11 and biological sensing,12,13 materials for energy conversion and

55

Figure 2.1. Schematic representation of the defining step of nanoskiving: mechanical

sectioning of microfabricated structures with an ultramicrotome to replicate arrays of

nanostructures embedded in thin polymeric slabs.

56

storage,14,15 and organic electronic devices16 require structures that are simpler, and less

expensive, than integrated circuits. Fabricating devices on non-planar, mechanically

compliant, or disposable substrates could lead to important applications that motivate the

development of “unconventional” methods of nanofabrication.10 Soft lithographic

printing and molding, and other methods, have begun to fill these niches in research

laboratories, and are expected to play important roles in manufacturing in the future.7

Nanoskiving. Nanoskiving is a technique for replicating nanostructures that

substitutes “cutting” for “printing” or “molding”. The process builds on two well-

established technologies: i) methods of forming thin films (e.g., physical- and chemical-

vapor deposition) and ii) patterning relief structures in polymers by molding (e.g., replica

molding17,18 and nanoimprint lithography19,20). There are two general strategies to form

nanostructures by nanoskiving: sectioning perpendicular to a planar thin film and

sectioning parallel to a pattern of relief structures supporting a thin film.1

The first process (Figure 2.2) produces nanowires.21 Stacking and sectioning

multiple films, separated by sacrificial layers (e.g., polymers or SiO2), which can be

etched in a later step, can produce closely spaced parallel nanowires.22 Topographic

patterning can introduce angles into these nanowires to form simple elements for

nanoelectronics, such as parallel electrodes.3 Patterning the thin film into stripes, and

sectioning perpendicular to the stripes, can produce nanowires with well-defined lengths

and spacings.23 The second process (Figure 2.3) produces 2D arrays of nanostructures.

The outlines of the molded features define the geometries of the structures produced by

sectioning.5

57

Figure 2.2. Summary of the process used for fabricating nanowires of loosely defined

length (> 100 µm) by sectioning thin films. (1) A piece of flat epoxy served as the

substrate for deposition of a metallic, polymeric, semiconducting, or oxide film. (2) A

rough cut provided a strip of this film supported by epoxy, which we embedded in

additional epoxy (3). (4) Ultrathin sectioning (nanoskiving) and removal of the epoxy

matrix formed nanowires in which each dimension was controlled by a different step of

the process.

58

Figure 2.3. Summary of the process used to generate two-dimensional arrays of metallic

crescent-shaped nanoparticles. (1) Soft lithographic molding afforded an array of epoxy

nanoposts. (2) Shadow evaporation, using a collimated source, deposited metal roughly

halfway around the circumferences of the nanoposts. (3) Additional epoxy embedded the

entire structure. (4) The ultramicrotome sectioned the embedded array into manipulable

slabs containing arrays of metallic crescents.

59

Ultramicrotomy. The microtome was invented in the 1770s to section specimens

of wood for microscopy.24 Microtomy, the practice of generating thin sections for

analysis, has since developed alongside microscopy.25 The invention of electron

microscopy (EM) in the 1950s motivated the development of the ultramicrotome, an

instrument that can section specimens into slabs < 100 nm in thickness under ambient

conditions.25 Ultramicrotomy is ubiquitous in histology and polymeric materials science.

In the science of hard materials, it is a complement to the more common techniques for

the preparation of samples of ion thinning and electropolishing.26 When equipped with a

diamond knife, the ultramicrotome can section even hard materials. Examples include:

industrial alumina;27 nanocrystalline pure elements, alloys, and ceramics;28 diamond

films;29 steel sheets;30 and stacked layers of semiconductors.31

Comparison of Nanoskiving with Materials Ultramicrotomy. Nanoskiving can

be treated as ultramicrotomy of materials having the goal of generating functional

nanostructures, rather than of preparing specimens for EM. There are at least four other

important differences. (1) In materials ultramicrotomy, the embedded specimen is usually

a bulk sample. In nanoskiving, the embedded specimen is usually a thin film. (2) In

materials ultramicrotomy, the requirement of transmission of electrons in EM imposes a

limit on the thickness of sections to < 100 nm. In nanoskiving, sections can be 30 nm –

10 µm thick, or greater. (3) In materials ultramicrotomy, small-area sections (< 200 µm in

width) are easier to obtain than large-area sections, because the force per unit length

exerted by the knife on the facet of a block is greater for small facets than for large facets,

and because it is easier to maneuver a small facet to an unworn region of the knife.32 In

nanoskiving, it is often desirable to fabricate arrays of nanostructures over as large an

60

area as possible (≥ 1 mm2). (4) In materials ultramicrotomy, extensively fractured

specimens of brittle materials or specimens damaged by artifacts of the sectioning

process can yield fragments that are large enough for analysis. In nanoskiving, any

fracturing can be catastrophic, and fracturing limits the materials that can be used. An

understanding of the properties of materials that make them amenable to nanoskiving

would enable it to be applied more broadly than it now is.

Mechanism of Sectioning. Sectioning involves a complicated interplay of

processes: compression of the sample; tension perpendicular to the plane of sectioning;

formation of new surfaces; bending, as the slab reorients from vertical to horizontal;

shearing stress; friction of the slabs on the knife; and heat.33 There are two possible

mechanisms for the sectioning process, i) true, or shear, sectioning, in which the edge of

the knife maintains intimate contact with both the facet of the block and the nascent slab

(Figure 2.4a), and ii) a mechanism of crack initiation and propagation, in which the knife

splits or cleaves the block (Figure 2.4b). There is evidence for both processes, and each

process can be present to different degrees in the same epoxy block, because each

material within the block (or different grains or phases within the same material) can

respond differently to the force of the knife (e.g., cleavage vs. shattering).

Processes resembling true sectioning (2.4a) appear to dominate for metals and

alloys.32 The shearing of the slab produces shear lamellae in the sections, perpendicular

to the direction of cutting. Microtomed sections of aluminum (thickness = 500 nm) and

micromachined chips of steel display this morphology, which is characteristic of unstable

plastic flow in metals and alloys.32 The appearance of scoring on both the top and bottom

surfaces of slabs of some specimens also suggests a mechanism like that of 2.4a.32

61

Figure 2.4. Schematic drawings of two mechanisms proposed for sectioning with an

ultramicrotome. (a) In “true” or “shear” sectioning, the edge of the knife maintains

contact with both the facet of the block and the underside of the epoxy slab. This

mechanism produces a region of intense shearing, which is responsible for shear lamellae

that are visible in micromachined chips of steel. This mechanism operates for metals and

soft materials. (b) The mechanism of crack initiation and propagation is active for brittle

materials, such as ceramics. The orange circle highlights the location within the

sectioning process where the two models are different.

62

Processes of crack initiation and propagation (2.4b) appear to operate for brittle

materials such as minerals and ceramics.26 The knife initiates a crack, which follows the

path of weakest molecular strength.34 Acetarin et al. argued in favor of some degree of

crack propagation for all specimens because of the observation of craze formation33

within slabs and also because it is impossible for a knife to be infinitely sharp—the radius

of curvature of diamond knives is typically 3 – 6 nm, or a few tens of atoms.35 The extent

to which crack propagation extends ahead of the knife edge depends on the amount of

energy the sample can absorb by plastic flow prior to fracture (as in a measurement of

hardness with a sharp indenter). Jésior postulated that the distance between the crack and

knife increases with increasing hardness of the specimen. He argued that the farther the

crack propagates ahead of the knife, the larger the radius over which the slab has to bend

to go from vertical to horizontal, and thus the smaller the compressive stress on the slab

while bending.34

The orientation of cleavage planes in crystalline samples also determines the

extent of fragmentation upon sectioning, as Antonovsky observed in samples of

alumina.27 True sectioning and crack propagation can apparently operate at the same time

on different grains within the same specimen. In samples of high-strength steel, Malis

observed regions with shear lamellae, which are consistent with true sectioning (2.4a), as

well as large defect-free regions, which is evidence of cleavage (2.4b).32

Ultramicrotomes, when equipped with sensors, have also been used as analytical tools to

measure fracture toughness and other properties of specimens.36

Mechanical Properties of Thin Films. The ability to obtain intact nanostructures

after sectioning depends on the mechanical properties of the film. These properties are

63

strongly influenced by the morphology of the film, which depends on the technique used

for deposition.37 In general, evaporated metallic films are polycrystalline and assume a

columnar grain structure whose columns are perpendicular to the substrate.37 Evaporation

of covalent solids, such as silicon and germanium, forms amorphous films.37Evaporated

films of oxides and nitrides can be depleted of oxygen and nitrogen. Compared to

evaporation, sputtering-coating generally affords the user more control over the

morphology of the thin film, and also preserves the stoichiometry.

Evaporated metallic films can be up to 100 times harder and stronger more than

their bulk, annealed counterparts.37 These polycrystalline films have two characteristics

that are responsible for their hardness and strength. (1) Work hardening: high densities of

dislocations (1010 – 1012 cm-2) correspond to those of extensively worked bulk metals. (2)

Grain-boundary strengthening: the sizes of grains in thin films (10 – 1000 nm) can be a

few orders of magnitude smaller than those of bulk samples. Work hardening and grain-

boundary strengthening combine to produce films in which the mobilities of dislocations

is impeded, plastic flow is restricted, and thus increased strength, hardness, and

brittleness.37

Scope. There is not an absolute criterion for the successful sectioning of materials

by ultramicrotomy. This paper should provide practical information for the selection of

materials and processes of deposition. It also describes the artifacts of the sectioning

process (and how to avoid them), and their influence on the nanostructures produced. For

reviews of the practical aspects of ultramicrotomy, see Goldstein and Newbury;38 for a

review of ultramicrotomy applied to the science of inorganic materials, see Malis and

Steele;32 for a survey of embedding resins, see Acetarin et al.;33 for a perspective on the

64

role of ultramicrotomy for biological applications, see Villiger;39 and for a history of

ultramicrotomy, see Pease and Porter.25

Experimental Design

Selection of Thin-Film Materials. We sectioned films of several different

materials, including metals, semiconductors, metal oxides, conjugated polymers

(semiconducting and conducting), and films of semiconductor nanocrystals. The

materials chosen occupy a range of different mechanical properties, and most have not

been used with nanoskiving before.

Selection of Processes of Deposition. For metals, our primary method of

deposition was electron-beam (e-beam) evaporation.40 We used sputter-deposition for

films of platinum and ceramic materials, to try to control the amount of fragmentation in

the nanostructures produced. We synthesized microplates (d ~ 10 – 70 µm) of single-

crystalline gold,41 which we sectioned into nanowires. Electroless deposition provided

amorphous films of nickel on epoxy substrates. We deposited ceramic films by

evaporation, sputter deposition, or plasma-enhanced chemical-vapor deposition

(PECVD). We deposited films of conjugated polymers and semiconductor nanocrystals

by spin-coating and drop-casting. We deposited polypyrrole electrochemically.

Fabrication of Nanowires by Sectioning Thin Films. A stringent test that

enabled us to determine the applicability of materials for nanoskiving was to form long

nanowires (>100 µm) by sectioning thin films, and to determine the extent of

fragmentation by scanning electron microscopy (SEM). We measured electrical

conductivity to verify the continuity of some of the nanowires. We used the process

65

summarized in Figure 2.2 for all of these experiments. We sectioned the thin films with a

direction of cutting perpendicular to the edge of the thin film. In this orientation, the

action of the knife compressed the slab along the short axis of the nanowire. We also

determined the effect of compression on the frequency of defects when the direction of

cutting was parallel to the edge of embedded film for some metals.

Fabrication of 2D Arrays of Crescents and Rings by Sectioning Arrays of

Nanoposts. We fabricated arrays of simple semicircular and circular structures using a

process summarized in Figure 2.3, which began by forming an array of epoxy nanoposts

by soft lithography.42 Line-of-sight deposition of metal on the sidewalls of these posts,

followed by embedding, and sectioning with the ultramicrotome, produced metallic

crescent-shaped nanostructures. Conformal coating by a sputtering process, followed by

embedding and sectioning, produced arrays of rings. Examination of the structures

produced by these processes provided i) the yield of unbroken structures and ii) the rate

of compression of the axis of the array parallel to the direction of cutting.

Selection of Embedding Resin. An embedding resin should have i) a relatively

high value of elastic modulus (~3 GPa, a block that is too elastic deflects from the knife,

rather than cleave), and ii) a high yield stress after which the material undergoes plastic

deformation (~70 MPa, otherwise the slab will deform upon sectioning).33 Epoxy resins

such as Epon, Araldite 502, and Epo-Fix (which is used for all experiments in this paper,

unless otherwise noted) possess most of the properties required for good sectioning. Epo-

Fix (Electron Microscopy Sciences) is a typical two-part epoxy containing polymers of

bisphenol-A diglycidyl ether and triethylene tetraamine. It has excellent adhesion to most

materials tested, and can be cured at room temperature (the best results were achieved for

66

60° C for 2 h) from a prepolymer with relatively low viscosity (5 × 10-2 Pa·s), which

facilitates impregnation of porous samples. Epo-Fix is relatively hard (75 Shore D)43 and

stiff (flexural modulus = 2.4 GPa). Epo-Fix exhibits significant compression along the

direction of cutting (15 – 20% for 100 nm sections). This rate of compression ensured

that the thin films would have to be sufficiently soft and compliant to accommodate the

strain due to the compression of the embedding resin, and thus imposed a strict criterion

for the success or failure of a given thin film.

Ultramicrotome and Knife. We used a Leica Ultracut UCT equipped with a 35°

diamond knife for all applications (Diatome Ultra 35°). We and others found less

compression and delamination of films from the resin with 35° knives than with 45°

knives.26

Results and Discussion

Fabrication of Nanowires by Sectioning Thin Metallic Films. We deposited,

embedded, and sectioned a series of ten different metallic thin films by e-beam

evaporation: aluminum, titanium, nickel, copper, palladium, silver, platinum, gold, lead,

and bismuth. We deposited three additional films of these materials using different

techniques—a polycrystalline film of platinum by sputter deposition, single-crystalline

microplates of gold by growth from solution,41 and an amorphous film of nickel by

electroless deposition. We deposited most films with a thickness of 50 – 100 nm, and

sectioned them with a programmed thickness of 80 nm at a velocity of 1 mm/s, with a

direction of cutting perpendicular to the plane of the film. The substrate for deposition of

films of silver, gold, palladium, and platinum was the polished surface of a silicon wafer.

67

After deposition, we transferred the film to epoxy by puddle-casting an epoxy

prepolymer, curing it, and peeling off the solid epoxy, along with the metallic film

(“template stripping”).44,45 This method produced very flat films (rms roughness < 1 nm)

on the surface templated by the silicon wafer.45 We deposited all other materials directly

on the flat side (rms roughness = 0.5 nm) of a cured epoxy substrate, prepared by puddle-

casting the prepolymer against the polished side of a silicon wafer.4 Figure 2.5 shows

representative spans of several of the nanowires produced by sectioning these films.

Aluminum (2.5a), nickel (2.5c), copper (2.5d), palladium (2.5e), silver (2.5f), gold (2.5i),

lead (2.5k), and bismuth (2.5l) formed nanowires with long unbroken spans, as

determined by SEM. Most nanowires were 300 µm – 1 mm long. We defined this length

loosely by cutting with a razor blade (see Figure 2.2, step 2).

Electrical Conductivity of Long Gold Nanowires. In order to verify that

inspection by SEM was an accurate method of determining physical continuity, we

measured the electrical conductivity of the longest nanowires produced: a gold nanowire

with dimensions of l = 1.8 mm, w = 80 nm, h = 120 nm. Figure 2.6 shows an SEM image

and the electrical characteristics of a representative span of the nanowire. Based on these

dimensions and the current at ±10 mV, we calculated a value of conductivity of 3.0 × 105

Ω-1 cm-1 (the literature value for bulk gold is 4.5 × 105 Ω-1 cm-1). This experiment

demonstrates that the process of sectioning introduces relatively few defects that are

electronically significant into the nanowires.

Correlation of Bulk Properties of Materials with Fragmentation of Thin

Films. Sectioning films of titanium (2.5b) and platinum (2.5g) produced fragmented

68

Figure 2.5. Scanning electron microscope (SEM) images of metallic nanowires formed

by sectioning thin metallic films deposited by electron-beam evaporation (all except h

and j), sputter deposition (h), and solution-phase chemical growth (j). Metals are

arranged by atomic mass. Soft and compliant materials—aluminum, copper, palladium,

silver, gold, lead, and bismuth—formed long, intact nanowires over distances > 100 µm.

Hard and rigid materials—titanium, nickel, and platinum—tended to fracture.

69

Figure 2.6. Scanning electron micrograph (top) of a long gold nanowire that was

electrically continuous over a span of 1.8 mm. The plot (bottom) shows current density v.

voltage of the nanowire across the span of 1.8 mm.

70

nanowires. We verified, by SEM, that the films were intact prior to embedding and

sectioning. In general, metals that formed intact nanowires upon sectioning were soft,

compliant, and low-melting, based on values of bulk properties in the literature. Those

that fragmented were hard, rigid, and high-melting. Metals as soft or softer than

palladium—that is, metals with bulk values of hardness of the less than approximately

500 MPa (Vickers hardness number)—formed intact nanowires. The next hardest metal

tested, platinum, broke into fragments with average lengths of 10 µm (2.5g). Evaporation

of nickel produced an exceptionally resilient film by evaporation, which did not

fragment. Titanium, the hardest material tested, fragmented extensively (2.5b).

Effect of Direction and Speed of Cutting on Fragmentation. The orientation of

an embedded thin film with respect to the direction of cutting had a profound effect on

the frequency of defects and the morphology of the nanostructures that were formed. For

example, sectioning with a direction of cutting perpendicular to the edge of an evaporated

film of nickel formed unfragmented nanowires (2.5c), while sectioning parallel to the

edge of the same film produced fragmented nanowires whose segments had lengths of 10

– 20 µm. We attribute the higher frequency of defects in nanowires oriented parallel to

the direction of cutting to compressive stress along the longitudinal axis of the nanowires.

The breaks in the nanowires do not appear at regular intervals. This observation suggests

that randomly located defects and thin areas (which arise from uneven chemical or

physical vapor deposition) influence the sites of fracture. Evaporated films of gold,

palladium, nickel, and platinum—which displayed no or little fragmentation using a

perpendicular direction of cutting—yielded fragments with lengths of approximately 100

µm, 10 µm, 10 µm, and 1 µm, when sectioned parallel to the edge of the film.

71

In addition to increasing the rate of defects in the nanowires, a direction of cutting

parallel to the edge of the film also imparted a roughened morphology to the nanowires.

Figures 2.7a and 2.7b show the two relative orientations used in this paper between the

direction of cutting and the edge of the embedded film: perpendicular (2.7a) and parallel

(2.7b). Figures 2.7c and 2.7b show two palladium nanowires obtained from the same

embedded film, but sectioned from orthogonal directions. The insets are close-up images

that show the smooth microstructure of the film cut perpendicular to its edge (7c), and the

rough microstructure of the nanowire cut parallel to its edge (7d). The rough

microstructure in 2.7d resembles the shear lamellae (parallel to the short axis of the

nanowire) observed in microtomed foils of bulk metals.32

Jésior published a series of papers on compression in ultramicrotomy, and how to

avoid it.34,46 He observed that the rate of compression of latex spheres was independent of

the compression of the epoxy matrix. The soft latex spheres formed ellipses upon

sectioning, and delaminated from the edges of the epoxy matrix.46 In our system,

however, delamination of metallic films from the epoxy matrix was rare. Adhesion of

metallic films to a compressible embedding resin (17% for 100-nm slabs of Epo-Fix)

forces the nanowire to compress at the same rate as the matrix, and could be a factor

responsible for the high rate of defects observed in hard and stiff films whose edges were

oriented parallel to the direction of cutting.

Jésior also observed that compression is independent of the speed of cutting.34

Figure 2.7e shows a platinum nanowire oriented parallel to the direction of cutting using

six speeds from 0.1 – 10 mm/s. We found that the frequency of defects was independent

of the speed of cutting, from 0.1 – 8 mm/s. (The highest speed tested, 10 mm/s, actually

72

decreased the frequency of defects. At this speed, the nanowire, and the surrounding

epoxy, accommodated the compressive strain by buckling into long-range, wave-like

distortions. See 2.7e, “10 mm/s”.)

Fabrication of Platinum Nanowires from Sputter-Deposited Films. Sputter-

deposition can yield films with different morphologies: from polycrystalline metallic

films with a range of mean grain size, to metallic glasses, and oxide films with

stoichiometry that matches that of the source material.37 We found that a sputter-coated

film of platinum was much more resistant to fracture than an evaporated film of the same

thickness (50 nm, compare Figure 2.5g and 2.5h). While the evaporated film of platinum

formed fragmented nanowires with 1-µm segments using a direction of cutting parallel to

the edge of the film, the sputter-deposited film fragmented into 10-µm segments. We

attribute the greater resilience of the sputtered film to morphological characteristics

(density of dislocations and the sizes of grains). We expect the parameters of deposition

can be tuned to reduce the rate of fragmentation upon sectioning hard materials.

Fabrication of Gold Nanowires from Single-Crystalline Microplates.

Polycrystalline films, deposited by physical vapor deposition, become polycrystalline

nanostructures upon sectioning. Many applications however, require structures with

smooth surfaces (e.g., metallic nanowires for plasmonic waveguiding).47,48 There is a

family of procedures for the synthesis of single-crystalline metallic micro- and

nanoparticles that provide control over the shapes and the compositions of these

particles.49 Our laboratory previously reported the plasmonic properties of single-

crystalline gold nanowires formed by nanoskiving single-crystalline microplates grown

by solution-phase synthesis.4,41 Observation of lattice fringes by TEM indicated that the

73

Figure 2.7. The effects of the direction of cutting on the morphology of nanostructures

and the extent of fragmentation. (a) A palladium nanowire (NW) sectioned with the

direction of cutting perpendicular to the edge of the film. (b) A nanowire sectioned with

the direction of cutting parallel to the edge of the film. Sectioning parallel to the edge

produces two effects: a rougher morphology (insets of a and b) and a higher frequency of

fragmentation. (c) The effect of speed of sectioning on the frequency of fragmentation in

platinum nanowires sectioned parallel to the plane of the film. There was no effect from

0.1 mm/s to 8 mm/s. With a speed of 10 mm/s, buckling accommodated some of the

compressive strain, and had the effect of reducing the frequency of defects.

74

crystallinity of the microplates was intact after sectioning. We have never observed

fragmentation of single-crystalline gold nanowires, even those with lengths up to 50 µm

(the longest we have produced). Comparing Figures 2.5i and 2.5j illustrates the difference

between polycrystalline and single-crystalline nanowires.

Fabrication of Nickel Nanowires from Amorphous Films. We also deposited

and sectioned an amorphous nickel film by electroless deposition. While this film formed

an unfragmented nanowire with a direction of cutting perpendicular to the embedded

film, it broke into fragments 500 nm – 1 µm long when cut with a parallel direction.

These fragments were an order of magnitude smaller than those derived from the

evaporated film of nickel. Correlation of the microstructures of the thin films, as

determined by TEM, will be required to understand the extent to which morphology

influences the extent of fragmentation.

Fabrication of Semiconductor Nanowires from Evaporated Films. Covalent

solids, such as silicon and germanium, form amorphous films when evaporated. The most

important result from the survey of polycrystalline metallic films was that soft materials

form relatively soft thin films, which tend to form intact nanostructures upon sectioning.

Initial results suggest that this generalization holds true for amorphous semiconductors,

as well as for metals. We sectioned films of germanium and silicon using a direction of

cutting perpendicular to the edge of the embedded thin film. The germanium film formed

intact nanowires over lengths of > 30 µm (Figure 2.8a), while the silicon film fractured

extensively (2.8b).

Fabrication of Nanowires from Films of Ceramic Films. Films of oxides,

aluminum oxide (Al2O3), silicon dioxide (SiO2), and tin-doped indium oxide (ITO),

75

prepared by evaporation, sputter deposition, and PECVD, were the hardest materials we

tested (Figures 2.8c – 2.8g). The only film that formed intact nanowires was a sputter-

coated film of SiO2. Applications that require long, intact spans of ceramic materials

should be prepared using a softer material that can be converted to the film in its final

form. For example, thermal oxidation and calcination can, in principle, convert soft

metallic films or sol-gel precursors—which section easily—into the desired ceramic

materials.50

Fabrication of Nanostructures of Solution-Processed Materials:

Semiconductor Nanocrystals and Conjugated Polymers. Organic polymers and

polymer-like materials are among the easiest materials to section with the

ultramicrotome.22 Figure 2.8h is a nanowire derived from sectioning a spin-coated film of

oleylamine-capped lead sulfide (PbS) nanocrystals,51 which formed long, intact segments.

Figures 2.8i – 2.8l are examples of conjugated (semiconducting and conducting)

polymers: poly(3-hexylthiophene) (P3HT, 2.8i), poly(3,4-ethylenedioxythiophene)

poly(styrenesulfonate) (PEDOT:PSS, 2.8j), poly(benzimidazobenzophenantrholine

ladder) (BBL, 2.8k), and poly(2-methoxy-5-(2'-ethyl-hexyloxy)-1,4-phenylene vinylene)

(MEH-PPV, 2.8l). All of these materials formed intact nanowires regardless of the

orientation of cutting.

Table 2.1 summarizes the materials, methods of deposition, and average intact

spans of each nanowire produced by nanoskiving. Entries labeled with an asterisk (*)

indicate that the nanowire did not fragment when sectioned with a direction of cutting

normal to the plane of the film, as determined by inspection of the entire length of the

76

Figure 2.8. Nanowires of semiconductors, ceramics, and conjugated (semiconducting

and conducting) polymers. All materials were deposited by e-beam evaporation unless

otherwise noted. Among amorphous, evaporated elemental semiconductors, germanium

remained intact (a), while silicon fragmented (b). All films of alumina and ITO deposited

by evaporation (c and g) and sputter deposition (not shown) were fragmented. The ability

to form intact nanowires of SiO2 depended strongly on the method used for deposition:

sputter-deposited films remained intact, while evaporated films, and those deposited by

plasma-enhanced chemical-vapor deposition (PECVD) fragmented. Spin-coated films of

lead sulfide nanocrystals (h) formed intact nanowires, as did all conjugated polymer films

tested. The film of PEDOT:PSS adhered poorly to the epoxy matrix, but it did not

fragment.

77

Table 2.1 Summary of the materials, methods of deposition, and average intact spans of

the nanowires formed by nanoskiving thin films.

*Indicates unbroken nanowires over the range examined.

Category Material Method of Deposition Avg. Span (µm)

Metals Al* evaporation > 40

Ti evaporation 1

Ni* evaporation > 40

Ni* electroless deposition > 30

Cu* evaporation > 30

Pd* evaporation > 100

Ag* evaporation > 50

Pt evaporation 10

Pt* sputter-deposition > 120

Au* evaporation > 80

Au* chemical growth > 50

Pb* evaporation > 30

Bi* evaporation > 40

Semiconductors Si evaporation 5

Ge* evaporation > 30

Oxides SiO2 evaporation 10

SiO2* sputter-deposition > 20

SiO2 PECVD 5

Al2O3 evaporation 1

Al2O3 sputter-deposition 4

ITO evaporation 10

ITO sputter-deposition 2

Polymers P3HT* spin-coating > 50

MEH-PPV* spin-coating > 100

BBL* spin-coating > 50

PEDOT:PSS* spin-coating > 40

Nanocrystals PbS* spin-coating > 60

78

nanowire by SEM. For these unbroken nanowires, the minimum span reported in the

table corresponds to the longest span (≥ 30 µm in the center of the nanowire) that could

be verified as unbroken in a single, high-resolution SEM image. For all other entries, we

calculated the average span by dividing the length of a representative span of the

nanowire of at least 30 µm by the number of fragments.

Fabrication of 2D Arrays of Rings and Crescents. One of the most promising

uses of nanoskiving is the fabrication and replication of 2D arrays of nanostructures

embedded in thin slabs, which can be transferred to another substrate,52 for optics or other

applications.5 Figure 2.3 summarizes the procedure used, which we reported previously.53

Figure 2.9a shows an array of epoxy nanoposts, in which the sidewalls are partially

coated with gold. Embedding these metalized posts in epoxy and sectioning parallel to

the plane of the array produced 2D arrays of metallic nanostructures. The two most

deleterious artifacts of mechanical sectioning that manifest themselves in 2D arrays

formed by nanoskiving are i) fracture of individual structures and ii) compression of

square arrays into rectangular arrays.

Yield of Intact Nanostructures in 2D Arrays. Figure 2.9b shows an array of

platinum crescents. The yield of intact nanostructures was 93%. Of 320 crescents in the

array shown, nine were broken in the center (see lower inset), and thirteen were broken at

the tips. Figure 2.9c shows counterfacing silver and silicon crescents patterned in the

same plane. We did not find any broken structures in the array. The high yield (> 99%) of

intact silicon crescents is unexpected, because the planar film of silicon, deposited using

the same conditions of evaporation, was extensively fragmented when cut into a

79

Figure 2.9. (a) An array of gold-coated epoxy nanoposts (the product of step 2, Figure

2). The posts are 8 µm tall, and 250 nm in diameter, at the top. (b) An array of platinum

crescent-shaped nanoparticles. (c) An array of counterfacing crescents of silver and

silicon. (d) A square array of nanostructures was compressed during the cutting

operation.

80

nanowire (see Figure 2.8b). A possible explanation is that small particles are not subject

to long-range tensile stress that could contribute to fragmentation of nanowires. Table 2.2

summarizes the results of 2D arrays we have fabricated using previously published

procedures.53

Compression of 2D Arrays. The second effect of the sectioning process that is

relevant to some applications is the compression of the array along the direction of

cutting. We formed an array of gold crescents and measured the dimensions of the array

and the angle between the axes (Figure 2.9d). The original array of posts, from which we

cut this array of crescents, was square with a pitch of 2 µm between features. After

sectioning (thickness = 100 nm) using a randomly selected direction of cutting, we

observed a shortening of both the vertical axis to 1.65 µm and the horizontal axis to 1.9

µm. We measured a direction of cutting of 73° from the horizontal row of crescents, as

determined by a score, which was created by a defect in the knife and parallel to the

direction of cutting. Because the direction of cutting was not parallel to either axis of the

array, the compression skewed the array such that the unit cells deformed from square to

diamond-shaped, with angles of 88.7° and 91.3°. The total compression in the direction

of cutting was 17%. A survey of embedding resins led us to perform the same experiment

with another epoxy, UVO-114 (a UV-curable resin obtained from Epoxy Technologies),

and we measured 8% compression. (Through the course of the experiments described in

this paper, and others, UVO-114 became our preferred resin for nanoskiving.) In general,

the rate of compression is decreases with thickness of the section and the hardness of the

resin.34

81

Table 2.2 Table summarizing the materials, methods of deposition, geometries, and

yields of two-dimensional arrays of nanostructures.

*Indicates unfragmented nanostructures.

Material Method of Deposition Geometry Yield (%)

Si* evaporation crescents > 99.7

Pd* evaporation crescents > 99.7

Ag* evaporation crescents > 99.7

Pt evaporation crescents 93.1

Au evaporation crescents 99.9

Au* sputter-deposition rings > 99.7

Polypyrrole/Au* electrodeposition rings > 99.8

PbS nanocrystals drop-casting rings 97.5

82

We summarized our findings and made predictions regarding the applicability of

materials we have not yet tested in Figure 2.10. We assumed all metallic films were

polycrystalline and all covalent solids were amorphous. Materials labeled “intact” are

those that displayed no visible fragmentation, or large unbroken spans when cutting

nanowires perpendicular to the plane of the embedded thin film. Films labeled

“fragmented” fractured extensively into segments < 10 µm. Those labeled “borderline”

are films whose rate of fragmentation depended strongly on the method of deposition, the

size and geometry of the structure (e.g., nanowires or crescents), and the orientation of

the thin film with respect to the direction of cutting. We extrapolated our observations to

make predictions for materials that we did not test, or for which we did not have enough

data. In the case of relatively hard, d-block metals, our predictions are conservative;

while we predict most would fragment, it might be possible to optimize the recipe for

deposition to form films soft enough to promote a high yield of unfragmented

nanostructures. In the case of alkali and alkaline earth metals, we based our predictions

on mechanical properties alone, and assumed that these films could be deposited and

sectioned in an inert, dry environment. In addition to the materials in Tables 2.1 and 2.2,

we sectioned chips (~ 10 µm thick) of highly oriented pyrolytic graphite. This material

produced large intact spans of parallel plates when sectioned with a direction of cutting

perpendicular to the edge of the chip; when sectioned parallel to the edge of the chip, the

sample fractured extensively, into ~ 1 µm fragments. We thus labeled graphite as

“borderline” in Figure 2.10.

Scoring. Chips in the edge of the diamond knife score the epoxy slabs and

damage embedded structures in the paths of the scoring. A freshly sharpened knife

83

Figure 10. Summary of findings and predictions regarding the abilities of elements,

oxides, polymers, and nanocrystals to form nanowires by sectioning thin films.

84

contains no chips. As the knife is used, contact with hard material (commonly small

inorganic dust particles from the laboratory, steel from the razor blade during trimming of

the block face, or particularly hard samples) breaks microscopic pieces from knife.32

These chips are inevitable, and knives must be re-sharpened after 6 – 12 months of

normal use, at about half the cost of a new knife. Scoring was visible under dark-field

optical microscopy, or by SEM (Figure 2.11a). Figure 2.11b shows a fracture caused by

scoring in an aluminum nanowire. A region of delamination, where the silicon substrate

is visible, accompanied the fracture. Scoring provides an exact marker of the direction of

cutting on the surface of the epoxy slab, and can be useful in distinguishing fractures due

to a damaged knife from those due to the brittleness of the thin films.

Delamination. Adhesion of an embedded film to the epoxy matrix provides

stability to the nanostructures during the process of sectioning. Distortion caused by

delamination of the films from the epoxy matrix can impose tensile stress on

nanostructures. The structures have higher tendencies to break in regions of delamination

than in regions in which the matrix supports the embedded film on all sides. We observed

the least amount of delamination by embedding freshly deposited films. Films that were

exposed to the ambient atmosphere for several days tended to delaminate upon sectioning

because of physisorption of adventitious organic material from the ambient air. A brief (~

5 s) exposure to an air plasma removed adventitious organic material from metals such as

gold, which would not be oxidized by plasma. Figure 2.11c shows a delaminated region

in an epoxy slab containing a nickel nanowire. The delaminated regions extend for

several microns, and contained the only fracturing we observed.

85

Figure 11. Examples of artifacts of the process of sectioning that are deleterious to the

structures produced. (a) A divot in the edge of the knife produced the score on the surface

of the epoxy matrix. The score damaged the gold nanowire at the point of intersection.

(b) A defect in an aluminum nanowire and concomitant delamination of the nanowire

from the epoxy matrix. The silicon support used for imaging by SEM is visible in the

gap. (c) Poor adhesion of a nickel film to the epoxy matrix caused delamination of the

nanowire from the slab. Delamination caused both sides of the nanowire to be supported

unevenly during cutting, and the nanowire broke. (d) Image of a gold nanowire sectioned

with a knife, without supersonic oscillation. (e) The same gold nanowire cut with

oscillation. The morphology and the roughness of the nanowire cut with oscillation are

smoother than the nanowire cut using a stationary knife.

86

Chatter. Chatter is another artifact of ultramicrotomy, in which vibrations in the

room or generated by friction between the knife and the sample produce parallel lines in

the slab parallel to the edge of the knife. The effect is exacerbated when the knife is not

securely fixed in the chuck. We rarely observed instances of chatter in these experiments,

and direct the interested reader to the review by Malis and Steele for a discussion of this

effect.32

Oscillating Knives. We expected that ultrasonic, oscillating knives would

produce even better results: Studer and Gnaegi have demonstrated that polystyrene

blocks, sectioned with a 45° knife oscillating at 2 kHz with 400 nm amplitude, produced

sections compressed by only 3.5%.54 We found that, in addition to reducing compression,

oscillating also produced nanostructures of higher quality than do stationary knives.

Figures 2.11d and 2.11e show gold nanowires cut with a knife without and with

oscillation, using the frequency (2 kHz) and amplitude (400 nm) used by Studer and

Gnaegi.54 The morphology and the edges of the nanowire are smoother if the knife is

oscillated.

Conclusions

This paper surveyed several materials and processes of thin-film deposition, many

of which had not been used with nanoskiving before. From this survey of materials and

methods, we found a simple correlation between the propensities of thin evaporated

metallic films to fragment, and their bulk values of hardness, tensile strength, elastic

modulus, and melting point. The materials that yielded intact nanostructures upon

sectioning tended to be soft, compliant, and low-melting. Understanding the exact

87

relationship between the morphology of thin films, their mechanical properties, and their

performance in nanoskiving will require further experimentation (e.g., a combination of

analysis using TEM, and measurements of the indentation hardness or tensile strength of

the thin films). We predict, however, that the rate of defects observed by microtoming

thin films would agree qualitatively with measurements of mechanical properties. We

also determined the effects of other parameters, such as the orientation of cutting with

respect to the embedded thin films, the condition of the diamond knife, and the adhesion

of the thin film to the matrix. These experiments provide a practical model for selecting

materials, methods of thin-film deposition, and parameters of sectioning, for nanoskiving.

This paper also described factors that could limit the extent to which nanoskiving

can be integrated with other techniques. Compression, for example, limits precise

registration of arrays with structures produced by conventional lithography and with

other arrays produced by nanoskiving. The effect of scoring is more pernicious,

particularly if nanoskiving is to be used for production. The effect of accumulated

damage to the edge of the knife if used continuously is unknown, but it would require re-

sharpening of knives more frequently than is required for sectioning routine biological or

inorganic materials. Complete avoidance of hard, inorganic particles requires samples to

be molded, coated, and sectioned with assiduous attention to cleanliness (or,

alternatively, require these steps to be performed in a cleanroom). Despite these

shortcomings, nanoskiving is still among the most general methods of fabricating

nanostructures with respect to different classes of materials, and is possibly the only

method that can pattern multiple materials in the same plane in a single step (as in Figure

2.9b).

88

There could be opportunities to exploit what we called “artifacts” in this paper to

fabricate structures that cannot be made easily with other methods. For example,

controlled fragmentation of nanowires produces nanogaps, which could be used to

concentrate electric fields;55 deliberate scoring using knives with engineered defects

could be used for perforation of nanostructures for the same purpose; and compression

and skewing of 2D arrays of nanostructures could be used to generate anisotropic arrays

using an isotropic master.42

Nanoskiving is the first technique to use mechanical sectioning as the key step in

the fabrication and replication of nanostructures singly or in arrays. This report should be

useful as nanoskiving develops and for designing other processes of nanomachining in

the future.

Methods

Materials. We obtained Epo-Fix embedding resin from Electron Microscopy

Sciences, and UVO-114 from Epoxy Technology, Inc. We obtained metals for

evaporation from the Kurt J. Lesker Company. All other chemicals were purchased from

Sigma Aldrich. We deposited almost all metals directly onto planar epoxy substrates,

which we prepared by mixing Epo-Fix in a ratio (v:v) of base to hardener of 7.5:1 in a

sealed centrifuge tube by shaking. We degassed the prepolymer in a vacuum desiccator.

We then poured the prepolymer over the polished side of a silicon wafer (we contained

the liquid prepolymer using a ring of PDMS) and cured the epoxy for 60 °C for 2 h. After

cooling to room temperature, we peeled the epoxy off the silicon using a razor blade. The

surface of the epoxy had a value of roughness (rms) of 0.5 nm (as determined by AFM).

89

To promote adhesion of the thin films to the epoxy substrate, we treated the surface of the

epoxy with an air plasma (100 W, 1 torr, 30 s). The highest-quality films of gold, silver,

palladium, and platinum were prepared by template stripping.45 In this procedure, the

film was evaporated on a silicon wafer and peeled off by curing epoxy against the

evaporated film.

Electron-Beam Evaporation. We evaporated films with in a chamber with a

base pressure of 1 × 10-6 torr, an accelerating voltage of 10 kV, a filament current of 0.6 –

1 A, and rates of deposition of ~ 1 Å/s for all materials. The substrate was placed 40 cm

above the source.

Sputter Deposition. We sputter-deposited films of platinum, aluminum oxide,

ITO, and silicon dioxide using an AJA model ATC sputtering system, which operated at

a base pressure of 8 × 10-7 torr. We introduced argon into the chamber at a rate of 40

sccm/s. The pressure during deposition was 4 mtorr. The platinum film was deposited at

450 V (DC) at 50% power, with a rate of deposition of 3.6 Å/s. The aluminum oxide film

was deposited at 198 V (RF) at 50% power, with a rate of deposition of 0.1 Å/s. The ITO

film was deposited at 198 V (RF) at 50% power, with a rate of deposition of 0.6 Å/s. The

silicon dioxide film was deposited at 450 V (DC) at 50% power, with a rate of 0.12 Å/s.

PECVD. We deposited a film of silicon dioxide by PECVD using a Surface

Technology Systems (STS) system operating at a base pressure of 1.2 × 10-5 torr and a

pressure of 4 mtorr during deposition. DC power (475 V) and RF power (150 V) were

running simultaneously during the deposition.

Spin-Coating of Conjugated Polymer Films. We deposited films of P3HT, and

BBL and MEH-PPV, were deposited by spin-coating using previously described

90

conditions.22,52 We deposited films of PEDOT:PSS from a 0.65% dispersion in water (by

diluting a 1.3% dispersion as obtained from Aldrich) by spin-coating at a rate of 1 krpm

and annealing in a vacuum oven at 125 °C for 2 h. We prepared a solution of lead sulfide

nanocrystals in hexanes by a reported procedure at a concentration of approximately 1016

nanocrystals/L.56 Spin-coating this solution onto an epoxy substrate (1 krpm), followed

by a 1 s exposure to air plasma (100 W, 500 mtorr), provided a mechanically resilient

100-nm-thick film of nanocrystals.

Embedding. We cut strips from the metallic, semiconducting, ceramic, and

polymeric films, supported by their epoxy substrates, into strips, ~ 5 mm long, and 0.3 –

1 mm long. We placed these strips in polyethylene embedding molds (Electron

Microscopy Sciences), filled the molds with mixed and degassed Epo-Fix prepolymer,

and cured the blocks at 60 °C for 2 h.

Ultramicrotomy. We sectioned all films with a Leica Ultracut UCT

ultramicrotome equipped with a Diatome Ultra 35° diamond knife with a 6° built-in

clearance angle. An additional, external 6° clearance angle produced a total cutting angle

of 47°. We cut all slabs using a cutting speed of 1 mm/s and a programmed thickness of

80 nm or 100 nm. The Supporting Information of the paper by Xu et al.5 contains a

description of the cutting process.

Acknowledgements

This research was supported by the National Science Foundation under award

PHY-0646094. The authors used the shared facilities supported by the NSF under

MRSEC (DMR-0213805 and DMR-0820484). This work was performed in part using the

91

facilities of the Center for Nanoscale Systems (CNS), a member of the National

Nanotechnology Infrastructure Network (NNIN), which is supported by the National

Science Foundation under NSF award no. ECS-0335765. CNS is part of the Faculty of

Arts and Sciences at Harvard University. D.J.L. acknowledges a Graduate Fellowship

from the American Chemical Society, Division of Organic Chemistry, sponsored by

Novartis. The authors thank Professors John Hutchinson and Frans Spaepen for helpful

discussions, and Qiaobing Xu and Richard Schalek for obtaining sections using the

oscillating knife.

References

(1) Xu, Q. B.; Rioux, R. M.; Dickey, M. D.; Whitesides, G. M. Acc. Chem. Res. 2008, 41, 1566-1577.

(2) Xu, Q.; Rioux, R. M.; Whitesides, G. M. ACS Nano 2007, 1, 215-227.

(3) Dickey, M. D.; Lipomi, D. J.; Bracher, P. J.; Whitesides, G. M. Nano Lett. 2008, 8, 4568-4573.

(4) Wiley, B. J.; Lipomi, D. I.; Bao, J. M.; Capasso, F.; Whitesides, G. M.

Nano Lett. 2008, 8, 3023-3028. (5) Xu, Q. B.; Bao, J. M.; Rioux, R. M.; Perez-Castillejos, R.; Capasso, F.;

Whitesides, G. M. Nano Lett. 2007, 7, 2800-2805. (6) Gates, B. D.; Xu, Q. B.; Stewart, M.; Ryan, D.; Willson, C. G.;

Whitesides, G. M. Chem. Rev. 2005, 105, 1171-1196. (7) Stewart, M. E. M., M. J.; Yao, J.; Thompson, L. B.; Nuzzo, R. B. Proc.

IMechE Part N: J. Nanoengineering and Nanosystems 2007, 220, 81-138.

(8) Willson, C. G.; Roman, B. J. ACS Nano 2008, 2, 1323-1328. (9) Pease, R. F.; Chou, S. Y. Proc. IEEE 2008, 96, 248-270.

(10) Stewart, M. E.; Anderton, C. R.; Thompson, L. B.; Maria, J.; Gray, S. K.; Rogers, J. A.; Nuzzo, R. G. Chem. Rev. 2008, 108, 494-521.

92

(11) Yun, M. H.; Myung, N. V.; Vasquez, R. P.; Lee, C. S.; Menke, E.; Penner,

R. M. Nano Lett. 2004, 4, 419-422. (12) Cui, Y.; Wei, Q. Q.; Park, H. K.; Lieber, C. M. Science 2001, 293, 1289-

1292. (13) Ramanathan, K.; Bangar, M. A.; Yun, M.; Chen, W.; Myung, N. V.;

Mulchandani, A. J. Am. Chem. Soc. 2005, 127, 496-497. (14) Boettcher, S. W.; Spurgeon, J. M.; Putnam, M. C.; Warren, E. L.; Turner-

Evans, D. B.; Kelzenberg, M. D.; Maiolo, J. R.; Atwater, H. A.; Lewis, N. S. Science 2010, 327, 185-187.

(15) Williams, S. S.; Hampton, M. J.; Gowrishankar, V.; Ding, I. K.;

Templeton, J. L.; Samulski, E. T.; DeSimone, J. M.; McGehee, M. D. Chem. Mater. 2008, 20, 5229-5234.

(16) Menard, E.; Meitl, M. A.; Sun, Y. G.; Park, J. U.; Shir, D. J. L.; Nam, Y.

S.; Jeon, S.; Rogers, J. A. Chem. Rev. 2007, 107, 1117-1160. (17) Xia, Y. N.; McClelland, J. J.; Gupta, R.; Qin, D.; Zhao, X. M.; Sohn, L.

L.; Celotta, R. J.; Whitesides, G. M. Adv. Mater. 1997, 9, 147-149. (18) Xia, Y. N.; Kim, E.; Zhao, X. M.; Rogers, J. A.; Prentiss, M.; Whitesides,

G. M. Science 1996, 273, 347-349. (19) Chou, S. Y.; Krauss, P. R.; Renstrom, P. J. J. Vac. Sci. Technol., B 1996,

14, 4129-4133.

(20) Willson, C. G. J. Photopolym. Sci. Technol. 2009, 22, 147-153.

(21) Xu, Q. B.; Gates, B. D.; Whitesides, G. M. J. Am. Chem. Soc. 2004, 126, 1332-1333.

(22) Lipomi, D. J.; Chiechi, R. C.; Dickey, M. D.; Whitesides, G. M. Nano

Lett. 2008, 8, 2100-2105. (23) Xu, Q. B.; Bao, J. M.; Capasso, F.; Whitesides, G. M. Angew. Chem., Int.

Ed. 2006, 45, 3631-3635.

(24) Hill, J. The Construction of Timber; Imperial Academy: London, 1770. (25) Pease, D. C.; Porter, K. R. J. Cell Biol. 1981, 91, S287-S292. (26) Malis, T. Microsc. Res. Tech. 1995, 31, 265-266.

93

(27) Antonovsky, A. Microsc. Res. Tech. 1995, 31, 300-307. (28) McMahon, G.; Malis, T. Microsc. Res. Tech. 1995, 31, 267-274. (29) Swab, P. Microsc. Res. Tech. 1995, 31, 308-310.

(30) Barreto, M. P.; Veillette, R.; Lesperance, G. Microsc. Res. Tech. 1995, 31, 293-299.

(31) Glanvill, S. R. Microsc. Res. Tech. 1995, 31, 275-284. (32) Malis, T. F.; Steele, D. Mater. Res. Soc. Symp. Proc. 1990, 199, 3-50.

(33) Acetarin, J. D.; Carlemalm, E.; Kellenberger, E.; Villiger, W. J. Electron Microsc. Tech. 1987, 6, 63-79.

(34) Jesior, J. C. J. Ultrastruct. Mol. Struct. Res. 1986, 95, 210-217.

(35) Matzelle, T. R.; Gnaegi, H.; Ricker, A.; Reichelt, R. J. Microsc. Oxford 2003, 209, 113-117.

(36) Atkins, A. G.; Vincent, J. F. V. J. Mater. Sci. Lett. 1984, 3, 310-312.

(37) Ohring, M. Materials Science of Thin Films, Deposition and Structure; 2nd ed.; Academic Press: San Diego, CA, 2002.

(38) Goldstein, J. N., D.; Joy, D.; Lyman, C.; Echlin, P.; Lifshin, E.; Sawyer,

L.; Michael, J. Scanning Electron Microscopy and X-Ray Analysis; 3 ed.; Springer, 2003.

(39) Villiger, W.; Bremer, A. J. Struct. Biol. 1990, 104, 178-188. (40) Graper, E. B. J. Vac. Sci. Technol., A 1987, 5, 2718-2723.

(41) Kan, C. X.; Zhu, X. G.; Wang, G. H. J. Phys. Chem. B 2006, 110, 4651-4656.

(42) Pokroy, B.; Epstein, A. K.; Persson-Gulda, M. C. M.; Aizenberg, J. Adv.

Mater. 2009, 21, 463-469. (43) Electron Microscopy Sciences, Technical Data Sheet: Epo-fix Cold-

Setting Embedding Resin. http://www.emsdiasum.com/microscopy/technical/datasheet/1232.aspx (accessed March 28, 2010).

94

(44) Nagpal, P.; Lindquist, N. C.; Oh, S. H.; Norris, D. J. Science 2009, 325, 594-597.

(45) Weiss, E. A.; Kaufman, G. K.; Kriebel, J. K.; Li, Z.; Schalek, R.;

Whitesides, G. M. Langmuir 2007, 23, 9686-9694.

(46) Jesior, J. C. J. Ultrastruct. Res. 1985, 90, 135-144.

(47) Ditlbacher, H.; Hohenau, A.; Wagner, D.; Kreibig, U.; Rogers, M.; Hofer, F.; Aussenegg, F. R.; Krenn, J. R. Phys. Rev. Lett. 2005, 95.

(48) Pyayt, A. L.; Wiley, B.; Xia, Y. N.; Chen, A.; Dalton, L. Nat.

Nanotechnol. 2008, 3, 660-665. (49) Xia, Y.; Xiong, Y. J.; Lim, B.; Skrabalak, S. E. Angew. Chem., Int. Ed.

2009, 48, 60-103. (50) Choi, S. Y.; Mamak, M.; Coombs, N.; Chopra, N.; Ozin, G. A. Adv.

Funct. Mater. 2004, 14, 335-344. (51) Cademartiri, L.; von Freymann, G.; Arsenault, A. C.; Bertolotti, J.;

Wiersma, D. S.; Kitaev, V.; Ozin, G. A. Small 2005, 1, 1184-1187. (52) Lipomi, D. J.; Ilievski, F.; Wiley, B. J.; Deotare, P. B.; Loncar, M.;

Whitesides, G. M. ACS Nano 2009, 3, 3315-3325. (53) Lipomi, D. J.; Kats, M. A.; Kim, P.; Kang, S. H.; Aizenberg, J.; Capasso,

F.; Whitesides, G. M. ACS Nano (accepted) 2010.

(54) Studer, D.; Gnaegi, H. J. Microsc. Oxford 2000, 197, 94-100.

(55) Cubukcu, E.; Yu, N. F.; Smythe, E. J.; Diehl, L.; Crozier, K. B.; Capasso, F. IEEE J. Sel. Top. Quantum Electron. 2008, 14, 1448-1461.

(56) Ghadimi, A.; Cademartiri, L.; Kamp, U.; Ozin, G. A. Nano Lett. 2007, 7,

3864-3868.

Chapter 3

Patterning the Cleaved Facets of Optical Fibers with Metallic Nanostructures Using

Nanoskiving

Darren J. Lipomi,1 Mikhail A. Kats,2 Ramses V. Martinez,1 Sung H. Kang,2 Philseok

Kim,2 Joanna Aizenberg,2 Federico Capasso,2 and George M. Whitesides1*

1Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street,

Cambridge, MA, 02138

2School of Engineering and Applied Sciences, Harvard University, 33 Oxford Street,

Cambridge, MA, 02138

96

Abstract

Convenient and inexpensive methods to pattern the facets of optical fibers with

metallic nanoparticles would enable many applications. This communication reports a

method to generate and transfer arrays of metallic nanoparticles to the cleaved facets of

optical fibers. The process relies on nanoskiving, in which an ultramicrotome, equipped

with a diamond knife, sections embedded thin films with an ultramicrotome. Sectioning

produces arrays of nanostructures embedded in thin epoxy slabs, which float on the

surface of water. The tips of optical fibers can be used to capture the floating slabs, and

the nanostructures they contain. Arrays of nanostructures can be transferred at a rate of

approximately 2 min-1, with 88% yield. Etching the epoxy matrices leaves arrays of

nanostructures supported directly by the facets of the optical fibers. Examples of

structures transferred include gold crescents, rings, and gratings of nanowires.

97

Introduction

This paper describes an integrated, rapid, and high-yielding approach to the

fabrication and transfer of two-dimensional (2D) arrays of metallic nanostructures from

the microtome used to make them to the cleaved facets of optical fibers. The process

combines nanoskiving—here, the thin sectioning of patterned epoxy microposts

supporting thin metallic films to produce arrays of metallic nanostructures embedded in

thin epoxy slabs1,2—with manual transfer of the slabs to the optical fibers. The ability to

pattern the facets of optical fibers and other small substrates with nanostructures could

enable several applications, which include sensing based on localized surface plasmon

resonances (LSPRs),3 label-free detection of extremely dilute chemical and biological

analytes4 using surface-enhanced Raman scattering (SERS),5,6 optical filters,7,8 and

diffraction gratings.9 The small sizes and mechanical flexibility of these “optrodes” allow

them to be inserted into the small volumes that are otherwise inaccessible (e.g., the

bloodstream).10

It is extremely challenging to pattern optical fibers, because spin-coating them

with resist is impossible, and mounting them in electron-beam writers and

photolithographic exposure tools is extremely awkward.11 Our process takes advantage of

the easily manipulated slabs of epoxy in which the nanostructures produced by

nanoskiving are embedded. These slabs can be transferred to essentially any smooth (but

not necessarily planar) substrate.

Background

98

Challenges Remaining in Nanofabrication. Nanofabrication is the collection of

methods that generates and arranges structures that have at least one lateral dimension

between 1 and 100 nm.12,13 Silicon integrated circuits, which contain nanoscale

components, are manufactured using scanning-beam and photolithography; these

methods will continue to produce less expensive and more powerful semiconductor

devices for at least the next decade.14,15 The problem of producing complex architectures

on planar, rigid substrates is, thus, largely solved for the foreseeable future.

The ability to pattern unusual substrates—those that are non-planar, mechanically

compliant, incompatible with the conditions and instrumentation of conventional

fabrication, very small, or disposable—could enable new applications in chemistry,16

biology,17 medicine,18 optics,19 and materials for energy conversion and storage.20

Materials such as organic semiconductors,21 carbon nanotubes,22 and wavy silicon,23,24

and methods such as printing25 and molding26 with elastomers, and nanoimprint

lithography,27 and other unconventional processes,12 are filling niches inaccessible to

conventional lithography. There nonetheless remain challenges in nanofabrication for

which a general solution does not yet exist.

Methods of Patterning Optical Fibers. The size and shape of an optical fiber

precludes the use of ordinary lithographic processes. Producing a uniformly thick coating

of resist is a particular challenge. Spin-coating produces a raised region around the

perimeter of a substrate (or edge bead), which for small substrates, can be as large as the

substrate itself.11 Kelkar et al. described a resist that can be deposited by evaporation.

Evaporated resists produce uniform coatings, but the process requires specialized vacuum

equipment, resist, and developer.28 Direct patterning of the optical fibers by focused-ion

99

beam milling is susceptible to contamination of the substrate with gallium atoms.

Inadvertent deposition of gallium can interfere with optical methods of sensing.29

Our laboratories recently described a method by which gold nanostructures

defined by electron-beam lithography (EBL) could be transferred to the facets of optical

fibers using a sacrificial polymeric film.11 While this method was successful in

transferring arbitrary patterns of gold nanostructures to hemispherical substrates and

optical fibers, and was used to produce a bi-directional optical probe for SERS,6 the

method is serial, and each array must be written individually.

Patterning substrates with arbitrary sizes and properties cannot be accomplished

with a single technique of nanofabrication. It is possible, however, to expand the range of

substrates that can be patterned, by dividing patterning into two independent processes:

generation and transfer of structures. This division ensures that the size, shape,

composition, or mechanical properties of the substrate do not affect the generation of the

pattern.

Fabrication and Transfer of Arrays Using Nanoskiving. Nanoskiving is a

simple method of nanofabrication whose key step is sectioning an embedded thin film

with an ultramicrotome equipped with a diamond knife. Sectioning produces slabs

containing nanostructures whose linewidths correspond to the thicknesses of the thin

films. If the film is deposited on an epoxy substrate bearing relief features, sectioning can

produce nanostructures whose geometries correspond to the outlines of the topographic

features. After sectioning, the structures remain embedded in thin epoxy slabs. The slabs

are macroscopic objects that preserve the orientations of the nanostructures within the

arrays and also provide physical handles by which the user can transfer the

100

nanostructures to substrates. After depositing the epoxy slabs on a substrate, the organic

matrix can be removed by etching with an oxygen plasma.

Experimental Design

Choice of Nanoskiving. There are very few methods of fabrication in which the

processes of generating and transferring patterns are completely decoupled, beside decal

transfer printing,30 in which the structures must be fabricated serially and transferred to a

stamp. We chose nanoskiving, because it offered an integrated approach to both the

fabrication and transfer of nanostructures: we reasoned that we could capture floating

slabs manually with the cleaved surfaces of optical fibers, in the same way that

microtomed specimens are retrieved using TEM grids.

Nanostructures. We chose three types of nanostructures to transfer to optical

fibers: crescents, rings, and a grating of nanowires. These structures are interesting for

their optical and plasmonic properties. Crescents and rings can be made by cutting an

array of metalized epoxy nanoposts,31 while gratings of parallel nanowires can be made

by cutting gold-coated molded epoxy washboards bearing long, rectangular ridges.2 We

used gold for all nanostructures because it has useful plasmonic properties, is not

oxidized in the ambient atmosphere, and it is soft enough to produce a high yield of non-

defective nanostructures when sectioned.32

Fibers. We chose a typical multimode optical fiber with a total diameter of 125

µm and a core diameter of 50 µm. We chose this fiber so that we could eventually

interrogate the nanostructures for SERS using 785 nm light. To prepare the fibers for the

101

transfer of slabs, we stripped the polymeric jacket, and cleaved the silica fiber manually

with a diamond scribe, or with an electronic fusion splicer.

Results and Discussion

Figure 3.1 (step 1) shows a schematic drawing of the process used to generate

nanostructures (gold rings, in the example shown). The block—containing embedded

gold-coated posts—contacts a single-crystalline diamond knife, which has a wedge angle

of 35° and whose edge has a radius of curvature of 3 – 6 nm.33 As the knife passes

through the block, the nascent slab (area ≤ 1 mm2) containing gold rings floats onto the

surface of a water-filled trough (step 1). We have produced as many as 60 slabs bearing

arrays of nanostructures from a single embedded array of nanoposts.31 The color of a

slab, produced by the interference of white light, can be used to estimate its thickness:

grey, < 60 nm; silver, 60 – 90 nm; gold, 90 – 150 nm; purple, 150 – 190 nm; blue, 190 –

240 nm; green, 240 – 280 nm; yellow, 280 – 320 nm.34 We used 80 – 150 nm slabs for all

experiments.

Steps 2 – 4 of Figure 3.1 summarize the procedure used to transfer the floating

arrays to the facets of optical fibers. We captured a floating slab with the tip of a fiber, by

holding fiber with tweezers, and pressing down manually on the slabs, from above, with

the tip of the fiber (step 2). As we drove the slab under the surface of the water using the

tip, the slab wrapped itself around the tip. When we withdrew the tip from the water bath,

the slab was attached irreversibly to the tip. We allowed the water to evaporate (step 3),

while we repeated the process with another fiber. Each transfer took approximately 30 s.

After the water evaporated, the appearance of a slab covering the tip of a fiber resembled

102

Figure 3.1. Summary of the procedure used to fabricate and transfer arrays of metallic

nanostructures to the facets of optical fibers.

103

a tablecloth draped over the top and sides of a round table. After allowing the tip to dry in

the ambient atmosphere, exposure to an air plasma using a bench-top plasma cleaner (1

torr, 100 W, 10 min) revealed free-standing nanostructures on the facet of the fiber (step

4). Figure 3.2a shows a low-magnification scanning electron microscope (SEM) image of

an optical fiber. Figure 3.2b is a close-up of the facet of a fiber bearing an array of gold

crescents. The pattern extends to the edge of the circular facet. The structures on the

curved sides of the fiber presumably fall away during the plasma etching of the epoxy

matrix.

The process can be used for any structure or array of structures that can be

produced using nanoskiving. Figure 3.2c shows a grating of parallel gold nanowires with

linewidths of 50 nm.2 The nanowires span the entire facet. Iterative capture of slabs,

followed by plasma etching, yielded overlapping arrays of gold nanostructures. Figure

3.2d shows the core of a fiber bearing an overlapping pattern of concentric gold rings (see

inset). We fabricated the rings using a previously published procedure.31

Defects. There are two classes of defects that occur during the process described.

The first class comprises those that occur because of the mechanical stresses of

sectioning, combined with the intrinsic brittleness of evaporated films, and the

compressibility of the epoxy matrix.32 Fracture of individual structures and global

compression in the direction of cutting are the two most prominent defects observed, and

were described in detail in an earlier report.32 The second class of defects includes those

that occur because of the transfer. Folding of the slabs, in which a crease runs across the

facet of a fiber, is the most prominent type of defect due to the transfer.

104

Figure 3.2. (a) An optical fiber bearing an array of gold crescents like the four shown in

the inset. (b) A close-up of the facet. The inset is a single gold crescent. (c) A facet

bearing a grating of gold nanowires. (d) The core of a fiber bearing two overlapping

arrays of concentric gold rings.

105

In one experiment, we transferred 16 arrays of 80-nm-thick slabs bearing gold

crescents, and examined the tips of the fibers for defects by SEM. The yield of successful

transfers was 14/16 (88%). We defined a successful transfer as one in which the core of

the optical fiber was completely covered by the array of nanostructures with no folds.

The two unsuccessful transfers of this group exhibited folds that ran across the core of the

optical fiber. Figure 3.3 shows an example of one of the two defective transfers. A fold in

the slab produced a boundary in the array, where the orientations of the axes of the arrays

were different on either side of the fold. After etching the epoxy with an air plasma, the

nanostructures were jumbled in the immediate vicinity of the fold. The tendency of the

slabs to fold is exacerbated by approaching the slab with the tip of the fiber from the

bottom surface, under the water. To suppress folding, and to ensure that the arrays of

nanostructures are distributed homogeneously across the surface of the fiber, the fiber

must approach the floating slab from a perpendicular trajectory. This motion is easier

when approaching from the top, and nearly impossible from the bottom (the length and

finite bending radius of the fiber make it impossible to lift the slabs out from the bottom

of the water-filled boat of the ultramicrotome).

In order to show that we could address the nanostructures optically, we coupled

the fiber to a halogen light source and obtained optical microscope images of the cleaved

facet of the fiber. Figure 3.4 shows light scattering from an array of gold crescents on top

of the core of the optical fiber. While the crescents covered the entire facet (core and

cladding), light scattered only from those on the core. Scatterers with well-defined shapes

could be used to couple light of specific frequencies into and out of the fiber. Metallic

106

Figure 3.3. A facet bearing a poorly transferred array of crescents. There was a fold in

the epoxy slab as it was transferred to the fiber. The inset shows a region of the fold,

which contains jumbled nanostructures.

107

Figure 3.4. Dark-field optical image of a cleaved facet of an optical fiber supporting an

array of gold crescents. The core of the fiber is illuminated using a halogen source. The

inset is a close-up of the facet. Light, supplied by the fiber, scatters from the gold

nanostructures on the core of the optical fiber, and not the cladding. The fiber is tilted ~

45° to isolate only the light scattered from the nanostructures (tilting prevents light from

the fiber from entering the objective of the microscope directly).

108

nanostructures with designed plasmon resonances would enable sensing based on SERS

or LSPR.6

Conclusions

Nanoskiving is well—perhaps uniquely—suited for fabricating and transferring

patterns of nanostructures to the tips of optical fibers and other small substrates. In

addition to producing arrays of metallic nanostructures that are potentially useful for

plasmonic applications,2,8 nanoskiving produces these arrays in transferrable epoxy slabs

that can be placed conveniently on substrates of many compositions, sizes, and

topographies. All structures produced by nanoskiving—two- and three-dimensional

arrays of metallic, dielectric, and semiconducting particles, gratings of nanowires,2

single-crystalline gold nanowires,35 and conjugated polymers36 —can be mounted on the

facets of fibers using the same process. We also believe that this process could be

extended to other delicate films (e.g., photoresist, conjugated polymers, and other

materials floated on the surface of water) that would be difficult to deposit on the fibers

directly.

Acknowledgements

This research was supported by the National Science Foundation under award

PHY-0646094. The authors used the shared facilities supported by the NSF under

MRSEC (DMR-0213805 and DMR-0820484). This work was performed in part using the

facilities of the Center for Nanoscale Systems (CNS), a member of the National

Nanotechnology Infrastructure Network (NNIN), which is supported by the National

109

Science Foundation under NSF award no. ECS-0335765. CNS is part of the Faculty of

Arts and Sciences at Harvard University.

References

(1) Xu, Q. B.; Rioux, R. M.; Dickey, M. D.; Whitesides, G. M. Acc. Chem. Res. 2008, 41, 1566-1577.

(2) Xu, Q.; Rioux, R. M.; Whitesides, G. M. ACS Nano 2007, 1, 215-227. (3) Sharma, A. K.; Jha, R.; Gupta, B. D. IEEE Sens. J. 2007, 7, 1118-1129. (4) Wolfbeis, O. S. Anal. Chem. 2008, 80, 4269-4283.

(5) Kneipp, K.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S. J. Phys. Condens. Matter 2002, 14, R597-R624.

(6) Smythe, E. J.; Dickey, M. D.; Bao, J. M.; Whitesides, G. M.; Capasso, F.

Nano Lett. 2009, 9, 1132-1138. (7) Paul, K. E.; Zhu, C.; Love, J. C.; Whitesides, G. M. Appl. Opt. 2001, 40,

4557-4561. (8) Xu, Q. B.; Bao, J. M.; Rioux, R. M.; Perez-Castillejos, R.; Capasso, F.;

Whitesides, G. M. Nano Lett. 2007, 7, 2800-2805. (9) Halpern, A. R.; Nishi, N.; Wen, J.; Yang, F.; Xiang, C. X.; Penner, R. M.;

Corn, R. M. Anal. Chem. 2009, 81, 5585-5592.

(10) Stokes, D. L.; Vo-Dinh, T. Sens. Actuators, B 2000, 69, 28-36.

(11) Smythe, E. J.; Dickey, M. D.; Whitesides, G. M.; Capasso, F. ACS Nano 2009, 3, 59-65.

(12) Stewart, M. E. M., M. J.; Yao, J.; Thompson, L. B.; Nuzzo, R. B. Proc.

IMechE Part N: J. Nanoengineering and Nanosystems 2007, 220, 81-138. (13) Gates, B. D.; Xu, Q. B.; Stewart, M.; Ryan, D.; Willson, C. G.;

Whitesides, G. M. Chem. Rev. 2005, 105, 1171-1196.

(14) Pease, R. F.; Chou, S. Y. Proc. IEEE 2008, 96, 248-270. (15) Willson, C. G.; Roman, B. J. ACS Nano 2008, 2, 1323-1328.

110

(16) Zhang, J. L.; Han, Y. C. Chem. Soc. Rev., 39, 676-693.

(17) Kane, R. S.; Takayama, S.; Ostuni, E.; Ingber, D. E.; Whitesides, G. M. Biomater. 1999, 20, 2363-2376.

(18) Hong, J.; Edel, J. B.; deMello, A. J. Drug Discovery Today 2009, 14, 134-

146. (19) Stewart, M. E.; Anderton, C. R.; Thompson, L. B.; Maria, J.; Gray, S. K.;

Rogers, J. A.; Nuzzo, R. G. Chem. Rev. 2008, 108, 494-521.

(20) Hochbaum, A. I.; Yang, P. D. Chem. Rev., 110, 527-546.

(21) Menard, E.; Meitl, M. A.; Sun, Y. G.; Park, J. U.; Shir, D. J. L.; Nam, Y. S.; Jeon, S.; Rogers, J. A. Chem. Rev. 2007, 107, 1117-1160.

(22) LeMieux, M. C.; Bao, Z. N. Nat. Nanotechnol. 2008, 3, 585-586. (23) Kim, D. H.; Rogers, J. A. Adv. Mater. 2008, 20, 4887-4892.

(24) Kim, D. H.; Ahn, J. H.; Choi, W. M.; Kim, H. S.; Kim, T. H.; Song, J. Z.; Huang, Y. G. Y.; Liu, Z. J.; Lu, C.; Rogers, J. A. Science 2008, 320, 507-511.

(25) Perl, A.; Reinhoudt, D. N.; Huskens, J. Adv. Mater. 2009, 21, 2257-2268.

(26) Xia, Y. N.; Kim, E.; Zhao, X. M.; Rogers, J. A.; Prentiss, M.; Whitesides, G. M. Science 1996, 273, 347-349.

(27) Willson, C. G. J. Photopolym. Sci. Technol. 2009, 22, 147-153.

(28) Kelkar, P. S.; Beauvais, J.; Lavallee, E.; Drouin, D.; Cloutier, M.; Turcotte, D.; Yang, P.; Mun, L. K.; Legario, R.; Awad, Y.; Aimez, V. J. Vac. Sci. Technol., A 2004, 22, 743-746.

(29) Fu, Y. Q.; Bryan, N. K. A. Appl. Phys. B 2005, 80, 581-585.

(30) Meitl, M. A.; Zhu, Z. T.; Kumar, V.; Lee, K. J.; Feng, X.; Huang, Y. Y.; Adesida, I.; Nuzzo, R. G.; Rogers, J. A. Nat. Mater. 2006, 5, 33-38.

(31) Lipomi, D. J.; Kats, M. A.; Kim, P.; Kang, S. H.; Aizenberg, J.; Capasso,

F.; Whitesides, G. M. ACS Nano (accepted) 2010. (32) Lipomi, D. J.; Martinez, R. V.; Rioux, R. M.; Cademartiri, L.; Reus, W.

F.; Whitesides, G. M. Submitted 2010.

111

(33) Matzelle, T. R.; Gnaegi, H.; Ricker, A.; Reichelt, R. J. Microsc. Oxford

2003, 209, 113-117.

(34) Peachey, L. D. J. Biophys. Biochem. Cytol. 1958, 4, 233-348.

(35) Wiley, B. J.; Lipomi, D. I.; Bao, J. M.; Capasso, F.; Whitesides, G. M. Nano Lett. 2008, 8, 3023-3028.

(36) Lipomi, D. J.; Chiechi, R. C.; Reus, W. F.; Whitesides, G. M. Adv. Funct.

Mater. 2008, 18, 3469-3477.

Chapter 4

Transistors Formed from a Single Lithography Step Using Information Encoded

in Topography

Michael D. Dickey,1 Kasey J. Russell,2 Darren J. Lipomi,1 Venkatesh Narayanamurti,2

and George M. Whitesides1

1Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford St.,

Cambridge, Massachusetts 02138, USA

2School of Engineering and Applied Sciences, Harvard University, 29 Oxford St.,

Cambridge, MA 02138, USA

113

Abstract

This paper describes a strategy for the fabrication of functional electronic

components (transistors, capacitors, resistors, conductors, and logic gates but not, at

present, inductors) that combines a single layer of lithography with angle-dependent

physical vapor deposition; this approach is named topographically encoded

microlithography (abbreviated as TEMIL). This strategy extends the simple concept of

‘shadow evaporation’ to reduce the number and complexity of the steps required to

produce isolated devices and arrays of devices, and eliminates the need for registration

(the sequential stacking of patterns with correct alignment) entirely. The defining

advantage of this strategy is that it extracts information from the 3D topography of

features in photoresist, and combines this information with the 3D information from

the angle-dependent deposition (the angle and orientation used for deposition from a

collimated source of material), to create ‘shadowed’ and ‘illuminated’ regions on the

underlying substrate. It also takes advantage of the ability of replica molding

techniques to produce 3D topography in polymeric resists. A single layer of patterned

resist can thus direct the fabrication of a nearly unlimited number of possible shapes,

composed of layers of any materials that can be deposited by vapor deposition. The

sequential deposition of various shapes (by changing orientation and material source)

makes it possible to fabricate complex structures—including interconnected

transistors—using a single layer of topography. The complexity of structures that can

be fabricated using simple lithographic features distinguishes this procedure from

other techniques based on shadow evaporation.

114

Introduction

This paper describes the fabrication of single transistors (and small groups of

transistors) by a process that combines a single level of topography (defined by

photolithography or molding, for example) on a substrate with multiple shadow

evaporations. We call this technique topographically encoded microlithography

(TEMIL). It encodes all the information needed to fabricate a complex structure—

such as a transistor—in the topography of patterned polymer. It has the conceptual

attraction that it replaces several steps of lithography—each of which require

registration—with a single step of patterning and several steps of oriented deposition.

Most lithographic techniques (e.g., photolithography, e-beam lithography, and imprint

lithography), are generally treated as two-dimensional (2D); that is, even though the

lithographically defined features of photoresist have height, it is the shape of the

opening in the photoresist in the 2D plane that defines the shape of the pattern on the

underlying substrate. Using both the opening and the height of the features

(topography) increases significantly the effective complexity of a pattern derived from

a single layer of topography. Although the demonstrations in this work involve only a

single level of topography, the same principles apply to more complex surfaces.

The progress of microelectronic technology described by Moore’s Law has

largely been driven by the economics of reductions in the cost per transistor.

Advances in photolithography have coupled “cost” and “feature size”: that is,

“smaller” has, for several decades, also been “less expensive”. There are, of course,

other ways of reducing cost (especially by reducing reliance on photolithography;

hence the active current interest in methods of fabrication based on molding and

115

imprinting1-4). Molding, in particular, has the ability to reproduce three-dimensional

(3D) features with nanoscale accuracy,5,6 and imprint lithography can pattern features

over large areas rapidly and economically.4,7-9 The question then becomes, “How does

one convert a surface with information encoded in topography into electrical (or

optical, or magnetic) function?” Most approaches to this question have transferred

patterns directly to the exposed regions of the substrate (i.e., those that are not covered

by polymer); the function of the lithographic features is therefore two-

dimensional. This paper demonstrates the value of the information coded in the third

dimension of a thin, structured polymer film to generate electronic function. The

structured film represents an abstraction of the final device—rather than a literal

image of the desired structure—because the information is coded in the 3D

topography rather than the 2D footprint.

As a demonstration of this method, we used multiple, angle-dependent

depositions (“shadow evaporations”) over a single layer of lithographically defined

features to fabricate single and interconnected transistors (metal-oxide semiconductor

field-effect transistors, MOSFETs, interconnected to form an AND gate), and

conductive pathways. The architectures of these devices are defined by the topography

of the photoresist, and by the orientation of the substrate with respect to the

evaporation source; each component of the transistor (source, drain, gate dielectric,

gate) is defined by a unique orientation. The composition of the transistor is defined

by the choice of evaporated materials. This method produces transistors without any

doping, etching, or lithographic alignment steps. In principle, a fully developed

116

process would reduce the time, materials, and complexity of equipment to form a

transistor compared to conventional methods (multi-step lithography).

It is well known that physical obstructions, such as topography or structures

suspended over a surface, cast shadows when placed in the path of a collimated

deposition source. Shadow deposition through stencil masks has been used to fabricate

nanostructures and devices (e.g, nanowires, rings, sensors), but the masks must be

registered with the substrate prior to each deposition, the apertures of the stencil

reduce in size during deposition, and the method is effectively as two-dimensional as

photolithography.10-18 Structures (“bridges”) suspended across lithographically defined

openings have been used to create simple structures by shadow evaporation.19 This

approach alleviates the need for registration, but fabricating these structures is

challenging. This approach is used to fabricate small (~0.010 mm2) metal-insulator-

metal tunnel junctions by depositing a metal wire that overlaps an oxidized metal

wire.20-24 The use of this technique to fabricate more complex structures has been

limited by its complexity and the inadvertent shadows cast by the suspended

structures. Shadow evaporation over the edges of photoresist has been used to deposit

simple nanowires (20-30 nm wide) that are smaller in size than the openings in the

resist.25 Shadow evaporation over topography (e.g., metal wires, arrays of self-

assembled spheres, the pores of alumina membranes) has also been used previously to

form small electrodes,26-32 arrays of nanotubes,33 magnetic wires,34 gradient

structures,35 molecular junctions,36 structures that show Coulomb blockades,37 and

simple nanostructures.38-40 TEMIL can fabricate complex (multi-layer, functional,

interconnected) microsystems without the issue of registration involved in the use of

117

stencil masks, and without the complexities associated with suspended bridge

structures. The method described here suggests a different approach to the use of

topography to encode information useful in the fabrication of functional,

interconnected micro- and nanosystems.

TEMIL

As a demonstration of TEMIL, we have fabricated individual MOSFET’s (in

arrays containing ~1600 devices), and a pair of transistors connected to form an AND

gate with resistive and conductive pathways, all using a single layer of

photolithography. Figure 4.1 illustrates how multiple shapes can be fabricated on a

substrate from a single simple lithographic feature (e.g., one involving only a single

layer of topography) by using TEMIL. Figure 4.1a depicts an opening patterned into a

film (e.g., a “resist”, which could either be a photoresist or a molded polymer) of

thickness Z. A collimated beam of evaporating metal (gold, as depicted) is aligned

parallel to dimension X and oriented at an angle F relative to the surface of the resist.

Although the opening in the film is a polygon with eight sides, the material deposited

on the substrate is a square. A square forms because shadow evaporation is a “line-of-

sight” technique41 in which a collimated beam deposits material on a substrate in a

pattern determined by (i) the topography on the substrate and (ii) the orientation of the

substrate with respect to the source of material. Shadow evaporation creates

simultaneously “shadowed” and “illuminated” regions on the portions of the substrate

not covered by polymer. In Figure 4.1a, Au deposits on only a fraction of the exposed

substrate; we refer to this portion as “illuminated” and the remaining portion as

118

“shadowed”. In general, the length of a shadow cast by a beam over topography is

Z/tan(F), in which Z is the height of the resist and F is the angle between the beam and

the plane of the resist. In Figure 4.1a, the shadow is longer than X2 but shorter than X1.

As a result, the substrate is only illuminated in region Y2 (and the patterned formed is

completely independent of dimension Y1). The deposited material therefore has

dimensions Y2 and X1- Z/tan(F).

Figure 4.1b illustrates how the shape of the material deposited on the substrate

changes with the angle F of the beam relative to the resist. As F increases (with all

other variables remaining constant), the length of the shadowed region decreases and

the illuminated region increases. Each change in F produces a shape with unique

dimensions. In some cases, the change in the shape arising from these changes can be

dramatic. For example, in Figure 4.1b, the shape changes from a rectangle to a square

to a polygon. When F reaches 90°, the beam is perpendicular to the substrate and the

shape of the illuminated region on the substrate replicates the shape of the opening;

this condition is used during conventional patterning.

Figures 1c illustrates how the shape of the illuminated region is affected by

varying the angular orientation (q) of the substrate (i.e., by rotating the substrate

within the X-Y plane) while keeping the beam orientation (F) constant. The angular

orientation q dictates the region of the substrate that is shadowed for a given beam

angle F, and changing q can also entirely change the pattern of deposited material

(from a triangle to a rectangle in Figure 4.1c). Complex structures, such as a

MOSFET, can be fabricated by combining multiple, sequential depositions of various

materials, each at a unique orientation. Using this method, each functional layer of the

119

Figure 4.1. Conceptualization of shadow evaporation. Multiple shapes can be formed

with one lithographic feature. (a) A rectangular-prism shaped relief structure is

defined on a substrate. The angle F defines the orientation of the vectorial path of

depositing gold atoms with respect to the substrate (or the X-Y plane). This beam is

parallel to the X dimension. The angle q defines the angular orientation of the

substrate within the X-Y plane. In this particular example, the photoresist casts a

shadow of length Z/tan(F) onto the base of the opening; thus, material deposits over a

length X1- Z/tan(F). The deposited material has a width of Y2 and is effectively

independent of dimension Y1. (b) Increasing the angle F, while keeping the other

parameters constant, decreases the length of the cast shadow and the resulting

structures vary accordingly (the features are depicted before and after removal of the

polymer). (c) Varying the orientation of the rotational orientation of the substrate, q,

while keeping the other parameters constant, determines the direction that the shadow

is cast.

120

Figure 4.1 (Continued)

121

device can be deposited independently, using a single level of topography.

Experimental Design

We fabricated MOSFETs using only one layer of lithography, and shadow

evaporation, to demonstrate the utility of TEMIL for fabrication. MOSFETs—

ubiquitous in modern integrated circuits—are complex, multicomponent structures.

MOSFETs typically contain regions directly under the source and drain pads that are

doped to levels that give conductivities greater than that in the channel (often the

inherent doping of the substrate). We did not incorporate contact-specific doping steps

beyond the inherent doping of the wafer. This approach has been used by others to

fabricate back-gated transistors.42 We used a “depletion mode” MOSFET design (see

supplemental Figure 4.6) in which charge carriers are depleted from the channel by a

bias applied to the gate.43 This approach requires the use of silicon-on-insulator (SOI)

wafers that have a thin (<0.2 mm thick) silicon layer, which confines the channel

through which the charge carriers conduct to a depth predefined by the thickness of

the wafer.

In this demonstration, we fabricated features on the 50-100 mm length scale

because these dimensions are (i) easy to fabricate using commonly available

lithographic tools (contact lithography) and (ii) simple to characterize electrically

because the source, drain, and gate pads can be addressed directly using a wire bonder

(wire bonds are ~50 mm in diameter). We designed the layout of the photoresist

features such that each transistor could be electrically addressed with a wire bonder

without concern of overlapping with neighboring devices (~200 mm separated the

transistors). The design was therefore intended to facilitate addressing rather than

maximizing feature density. We typically fabricated arrays of ~1600 transistors on

122

each substrate (1 x 1”). The supporting information contains an image of an array of

transistors formed using TEMIL. To show that more than one transistor formed by

shadow evaporation could be linked together using features in the same single layer of

patterned resist, we fabricated AND devices by connecting the source and drain pads

of adjacent transistors.

The experimental section contains experimental details; a brief summary

follows. Onto a 1 x 1” piece of SOI wafer (~0.2 mm Si on ~1 mm oxide), we spin-

coated a thin film (~1 mm) of poly(methylglutarimide) (LOR 10 B, Microchem). On

top of this film, we patterned features in a 50-mm-thick film of SU-8 2025

(Microchem) using contact photolithography. A typical pattern was a 40 x 40 array of

transistors. The target height of the features (i.e., the thickness of the SU-8 film)

ultimately depended on the desired angle of evaporation and the dimensions of the

patterned features. The poly(methylglutarimide) was necessary for the lift-off process,

because SU-8 is an insoluble, highly crosslinked polymer that has excellent adhesion

to silicon wafers. Bilayer lithography schemes are used routinely for other patterning

techniques such photolithography,44 imprint lithography,8 and self-assembled block

copolymer lithography,45 so this approach is not limited to SU-8, and should be

applicable to patterning layers made of other polymers, ceramics (e.g., thick SiO2 or

other metal oxides), or even metals.

Figure 4.2 outlines the sequence of steps involved in the fabrication of a

MOSFET using one layer of lithography. The orientation of the substrate is critical

during fabrication. We define fixed laboratory coordinates (X, Y, Z) in Figure 4.2 such

that the Y-axis is always perpendicular to the vector path of the beam of depositing

material. Changes in the orientation of the beam vector, F, therefore occur in the X-Z

123

Figure 4.2. Top-down schematic of the formation of a transistor by shadow

evaporation on a silicon-on-insulator substrate. (i) We first defined features by

photolithography. (ii) We evaporated aluminum to form the source pad. We oriented

the substrate such that the evaporating material would follow the path denoted by the

dotted arrows (q = 0°). The topographical resist features cast shadows such that the

metal only reaches the substrate in a defined region. (iii) To define the drain, we

rotated the sample 180° in the plane of the substrate and evaporated aluminum. (iv)

We then rotated the sample 90° and evaporated oxide to electrically insulate the

source and drain pads from the gate. (v) We rotated the sample 180° and evaporated

oxide to provide more dielectric insulation, and without changing the orientation, (vi)

evaporated aluminum to define the gate electrode. We lifted-off the resist features

using solvent, leaving behind a MOSFET architecture. The surrounding silicon could

be selectively etched to improve the performance of the device by confining electron

transport to regions under the deposited features.

124

Figure 4.2 (Continued)

125

plane. Prior to each deposition, the substrate is rotated to a specific orientation within

the X-Y plane. We arbitrarily define q = 0° as the orientation of the substrate depicted

in the first processing step in Figure 4.2. We typically deposited the gate dielectric

from two opposing angles (θ = 90° and 270°) to prevent electrical shorting between

the gate metal and the substrate.

Following the final deposition, immersion of the substrate in acetone lifted-off

the resist. A subsequent immersion in an etching solution of tetramethyl ammonium

hydroxide (TMAH), dissolved Si, and ammonium peroxydisulfate ((NH4)2S2O8)

removed selectively the Si surrounding the features of the transistor,46,47 thereby

eliminating any conductive pathways around the periphery of the device. This

formulation etches Si preferentially over silicon oxide, aluminum oxide, and

aluminum, allowing the device itself to be used as an etch mask and the oxide layer of

the SOI wafer to be used as an etch stop. An optical micrograph of a completed device

is shown in Figure 4.3. We made individual electrical contacts from the source, drain,

and gate of each MOSFET using an aluminum wire bonder and characterized the

device electrically using a semiconductor parameter analyzer.

Results and Discussion

MOSFET. Figure 4.4 shows the traces of current vs. applied voltage (I-V)

between the drain and source for various gate voltages, and shows its output

characteristics. We achieved an on:off ratio (defined as the ratio of current at VG = 0 V

and VG = -40 V at VDS = 1 V) of 4.5 x 102. We measured the leakage current through

the gate to the drain at VG = 40 (VDS=0) to be four orders of magnitude lower than that

126

Figure 4.3. (left) Optical micrograph of a field effect transistor formed by combining

one layer of lithography and shadow evaporation. This image depicts one of an array

of 1600 identical devices. (right) A schematic illustration to clarify the features

depicted in the micrograph.

127

Figure 4.4. Electrical characterization of a MOSFET formed by shadow evaporation.

Current, IDS (mA) between the drain and source as a function of drain-source bias

(VDS) for various gate voltages.

128

through the channel for VDS = 1 V (VG=0 V); this result suggests that the gate

dielectric electrically insulated the gate from the substrate. Both the transconductance

and gate leakage could presumably be improved by optimizing the properties of the

gate dielectric. The substrates typically contained ~1600 transistors. Based on visual

inspection, approximately 10-20% of these transistors had obvious flaws (e.g.,

particles) that presumably arose from extensive processing outside of a cleanroom.

AND Gate. TEMIL can also be used to fabricate interconnected MOSFETs

using a single layer of lithography. We fabricated two connected devices using the

same process that was used for the fabrication of the MOSFETs. As a proof of

principle, we fabricated an AND gate that incorporates the same basic MOSFET

design shown in Figure 4.2, and included an electrically conductive connector between

the drain pad of one device and the source pad of the adjacent device. In principle, this

approach could be adapted to connect a large number of transistors; developing design

rules for such systems will require integrating materials and topography with circuit

design. Figure 4.5a is a circuit diagram for the AND gate fabricated here.

Figure 4.5b outlines the procedure used to fabricate the electrical connector (a

continuous wire) between adjacent transistors. The connector consists of three sections

that are fabricated over the course of four depositions. One section is deposited with

metal concurrently with the source and drain depositions (θ = 0° and 180°) and the

other sections consist of two symmetrical, shorter pathways that are deposited with

metal at an incident angle of F = 16° and orientations of q = 45° and 135°. These

depositions link the sections to the source and drain pads of adjacent transistors. All of

the sections of the connector are shadowed during deposition of the gate dielectric

(i.e., when the substrate is oriented at θ = 90° and 270°) to ensure they are electrically

129

Figure 4.5. One layer of lithography and shadow evaporation produced an AND logic

device by connecting two transistors in series. (a) A circuit diagram of the device. (b)

Schematic illustration depicting the way by which a drain pad from one transistor

connects to the source of an adjacent transistor. The region of the connection that is

parallel to the source and drain is formed during the evaporation of the source and

drains. The remaining two portions of the connectors are formed by two separate

evaporations (q = 45°, as depicted by line labeled “deposition path” and q = 135°). (c)

A top-down optical micrograph of two transistors with an ohmic connection.

130

continuous. (The Supporting Information includes a description of the fabrication of a

resistor—three hundred times more resistive than the channel at zero gate bias—by

incorporating insulating materials in the connector.) The conductivity of the metallic

connector is 2.4 x 107 W-1 m-1 (the value of bulk aluminum is 3.7 x 107 W-1 m-1). The

additional resistance is expected since the substrate is exposed to atmosphere between

depositions. Our experimental configuration required us to open the evaporation

chamber to orient the substrate between depositions; the exposure to oxygen likely

resulted in imperfect junctions between the individual sections of the connector.

Figure 4.5c is an optical micrograph of a completed AND gate. The measured

current from the input and output (Figure 4.5a) as a function of voltage applied to each

gate is listed in Table 4.1. This result demonstrates AND gate functionality, although

practical use would require improved performance for the (1,0) state.

Advantages and Limitations of TEMIL

Conventional transistor fabrication requires the registration of multiple

lithographic steps. TEMIL requires only one lithographic step and no registration, but

it does require a collimated beam for metal deposition, and alignment of the substrate

relative to the beam (q, F). In all of our experiments, we set the angular orientation of

the substrate (q) and the beam orientation relative to the substrate (F) crudely, by eye,

with the aid of a protractor. We designed the depositions to have a tolerance of D F ~

2°. This tolerance is effectively constant regardless of feature scaling, whereas

conventional registration becomes increasingly difficult with reduced feature size.

Simply scaling all of the dimensions of the structured polymer film (X, Y, and Z) by

131

Table 4.1. Truth Table for an AND Device (for VDS = 2 V).

132

the same factor results in the proportionate scaling of the deposited features without

the need to adjust the deposition orientation (F, q). To some limit, a topographical

design for shadow evaporation should therefore work as well for small as for large

features, provided that the features are proportional.

There is, of course, a lower limit to the size of features that are practical to

fabricate using shadow evaporation. In principle, the resolution of shadow evaporation

at small (<100 nm) length scales is limited by several factors, all of which broaden and

blur the edge of the illuminated region: (i) imperfect collimation of the evaporation

beam; (ii) uncertainty in F and q; (iii) surface migration of the depositing species on

the substrate;48 iv) irregularities in the topography of the patterned film and the edge

of the pattern; and v) change in the topography of the shadowing edge as deposition

proceeds. These factors are discussed in the supporting information. Despite these

junctions with sub-10-nm gaps.26,29,30 Stencil masks, which also cast shadows, have

also been used to make 10-nm features.12

We discuss other advantages, limitations, and considerations of TEMIL in the

Supporting Information.

Conclusions

Topographically encoded microlithography (TEMIL) is an integrated strategy

for fabricating microscale devices combining a single lithographic step with multiple

shadow evaporations. The current level of demonstration of this methodology is

limited to simple components: a MOSFET, an AND gate (two connected MOSFETS),

a capacitor (i.e., the gate of the transistor), and a conducting wire with several angles

133

and connections. We have, however, fabricated ~1600 of these components in parallel.

The process is scalable, applicable to a wide variety of devices and materials, and

compatible with any patterning technique that produces topography (photolithography,

soft lithography,49 step-and-flash imprint lithography,1 imprint lithography,3

nanoskiving,50 anisotropic etching,51 and block copolymer lithography52).

Conventional multi-layer photolithography requires registration (the sequential

stacking of patterns with correct alignment) and has poor tolerance for lateral drift

(i.e., the deviation from the designed spacing between features on a single level of

lithography). Soft molding techniques (e.g., soft lithography), which are particularly

attractive because of their low cost and high resolution,5,6 are particularly subject to

lateral drift because of the softness of the stamp; Step-and-Flash Lithography (SFIL)

developed by Willson combines many of the desirable features of both

photolithography and soft lithography. By encoding the information required for

fabrication in the topography of a single layer of polymer, TEMIL has the potential to

eliminate multiple steps of lithography and pattern registration; it may also be tolerant

of the lateral drift that limits soft lithographic techniques.

TEMIL has the ability to accommodate a wide variety of materials. The

“illuminated” region created by shadow evaporation can be used directly as a feature

(as demonstrated here), or can be used to form a mask to protect the underlying

substrate while the “shadowed” region was removed by etching. In our hands, the

throughput has been limited by the requirements of manual changes of evaporation

sources and manual setting of angles, but the efficiency and versatility of the process

could be easily improved by using multiple beams simultaneously or sequentially from

134

different values of F and q using an automated deposition chamber. More importantly,

hierarchical resist structures (i.e., structures with two or more heights), would cast

shadows of differing lengths depending on the height of each structure; using 3D

shadow masks would provide another variable that can be used to alter the geometry

of the deposited structure.

Experimental

Design. The schematic diagram of a MOSFET in Figure 4.6 consists of two

aluminum pads, separated by a defined distance. The gate electrode is also aluminum

and is separated from the substrate by a thin dielectric (e.g. SiO2). A bias is applied

across the source and drain and the current that flows between the two pads is

modulated by the bias applied to the gate. The bias of the gate controls the

concentration of charge carriers in the channel (defined as the region of doped silicon

connecting the source and drain). We removed the extra silicon surrounding the device

(not shown) to eliminate alternative conduction pathways between the source and

drain.

Resistors. We formed a resistor using an approach similar to that used to form

the ohmic connector (Figure 4.5). The MOSFETs are fabricated by the sequential

deposition of metal (conductor) and oxide (dielectric). We therefore sought a material

with an intermediate conductivity –the thin, top layer of silicon of the SOI substrate—

to form the resistive elements. A wet etch removes all of the exposed silicon during

the final

135

Figure 4.6. A cut-away schematic of the transistor architecture that we sought to

fabricate by shadow evaporation. The source, drain, and gate pads are metal (e.g. Al).

The gate pad is separated from the substrate by an insulating dielectric layer (e.g.

SiO2). The substrate is a silicon-on-insulator (SOI) wafer and the insulating layer

confines the carriers to the top silicon layer. Carriers conduct through the channel

between the source and drain and the current is modulated by the bias applied to the

gate.

136

step of the process used to fabricate the MOSFETs. We defined the resistive pathways

of silicon by depositing oxide (via shadow evaporation) to protect the underlying

silicon during the etch step; the silicon that remains under the oxide after etching

ultimately defines the resistive element.

We fabricated a resistor ~100 microns long and ~20 microns wide and measured a

resistivity of 64 W·cm (the listed resistivity of the wafer was 20 W·cm). The resistance

through the resistor was 300 times greater than that through the channel at zero gate

bias. This particular approach to fabrication demonstrates the ability to define

structures in a subtractive manner. The shadow deposition defines a passive element

(here, oxide) that serves as an etch mask to define structures in the underlying

materials during subsequent etch steps.

Figure 4.7 illustrates an alternate design of a resistive element that can be

fabricated on the same substrate as the MOSFET devices and ohmic connectors

described in the manuscript (Figures 2-5). An opening in the resist allows the pathway

to be illuminated during oxide deposition (i.e., when the substrate is oriented at θ =

90° and 270°) and shadowed during the metal deposition. During the deposition of the

gate oxide, these regions became covered with oxide. The oxide insulates the

underlying Si from the parallel conductive pathways (Figure 4.7 (c)) such that

electrons have to travel through the underlying Si, which is significantly more resistive

than the metal. The conductive pathways in Figure 4.7 can connect adjacent

transistors, as demonstrated by the AND device in the manuscript.

137

Figure 4.7. Top-down schematic of a void pattern that produces a resistor while

simultaneously producing a transistor. (a) The source and drain depositions are done

parallel to the X-axis and simultaneously produce conductive sections that are parallel

to the X-axis. (b) The gate oxide is deposited parallel to the Y-axis and also deposits

oxide in the “resistive section” because it is parallel to the Y-axis. Deposition of the

gate metal (and the 45 degree connector) does not result in metal connecting the two

conductive elements; thus, a resistive section (c) is produced between the conductive

sections.

138

We used a modified recipe to process SU-8 (Microchem) to improve adhesion

to the substrate and to minimize cracking of the resist. We cleaned the substrates

(typically, silicon-on-insulator wafer from Ultrasil Corporation) using an oxygen

plasma (20 sccm O2, 80 W, 5 min, Technics Plasma Cleaner).

We coated the substrate with LOR 10B by spin coating (3000 RPM, 60 s) and

baked it at 195 °C for 5 min. We then coated the substrate with SU-8 by spin coating

(2400 RPM, 30 sec). Profilometry verified the thickness to be ~ 50 microns. We baked

the resist on a hot plate by ramping the plate to 75 °C at ~450 °C/hr, baking at 75 °C

for 5 min, and then cooling to room temperature. We exposed the resist through a

transparency mask (CAS Outputcity) for 11.8 sec at 25 mW/cm2 (ABM Mask Aligner)

and baked the resist on a hot plate at 65 °C for 30 minutes using heating and cooling

ramps. We developed the features in SU-8 Developer (Microchem) for 6 min.

To undercut the LOR 10 B, we soaked the wafer in a 1:4 solution of 400K:DI

water for 6.5 min, rinsed with DI water, and dried the substrate with nitrogen. We

removed any residual polymer using an oxygen plasma (Technics, 60 W, 1 min, 100

mT, 20 sccm O2) and removed the native oxide using reactive ion etching (30 sccm

CF4, 10 sccm Ar, 70 mT, 100 W, 30 sec). We immediately placed the substrate in an

e-beam evaporation chamber. We used e-beam evaporation because it is capable of

achieving high rates of deposition (~1 nm/s) and it provides a collimated source of

material for deposition; collimation is critical for shadow evaporation.

We deposited 50 nm of Al for the source and drain, ~80 nm of oxide (SiO2 or

Al2O3) for the first gate oxide layer, and 80-300 nm of oxide for the second gate oxide

layer, and 80 nm of Al for the gate. For the AND gate, we deposited ~50 nm of Al on

139

the connectors. We removed the resist using PG Remover (Microchem) for 60 min at

60 °C, rinsed the substrate with isoproponal, and dried it with nitrogen.

To remove the Si surrounding the device, we placed the substrate in an

aqueous solution of tetramethyl ammonium hydroxide (5 wt%), Si (16.5 g/L) and

ammonium peroxydisulfate (4 g/L) for 1.5 hr at 80 °C. This formulation selectively

etches Si in the presence of SiO2 and Al (and Al2O3); the device itself therefore serves

as an etch mask. This etch step improved the performance of the devices (e.g., by

confining charge transport to the channel between the source and drain), but caused

some overetching and underetching of the Si in some of the devices. A dry etch

process would be preferable, but the wet etch process used here was sufficient for

proof of principle.

We placed the substrate in a vacuum oven at 80 °C for 12 hours to remove any

solvent or moisture on the device before wire bonding with Al. We characterized the

devices with a semiconductor parameter analyzer (Agilent 4156C).

Arrays

We typically use 1 x 1” substrates because they are easy to handle during

processing. A substrate of this size can accommodate approximately 1600 transistors.

Figure 4.8 contains images of a typical array. Figure 4.8 (a) is an optical micrograph

of the substrate after the oxide deposition. There is some apparent roughness on the

surface of the resist primarily due to the contact lithography process. Figure 4.8 (b) is

a scanning electron micrograph of an array of transistors after the lift off step. Based

140

Figure 4.8. Top-down images of an array of transistors formed by TEMIL. (a) An

optical micrograph of the substrate after the oxide deposition. The photoresist appears

rough due to the contact lithography step. (b) A scanning electron micrograph of an

array of transistors after the lift-off step.

141

on visual inspection, approximately 10-20% of these transistors had flaws (e.g.,

particles) that presumably arose from extensive processing outside of a cleanroom.

Advantages and Disadvantages of Shadow Evaporation

Alignment and Registration

Conventional transistor fabrication requires the registration of multiple

lithographic steps. Lithography is expensive and registration is technically

challenging. In contrast, TEMIL only requires one lithographic step and no

registration. Shadow evaporation does, however, require alignment of the substrate

relative to the depositing beam (q, F). In all of our experiments, we set the angular

orientation of the substrate (q) and the beam orientation relative to the substrate (F)

crudely by eye with the aid of a protractor. We designed the depositions to have a

tolerance of D F ~ 2°. This tolerance is effectively constant regardless of scaling (see

below), whereas conventional registration becomes increasingly difficult with reduced

feature size.

Scaling

An appealing characteristic of this method is the way in which the deposited

features scale with the geometry of the voids defined in the film of polymer. Simply

scaling all of the dimensions of the structured polymer film (X, Y, and Z) by the same

factor results in the proportionate scaling of the deposited features without the need to

adjust the deposition orientation (F, q). The length of the features defined by shadow

evaporation (X - Z/tan F, Figure 4.1a) depends on the height (Z) and width (X) of the

142

lithographically defined void. These two parameter typically scale proportionally in

lithography – that is, smaller features (X) are typically patterned with thinner films (Z)

because features with large aspect ratios (Z:X) are difficult to pattern by imprint or

photolithography. To some limit, a topographical design for shadow evaporation

should therefore work as well for small as for large features, provided that the ratio of

Z:X is constant.

Resolution

There is, of course, a lower limit to the size of features that are practical to

fabricate using shadow evaporation. In principle, the resolution of shadow evaporation

at small (<100 nm) length scales is limited by several factors, all of which broaden and

blur the edge of the illuminated region: (i) imperfect collimation of the evaporation

beam; (ii) uncertainty in F and q; (iii) surface migration of the depositing species on

the substrate;48 and iv) irregularities in the topography of the patterned film and the

edge of the pattern. The “spread” in F due to imperfect collimation should contribute

minimally to the resolution limitations since the mean free path in a vacuum of ~1x10-

6 torr (approximately the pressure in our experiments) is ~ 50 m,53 whereas typical

evaporators have source to sample distances less than 0.5 m). There is, however, a

small yet predictable spread in F since the source (i.e., the crucible liner) is not a

perfect point source; we accounted for this approximate one degree deviation during

the alignment of the substrate prior to deposition. Substrate vibration (tip / tilt) can

alter the orientation F, q, but these effects should also have minimal effect with

substrate holders that are designed carefully. Some studies suggest that surface

143

migration—facilitated by the momentum of the depositing material—may alter the

deposited geometry from the theoretical line-of-sight value by as much as 7 nm.54

Despite these limitations, shadows cast over topography have been used to make two

electrode junctions with sub-10-nm gaps.26,29,30 Stencil masks, which also cast

shadows, have also been used to make 10-nm features.12

Deviations from Ideality

The deposited features can deviate from ideality due to (i) the roughness of the

lithographically defined edge that defines the shadow, (ii) imperfect initial alignment

of the substrate relative to the beam, and (iii) changes in the height of the topography

during the deposition. Figure 4.8 depicts several of these sources of deviation from

ideality. In principle, the edge can have roughness in the plane of and perpendicular to

the plane of the substrate. Edge roughness perpendicular to the plane should be

minimal since films formed by spin-coating are extremely smooth. Edge roughness in

the plane of the substrate is akin to the issue of line edge roughness in

photolithography; state of the art lithographic features have line edge roughness of ~3

nm.55 A substrate holder that allows the user to accurately orient the substrate should

minimize error associated with poor alignment (in this work, we performed all

alignments by naked eye). The height (Z in Figure 4.1) of the polymeric features

increases as material deposits on top of the features; this factor will increase the length

of the shadow as a function of deposition time and will only be of significance for thin

polymer films.

144

Figure 4.8. A schematic depiction of potential sources of error during TEMIL. (i)

Imperfect collimation or uncertainty in the alignment of the beam relative to the

substrate can result in a deviation from the intended beam path. (ii) Imperfections and

irregularities in the edge of the topography can result in deviations from the desired

beam path.

145

Another consideration of shadow evaporation is that the orientation of the

beam with respect to the substrate varies across the substrate because the source is

effectively a point source (in which the emission of evaporated material from the

source follows a cosine function56), whereas the substrate is effectively a two

dimensional plane. This variation can be accounted for with proper layout of the

features.57

Film Quality

Evaporated films are generally of lower quality (e.g. lower density) than those

formed by atomic layer deposition or chemical vapor deposition; in some cases, the

quality of the film can be improved by using specialized techniques, such as heating

the substrate during deposition.58 Heating can only be performed within the limits of

stability of the organic photoresist and may deform the photoresist features due to

thermal expansion.59 Alternatively, the thin-film structures deposited by shadow

evaporation can be used as an etch mask to define structure in an underlying substrate

or film in a subtractive manner; this approach greatly increases the quality and number

of materials that can be patterned. We demonstrated this approach to remove the

excess silicon of the SOI substrate by using the transistor itself as an etch mask

(Figure 4.2, step 11).

Acknowledgements

This work was supported by NSF award PHY-064609 and CHE-0518055. We

used shared facilities supported by the MRSEC (DMR-0213805). This work was

146

performed, in part, using the facilities of the Center for Nanoscale Systems (CNS), a

member of the National Nanotechnology Infrastructure Network (NNIN), which is

supported by the National Science Foundation under NSF award ECS-0335765. CNS

is part of the Faculty of Arts and Sciences at Harvard University. We thank Prof.

Marko Lončar (Harvard University) for the SOI wafers.

References and Notes

(1) Colburn, M.; Johnson, S.; Stewart, M.; Damle, S.; Bailey, T. C.; Choi, B.; Wedlake, M.; Michaelson, T.; Sreenivasan, S. V.; Ekerdt, J.; Willson, C. G. Proceedings of SPIE-The International Society for Optical Engineering 1999, 3676, 379-389.

(2) Willson, C. G. J. Photopolym. Sci. Technol. 2009, 22, 147-153. (3) Chou, S. Y.; Krauss, P. R.; Renstrom, P. J. Science 1996, 272, 85-7. (4) Chou, S. Y. MRS Bulletin 2001, 26, 512-517.

(5) Xu, Q.; Mayers, B. T.; Lahav, M.; Vezenov, D. V.; Whitesides, G. M. J. Am. Chem. Soc. 2005, 127, 854-855.

(6) Hua, F.; Sun, Y.; Gaur, A.; Meitl, M. A.; Bilhaut, L.; Rotkina, L.;

Wang, J.; Geil, P.; Shim, M.; Rogers, J. A.; Shim, A. Nano Lett. 2004, 4, 2467-2471.

(7) Guo, L. J. J. Phys. D: Appl. Phys. 2004, 37, R123-R141.

(8) Costner, E. A.; Lin, M. W.; Jen, W. L.; Willson, C. G. Annu. Rev. Mater. Res. 2009, 39, 155-180.

(9) Schift, H. Journal of Vacuum Science & Technology B 2008, 26, 458-

480.

(10) Racz, Z.; He, J.; Srinivasan, S.; Zhao, W.; Seabaugh, A.; Han, K.; Ruchhoeft, P.; Wolfe, J. J. Vac. Sci. Technol., B: Microelectron. Nanometer Struct.--Process., Meas., Phenom. 2004, 22, 74-76.

(11) Egger, S.; Ilie, A.; Fu, Y.; Chongsathien, J.; Kang, D.-J.; Welland, M.

E. Nano Lett. 2005, 5, 15-20.

147

(12) Champagne, A. R.; Couture, A. J.; Kuemmeth, F.; Ralph, D. C. Appl. Phys. Lett. 2003, 82, 1111-1113.

(13) Luthi, R.; Schlittler, R. R.; Brugger, J.; Vettiger, P.; Welland, M. E.;

Gimzewski, J. K. Appl. Phys. Lett. 1999, 75, 1314-1316.

(14) Deshmukh, M. M.; Ralph, D. C.; Thomas, M.; Silcox, J. Appl. Phys. Lett. 1999, 75, 1631-1633.

(15) Noguchi, Y.; Sekitani, T.; Someya, T. Appl. Phys. Lett. 2007, 91,

133502/1-133502/3.

(16) Zhou, Y. X.; Johnson, A. T., Jr.; Hone, J.; Smith, W. F. Nano Lett. 2003, 3, 1371-1374.

(17) Vazquez-Mena, O.; Villanueva, G.; Savu, V.; Sidler, K.; van den

Boogaart, M. A. F.; Brugger, J. Nano Lett. 2008, 8, 3675-3682.

(18) Kohler, J.; Albrecht, M.; Musil, C. R.; Bucher, E. Physica E 1999, 4, 196-200.

(19) Dolan, G. J. Appl. Phys. Lett. 1977, 31, 337-339. (20) Jelks, E. C.; Kerber, G. L. Appl. Phys. Lett. 1981, 38, 933-935. (21) Bylander, J.; Duty, T.; Delsing, P. Nature 2005, 434, 361-364.

(22) Pashkin, Y. A.; Nakamura, Y.; Tsai, J. S. Appl. Phys. Lett. 1999, 74, 132-134.

(23) Weimann, T.; Wolf, H.; Scherer, H.; Niemeyer, J.; Krupenin, V. A.

Appl. Phys. Lett. 1997, 71, 713-715.

(24) Harada, Y.; Haviland, D. B.; Delsing, P.; Chen, C. D.; Claeson, T. Appl. Phys. Lett. 1994, 65, 636-8.

(25) Weber, R.; Kruger, U.; Martienssen, W.; Niemeyer, J. Phys. Stat. Sol.

(a) 1993, 136, K41-K45.

(26) Philipp, G.; Weimann, T.; Hinze, P.; Burghard, M.; Weis, J. Microelectron. Eng. 1999, 46, 157-160.

(27) Javey, A.; Qi, P.; Wang, Q.; Dai, H. Proc. Natl. Acad. Sci. U. S. A.

2004, 101, 13408-13410.

148

(28) Holdeman, L. B.; Barber, R. C.; Abita, J. L. Journal of Vacuum Science & Technology, B: Microelectronics and Nanometer Structures 1985, 3, 956-8.

(29) Lefebvre, J.; Radosavljevic, M.; Johnson, A. T. Appl. Phys. Lett. 2000,

76, 3828-3830. (30) De Poortere, E. P.; Stormer, H. L.; Huang, L. M.; Wind, S. J.; O'Brien,

S.; Huang, M.; Hone, J. Appl. Phys. Lett. 2006, 88, 143124/1-143124/3.

(31) Xu, T.; Metzger, R. M. Nano Lett. 2002, 2, 1061-1065. (32) Chung, J.; Lee, K.-H.; Lee, J. Nano Lett. 2003, 3, 1029-1031.

(33) Dickey, M. D.; Weiss, E. A.; Smythe, E. J.; Chiechi, R. C.; Capasso, F.; Whitesides, G. M. ACS Nano 2008, 2, 800-808.

(34) Park, Y. D.; Caballero, J. A.; Cabbibo, A.; Childress, J. R.; Hudspeth,

H. D.; Schultz, T. J.; Sharifi, J. R. J. Appl. Phys. 1997, 81, 4717-4719. (35) Chiang, J.-C.; Hwu, J.-G. Appl. Phys. Lett. 2007, 90, 102902/1-

102902/3. (36) Jones, E. T. T.; Chyan, O. M.; Wrighton, M. S. J. Am. Chem. Soc.

1987, 109, 5526-5528. (37) Kubatkin, S. E.; Danilov, A. V.; Bogdanov, A. L.; Olin, H.; Claeson, T.

Appl. Phys. Lett. 1998, 73, 3604-3606. (38) Xu, Q.; Bao, J.; Rioux, R. M.; Perez-Castillejos, R.; Capasso, F.;

Whitesides, G. M. Nano Lett. 2007, 7, 2800-2805. (39) Kosiorek, A.; Kandulski, W.; Chudzinski, P.; Kempa, K.; Giersig, M.

Nano Lett. 2004, 4, 1359-1363. (40) Sordan, R.; Burghard, M.; Kern, K. Appl. Phys. Lett. 2001, 79, 2073-

2075.

(41) Yao, S. K. J. Appl. Phys. 1979, 50, 3390-3395.

(42) Ahn, H.; Lee, K. J.; Childs, W. R.; Rogers, J. A.; Nuzzo, R. G.; Shim, A. J. Appl. Phys. 2006, 100, 084907/1-084907/7.

(43) Sze, S. M.; Ng, K. K. Physics of semiconductor devices; 3rd ed.;

Wiley-Interscience: Hoboken, N.J., 2007.

149

(44) Ito, H. Adv. Polym. Sci. 2005, 172, 37-245.

(45) Park, O.-H.; Cheng, J. Y.; Hart, M. W.; Topuria, T.; Rice, P. M.; Krupp, L. E.; Miller, R. D.; Ito, H.; Kim, H.-C. Adv. Mat. 2008, 20, 738-742.

(46) Pandy, A.; Landsberger, L. M.; Nikpour, B.; Paranjape, M.; Kahrizi,

M. J. Vac. Sci. Technol., A 1998, 16, 868-872. (47) Tabata, O.; Asahi, R.; Funabashi, H.; Shimaoka, K.; Sugiyama, S. Sens.

Actuators, A 1992, A34, 51-7. (48) Racz, Z.; Seabaugh, A. J. Vac. Sci. Technol., B: Microelectron.

Nanometer Struct.--Process., Meas., Phenom. 2007, 25, 857-861.

(49) Xia, Y.; Whitesides, G. M. Angew. Chem., Int. Ed. 1998, 37, 550-575. (50) Xu, Q.; Rioux, R. M.; Dickey, M. D.; Whitesides, G. M. Acc.

Chem. Res. 2008, 41, 1566-1577.

(51) Xu, Q.; Tonks, I.; Fuerstman, M. J.; Love, J. C.; Whitesides, G. M. Nano Lett. 2004, 4, 2509-2511.

(52) Kim Sang, O.; Solak Harun, H.; Stoykovich Mark, P.; Ferrier Nicola,

J.; De Pablo Juan, J.; Nealey Paul, F. Nature 2003, 424, 411-4. (53) The mean free path of evaporated material at vacuum pressures below

0.1 mTorr is longer than the distance between the hearth and stage of most evaporation systems.

(54) Chopra, N.; Xu, W.; De Long, L. E.; Hinds, B. J. Nanotech. 2005, 16,

133-136.

(55) International Technology Roadmap for Semiconductors. 2007, San Jose, CA: Semiconductor Industry Association.

(56) Maissel, L. I.; Glang, R., Handbook of thin film technology. 1970, New

York, McGraw-Hill.

(57) Bai, J. G.; Chang, C.-L.; Chung, J.-H.; Lee, K.-H. Nanotech. 2007, 18, 405307/1-8.

(58) Smith, D. L., Thin-film deposition: principles and practice. 1995, New

York: McGraw-Hill. (59) Feng, R.; Farris, R. J. J. Mater. Sci. 2002, 37, 4793-4799.

150

Appendix I

Laterally Ordered Bulk Heterojunction of Conjugated Polymers: Nanoskiving a

Jelly Roll

Darren J. Lipomi, Ryan C. Chiechi, William F. Reus, and George M. Whitesides

Department of Chemistry and Chemical Biology, Harvard University

12 Oxford St., Cambridge, Massachusetts, 02138 (USA)

Reproduced with permission from

Adv. Funct. Mater. 2008, 18, 3469-3477

Copyright 2008, Wiley-VCH Verlag GmbH & Co. KGaA

151

DOI: 10.1002/adfm.200800578

Laterally Ordered Bulk Heterojunction of Conjugated Polymers:Nanoskiving a Jelly Roll**

By Darren J. Lipomi, Ryan C. Chiechi, William F. Reus, and George M. Whitesides*

1. Introduction

This paper describes the fabrication of a heterojunction oftwo conjugated polymers in which laterally thin (!15 to100 nm) but vertically tall (100 to 1000 nm) phases areintimately packed and oriented perpendicularly to a substrate.The process used in the fabrication has three steps: i) spin-coating a composite film with 100 alternating layersof poly(benzimidazobenzophenanthroline ladder) (BBL,‘‘n-type’’) and poly(2-methoxy-5-(20-ethylhexyloxy)-1,4-phenylenevinylene) (MEH-PPV, ‘‘p-type’’); ii) rolling thismultilayer film into a cylinder (a ‘‘jelly roll’’); and iii) sectioningthe jelly roll with an ultramicrotome (nanoskiving,[1–6] Fig. 1).The cross-section of a slab of the jelly roll has an interdigitatedarrangement of the two polymers. The thickness of the slab is

determined by the ultramicrotome and the spacing betweenthe two materials is determined by spin-coating.

Heterojunctions with designed order have been proposedfor organic photovoltaic (OPV) devices, for which nano-structuring of the n-type and p-type phases with a spacing closeto the exciton diffusion length (5 to 20 nm) within thephotoactive layer would facilitate efficient separation ofcharges.[7] The structure described here provides an exampleof a rationally ordered heterojunction composed entirely ofconjugated polymers and arranged on the length scale thatcharacterizes exciton diffusion. We suggest that this approachto such structures could be useful in photophysical studies, andmight ultimately suggest new approaches to OPV devices.

1.1. Background

1.1.1. Conjugated Polymer Heterostructures

The tunable optical and electronic properties, mechanicalflexibility, and relatively low cost of conjugated polymers havemotivated research into their use as the active components ofmany devices traditionally associated with inorganic semicon-ductors: particular interest has focusedonpolymer light-emittingdevices,[8,9] field-effect transistors,[10] nanowires,[11] and photo-voltaic devices.[12] Conjugated polymers, however, are oftenfundamentally incompatible with traditional methods for nano-fabrication developed for inorganic semiconductors. Severalcreative techniques now exist for the fabrication of single-component structures.[13,14] This work is focused on developingroutes to structures comprising multiple components.

FULL

PAPER

[*] Prof. G. M. Whitesides, D. J. Lipomi, Dr. R. C. Chiechi, W. F. ReusDepartment of Chemistry and Chemical Biology, Harvard University12 Oxford St., Cambridge, Massachusetts, 02138 (USA)E-mail: gwhitesides@gmwgroup.harvard.edu

[**] This work was supported by the DOE under DE-FG02-00ER45852 andthe NSF under CHE-0518055. The authors used shared facilitiessupported by the NSF through the MRSEC program under awardDMR-0213805 and through the NSEC program under award PHY-0117795. The authors thank Dr. Emily A. Weiss and Dr. Michael D.Dickey for helpful discussions and Dr. Richard Schalek for training onthe ultramicrotome. W. F. R. acknowledges a training grant from NIHaward number T32 GM007598. This work was performed in part usingthe facilities of the Center for Nanoscale Systems (CNS), a member ofthe National Nanotechnology Infrastructure Network (NNIN), whichis supported by the National Science Foundation under NSFaward no.ECS-0335765. CNS is part of the Faculty of Arts and Sciences atHarvard University.

This paperdescribes the fabricationof ananostructuredheterojunctionof twoconjugatedpolymersbya three-stepprocess: i) spin-coating a multilayered film of the two polymers, ii) rolling the film into a cylinder (a ‘‘jelly roll’’) and iii) sectioning the filmperpendicular to the axis of the roll with an ultramicrotome (nanoskiving). The conjugated polymers are poly(benzimidazoben-zophenanthroline ladder) (BBL, n-type) and poly(2-methoxy-5-(20-ethylhexyloxy)-1,4-phenylenevinylene) (MEH-PPV, p-type).Theprocedureproduces sectionswithan interdigitated junctionof the twopolymers.The spacingbetween thephases isdeterminedby spin-coating (!15 nm to 100 nm) and the thickness of each section is determined by the ultramicrotome (100 to 1000 nm). Theminimumwidthof theMEH-PPVlayersaccessiblewith this technique(!15 nm)isclose toreportedexcitondiffusion lengths for thepolymer.Whenplaced ina junctionbetween twoelectrodeswithasymmetricwork functions (tin-doped indiumoxide (ITO)coatedwith poly(3,4-ethylenedioxythiophene:poly(styrenesulfonate) (PEDOT:PSS), and eutectic gallium-indium, EGaIn) the hetero-structures exhibit a photovoltaic response under white light, although the efficiency of conversion of optical to electrical energy islow. Selective excitation of BBL with red light confirms that the photovoltaic effect is the result of photoinduced charge transferbetween BBL and MEH-PPV.

Adv. Funct. Mater. 2008, 18, 3469–3477 ! 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim 3469

152

FULL

PAPER

1.1.2. Mechanism of OPV Devices

When a photon is absorbed in an organic semiconductor, thelow dielectric constant of the medium impedes the dissociationof the resulting electron-hole pair (called an exciton). Theexciton can diffuse a characteristic length – the excitondiffusion length (LD) – before it either decays, or reaches aboundary with another material. Devices that work by this so-called excitonic mechanism[15] require an interface betweenan electron donor (‘‘p-type’’) and an electron acceptor(‘‘n-type’’)[16] with offset frontier molecular orbitals (HOMO–LUMO energy levels) to enable the generation of free chargecarriers. An excited electron in the LUMO of the p-typematerial transfers to the lower LUMO of the n-type material.Conversely, a hole created in the HOMO of the n-typematerial transfers to the higher HOMO of the p-type material.These changes in free energy provide the driving forces for thedissociation of excitons into free charge carriers. Oncedissociated from each other, the electrons migrate by hoppingamong the LUMO(s) of the n-type material toward a low-work-function electrode (LWFE) and the holes migratethrough the HOMO(s) of the p-type material toward a high-work-function electrode (HWFE). Ideally, all regions in the

active layer should be situated less than one LD (typically 5 to20 nm) from an interface between phases.

A competing criterion for efficient harvesting of photons,however, requires that the active layer have sufficient thicknessto absorb the majority of incident photons (typically 100 to200 nm).[17] Most organic heterojunctions are of two generalconfigurations: the planar heterojunction and the bulkheterojunction. Planar heterojunctions consist of stacked thinfilms in the structure of HWFE/p-type/n-type/LWFE, where‘‘p-type/n-type’’ denotes a 2D interface within the photoactivelayer. In the planar configuration, only excitons created nearthe interface can contribute to the photovoltaic effect.[7] Thebulk heterojunction has the form HWFE/p-type:n-type/LWFE, where ‘‘p-type:n-type’’ indicates a disordered, co-deposited layer of materials. The photoactive layer is usually aconjugated polymer (p-type) combined with a fullerenederivative (n-type). Co-deposition of the active layer increasesthe amount of interfacial area within the photoactive layer, butalso destroys the complete continuity of each phase andprovides little control over which material is in contact withwhich electrode. Despite these shortcomings, and the fact thatthe efficiencies of these devices are extremely sensitive toprocessing conditions,[7,18,19] bulk heterojunctions can exhibitquantitative photoluminescence quenching[20] and can be fairlyefficient when incorporated into OPV devices (!5%).[21,22]

1.1.3. The Ordered Bulk Heterojunction

Recent reviews[7,17] and theoretical studies[23,24] havesuggested that the ideal heterojunction would have ananostructured network of the n-type and p-type materialspreserving the physical continuity of each material both withinthe photoactive layer and to the proper electrodes. Structuresthat meet these criteria are called ‘‘ordered bulk heterojunc-tions’’[17] (Fig. 2). The length scale of the nanostructuringshould be close toLD, in order to maximize the probability thatan exciton formed in one material would reach the interfacewith the complementary material before de-excitation. A feworganic-inorganic hybrid devices have been described in which

D. J. Lipomi et al. / ‘‘Nanoskiving’’ a Jelly Roll

Figure 1. Brief summary of the procedure used to fabricate nanostructuredheterojunctions from sectioning a jelly roll made of conjugated polymers.Spin-coating in an alternating fashion yields a composite film of theconjugated polymers, BBL (n-type) and MEH-PPV (p-type). Rolling thiscomposite film into a jelly roll increases the density of material in the cross-section. An ultramicrotome sections the jelly roll into thin slabs. The cross-section of an individual slice has a structure with an interdigitated arrange-ment of the two polymers. The ultramicrotome determines the thickness ofeach slab, while spin-coating determines the width of each material withinthe heterojunction.

Figure 2. Schematic drawingof the cross-section of an orderedbulk hetero-junctionproposed forOPVdevices. Thearchitecturehasa cross-sectionwithan interdigitatedarrangementofn-typeandp-typephases. Thewidthof eachphase should be close to the exciton diffusion length (5 to 20 nm), while thethickness of the device should allow efficient collection of photons (100 to200 nm for many conjugated polymers). This arrangement maximizes theprobability thananexcitonwill reachan interface,where it candissociate intotwo charge carriers, a hole (hþ) and an electron (e#).

3470 www.afm-journal.de ! 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2008, 18, 3469–3477

153

FULL

PAPER

a conjugated polymer combines with an inorganic electronacceptor in an ordered fashion on the nanometric scale. Forexample, Alivisatos and co-workers cast poly(3-hexylthio-phene) over vertically oriented CdTe nanocrystals,[25] andMcGehee and co-workers infiltrated this polymer intonanoporous TiO2.

[26] These processes give photoactive layerswith well-defined networks and straight, uninterrupted path-ways to the electrodes. All-organic devices, in contrast, havenot achieved the level of control attainable with organic-inorganic systems, although the use of block copolymers,[27]

polymer demixing,[28] nanoimprinting,[29] controlled organicvapor-phase deposition[30] and photoinduced mass transportusing an all-optical technique[31] have yielded interesting newheterostructures that satisfy some of the criteria required for anordered bulk heterojunction.

1.2. Experimental Design

1.2.1. Nanoskiving

‘‘Nanoskiving’’ is the name we have given to the use of anultramicrotome for creating functional nanostructures bysectioning thin films;[1–6,32] it is a form of edge lithography.[33]

We have applied nanoskiving to the fabrication of an orderedbulk heterojunction by spin-coating a composite film ofalternating layers of p-type and n-type polymers on a planarsubstrate (in which the thickness of each layer is LD! 5 to20 nm), rolling the film to increase the density of the alternatinglayers within the structure, and obtaining sections of thethickness at which light absorption is optimal (100 to200 nm).[7] This procedure would allows us, in principle, to‘‘dial in’’ the spacing between the two materials (using spin-coating) and the thickness of the heterojunction (usingnanoskiving).

1.2.2. Selection of Conjugated Polymers

The first step in the procedure was the generation of a free-standing, composite film of n-type and p-type materials. Inchoosing the two polymers, it was essential that we coulddeposit one on top of the other in a process that left theproperties of both intact. The work of Jenekhe and coworkersestablished poly(benzimidazobenzophenanthroline ladder)(BBL) and poly(2-methoxy-5-(20-ethylhexyloxy)-1,4-phenylenevinylene) (MEH-PPV) as one of the most effective n-type/p-type pairs that can be processed from orthogonal solvents tomake a planar OPV device.[34] BBL is an n-type conjugatedladder polymer that has an ionization energy (HOMO level) of5.9 eV, an electron affinity (LUMO level) of 4.0 eV, excellentthermal stability in air ("500 8C),[35] and exceptionally highfield-effect electron mobility.[36] MEH-PPV is a highlyfluorescent p-type polymer with HOMO level of 5.1 eV anda LUMO level of 2.9 eV.[34] The exciton diffusion length ofMEH-PPV has been measured using a variety of techniques inthe literature,[37] but typically falls between 5 and 14 nm.[38,39]

BBL and MEH-PPV are processed from methanesulfonic acid

and chloroform, respectively. These materials could beiteratively spin-coated on top of each other, because chloro-form neither swells nor dissolves BBL and methanesulfonicacid neither swells nor dissolves MEH-PPV.

1.2.3. Selection of Electrodes

We used two electrodes with different work functions.Tin-doped indium oxide (ITO, work function¼ 4.7–4.8[40]),spin-coated with a thin film of the hole-selective polymerblend, poly(3,4-ethylenedioxythiophene):poly(styrenesulfo-nate) (PEDOT:PSS) was the HWFE. This electrode is highlytransmissive in the visible region. PEDOT:PSS smoothes thesurface of ITO and facilitates the injection of holes into thejelly roll, but does not itself produce a photovoltaicresponse.[41] We also required an electrode with a workfunction lower than that of ITO, in order to break thesymmetry of the jelly roll and bias the photogenerated chargecarriers to drift toward the proper electrodes.[42] For theLWFE we used the liquid eutectic gallium indium (EGaIn);this material substitutes for evaporated Al, which is commonlyused.[43] EGaIn is conformal, convenient, and does not requirethe potentially damaging step of physical vapor deposition.[44]

Figure 3A shows the positions of the work functions (for ITO,PEDOT:PSS and EGaIn) and the HOMOs and LUMOs (forBBL andMEH-PPV). Figure 3B shows a schematic illustration

D. J. Lipomi et al. / ‘‘Nanoskiving’’ a Jelly Roll

Figure 3. A) Energy level diagram showing the vacuum-level positions ofwork functions (for ITO, PEDOT:PSS and EGaIn) andHOMOs and LUMOs(for BBL and MEH-PPV). B) Schematic illustration of the junction used tomeasure a photovoltaic response of a jelly roll. Under short-circuit con-ditions, a weak electric field E develops across the junction that biases thedrift of photogenerated electrons (e$) and holes (hþ) toward the EGaIn andthe ITO.

Adv. Funct. Mater. 2008, 18, 3469–3477 ! 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.afm-journal.de 3471

154

FULL

PAPER

of the experimental setup and the direction of charge carriersin the junction. The asymmetry of the electrodes creates a weakelectric field, E, which, in principle, causes electrons (e!) andholes (hþ) to drift toward the proper electrodes.

2. Fabrication

We fabricated two jelly rolls with different characteristics(Fig. 4). We made the first by spin-coating relatively thicklayers (100 nm) of the conjugated polymers: sectioning thisstructure would allow proof of principle, and be easy to image.The second jelly roll tested how thin we could spin-coat layersof the conjugated polymers. We formed 50 layers of BBL

(#100 nm), alternating with 50 layers ofMEH-PPV (#100 nm),onto glass, by successive cycles of spin-coating. The substratewas immersed in deionized water after each layer of BBL (toremove methanesulfonic acid) and dried with a stream of N2.After annealing the substrate at 125 8C under vacuum,sonication in methanol partially separated the layered filmfrom the glass. We used tweezers to place a rectangular($5% 10mm) piece of the film on a flat piece of poly(dimethylsiloxane) (PDMS), and dragged a second piece of flatPDMS over the top of the film. The film rolled into a jelly roll($5 to 10mm long and $500mm in diameter). We embeddedthe jelly roll in epoxy and sectioned it with an ultramicrotomeequippedwith a diamond knife into square slices (h¼ 150 nm, l,w$ 1mm).[3]

3. Results and Discussion

3.1. Imaging

The first (‘‘thick’’) jelly roll yielded a spiral structure whenembedded in epoxy and sectioned with the ultramicrotome(see optical image, Fig. 5A). Figure 5B is a scanning electron

D. J. Lipomi et al. / ‘‘Nanoskiving’’ a Jelly Roll

Figure 4. Summary of the procedure used to fabricate the polymer jellyroll. We spin-coated a free-standing film incorporating 50 layers of BBLalternating with 50 layers of MEH-PPV onto glass. We peeled the compo-site film from the substrate, and transferred it to a slab of poly(dimethyl-siloxane) (PDMS). We dragged a second piece of PDMS over the top of thefilm. This action rolled the film into a loose cylinder, which we subsequentlyembedded in epoxy. Sectioning of the film with the ultramicrotome yieldedindividual slices (l¼ 1mm, w¼ 1mm, h¼ 150 nm).

Figure 5. Images of the polymer jelly rolls. A) Optical (bright-field) imageof a 150-nm-thick slice of the ‘‘thick’’ jelly roll embedded in an epoxymembrane. B) Scanning electron micrograph (SEM) close-up of a regionlike the one indicated by the white box in (A). The exposed, 1-mm-thick, 100-layer film contains clearly defined, alternating layers of BBL and MEH-PPV.The average thickness of each phase is 100 nm. The inset is an atomic forcemicrograph (AFM tapping mode, phase image, range¼ 30 8) of a region ofthe exposed composite film, which exhibits sharp boundaries between thelayers. C) Optical image of the ‘‘thin’’ jelly roll. The composite film fromwhich this structure was rolled was 2.5mm thick and was composed of 50layers of BBL ($35 nm) alternating with 50 layers of MEH-PPV ($15 nm).The inset is a SEM close-up region of three 2.5-mm strands, closely packed(the region shown has $300 parallel structures across). D) AFM heightimage of a region of the exposed BBL/MEH-PPV composite film shown in(B) (range¼ 52.5 nm). The inset is an AFM phase image of the exposedcomposite film (range¼ 30 8). We measured a surface roughness (rms) of6 nm for the exposed film and 0.5 nm for the surrounding epoxy. The insetis a close-up phase image of the exposed composite film.

3472 www.afm-journal.de ! 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2008, 18, 3469–3477

155

FULL

PAPER

micrograph (SEM) that shows two strands of the exposedsurface of the BBL/MEH-PPV film. The exposed cross sectionof the film comprises 50 100-nm-thick layers of BBL and 50 ofMEH-PPV. The structure is essentially a bicontinuousheterostructure of parallel nanowires. The inset is an atomicforce micrograph (AFM, phase) of a region of the exposedpolymer film and shows the clean separation between the BBLand the MEH-PPV phases.

The second (‘‘thin’’) jelly roll rolled into a tighter structurethan the ‘‘thick’’ jelly roll (Fig. 5C). For the ‘‘thin’’ jelly roll, theaverage thickness of each layer was 25 nm. We estimated fromthe SEM that the BBL layers were closer to !35 nm and theMEH-PPV layers were !15 nm, although the accumulation ofimperfections in the composite film led to non-uniform spacingof the two polymers. Independent measurement by profilo-metry of these films spin-coated under the same conditions on aSi/SiO2 wafer gave heights of 40 nm and 20 nm for BBL andMEH-PPV.We estimate that the roughly elliptical area definedby the jelly roll in 5C contains "10% embedding epoxy.

The surface profile of the active material in the sectionwould influence the ability to contact the top and bottom of thestructure with electrodes. We obtained AFM profiles of the‘‘thin’’ jelly roll (see Fig. 5D and the inset phase image): therms roughness of the BBL/MEH-PPV filmwas 6 nm; that of theepoxy matrix was 0.5 nm.

We examined the cross section of the ‘‘thin’’ jelly roll bycutting a perpendicular cross section of a 1-mm-thick slice ofthe original structure. Figure 6 shows the interdigitatedarrangement of the BBL/MEH-PPV composite film (compareFig. 6 to Fig. 2). The image also qualitatively verifies theroughness (as seen by AFM) of the top and bottom of thecomposite film. Sectioning with the diamond knife does notappear to smear the surfaces of the polymer films.

3.2. Evidence of Photoinduced Charge Transfer withinJelly Roll by Measurement of Photovoltaic Response

We screened the heterostructures for a photovoltaicresponses by placing sections of the ‘‘thin’’ jelly roll betweena transparent electrode composed of an ITO-coated glass slidewith a thin transparent film of PEDOT:PSS (20 nm) and a drop

ofEGaIn (Fig. 7A).Weused apoly(dimethylsiloxane) (PDMS)membrane containing a circular hole to prevent EGaIn fromspilling over the jelly roll (shorting the device).[45–47]

We illuminated jelly rolls using white light (halogen source,flux !100mW cm#2).[48] Figure 7B shows a representativeplot of the current density (J) versus voltage (V). Wedetermined the open-circuit voltage (Voc, V) of the deviceby measuring the applied voltage required to bringthe current to zero. The short-circuit current density(Jsc, mA cm#2) is the current density that flows under zeroapplied voltage. We measured a Voc of 225mV and a Jsc of0.45mA cm#2. We approximated the area of the jelly roll forthe calculation of the current density by assuming its shape waselliptical and by measuring the semi-major axes with the SEM(area¼ 3.2% 10#4 cm2).[49] The area was not corrected forincluded epoxy, which made up "10% of the area of the

D. J. Lipomi et al. / ‘‘Nanoskiving’’ a Jelly Roll

Figure 6. SEM of a cross-section of a 1-mm-thick section of the jelly rollderived from the 2.5-mm film (shown in Figure 5C and D). This imageshows the orientation that the BBL/MEH-PPV layers would have in an OPVdevice. The inset is a close-up, which shows the dense packing ofBBL (lighter shades) and MEH-PPV (darker shades) within the cross-section.

Figure 7. A) Schematic drawing of the electronic setup to measure thephotovoltaic response of a jelly roll. We illuminated the junction from thebottom. B) Representative current versus voltage ( J–V) data of the PVresponse from a 150-nm-thick section of the ‘‘thin’’ jelly roll in the dark(squares) and under white light illumination from a halogen source(diamonds). C) A plot of logjJj versus V for a different junction in thedark and illuminated by a red LED with lmax¼ 660 nm.

Adv. Funct. Mater. 2008, 18, 3469–3477 ! 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.afm-journal.de 3473

156

FULL

PAPER

structure. The fill factor (FF) – the figure of merit thatcorresponds to the tendency of charge carriers to reach theelectrodes rather than recombine – was 17%.

We tested >30 devices with this or similar configurations,and about half produced photovoltaic responses. The mostcommon malfunctions were electrical shorting, probablythrough small cracks in the epoxy section or holes in a jellyroll through which the EGaIn made direct contact with theITO/PEDOT:PSS electrode. The J–V curves typically over-lapped upon repeated cycles (up to three) of applied voltage ona single junction. From junction to junction on a singlesubstrate, the values of Voc varied only slightly (!50mV) butthe values of Jsc were reproducible within an order ofmagnitude.

3.2.1. Controls

We needed to demonstrate that i) the jelly roll itselfproduced the photovoltaic response (rather than PEDOT:PSS)and ii) photoinduced charge transfer within the jelly rollcontributed to the photovoltaic response (rather than BBL orMEH-PPV acting independently). A control experiment withthe junction ITO/PEDOT:PSS/EGaIn produced no photo-voltaic response; this experiment demonstrated (i). Designinga control experiment for (ii) was necessary becauseMEH-PPValone generates a Voc in the configuration ITO/MEH-PPV/Alunder white light illumination, whereas ITO/BBL/Al[34,40] andITO/PEDOT:PSS/BBL/EGaIn do not. Illumination of ajunction containing a jelly roll with a red light-emitting diode(LED) with lmax¼ 660 nm (flux¼ 4.5mW cm#2), below theHOMO–LUMO gap of MEH-PPV, still produced a photo-voltaic response. Illuminating a junction with the configurationITO/PEDOT:PSS/MEH-PPV/EGaIn with the same LED didnot produce a photovoltaic response. (As expected, white lightdid produce a weak photovoltaic response.) The only way,therefore, for the jelly roll to have produced a photovoltaicresponse under red light was for an exciton to be created inBBL, to reach an interface with MEH-PPV, and to transfer ahole toMEH-PPV. These observations, combined with the factthat BBL quenches$80%of photoluminescence inMEH-PPVfilms%20-nm thick,[34] were consistent with the hypothesis thatphotoinduced charge transfer within the heterojunctioncontributed to or dominated the photovoltaic response ofthe jelly roll.

3.2.2. The Effect of Buffer Layers on Photovoltaic

Performance

We investigated the use of ‘‘buffer layers’’ of MEH-PPVbetween the ITO and the jelly roll, and BBL between the jellyroll and the top electrode, as a first step toward improving thephotovoltaic properties of these junctions. This experimentwould ensure that the p-type and n-type phases made contactwith only the HWFE and LWFE, respectively (compareFig. 8A to the ‘‘ordered bulk heterojunction’’ of Fig. 2). Wespin-coated a thin film of MEH-PPV (20 nm) on an ITO slide,

deposited a section of a jelly roll on the substrate, and spin-coated a layer of BBL on top of the jelly roll. Evaporation of anAu contact pad through a PDMS stencil finished the device.[50]

See Figure 8A for a schematic drawing. A typical devicedisplayed the following figures of merit (taken from Fig. 8B):Voc¼ 500mV, Jsc¼ 0.15mA cm#2, and FF¼ 27% (powerconversion efficiency &0.02%). Note that Jsc is nearly 103

times higher than in the ‘‘no-buffer-layer’’ case. We attributeour relatively low values of Jsc to non-conformal contact of thejelly roll to the substrate. We tested >200 devices using bufferlayers. The yield of devices that produced photovoltaicresponses was over 60%. This configuration gave morereproducible J–V data from device to device than the ‘‘no-buffer-layer’’ configuration. When Au was used as the topcontact, the values of Voc varied between 500 and 550mV,while the values of Jsc varied between 0.12 and 0.18mA cm#2.

4. Conclusions

This paper demonstrates nanoskiving as a technique fornanofabrication in thin-film polymer science, and suggests apotential application in organic photovoltaics. The techniqueconverts the edge of a multilayered film into a densely packed

D. J. Lipomi et al. / ‘‘Nanoskiving’’ a Jelly Roll

Figure 8. A) Schematic drawing of the electronic setup to measure thephotovoltaic response of a jelly roll with buffer layers included between thejelly roll and the electrodes such that the BBL and the MEH-PPV madeexclusive contact with the EGaIn and the ITO. B) A representative J–V plotshows that the devices produce a photovoltaic effect.

3474 www.afm-journal.de ! 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2008, 18, 3469–3477

157

FULL

PAPER

structure of macroscopic proportions (visible to the naked eye)that can be placed on almost any substrate for characterization.Further, it expands the capabilities of nanoskiving frommetallic structures for optical applications to include organiccomponents for electronic devices. The fabrication of theheterostructures is experimentally straightforward because itrequires only a spin-coater and an ultramicrotome (instru-ments to which most researchers already have access). Wewere able to execute the entire process, from spin-coating tophotovoltaic measurement, in about two days. We believe,therefore, that this technique will facilitate photovoltaic andphotophysical investigation of n-type/p-type pairs of conju-gated polymer structures arranged on the length scale ofexciton diffusion.

This technique is not restricted to BBL andMEH-PPV – any‘‘stackable’’ materials are suitable. They include otherconjugated polymer pairs,[51,52] conjugated polyelectrolytes,[53]

semiconductor nanocrystals,[47,54] physically deposited smallmolecules,[29] sol-gel precursors[55] and metal films alternatingwith metal oxides.[4] It should be possible, therefore, to formnanostructured heterojunctions of different compositions fordifferent purposes. For example, this method would enablestudies of photoluminescence quenching on pairs of materialsthat would not otherwise form an intimate heterojunction. Webelieve that the ultramicrotome is ideally suited to section thin-film electronic materials for the purposes of the characteriza-tion of materials or for the fabrication of test devices, and thatnanoskiving could play an important role in the developmentof organic nanostructures.

5. Experimental

Fabrication of the ‘‘Thick’’ Jelly Roll (Fig. 5A and B and Photo-voltaic Data in Fig. 7C): A glass slide was cut into a 2-cm square andspin-coated with BBL (obtained from Aldrich, made into a 0.5wt%solution in methanesulfonic acid (MSA) (Fluka) prepared bydissolving 370mg of polymer in 50mL of MSA) at 3 krpm with aramp rate of 1 krpm s!1 for 30 s. (MSA causes burns and should be spin-coated in a fume hood with the sash down. We generally used ahomemade high-density polyethylene liner for the basin of the spin-coater, as MSA reacts slowly with aluminum foil.) The substrate wasremoved from the spinner with tweezers and immersed in deionizedwater for 5 s to remove MSA. The BBL film was dried with an N2 gun,during which time the film changed from dark purple to light purplewith a metallic gold luster. On top of the BBL film, we spin-coatedMEH-PPV (purchased from Aldrich, avg. MW¼ 70,000–100,000,made into a 0.6wt% solution in chloroform, prepared by dissolving444mg of polymer in 50mL of chloroform) at 3 krpm with a ramp rateof 1 krpm s!1. BBL and MEH-PPV films were stacked in this mannerfifty times for 100 total layers of polymer with average thicknessof 100 nm for each layer. The composite film was 10-mm-thick, asdetermined by SEM. The filmwas annealed under vacuum at 125 8C for5min, and scored around the edges of the glass substrate with a scalpel(in a square #1mm from the edge of the glass). The substrate with thepolymer film was immersed in methanol and placed in a sonicator bathfor #20 s. This action delaminated the edges of the film from the glass.The film was then easily peeled off with tweezers, removed from themethanol and placed on a flat piece of poly(dimethylsiloxane) (PDMS)(Dow Coring Sylgard 184 kit, mixing cross-linker and prepolymer in a

ratio of 1:10). The multilayered film was cut into 1-cm squares with arazor blade. A 1-cm square of the film was lubricated with a few dropsof ethanol. A second piece of PDMS was dragged over the top of thefilm about 5 times in the same direction such that the film rolled into acylinder. The ‘‘jelly roll’’ was embedded in epoxy prepolymer (Epo-Fix, obtained from Electron Microscopy Sciences, mixed and degassedbefore use), pressed with a wooden applicator to remove air bubbles,and cured at 60 8C for 2 h in a polyethylene mold (ElectronMicroscopySciences). The cooled block was cut with a hand saw to expose the crosssection of the jelly roll. The block was trimmed and sectioned with theultramicrotome (Leica Ultracut UCT, equipped with a diamond knifeDiatome Ultra 358) as described previously [3].

Fabrication of the ‘‘Thin’’ Jelly Roll (Figs. 5C, D, and 6; and Photo-voltaic Data in Fig. 7B): This jelly roll was fabricated the same wayas the first jelly roll, with the following modifications. 1) The first layerof BBL was spin-coated as before, successive layers were spin-coatedfrom a 0.25wt% solution at 3 krpm with a ramp rate of 1 krpm for 10 s(the BBL film directly touching the glass substrate had to be thick,otherwise it cracked during spinning). 2) MEH-PPV was spin-coatedfrom a 0.12wt% solution at 3 krpm with a ramp rate of 1 krpm for 10 s.The total thickness of the 100-layer film was 2.5mm (as measured bySEM). The average thickness of each layer was 25 nm. Althoughindividual layers were cast without major defects, the accumulation ofminor imperfections in the composite filmmade the individual layers ofBBL and MEH-PPV somewhat inhomogeneous, so it was difficult tomeasure accurate thicknesses of each layer. We estimate that thethickness of each BBL layer was approximately 35 nm and of eachMEH-PPV layer was 15 nm. Profilometry (Veeco Dektak 6M StylusProfilometer) of these films spin-coated under the same conditions on aSi/SiO2 wafer gave heights of 40 nm and 20 nm for BBL and MEH-PPV.

Imaging: Optical images (Fig. 5A, C) were obtained using anoptical microscope in bright field (Leica DMRX). Scanning electronmicroscope (SEM) images (Fig. 5B inset, Fig. 5C and inset, and Fig. 6)of the epoxy sections containing slices of the jelly roll were acquiredwith a LEO 982, Zeiss Ultra55, or Supra55 VP FESEM at 2 or 5 kV at aworking distance of 2–6mm. Before SEM imaging, some epoxysections were placed on a silicon wafer and sputter coated with Pt/Pd at60mA for 15–45 s. Atomic force microscope (AFM) height (Fig. 5D)and phase (Fig. 5B inset, Fig. 5D inset) were obtained with a VeecoDimension 3100 instrument using tapping mode.

Photovoltaic Measurements (Fig. 7): The ‘‘thin’’ jelly roll was usedto obtain the photovoltaic data of Figure 7B. The ‘‘thick’’ jelly roll wasused for Figure 7C. An ITO/SiO2 slide (Delta Technologies, Ltd.,0.7mm SiO2, Rs¼ 4–8V) was cut into a 2.5-cm square, washed withethanol or acetone, and treated with oxygen plasma (1min) prior touse. The slide was spin-coated with PEDOT:PSS (supplied by Aldrichas a 1.3wt% dispersion in water, diluted by us 1:1 with deionizedwater) at 3 krpm with a ramp rate of 1 krpm s!1 for 60 s. ThePEDOT:PSS was annealed at 125 8C in a vacuum oven for 15min. Thesubstrate was treated with oxygen plasma for 10 s in order to increasethe wettability of the substrate. This action facilitates the transfer ofepoxy sections from the water boat of the ultramicrotome to thesubstrate. Epoxy sections floating in the water bath of the diamondknife of the ultramicrotome were transferred with the Perfect Looptool (Electron Microscopy Sciences) to the surface of the MEH-PPV-coated substrate. The substrate was placed in a vacuum desiccator untilthe water evaporated, leaving the epoxy sections adhered flatly to thesubstrate by way of capillary forces. Sections (5–10 per substrate)were then annealed in a vacuum oven at 125 8C for 15 to 60min inorder to remove wrinkles in the epoxy sections. We obtained aPDMS membrane patterned with circles (r¼ 0.5mm) by a proceduredescribed previously [45]. Pieces of themembranewere placed over theepoxy sections, such that the jelly rolls were exposed through thecircular holes. Drops of EGaIn (Aldrich) were placed with a syringe ontop of the exposed jelly rolls. The PDMS membrane prevented theEGaIn from spilling onto the substrate. Copper wires were placedin each drop of EGaIn and secured to the substrate with drops of

D. J. Lipomi et al. / ‘‘Nanoskiving’’ a Jelly Roll

Adv. Funct. Mater. 2008, 18, 3469–3477 ! 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.afm-journal.de 3475

158

FULL

PAPER

5-Minute Epoxy (Devcon). Devices were screened indoors forphotovoltaic effect using a Keithley 6430 source meter and a halogenlamp with a flux of 100mW cm!2 as estimated (for Fig. 7B) using aDaystar Solar Meter and (for Fig. 7C) with a red LED withlmax¼ 660nm and flux¼ 4.5mW cm!2 as determined using an opticalpower meter (ThorLabs DET 110). The ITO was the anode underpositive bias, while the EGaIn was grounded.

Photovoltaic Measurements with ‘‘Buffer Layers’’ (Fig. 8): Thesecond configuration differs from the first configuration in three ways.1) Instead of PEDOT:PSS, MEH-PPV coated the ITO/SiO2 substrate.The substrate was spin-coated with MEH-PPV (0.12wt% solution inchloroform) at 3 krpm with a ramp rate of 1 krpm s!1 and treated withoxygen plasma for 1 s in order to assist transfer of the epoxy sections tothe substrate. 2) After placement of the epoxy sections, the substratewas spin-coated with BBL (0.5wt% in MSA, at a spin rate of 6 krpmwith a ramp of 1 krpm s!1 for 30 s). The substrate was immersed indeionized water for at least 5 h, blown dry with a stream of N2, andannealed under vacuum at 125 8C for 15min. 3) After placing thePDMS membranes over the jelly rolls, the exposed regions of thesubstrate (except for the jelly rolls) were covered with sticky tape.The substrate was sputter-coated with Au (#100nm). The PDMSmembranes and the sticky tape were removed and placed fresh PDMSmembranes over the jelly rolls, covered with a circular thin film of Au(#100 nm). The ends of several copper wires were dipped in graphiteink and placed in contact with the evaporated Au contacts. Thegraphite ink was allowed to dry overnight and the wires were secured tothe substrate with drops of 5-minute epoxy. Photovoltaic measure-ments were carried out as in the first configuration.

Received: April 28, 2008Published online: October 7, 2008

[1] Q. Xu, R. M. Rioux, G. M. Whitesides, ACS Nano 2007, 1, 215.[2] Q. B. Xu, J. M. Bao, F. Capasso, G. M. Whitesides,Angew. Chem. Int.

Ed. 2006, 45, 3631.[3] Q. B. Xu, J. M. Bao, R. M. Rioux, R. Perez-Castillejos, F. Capasso, G.

M. Whitesides, Nano Lett. 2007, 7, 2800.[4] Q. B. Xu, B. D. Gates, G. M. Whitesides, J. Am. Chem. Soc. 2004, 126,

1332.[5] Q. B. Xu, R. Perez-Castillejos, Z. F. Li, G. M. Whitesides, Nano Lett.

2006, 6, 2163.[6] D. J. Lipomi, R. C. Chiechi, M. D. Dickey, G. M. Whitesides, Nano

Lett. 2008, 8, 4100–4105.[7] S. Gunes, H. Neugebauer, N. S. Sariciftci, Chem. Rev. 2007, 107, 1324.[8] J. H. Burroughes, D. D. C. Bradley, A. R. Brown, R. N. Marks, K.

Mackay, R. H. Friend, P. L. Burns, A. B. Holmes, Nature 1990, 347,539.

[9] I. D. W. Samuel, G. A. Turnbull, Chem. Rev. 2007, 107, 1272.[10] H. Sirringhaus, P. J. Brown, R. H. Friend, M. M. Nielsen, K. Bech-

gaard, B. M.W. Langeveld-Voss, A. J. H. Spiering, R. A. J. Janssen, E.W. Meijer, P. Herwig, D. M. de Leeuw, Nature 1999, 401, 685.

[11] A. K. Wanekaya, W. Chen, N. V. Myung, A. Mulchandani, Electro-analysis 2006, 18, 533.

[12] B. C. Thompson, J. M. J. Frechet, Angew. Chem. Int. Ed. 2008, 47, 58.[13] S. Holdcroft, Adv. Mater. 2001, 13, 1753.[14] E. Menard, M. A.Meitl, Y. G. Sun, J. U. Park, D. J. L. Shir, Y. S. Nam,

S. Jeon, J. A. Rogers, Chem. Rev. 2007, 107, 1117.[15] B. A. Gregg, J. Phys. Chem. B 2003, 107, 4688.[16] We adopt the terms ‘‘p-type’’ or ‘‘n-type’’ to mean that the materials

are principally used to transport holes or electrons, using typicalelectrode materials. For a discussion of the factors that influencecharge transport in organic semiconductor, see: V. Coropceanu, J.

Cornil, D. A. da Silva Filho, Y. Olivier, R. Silbey, J.-L. Bredas, Chem.

Rev. 2007, 107, 926.[17] K. M. Coakley, M. D. McGehee, Chem. Mater. 2004, 16, 4533.[18] F. Padinger, R. S. Rittberger, N. S. Sariciftci, Adv. Funct. Mater. 2003,

13, 85.[19] K. Sivula, C. K. Luscombe, B. C. Thompson, J. M. J. Frechet, J. Am.

Chem. Soc. 2006, 128, 13988.[20] N. S. Sariciftci, L. Smilowitz, A. J. Heeger, F. Wudl, Science 1992, 258,

1474.[21] W. L. Ma, C. Y. Yang, X. Gong, K. Lee, A. J. Heeger, Adv. Funct.

Mater. 2005, 15, 1617.[22] J. Peet, J. Y. Kim, N. E. Coates, W. L. Ma, D. Moses, A. J. Heeger, G.

C. Bazan, Nat. Mater. 2007, 6, 497.[23] B. Kannan, K. Castelino, A. Majumdar, Nano Lett. 2003, 3, 1729.[24] L. G. Yang, L. Chen, R. Bai, F. Yang, M. Wang, H. Z. Chen, Sol.

Energy Mater. Sol. Cells 2007, 91, 1110.[25] I. Gur, N. A. Fromer, A. P. Alivisatos, J. Phys. Chem. B 2006, 110,

25543.[26] C. Goh, K. M. Coakley, M. D. McGehee, Nano Lett. 2005, 5, 1545.[27] U. Stalmach, B. de Boer, C. Videlot, P. F. van Hutten, G. Hadziioan-

nou, J. Am. Chem. Soc. 2000, 122, 5464.[28] F. A. Castro, H. Benmansour, C. F. O. Graeff, F. Nuesch, E. Tutis, R.

Hany, Chem. Mater. 2006, 18, 5504.[29] D. M. Nanditha, M. Dissanayake, A. Adikaari, R. J. Curry, R. A.

Hatton, S. R. P. Silva, Appl. Phys. Lett. 2007, 90.[30] F. Yang, M. Shtein, S. R. Forrest, Nat. Mater. 2005, 4, 37.[31] C. Cocoyer, L. Rocha, L. Sicot, B. Geffroy, R. de Bettignies, C.

Sentein, C. Fiorini-Debuisschert, P. Raimond, Appl. Phys. Lett.

2006, 88.[32] Q. B. Xu, R. M. Rioux, M. D. Dickey, G. M. Whitesides, Acc. Chem.

Res. in press.[33] B. D. Gates, Q. B. Xu, M. Stewart, D. Ryan, C. G. Willson, G. M.

Whitesides, Chem. Rev. 2005, 105, 1171.[34] M. M. Alam, S. A. Jenekhe, Chem. Mater. 2004, 16, 4647.[35] S. Y. Hong, M. Kertesz, Y. S. Lee, O. K. Kim, Macromolecules 1992,

25, 5424.[36] A. Babel, S. A. Jenekhe, J. Am. Chem. Soc. 2003, 125, 13656.[37] Scully and McGehee noted that exciton diffusion lengths are often

overestimated due to effects of optical interference and energy trans-fer. Theymeasured a diffusion length of (6$ 1) nm forMDMO-PPV, astructural relative of MEH-PPV. See: S. R. Scully, M. D. McGehee, J.Appl. Phys. 2006, 100, 034907.

[38] A. J. Lewis, A. Ruseckas, O. P. M. Gaudin, G. R. Webster, P. L. Burn,I. D. W. Samuel, Org. Electron. 2006, 7, 452.

[39] D. E. Markov, C. Tanase, P. W. M. Blom, J. Wildeman, Phys. Rev. B2005, 72.

[40] S. A. Jenekhe, S. J. Yi, Appl. Phys. Lett. 2000, 77, 2635.[41] F. Zhang, O. Inganas, in Organic Photovoltaics: Mechanisms,

Materials, and Devices (Eds: S. S. Sun, N. S. Sariciftci), CRC Press,Boca Raton, FL 2005, p. 479.

[42] The use of electrodes with asymmetric workfunctions is also necessaryin bulk heterojunction devices, where the photoactive layer is adisordered network with both types of charge carriers distributedrandomly throughout the device.

[43] A. Du Pasquier, S. Miller, M. Chhowalla, Sol. Energy Mater. Sol. Cells

2006, 90, 1828.[44] M. D.Dickey, R. C. Chiechi, R. J. Larsen, E. A.Weiss, D. A.Weitz, G.

M. Whitesides, Adv. Funct. Mater. 2008, 18, 1.[45] R. J. Jackman, D. C. Duffy, O. Cherniavskaya, G. M. Whitesides,

Langmuir 1999, 15, 2973.[46] R. C. Chiechi, E. A. Weiss, M. D. Dickey, G. M. Whitesides, Angew.

Chem. Int. Ed. 2008, 47, 142.[47] E. A. Weiss, R. C. Chiechi, S. M. Geyer, V. J. Porter, D. C. Bell, M. G.

Bawendi, G. M. Whitesides, J. Am. Chem. Soc. 2008, 130, 74.

D. J. Lipomi et al. / ‘‘Nanoskiving’’ a Jelly Roll

3476 www.afm-journal.de ! 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Funct. Mater. 2008, 18, 3469–3477

159

FULL

PAPER

[48] We estimated the flux of the halogen source with a commercial powermeter calibrated for the solar spectrum.

[49] This method includes some of the epoxy membrane, which doesnot contribute to the photocurrent. Using only the area of the expos-ed BBL/MEH-PPV film would yield higher Jsc values (we estimate by< 10%), but would not be as relevant from an engineering standpoint,where density of photoactive material within a section is critical.

[50] The asymmetry within the active layer imposed by the buffer layersallowed us to use Au as a top electrode, even though it has a higherwork function than EGaIn. See ref. [15].

[51] M. Onoda, K. Tada, A. A. Zakhidov, K. Yoshino, Thin Solid Films

1998, 331, 76.[52] S. R. Scully, P. B. Armstrong, C. Edder, J. M. J. Frechet, M. D.

McGehee, Adv. Mater. 2007, 19, 2961.[53] A. Garcia, R. Yang, Y. Jin, B. Walker, T. Q. Nguyen,Appl. Phys. Lett.

2007, 91.[54] I. Gur, N. A. Fromer, M. L. Geier, A. P. Alivisatos, Science 2005, 310,

462.[55] J. Y. Kim, S. H. Kim, H. H. Lee, K. Lee, W. L. Ma, X. Gong, A. J.

Heeger, Adv. Mater. 2006, 18, 572.

D. J. Lipomi et al. / ‘‘Nanoskiving’’ a Jelly Roll

Adv. Funct. Mater. 2008, 18, 3469–3477 ! 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.afm-journal.de 3477

160

Appendix II

Fabrication of Conjugated Polymer Nanowires by Edge Lithography

Darren J. Lipomi, Ryan C. Chiechi, Michael D. Dickey, and George M. Whitesides

Department of Chemistry and Chemical Biology, Harvard University

12 Oxford St., Cambridge, Massachusetts, 02138 (USA)

Reproduced with permission from

Nano Lett. 2008, 8, 2100-2105

Copyright 2008, American Chemical Society

161

Fabrication of Conjugated PolymerNanowires by Edge LithographyDarren J. Lipomi, Ryan C. Chiechi, Michael D. Dickey, andGeorge M. Whitesides*

Department of Chemistry and Chemical Biology, HarVard UniVersity, 12 OxfordStreet, Cambridge, Massachusetts 02138

Received April 2, 2008; Revised Manuscript Received May 8, 2008

ABSTRACT

This paper describes the fabrication of conjugated polymer nanowires by a three stage process: (i) spin-coating a composite film comprisingalternating layers of a conjugated polymer and a sacrificial material, (ii) embedding the film in an epoxy matrix and sectioning it with anultramicrotome (nanoskiving), and (iii) etching the sacrificial material to reveal nanowires of the conjugated polymer. A free-standing, 100-layer film of two conjugated polymers was spin-coated from orthogonal solvents: poly(2-methoxy-5-(2!-ethylhexyloxy)-1,4-phenylenevinylene)(MEH-PPV) from chloroform and poly(benzimidazobenzophenanthroline ladder) (BBL) from methanesulfonic acid. After sectioning the multilayerfilm, dissolution of the BBL with methanesulfonic acid yielded uniaxially aligned MEH-PPV nanowires with rectangular cross sections, andetching MEH-PPV with an oxygen plasma yielded BBL nanowires. The conductivity of MEH-PPV nanowires changed rapidly and reversibly by>103 upon exposure to I2 vapor. The result suggests that this technique could be used to fabricate high-surface-area structures of conductingorganic nanowires for possible applications in sensing and in other fields where a high surface area in a small volume is desirable.

We have developed a method for the fabrication of electri-callyconductivenanowiresoftwoconjugatedpolymersspoly(2-methoxy-5-(2!-ethylhexyloxy)-1,4-phenylenevinylene) (MEH-PPV) and poly(benzimidazobenzophenanthroline ladder)(BBL)sby sectioning spin-coated multilayer films with anultramicrotome (“nanoskiving”, see Figure 1 for chemicalstructures).1–5 We measured conductivity through groups of50 to several hundred of these nanowires, which can havecross sectional widths and heights of ∼100 nm, overdistances as long as 100 µm. Polymer nanowires are versatilestructures that are sensitive chemical6,7 and biological8

sensors, field-effect transistors,9–11 interconnects in electroniccircuitry,12 and tools for studying one-dimensional chargetransport in materials.13 Of the methods available for thefabrication of conjugated polymer nanowires, most arespecific to certain combinations of materials and substrates,and many require expensive or specialized equipment.9 Withthe goal of developing inexpensive and simple methodologyfor nanofabrication, we have developed a technique for thefabrication of conjugated polymer nanowires that requiresonly a spin-coater and an ultramicrotome; the only require-ment of the conjugated polymers is that they form thin films.After sectioning, the nanowires remain embedded in amacroscopic slab of embedding epoxy, which can bepositioned manually on a desired substrate.

Amorphous organic semiconductors such as conjugatedpolymers possess many of the useful electrical properties oftraditional crystalline semiconductors (most importantly,electroluminescence,14 photovoltaic response,15 and modula-tion of conductivity by gate voltage16 or by doping17).Polymers are more mechanically flexible and less expensiveto produce and process than crystalline semiconductormaterials. As a result, there is a vast literature devoted tothe fabrication and patterning of functional conjugatedpolymer structures.18 One of the simplest and most usefulsmall-dimensional geometries is the nanowire.9,19

Nanowires composed of conjugated polymers are well-suited for chemical sensing because they have a high ratioof area to volume; this feature permits rapid diffusion of ananalyte into and out of a wire (or adsorption/desorption fromits surface).6,20 These characteristics allow electrical responseand recovery rates that are superior to those of devices basedon thin films or fibrous networks. Incorporation of molecularrecognition elements into conjugated polymer nanowires isrelatively straightforward by synthesis; analogous modifica-tions of carbon nanotubes and inorganic nanowires require

* Corresponding author. Telephone: (617) 495-9430. Fax: (617) 495-9857. E-Mail: gwhitesides@gmwgroup.harvard.edu.

Figure 1. Chemical structures of MEH-PPV and BBL.

NANOLETTERS

2008Vol. 8, No. 72100-2105

10.1021/nl8009318 CCC: $40.75 " 2008 American Chemical SocietyPublished on Web 06/03/2008

162

surface reaction(s) carried out postfabrication.8 Other possibleuses for conjugated polymer nanowires are as tools forstudying one-dimensional charge transport21 or as field-effecttransistors,11 actuators,22 or interconnects.12

Despite the growing interest in conjugated polymernanowires, there is not yet a truly general technique for thefabrication of these structures. Two of the most versatiletechniques for the fabrication of conjugated polymer nanow-ires are templated electrodeposition and electrospinning.Penner and co-workers have developed a technique ofelectrodeposition in a template that can produce electricallyaddressable nanowires confined to trenches,23 but it requireselectron-beam lithography, is limited to materials thatpolymerize in situ, and is only compatible with rigidsubstrates, to which the nanowires are permanently attached.Craighead and co-workers have used scanned electrospin-ning24 to deposit single nanowires of polyaniline6 and poly(3-hexylthiophene)11 on a rotating substrate, while Xia and co-workers have developed an approach to deposit uniaxialcollections of nanofibers of a range of inorganic and organicmaterials.25,26 Processes involving two or more techniquesin combination have also been reported: Chi et al. havepatterned high-density arrays of polypyrrole and polyanilineby a process comprising electron-beam and nanoimprintlithographies.27 Lastly, dip-pen nanolithography, a versatiletool created by the Mirkin Laboratory, has been shown toform nanowires of charged conjugated polymers.28

Nanoskiving. We have developed a technique, nanosk-iving, for fabricating nanostructures by sectioning patternedor stacked thin films of inorganic materials with an ultra-microtome.4 It is a form of edge lithography: the ultrami-crotome exposes the cross section of an embedded thin filmto form the lateral dimension of a nanostructure.29 Thetechnique is amenable to conjugated polymers because mostof these materials (i) form thin films (by spin-coating), (ii)are tough enough to be sectioned by a diamond knife withoutfracture, and (iii) adhere to the embedding matrix (usuallyepoxy). The nanostructures, after sectioning, remain embed-ded in a thin slab of the epoxy matrix. The slab is a flexiblemacroscopic object (mm2 range) that one can manipulate andposition/orient on a substrate containing, for example,patterned electrodes. The combination of forming thin filmsand nanoskiving can create structures with high aspect ratiosand cross-sectional dimensions of 100 nm or less.

Choice of Polymers. The goal of this work was to formnanowires of conjugated polymers by sectioning spin-coatedfilms with an ultramicrotome. We reasoned that we couldfabricate multiple nanowires in this way by spin-coating acomposite film of two or more polymers, in which everyother layer would be a sacrificial material that we couldremove from the final structure (see Figure 2 for anoverview).

We explored two polymers, poly(2-methoxy-5-(2!-ethyl-hexyloxy)-1,4-phenylenevinylene) (MEH-PPV) and poly-(benzimidazobenzophenanthroline ladder) (BBL); these poly-mers have different properties and are processed in differentways. MEH-PPV is a “p-type”, solution-processible conju-gated polymer that is soluble in organic solvents (including

chloroform, tetrahydrofuran, and toluene). The conductivityof MEH-PPV reversibly increases several orders of magni-tude upon exposure to oxidizing agents, such as I2 vapor.30

This feature makes it a model for monitoring the electricalresponse of nanowires to a chemical stimulus. It is preparedchemically by an anionic polymerization and thus is notamenable to templated electrodeposition.

BBL is a ribbon-like ladder polymer and a rare exampleof a conjugated polymer that exhibits high electron mobility(“n-type”).31 It has high tensile strength and is stable in airat high T (g500 °C).32 Further, it is regarded as one of themost promising candidates for n-channel performance infield-effect transistors31 and photovoltaic cells.33 Its exclusivesolubility in neat methanesulfonic acid (MSA) or highlyLewis-acidic solutions requires exhaustive aqueous rinsingof the nonvolatile MSA or decomplexation of Lewis acidsfrom films during processing, and thus it is unlikely that anyexisting techniques are capable of forming nanowires ofBBL.34 We reasoned that, by using BBL as the sacrificialmaterial for MEH-PPV and MEH-PPV as the sacrificialmaterial for BBL, we could fabricate nanowires of eitherpolymer from the same precursor film (because MEH-PPV

Figure 2. Summary of the procedure used for fabrication of multiplenanowires of MEH-PPV or BBL.

Nano Lett., Vol. 8, No. 7, 2008 2101

163

is unaffected by MSA, and BBL is unaffected by commonorganic solvents and remarkably resistant to plasma etching).There are, additionally, other important potential uses ofbicontinuous, densely packed nanostructures of p-type/n-typeconjugated polymers, in such devices as ambipolar field-effect transistors35 and photovoltaic cells.36 In principle, thereare many possible sacrificial materials (including photoresistor metal films) depending on the sensitivity of the conjugatedpolymer to the conditions required to remove the sacrificialmaterial.

Fabrication. Figure 2 summarizes the procedure we usedto fabricate nanowires of conjugated polymers. Spin-coatingalternating layers of MEH-PPV and BBL onto glass gave acomposite film (of up to 100 total layers), which weimmersed in methanol, sonicated for ∼5 s and gently peeledoff the substrate (step 1). (Each layer of BBL had to beimmersed in water for 30 s to remove MSA and dried witha stream of N2 before deposition of the next layer of MEH-PPV.) Then, we either (i) cut <1 mm wide, ∼5 mm longstrips from the film with scissors or (ii) folded the film inhalves ∼5 times in order to fit the entire film in a mold forembedding (step 2). A thermally curable, epoxy prepolymer(Epo-Fix, Electron Microscopy Sciences) served to embedindividual strips of the film. We sectioned the cured epoxyblock with an ultramicrotome (step 3), which yielded slicesof the conjugated polymer film, framed by an epoxy matrix.Using a metal loop designed to suspend thin sections in afilm of water by exploiting surface tension, we manuallytransferred the sections from the water boat of the ultrami-crotome to photolithographically patterned Au electrodes ona SiO2 substrate (step 4). We controlled the position as wellas the orientation of the section on the substrate using largesections that stuck reversibly to the perimeter of the metalloop, such that the section was not allowed to rotate duringthe transfer. Thermal annealing (at 125 °C under vacuum,step 5) improved the physical contact between the thinsections and the substrate. Selective etching of BBL withMSA (step 6a) or MEH-PPV (and epoxy) with oxygenplasma (step 6b) gave MEH-PPV or BBL nanowires,respectively.

Concern that the oxygen plasma would destroy theelectronicpropertiesofBBLledus toobtainultraviolet-visibleabsorption spectra of two films: (i) as-cast and (ii) afteretching (Figure 3). The as-cast film was 30 nm thick and

displayed an absorption spectrum consistent with previouslypublished data.33 We exposed the second film to plasma for5 min at 1 torr and 100 W and measured a final thickness of15 nm. The absorption intensity of the etched film decreasedby one half but exhibited maxima at the same wavelengthsas the as-cast film. These results suggested that the electronicstructure of the bulk of BBL was unchanged by the etchingstep and that the oxidation of the polymer occurred only atthe surface.

Characterization of Polymer Nanowires. Figure 4Ashows the transition between the composite MEH-PPV/BBLfilm and the free MEH-PPV nanowires. We obtained theimage by covering a portion of the epoxy section with aconformal slab of poly(dimethylsiloxane) (PDMS) andtreating the uncovered portion with a drop of MSA (as shownschematically on the left-hand side of Figure 4A) for ∼5 s.We rinsed the MSA off the substrate with ethanol andremoved the slab of PDMS.

Simple dissolution of MEH-PPV by its processing solvent,chloroform, did not completely remove MEH-PPV to revealfree BBL nanowires. We observed that the MEH-PPVbecomes partially insoluble after spin-coating, sectioning andthermally annealing. (To use selective wet etching, a differentsacrificial material would have to take the place of MEH-PPV.) Instead of wet etching, however, we chose to exploitthe relative rates of dry etching by an oxygen plasma ofMEH-PPV, epoxy, and BBL. We etched thin films of MEH-PPV, epoxy, and BBL in a 100 W etcher at 1 torr of ambientair and found relative rates of etching of 12:9:1 (heights

Figure 3. Absorption spectra of two BBL films: as-cast (30 nm)and after etching with oxygen plasma (15 nm). The absorptioncharacteristics of the etched film changed in intensity only(decreased by one half), while it retained the features of the as-cast film.

Figure 4. (A) Scanning electron micrograph (SEM) showing thetransition between the composite MEH-PPV/BBL film and the freeMEH-PPV nanowires (NWs). The image was obtained by coveringa portion of the epoxy section with a conformal slab of poly(dim-ethylsiloxane) (PDMS) and treating the uncovered portion with adrop of methanesulfonic acid (MSA), as shown schematically onthe left-hand side. The slab of PDMS was removed before acquiringthe images. The solvent front is the borderline between the etchedand the intact polymer film. The fibers of BBL connecting theMEH-PPV nanowires disappear after a few successive rinsings infresh MSA. (B) Transition between the intact composite film (right-hand side) and the free BBL nanowires after dry etching of theMEH-PPV and epoxy matrix (left-hand side).

2102 Nano Lett., Vol. 8, No. 7, 2008

164

measured by profilometry). Figure 4B shows the transitionbetween the intact composite film (right-hand side) and thefree BBL nanowires after dry etching of the MEH-PPV andepoxy matrix (left-hand side).

Nanowires Bridging Electrodes. We obtained sectionslike the one shown in the optical image of Figure 5A. Theepoxy sections contained a ∼1.5 cm long, 100 layer film ofMEH-PPV and BBL and were 150 or 200 nm thick. Wefolded the film in order to ensure that it would span a set ofparallel electrodes no matter what orientation we depositedthe epoxy slab.37 Figure 5B shows an SEM image of tworegions of the folded composite film over Au electrodes (thegap is 50 µm). The epoxy section that contains the compositefilm is “transparent” because we obtained the image at arelatively high accelerating voltage (5 kV). The inset showsa global view of the section over electrodes; the white boxindicates the region shown in the main image.

Figure 5C shows uniaxially aligned MEH-PPV nanowires,after wet etching of BBL, spanning electrodes. This imagewas acquired at 2 kV, so the surrounding epoxy is “opaque”and clearly visible. The insets are of a group of nanowires(top right) and an isolated nanowire (bottom left). The edgesof each individual nanowire are sharp, and the nanowireshave the expected rectangular cross sections. The MEH-PPVnanowires have a tendency to aggregate, due to capillary

Figure 5. (A) Optical micrograph showing a 150 nm thick epoxysection containing a 1.5 cm long, folded, 100 layer composite filmof MEH-PPV and BBL. (B) SEM of two regions of the compositefilm bridging Au electrodes on a SiO2/Si substrate. The inset is aglobal view of the epoxy slab placed over the electrodes. The whitebox indicates the area shown in the main image. (C) SEM of 50MEH-PPV nanowires spanning electrodes after etching of BBL withMSA. The insets (top, right) show a close-up of the nanowires and(bottom, left) an isolated nanowire. The wires have well-definedcorners and rectangular cross sections. (D) SEM of 50 BBLnanowires spanning a gap between electrodes.

Figure 6. (A) Current density vs applied voltage (J-V) plot of300 MEH-PPV nanowires in the proximity of an I2 crystal. (B)The response of the same set of nanowires upon removal of the I2crystal. (C) Current vs applied voltage (I-V) of 400 BBL nanowirescompared with that of the SiO2 substrate.

Nano Lett., Vol. 8, No. 7, 2008 2103

165

forces, while the rinsing solvent, ethanol, evaporates. Figure5D shows BBL nanowires spanning a 25 µm gap betweenelectrodes after dry etching the MEH-PPV layers and theepoxy matrix.

Application to Sensing by Reversible Doping. I2 revers-ibly increases the conductivity of many conjugated polymersby several orders of magnitude.28 We measured the electricalresponse of a group of 300 MEH-PPV nanowires (crosssection: 200 nm × 100 nm, each) to I2 vapor by placing acrystal of solid I2 about 2 mm from the nanowires.38 Threesuccessive cycles of applied voltage (0 V f 1 V f -1 Vf 0 V) yielded overlapping plots and no hysteresis (Figure6A). Upon removal of the I2, we observed a decrease incurrent density by a factor of 103 in the time it took to removethe I2 and obtain another measurement (∼10 s; see Figure6B). The conductivity continued to decrease over fiveconsecutive cycles of applied voltage (labeled 1 through 5in Figure 6B; the experiment lasted 135 s total). We attributethe loss of conductivity to the loss of I2 from the MEH-PPVnanowires and a concomitant decrease in the density ofcharge carriers in the material. When we replaced the I2 nearthe wires, the MEH-PPV nanowires recovered the originalconductivity of Figure 6A. From the current density versusapplied voltage (J-V) data for this representative group ofnanowires, we calculated an approximate conductivity of 4S cm-1 (approximate because the orientation of the nanowireswas not always exactly perpendicular to the electrodes). Ourvalue for doped conductivity is reasonable for a doped filmof MEH-PPV.30

Although BBL is not sensitive to oxidants such as I2, wewere able to distinguish pairs of electrodes that were spannedby BBL nanowires from those that were not. Figure 6C is acurrent versus voltage (I-V) plot; we attribute the conductiv-ity to a set of 400 BBL nanowires that spans the 50 µm gapbetween the electrodes. The second plot in Figure 6C(“SiO2”) is a hysteretic curve of a pair of electrodes that arenot spanned by nanowires. It displays much lower conduc-tance, characteristic of conductivity across the bare SiO2

substrate. In the linear region in the center of the BBL curve,we calculated an approximate conductivity of 10-6 S cm-1.The conductivity of BBL is a strong function of its processinghistory and ranges over 14 orders of magnitude in theliterature: 10-14 to 10-12 S cm-1 for pristine films,39 10-6 Scm-1 for annealed films,40 and 100 S cm-1 for p-type-dopedfilms (AsF5).40 Residual MEH-PPV could not have beenresponsible for the conductivity, because at the end of theetching step, the junction no longer responded to I2. Theshape of the I-V plot shown in Figure 6C for BBL nanowiresis consistent with those of thin films that we have measured.

In conclusion, we have developed a general technique forthe fabrication of conjugated polymer nanowires. We believethat this technique is at least as simple, conceptually andoperationally, as any existing method. It has the potential toreplace other techniques in many circumstances, particularlywhere sophisticated processes such as electron-beam lithog-raphy are not available and when relatively small lengths(<1 mm) of nanostructured material are required. We areunaware of a film-forming conjugated polymer that, in

principle, cannot be made into nanowires by this method. Itshould be possible to select sacrificial materials that haveorthogonal processing conditions (e.g., rates of wet or dryetching) to many conjugated polymers of interest. The entireprocess, from spin-coating to electrical characterization, canbe executed in a single day and does not require access to acleanroom. We believe that nanoskiving can address someof the limitations of current techniques for the fabricationof conjugated polymer nanostructures, and that it will enablenew architectures that have been otherwise difficult to obtain.

Acknowledgment. This work was supported by theNational Science Foundation under award CHE-0518055.The authors used the shared facilities supported by the NSFunder NSEC (PHY-0117795 and PHY-0646094) and MR-SEC (DMR-0213805). This work was performed in partusing the facilities of the Center for Nanoscale Systems(CNS), a member of the National Nanotechnology Infra-structure Network (NNIN), which is supported by theNational Science Foundation under NSF Award No. ECS-0335765. CNS is part of the Faculty of Arts and Sciences atHarvard University. The authors thank Dr. Richard Schalekfor training on the ultramicrotome.

Supporting Information Available: Detail of the fabrica-tion process. This information is available free of charge viathe Internet at http://pubs.acs.org.

References(1) Xu, Q. B.; Bao, J. M.; Capasso, F.; Whitesides, G. M. Angew. Chem.,

Int. Ed. 2006, 45, 3631–3635.(2) Xu, Q. B.; Gates, B. D.; Whitesides, G. M. J. Am. Chem. Soc. 2004,

126, 1332–1333.(3) Xu, Q. B.; Perez-Castillejos, R.; Li, Z. F.; Whitesides, G. M. Nano

Lett. 2006, 6, 2163–2165.(4) Xu, Q. B.; Rioux, R. M.; Whitesides, G. M. ACS Nano 2007, 1,

215227.(5) Xu, Q. B.; Bao, J. M.; Rioux, R. M.; Perez-Castillejos, R.; Capasso,

F.; Whitesides, G. M. Nano Lett. 2007, 7, 2800–2805.(6) Liu, H. Q.; Kameoka, J.; Czaplewski, D. A.; Craighead, H. G. Nano

Lett. 2004, 4, 671–675.(7) Hernandez, S. C.; Chaudhuri, D.; Chen, W.; Myung, N. V.; Mulchan-

dani, A. Electroanalysis 2007, 19, 2125–2130.(8) Ramanathan, K.; Bangar, M. A.; Yun, M.; Chen, W.; Myung, N. V.;

Mulchandani, A. J. Am. Chem. Soc. 2005, 127, 496–497.(9) Wanekaya, A. K.; Chen, W.; Myung, N. V.; Mulchandani, A.

Electroanalysis 2006, 18, 533–550.(10) Wanekaya, A. K.; Bangar, M. A.; Yun, M.; Chen, W.; Myung, N. V.;

Mulchandani, A. J. Phys. Chem. C 2007, 111, 5218–5221.(11) Liu, H. Q.; Reccius, C. H.; Craighead, H. G. Appl. Phys. Lett. 2005,

87.(12) Samitsu, S.; Shimomura, T.; Ito, K.; Fujimori, M.; Heike, S.;

Hashizume, T. Appl. Phys. Lett. 2005, 86.(13) He, H. X.; Li, C. Z.; Tao, N. J. Appl. Phys. Lett. 2001, 78, 811–813.(14) Sirringhaus, H.; Tessler, N.; Friend, R. H. Science 1998, 280, 1741–

1744.(15) Yu, G.; Gao, J.; Hummelen, J. C.; Wudl, F.; Heeger, A. J. Science

1995, 270, 1789–1791.(16) Sirringhaus, H.; Brown, P. J.; Friend, R. H.; Nielsen, M. M.;

Bechgaard, K.; Langeveld-Voss, B. M. W.; Spiering, A. J. H.; Janssen,R. A. J.; Meijer, E. W.; Herwig, P.; de Leeuw, D. M. Nature 1999,401, 685–688.

(17) Chiang, C. K.; Fincher, C. R.; Park, Y. W.; Heeger, A. J.; Shirakawa,H.; Louis, E. J.; Gau, S. C.; Macdiarmid, A. G. Phys. ReV. Lett. 1977,39, 1098–1101.

(18) Menard, E.; Meitl, M. A.; Sun, Y. G.; Park, J. U.; Shir, D. J. L.; Nam,Y. S.; Jeon, S.; Rogers, J. A. Chem. ReV. 2007, 107, 1117–1160.

(19) Cui, Y.; Wei, Q. Q.; Park, H. K.; Lieber, C. M. Science 2001, 293,1289–1292.

2104 Nano Lett., Vol. 8, No. 7, 2008

166

(20) McQuade, D. T.; Pullen, A. E.; Swager, T. M. Chem. ReV. 2000, 100,2537–2574.

(21) Duvail, J. L.; Retho, P.; Fernandez, V.; Louarn, G.; Molinie, P.;Chauvet, O. J. Phys. Chem. B 2004, 108, 18552–18556.

(22) Smela, E. AdV. Mater. 2003, 15, 481–494.(23) Yun, M. H.; Myung, N. V.; Vasquez, R. P.; Lee, C. S.; Menke, E.;

Penner, R. M. Nano Lett. 2004, 4, 419422.(24) Kameoka, J.; Czaplewski, D.; Liu, H. Q.; Craighead, H. G. J. Mater.

Chem. 2004, 14, 1503–1505.(25) Li, D.; Babel, A.; Jenekhe, S. A.; Xia, Y. N. AdV. Mater. 2004, 16,

2062–2066.(26) McCann, J. T.; Chen, J. I. L.; Li, D.; Ye, Z. G.; Xia, Y. N. Chem.

Phys. Lett. 2006, 424, 162–166.(27) Dong, B.; Lu, N.; Zelsmann, M.; Kehagias, N.; Fuchs, H.; Torres,

C. M. S.; Chi, L. F. AdV. Funct. Mater. 2006, 16, 1937–1942.(28) Lim, J. H.; Mirkin, C. A. AdV. Mater. 2002, 14, 1474–1477.(29) Gates, B. D.; Xu, Q. B.; Stewart, M.; Ryan, D.; Willson, C. G.;

Whitesides, G. M. Chem. ReV. 2005, 105, 1171–1196.(30) Wudl, F.; Srdanov, G. Conducting Polymer Formed Of Poly(2-

methoxy,5-(2!-ethylhexyloxy)-p-phenylenevinylene. United States Patent5,189,136.

(31) Babel, A.; Jenekhe, S. A. J. Am. Chem. Soc. 2003, 125, 13656–13657.

(32) Arnold, F. E.; Vandeuse, R. Macromolecules 1969, 2, 497–&.(33) Alam, M. M.; Jenekhe, S. A. Chem. Mater. 2004, 16, 4647–4656.(34) Roberts, M. F.; Jenekhe, S. A. Polymer 1994, 35, 4313–4325.(35) Babel, A.; Zhu, Y.; Cheng, K. F.; Chen, W. C.; Jenekhe, S. A. AdV.

Funct. Mater. 2007, 17, 2542–2549.(36) Halls, J. J. M.; Walsh, C. A.; Greenham, N. C.; Marseglia, E. A.;

Friend, R. H.; Moratti, S. C.; Holmes, A. B. Nature 1995, 376, 498–500.

(37) Alternatively, we have used epoxy blocks that contain individual stripsof the composite film that, after sectioning, we addressed electricallywith evaporated top contacts through a stencil mask.

(38) We note that our two-point measurement does not decouple contactresistance from material resistance. To determine the contribution ofcontact resistance, a four-point measurement would be required (seeref. 6).

(39) Alam, M. M.; Jenekhe, S. A. J. Phys. Chem. B 2002, 106, 11172–11177.

(40) Coter, F.; Belaish, Y.; Davidov, D.; Dalton, L. R.; Ehrenfreund, E.;McLean, M. R.; Nalwa, H. S. Synth. Met. 1989, 29, E471-E476.

NL8009318

Nano Lett., Vol. 8, No. 7, 2008 2105

167

SI - 1

Supporting Information:

Fabrication of Conjugated Polymer Nanowires by Edge Lithography

Darren J. Lipomi, Ryan C. Chiechi, Michael D. Dickey, and George M. Whitesides*

Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street,

Cambridge, Massachusetts, U.S.A. 02138

*Corresponding Author

Telephone Number: (617) 495-9430

Fax Number: (617) 495-9857

Email Address: gwhitesides@gmwgroup.harvard.edu

168

SI - 2

EXPERIMENTAL SECTION

Fabrication of MEH-PPV and BBL nanowires. A glass slide was cut into a 2-cm

square and spin-coated with BBL (obtained from Aldrich, made into a 0.5 wt. % solution in

methanesulfonic acid (MSA) (Fluka) prepared by dissolving 370 mg of polymer in 50 mL of

MSA) at 3 krpm with a ramp rate of 1 krpm/s for 30 s. (MSA causes burns and should be spin-

coated in a fume hood with the sash down. We generally used a homemade high-density

polyethylene liner for the basin of the spin-coater, as MSA reacts slowly with aluminum foil.)

The substrate was removed from the spinner with tweezers and immersed in deionized water for

5 s to remove MSA. The BBL film was dried with an N2 gun, during which time the film

changed from dark purple to light purple with a metallic gold luster. On top of the BBL film, we

spin-coated MEH-PPV (purchased from Aldrich, avg. MW = 70,000 – 100,000, made into a 0.6

wt. % solution in chloroform, prepared by dissolving 444 mg of polymer in 50 mL of

chloroform) at 3 krpm with a ramp rate of 1 krpm/s. BBL and MEH-PPV films were stacked in

this manner fifty times for 100 total layers of polymer with average thickness of 100 nm for each

layer. The composite film had a thickness of 10 µm (as determined by SEM). The film was

annealed under vacuum at 125 °C for 5 min, and scored around the edges of the glass substrate

with a scalpel (in a square ~1 mm from the edge of the glass). The substrate with the polymer

film was immersed in methanol, and placed in a sonicator bath for ~20 s. This action

delaminated the edges of the film from the glass. The film was then easily peeled off with

tweezers, removed from the methanol. The film was cut into a 1 x 2 cm rectangle, which was

folded in half several times. The folded structure was embedded in epoxy prepolymer (Epo-Fix,

obtained from Electron Microscopy Sciences, mixed and degassed at <10 torr for 30 min before

use), and cured at 60 °C for 2 h in a polyethylene mold (Electron Microscopy Sciences). The

169

SI - 3

folded film relaxed during curing to yield the cross section shown in Figure 5A. Alternatively,

instead of folding, we cut the composite film into strips (5 mm x 1 mm) with scissors. The

cooled block was cut with a hand saw to expose the cross section of the film. The block was

trimmed and sectioned with the ultramicrotome (Leica Ultracut UCT, equipped with a diamond

knife Diatome Ultra 35) as described previously (see Supporting Information of ref. 5).

Imaging and Profilometry. Optical imaging (Figure 5A) was performed using an

upright optical microscope (Leica DMRX). Scanning electron microscope (SEM) images

(Figure 5B-5D) of the epoxy sections were acquired with a Zeiss Ultra55 or Supra55 VP FESEM

at 2 or 5 kV at a working distance of 2-6 mm. Profilometry data was obtained on a Veeco

Dektak 6M Stylus Profilometer.

Electrical Characterization. The contact pads were patterned by a typical lift-off

process, as follows. A 300-nm thermal oxide was grown on an N/Phos test grade silicon wafer,

bulk resistivity 1 to 10 ! cm and <100> orientation. The wafer was cleaned with an oxygen

plasma, treated with hexamethyldisilazane, and coated with Shipley 1818 positive photoresist.

The resist was patterned by UV irradiation through a transparency mask in contact mode. After

development of the exposed areas, the substrate was coated with a 5-nm-thick Ti adhesion layer,

followed by a 50-nm gold layer, by electron-beam evaporation. The substrate was placed in a

beaker containing acetone and immersed in a sonication bath. The metal film in the unexposed

areas detached from the substrate.

Individual epoxy sections containing MEH-PPV and BBL nanowires were placed on the

electrodes with the Perfect Loop tool (Electron Microscopy Sciences). The orientation of the

sections could be controlled by trimming the block face, prior to sectioning, such that the area of

the sections was slightly larger than the perimeter of the Perfect Loop tool. In this way, the

170

SI - 4

edges of the section stuck to the Perfect Loop and did not rotate during the transfer of the

sections from the water bath of the ultramicrotome to the substrate.

171

Appendix III

Electrically Addressable Parallel Nanowires with 30 nm Spacing from

Micromolding and Nanoskiving

Michael D. Dickey, Darren J. Lipomi, Paul J. Bracher, and George M. Whitesides

Department of Chemistry and Chemical Biology, Harvard University

12 Oxford St., Cambridge, Massachusetts, 02138 (USA)

Reproduced with permission from

Nano Lett. 2008, 8, 4568-4573

Copyright 2008, American Chemical Society

172

Electrically Addressable ParallelNanowires with 30 nm Spacing fromMicromolding and NanoskivingMichael D. Dickey, Darren J. Lipomi, Paul J. Bracher, and George M. Whitesides*

Department of Chemistry and Chemical Biology, HarVard UniVersity, 12 OxfordStreet, Cambridge, Massachusetts 02138

Received September 16, 2008; Revised Manuscript Received October 20, 2008

ABSTRACT

This paper describes the fabrication of arrays of parallel, electrically addressable metallic nanowires by depositing alternating layers of thinfilms of metal and polymersboth planar and topographically patternedsand sectioning the laminated structures with an ultramicrotome(nanoskiving). The structures that resulted from this process had two distinct regions: one in which parallel Au nanowires were separated bya minimum distance of 30 nm, and one in which the nanowires diverged such that the distal ends were individually addressable by low-resolution (g10 µm) photolithography. Conductive polyaniline (PANI) was electrochemically deposited across the nanowire electrodes todemonstrate their electrical addressability, continuity, and physical separation. Before deposition, the wires were electrically isolated; with thePANI, they were electrically connected. After dry etching to remove the polymer, the gap between the nanowire electrodes returned to aninsulating state. This procedure provides a method for making wires with dimensions and separations of <50 nm without the use of e-beamor focused-ion-beam “writing” and opens applications in organic and molecular electronics, chemical and biological sensing, and other fieldswhere nanoscale distances between parallel conductive electrodes are desirable.

This paper describes the use of nanoskiving,1-4 a techniquethat uses an ultramicrotome to section thin films of metalembedded in a polymer matrix, to fabricate parallel nanow-ires with controlled spacing (gaps as narrow as 30 nm) whosedistal ends diverge in a way that makes it possible to addresseach wire electrically using low resolution (g10 µm)photolithography. The process can be used to fabricate twoor three (or more) nanowires that are parallel over distancesof ∼50 µm and comprises five basic steps: (i) deposition ofa thin (<100 nm) metal film on a flat polymeric substrate;(ii) deposition of a thin polymer film onto the metal by spin-coating; (iii) molding microscale parallel lines of polymeron top of the composite structure; (iv) shadow evaporationof a thin metal film on the composite structure; and (v)nanoskiving (see Figure 1 for an overview of the process; adetailed description of it follows). The geometries of thestructures accessible by this technique resemble those thatare ordinarily made using significantly more sophisticated,expensive, and slower techniques (such as electron- andfocused-ion-beam (FIB) lithography) than those used in thisstudy. The technique is particularly useful for manufacturingindistinguishable copies (thousands per hour, in principle)of addressable electrodes for the characterization of electronicmaterials and the fabrication of devices that rely on chargetransport over nanoscale dimensions (e.g., sensors, capacitors,

resistors, photovoltaics, transistors, diodes, and molecularjunctions). The use of low-resolution photolithography makesthe method accessible to general users who do not haveaccess to high-resolution, direct-write techniques, or whowish to use materials that are not allowed in a cleanroomdedicated to solid-state electronics.

Background. Nanowire electrodes that are separated bysmall (<100 nm) gaps and are electrically addressableindividually are useful in sensors,5,6 as electrodes for dielec-trophoresis (used to entrap nanostructures and molecules7,8)and electrochemistry,9,10 in molecular junctions,11,12 and astest-bed structures for studying nanoscale phenomena or newnanoscale architectures.13

Electrical characterization of nanostructures and theirincorporation into functional devices depends on the forma-tion of stable electrical contacts between the nanostructuresand external electrical circuits; most applications require aminimum of two electrodes for simple electronic function.It is challenging to address multiple individual nanowireselectrically that are in close proximity (e.g., in parallel) andseparated by a nanoscale gap (<100 nm) without inadvert-ently contacting multiple wires with a single contact pad.

There are only a small number of techniques capable ofgenerating nanowire electrodes with nanoscale separation thatare easy to address electrically. Of these techniques, fewcombine useful levels of generality and simplicity. “Direct-write” approaches to nanofabrication, most prominently,

* To whom correspondence should be addressed. Tel: (617) 495-9430.Fax: (617) 495-9857. E-mail: gwhitesides@gmwgroup.harvard.edu.

NANOLETTERS

2008Vol. 8, No. 12

4568-4573

10.1021/nl8028174 CCC: $40.75 ! 2008 American Chemical SocietyPublished on Web 11/04/2008

173

e-beam and FIB lithography, are capable of forming addres-sable nanowires but are expensive, technically challenging,damaging to organic materials, and limited to rigid planarsubstrates.14,15 Addressing parallel nanowires using thesetechniques becomes increasingly difficult as the spacingbetween the wires decreases, and it is physically impossiblewhen the spacing is smaller than the resolution of the writer(∼10 nm).

Photolithography is, by far, the most common method ofpattern replication in the semiconductor industry and inresearch laboratories. State-of-the-art tools for photolithog-raphy are capable of patterning sub-100 nm contacts withprecise alignment, but most research laboratories are limitedby equipment that has relatively low resolution (∼1 µm) andalignment precision. In practice, it is nearly impossible inan academic laboratory to align a contact pad usingphotolithography such that the pad only contacts one structurethat is in close proximity (<100 nm) to a neighboring one.

Nonlithographic methods that have been developed tomake addressable, narrow junctions between electrodesinclude shadow mask evaporation,16 mechanical break junc-tion techniques,17 local oxidative cutting of carbon nano-tubes,18 electromigration,19 and on-wire lithography.20 Ingeneral, these methods are restricted to two electrodes withlimited geometrical configurations (typically, nanoscale

breaks in a single wire), and in some cases the techniquesare challenging experimentally. Nonlithographic methodsgenerally are not practical for creating addressable parallelnanowires with separations close to the thickness of the wires.

Nanoskiving. The goal of this work was to develop asimple method to produce nanowires in close proximity thatcan be addressed individually without the use of direct-writetechniques. “Nanoskiving” is the use of an ultramicrotometo generate nanostructures from planar or topographicallypatterned thin films.2-4 The technique is attractive as anapproachfornanofabricationbecauseofitssimplicitysnanoskivingconverts features that are thin in the vertical dimension (thinfilms) to features that are thin in the lateral dimension (“edgelithography”).21 Our laboratory has used nanoskiving tofabricate arrays of metallic nanostructures for optical ap-plications, as well as nanowires of conjugated polymers forelectronic and optoelectronic applications.22-24 We chose touse nanoskiving because it is capable of forming nanowireswithout the use of sophisticated lithographic techniques;nanoskiving offers a means of producing uniform nanostruc-tures reproducibly and simply. It is easily capable ofproducing multiple indistinguishable copies (that is, con-secutively cross-sectioned slabs) of a parent structure. Wesought to produce parallel nanowires whose lateral arrange-ment with respect to each other would have two distinct

Figure 1. Schematic representation of the procedure used for the fabrication of electrically addressable nanowires. Spin-coating producesa thin-film of polymer on top of a flat polymer substrate coated with gold. A microchannel defined in PDMS is placed onto the substrateand is filled with prepolymer that is then cured with UV light. The PDMS is removed and a thin layer of gold is deposited onto the substrateby evaporation. Embedding the substrate in polymer and slicing it with an ultramicrotome creates parallel nanowires with addressableregions. The polymer can be removed by an oxygen plasma.

Nano Lett., Vol. 8, No. 12, 2008 4569

174

regions: one in which parallel nanowires were separated bya small (<100 nm wide), defined gap, and another in whichthe wires diverged until they were separated by a largeenough spacing (>10 µm) to be addressed individually bycontact photolithography.

Choice of Materials. Nanoskiving takes advantage of theprecise control of thickness provided by techniques used forthin-film deposition. We utilized e-beam evaporation todeposit the Au that would ultimately become the wires, andspin-coating to define the dielectric spacer layer between thewires. We chose Au because it is easy to deposit, sectionswell, and does not oxidize. For the spacer layer (in theparallel region), we utilized a photocurable prepolymer whosethickness was defined by spin-coating.

Micromolding in Capillaries. To create the addressableregion of the structure, we used micromolding in capillaries(MIMIC).25 MIMIC is a process that forms structures on asubstrate by drawing a fluid into an elastomeric mold. Wefilled a mold with optical adhesive prepolymer, cured thefluid, and removed the mold to produce topographic featuresthat were an inverse replica of the mold. The use ofpolydimethylsiloxane (PDMS) ensured conformal contact ofthe mold with the substrate during the filling process andfacile removal of the mold from the substrate due to the lowsurface energy of PDMS. We fabricated the molds using softlithographic techniques.26,27 We used a commercially avail-able optical adhesive (Norland Optical Adhesive 61) as theembedding material for nanoskiving because it contains thiolgroups that promote adhesion to the Au, it cures rapidly (<5min) by exposure to UV radiation, it is inexpensive, and itsections well in the ultramicrotome (it displays minimalcompression or other distortion during cutting).

Fabrication. Figure 1 summarizes the process used forfabrication. We evaporated 40-70 nm of Au onto a flat pieceof cured optical adhesive; this layer of gold ultimatelybecame one of the nanowires. Onto the gold, we spin-coateda thin film of a prepolymer solution containing 2.5 wt % ofa 1:1 (w/w) mixture of a multifunctional thiol (pentaerythritoltetrakis(3-mercaptopropionate)) and an acrylate (di(trimethy-lolpropane)tetraacrylate) diluted in solvent (propylene glycolmethyl ether acetate, PGMEA). This layer could be removed(as a sacrificial material) after sectioning by etching in anoxygen plasma or could serve as a functional material (as adielectric) between the two nanowires. We used this materialbecause optical adhesive, which serves as the embeddingmaterial for microtoming, does not form smooth films at<100 nm thickness.

We filled microfluidic channels in a PDMS mold withoptical adhesive and cured the adhesive with UV light. Thecross-sectional shape of the channels would ultimately definethe shape of the addressable region of the wires. We foundthat smooth, semicylindrical features sectioned better thanthose with sharp, rectangular features, presumably becauseof the way in which mechanical stress is distributed inrounded features during sectioning. We fabricated the PDMSchannels by two methods: (i) replica molding lines of positivephotoresist (50 µm wide, 30 µm tall lines of AZ P4903 ona 100 µm pitch), then reflowing the resist at 130 °C for 1 h

to round the tops of the features, and (ii) replica moldinglines of negative photoresist (50 µm wide, 30 µm tall linesof SU-8 resist on a 100 µm pitch), then spin-coating andcuring an additional 10 µm thick layer of SU-8 over thesubstrate to round the features. After removing the PDMSmold, we evaporated 40-70 nm of Au onto the substrate ata glancing angle (∼60° angle between the beam and thesubstrate) to cover the region of the substrate between themolded features and one of the two sets of sidewalls of themolded features.

We cut the substrate with a razor blade into small strips(∼2 mm wide) parallel to the long axes of the ridges definedby the MIMIC process. We placed the strips into polyeth-ylene molds (Electron Microscopy Sciences), covered themin optical adhesive, and cured the adhesive with a mercurylamp. We trimmed the face of the resulting block (the faceis perpendicular to both the gold film and the lines definedby MIMIC, see Figure 1) with a razor blade to an area of∼1 mm × 1 mm and then sliced thin sections of the sampleusing an ultramicrotome (Leica Ultracut UCT) fitted with adiamond knife (DiatomeTM Ultra 35°). The sectioning speed(i.e., the speed of the knife through the sample block) was1 mm·s-1 for a selected section thickness (typically 70 nm).The sectioned slices of polymer containing the embeddednanostructures floated on the surface of the water in thereservoir of the knife. We transferred the sections (using thePerfect Loop tool, Electron Microscopy Sciences) to thesurfaces of Si wafers bearing ∼600 nm layers of thermallygrown oxide that prevented electrical shorting of the nanow-ires through the substrate.

We placed the samples on a hot plate at 150 °C to improvethe adhesion of the section to the substrate. A 10 minexposure to oxygen plasma (0.2 Torr, 150 W barrel etcher)removed the embedding material from the sections, whilethe metal nanowires remained on the insulating substrate.We addressed the wires by contact lithography and a typicallift-off process. Briefly, we spin-coated a thin (∼3 µm) layerof Shipley 1822 onto the substrate, and baked the photoresistfor 3 min at 110 °C. After cooling the substrate to roomtemperature, we aligned the contact pads on the photomaskwith the addressable regions of the wire and irradiated thephotoresist (ABM mask aligner, ∼70 W). Developing (inCD-30) for 30 s produced a pattern onto which we deposited4 nm of Ti and 40 nm of Au. Acetone removed the resistand extra metal (without sonication), leaving behind Auelectrodes on the individual nanowires.

Before electrical characterization, we secured Cu wiresto the Au contact pads with drops of graphite ink (Ercon3456). After drying the ink, we covered the graphite andexposed regions of the Cu wire in five-minute epoxy. Wecharacterized the wires electronically using a Keithley6430 subfemtoammeter. We deposited polyaniline elec-trochemically from an aqueous solution of 0.1 M aniline(freshly distilled at 90 °C under reduced pressure), 0.5 MNa2SO4, and 0.1 M H2SO4 (pH ∼ 1). We used abipotentiostat (Pine, AFCBP1) to electropolymerize theaniline at +0.8 V versus Ag/AgCl with platinum foilserving as the counterelectrode.

4570 Nano Lett., Vol. 8, No. 12, 2008

175

Results and Discussion. Figure 2 is a series of images ofelectrically addressable gold nanowires on a Si/SiO2 sub-strate. Figure 2a is an optical micrograph of a polymeric slabthat contains six embedded addressable structures. Figure2b is a scanning electron micrograph (SEM) of a singleaddressable structure after removal of the polymer slab byoxygen plasma. The structure consists of a parallel regionwhere two wires (∼70 nm wide) are separated by ∼30 nm,and a diverging, addressable region where the wires arespaced by ∼30 µm (i.e., approximately 3 orders of magnitudegreater spacing than the parallel region). The parallel regionin Figure 2c is ∼50 µm long. We produced parallel regionsas long as ∼150 µm, but in principle the parallel region couldbe as long as the width of the slab (∼1 mm).

The cross-sectional profile of the microfluidic channel usedduring the MIMIC step defines the 2D geometry of theaddressable region of the nanowire electrodes. We found thatrectangular addressable regions tended to have discontinuities(breaks) in the wires primarily along portions of the wiresperpendicular to the edge of the diamond knife (i.e., parallelto the direction of cutting). We believe that a wire is subjectto compression and breakage when it is aligned parallel tothe direction of cutting because of differences in themechanical properties of the gold and the embeddingpolymer. We tested this hypothesis by observing the effectof the orientation of a sample (consisting of a thin gold filmembedded in polymer) during cutting; wires oriented parallelto the direction of the sectioning had significantly moredamage and breaks than those formed by cutting perpen-dicular to the wires. Sectioning at an intermediate angle (e.g.,a 45° angle between the wire and the blade) also reduced

the amount of total damage to the wires, but we still observedsome breakage in the longer portions of the wires. On thebasis of these observations, we aligned the knife perpen-dicular to the parallel region of the wires and chose to userounded addressable regions (the addressable region in Figure2 is effectively a semicircle) to minimize the portion of thewires aligned parallel to the direction of cutting. Otherdesigns of the addressable region are discussed in theSupporting Information.

The yield of a representative batch of 49 pairs ofaddressable, parallel nanowires was 73%; this value includedonly those structures that were both (i) spatially isolated(individually) and (ii) physically continuous (independently)along the entire length of either nanowire (∼100 µm).Independent experiments indicated that nanowires that ap-peared physically continuous by SEM analysis were alsoelectrically continuous. (See Supporting Information for ademonstration of electrical continuity along single nanow-ires.) The remaining 27% of structures contained defects suchas shorts (where the two wires touched) and breaks. Weattribute the defects to damage to the nanostructures bydefects in the knife blade, delamination of the Au films fromthe polymeric matrix during the sectioning process, and otherstresses of cutting (see Supporting Information for descrip-tions of the defects).28 An individual epoxy section istypically ∼1 mm wide, and an individual structure (Figure2a) is approximately 100 µm long; we can therefore produce6-10 addressable samples (of the type shown in Figure 2a)per section.

Electrical Characterization. We electrically addressed theparallel wires by standard contact lithographic and lift-offprocedures. A simple experiment demonstrated the con-nectivity of our lithographically defined contact pads to thenanowires, the continuity of the nanowires, and the physicalseparation of the nanowires. We potentiostatically depositeda conductive polymer, polyaniline (PANI), in the gapbetween the nanowires (Figure 3a). Electrodeposition ofconductive polymers in the trenches of hard substrates(defined by e-beam lithography) has been used to fabricatepolymeric nanowires; these nanowires behave as high-surface-area chemical and biological sensors.5,6,29-31 The goldnanowires (∼100 nm wide with ∼40 nm gap) served as theworking electrodes; the polymer deposited on the wires andbetween them met in the center. Traces of current versusvoltage (I-V) for the nanowires joined with PANI yieldedlinear functions without noticeable hysteresis (Figure 3c).After we exposed the substrate to an oxygen plasma toremove the polyaniline (Figure 3b), a subsequent I-V tracedisplayed much lower conductivity and hysteresis; weattribute the residual conductivity to the insulating substrate.The conductivity after removing the polyaniline was similarto that obtained for samples that were not exposed topolyaniline or oxygen plasma.

Multiple Nanowires. An advantage of using nanoskiving,compared to other methods of making nanoscale gapsbetween electrodes, is that the technique is not limited to asingle geometry. As a proof of principle, we modified theprocedure in Figure 1 to include two additional steps

Figure 2. Images of electrically addressable Au nanowires. (a)Optical micrograph of a slab of polymer containing six addressablestructures. (b) SEM overview of a single addressable structure. (c)Close-up of structure reveals two parallel nanowires that diverge.The point of divergence corresponds to the semicylinder of opticaladhesive formed using MIMIC. (d) Close-up image of two parallelgold nanowires separated by a narrow gap (∼30 nm).

Nano Lett., Vol. 8, No. 12, 2008 4571

176

(summarized in Figure 4). After the second deposition ofgold (Figure 1), we (i) spin-coated a thin film of polymer toserve as an additional spacer layer and (ii) deposited a thirdlayer of gold (which ultimately became the third wire) whileorienting the sample such that gold would deposit on theparallel region and the other side of the protruding addres-sable region. Figure 4 includes an SEM of this three-wirestructure. The ability to fabricate three addressable terminalsis useful for incorporating additional functionality into anelectrical device, such as a transistor with a source, drain,and gate.

Conclusions. This paper describes a simple method tofabricate individually addressable nanowires without usingdirect-write lithographic tools. The technique combinesnanoskiving and MIMIC to form nanowires that are parallelin one region of the structure, but diverge to create a regionin which the wires can be addressed individually using low-resolution lithography.

We believe the technique will be of greatest interest toresearchers who want to create nanoscale test structures withoutthe use of sophisticated equipment or techniques. The ability

to create many (hundreds to thousands of) easily manipulatedpolymer slabs containing these nanoscale structures (by suc-cessive sectioning) makes nanoskiving a very practical proce-dure. We fabricated parallel wires with a sacrificial polymericspacer between the wires, but in principle, active materials (e.g.,dielectrics, conducting polymers, and oxides) could be incor-porated into that region during preparation of the sample. Withfurther development, it may be possible to adapt this processto form narrower gaps, which could be used, for example, formolecular junctions. The nanowires may also be useful to createlarge local electric fields with the application of small voltages(e.g., ∼2 V applied across a 30 nm gap would exceed thedielectric breakdown of air). In addition, the structures featuringthree-terminal (or more) electrodes may be adapted to formsophisticated test-bed devices, such as nanoscale transistors(source, gate, drain).

Acknowledgment. This work was supported by NSF awardsPHY-064609 and CHE-0518055. We used shared facilitiessupported by the MRSEC (DMR-0213805). This work wasperformed, in part, using the facilities of the Center forNanoscale Systems (CNS), a member of the National Nano-technology Infrastructure Network (NNIN), which is supportedby the National Science Foundation under NSF award ECS-0335765. CNS is part of the Faculty of Arts and Sciences atHarvard University. D.J.L. acknowledges a graduate fellowshipfrom the American Chemical Society, Division of OrganicChemistry, sponsored by Novartis. P.J.B. thanks the NSF andHarvard Origins Initiative for graduate fellowships.

Supporting Information Available: More images ofwires using different spacer geometries (Figure S1), ad-ditional experimental details, an image of the contact pads(Figure S2), a demonstration of electrical continuity along asingle wire (Figure S3), and images of damaged wires (FigureS4). This material is available free of charge via the Internetat http://pubs.acs.org.

References(1) Xu, Q.; Rioux, R. M.; Dickey, M. D.; Whitesides, G. M. Acc. Chem.

Res., published online July 23, 2008, http://dx.doi.org/10.1021/ar700194y.

(2) Xu, Q.; Perez-Castillejos, R.; Li, Z.; Whitesides, G. M. Nano Lett.2006, 6, 2163–2165.

(3) Xu, Q.; Bao, J.; Capasso, F.; Whitesides, G. M. Angew. Chem., Int.Ed. 2006, 45, 3631–3635.

(4) Xu, Q.; Gates, B. D.; Whitesides, G. M. J. Am. Chem. Soc. 2004,126, 1332–1333.

(5) Ramanathan, K.; Bangar, M. A.; Yun, M.; Chen, W.; Myung, N. V.;Mulchandani, A. J. Am. Chem. Soc. 2005, 127, 496–497.

(6) Yun, M.; Myung, N. V.; Vasquez, R. P.; Lee, C.; Menke, E.; Penner,R. M. Nano Lett. 2004, 4, 419–422.

(7) Bezryadin, A.; Dekker, C.; Schmid, G. Appl. Phys. Lett. 1997, 71,1273–1275.

(8) Holzel, R.; Calander, N.; Chiragwandi, Z.; Willander, M.; Bier, F. F.Phys. ReV. Lett. 2005, 95, 128102/1–128102/4.

(9) Murray, R. W. Chem. ReV. 2008, 108, 2688–2720.(10) Arrigan, D. W. M. Analyst 2004, 129, 1157–1165.(11) Tao, N. J. Nat. Nanotechnol. 2006, 1, 173–181.(12) Nitzan, A.; Ratner, M. A. Science 2003, 300, 1384–1389.(13) Snider, G.; Kuekes, P.; Hogg, T.; Williams, R. S. Appl. Phys. A 2005,

80, 1183–1195.(14) Carcenac, F.; Malaquin, L.; Vieu, C. Microelectron. Eng. 2002, 61-

62, 657–663.(15) Gates, B. D.; Xu, Q.; Love, J. C.; Wolfe, D. B.; Whitesides, G. M.

Annu. ReV. Mater. Res. 2004, 34, 339–372.

Figure 3. SEM images of electrically addressable Au nanowires.(a) After polyaniline was grown electrochemically across the wires.(b) Oxygen plasma removed the polymer, restoring pristine wires.(c) Electrical characterization of the wires. Inclusion of the polymerbetween the gaps increases the conductivity of the junctions. Inthe absence of polymer, the current decreases significantly.

Figure 4. Three electrically addressable nanowires. (a) Schematicdepicting the architecture of the wires. (b) Close up image of threeparallel wires.

4572 Nano Lett., Vol. 8, No. 12, 2008

177

(16) Tang, J.; Wang, Y.; Klare, J. E.; Tulevski, G. S.; Wind, S. J.; Nuckolls,C. Angew. Chem., Int. Ed. 2007, 46, 3892–3895.

(17) Ohnishi, H.; Takayanag, K. Nature 1998, 395, 780–783.(18) Guo, X.; Small, J. P.; Klare, J. E.; Wang, Y.; Purewal, M. S.; Tam,

I. W.; Hong, B. H.; Caldwell, R.; Huang, L.; O’Brien, S.; Yan, J.;Breslow, R.; Wind, S. J.; Hone, J.; Kim, P.; Nuckolls, C. Science 2006,311, 356–359.

(19) Park, H.; Lim, A. K. L.; Alivisatos, A. P.; Park, J.; McEuen, P. L.Appl. Phys. Lett. 1999, 75, 301–303.

(20) Qin, L.; Park, S.; Huang, L.; Mirkin, C. A. Science 2005, 309, 113–115.

(21) Gates, B. D.; Xu, Q.; Stewart, M.; Ryan, D.; Willson, C. G.;Whitesides, G. M. Chem. ReV. 2005, 105, 1171–1196.

(22) Xu, Q.; Bao, J.; Capasso, F.; Whitesides, G. M. Angew. Chem., Int.Ed. 2006, 45, 3631–3635.

(23) Xu, Q.; Bao, J.; Rioux, R. M.; Perez-Castillejos, R.; Capasso, F.;Whitesides, G. M. Nano Lett. 2007, 7, 2800–2805.

(24) Lipomi, D. J.; Chiechi, R. C.; Dickey, M. D.; Whitesides, G. M. NanoLett. 2008, 8, 2100–2105.

(25) Xia, Y.; Kim, E.; Whitesides, G. M. Chem. Mater. 1996, 8, 1558–1567.

(26) Xia, Y.; Whitesides, G. M. Angew. Chem., Int. Ed. 1998, 37, 550–575.

(27) Xia, Y.; Whitesides, G. M. Annu. ReV. Mater. Sci. 1998, 28, 153–184.

(28) Defects primarily consisted of breaks in the wires, shorts (i.e., touchingbetween adjacent wires), and delamination of the wires from thesurrounding polymer.

(29) Ramanathan, K.; Bangar, M. A.; Yun, M.; Chen, W.; Mulchandani,A.; Myung, N. V. Nano Lett. 2004, 4, 1237–1239.

(30) Ramanathan, K.; Bangar, M. A.; Yun, M.; Chen, W.; Mulchandani,A.; Myung, N. V. Electroanalysis 2007, 19, 793–797.

(31) Favier, F.; Walter, E. C.; Zach, M. P.; Benter, T.; Penner, R. M. Science2001, 293, 2227–2231.

NL8028174

Nano Lett., Vol. 8, No. 12, 2008 4573

178

Supporting Information for:

Electrically Addressable Parallel Nanowires with

30-nm Spacing from Micromolding and

Nanoskiving

Michael D. Dickey, Darren J. Lipomi, Paul J. Bracher, George M. Whitesides*

Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford St.,

Cambridge, Massachusetts 02138, USA

179

Addressable Region:

We formed the addressable region of the samples by filling a microfluidic channel

with a photocurable pre-polymer (NOA 61) followed by ultraviolet irradiation. To

minimize breaks in the wire, we rounded the channel walls by creating masters for the

microfluidic channels by either (i) reflowing lines of positive resist (AZ P4903) or (ii)

spin coating a conformal film of additional negative resist (SU-8) over pre-existing resist

lines. Figures S1 is an SEM image of wires formed using the negative resist method.

Figure S1. SEM images of electrically addressable Au nanowires. The diverging,

addressable region of the wires was formed using a microfluidic channel created against a

master defined by spinning negative photoresist over photolithographically defined lines.

Addressing the Nanowires:

We addressed the nanowires using contact photolithography, and conventional

lift-off procedures. Briefly, we annealed the polymeric slabs (which contained the

nanowires) on a hot plate at 150 °C for 1 min to promote contact with the substrate and

improve the adhesion of the wires to the substrate. Placing the substrate in an oxygen

plasma (8 min, 80 W, 20 sccm O2, 200 mT) removed the sacrificial polymer. Spin-

casting (2500 RPM, 30 s) coated the substrate with positive photoresist (Shipley 1822).

180

After baking on a hot plate (3 min, 110 °C), we used a mask aligner to bring the contact

pads into alignment with the addressable region of the nanowires, exposed the photoresist

(~80 W), and developed it using Shipley CD-30 for 30 sec. We rinsed the substrate with

deionized water and dried it with a stream of nitrogen. We deposited an adhesion layer

of Ti (4 nm) followed by Au (50 nm) using electron-beam evaporation. The substrate

was developed using a typical lift-off procedure (in acetone). Figure S2 is an optical

micrograph of an addressed nanowire. The pads could be aligned with the wires with

100% accuracy (i.e., each pad touched only one wire).

Figure S2. Optical micrograph of gold electrodes in contact with the addressable region

of the parallel nanowires.

Electrical Continuity in a Single Wire:

We tested single nanowires for electrical continuity by obtaining traces of current

v. voltage of individual wires from a representative batch. First, we placed an individual

section (thickness = 80 nm) containing a nanowire (width = 80 nm) on a Si wafer bearing

a thermal oxide of thickness >300 nm. The oxide layer served to insulate electrically the

nanowires from the substrate. Next, we defined contact pads (separated by 8 !m) on the

ends of each wire using a stencil mask (a single fiber of glass wool placed perpendicular

181

to the long axis of the nanowire and attached to the substrate by adhesive tape). E-beam

evaporation deposited an 80 nm film of Au on the substrate. The shadow cast by the

fiber and the tape defined contact pads spanned by a single nanowire within the section.

Figure S3a is an SEM image of a representative single Au nanowire that spans a gap

between Au electrodes. We obtained linear plots of current density v. voltage (up to |10

mV|) for each nanowire. Figure S3b shows a representative plot of current density v.

voltage. Six of the eight wires we measured were electrically continuous. We calculated

an average conductivity of 3.5 x 105 S cm-1 (standard deviation = 0.91 x 105 S cm-1). The

bulk conductivity of Au is 4.5 x 105 S cm-1. The measured conductivity was likely lower

than the bulk conductivity due to imperfections (e.g. roughness or granularity) along the

long axis of the wire, which would constrict the conductive pathway relative to a

nanowire that has a uniform rectangular cross section at every point in the direction of

charge transport. Nanowires could be destroyed in two ways: (i) by passing sufficiently

large currents (9 mA at 300 mV) through a nanowire and (ii) by scratching away a

conductive nanowire with a surgical blade.

182

Figure S3. A demonstration of electrical continuity in a single gold wire produced by

nanoskiving. (a) SEM of gold contact pads in contact with a single gold nanowire. (b)

A trace of current density as a function of applied voltage between the two pads

demonstrates that the wire is electrically continuous and ohmic.

183

Defects in the Nanowires:

Nanoskiving relies on the cutting mechanism of the microtome to produce

nanostructures. This mechanical approach is subject to producing defects in the

structures. Figure S4 shows some representative defects. There are four types of defects.

1. Scoring: Nanoskiving utilizes a knife (typically diamond) to create the nanostructures

by sectioning blocks of polymer containing embedded metal features. Microscopic

damage on the surface of the knife is transferred to the face of the slabs during

sectioning. Often, this scoring damages the nanostructures.

2. Delamination: Delamination can occur between the metal and the polymer in the

slabs during sectioning. We used a thiol-containing polymer to minimize this

delamination (when, for example, we used polymethylmethacrylate in contact with the

gold, delamination was common).

3. Breaks: Breaks in the wire can occur in portions of the nanowires that are

perpendicular to the blade (or, equivalently, parallel to the direction of sectioning). We

minimized these breaks by using a design in which the wires gradually diverged (e.g.,

Figure S1) from the parallel region to the addressable region of the wires. We believe

these breaks are due to the mismatch in mechanical modulus between the gold and the

surrounding polymer.

4. Shorts: Shorts between the two parallel wires are not common, but occur in some of

the samples. The shorts may occur due to “smearing” of the material by the knife, or due

to imperfections (e.g., holes) in the polymeric spacer layer that could occur during spin

casting.

184

Quantification of Defects: Scoring and breaks were the most common form of

observed defects. In general, the number of scores on a sample depended on the quality

of the diamond knife used for sectioning. Pristine knives produced no observable scores

on the resulting slab, but damaged knives (e.g., knives with nicks on the cutting edge)

consistently scored the slabs in the same spot on each slab. Breaks occurred in ~15-20%

of the fabricated structures (with wires having widths of 70 nm, and separated by 30 nm).

As discussed previously, the number of breaks depended on the orientation of the knife

with respect to the sample (aligning the wires parallel to the blade is preferable) and the

shape of the addressable region (rounded shapes are preferable). Approximately 5-10%

of the structures had shorted wires (i.e., the wires were touching). We minimized

delamination by using a thiol-containing polymer between the wires, but it still occurred

occasionally (in 5-10% of the samples).

185

Figure S4. Representative defects observed in the structures. (a) Imperfections on the

surface of the diamond knife of the microtome scored the slabs and induced defects in a

repeatable manner from section to section. (b) Occasionally, poor interfacial adhesion

between the materials in the sample block resulted in delamination at interfaces during

sectioning. (c) Portions of the metallic components aligned perpendicular to the cutting

motion of the blade tended to have breaks. (d) Regions of the parallel nanowires

occasionally touched, resulting in electrical shorts. This type of defect was the least

common of the ones listed.

186

Appendix IV

Fabrication of Surface Plasmon Resonators by Nanoskiving Single-Crystalline Gold

Microplates

Benjamin J. Wiley,1 Darren J. Lipomi,1 Jiming Bao,2 Federico Capasso,2

and George M. Whitesides1

1Department of Chemistry and Chemical Biology, Harvard University

12 Oxford St., Cambridge, Massachusetts, 02138 (USA)

2School of Engineering and Applied Sciences, Harvard University

33 Oxford St., Cambridge, Massachusetts, 02138 (USA)

Reproduced with permission from

Nano Lett. 2008, 8, 3023-3028

Copyright 2008, American Chemical Society

187

Fabrication of Surface PlasmonResonators by NanoskivingSingle-Crystalline Gold MicroplatesBenjamin J. Wiley,† Darren J. Lipomi,† Jiming Bao,‡ Federico Capasso,‡and George M. Whitesides*,†

Department of Chemistry and Chemical Biology, HarVard School of Engineering andApplied Sciences, HarVard UniVersity, 12 Oxford Street, Cambridge, Massachusetts 02138

Received July 25, 2008; Revised Manuscript Received August 4, 2008

ABSTRACT

This paper demonstrates the sectioning of chemically synthesized, single-crystalline microplates of gold with an ultramicrotome (nanoskiving)to produce single-crystalline nanowires; these nanowires act as low-loss surface plasmon resonators. This method produces collinearly alignednanostructures with small, regular changes in dimension with each consecutive cross-section: a single microplate thus can produce a numberof “quasi-copies” (delicately modulated variations) of a nanowire. The diamond knife cuts cleanly through microplates 35 µm in diameter and100 nm thick without bending the resulting nanowire and cuts through the sharp edges of a crystal without deformation to generate nanoscaletips. This paper compares the influence of sharp tips and blunt tips on the resonator modes in these nanowires.

Introduction. This paper describes the fabrication of single-crystalline gold nanowires by sectioning chemically synthe-sized single-crystalline microplates with an ultramicrotome(nanoskiving) and demonstrates that these wires can act assurface plasmon resonators. They are the first gold nanowiresto exhibit this property (silver nanowires have been shownby Ditlbacher et. al1 to be resonators); the single-crystallinegold nanowires fabricated here have much lower radiativeloss than polycrystalline nanowires. Nanowires produced bynanoskiving are collinearly aligned and have small, regularvariations in size and interwire distances from section tosection; they are thus exceptionally useful as subjects forspectroscopic studies involving comparisons of structuresdiffering in nanometer scale dimensions. By combiningchemical synthesis of single-crystalline metallic plates withnanoskiving, we demonstrate a new strategy for preparingsingle-crystalline nanostructures and for studying the cou-pling of light between these nanostructures. These studiessuggest new types of nanophotonic devices based on low-loss single-crystalline plasmonic waveguides.

Background. The miniaturization of optical waveguides,lenses, and resonators based on dielectric materials is limitedby diffraction to about half the wavelength of light in thematerial. One way to circumvent the limits of diffraction isto couple light into a surface plasmon mode confined to a

metal-dielectric interface.2 Strips of evaporated metal withsubwavelength (nanoscale) diameters are capable of guidinglight at the nanoscale, but these waveguides suffer from highloss (the propagation length of light in these structures isaround 2.5 µm).3,4 In chemically synthesized silver nanow-ires, this propagation length increases to 10 µm because oftheir smooth surfaces.1 The low loss of these silver nanowiresgives them the ability to act as surface plasmon resonators,a characteristic that polycrystalline metal nanowires do notpossess.1 A disadvantage of using silver nanowires for opticalapplications is that they oxidize after several days underambient conditions.5 The ability to prepare and use single-crystalline gold nanowires would circumvent this problem,but there have been no studies of the waveguiding andresonator properties of such structures.

Experimental Design. In nanoskiving, an ultramicrotomecuts a slab of a polymer block containing an embedded,vertically thin first structure and forms a laterally thin secondstructure.6 We have used nanoskiving to fabricate wires,rings, and periodic arrays of shapes that possessed a rangeof optical properties.7-10 All of these previous applicationshave started with physically deposited, polycrystalline thinfilms, which in turn generated polycrystalline nanostructures.To increase the quality of the nanostructures made bynanoskiving, we have developed a tandem technique com-prising two steps: (i) chemical synthesis of single-crystallinegold microplates, and (ii) sectioning of these plates perpen-dicular to their plane to yield single-crystalline gold nanow-ires (Figure 1 summarizes the procedure). Remarkably, the

* To whom correspondence should be addressed. E-mail: gwhitesides@gmwgroup.harvard.edu.

† Department of Chemistry and Chemical Biology, Harvard University.‡ Harvard School of Engineering and Applied Sciences, Harvard

University.

NANOLETTERS

2008Vol. 8, No. 93023-3028

10.1021/nl802252r CCC: $40.75 ! 2008 American Chemical SocietyPublished on Web 08/23/2008

188

sectioning leaves the crystalline order of the gold plateslargely or completely intact.

The gold microplates grew in a solution containingethylene glycol (EG), HAuCl4, and polyvinylpyrrolidone(PVP).11 We prepared a flat piece of epoxy (Epo-fix, ElectronMicroscopy Sciences) by puddle-casting the prepolymeragainst the flat surface of a Si/SiO2 wafer. After curing, wedetached the epoxy from the Si/SiO2 and deposited asuspension of the microplates on the flat side of the epoxysubstrate. We adjusted the concentration of plates empiricallyto give the desired coverage of the surface (approximately25%). After drying and rinsing to remove PVP, we embeddedthis substrate (epoxy and microplates) in additional epoxyand sectioned it with an ultramicrotome equipped with adiamond knife (Diatome Ultra 35).8 It was straightforwardto transfer the resulting thin (50 nm) epoxy slabs (e1 mm2)containing nanowires onto almost any substrate; we usedglass coverslips or Si/SiO2 wafers. Exposure to an oxygenplasma removed the embedding epoxy and left free-standingnanowires. Chemical synthesis determined the width (90-200nm) and the length (e40 µm) of the nanowires, while theultramicrotome determined the height (tunable from 10 to1000 nm).

Results and Discussion. Sectioning the embedded mi-croplates gave 50-nm-high gold nanowires embedded inepoxy; these wires were easily resolved using dark-field

optical microscopy (Figure 1A). The random spacing anddimensions (lengths, widths) of nanowires, together with thefact that there were few wires along the slab, greatlysimplified correlating scanning electron microscope (SEM)images with optical images of the same nanowires (Figure1B). Note that the nanowire on the far left of Figure 1Ascatters more red light, and that on the far right scatters moregreen light. The magnified views of nanowires in the insetsof Figure 1B (scale bar ) 1 µm) show that two nanowiresstuck together scatter red light (bottom left), and a nanowirealigned with a nanoparticle scatters green light (upper right).A defect in the diamond knife created a score mark (greenline) across the epoxy slab and through the nanowire on thefar right.

Morphology of Microplates and Nanowires. The single-crystalline gold microplates obtained from the reaction anddeposition step consisted of hexagonal, triangular, andribbon-like structures with diameters ranging from 1 to 40µm (Figure 2A). The thickness of these microstructures wasusually between 90 and 200 nm. The inset in Figure 2A isa magnified image of a representative microplate: it is 25µm in diameter and ca. 100 nm thick. This thicknesscorresponds to the width of the gold nanowire in Figure 2B;this wire was obtained by sectioning a microplate 35 µm indiameter. Although this microplate has a ratio of length tothickness of 350, the diamond knife cuts cleanly through itwithout bending the resulting nanowire. The inset gives amagnified view of a short rod along the side of the largerwire; the thin white line running parallel to the long wire isthe edge of the epoxy that has delaminated from the wire.This arrangement of paired rod and wire reflects thesectioning of a small plate on top of a larger plate.

Treating the samples with oxygen plasma removed theepoxy matrix and generated freestanding, gold nanowires.This action made it possible to examine the sides of thenanowires. The SEM image of a gold nanowire in Figure2C shows that the facet cut by the diamond knife (top of thewire) appears to be as smooth as the sides of the originalcrystal. The sharp tip at the end of the rod (inset) demon-strates that the diamond knife can cut through sharp edgesof microplates with minimal disturbance and generatenanoscale tips free of bends or deformation. The lack ofcontrast across a nanowire when viewed under the transmis-sion electron microscope (TEM) suggests it retained thesingle-crystalline structure of the original microplate (Figure2D). The inset is a magnified view of the edge of the plate,which reveals defect free (111) lattice fringes and a nearlyatomically smooth sidewall.

Figure 3 shows SEM images of gold nanostructures in five(of six; one slice is missing between B and C) consecutivesections to illustrate a unique aspect of nanoskiving. Theposition and dimensions of the large wire on the left remainedroughly the same because the microtome cut thin (e50 nm)sections of a crystal 8 µm in diameter. Successive slabsproduced controlled, regular, small variations in size for thesmaller structure on the right, which are probably sectionsfrom the edge of a smaller, triangular microplate. Themagnified images of the smaller structure (Figure 3F,G) again

Figure 1. Schematic drawing of the procedure used to fabricatesingle-crystalline nanowires. We deposited microplates from solu-tion onto a thin slab of epoxy and embedded a piece of this slab sothat the plates were entirely surrounded by epoxy. (A) Optical imageof a 100-nm-thick section in which aligned, single-crystalline goldnanowires are embedded. (B) SEM image of the same region shownin the optical image.

3024 Nano Lett., Vol. 8, No. 9, 2008

189

show that nanoskiving of crystals can result in remarkablysharp nanoscale tips. Crystals with sharp tips (similar toFigure 2C or Figure 3F,G) could be found in greater than90% of the slabs; roughly 20% of nanowires had sharp tips.

The spacing between the two structures varied by less than100 nm for all the slices.

Because each slab in nanoskiving can be registered withthe following and preceding slabs, this method allows thegeneration of a sequence of structures with highly regularslab-to-slab changes. Precise, end-to-end alignment of nano-wires deposited from solution is very difficult, but nanowiresproduced from nanoskiving are aligned collinearly every timewith any other nanocrystals codeposited on the flat epoxy.This method is, thus, ideal for generating aligned arrays ofnanostructures with spacings that are smoothly modulated(at least over several slabs), if presently random. Thesestructures are very attractive for research in plasmonicsbecause they allow detailed examination of dimensions andspacings at the scale of 10 nm.

Characterization of the Plasmonic Properties of theNanowire. The smooth sidewalls of the gold nanowiressuggested they might act as plasmon resonators. We testedthis hypothesis by illuminating nanowires with unpolarized,focused white light from a tungsten lamp through a glassprism under total internal reflection (Figure 4A). Epoxy slabscontaining nanowires were placed on a glass coverslip, etched

Figure 2. (A) SEM image of a substrate with a dense coverage of gold microplates to illustrate the variety of shapes and sizes producedby the synthesis. A slightly less dense, submonolayer coverage of microplates is best for producing isolated nanowires. The inset shows the100-nm-thick edge of a 25-µm-wide microplate. (B) SEM of a gold nanowire 35 µm long and 100 nm wide obtained by nanoskiving amicroplate. (C) SEM image of a gold nanowire with pointed tips obtained by nanoskiving. We removed the epoxy by etching with oxygenplasma to reveal the smooth sidewalls of the nanorod. (D) TEM image of a gold nanowire. The inset shows defect-free (111) lattice fringesat the edge of the nanowire. The fringes indicate that the nanowire retained the crystal structure of the original synthesized microplate.

Figure 3. (A-E) SEM images of successive slabs illustratenanoskiving can approximately reproduce the size and position ofa nanowire as well as produce small, regular variations in size fromsection to section. (F,G) Magnified views of two nanoparticles withsharp tips collinearly aligned with the nanowire.

Nano Lett., Vol. 8, No. 9, 2008 3025

190

with oxygen plasma to remove the epoxy, and opticallycoupled to the prism with index-matching silicone oil.

We oriented the nanowires to be parallel to the wave vectorof the evanescent wave generated at the interface betweenair and the glass coverslip. Although the nanowires areilluminated uniformly with a spot ca. 1 mm in diameter, lightcan couple into a propagating plasmon only at the end ofthe nanowire pointing toward the illumination source (definedas the input end in Figure 4A).1 Plasmons can scatter as lightat the ends of the wire; light does not radiate from the middleof the nanowire because the wave vector of a surface plasmonis greater than that of light in air. Significant roughness witha Fourier component corresponding to the momentummismatch between the surface plasmon and light in air isrequired to couple light out from the middle of the wire.We collected the light radiated from the ends of the nanowirewith a microscope objective (Mitutoyo SL50, numericalaperture (NA) ) 0.55). The collected light passed through apolarizer oriented parallel to the long axis of the nanowire,a 665 nm long-pass colored glass filter, and focused on theplane of the entrance slit of a single-grating monograph(Jobin Yvon Horiba Triax 550); this apparatus allowed usto observe the nanowires on a monitor and measure theirspectrum. We closed the entrance slits until only the lightemitting from one end of the nanowire could be observedon the monitor before taking spectra.

Surface plasmons can reflect off the ends of the nanowiremultiple times before radiating from an end as light; thisgeometry is similar to that of a Fabry-Perot etalon, i.e., adielectric slab in which light undergoes multiple reflections.The only modes within a lossless resonator that can perpetu-ate are those that reproduce themselves after a single roundtrip; all other waves undergo deconstructive interference. Thelight transmitted through a surface plasmon resonator (andthat reflected by it) consists of peaks separated by awavelength difference, ∆λ, given by eq 1, where λ is thefree space

∆λ)λ2νp

2dc(1)

wavelength, νp is the phase velocity of the plasmons, d isthe length of the nanowire, and c is the speed of light.

The spectra taken from the ends of a nanowire 5.1 µm inlength (SEM image in the inset) exhibited peaks with aperiodic spacing characteristic of Fabry-Perot resonatormodes (Figure 4B). The minima in the spectrum from theinput end, and maxima in the spectrum of from the outputend, correspond to the wavelengths at which maximumtransmission occurred due to constructive interference ofplasmons within the nanowire resonator.4

The spacing between the resonator modes is not constantbecause the phase velocity of the plasmons varies as afunction of wavelength (Figure 4C). We calculated the phase

Figure 4. (A) Illustration showing the coupling of light into gold nanowires with illumination by total internal reflection on a prism. Theinput and output tip faced toward and away from the light source, respectively. Light radiated only from the ends of the nanowire, and thislight was directed one end at a time into a spectrometer. (B) Spectra of the light reflected from the input and transmitted through the outputends of the single-crystalline nanowire shown in the inset SEM image. The wavelength-dependent periodic modulation of light and thecorrespondence between the reflection minima and the transmission maxima indicate the nanowire acted as a surface plasmon resonator.(C) The relative phase velocity of the surface plasmons decreased as the frequency increased on gold nanowires with three different lengths.Insets show the difference in the wavelength of the surface plasmon (induced surface charge) at the lowest and highest frequency. (D)Spectra of the light emitted from the ends of a polycrystalline wire indicate the surface plasmon was damped relative to the single-crystalline nanowire. The damping was due to the greater surface roughness of the polycrystalline wire, shown in the SEM image in theinset.

3026 Nano Lett., Vol. 8, No. 9, 2008

191

velocity as a function of wavelength with eq 1, using thewavelength difference between the peaks in the spectra fromthree nanowires of different lengths (4.4, 5.1, and 5.2 µm).The wavelength of the plasmon is approximately 283 nm at1.75 eV and 491 nm at 1.38 eV. The wavelength of aplasmon is shorter than the wavelength of light in vacuumat the same frequency, and the rate of change of thewavelength per unit frequency is less than that of light in avacuum. These results are similar to those obtained for silvernanowires.12 The insets of Figure 4C illustrate the changein the plasmon wavelength of about 60% across thisfrequency range.

To compare the morphology and optical properties ofsingle-crystalline nanowires produced by nanoskiving withthose of polycrystalline nanowires, we fabricated polycrys-talline nanowires by a three-step process: (i) patterning 5-µm-wide gold lines on an epoxy substrate by electron-beamevaporation and a photolithographic lift-off process, (ii)embedding the patterned film in additional epoxy, and (iii)sectioning it with the ultramicrotome. The surface of thepolycrystalline wire in Figure 4D appears much rougher thanthat of the single-crystalline nanowires. The polycrystallinewire also bent due to the inhomogeneous stress applied bythe microtome blade. The remarkable smoothness andstraightness of the single-crystalline wires indicates that theroughness and bending of polycrystalline wires is due to theirpolycrystallinity rather than the nanoskiving process itselfand that nanoskiving produces nanostructures with the samesmooth surfaces, straight edges, and single-crystalline struc-ture as synthetic methods when performed on a synthetic(single-crystalline) starting material.

The spectrum from the input end of the polycrystallinewire has some small, broad periodic peaks, and the outputspectrum displays no periodic peaks. These spectra indicatethe polycrystalline plasmon resonator has greater loss thanthe single-crystalline wire. The relative greater loss is dueto the scattering of surface plasmons from the rough surfaceof the polycrystalline wire; this emission is not allowed froma smooth surface due to the wavevector mismatch betweenthe surface plasmons and light in air.13

The radiation of the surface plasmons from rough surfacescan be understood as polarization currents normal to thesurface radiating as Hertzian dipoles.14 A quantitativedescription of the emitted light dI per solid angle elementdΩ and per incident power I0 from a rough gold film is givenby Kretschmann14 as eq 2,

dII0dΩ

) (πλ )4 4!ε

cos θ0|tp(θ0)|

2|W|2|sk-k0|2 (2)

where λ is the wavelength of light in vacuum, ε is thedielectric constant of the substrate (e.g., glass), θ0 is the angleof incident light from normal, tp describes the amplitude ofthe electric field at the metal-air interface, W describes theangular intensity of radiated light, and sk-k0 is the Fouriertransform of the roughness function. Equation 2 indicatesthat radiative loss should be greater at shorter wavelengths,and this relationship can be observed qualitatively in thespectra of Figure 4D.

The sharp tips at the ends of some nanowires present anopportunity to study how these tips affect the opticalproperties of the nanowires. Figure 5 shows spectra from ananowire 8 µm in length with one blunt end and one sharpend. If the blunt end served as the input (Figure 5A), thespectra from the blunt end and sharp end both exhibited peakswith a periodic frequency spacing, but the minima of theinput were offset from the maxima of the output, with agreater offset at shorter wavelengths. The asymmetry of thisnanowire resonator caused a mismatch between the maximaof the transmittance and minima of the reflectance spectra.

If the wire was rotated 180°, such that the sharp end servedas the input, the scattering spectra exhibited a broad peakcentered at 825 nm, with a superimposed Fabry-Perotinterference pattern. This result suggests that the sharp tipacted as a local plasmon resonator, or optical antennae,strongly scattering light at a frequency red-shifted from thesurface plasmon resonance frequency of bulk gold films(∼610 nm).15 Gold nanorods with an aspect ratio of ∼4.5

Figure 5. (A) Spectra from a nanowire with a blunt input and sharpoutput. The lack of correspondence between the minima of thereflection spectra (input) and the maxima of the transmission spectra(output) was caused by the asymmetry of the resonator. (B) Spectrafrom the same nanowire rotated by 180° shows that the sharp tipacted as an optical antennae, with a longitudinal resonancefrequency of ∼825 nm. (C) The correspondence between theminima from the blunt input and the maxima from the blunt outputconfirm that the sharp tip caused the mismatch in (A).

Nano Lett., Vol. 8, No. 9, 2008 3027

192

strongly scatter light at a similar frequency as the tip of thisnanowire due to longitudinal plasmon resonance.16 Theincrease in scattered light from the blunt output at 825 nmsuggests that the localized plasmon resonance of the sharp,nanoscale antennae increased the coupling of light into thepropagating plasmon mode along the nanowire.

To confirm that the sharp tip caused the mismatch betweenthe transmittance and reflectance spectra, we plotted thespectra from the blunt input with the spectra from the bluntoutput (Figure 5C). The spectra exhibited the expectedcorrespondence between the minima of the input and maximaof the output.

Conclusion. Nanoskiving of chemically synthesized, single-crystalline microplates of gold produces collinearly aligned,single-crystalline nanowires with very smooth surfaces (muchsmoother than those obtained by nanoskiving polycrystallinethin films). Successive sections from microplates producenanowires with smoothly modulated dimensions. Correlationof the optical properties of nanowires in each section (<1mm2) with SEM images of their structure is greatly simplifiedby the low number of nanowires in each section, theircollinear alignment, and their random but reproduciblespacing. The smooth surfaces of the gold nanowires producedby this process allow them to serve as surface plasmonresonators.

Nanoskiving of chemically synthesized, single-crystallinematerials makes possible the fabrication and study of newstructures for nanophotonics and nanomaterials research. Forexample, in addition to preparation of collinear nanowires,one might also deposit a thin dielectric layer between twolayers of microplates to make plasmonic slot waveguides.Nanoskiving of other two-dimensional (silver and palladiumnanoplates) and one-dimensional nanostructures (metal andsemiconductor nanowires) will, we believe, bring a new levelof control to their dimensions and relative positions and helpelucidate the effect of size and shape on their optical andelectronic properties.

Acknowledgment. We acknowledge support from theDefense Advanced Research Projects Agency (DARPA)under award no. HR0011-04-1-0032 and The CaliforniaInstitute of Technology. F.C. acknowledges support from theAir Force Office of Scientific Research (AFOSR) MURI onPlasmonics, and G.M.W. and F.C. both acknowledge supportfrom the Harvard Nanoscale Science and Engineering Center(NSEC). This work was performed in part at the Center forNanoscale Systems (CNS), a member of the NationalNanotechnology Infrastructure Network (NNIN), which issupported by the National Science Foundation under NSFaward no. ECS-0335765. CNS is part of the Faculty of Artsand Sciences at Harvard University.

Supporting Information Available: Experimental sec-tion. This material is available free of charge via the Internetat http://pubs.acs.org.

References(1) Ditlbacher, H.; Hohenau, A.; Wagner, D.; Kreibig, U.; Rogers, M.;

Hofer, F.; Aussenegg, F.; Krenn, J. Phys. ReV. Lett. 2005, 95, 257403.(2) Barnes, W. L.; Dereux, A.; Ebbesen, T. W. Nature 2003, 424, 824.(3) Yatsui, T.; Kourogi, M.; Ohtsu, M. Appl. Phys. Lett. 2001, 79, 4583.(4) Krenn, J. R.; Lamprecht, B.; Ditlbacher, H.; Schider, G.; Salerno, M.;

Leitner, A.; Aussenegg, F. R. Europhys. Lett. 2002, 60, 663.(5) Elechiguerra, J. L.; Larios-Lopez, L.; Liu, C.; Garcia-Gutierrez, D.;

Camacho-Bragado, A.; Yacaman, M. J. Chem. Mater. 2005, 17, 6042.(6) Xu, Q.; Rioux, R. M.; Whitesides, G. M. ACS Nano 2007, 1, 215.(7) Xu, Q.; Bao, J.; Capasso, F.; Whitesides, G. Angew. Chem., Int. Ed.

2006, 45, 3631.(8) Xu, Q.; Bao, J.; Rioux, R. M.; Perez-Castillejos, R.; Capasso, F.;

Whitesides, G. M. Nano Lett. 2007, 7, 2800.(9) Xu, Q.; Gates, B. D.; Whitesides, G. M. J. Am. Chem. Soc. 2004,

126, 1332.(10) Xu, Q.; Perez-Castillejos, R.; Li, Z.; Whitesides, G. M. Nano Lett.

2006, 6, 2163.(11) Kan, C.; Zhu, X.; Wang, G. J. Phys. Chem. B 2006, 110, 4651.(12) Allione, M.; Temnov, V. V.; Fedutik, Y.; Woggon, U.; Artemyev,

M. V. Nano Lett. 2008, 8, 31.(13) Raether, H., Surface Plasmons on Smooth and Rough Surfaces and

on Gratings; Springer-Verlag: Berlin, 1988.(14) Kretschmann, E. Opt. Commun. 1972, 5, 331.(15) Kapitza, H. Opt. Commun. 1976, 16, 73.(16) Huang, X.; El-Sayed, I. H.; Qian, W.; El-Sayed, M. A. J. Am. Chem.

Soc. 2006, 128, 2115.

NL802252R

3028 Nano Lett., Vol. 8, No. 9, 2008

193

Supporting Information:

Fabrication of Surface Plasmon Resonators by Nanoskiving Single-Crystalline Gold Microplates

Benjamin J. Wiley1, Darren J. Lipomi1, Jiming Bao2, Federico Capasso2, George M. Whitesides1,*

1. Department of Chemistry and Chemical Biology

2. Harvard School of Engineering and Applied Sciences

Harvard University, 12 Oxford St., Cambridge, MA 02138, U.S.A.

* Author to whom correspondence should be addressed

194

EXPERIMENTAL SECTION:

Materials

Epo-fix epoxy resin and hardener were purchased from Electron Microscopy Sciences.

HAuCl4 and polyvinylpyrrolidone (PVP, MW=55,000) were purchased from Aldrich.

Ethylene glycol (EG) was purchased from J.T. Baker.

Synthesis of Microplates

To grow microplates of gold, we heated 5 mL of EG in a disposable glass vial submerged

in an oil bath set to 160 °C for 1 hour before adding 0.83 ml of 0.2 M HAuCl4 in EG and

2.5 ml of a 222 mg/mL solution of PVP (MW=55,000) in EG. We heated this mixture

with stirring for about 1 hour before stopping the reaction, at which time the solution

contained a mixture of microplates and nanoparticles. We then immediately added the

reaction mixture to 10 mL of a 0.25 g/mL solution of PVP in ethanol. After mixing, we

layered this suspension of gold microplates and nanoparticles on top of 30 ml of a 0.25

g/mL aqueous solution of PVP. Over the course of 6 hrs, the larger microplates settled to

the bottom of the centrifuge tube while the nanoparticles stayed in the top ethanol phase.

We then discarded the supernatant and resuspended the microplates in a 0.25 g/mL

aqueous solution of PVP.

Procedure for Nanoskiving Microplates

195

The procedure for nanoskiving microplates is illustrated in Figure 1. We first deposited

them from solution onto a thin (l,w,h ~ 2 cm, 2 cm, 2 mm) slab of epoxy (Epo-fix,

obtained from Electron Microscopy Sciences). After drying, we immersed the slab with

microplates in deionized water to wash away the PVP, being careful not to wash the

microplates off the slab as well. After drying a second time, the slab with microplates

was cut into thin strips with a razor blade. These strips were placed in polyethlyene

molds, embedded in epoxy prepolymer and cured to form epoxy blocks. After trimming

away excess epoxy, these blocks were sectioned into 50-nm-thick slices. Readers are

referred to the supplementary information of our previous work for a detailed description

of microtome section alignment (8).

Microscopy

Dark-field optical microscopy was performed with a Leica DMRX upright optical

microscope. Epoxy slabs containing nanowires were placed on pieces of a Si wafer for

scanning electron microscopy, and on a carbon-coated copper grid for transmission

electron microscopy. SEM images were acquired with a Zeiss Supra55VP field emission

scanning electron microscope operated at 5 kV. For imaging of freestanding gold

nanowires, epoxy sections were etched by oxygen plasma (1 Torr, 70 W barrel etcher) for

40 min. TEM images were acquired with a Jeol JEM-2100 LaB6 transmission electron

microscope operated at 200kV.

196

Appendix V

Integrated Fabrication and Magnetic Positioning of Metallic and Polymeric Nanowires

Embedded in Thin Epoxy Slabs

Darren J. Lipomi,1 Filip Ilievski,1 Benjamin J. Wiley,1 Parag B. Deotare,2 Marko Lončar,2

and George M. Whitesides1

1Department of Chemistry and Chemical Biology, Harvard University

12 Oxford St., Cambridge, Massachusetts, 02138 (USA)

2School of Engineering and Applied Sciences, Harvard University

33 Oxford St., Cambridge, Massachusetts, 02138 (USA)

Reproduced with permission from

ACS Nano 2009, 3, 3315-3325

Copyright 2009, American Chemical Society

197

Integrated Fabrication and MagneticPositioning of Metallic and PolymericNanowires Embedded in Thin EpoxySlabsDarren J. Lipomi,† Filip Ilievski,† Benjamin J. Wiley,† Parag B. Deotare,‡ Marko Loncar,‡ andGeorge M. Whitesides†,*†Harvard University, Department of Chemistry and Chemical Biology, 12 Oxford Street, Cambridge, Massachusetts 02138, and ‡Harvard University, School of Engineeringand Applied Sciences, 33 Oxford Street, Cambridge, Massachusetts 02138

T his paper describes an integratedapproach to the fabrication and po-sitioning of nanowires embedded in

thin slabs of polymer. The procedure com-bines nanoskivingOa technique for the fab-rication of nanostructures that uses an ultra-microtome to cut thin slabs of nonmagneticand ferromagnetic materials embedded in apolymeric matrix1,2Owith noncontact, mag-netic manipulation of the polymeric slabscontaining ferromagnetic particles (Figure1 summarizes the process, which we havenamed “magnetic mooring”). Magneticmooring exploits an important characteris-tic of nanoskiving; that is, after sectioning,the nanostructures remain embedded inthin slabs of polymer, which float on thesurface of water. The user can collect theslabs by transferring them to a substrate,along with !5 "L of water. The water formsa pool on which the slabs float and acrosswhose surface they can be moved usingmagnetic interactions. As the water evapo-rates, capillary interactions cause the slabsto adhere to the substrate. In this work, wecoembedded Ni strips or powder with thenanowires in order to make the floatingslabs (along with the nanostructures theycontain) magnetically responsive to thefield created by a movable external mag-net. The accuracy of registration of the em-bedded nanostructures with predepositedobjects on a substrate was typically 5#25"m.

Magnetic mooring can form geometriesof nanostructures for electronic and photo-nic applications that would be difficult orimpossible to arrange using other tech-niques. We crossed nanowires of Au, Pd,

and conjugated polymers, and were ableto place individual single-crystalline Aunanowires on dielectric waveguides. Theprocedure is nonphotolithographic, and re-quires only methods for the deposition ofthin films, an ultramicrotome, a microscope,and a movable stage for positioning. It iscompatible with conventional methods oflithography for subsequent electrical or op-tical characterization. The process providesa way of fabricating, positioning, and inte-grating nanostructures with each other andwith instruments in the laboratory.

BACKGROUNDPositioning Nanoscale Objects. Nanoscience

and nanotechnology require methods toaddress and manipulate nanostructures

*Address correspondence togwhitesides@gmwgroup.harvard.edu.

Received for review August 13, 2009and accepted September 8, 2009.

Published online September 16,2009.10.1021/nn901002q CCC: $40.75

© 2009 American Chemical Society

ABSTRACT This paper describes a process for the fabrication and positioning of nanowires (of Au, Pd, and

conjugated polymers) embedded in thin epoxy slabs. The procedure has four steps: (i) coembedding a thin film

of metal or conducting polymer with a thin film of nickel metal (Ni) in epoxy; (ii) sectioning the embedded

structures into nanowires with an ultramicrotome (“nanoskiving”); (iii) floating the epoxy sections on a pool of

water; and (iv) positioning the sections with an external magnet to a desired location (“magnetic mooring”). As

the water evaporates, capillary interactions cause the sections to adhere to the substrate. Both the Ni and epoxy

can be etched to generate free-standing metallic nanowires. The average translational deviation in the positioning

of two nanowires with respect to each other is 16 ! 13 "m, and the average angular deviation is 3 ! 2°.

Successive depositions of nanowires yield the following structures of interest for electronic and photonic

applications: electrically continuous junctions of two Au nanowires, two Au nanowires spanned by a poly(3-

hexylthiophene) (P3HT) nanowire; single-crystalline Au nanowires that cross; crossbar arrays of Au nanowires;

crossbar arrays of Au and Pd nanowires; and a 50 # 50 array of poly(benzimidazobenzophenanthroline ladder)

(BBL) nanowires. Single-crystalline Au nanowires can be placed on glass wool fibers or on microfabricated

polymeric waveguides, with which the nanowire can be addressed optically.

KEYWORDS: nanoskiving · nanowires · nanofabrication ·conjugated polymers · microtome · nanophotonics · nanowire positioning ·magnetic positioning

ARTIC

LE

www.acsnano.org VOL. 3 NO. 10 3315–3325 2009 3315

198

individually and in groups. There are many strategiesfor fabricating nanostructures and assembling theminto useful geometries. Scanning-beam techniquessuch as electron-beam lithography (EBL) are well-established and can generate nearly arbitrary patternsin resist, which can be transferred to metallic thin filmsor other materials. Focused-ion-beam (FIB) milling cancarve patterns into materials directly and is used exten-sively in research laboratories for the fabrication of teststructures. These techniques, powerful as they are, havehigh costs, low throughput, limited accessibility to gen-eral users, and limited flexibility with respect to thetypes of substrates they can pattern directly. In this dis-cussion, we focus on structures fabricated through non-lithographic means, and then deposited on a substratein a desired geometry. This paper uses structures thatare one-dimensional3 (nanowires, nanorods, etc.), butthe process we describe would be applicable to otherstructures or arrays of structures as well.

Nonlithographically generated nanostructures areusually deposited by random assembly. Once depos-ited, nanostructures arranged by chance in a desiredgeometry can be addressed lithographically. Alterna-tively, nanostructures can be deposited on a substratealready bearing lithographically patterned features. Aserendipitously positioned nanowire (spanning twoelectrodes or sitting on an optical waveguide) can thensometimes be addressed. This method proceeds withlow yields and provides little control over the orienta-tion of nanostructures.

Nanoscience often requires the inter-actions of multiple structures in closeproximity in well-defined geometries,for example, specific structures compris-ing nanowires, quantum dots, electrodes,waveguides, and other structures infunctional forms.4 The more elements asystem has, however, the lower the prob-ability that a desired geometry can begenerated by random assembly. Systemsin which geometry is important, and ofcourse, any engineering application re-quire methods of fabricating multiele-ment structures that do not rely onchance.

Existing Methods for Positioning Nanowires.Existing methods for positioning nano-wires have one of two goals: (i) to aligna large number of nanowires over a largearea (cm2)5,6 or (ii) to position individualnanowires one-by-one.

Shear alignment of nanowires sus-pended in fluids7 is a common methodto fabricate useful geometries of nano-wires of several classes of materials.8!11

Lieber and co-workers have recently ex-tended shear alignment of nanowires to

the scale of cm2 by suspending and depositing themin bubble-blown films.12 Brushing suspensions ofnanowires over a lithographically patterned substratecreates highly aligned regions of nanowires on exposedareas of the substrate.13 Alignment of nanowires in aLangmuir!Blodgett trough can form highly anisotro-pic films that can be cast over large areas.14 In general,shear alignment is capable of manipulating the orienta-tion of many nanowires at once but with limited con-trol of the position of individual nanowires.

Optical tweezing can manipulate single semicon-ducting nanowires in a liquid environment with a hightheoretical accuracy.15 Opto-electronic tweezing is anew technique in which an optical signal creates anelectrical potential on a photoconductive layer on thebottom of a liquid-filled cell to yield groups of nano-wires aligned perpendicular to the substrate in arbitrarylocations.16 Optical methods of manipulation, in gen-eral, depend on the size, shape, and composition of thenanostructures, and on the presence of a fluid.

Methods of manipulation by direct contact withscanning probe tips17 and micromanipulators18 can pro-vide control over individual nanowires, but are depen-dent on the size and composition of the nanowires andthe topography of the substrate. Electrophoretic align-ment of nanowires over prepatterned electrodes hasthe potential for integration over large areas,19 but it re-quires an extensively processed substrate. The pitch ofthe nanowires, further, may only be as high as that ofthe spacing between electrodes. Templated elec-

Figure 1. Schematic representation of the procedure used for fabrication (I. nanoskiving)and positioning (II. magnetic mooring) of nanowires.

ART

ICLE

VOL. 3 NO. 10 LIPOMI ET AL. www.acsnano.org3316

199

trodeposition in lithographically defined trenches,20 orby self-assembled structures of block copolymers,21 al-low integrated fabrication and alignment of structures,but these methods are limited to materials grownelectrochemically.

Use of Magnetic Forces in Nanoscience. The use of magne-tism to manipulate nanostructures has usually beenlimited to structures that are themselves magnetic.22!25

For example, Hangarter et al. prepared Au and Binanowires capped with Ni and aligned them on mag-netic electrodes.26

Magnetism is generally regarded as nonharmful tobiological systems, at low fields. This characteristic hasinspired biomedical applications of magnetic nanopar-ticles that include attachment to biomolecules27 andmicrotubules,28 actuation of magnetic particles internal-ized by living cells to promote cell death,29 and the fab-rication of particles that could be localized using mag-netic fields for drug delivery.30

Combination of magnetic forces with self-assemblyprovides another method to position small structures.Yellen et al. demonstrated the self-assembly of non-magnetic particles when suspended in a medium con-taining magnetic particles arranged by a program-mable, magnetic substrate.31 In another example,Lapointe et al. suspended Ni nanowires in a bed of ne-matic liquid crystals, which was patterned into zones inwhich the molecular orientation (director) pointed indifferent directions. At equilibrium, the nanowires self-aligned with their long axes parallel to the director.When the authors reoriented nanowires with a mag-netic field, the nanowires migrated to a different regionof the pattern such that their long axes were again par-allel to the director.32

We reasoned that any nanowire, including those ofnonmagnetic materials, could be manipulated by an ex-ternal magnet if the wire could be tethered physicallyto a magnetic particle. For example, Shi et al. modifiedthe surfaces of glass fibers with Fe3O4 nanoparticles us-ing layer-by-layer assembly and observed the motionof the fiber on the surface of water in the presence ofa magnetic field.33

Nanoskiving. Nanoskiving is a technique based on sec-tioning thin structures (e.g., films or microplates) withan ultramicrotome.1 We have used this process to gen-erate nanowires from polymeric thin films formed byspin-coating,34,35 metallic thin films formed by physicalvapor deposition,36 and chemically grown, single-crystalline microplates.37 Our laboratory has previouslydescribed an approach to orient structures produced bynanoskiving, by stacking successive slabs manually (byhand, with an eyelash, or another tool).1

EXPERIMENTAL DESIGNOur goal was to develop a method to position poly-

meric slabs containing nanowires on flat or topographi-cally patterned substrates. At its core, this process re-

quired forming thin slabs of polymer that embeddedboth the nanowires and a sacrificial ferromagnetic ma-terial. These slabs, floating on the surface of a pool ofwater, would be mobile under the influence of an exter-nal permanent magnet. As the water evaporated, thepolymeric slabs would become docked (“moored”) tothe substrate in the position fixed by the user. Whilethere are potentially many ways of generating suchfilms, we chose to combine this technique with nano-skiving. Even though nanoskiving is capable of formingseveral types of nanostructures, all of which would beamenable to positioning by the process we are describ-ing, we focused on nanowires for the following rea-sons: (i) they are straightforward to fabricate by section-ing thin films or microplates, and thus make goodsystems with which to characterize this method; (ii)the lines they make in epoxy slabs make them easy tolocate using optical microscopy (which would facilitatepositioning) and (iii) they are important components ofnanoelectronic and nanophotonic devices.3

We chose Ni as the sacrificial ferromagnetic ma-terial to coembed with the nonmagnetic nanowires inthe epoxy matrix. While Ni has a lower magnetic perme-ability than other ferromagnetic materials (e.g., Fe), Niis mechanically softer. Softer materials are less likelythan harder materials to damage the diamond knifewe use in the ultramicrotome.

We reasoned that the best geometry of Ni particlewith which to embed the nanowires would be a longstrip for magnetostatic considerations.38 An externalmagnet would magnetically saturate the Ni strip in thelong direction. Two separate effects would account forthe translational and rotational positioning of the epoxyslabs in the magnetic field. The first effect, maximizingthe flux of the field through the Ni strip, would governthe initial capture and translation of the epoxy slabon the pool of water. An epoxy slab, mobile on the sur-face of the droplet of water, would move over the ex-ternal magnet until the Ni strip reached the region withthe highest flux. The second effect, the torque actingon the magnetic moment of the polarized Ni strip in theexternal field, would enable rotational alignment. Wechose the dimensions of the Ni strips (l " 102 #m andw " 2 #m) as a compromise between two differentgoals: (i) maximizing the absolute strength of interac-tion with an external magnetic field (which is propor-tional to the volume of the ferromagnetic material) and(ii) minimizing the thickness of the Ni film (a film thatis too thick can damage the diamond knife). The thick-ness of the epoxy slabs became the height of the Nistrips, which constrained that dimension to 100 nm forall experiments in this paper. An alternative method ofadding ferromagnetic particles to the epoxy slabs wasto mix Ni nanopowder (particle size " 200 nm) into theepoxy prepolymer (2% by mass Ni); this method wouldbe convenient but would sacrifice some control in posi-tioning slabs.

ARTIC

LE

www.acsnano.org VOL. 3 NO. 10 3315–3325 2009 3317

200

We designed the experimental apparatus for sim-plicity. It required a microscope and two stages: the up-per stage held the substrate, typically a Si wafer, sit-ting over a circular hole drilled into the stage with adiameter (d ! 1.5 in.) smaller than that of the Si wafer(d ! 2 in.); the lower stage had a movable platform,with three degrees of freedom (x, y, and "), which sup-ported the magnets (grade N42, NdFeB, cylindrical,d ! 0.125 in., l ! 0.375 in.) and enabled translationand rotation of the magnetic field. We assembled sixof these magnets into two parallel columns of threemagnets each (to increase the height of the magneticcolumn), with the magnetization of each column point-ing in the opposite direction. The top of the dual col-umn of magnets was positioned so that it was #0.5 mmfrom the bottom of the substrate. The opposite polar-ization of the top of each of the two magnets magne-tized the strips of Ni in the floating slabs and allowed formanipulation of the slabs by translation and rotationof the magnetic field (see the Supporting Informationfor photographs of the apparatus).

Fabrication. Figure 1 summarizes both phases of theprocedure: fabrication (nanoskiving) and positioning(magnetic mooring). We illustrated the process for thesimple case of generating crossed Au nanowires, but itis easily amenable to structures of any material that canbe fabricated by nanoskiving. We deposited two thinfilms, one of Ni (2 $m thick) and one of Au (80 nm thick),on flat epoxy substrates by electron-beam evaporation(step 1). Then, using a razor blade, we cut strips (#1 mm% 5 mm) of Ni and Au supported by their epoxy sub-strates (step 2). We embedded the strips together in ep-oxy prepolymer (step 3). We sectioned the blocks withthe ultramicrotome into epoxy slabs (step 4), which wetransferred to the substrate using a metallic loop thatsuspended the slab by surface tension in a thin film ofwater (the “Perfect Loop” tool, inner diameter ! 2 mm,obtained from Electron Microscopy Sciences; step 5).The substrates used in this work were typically test-grade Si wafers bearing a native layer of SiO2, cleanedwith an air plasma (500 mtorr, 100 W, 30 s). Optionally,selectively etching the epoxy in an air plasma (100 W,1 Torr, 15 min), and the Ni with a commercial etchant,left behind a free-standing nanowire on the surface(step 6). We transferred a second epoxy slab to the sub-strate by hand using the Perfect Loop. When touchedto the substrate, the loop released the slab along witha droplet of water of &5 $L, which spread into a pool&1 cm in diameter (step 7). We captured the floatingslab in the magnetic field of the column of magnetsmounted on the movable stage. By manipulating theposition of the magnets, we guided the second slabover the first nanowire in the desired orientation (across, step 8). As the water evaporated, capillary forcescaused the slab to adhere to the substrate (step 9). Theepoxy and Ni could be etched, as before; this actionyielded free-standing, crossing nanowires (step 10).

RESULTS AND DISCUSSIONThin Epoxy Slabs Containing Nanowires and Ni Particles. Fig-

ure 1 shows the preferred method of incorporating Niparticles into epoxy slabs containing nanowires by co-embedding evaporated Ni films. Sectioning 2-$m-thickfilms of Ni formed strips in the epoxy slabs that en-abled manipulation by an external magnet. (Section-ing thinner Ni films (100 nm) provided extensively frac-tured wires that did not have sufficient volume forcapture and manipulation of the slabs by the externalmagnet.) The alternative method was to make a “mag-netic epoxy” by mixing Ni powder into the epoxy pre-polymer. Either method of incorporating Ni (strips orpowder) into the epoxy provided 100'500 pg of Ni ina typical slab with dimensions l ( w ( 500 $m, h ! 100nm. Figure 2 shows optical micrographs of typical ep-oxy slabs containing nanowires and Ni particles. Thefeatures stand out most clearly under dark field.

Magnetic Manipulation of Thin Epoxy Slabs. We were ableto manipulate slabs containing Ni along with metallicand polymeric nanowires with maximum velocities of&100 $m s'1. It took about 2 min to position each float-ing slab on the surface of the water. The 5-$L pool ofwater typically evaporated in 10 min if the stage washeated gently to &35 °C with a halogen lamp placed 10cm away. We found that it was more difficult to rotateslabs containing Ni powder than it was to rotate thosecontaining Ni strips; as we rotated the column of per-manent magnets, the epoxy slabs bearing Ni powderslipped unpredictably from one equilibrium position toanother, though some control was possible. We at-tribute the difference in the ability to control the posi-tion of slabs containing Ni strips and those containingNi powder to differences in shape of the embeddedmagnetic particles. The Ni powder consists roughly ofspheres, which are weakly interacting with each otherand can be magnetized easily in any direction. In con-trast, the anisotropic shape of a Ni strip forces the mag-netization along the long axis and makes it more diffi-cult to demagnetize than Ni powder by a movingmagnetic field. Consequently, the slabs embeddedwith powder are more difficult to capture and controlwith an applied field than those containing strips.

Electrical Continuity of Two Crossing Au Nanowires. The firstgeometry of nanowires that we demonstrated com-prised two Au nanowires that crossed at 90°; we thencharacterized the electrical conductivity of the wiresand the junction. On a Si wafer bearing &300 nm ofthermally grown SiO2, we moored two 100-nm-thick ep-oxy slabs, which contained one Au nanowire each (l (200 $m, w ! 80 nm, h ! 100 nm). After positioning, theslabs were heated to 125 °C for &15 min in an oven toimprove adhesion of the epoxy to the substrate. Next,we defined contact pads using a hand-cut conformalstencil mask made of a poly(dimethylsiloxane) (PDMS)membrane (100 $m thick), through which we depos-ited 50 nm of Au by electron-beam evaporation. Before

ART

ICLE

VOL. 3 NO. 10 LIPOMI ET AL. www.acsnano.org3318

201

characterization, the epoxy matrix was removed by atreatment in an air plasma. While this procedure can beperformed at any point, we left the epoxy matrix untilthe end so that it would provide structural integrityagainst the conformal PDMS mask, which might other-wise damage free-standing nanowires. We measuredthe electrical characteristics of the crossing nanowiresafter removal of the epoxy matrix. We were able to ap-ply up to !10 mV without failure of the nanowires;when we increased the bias beyond !10 mV, the cir-cuit failed, presumably by melting at a thin area of oneof the wires (the crossing region itself remained intact,as determined by scanning electron microscopy). Fig-ure 3a shows the plot of current density vs voltage(J"V). In the calculation of current density, we assumeda uniform rectangular cross section of the wires of 80nm # 100 nm, which spanned a length of 100 $m, asdetermined by SEM. The current density was 0.4 of thetheoretical maximum based on the conductivity of bulkAu. We attribute the lower effective conductivity tononuniform cross sections of the wires, a nonconfor-mal junction between them, the graininess of the par-ent thin films, and possible contamination of organicmaterial between the nanowires.

Au Nanowire Electrodes Spanned by a Conjugated PolymerNanowire. To demonstrate the ability to mix differenttypes of materials, we generated a system that spannedtwo Au nanowires with a poly(3-hexylthiophene) (P3HT)nanowire. This geometry could be useful in measuringnanoscale charge transport in optoelectronic polymersand in the fabrication of chemical sensors39 or field-effect transistors based on single nanowires.40 Poly(3-hexylthiophene) (which we synthesized using estab-lished methods41) undergoes an insulator-to-metaltransition upon exposure to I2.42 We began by deposit-ing two parallel Au nanowires, which were embeddedin the same epoxy slab (thickness % 100 nm). We re-moved 10"20 nm of the epoxy surrounding free Aunanowires by brief etching with an air plasma. We chosenot to etch the epoxy completely because occasion-ally, immersion in water during the mooring step dis-lodged Au nanowires that were unsupported by epoxyslabs (Au adheres poorly to SiO2). We fabricated a P3HTnanowire (100 nm # 100 nm cross section), coembed-ded with Ni powder in epoxy, and moored it in a posi-tion that spanned the 50-$m gap between Au nano-wires. We did not remove the epoxy matrix surroundingthe P3HT wire, since to do so would have destroyedthe P3HT. We deposited contact pads though a stencilmask as described previously. In the absence of I2, thecurrent of the nanowires at !1 V was too low to be de-tected by our electrometer (Keithley 6430 Femtoamme-ter). When we placed an I2 crystal &1 mm away fromthe P3HT nanowire (uncovered), the conductivity of theP3HT increased within 10 s (the time it took to acquirean I"V plot). Figure 3b shows the conductivity of thenanowire when exposed to I2 (“doped”) and the con-

ductivity after removal of the I2 (“undoped”). We didnot try to achieve the maximum doping level reportedfor P3HT.42 It should be possible to fabricate arrange-ments of nanowire electrodes for four-terminal mea-surements; this geometry would allow decoupling ofthe contact resistance from the true resistance of ananowire.

Accuracy of Positioning. To determine the accuracy withwhich we could position and orient nanostructures ontop of one another, we formed crosses of single-crystalline Au nanowires (which we obtained bynanoskiving chemically synthesized Au microplates37),with the goal of superimposing the center of eachnanowire with a crossing angle of 90°. We used Ni stripsas the sacrificial ferromagnetic material for this experi-ment. Figure 4a is an SEM image of two crossingnanowires, which have a center-to-center distance (de-viation) of 2.2 $m. The average center-to-center dis-

Figure 2. Optical micrographs of epoxy slabs containing nanostruc-tures and sacrificial Ni particles. All slabs are 100 nm thick. (a) An ep-oxy slab containing an Au NW and strips of Ni; the features stand outin the dark-field image shown in panel b. The Ni strips look damagedbecause of buckling of the Ni film, possibly due to thermal expansionand contraction of the epoxy substrate during the process of evapora-tion. (c) An epoxy slab containing two parallel Au nanowires coembed-ded with Ni nanopowder. The dark-field image (d) shows light scatter-ing off of the nanowires and the powder. The upper Au nanowireappears thicker because the epoxy matrix is delaminated from theAu (the interface scatters light strongly). A defect in the cutting edgeof the diamond knife created a line of scoring parallel to the directionof cutting, which stands out under dark field. (e) Two crossing epoxyslabs positioned by mooring. (f) Dark-field image of five parallel Aunanowires crossing five parallel Pd nanowires (the spacing is too smallto resolve the Pd nanowires individually). The Au and the Pd nano-wires were deposited in single epoxy slabs containing five nanowireseach.

ARTIC

LE

www.acsnano.org VOL. 3 NO. 10 3315–3325 2009 3319

202

tance over 10 attempted crossings (N ! 10) was 16 "13 #m, and the average angular deviation (from per-pendicular) was 3 " 2°. The Supporting Informationcontains images of all 10 attempted crossings in thedata set. The mooring process is subject to human er-ror; the registration is performed by eye under 95$magnification and the slabs are manipulated using ananalog micromanipulator. There are many possibleways to improve the accuracy. For example, using longnanowires (%100 #m) and an eyepiece containing asquare grid, it was possible to reduce the average angu-lar deviation to 0.7 " 0.5° (N ! 4). We estimated thatthe amplitude of vibrations visible in the microscopewas 1&5 #m (due to noise, air currents, and other vibra-tions in the room). The best realizable accuracy de-pends on the interplay between the floating slab andthe receding edge of the drop of water (the interfacebetween the drop of water and the dry substrate). Fora clean surface (e.g., a Si wafer or glass slide cleanedwith a brief exposure to an air plasma), the water will re-cede toward the floating slab smoothly as the waterevaporates. As the edge of the drop advances towardthe floating slab, the slab tends to travel toward theedge, as if sliding downhill. This perturbation is typi-cally '10 #m and can be corrected by the user. For acontaminated surface (one that has been left open tothe ambient air for a day or more), the drop edge doesnot recede smoothly. Rather, episodes of abrupt dewet-ting of the substrate at the drop edge create vibra-tions that can displace the slab away from its equilib-

rium position by 10 #m or more. The addition ofsurfactants increases the wettability of the substrateby the water, but leaves residue upon evaporation. Theuse of solvents with low surface tension in addition toor instead of water could make the process amenable tohydrophobic substrates without contaminating thesurface.

Stacking more than two slabs should be possible aswell. The only caution is that a predeposited epoxy slabdewets with a different rate than does the SiO2 sub-strate. We found that abrupt dewetting of the drop overepoxy slabs attached to the substrate often displacedthe floating slab; we obtained the most accurate resultsby ensuring that the retreating drop edge intersectedthe floating slab over the SiO2, rather than over a fixedslab. Any amount of overhang (of the floating slababove the fixed slab) was sufficient to bypass the ef-fect of rapid dewetting over slabs attached to the sur-face. We have not observed that the magnetic field ofthe predeposited Ni particles interferes with the moor-ing process.

Crossbars. Crossbar arrays of long nanowires withspacing between nanowires approximately equal tothe widths of the nanowires can be made by nanoskiv-ing and magnetic mooring. We fabricated crossbars ofAu (Figure 4b), of Au and Pd (Figure 4c), and a 50 $ 50square array of the conjugated polymer poly(benzimi-dazobenzophenanthroline) ladder (BBL, Figure 4d). Ineach case, the vertical nanowires were fabricated andmoored as a group over the horizontal nanowires. We

Figure 3. Images of junctions of single nanowires (NWs) fabricated by nanoskiving and magnetic mooring, and demonstra-tion of electrical continuity. (a) Schematic illustration and scanning electron micrograph (SEM) of perpendicular Au nano-wires and a plot of current density vs voltage (J!V) that demonstrates conductivity through the junction. The SEMs in pan-els a and b were obtained before the deposition of contact pads. (b) Illustration and SEM of parallel Au nanowires separatedby 50 "m and spanned by a P3HT nanowire embedded in an epoxy slab. The epoxy was not etched because the air plasmawould have destroyed the P3HT. The inset is a close-up of the junction (the Au nanowire is out of focus because the epoxyslab that contains the P3HT nanowire obstructs the electron beam; a particle of dust and a hazy vertical artifact of the im-age are labeled). The plot of current vs voltage (I!V) demonstrates that the P3HT increased in conductivity upon exposureto I2 (“doped”). Upon removal of the I2, the conductivity of the P3HT nanowire decreased (“undoped”).

ART

ICLE

VOL. 3 NO. 10 LIPOMI ET AL. www.acsnano.org3320

203

used Ni powder as the sacrificial ferromagnetic ma-terial in Figure 4b!d.

Mooring Nanowires onto Topographic Features. We wereable to moor nanowires on top of topographic fea-tures on a substrate, as long as the pool of water sub-merged the features during the process of positioning.We deposited a 15-"m-long, single-crystalline Au nano-wire on a single fiber of glass wool (Figure 5a), as wellas an 8-"m-long nanowire on a 1-cm-long, 10-"m-wide, 2-"m-high, waveguide fabricated by EBL in SU-8,negative-tone resist (Figure 5b).

Scattering from a Nanowire on a Waveguide. To show thatlight could be coupled into the nanowire by the eva-nescent field near the surface of a polymer wave-guide, we used an optical fiber to couple light intothe waveguide and observed scattering from the ter-mini of the nanowire. Figure 5c is a schematic draw-ing of the arrangement between the nanowire,waveguide, and optical fiber. The micrograph is ofthe nanowire on top of the waveguide, illuminatedby an external halogen lamp. Figure 5d shows scat-tering from the ends of the nanowires when lightfrom the optical fiber coupled into the waveguide.The effect is similar to that observed by Pyayt et al.,who found scattering from randomly deposited Ag

nanowires coupled to SU-8 waveguides.43 The easeof integrating these single-crystalline nanowireswith microfabricated dielectric waveguides, alongwith the observation of scattering from the ends ofthe nanowires, suggests that the wires could be usedas plasmonic waveguides in a nanophotonicdevice.44

CONCLUSIONSThe combination of nanoskiving and magnetic

mooring is useful for the assembly of nanostructuresfor simple, multicomponent electronic or optical de-vices. The technique is complementary to existing tech-niques for positioning and orienting nanostructures.Optical tweezing, for example, has control over indi-vidual structures with high accuracy, but is unable tocontrol groups of structures and relies on a liquid me-dium. Methods of fluid-assisted alignment can alignnanowires over large areas, but do not provide controlover individual structures. Integrated fabrication andpositioning is not provided by any other nonlitho-graphic method.

Magnetic mooring is not limited to structuresthat can be produced by nanoskiving; any delicatefilm that contains ferromagnetic particles could be

Figure 4. Crossbar structures formed by nanoskiving and magnetic mooring. (a) SEM image of crossing single-crystallineAu nanowires. The vertical nanowire has dimensions of l ! 10 "m, w ! 300 nm, h ! 100 nm; the horizontal nanowire has di-mensions of l ! 8 "m, w ! 290 nm, h ! 100 nm. The centers of the nanowires are separated by 2.2"m, and the long axes de-viate from perpendicular by 2.6°. The inset was obtained at a tilt of 45°. (b) An array of Au nanowires with dimensions of in-dividual nanowires of w ! 80 nm and h ! 100 nm. (c) An array of Pd nanowires (w ! 60 nm, h ! 80 nm) crossed by Aunanowires (w ! 80 nm, h ! 100 nm). (d) A 50 # 50 square array of poly(benzimidazobenzophenanthroline ladder) (BBL)nanowires. The pitch is 200 nm. We attribute the defects in which the nanowires diverge, touch, or both to delamination ofthe thin films from which the nanowires are made, during the sectioning process.

ARTIC

LE

www.acsnano.org VOL. 3 NO. 10 3315–3325 2009 3321

204

positioned. Nanoskiving is, however, a convenientway of making such films. Any of the structures thathave been fabricated using nanoskiving could bepositioned and oriented by magnetic mooring. Thepresence of the epoxy slab preserves the spatial re-lationship between structures within each slab, acharacteristic that could lead to more complex ar-rangements of nanostructures than those demon-strated. Stacking of 2D arrays of plasmonic resona-tors, for example, could provide a new route towardthe fabrication of 3D metamaterials.1

Crossbar structures of metallic nanowires with a sub-micrometer pitch could be valuable for a variety of ap-plications, including memory devices, tunnel diodes,and optical antennae.45 Square arrays of conjugatedpolymer nanowires could be useful as high-surface-areaorganic semiconductors in heterojunction photodetec-tors, for example.46 We believe that our process is par-ticularly useful for assembling structures for nanopho-tonic applications. We showed that it was possible toposition Au nanowires on glass fibers and photoresistfeatures; these structures can serve as opticalwaveguides for photonic circuits, while single-crystalline metallic nanowires can serve as sub-! plas-monic waveguides. Arrangements of these compo-nents in arbitrary geometries could enable the fabrica-tion of nanophotonic devices comprising metallicnanowires,47 semiconducting nanowires,48 opticalwaveguides,43 and single-photon emitters4 or the fabri-cation of apertureless near-field optical probes.44

The most important limitations of magnetic moor-ing are the average positional deviation (16 "m) andthe serial nature of depositing polymeric slabs one-by-one. We do not believe that we have achieved the high-est accuracy possible for this technique. A combina-tion of (i) increasing the strength of interaction betweenthe slabs and the external magnets for faster correc-tion of the errors caused by abrupt dewetting of thesubstrate as the drop edge approaches the floatingslab, (ii) controlling the wettability of the substrates orthe surface tension of the pool of liquid, or (iii) design-ing topographic features that dock the slabs before thereceding edge of the drop of water influences their po-sitions could increase the accuracy of positioning. Anideal apparatus would include dark-field optics, a cam-era, and a closed-loop system with a piezoelectronicallycontrolled manipulator.15 In the long term, magnetic in-teractions between thin polymeric slabs and externalmagnets (or between the slabs themselves) might beamenable to programmed or templated self-assembly.

METHODSFabrication of Au Nanowires Coembedded with Ni Strips (Figures 1 and

2a!e). We began by puddle casting an epoxy prepolymer (Epo-Fix, obtained from Electron Microscopy Sciences) against a test-grade Si wafer bearing no surface treatment. We generally useda ring of PDMS to contain the epoxy prepolymer. Thermal curingof the epoxy at 60 °C for 2 h, cooling to room temperature (rt),and separation of the cured epoxy from the Si template provideda smooth epoxy surface (rms roughness # 0.5 nm by atomicforce microscopy). We coated this epoxy substrate with an 80-nm-thick film of Au by e-beam evaporation at a rate of 1$5 Å s$1.We coated a second epoxy substrate with a 2-"m-thick film ofNi by e-beam evaporation at a rate of %10 Å s$1. This film dis-played buckling that did not adversely affect subsequent stepsof the process. We cut the Au and Ni films on their epoxy sub-strates into strips (l & 5 mm, w & 300 "m) using a razor bladeand a hammer. We placed Au and Ni strips face-to-face and em-bedded them in more epoxy (the two films were separated by

epoxy, not air). After curing, we placed this roughly embeddedstructure into a 1-mL polyethylene centrifuge tube and embed-ded it in additional epoxy. This action provided a block thatcould be secured in the ultramicrotome for sectioning. We ex-posed the cross section of the Au and Ni films using a jeweler’ssaw and trimmed the block facet into a rectangle with sides of200$500 "m. We used an ultramicrotome (Leica Ultracut UCT)equipped with a 35° diamond knife (Diatome Ultra 35 with 1.8 or2.4 mm length cutting edge) set to a clearance angle of 6°. Allsections were collected at ambient temperature on the surfaceof deionized water (see the Supporting Information of Xu et al.49

for a detailed description of the operation of the ultramicro-tome). After sectioning, the slabs floated on the surface of awater-filled trough, were collected by hand using the PefectLoop tool (Electron Microscopy Sciences), and were transferredto the substrate. The droplet of water spread into a pool with adiameter of %1 cm. At this point, we either placed the substratein the apparatus for magnetic mooring or allowed the water toevaporate. Following deposition, optional etches of the epoxy

Figure 5. Optical micrographs of single-crystalline nanowires posi-tioned on top of topographic features. (a) A nanowire (l " 15 #m, w" 150 nm, h " 100 nm) lying on the side of a fiber of glass wool. The in-set is a magnified view of the same image. (b) A nanowire (l " 8 #m,w " 290 nm, h " 100 nm) placed on top of a polymeric opticalwaveguide microfabricated in SU-8 photoresist. The inset is an SEMimage of the nanowire on top of the waveguide. (c) Schematic illustra-tion and optical micrograph of the nanowire from panel b on themicrofabricated waveguide. The image was obtained under external il-lumination from a halogen light source. (d) Schematic illustration andoptical micrograph of the nanowire obtained by coupling light intothe waveguide using an optical fiber. The only light visible was thatwhich scattered from the ends of the nanowire.

ART

ICLE

VOL. 3 NO. 10 LIPOMI ET AL. www.acsnano.org3322

205

(SPI Plasma Prep II benchtop etcher, 100 W, 1 Torr ambient air,15 min) and/or the Ni (Nickel Etchant, type TFB, Transene Com-pany, Inc., etch rate 3 nm s!1 at 25 °C) generated free-standingnanowires.

Fabrication of Parallel Au Nanowires (Figures 2c!f, 3b, and 4b,c). Wegenerated multiple parallel nanowires of Au in the same epoxyslab by sectioning parallel films of Au. We laminated multiple Aufilms together by evaporating an Au film on a flat epoxy sub-strate, applying a drop of epoxy prepolymer to the film, and cur-ing it under compression by a flat slab of PDMS. The PDMS slabwas pressed into the epoxy prepolymer by gravity or with abinder clip. After thermal curing of the epoxy, the substratecould again be coated with Au. This process could be repeatedto form several parallel films. We cut this film into strips, embed-ded it with a Ni film or powder, trimmed the block, and sec-tioned it, as before.

Fabrication of Parallel Pd Nanowires (Figures 2f and 4c). We formedepoxy blocks containing multiple parallel films of Pd separatedby thin epoxy layers by iterative stripping of an evaporated filmof Pd (60 nm) off of a Si wafer. We placed a drop of epoxy pre-polymer onto the surface of the Pd film. We placed a cured pieceof epoxy of the same type against the drop of prepolymer, ap-plied pressure with two binder clips, and thermally cured the ep-oxy in an oven at 200 °C for 15 min. After cooling with a fan for5 min, release of the epoxy support transferred a region of the Pdfilm to the epoxy substrate. We extruded a second drop of ep-oxy on the wafer and again placed the support in contact withit, under pressure. Repetition of these steps provided a lami-nated structure with five layers of Pd separated by four layersof epoxy. We cut the laminated structure into strips with a razorblade, embedded a strip in additional epoxy prepolymer to forma block, and sectioned the block with the ultramicrotome. Etch-ing in an air plasma removed the epoxy matrix and liberated fivefree-standing nanowires. The spacing between the wires wascontrolled by the pressure on the epoxy support and the viscos-ity of the epoxy prepolymer (initial viscosity " 0.05 Pa s). Wewere able to achieve a minimum spacing between nanowiresof 70 nm.

Fabrication of P3HT Nanowires (Figure 3b). We synthesized regio-regular P3HT using the McCullough method of polymeriza-tion.42 We dissolved this material in chloroform at a concentra-tion of 17 mg mL!1 and spin-coated it on a flat epoxy substrateat 1 krpm. Thermal annealing and removal of the solvent at 125°C in a vacuum oven for 30 min produced a red film with a me-tallic luster. We cut the film into strips, embedded it with a Ni filmor powder, trimmed the block, and sectioned it, as before.

Fabrication of Single-Crystalline Au Nanowires by Nanoskiving ChemicallySynthesized Microplates. We described the fabrication of the single-crystalline nanowires in a previous report from our laboratory.37

Briefly, we generated single crystalline microplates by heating asolution of HAuCl4 in the presence of poly(vinylpyrrolidone) inethylene glycol. We deposited microplates grown by thismethod onto flat epoxy substrates. We cut the substrates bear-ing Au microplates into strips, embedded it with a Ni film or pow-der, trimmed the block, and sectioned it, as before.

Fabrication of BBL Nanowires (Figure 4d). We fabricated groups of50 parallel BBL nanowires using a previously published proce-dure.34 Briefly, 100 total layers of BBL and a sacrificial polymerwere spin-coated onto a glass slide, embedded in epoxy, andsectioned with the ultramicrotome. Etching with an air plasmaremoved the epoxy matrix and the sacrificial polymer. This pro-cess generated 50 parallel, free-standing BBL nanowires.

Apparatus for Magnetic Mooring. The apparatus for the processof positioning nanostructures embedded in thin films was con-structed on a floating (antivibration) optical table (see Support-ing Information for photographs of the apparatus). We used astereomicroscope under 95# magnification, mounted on aboom stand, to monitor all positioning. Beneath the objectivewe mounted two stages to the optical table. All optical equip-ment was obtained from ThorLabs. The lower stage wasequipped with the micromanipulators used for translation (PT1translation stages for x and y) and rotation (PR01 high precisionrotation mount for $) of the magnets, which were two parallelcolumns of three cylindrical permanent magnets (d " 0.125 in.,l " 0.375 in., grade N42, NdFeB) with the polarization of each col-

umn pointing in opposite directions. The upper stage was an alu-minum slab (MB6 aluminum breadboard, 6 in. # 6 in. # 0.5 in.)containing a circular hole (d " 1.5 in.) on top of which the typi-cal substrate, a Si wafer (d " 2 or 3 in.), sat. The upper stage wasbrought down toward the lower stage such that the column ofmagnets sat %0.5 mm from the bottom of the Si wafer, throughthe hole in the upper stage. The maximum values of magneticfield in the transverse direction were 2.2 kG and !2.2 kG, abovethe centers of the magnets in the plane of the substrate, as de-termined by a Gauss meter. The substrate and upper stage wereheated to 36 °C to quicken the evaporation of water using a halo-gen lamp placed &10 cm from the stage. The substrate was cov-ered by a Petri dish cover to protect the floating slabs from dis-turbances by air currents in the room. Holes drilled into the Petridish cover with a heated syringe needle allowed water vapor toescape.

Imaging. Optical imaging (Figures 2 and 5) was performed us-ing an upright optical microscope (Leica DMRX). Scanning elec-tron microscope (SEM) images (Figures 3, 4, and 5b) of the epoxysections were acquired with a Zeiss Ultra55 or Supra55 VP field-emission SEM at 5 kV with a working distance of 2!6 mm.

Coupling of Light into Au Nanowires Using Polymeric Waveguides (Figure5b!d). We fabricated waveguides of SU-8 2002 negative resiston a Si wafer bearing 3 'm of SiO2 using e-beam lithography at100 kV (Elionix 7000). We moored a single-crystalline nanowireon top of the polymeric waveguide; we did not etch the epoxymatrix nor the Ni strip as they did not interfere with the observa-tion of scattering from the termini of the nanowires. Light froma supercontinuum source (Koheras) was coupled to thewaveguide using a tapered lensed fiber (Nanonics Inc.).

Acknowledgment. This work was supported by the NationalScience Foundation under award CHE-0518055. The authorsused the shared facilities supported by the NSF under NSEC(PHY-0117795 and PHY-0646094) and MRSEC (DMR-0213805and DMR-0820484). This work was performed in part using thefacilities of the Center for Nanoscale Systems (CNS), a member ofthe National Nanotechnology Infrastructure Network (NNIN),which is supported by the National Science Foundation underNSF Award No. ECS-0335765. CNS is part of the Faculty of Artsand Sciences at Harvard University. The authors thank W. Reusfor help in preparing the Pd nanowires. D.J.L. acknowledges agraduate fellowship from the American Chemical Society, Divi-sion of Organic Chemistry, sponsored by Novartis.

Supporting Information Available: Photographs of the appara-tus used for magnetic micromooring and images of nanowiresfrom which the positional and rotational accuracy of the tech-nique was calculated. This material is available free of charge viathe Internet at http://pubs.acs.org.

REFERENCES AND NOTES1. Xu, Q.; Rioux, R. M.; Whitesides, G. M. Fabrication of

Complex Metallic Nanostructures by Nanoskiving. ACSNano 2007, 1, 215–227.

2. Xu, Q. B.; Rioux, R. M.; Dickey, M. D.; Whitesides, G. M.Nanoskiving: A New Method To Produce Arrays ofNanostructures. Acc. Chem. Res. 2008, 41, 1566–1577.

3. Xia, Y. N.; Yang, P. D.; Sun, Y. G.; Wu, Y. Y.; Mayers, B.; Gates,B.; Yin, Y. D.; Kim, F.; Yan, Y. Q. One-DimensionalNanostructures: Synthesis, Characterization, andApplications. Adv. Mater. 2003, 15, 353–389.

4. Akimov, A. V.; Mukherjee, A.; Yu, C. L.; Chang, D. E.; Zibrov,A. S.; Hemmer, P. R.; Park, H.; Lukin, M. D. Generation ofSingle Optical Plasmons in Metallic Nanowires Coupled toQuantum Dots. Nature 2007, 450, 402–406.

5. Cao, Q.; Rogers, J. A. Ultrathin Films of Single-WalledCarbon Nanotubes for Electronics and Sensors: A Reviewof Fundamental and Applied Aspects. Adv. Mater. 2009,21, 29–53.

6. Heo, K.; Kim, C. J.; Jo, M. H.; Hong, S. Massive Integration ofInorganic Nanowire-Based Structures on Solid Substratesfor Device Applications. J. Mater. Chem. 2009, 19, 901–908.

ARTIC

LE

www.acsnano.org VOL. 3 NO. 10 3315–3325 2009 3323

206

7. Stover, C. A.; Koch, D. L.; Cohen, C. Observations of FiberOrientation in Simple Shear-Flow of SemidiluteSuspensions. J. Fluid Mech. 1992, 238, 277–296.

8. Zhong, Z. H.; Wang, D. L.; Cui, Y.; Bockrath, M. W.;Lieber, C. M. Nanowire Crossbar Arrays as AddressDecoders for Integrated Nanosystems. Science 2003,302, 1377–1379.

9. Cui, Y.; Lieber, C. M. Functional Nanoscale ElectronicDevices Assembled Using Silicon Nanowire BuildingBlocks. Science 2001, 291, 851–853.

10. Huang, Y.; Duan, X. F.; Wei, Q. Q.; Lieber, C. M. DirectedAssembly of One-Dimensional Nanostructures intoFunctional Networks. Science 2001, 291, 630–633.

11. Messer, B.; Song, J. H.; Yang, P. D. Microchannel Networksfor Nanowire Patterning. J. Am. Chem. Soc. 2000, 122,10232–10233.

12. Yu, G. H.; Cao, A. Y.; Lieber, C. M. Large-Area Blown BubbleFilms of Aligned Nanowires and Carbon Nanotubes. Nat.Nanotechnol. 2007, 2, 372–377.

13. Fan, Z. Y.; Ho, J. C.; Jacobson, Z. A.; Yerushalmi, R.; Alley,R. L.; Razavi, H.; Javey, A. Wafer-Scale Assembly of HighlyOrdered Semiconductor Nanowire Arrays by ContactPrinting. Nano Lett. 2008, 8, 20–25.

14. Jin, S.; Whang, D. M.; McAlpine, M. C.; Friedman, R. S.; Wu,Y.; Lieber, C. M. Scalable Interconnection and Integrationof Nanowire Devices without Registration. Nano Lett.2004, 4, 915–919.

15. Pauzauskie, P. J.; Radenovic, A.; Trepagnier, E.; Shroff, H.;Yang, P. D.; Liphardt, J. Optical Trapping and Integration ofSemiconductor Nanowire Assemblies in Water. Nat. Mater.2006, 5, 97–101.

16. Jamshidi, A.; Pauzauskie, P. J.; Schuck, P. J.; Ohta, A. T.;Chiou, P. Y.; Chou, J.; Yang, P. D.; Wu, M. C. DynamicManipulation and Separation of Individual Semi-conducting and Metallic Nanowires. Nat. Photonics 2008,2, 85–89.

17. Postma, H. W. C.; Sellmeijer, A.; Dekker, C. Manipulationand Imaging of Individual Single-Walled CarbonNanotubes with an Atomic Force Microscope. Adv. Mater.2000, 12, 1299–1302.

18. Sirbuly, D. J.; Law, M.; Pauzauskie, P.; Yan, H. Q.; Maslov,A. V.; Knutsen, K.; Ning, C. Z.; Saykally, R. J.; Yang, P. D.Optical Routing and Sensing with NanowireAssemblies. Proc. Natl. Acad. Sci. U.S.A. 2005, 102,7800–7805.

19. Smith, P. A.; Nordquist, C. D.; Jackson, T. N.; Mayer, T. S.;Martin, B. R.; Mbindyo, J.; Mallouk, T. E. Electric-FieldAssisted Assembly and Alignment of Metallic Nanowires.Appl. Phys. Lett. 2000, 77, 1399–1401.

20. Yun, M. H.; Myung, N. V.; Vasquez, R. P.; Lee, C. S.; Menke,E.; Penner, R. M. Electrochemically Grown Wires forIndividually Addressable Sensor Arrays. Nano Lett. 2004, 4,419–422.

21. Chai, J.; Buriak, J. M. Using Cylindrical Domains of BlockCopolymers to Self-Assemble and Align MetallicNanowires. ACS Nano 2008, 2, 489–501.

22. Tanase, M.; Silevitch, D. M.; Hultgren, A.; Bauer, L. A.;Searson, P. C.; Meyer, G. J.; Reich, D. H. Magnetic Trappingand Self-Assembly of Multicomponent Nanowires. J. Appl.Phys. 2002, 91, 8549–8551.

23. Tanase, M.; Bauer, L. A.; Hultgren, A.; Silevitch, D. M.; Sun,L.; Reich, D. H.; Searson, P. C.; Meyer, G. J. MagneticAlignment of Fluorescent Nanowires. Nano Lett. 2001, 1,155–158.

24. Love, J. C.; Urbach, A. R.; Prentiss, M. G.; Whitesides, G. M.Three-Dimensional Self-Assembly of Metallic Rods withSubmicron Diameters Using Magnetic Interactions. J. Am.Chem. Soc. 2003, 125, 12696–12697.

25. Zhang, S.; Zhu, H. Y.; Hu, Z. B.; Liu, L.; Chen, S. F.; Yu, S. H.Multifunctional Necklace-Like Cu@Cross-Linked Poly(vinylalcohol) Microcables with Fluorescent Property and TheirManipulation by an External Magnet. Chem. Commun.2009, 2326–2328.

26. Hangarter, C. M.; Myung, N. V. Magnetic Alignment ofNanowires. Chem. Mater. 2005, 17, 1320–1324.

27. Fu, A. H.; Hu, W.; Xu, L.; Wilson, R. J.; Yu, H.; Osterfeld, S. J.;Gambhir, S. S.; Wang, S. X. Protein-FunctionalizedSynthetic Antiferromagnetic Nanoparticles forBiomolecule Detection and Magnetic Manipulation.Angew. Chem., Int. Ed. 2009, 48, 1620–1624.

28. Platt, M.; Muthukrishnan, G.; Hancock, W. O.; Williams, M. E.Millimeter Scale Alignment of Magnetic NanoparticleFunctionalized Microtubules in Magnetic Fields. J. Am.Chem. Soc. 2005, 127, 15686–15687.

29. Fung, A. O.; Kapadia, V.; Pierstorff, E.; Ho, D.; Chen, Y.Induction of Cell Death by Magnetic Actuation of NickelNanowires Internalized by Fibroblasts. J. Phys. Chem. C2008, 112, 15085–15088.

30. Liong, M.; Lu, J.; Kovochich, M.; Xia, T.; Ruehm, S. G.; Nel,A. E.; Tamanoi, F.; Zink, J. I. Multifunctional InorganicNanoparticles for Imaging, Targeting, and Drug Delivery.ACS Nano 2008, 2, 889–896.

31. Yellen, B. B.; Hovorka, O.; Friedman, G. Arranging Matter byMagnetic Nanoparticle Assemblers. Proc. Natl. Acad. Sci.U.S.A. 2005, 102, 8860–8864.

32. Lapointe, C. P.; Reich, D. H.; Leheny, R. L. Manipulationand Organization of Ferromagnetic Nanowires byPatterned Nematic Liquid Crystals. Langmuir 2008, 24,11175–11181.

33. Shi, F.; Liu, S. H.; Gao, H. T.; Ding, N.; Dong, L. J.; Tremel, W.;Knoll, W. Magnetic-Field-Induced Locomotion of GlassFibers on Water Surfaces: Towards the Understanding ofHow Much Force One Magnetic Nanoparticle Can Deliver.Adv. Mater. 2009, 21, 1927–1930.

34. Lipomi, D. J.; Chiechi, R. C.; Dickey, M. D.; Whitesides, G. M.Fabrication of Conjugated Polymer Nanowires by EdgeLithography. Nano Lett. 2008, 8, 2100–2105.

35. Lipomi, D. J.; Chiechi, R. C.; Reus, W. F.; Whitesides, G. M.Laterally Ordered Bulk Heterojunction of ConjugatedPolymers: Nanoskiving a Jelly Roll. Adv. Funct. Mater. 2008,18, 3469–3477.

36. Xu, Q. B.; Gates, B. D.; Whitesides, G. M. Fabrication ofMetal Structures with Nanometer-Scale LateralDimensions by Sectioning Using a Microtome. J. Am.Chem. Soc. 2004, 126, 1332–1333.

37. Wiley, B. J.; Lipomi, D. J.; Bao, J. M.; Capasso, F.; Whitesides,G. M. Fabrication of Surface Plasmon Resonators byNanoskiving Single-Crystalline Gold Microplates. Nano Lett.2008, 8, 3023–3028.

38. O’Handley, R. C. Modern Magnetic Materials: Principles andApplications; Wiley: New York, 2000; pp 42!43.

39. Liu, H. Q.; Kameoka, J.; Czaplewski, D. A.; Craighead, H. G.Polymeric Nanowire Chemical Sensor. Nano Lett. 2004, 4,671–675.

40. Liu, H. Q.; Reccius, C. H.; Craighead, H. G. SingleElectrospun Regioregular Poly(3-hexylthiophene)Nanofiber Field-Effect Transistor. Appl. Phys. Lett. 2005, 87,253106-1–253106-3.

41. Loewe, R. S.; Ewbank, P. C.; Liu, J. S.; Zhai, L.; McCullough,R. D. Regioregular, Head-to-Tail Coupled Poly(3-alkylthiophenes) Made Easy by the GRIM Method:Investigation of the Reaction and the Origin ofRegioselectivity. Macromolecules 2001, 34, 4324–4333.

42. McCullough, R. D. The Chemistry of ConductingPolythiophenes. Adv. Mater. 1998, 10, 93–116.

43. Pyayt, A. L.; Wiley, B.; Xia, Y. N.; Chen, A.; Dalton, L.Integration of Photonic and Silver Nanowire PlasmonicWaveguides. Nat. Nanotechnol. 2008, 3, 660–665.

44. Prasad, P. N. Nanophotonics; John Wiley & Sons, Inc.:Hoboken, NJ, 2005; pp 62!65.

45. Berland, B. Photovoltaic Technologies Beyond the Horizon:Optical Rectenna Solar Cell; Subcontractor Report forNational Renewable Energy Laboratory, 2003.

46. Yang, F.; Shtein, M.; Forrest, S. R. Controlled Growth of aMolecular Bulk Heterojunction Photovoltaic Cell. Nat.Mater. 2005, 4, 37–41.

47. Ditlbacher, H.; Hohenau, A.; Wagner, D.; Kreibig, U.; Rogers,M.; Hofer, F.; Aussenegg, F. R.; Krenn, J. R. Silver Nanowiresas Surface Plasmon Resonators. Phys. Rev. Lett. 2005, 95,257403-1–257403-4.

ART

ICLE

VOL. 3 NO. 10 LIPOMI ET AL. www.acsnano.org3324

207

48. Falk, A. L.; Koppens, F. H. L.; Yu, C. L.; Kang, K.; Snapp, N. D.;Akimov, A. V.; Jo, M. H.; Lukin, M. D.; Park, H. Near-FieldElectrical Detection of Optical Plasmons and SinglePlasmon Sources. Nat. Phys. 2009, 5, 475–479.

49. Xu, Q. B.; Bao, J. M.; Rioux, R. M.; Perez-Castillejos, R.;Capasso, F.; Whitesides, G. M. Fabrication of Large-AreaPatterned Nanostructures for Optical Applications byNanoskiving. Nano Lett. 2007, 7, 2800–2805.

ARTIC

LE

www.acsnano.org VOL. 3 NO. 10 3315–3325 2009 3325

208

1

Supporting Information

Integrated Fabrication and Magnetic Positioning of Metallic and Polymeric Nanowires

Embedded in Thin Epoxy Slabs

Darren J. Lipomi1, Filip Ilievski1, Benjamin J. Wiley1, Parag B. Deotare2, Marko Lon!ar2, and

George M. Whitesides1*

1Harvard University, Department of Chemistry and Chemical Biology

12 Oxford Street, Cambridge, MA 02138

2Harvard University, School of Engineering and Applied Sciences,

33 Oxford Street, Cambridge, MA 02138 *Corresponding Author

Telephone Number: (617) 495-9430

Fax Number: (617) 495-9857

Email Address: gwhitesides@gmwgroup.harvard.edu

209

2

Figure S1. Photograph of the apparatus used for magnetic mooring.

210

Appendix VI

Fabrication and Replication of Arrays of Single- or Multicomponent

Nanostructures by Replica Molding and Mechanical Sectioning

Darren J. Lipomi,1 Mikhail A. Kats,2 Philseok Kim,1,2 Sung H. Kang,2 Joanna

Aizenberg,1,2 Federico Capasso,2 and George M. Whitesides1

1Department of Chemistry and Chemical Biology, Harvard University

12 Oxford St., Cambridge, Massachusetts, 02138 (USA)

2School of Engineering and Applied Sciences, Harvard University

33 Oxford St., Cambridge, Massachusetts, 02138 (USA)

Reproduced with permission from

ACS Nano 2010, ASAP Article

Copyright 2010, American Chemical Society

211

Fabrication and Replication of Arrays ofSingle- or MulticomponentNanostructures by Replica Molding andMechanical SectioningDarren J. Lipomi,† Mikhail A. Kats,‡,§ Philseok Kim,†,‡,§ Sung H. Kang,‡ Joanna Aizenberg,†,‡

Federico Capasso,‡ and George M. Whitesides†,*†Department of Chemistry and Chemical Biology and ‡School of Engineering and Applied Sciences, Harvard University, 29 Oxford Street,Cambridge, Massachusetts 02138. §These authors contributed equally to this work.

T he interaction of electromagneticfields with metallic nanostructurescan produce collective oscillations of

the conduction electrons atmetal!dielectric interfaces. These oscilla-tions are known as localized surface plas-mon resonances (LSPRs) and propagatingsurface plasmon polaritons (SPPs) and areresponsible for most of the phenomena inthe field of plasmonics.1 Two- and three-dimensional arrays of metallic nanostruc-tures could play key roles in functional ma-terials. Optical filters,2,3 substrates for opticaldetection of chemical and biological ana-lytes using LSPRs4 or surface-enhanced Ra-man scattering (SERS),5!7 substrates for en-hanced luminosity,8 materials to augmentabsorption in thin-film photovoltaicdevices,9,10 metamaterials11,12 with negativemagnetic permeabilities13 and negative re-fractive indices,14 and surfaces for perfectlenses15 and invisibility cloaking16 are ex-amples of possible and realized applica-tions. There are, however, significant techni-cal challenges in generating arbitrarypatterns of metallic, dielectric, and semicon-ducting nanostructures for research andfor potential commercial devices. The mostsophisticated patterns are fabricated usingelectron-beam lithography (EBL),17 focused-ion beam (FIB) milling and lithography,14 ordirect laser writing.18 These techniques cangenerate nearly arbitrary patterns in resists(e.g., EBL) or hard materials (e.g., FIB milling),but they are serial, expensive, and requireaccess to a cleanroom.

A number of techniques have emergedthat have begun to address the challengesassociated with conventional scanning-

beam lithographic tools.19 Xia and co-workers recently reviewed synthetic meth-ods that can produce large quantities ofhigh-quality metallic structures for plas-monic applications.20 These materials, how-ever, are difficult to arrange in the orderedarrays that are required for many applica-tions in optics.21 Van Duyne and co-workershave developed an approach to generateordered arrays of metallic particles called“nanosphere lithography”, which uses amonolayer of colloidal crystals as a stencilmask; the triangular voids direct the deposi-tion of metal on the substrate by evapora-tion.22 Giessen and co-workers used thesame voids as apertures through which toproduce split-ring resonators by rotatingthe substrate at an angle during evapora-tion.23 The laboratories of Rogers, Odom,Nuzzo, and others have used soft litho-graphic processes to fabricate large-areapatterns of metallic nanostructures.

*Address correspondence togwhitesides@gmwgroup.harvard.edu.

Received for review May 6, 2010and accepted May 26, 2010.

10.1021/nn100993t

© XXXX American Chemical Society

ABSTRACT This paper describes the fabrication of arrays of nanostructures (rings, crescents, counterfacing

split rings, cylinders, coaxial cylinders, and other structures) by a four-step process: (i) molding an array of epoxy

posts by soft lithography, (ii) depositing thin films on the posts, (iii) embedding the posts in epoxy, and (iv)

sectioning in a plane parallel to the plane defined by the array of posts, into slabs, with an ultramicrotome

(“nanoskiving”). This work demonstrates the combination of four capabilities: (i) formation of structures that are

submicrometer in all dimensions; (ii) fabrication of 3D structures, and arrays of structures, with gradients of height;

(iii) patterning of arrays containing two or more materials, including metals, semiconductors, oxides, and

polymers; and (iv) generation of as many as 60 consecutive slabs bearing contiguous arrays of nanostructures.

These arrays can be transferred to different substrates, and arrays of gold rings exhibit plasmonic resonances in

the range of wavelengths spanning 2!5 "m.

KEYWORDS: nanoskiving · nanofabrication · plasmonics · metamaterials · softlithography · ultramicrotomy

ARTIC

LE

www.acsnano.org VOL. XXX NO. XX 000–000 XXXX A

212

These processes include phase-shifting edge lithogra-phy24 and molding UV-curable imprint resists withtransparent poly(dimethylsiloxane) (PDMS) stamps25 toform arrays of plasmonic nanostructures, such as arraysof apertures in metallic films26 or pyramidal shells.27

One particular challenge among methods of fabrica-tion is the ability to generate patterns comprising twoor more materials (such as metals, semiconductors, anddielectrics) in the same plane. Multicomponent pat-terns would be useful for building metamaterials andother optical devices. Tserkezis et al. produced arrays ofsandwich-like, metal!dielectric!metal nanostructures,which exhibited negative magnetic permeabilitieswhen exposed to visible and near-IR radiation,28 whileSu et al. used the same type of structures as substratesfor efficient SERS.29 Engheta has recently described op-tical nanocircuits inspired by metamaterials, in whichpatterns of mixed metallic and dielectric structures be-have as nanoinductors and nanocapacitors, in closeanalogy to microelectronic systems.30 Top-down pat-terning of vertical stacks of alternating layers of materi-

als has been the principal method used to fabricatemultimaterial optical nanostructures. The use offocused-electron-beam (FEB) deposition, or FIB milling,to produce patterns comprising two or more materialsin the same plane is possible, in principle,31 but it hasnot been developed.

Nanoskiving is a process whose key step is thin sec-tioning with an ultramicrotome.32 When combined withsoft lithography and thin-film deposition, it introduces“cutting” as a method of replicating patterns that iscomplementary to the established techniques of print-ing and molding.33 It converts the perimeters of moldedrelief features into the geometries of the nanostruc-tures (see Figure 1). The procedure used for the two-dimensional patterning required to create the three-dimensional master determines the geometry of thefeatures; the thicknesses of the thin films determine theline widths of the features; and the ultramicrotome de-termines the height of the features (30!2000 nm). Inprevious work, Xu et al. fabricated arrays of open andclosed-loop structures of gold using a process that com-bined photolithography, replica molding, and nano-skiving.34 These arrays served as mid-IR, frequency-selective surfaces. The structures had relatively largeouter diameters (2 "m) and were sectioned from anembedded structure with relatively shallow relief (#2"m); this topography did not allow a large number ofcross sections to be made. The ability to generate alarge number of copies ($20) of structures that are sub-micrometer in all dimensions requires reusable, high-aspect-ratio masters of the type that can be made byEBL followed by deep reactive-ion etching (DRIE).35

Soft lithographic molding has improved dramati-cally in the past few years, in terms of both the abso-lute sizes and aspect ratios of molded features that itcan generate. For example, (PDMS), molded over acrack in a silicon wafer, replicated a step height of 0.4nm.36 We have shown earlier that replica molding canbe used to transfer an array of high-aspect-ratio nano-posts from silicon to epoxy, through a PDMS intermedi-ate.35 Here we show that, in combination with nano-skiving, these high-aspect-ratio structures provide abasis for the fabrication of arrays of significantly smallerand more complex structures, and in greater numbersof replicas, than has been possible previously.

Mastering and Molding. We obtained a silicon masterbearing an array of cylindrical posts by EBL followedby Bosch DRIE.37 The total area of the array was !1 cm2.We replicated these arrays of cylindrical posts in a UV-curable epoxy using a PDMS mold as an intermediate.35

Materials. Many methods of fabrication focus onstructures of gold because it has useful plasmonic andelectronic properties, it does not oxidize under ambientconditions, and it is easily deposited by evaporation.19

To demonstrate that our process could be used withmaterials in addition to gold, we formed nanostructuresof silver, silicon, palladium, platinum, silicon dioxide,

Figure 1. Summary of the procedure used to fabricate concentric ringsby thin-film deposition and thin sectioning of high-aspect-rationanoposts.

ART

ICLE

VOL. XXX NO. XX LIPOMI ET AL. www.acsnano.orgB

213

the conducting organic polymer poly(pyrrole) (PPy),and films of lead sulfide (PbS) nanocrystals. We alsodemonstrated the ability to fabricate structures of twoor more materials in the same array.

Deposition of Thin Films. We deposited conformal filmsof metal on the nanoposts using a benchtop sputter-coater. In many cases, it was desirable to leave regionsof the nanoposts unmetalized, for example, when mak-ing split rings. In these situations, we used electron-beam evaporation. Evaporation produces a collimatedbeam of metal atoms, which metalizes only the sidewalls of topographic features in its path. We depositedPPy films by electrodeposition and films of PbS nano-crystals by drop-casting.

Nanoskiving. To produce two-dimensional arrays, wesectioned most blocks into slabs !1 mm2 in area and80"150 nm thick, under ambient conditions. To pro-duce quasi-three-dimensional arrays, very high-aspect-ratio structures, and wedge-shaped slabs, we cut slabs

up to 2 #m thick. We used a 35° diamond knife, whichsections materials with less compression than does a(more common) 45° knife.38,39 We used a UV-curable ep-oxy (UVO-114, obtained from Epoxy Technology Inc.)because of its strong adhesion to metals, and resistanceto compression during sectioning.

Measurement and Simulation. We transferred our nano-structures onto chemically vapor-deposited ZnSe,which is a common substrate for infrared applications.It has a wide band of transmission spanning wave-lengths from 500 nm to 20 #m (which includes the re-gion in which we expected the nanostructures to reso-nate) and a refractive index around 2.4 at near- tomid-IR frequencies. We characterized some of thestructures produced by this method using Fouriertransform infrared (FTIR) spectroscopy in transmis-sion mode. We compared the experimental spectrato those calculated by the finite-difference time-domain (FDTD) method, a standard technique

Figure 2. Scanning electron microscope (SEM) images (a,b,d!f) and optical images (c) of arrays of structures, before and af-ter nanoskiving. (a) Array of nanoposts after conformal coating with gold (Au), polypyrrole (PPy), and gold. (b) Array of mi-croposts bearing corrugated side walls. (c) A !1 mm2 array of Au microrings. The interference-like pattern across the array isdue to the corrugated side walls of the master used. The diameters of the rings are!10% larger in the bright regions thanthey are in the dark regions, as determined by SEM. The inset is a high-magnification image of the features. (d) Two-dimensional array of single Au rings. (e) Array of concentric Au rings separated by a layer of electrochemically grown PPy.(f) Array of concentric Au rings after etching the organic components.

ARTIC

LE

www.acsnano.org VOL. XXX NO. XX 000–000 XXXX C

214

for solving Maxwell’s equations in the timedomain.40,41

RESULTS AND DISCUSSIONFabrication of Nanopost Arrays. We created two types of

arrays of epoxy posts. The first (“nanoposts”) was asquare array with d ! 250 nm, h " 8 #m, and pitch "2 #m. The array covered 1 cm2 (though we would onlybe able to cut slabs with sides !2.4 mm because thatwas the length of our diamond knife). The second mas-ter (“microposts”) had features with d ! 1 #m, h " 9#m, and pitch " 3 #m. The array covered 8 cm2. Wefound that the silicon masters could be used indefi-nitely to prepare PDMS molds, which, in turn, could pro-duce multiple epoxy replicas.

Fabrication of Arrays of Metallic Nanostructures by SectioningNanoposts Bearing Metallic Films. Figure 1 summarizes theprocedure used to generate arrays of nanoposts coatedwith two gold films, separated by a film of conductingpolymer. We began by sputter-coating an array of nan-oposts with gold (step 1). This film served as the work-

ing electrode for the conformal electrodeposition ofPPy (step 2). A second deposition of gold provided anarray of four-component, coaxial nanoposts (step 3). Weembedded this structure in additional epoxy to form ablock (step 4). Sectioning this block yielded epoxy slabscontaining the nanostructures (step 5). The slabs couldbe transferred from the water bath on which theyfloated to a wide variety of substrates (not shown).42

Treatment with an air plasma removed both the epoxymatrix and the PPy between the gold rings (step 6). Fig-ure 2a,b shows arrays of nanoposts and microposts be-fore sectioning. The microposts had scalloped sidewalls; this topography is a trait of the Bosch DRIE pro-cess used to produce the silicon master. The corrugatedside walls (scalloping) had a period of !400 nm andan amplitude of !100 nm. We were able to smooth theside walls of the posts and to create posts with thinnedor thickened diameters, without fabricating a new sili-con master, by etching or coating an epoxy replica andreusing it as a master (see Supporting Information). Fig-ure 2c shows a dark-field optical image of an array of

Figure 3. SEM images of two-dimensional (2D) arrays of gold nanostructures. (a) Gold crescents. (b) Gold split rings. (c)Concentric split rings. The array contains a mixture of the two structures shown in the insets. (d) High-aspect-ratio concen-tric rings (image obtained at 45°). (e) Very high-aspect-ratio coaxial cylinders of gold.

ART

ICLE

VOL. XXX NO. XX LIPOMI ET AL. www.acsnano.orgD

215

gold rings covering an area of !1 mm2. Figure 2dshows a scanning electron microscope (SEM) image ofan array of gold rings.

Fabrication of Concentric Rings of Gold. We obtained con-centric rings of gold separated by PPy by following theprocedure summarized in Figure 1. Figure 2e is an im-age of the array of these discs, which are embedded inan epoxy matrix. Figure 2f is an image of concentricrings separated by empty space, created by etching theorganic components of the structures shown in Figure2e. To determine the yield of the process, we obtainedan array by sectioning a 140 nm thick slab of thesestructures over an area of 1 mm2. We chose a randomregion 2000 "m2 in area (which contained 486 concen-tric rings) and counted the defective structures by SEM.The yield of unbroken structures was 485 (99.8%). Bro-ken structures had gaps in the rings. Ten structures (2%)appeared to have a metal fragment bridging the twogold rings (see Supporting Information, and Figure S6,for details).

Number of Sections. To determine the number of sec-tions we could obtain from a single block of nano-posts, we cut through an 8 "m tall, 1 mm2 in area, ar-ray of nanoposts in 100 nm increments (the maximumnumber of perfect slabs would be 80). We counted 27consecutive slabs in which at least 90% of the area de-fined by the slab was covered by a contiguous array ofnanostructures, and 60 slabs (including the first 27) inwhich at least 50% of the area was covered, as esti-mated by optical microscopy. We attribute the loss inyield to the imperfect, manual alignment of the embed-

ded nanoposts with the diamond knife. Fiduciary mark-ers within the epoxy block or an automated alignmentsystem would increase the yield of replications.

Fabrication of Crescents and Split Rings. Using a modifica-tion of the procedure shown in Figure 1, it was pos-sible to obtain arrays of gold crescents and split rings.We began with an array of epoxy nanoposts, but in-stead of coating the nanoposts conformally, we depos-ited only partially around the circumferences of thenanoposts by shadow evaporation. Placement of thesubstrate at a 45° angle from the source of evapora-tion afforded the array of crescents shown in Figure 3a.Iterative evaporation and rotation of the substrate inthe evaporator produced the array of split rings shownin Figure 3b.

Fabrication of Counterfacing Concentric Split Rings. Counter-facing, concentric split-ring resonators have been pre-dicted to have negative effective magnetic permeabili-ties and are thus a possible component of negative-index metamaterials.13 Shadow evaporation androtation of the nanopost array within the evaporatorproduced a gold film around the nanoposts in the formof split cylinders. Electrodeposition of PPy, whichbridged the opening and controlled the spacing be-tween the two counterfacing split rings, followed by an-other metallization oriented 180° to the first, formed asecond split cylinder enclosing the first. Nanoskivingthis array produced counterfacing split rings (Figure 3c).The array shown is a mixture of the two types of struc-tures shown in the insets: counterfacing split rings andstructures in which the inner ring is closed.

Figure 4. SEM images of 2D arrays of multicomponent nanostructures. All structures are 80!100 nm in height. (a)Crescents of platinum. (b) Counterfacing crescents of silver and silicon. (c) Crescents of gold and palladium sepa-rated by a layer of silicon dioxide. (d) Microrings composed of PbS nanocrystals, which are continuous around the cir-cumference of the rings.

ARTIC

LE

www.acsnano.org VOL. XXX NO. XX 000–000 XXXX E

216

Fabrication of High-Aspect-Ratio Structures. One of the ad-vantages of nanoskiving is that it can produce high-aspect-ratio structures, by cutting micrometer-thickslabs (up to 10 !m thick is possible with an ultramicro-tome). Figure 3d shows an array of high-aspect-ratioconcentric cylinders obtained by cutting a 600 nm thickslab from an array of posts, such as those shown in Fig-ure 2a. A 20 min exposure to an air plasma (100 W,1 torr) etched the epoxy matrix and the PPy betweenthe two gold shells. Figure 3e is an array of coaxial cyl-inders with very high aspect ratios. Even the talleststructures in the array ("2 !m, as shown in Figure 3e)did not fall over.

Fabrication of Nanostructures of Other Materials. In order toshow that it was possible to produce structures of ma-terials other than gold, we formed crescents of plati-num (Figure 4a). We also formed opposing crescents

of silver and silicon (Figure 4b) and an array of cres-cents composed of gold, silicon dioxide, and palladium(Figure 4c) to demonstrate the formation of patternscomprising precisely registered (or touching) nano-structures of two or more materials. Nanoskiving is notlimited to structures deposited by physical vapor depo-sition. Semiconductor nanocrystals can be coated onthe side walls of epoxy posts and sectioned into arraysof rings. Drop-casting a solution of oleylamine-cappedPbS nanocrystals in hexanes on an array of epoxy micro-posts, followed by plasma oxidation of the ligands(which rendered the film insoluble),43 embedding thearray in epoxy, and sectioning, produced the array ofrings shown in Figure 4d.

Transmission Spectra of Arrays of Single and Double Rings. Inorder to demonstrate that the arrays of nanostructuresproduced by nanoskiving are of sufficient quality for op-tical applications, we obtained transmission spectra ofan array of single rings and double, concentric rings(Figure 5) on a ZnSe substrate. The dimensions of thesingle rings were as follows: d # 335 $ 26 nm, thick-ness # 34 $ 5 nm, and h # 114 $ 19 nm (N # 7). Thedimensions of the double rings were as follows:dinner ring # 330 $ 19 nm, thicknessinner ring # 40 $ 5nm, douter ring # 725 $ 48 nm, thicknessouter ring # 38 $8 nm, and h # 137 $ 10 nm (N # 7). The diameter of theouter ring in the direction of cutting was 10% smallerthan it was along the perpendicular, uncompressed axisbecause the PPy spacer layer was more compressiblethan was the epoxy matrix. The compression of the in-ner ring was insignificant. The source of irradiation wasa focused globar, which was polarized perpendicular tothe compressed axis (direction of cutting). We placedthe sample at the beam waist to approximate excita-tion by a plane-wave. The transmission spectrum of thesingle rings displayed one dip in transmission, whilethe spectrum of the double rings exhibited two (Fig-ure 5a). These features in the spectra corresponded tothe dipole resonances of the rings.44 As expected, thesmaller ring produced a higher energy resonance (% "2.5 !m) than did the larger ring (% " 5 !m). The posi-tions of the resonances in the results of the FDTD simu-lations approximately matched those of the measuredspectra (Figure 5b). Figure 5c shows the simulated in-tensities of the electric field in the near-field for each ofthe dipolar resonances of the rings.

We attribute the red shift of the resonances in theexperimental spectra relative to those of the simulatedspectra to the mechanical deformation of the rings intoellipses during sectioning. Because the polarization ofthe excitation light was perpendicular to the compres-sion direction, the incident light encountered a slightlylarger effective ring diameter. This effect had the conse-quence of red-shifting the resonances compared tothose found in the simulated spectra. Furthermore, thehigher-wavelength resonance is more red-shifted thanthe lower-wavelength resonance because the PPy

Figure 5. Comparison of transmission spectra of arrays ofgold nanostructures produced by nanoskiving, and FDTDsimulations of the nanostructures with idealized geometries.(a) Transmission spectra as a function of free-space wave-lengths for two types of arrays of gold nanostructures (area" 0.5 mm2) mounted on a ZnSe substrate. (b) Correspondingsimulated spectra (FDTD) of 5 ! 5 arrays, for which we as-sumed the rings were identical and exactly circular. (c) Simu-lated near-field profiles of intensity of the electric field,viewed as a cross section through the center of the rings.The wavelengths of incident light that excite these modescorrespond to the dips in transmission in (b).

ART

ICLE

VOL. XXX NO. XX LIPOMI ET AL. www.acsnano.orgF

217

spacer layer between the two rings was more com-pressible than the epoxy encircled by the inner ring.Nevertheless, the closeness of the experimental andtheoretical results demonstrates that nanoskiving pro-duces structures of sufficient quality for opticalapplications.

While the experimental and simulated results matchwell for the dips in transmission due to the smaller ring,the dip in transmission due to the larger ring is broaderand less intense in the experimental data than in theFDTD simulations. We attribute this effect to the re-duced quality (e.g., roughness and holes; see Figure 3d)of the gold film deposited on the PPy sacrificial layer,which forms the larger ring. The decrease in the qual-ity of the film reduces the quality factor of the resonantmode and also reduces the homogeneity of the ringswith respect to the other rings in the array. Both ofthese effects can decrease the amplitude and increasethe width of a resonance. In general, the roughnessshould not change the overall shape of the resonances,provided the roughness is small compared to the wave-length. The effect of point defects depends on the na-ture of the defect. For example, rings of slightly differ-ent sizes will resonate at different frequencies, and theglobal resonances observed will again be wider andshallower than in the ideal case (inhomogeneousbroadening). Possible methods to improve the qualityof the film of the larger ring would be to use a sacrifi-

cial material whose morphology is easier tocontrol than is that of PPy (e.g., silicon diox-ide, etched by HF or CF3H, or amorphous sili-con, etched by XeF2). The simulated spectrarepresent the responses of 5 ! 5 arrays ofstructures (see the Supporting Informationfor details of the optical characterization andFDTD simulations).

Fabrication of Arrays with Gradients of Height and3D Nanostructures. Figure 6a is a schematic illus-tration of the procedure used to generate ar-rays of nanostructures with a continuous gra-dient of height. Tilting the block face forwardby "2° with respect to the diamond knife pro-duced a wedge-shaped slab with a continu-ous increase in height from 0 to up to 2 #m.Figure 6b is an array of concentric cylindersformed from microposts. Figure 6c,d showsimages taken at tilt of 45° from the thick andthin regions, respectively. The effect of the cor-rugated side walls of the posts (see Figure 2b,inset) produced the structure shown in Figure6cOthe gold film collected on the thick partsof the epoxy posts. Such an approach, similarto “side wall corrugation lithography”,45 couldbe used to generate more sophisticated 3Dstructures.17

Limiting Factors. There are several factorsthat limit the types of structures that can

be made by molding, thin-film deposition, and sec-tioning. The resolution of EBL combined with theBosch process limits the geometries and sizes ofstructures that can be made. The use of the perim-eters of molded features limits the structures to linesegments that cannot cross. There is also a trade-off between the pitch and the heights of the moldedfeatures: features that are too tall must be sepa-rated in order to permit collimated beams of metalatoms to reach the bottoms of the posts. The me-chanical properties of thin films, in principle, limitthe materials that can be used; we have observed,for example, that hard metals (e.g., platinum) frac-ture more extensively than soft metals (e.g., gold).The rough edges and holes in the side walls of somestructures are traits of the underlying roughness ofthe epoxy features, as well as the conditions used forphysical vapor deposition. We have found that ultra-sonic, oscillating knives produce smoother featuresthan stationary knives, but these knives exacerbatethe effects of chips in the knives, by spreading theirdamage over a path whose width equals the ampli-tude of oscillation ("400 nm).46 Preliminary observa-tions suggest that the speed of cutting, betweenthe programmable range of 0.1$10 mm/s, has no ef-fect on the frequency of defects. We believe that fur-ther optimization of the process could yield struc-

Figure 6. Fabrication of structures with a gradient of heights by ob-taining wedge-shaped slabs of nanopost arrays. (a) Schematic illustra-tion of the process. (b!d) Array of stacked concentric rings with in-ner diameters of "1 "m. We achieved partial separation of the ringsby evaporating against microposts with corrugated side walls (see Fig-ure 2b) at normal incidence; this procedure deposited metal only onthe wide segments of the microposts.

ARTIC

LE

www.acsnano.org VOL. XXX NO. XX 000–000 XXXX G

218

tures with qualities close to those produced byconventional lithography.

CONCLUSIONSThe combination of processes described in this

paperOreplica molding of high-aspect-ratio nano-posts, deposition of thin films, and nanoskivingOprovides an effective new means of replicating arraysof nanostructures over areas 1 mm2. (The maximumarea is limited, at present, to the size of the availablediamond knives.) The geometries of individual nano-structures are not limited to circles and semicircles, suchas those demonstrated in this paper. Rather, the perim-eter of essentially any structure defined by lithogra-phy, and etching, can form a pattern by nanoskiving.Angle-dependent deposition and the use of sacrificialthin films as spacers can generate complex geometriescomprising two or more materials in the same array.Any material that can be deposited on an epoxy rep-lica by evaporation, sputtering, electrochemical growth,or drop-casting can define the composition of the finalarray of nanostructures. We have demonstrated struc-tures composed of metals, semiconductors, dielectrics,and conducting polymers, singly or in combination witheach other, in a variety of geometries, with low or highaspect ratios, and cut as many as 60 contiguous arraysfrom a single embedded structure.

We believe the combination of molding and nano-skiving has the potential to produce structures formetamaterials, sensing based on SERS or LSPR, and forbasic study of optical properties of structures that can-not be made using existing methods (e.g., very high-aspect-ratio structures, or those comprising two ormore materials in the same plane). The closeness ofthe experimental data to the FDTD simulations for thearrays of single and double rings suggests that at leastthe basic optical properties of these arrays are calcu-lable and predictable. The extent to which desirableproperties can be programmed into the structures de-pends on the error introduced by rough edges andquality of the films, damage to the structures due tomechanical cleavage, and the thickness-dependentcompression in the direction of cutting. We believe thatthe production of masters whose relief features have ar-bitrary geometries and smooth side walls, the deposi-tion of smooth thin films, and proficient use of the ul-tramicrotome will enable the production of a largenumber of structures for basic research and eventualapplications.

The most important impediment to transformingnanoskiving from a technique for research to one ofmanufacturing is replacing the manual stepsOroughcutting the arrays to a size small enough (!1 mm " 1mm) for the knife of the ultramicrotome, aligning theembedded structures with the knife edge, and collect-ing the sections from the water-filled troughOwith au-tomated procedures. The manual steps contribute to

slab-to-slab variability. Differences in the slabs cut at dif-ferent depths within the arrays of nanoposts arise fromthe imperfect, manual alignment of the embedded ar-ray to the edge of the knife, and plastic deformation(bowing) of the array while rough cutting and embed-ding. A recent technological developmentOreel-to-reellathing ultramicrotomyOstands out as potentially use-ful for high-throughput and large-area nanoskiving.47

The use of custom knives longer than 4 mm could alsoenable fabrication over larger areas than is possiblewith knives designed for small-area sections for trans-mission electron microscopy.

One of the original motivations for developingnanoskiving was to provide a simple method of mak-ing nanostructures similar to those that could, in prin-ciple, be made by scanning-beam lithographies, such asEBL and FIB.32 Nanoskiving would thus be a simple, al-ternative method of nanofabrication accessible to gen-eral users in fields outside of electrical engineering andsolid-state physics, such as chemistry and biology. Thispaper, however, demonstrates that the combination ofcharacteristics of arrays of structures formed bynanoskivingOmultimaterial, high-aspect-ratio, three-dimensional, flexible, manipulable, and replicableOarenot found in structures formed by other techniques. Webelieve that research on plasmonic materials is thearea to which the structures produced by nanoskivingcan be most quickly applied, though there could beother applications in, for example, surfaces with engi-neered interfacial properties48 and devices for energyconversion and storage.9,49 Nanoskiving might, ulti-mately, suggest new ways of nanomanufacturing bycutting.

Acknowledgment. This research was supported by the Na-tional Science Foundation under award PHY-0646094. F.C. ac-knowledges a DOD/DARPA Contract Award No. HR 0011-06-1-0044. The authors used the shared facilities supported by theNSF under MRSEC (DMR-0213805 and DMR-0820484). This workwas performed in part using the facilities of the Center for Nano-scale Systems (CNS), a member of the National NanotechnologyInfrastructure Network (NNIN), which is supported by the Na-tional Science Foundation under NSF Award No. ECS-0335765.CNS is part of the Faculty of Arts and Sciences at Harvard Univer-sity. D.J.L. acknowledges a Graduate Fellowship from the Ameri-can Chemical Society, Division of Organic Chemistry, sponsoredby Novartis. The authors acknowledge Romain Blanchard andBenjamin Wiley for helpful discussions, Ludovico Cademartiri forsynthesizing the PbS nanocrystals, and Christian Pflugl for assis-tance with the optical setup and general advice.

Supporting Information Available: Details of the fabrication,optical characterization, and simulation. This material is avail-able free of charge via the Internet at http://pubs.acs.org.

REFERENCES AND NOTES1. Maier, S. A; Atwater, H. A. Plasmonics: Localization and

Guiding of Electromagnetic Energy in Metal/DielectricStructures. J. Appl. Phys. 2005, 98, 011101.

2. Love, J. C.; Paul, K. E.; Whitesides, G. M. Fabrication ofNanometer-Scale Features by Controlled Isotropic WetChemical Etching. Adv. Mater. 2001, 13, 604–607.

3. Wu, D. M.; Fang, N.; Sun, C.; Zhang, X.; Padilla, W. J.; Basov,

ART

ICLE

VOL. XXX NO. XX LIPOMI ET AL. www.acsnano.orgH

219

D. N.; Smith, D. R.; Schultz, S. Terahertz Plasmonic HighPass Filter. Appl. Phys. Lett. 2003, 83, 201–203.

4. Sharma, A. K.; Jha, R.; Gupta, B. D. Fiber-Optic SensorsBased on Surface Plasmon Resonance: A ComprehensiveReview. IEEE Sens. J. 2007, 7, 1118–1129.

5. Kneipp, K.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S.Surface-Enhanced Raman Scattering and Biophysics. J.Phys.: Condens. Matter 2002, 14, R597–R624.

6. Smythe, E. J.; Dickey, M. D.; Bao, J. M.; Whitesides, G. M.;Capasso, F. Optical Antenna Arrays on a Fiber Facet for InSitu Surface-Enhanced Raman Scattering Detection. NanoLett. 2009, 9, 1132–1138.

7. Liu, X. F.; Sun, C. H.; Linn, N. C.; Jiang, B.; Jiang, P. Wafer-Scale Surface-Enhanced Raman Scattering Substrates withHighly Reproducible Enhancement. J. Phys. Chem. C 2009,113, 14804–14811.

8. Fort, E.; Gresillon, S. Surface Enhanced Fluorescence. J.Phys. D: Appl. Phys. 2008, 41, 013001.

9. Atwater, H. A.; Polman, A. Plasmonics for ImprovedPhotovoltaic Devices. Nat. Mater. 2009, 9, 205–213.

10. Peumans, P.; Bulovic, V.; Forrest, S. R. Efficient PhotonHarvesting at High Optical Intensities in Ultrathin OrganicDouble-Heterostructure Photovoltaic Diodes. Appl. Phys.Lett. 2000, 76, 2650–2652.

11. Klar, T. A.; Kildishev, A. V.; Drachev, V. P.; Shalaev, V. M.Negative-Index Metamaterials: Going Optical. IEEE J. Sel.Top. Quantum Electron. 2006, 12, 1106–1115.

12. Cubukcu, E.; Yu, N. F.; Smythe, E. J.; Diehl, L; Crozier, K. B.;Capasso, F. Plasmonic Laser Antennas and RelatedDevices. IEEE J. Sel. Top. Quantum Electron. 2008, 14, 1448–1461.

13. Pendry, J. B.; Holden, A. J.; Robbins, D. J.; Stewart, W. J.Magnetism From Conductors and Enhanced NonlinearPhenomena. IEEE Trans. Microwave Theory Tech. 1999, 47,2075–2084.

14. Valentine, J.; Zhang, S.; Zentgraf, T.; Ulin-Avila, E.; Genov,D. A.; Bartal, G.; Zhang, X. Three-Dimensional OpticalMetamaterial with a Negative Refractive Index. Nature2008, 455, 376–380.

15. Pendry, J. B. Negative Refraction Makes a Perfect Lens.Phys. Rev. Lett. 2000, 85, 3966–3969.

16. Pendry, J. B.; Schurig, D.; Smith, D. R. ControllingElectromagnetic Fields. Science 2006, 312, 1780–1782.

17. Liu, N.; Guo, H. C.; Fu, L. W.; Kaiser, S.; Schweizer, H.;Giessen, H. Three-Dimensional Photonic Metamaterials atOptical Frequencies. Nat. Mater. 2008, 7, 31–37.

18. Gansel, J. K.; Thiel, M.; Rill, M. S.; Decker, M.; Bade, K.; Saile,V.; von Freymann, G.; Linden, S.; Wegener, M. Gold HelixPhotonic Metamaterial as Broadband Circular Polarizer.Science 2009, 325, 1513–1515.

19. Stewart, M. E.; Anderton, C. R.; Thompson, L. B.; Maria, J.;Gray, S. K.; Rogers, J. A.; Nuzzo, R. G. NanostructuredPlasmonic Sensors. Chem. Rev. 2008, 108, 494–521.

20. Xia, Y.; Xiong, Y. J.; Lim, B.; Skrabalak, S. E. Shape-Controlled Synthesis of Metal Nanocrystals: SimpleChemistry Meets Complex Physics. Angew. Chem., Int. Ed.2009, 48, 60–103.

21. Ahmadi, A.; Ghadarghadr, S.; Mosallaei, H. An OpticalReflectarray Nanoantenna: The Concept and Design. Opt.Express 2010, 18, 123–133.

22. Haynes, C. L.; Van Duyne, R. P. Nanosphere Lithography: AVersatile Nanofabrication Tool for Studies of Size-Dependent Nanoparticle Optics. J. Phys. Chem. B 2001,105, 5599–5611.

23. Gwinner, M. C.; Koroknay, E.; Fu, L. W.; Patoka, P.;Kandulski, W.; Giersig, M.; Giessen, H. Periodic Large-AreaMetallic Split-Ring Resonator Metamaterial FabricationBased on Shadow Nanosphere Lithography. Small 2009, 5,400–406.

24. Paul, K. E.; Zhu, C.; Love, J. C.; Whitesides, G. M. Fabricationof Mid-Infrared Frequency-Selective Surfaces by SoftLithography. Appl. Opt. 2001, 40, 4557–4561.

25. Stewart, M. E.; Mack, N. H.; Malyarchuck, V.; Soares,J. A. N. T.; Lee, T. W.; Gray, S. K.; Nuzzo, R. G.; Rogers, J. A.

Quantitative Multispectral Biosensing and 1D ImagingUsing Quasi-3D Plasmonic Crystals. Proc. Natl. Acad. Sci.U.S.A. 2006, 103, 17143–17148.

26. Henzie, J.; Barton, J. E.; Stender, C. L.; Odom, T. W. Large-Area Nanoscale Patterning: Chemistry Meets Fabrication.Acc. Chem. Res. 2006, 39, 249–257.

27. Lee, J.; Hasan, W.; Stender, C. L.; Odom, T. W. Pyramids: APlatform for Designing Multifunctional Plasmonic Particles.Acc. Chem. Res. 2008, 41, 1762–1771.

28. Tserkezis, C.; Papanikolaou, N.; Gantzounis, G.; Stefanou, N.Understanding Artificial Optical Magnetism of PeriodicMetal-Dielectric-Metal Layered Structures. Phys. Rev. B2008, 78, 165114.

29. Su, K. H.; Durant, S.; Steele, J. M.; Xiong, Y.; Sun, C.; Zhang,X. Raman Enhancement Factor of a Single TunableNanoplasmonic Resonator. J. Phys. Chem. B 2006, 110,3964–3968.

30. Engheta, N. Circuits with Light at Nanoscales: OpticalNanocircuits Inspired by Metamaterials. Science 2007, 317,1698–1702.

31. Hoffman, P.; Utke, I.; Perentes, A.; Bret, T.; Santschi, C.;Apostolopoulos, V. Comparison of Fabrication Methods ofSub-100 nm Nano-Optical Structures and Devices. Proc.SPIE 2007, 5925.

32. Xu, Q. B.; Rioux, R. M.; Dickey, M. D.; Whitesides, G. M.Nanoskiving: A New Method To Produce Arrays ofNanostructures. Acc. Chem. Res. 2008, 41, 1566–1577.

33. Gates, B. D.; Xu, Q. B.; Stewart, M.; Ryan, D.; Willson, C. G.;Whitesides, G. M. New Approaches to Nanofabrication:Molding, Printing, and Other Techniques. Chem. Rev. 2005,105, 1171–1196.

34. Xu, Q. B.; Bao, J. M.; Rioux, R. M.; Perez-Castillejos, R.;Capasso, F.; Whitesides, G. M. Fabrication of Large-AreaPatterned Nanostructures for Optical Applications byNanoskiving. Nano Lett. 2007, 7, 2800–2805.

35. Pokroy, B.; Epstein, A. K.; Persson-Gulda, M. C. M.;Aizenberg, J. Fabrication of Bioinspired ActuatedNanostructures with Arbitrary Geometry and Stiffness.Adv. Mater. 2009, 21, 463–469.

36. Xu, Q. B.; Mayers, B. T.; Lahav, M.; Vezenov, D. V.;Whitesides, G. M. Approaching Zero: Using FracturedCrystals in Metrology for Replica Molding. J. Am. Chem.Soc. 2005, 127, 854–855.

37. McAuley, S. A.; Ashraf, H.; Atabo, L.; Chambers, A.; Hall, S.;Hopkins, J.; Nicholls, G. Silicon Micromachining Using aHigh-Density Plasma Source. J. Phys. D: Appl. Phys. 2001,34, 2769–2774.

38. Jesior, J. C. How to Avoid CompressionOA Model Study ofLatex Sphere Grid Sections. J. Ultrastruct. Res. 1985, 90,135–144.

39. Jesior, J. C. How to Avoid Compression 2OThe Influenceof Sectioning Conditions. J. Ultrastruct. Mol. Struct. Res.1986, 95, 210–217.

40. Yee, K. S. Numerical Solution of Initial Boundary ValueProblems Involving Maxwell’s Equations in IsotropicMedia. IEEE Trans. Antennas Propag. 1966, AP14, 302.

41. Taflove, A.; Hagness, S. C. Computational Electrodynamics:The Finite-Difference Time-Domain Method; Artech HousePublishers: Boston, MA, 2005.

42. Xu, Q.; Rioux, R. M.; Whitesides, G. M. Fabrication ofComplex Metallic Nanostructures by Nanoskiving. ACSNano 2007, 1, 215–227.

43. Cademartiri, L.; von Freymann, G.; Arsenault, A. C.;Bertolotti, J.; Wiersma, D. S.; Kitaev, V.; Ozin, G. A.Nanocrystals as Precursors for Flexible Functional Films.Small 2005, 1, 1184–1187.

44. Aizpurua, J.; Hanarp, P.; Sutherland, D. S.; Kall, M.; Bryant,G. W.; de Abajo, F. J. G. Optical Properties of GoldNanorings. Phys. Rev. Lett. 2003, 90, 057401.

45. Kostovski, G.; Mitchell, A.; Holland, A.; Austin, M. SidewallCorrugation Lithography: Bulk Fabrication of OrderedNanowires, Nanoribbons, and Nanorings. Appl. Phys. Lett.2008, 92, 057401.

ARTIC

LE

www.acsnano.org VOL. XXX NO. XX 000–000 XXXX I

220

46. Studer, D.; Gnaegi, H. Minimal Compression of UltrathinSections with Use of an Oscillating Diamond Knife. J.Microsc. Oxford 2000, 197, 94–100.

47. Kasthuri, N.; Hayworth, K.; Tapia, J. C.; Schalek, R.; Nundy,S.; Lichtman, J. W. The Brain on Tape: Imaging an Ultra-Thin Section Library (UTSL). Soc. Neurosci. Abstr. 2009.

48. Pokroy, B.; Kang, S. H.; Mahadevan, L.; Aizenberg, J. Self-Organization of a Mesoscale Bristle into Ordered,Hierarchical Helical Assemblies. Science 2009, 323,237–240.

49. Lipomi, D. J.; Chiechi, R. C.; Reus, W. F.; Whitesides, G. M.Laterally Ordered Bulk Heterojunction of ConjugatedPolymers: Nanoskiving a Jelly Roll. Adv. Funct. Mater. 2008,18, 3469–3477.A

RTICLE

VOL. XXX NO. XX LIPOMI ET AL. www.acsnano.orgJ

221

SI 1

Supporting Information

Fabrication and Replication of Arrays of Single- or Multi-Component Nanostructures by Replica Molding and Mechanical Sectioning

Darren J. Lipomi,1 Mikhail A. Kats,2§ Philseok Kim,1,2§ Sung H. Kang,2 Joanna Aizenberg,1,2

Federico Capasso,2 and George M. Whitesides1*

1Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street, Cambridge, MA, 02138

2School of Engineering and Applied Sciences, Harvard University, 29 Oxford Street, Cambridge, MA, 02138

§These authors contributed equally to this work. *Author to whom correspondence should be addressed.

222

SI 2

1. Mastering, Molding, Embedding, and Sectioning

Mastering and Molding. Silicon masters of the micropost and nanopost arrays were

fabricated using electron-beam lithography (EBL) followed by the Bosch DRIE process as

described elsewhere.1 The surfaces of the masters were passivated with heptadecafluoro-1,1,2,2-

tetrahydrodecyltrichlorosilane by exposure to a vapor. The elastomeric mold (a negative replica)

of this master structure was prepared by pouring a mixed and degassed elastomer,

poly(dimethylsiloxane) (PDMS, Dow-Corning Sylgard 184, mixed in a ratio of 10:1 base to

hardener), over the mold. The mold was cured for 2 h at 70 °C, after which the mold was cut

from the master with a razor blade. Figure S1 shows photographs of structures used during the

process of molding and embedding. Figure S1a is a photograph of a PDMS mold. To form an

epoxy replica, a UV-curable, single-component, epoxy prepolymer, UVO-114 (Epoxy

Technology, Billercia, MA), was poured over the mold and placed in a vacuum desiccator and

degassed for ~ 1 min to infiltrate the prepolymer into the mold. We poured the prepolymer so

that the liquid rose above the surface of the mold by about 2 mm. This material provided a rigid

backing after curing. The prepolymer was cured by irradiation with UV light (~ 100 W, with an

i-line bandpass filter) for 20 min. The epoxy replica was released from the mold after cooling to

room temperature (Figure S1b). These substrates were coated by evaporation, sputtering,

electrochemical deposition, or drop-casting, in a manner that depended on the desired geometry

and composition of the final structure. Details of specific procedures for producing structures

shown in Figures 2, 3, and 4 are written in the following sections. Epoxy substrates coated with

thin films (S1c) were cut into ~ 1-mm-wide strips (S1d) with a razor blade and a hammer, which

were again cut into 1-mm squares (S1e). These pieces were placed, face-up, on a scrap of PDMS

and treated with a brief exposure to an air plasma (5 s, 100 W, 500 mtorr, SPI Plasma Prep II),

223

SI 3

covered with an epoxy prepolymer, and degassed in a vacuum desiccator, such that the epoxy

prepolymer infiltrated the spaces between nanoposts (S1f). Separately, PDMS block molds were

prepared by casting a 1-cm-thick slab of PDMS in a Petri dish and cutting the PDMS into

squares. We cut a square hole (~ 0.5 cm × 0.5 cm × 1 cm, S1g) into each square of PDMS. The

squares of epoxy nanoposts coated with thin films were transferred, face-down, along with a

coating of uncured epoxy prepolymer, to flat pieces of PDMS (S1h). The epoxy was cured with

UV light for at least 20 min; this action glued the epoxy squares to the PDMS substrates. The

PDMS block molds were placed over the epoxy squares. The block molds formed a reversible

seal to the PDMS substrates (S1j). The molds were filled with additional epoxy prepolymer

(S1k) and again cured under UV radiation, for 20 min. The substrates were inverted and

irradiated a third time, in order to ensure complete curing of the epoxy in between the nanoposts.

The block molds were disassembled and the epoxy blocks containing the arrays of epoxy

nanoposts were removed (S1l). Figure S2 is a schematic diagram of the process used to embed

coated epoxy nanoposts so that the plane defined by the array of nanoposts was parallel with the

plane of the facet of the epoxy block.

The epoxy blocks were mounted in the sample chuck of the ultramicrotome (Leica

Ultracut UCT) and trimmed with a razor blade to define an area of ~ 1 mm2 around the array of

embedded nanoposts. The facets of the blocks were aligned under the stereomicroscope of the

ultramicrotome so that they were parallel to the edge of the knife. The tops of the nanoposts were

beneath the surface of the facet of the block by 1 – 5 !m. The material produced blank slabs of

epoxy. The point at which the knife cut off the tops of the nanoposts, which were covered in

gold, signaled the position in the block where useable sections containing nanostructures would

be cut. The blocks were sectioned at a speed of 1 mm/s with a typical set thickness of 100 nm

224

SI 4

and collected by surface tension in a loop tool and placed on pieces of an Si wafer for imaging,

or on a ZnSe window (ISP Optics) for optical characterization. Figure S3a is a photograph of a

Leica UC6 ultramicrotome, which is similar to the one we used. Figure S3b is a side view of the

sample chuck and knife holder as the epoxy block impinges upon the knife. Figure S3c is top

view of the same action. The single-crystalline diamond blade and the water-filled trough are

visible.

Figure S1. Photographs of structures used for molding and embedding.

225

SI 5

Figure S2. Schematic drawing of the process used to embed epoxy nanoposts in epoxy so that the epoxy nanoposts are parallel to the surface.

226

SI 6

Figure S3. Photographs and schematic drawings of the tools of ultramicrotomy and nanoskiving. Images courtesy of Dr. Ryan C. Chiechi.

2. Procedures for Depositing Thin Films for Structures Shown in Figures 2, 3, and 4.

Fabrication of Single Gold Rings (Figure 2d). The epoxy replica was coated with gold

at using a bench-top sputter-coater (Model 208HR, Cressington, Watford, UK) at 20 mA for

1000 s while rotating and tiling the epoxy replica. The substrate was trimmed, embedded, and

sectioned, as described in the previous part.

Fabrication of Double Gold Rings (Figure 2e and 2f). A gold-coated array of epoxy

nanoposts, described in the previous paragraph, was cleaned by 100 W oxygen plasma (Model

227

SI 7

Femto, Diener GmbH, Nagold, Germany) for 10 s. Separately, we prepared a solution for

electrochemical growth of polypyrrole (PPy), which contained 0.1 M pyrrole (obtained from

Sigma-Aldrich, purified using an alumina column) and 0.1 M sodium dodecylbenzene sulfonate

(Sigma-Aldrich). The solution was purged by dry nitrogen for 10 min. We added the gold-coated

array of nanoposts to this solution. The array served as the working electrode in a standard three-

electrode configuration. An anodic potential of +0.55 V vs. Ag/AgCl (saturated with NaCl) was

applied potentiostatically and a platinum mesh was used as a counter electrode. The rate of

growth was ~ 0.5 nm/s. Withdrawing the sample at a constant rate during the deposition

produced an array with a gradient of separation between the two gold layers. The freshly

deposited PPy layer was treated by applying a reductive potential of -0.5 V for 60 s. This

treatment reduced the roughness (rms) of the PPy layer from 4.5 nm to 3.0 nm (scan area: 2 !m

× 2 !m). Finally, we washed the deposited PPy layer with deionized water and dried with a

stream of compressed air. We deposited the second gold layer (the outer ring) on the

electrochemically deposited PPy using the same sputtering method described above, except that

we used a time of deposition of 1500 s, to form a continuous film over the PPy, which was

rougher than the original surface of the epoxy nanopost (the longer time of deposition filled in

some of the gaps in coverage caused by the roughness of the PPy layer). Embedding and

sectioning this substrate produced arrays of concentric rings separated by a spacer layer of PPy

embedded in a slab of epoxy (Figure 2e). Etching the epoxy and the PPy in an air plasma (1 torr,

100 W, 10 min) produced free-standing concentric rings (Figure 2f).

Fabrication of Gold Crescents (Figure 3a). We used e-beam evaporation to deposit

gold only partially around the perimeters of the nanoposts. We loaded the array of epoxy

nanoposts in the chamber directly over the crucible (distance of 40 cm) at an angle of 45° and

228

SI 8

evaporated a nominal thickness of 50 nm, as determined by the quartz crystal microbalance

(QCM) detector. We embedded, sectioned, and etched this array, as before.

Fabrication of Gold Split Rings (Figure 3b). To deposit gold around most of the

circumference of the epoxy posts, we performed three evaporations, and rotated the substrate by

approximately 80° between each one. We deposited nominal thicknesses of 30 nm. We

embedded, sectioned, and etched this array, as before.

Fabrication of Counterfacing Split Rings of Gold (Figure 3c). Starting with the array

of nanoposts from which we derived single split ring, we electrochemically deposited

polypyrrole on the partially gold-coated cylinder, and performed a second step of three e-beam

evaporations 180° to the first. We embedded, sectioned, and etched this array, as before.

Fabrication of High-Aspect-Ratio Concentric Rings of Gold (Figure 3d and 3e). We

fabricated this structure the same way as we fabricated the concentric rings, except that we

obtained sections 600 nm thick. We etched this array using a long exposure to an air plasma (1

torr, 100 W, 30 min).

Fabrication of Platinum Crescents (Figure 4a). We fabricated arrays of platinum

crescents using the same procedure as for the gold crescents, except that we evaporated platinum

instead of gold.

Fabrication of Silver/Silicon Counterfacing Crescents (Figure 4b). We fabricated this

structure by performing two evaporations. First, we evaporated silver, and then rotated the

substrate by 180°, and then evaporated silicon. We did not etch the epoxy in this case because

silver is sensitive to oxidation.

229

SI 9

Fabrication of Three-Layer Crescents of Gold, SiO2, and Pd (Figure 4c). We

performed three successive evaporations without rotating the substrate within the chamber. The

structure was embedded, sectioned, and etched, as before.

Fabrication of Arrays of Rings of PbS Nanocrystals (Figure 4d). We prepared a

solution of PbS nanocrystals in hexanes by a reported procedure at a concentration of

approximately 1016 nanocrystals/L.2

Measurement and Simulation of Arrays of Rings and Concentric Rings.

We transferred arrays of single and double rings to a 1-mm thick chemically vapor deposited

ZnSe window (ISP Optics) for optical characterization. We etched the epoxy slab using an air

plasma (1 torr, 100 W, 10 min). We covered the entire substrate except for a window around the

array by drilling a circular aperture (d = 400 µm) in copper tape, and placing the hole over the

array. We measured the transmission spectrum through the array. We performed transmission

measurements by placing the sample at the focus of a mid-IR polarized, incoherent beam from a

globar inside of the sample compartment of a Bruker Vertex 80V Fourier Transform Infrared

(FTIR) spectrometer. The polarization was set to be perpendicular the direction of compression

of the sample. The signal was obtained using a Deuterated Triglycine Sulfate (DTGS) detector

with a resolution of 16 cm-1 and a mirror scanning velocity of 1.6 kHz. The sample compartment

was placed under vacuum to remove any water absorption lines in the spectrum. Because the

signal measured by the DGTS detector was somewhat weak, we averaged 2000 scans per

spectrum for every sample. Averaging provided clean data in the range of wavelengths of 1.5 – 6

µm. For each sample, we normalized the data with a reference spectrum taken through a ZnSe

substrate by covering a blank area of the ZnSe substrate with a circular aperature (d ~ 400 µm)

drilled into the copper tape (transmission spectrum = raw sample spectrum / reference spectrum).

230

SI 10

We performed all simulations using the finite difference time domain (FDTD) method

using commercial FDTD software (Lumerical Solutions). We performed high resolution 3D

simulations using a broadband linearly polarized plane wave source in total field - scattered field

(TFSF) mode to remove the effect of the limited simulation size. Perfectly matched layers (PML)

served as boundary conditions. A graded mesh was used, designed to discretize accurately the

geometries of the rings. We defined the smaller ring with a resolution of < 3 nm in the plane of

the ring, the large ring with a resolution of < 4.5 nm, and the free space/substrate with a

resolution of < 20 nm. The vertical resolution of the rings was 3 nm. The mesh resolution was

enforced to change by at most a factor of 1.4 per mesh cell. The highly dispersive complex

dielectric function of gold was approximated by a three-coefficient polynomial fit of data from

Palik3 over the range of wavelengths of 1.5 µm – 6 µm. The slightly dispersive dielectric

function of ZnSe was approximated by a two-coefficient polynomial fit of data provided by the

supplier, ISP Optics.4 Due to the symmetry of our structures, we were able to apply symmetric

boundary conditions. The symmetry allowed us to reduce the requirement for computational

resources by a factor of four.

The distances between locations in the array in the simulation were adjusted for the

experimentally observed compression. That is, the arrays were rectangular, and the axis in the

direction of cutting was shorter than the uncompressed axis by 8%. However, the shapes of the

rings were not adjusted for the compression found in the experimental structures (where the rings

were slightly elliptical) in the simulations (we used circular rings). This discrepancy is probably

the greatest source of deviation between the experimental and simulated data.

Because of the finite size of the array, and because the exact ring size, shape, and position

differed slightly from unit cell to unit cell in our fabricated structures, we did not expect to see

231

SI 11

narrow array resonances corresponding to the photonic modes of the grating created by our array

of nanostructures.5 However, if an infinite array of structures is simulated with periodic

boundary conditions, these array resonances do appear. To decouple this photonic array

resonance from the plasmonic response of the gold rings, we simulated finite arrays and used

those as our basis for comparison. We found that a 5 × 5 array of rings was sufficient to account

for the very slight near-field coupling between neighboring rings, so we used simulations of a 5

× 5 arrays as our benchmark. Figure S4 contains the normalized transmission through a 1 × 1

array, a 3 × 3 array, a 5 × 5 array, and a 7 × 7 array, showing that simulating an array larger than

5 × 5 does not add significantly to the accuracy of our simulations.

Figure S4. Simulated transmission spectra for a single set of double rings (blue), a 3 × 3 array (black), a 5 × 5 array (red), and a 7 × 7 array (cyan). There almost no discernable difference between the 5 × 5 and the 7 × 7 case.

Each gold ring by itself (in the absence of the other one) supports a dipole resonance at

some resonant frequency. Charges on the surface of the metal move to cancel out (and reverse)

the direction of the field in the vicinity of the ring. These charges also create a high field

enhancement very close to the ring. Simulated field and charge profiles of the single and double

ring structures from Figure 5 are plotted in Figure S5.

232

SI 12

Figure S5. Simulated instantaneous electric field and charge profiles of the single ring at resonance (a), and at the lower (b) and higher (c) wavelength resonance of the double ring structure. The resonant wavelength is defined as the wavelength of lowest transmission through the array. The simulation data was monitored through the center of the ring structures. Arrows indicate the direction and magnitude of the electric field, red identifies areas of greater positive charge, while blue identifies areas of greater negative charge. Note that the charges are localized on the surfaces of the rings.

We note that the fields and charge distributions of the lower-wavelength resonance of the

double ring structure (Figure S5b) are very similar to those of a dipole resonance in a single ring

structures (such as one in Figure S5a). However in the higher-wavelength resonance of the

double structure, the fields inside the ring encounter the inner ring and are screened, creating a

charge density on the outer edge of the inner ring and nearly-zero fields inside. This interruption

slightly alters the mode profile of the higher-wavelength resonance.

233

SI 13

Fabrication of Three-Dimensional Structures and Arrays of Structures Embedded

in Prism-Shaped Slabs. One of the silicon masters we used had pronounced periodic variations

in the diameters of the posts (scalloping or sidewall corrugation). Epoxy replicas derived from

these masters retained this trait. When sputter-coated with gold, the metal collected around the

segments of the posts where the diameters were the largest. The sidewalls remained corrugated

even after electrodeposition of PPy. A second metallization also deposited selectively on the

wide segments of the posts. Embedding an array prepared in this way and sectioning it such that

the block was angled toward the knife by ~ 2° produced an array with a gradient of heights, as

shown in Figure 6.

Compression. Compression is an artifact of mechanical sectioning in which the slab is

shorter in the direction of cutting than it is in the direction parallel to the edge of the knife. We

measured a distance between sites in an array along the axis of cutting that was 8.5% shorter than

in the orthogonal axis. We calculated the same value using two different batches of the same

epoxy containing two different types of embedded structures. It should be possible to cancel out

the effect of compression by designing masters such that the axis in the direction of cutting is

elongated, so that the compression produces the desired structure. Deformation of the PDMS

mold during the curing of the epoxy, for example, would generate a skewed array which could be

“unskewed” by the compression.6 Gnaegi has shown that the use of ultrasonic, oscillating knives

produces immeasurably small compression of slabs, but we have not yet used oscillating knives.7

Yield. Figure S6 shows an array of double rings from which yield was calculated. The

array is mounted on a ZnSe crystal and contains 18 × 27 = 486 rings. They are 140 nm in height.

There is one broken structure (labeled) and ten structures whose inner and outer rings are

touching, or appear to have a metal fragment bridging the two rings (labeled).

234

SI 14

Figure S6. Array of 18 × 27 = 486 concentric double rings over an area of 2000 !m2.

Number of Consecutive Cross Sections. In order to determine the number of

consecutive cross sections of arrays of gold rings we could obtain by repeated sectioning, we

sectioned an 8-!m-tall array of gold-coated nanoposts into 100-nm slabs. We obtained 73 slabs

containing contiguous arrays of at least 100 !m × 100 !m in area. Figure S7 shows

approximately the 10th, 20th, 30th, and 50th consecutive slab. The dark grey regions contain the

arrays of gold rings; the light grey regions are blank epoxy. Two effects account for the non-

uniformity of the rings across the entire area. (1) The block face was aligned to the knife by

hand, and imperfect registration allowed the nanoposts on the left-hand side to section first. (2)

Deformation of the array during rough cutting and embedding caused the edges to bow. This

235

SI 15

bowing is why the area without nanostructures in the center of the right-hand side of the slab

persists through the 50th slab. The 50th slab is smaller than the others, because as the knife

approached the base of the posts on the left-hand side of the block, the array delaminated from

the block. The probability that part of the array will delaminate increases with depth.

Figure S7. Images of four slabs obtained from a single array of gold-coated embedded epoxy nanoposts. The scum is residue from a droplet of water that evaporated on top of the slab.

Thinning and Thickening the Epoxy Posts. It was possible to change the diameter of

the nanoposts without fabricating a new silicon master. To increase the radius, we fabricated a

secondary master of epoxy, and thickened it by sputter coating metal films of controlled

thickness using an AJA sputtering system. This secondary master, in turn, templated a new

PDMS mold with larger diameters of pores. Epoxy replicas made from these molds had

236

SI 16

correspondingly larger diameters. To reduce the diameters of the nanoposts, we made a different

secondary master from an epoxy replica that had been thinned by plasma etching (Femto plasma

cleaner, Diener electronic GmbH). PDMS molds made from these secondary masters had

correspondingly smaller diameters, which templated epoxy replicas with thinner diameters.

Figure S8. Summary of the procedure used to fabricate thickened and thinned epoxy posts from a single master structure.

Reducing the Scalloping from the Epoxy Posts. To reduce the scalloping of the epoxy

replica, we deposited a 300 nm silver film at the deposition rate of 4 !/s by DC sputtering (AJA

sputtering system). Then, a new PDMS mold was made using the metalized replica as a master.

A new replica, generated from this PDMS mold, was thinned by oxygen plasma to produce an

array of epoxy posts with the desired diameters and reduced scalloping. Figure S8 shows a

schematic drawing of the process used.

237

SI 17

References

1. McAuley, S. A.; Ashraf, H.; Atabo, L.; Chambers, A.; Hall, S.; Hopkins, J.; Nicholls, G.,

Silicon micromachining using a high-density plasma source. J. Phys. D, Appl. Phys. 2001, 34,

2769-2774.

2. Ghadimi, A.; Cademartiri, L.; Kamp, U.; Ozin, G. A., Plasma within templates: Molding

flexible nanocrystal solids into multifunctional architectures. Nano Lett. 2007, 7, 3864-3868.

3. Palik, E. D., Handbook of Optical Constants of Solids. Academic Press: New York, 1985.

4. ISP Optics, Optical Materials Specifications.

http://www.ispoptics.com/OpticalMaterialsSpecs.htm (accessed March 25).

5. Zou, S. L.; Janel, N.; Schatz, G. C., Silver nanoparticle array structures that produce

remarkably narrow plasmon lineshapes. J. Chem. Phys. 2004, 120, 10871-10875.

6. Pokroy, B.; Epstein, A. K.; Persson-Gulda, M. C. M.; Aizenberg, J., Fabrication of

Bioinspired Actuated Nanostructures with Arbitrary Geometry and Stiffness. Adv. Mater. 2009,

21, 463-469.

7. Studer, D.; Gnaegi, H., Minimal compression of ultrathin sections with use of an

oscillating diamond knife. J. Microsc. Oxford 2000, 197, 94-100.

238

Appendix VII

Micro- and Nanopatterning of Inorganic and Polymeric Substrates by Indentation

Lithography

Jinlong Gong,1 Darren J. Lipomi,1 Jiangdong Deng,2 Zhihong Nie,1 Xin Chen,1 Nicholas

X. Randall,3 Rahul Nair,3 and George M. Whitesides1

1Department of Chemistry and Chemical Biology, Harvard University

12 Oxford St., Cambridge, Massachusetts, 02138 (USA)

2Center for Nanoscale Systems, Harvard University, 9 Oxford Street, Cambridge,

Massachusetts 02138

3CSM Instruments Inc., 197 First Avenue, Suite 120, Needham, Massachusetts 02494

Reproduced with permission from

Nano Lett. 2010, ASAP Article

Copyright 2010, American Chemical Society

239

Micro- and Nanopatterning of Inorganic andPolymeric Substrates by IndentationLithographyJinlong Gong,† Darren J. Lipomi,† Jiangdong Deng,‡ Zhihong Nie,† Xin Chen,†Nicholas X. Randall,§ Rahul Nair,§ and George M. Whitesides†,*†Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street,Cambridge, Massachusetts 02138, ‡Center for Nanoscale Systems, Harvard University, 9 Oxford Street,Cambridge,Massachusetts02138,and§CSMInstrumentsInc.,197FirstAvenue,Suite120,Needham,Massachusetts02494

ABSTRACT This paper describes the use of a nanoindenter, equipped with a diamond tip, to form patterns of indentations on planarsubstrates (epoxy, silicon, and SiO2). The process is called “Indentation Lithography” (IndL). The indentations have the form of pitsand furrows, whose cross-sectional profiles are determined by the shapes of the diamond indenters, and whose dimensions aredetermined by the applied load and hardness of the substrate. IndL makes it possible to indent hard materials, to produce patternswith multiple levels of relief by changing the loading force, and to control the profiles of the indentations by using indenters withdifferent shapes. This paper also demonstrates the transfer of indented patterns to elastomeric PDMS stamps for soft lithography,and to thin films of evaporated gold or silver. Stripping an evaporated film from an indented template produces patterns of gold orsilver pyramids, whose tips concentrate electric fields. Patterns produced by IndL can thus be used as substrates for surface-enhancedRaman scattering (SERS) and for other plasmonic applications.

KEYWORDS Nanoindentation, surface patterning, lithography, SERS, nanofabrication

This paper describes the use of a commercial nanoin-denter to generate topographic patterns on substratessuch as epoxy, silicon, and SiO2. We call the process

“Indentation Lithography” and abbreviate it as IndL. Thepatterns comprise features of indentations and furrows withwell-defined sidewalls and dimensions and are created usinga diamond indenter, controlled using commercial software.This process is a new method for fabricating patterns ofmicro- and nanoscale features for soft lithography. It has fourcharacteristics not found in other techniques (e.g., electron-beam lithography (EBL) and photolithography) used tofabricate patterns of nanostructures.1 (i) It can generatemultiple levels of relief (different values of height) easily bychanging the loading force. (ii) It can control the profiles ofthe indentations by using indenters with different shapes.(iii) It can indent hard materials (thermally grown SiO2, glass,and metals) because it uses a diamond tip. (iv) It can producefeatures with a wide range of depths (e.g., from a few toseveral hundred nanometers). These characteristics combineto produce patterns with three-dimensional relief in hardmaterials; this type of typography has not been exploitedand is difficult or impossible to achieve using lithographictechniques in which features in resist can have only onevalue of height and approximately vertical sidewalls. Thetechnique does not require chemical processing and pro-

ceeds in an ambient environment. Nanoindentation is aversatile and ubiquitous tool for metrology,2 but has notbeen previously explored as a lithographic tool (aside fromconceptually related, but practically very different, scanningprobe-based techniques3).

Background. “Nanofabrication” is, in principle, any pro-cess that generates patterns of structures with sizes less than100 nm in at least one dimension. Scanning-beam tech-niques, such as EBL and focused-ion-beam (FIB) writing, arethe principal methods of generating arbitrary nanoscalepatterns (mastering); photolithography is the principal methodof transferring these patterns from one substrate to another(replication).4,5 These techniques, while the workhorses ofnanofabrication, come with high capital and operating costs,limited accessibility to general users (and users of materialsconsidered incompatible with electronics fabrication). Pho-tolithography and EBL are generally only applicable to thetwo-dimensional patterning of resist materials on planarsubstrates.4

Soft lithography is a collection of techniques that transferspatterns by printing or molding using an elastomericstamp.4,6-9 Fabricating these stamps involves casting andcuring a prepolymer (typically polydimethylsiloxane, PDMS)against a rigid master presenting topographic features. Thecured stamp is an inverse replica of the master. A largenumber of techniques have successfully generated mastersfor soft lithography,6,10 and this technique owes much of itsutility in patterning on the microscale to the use of inexpen-sive, printed transparency masks combined with photoli-

* To whom correspondence should be addressed. E-mail: gwhitesides@gmwgroup.harvard.edu.Received for review: 05/11/2010Published on Web: 00/00/0000

pubs.acs.org/NanoLett

© XXXX American Chemical Society A DOI: 10.1021/nl101675s | Nano Lett. XXXX, xxx, 000–000

240

thography.11 Generating patterns with submicrometer fea-tures, however, is more challenging. The most commonmethod of generating masters with arbitrary nanoscalepatterns is scanning-beam lithography (EBL, FIB, and directlaser writing).9,12 A simple alternative of producing surfacepatterns with nanoscale features would expand the capabili-ties of soft lithography for nanopatterning.

Nanoindentation is a technique normally used to measurethe mechanical properties (e.g., hardness and elastic modu-lus) of materials, thin films, and coatings.2,13 Modern nanoin-dentation systems equipped with diamond-tipped indentersare precise instruments that, when equipped with suitablesoftware, are capable of producing indentations and furrowsin a variety of geometries. The shapes of the features dependon the shapes of the probe tips used. Indentation systems,equipped with a diamond indenter, can indent essentiallyany material. With routine use of approximately one hun-dred indentations per day on polymeric samples, the lifetimeof a tip is about two years. Indenting metallic or ceramicsubstrates dulls a tip more quickly than indenting softmaterials, but a tip still retains its sharpness for severalmonths with routine usage. A precision nanoindentationsystem with a programmable x, y stage costs ∼$150 K(2010), and while expensive, it is less expensive than an FIBsystem or e-beam writer.

Figure 1a depicts the use of nanoindentation as a tool formicro- and nanofabrication and summarizes the procedureused to transfer the patterns to elastomeric stamps, whichin turn can be used to pattern other materials. We also usedindented silicon dioxide (thermally grown on Si(110) sub-strates) as templates for an evaporated gold or silver film,which can be “template stripped” (Figure 1b).14 This actionproduced smooth metallic films with pyramidal structuresthat can be used to concentrate electric fields. These struc-tures can thus be used as substrates for SERS and for otherplasmonic applications.

Comparison of IndL with Scanning-ProbeLithography. The use of a diamond probe for lithographyhas many conceptual similarities to scanning probe lithog-raphy (SPL, which uses, for example, an atomic forcemicroscope, AFM15-20), and other techniques, such as dip-pen nanolithography,16 nanoshaving, and nanografting.17

These techniques use a probe whose position can be scannedrelative to a substrate to write a pattern. There are, however,four important differences between IndL and SPL. (i) IndLcan be used to pattern hard materials, because it uses adiamond tip that is driven vertically into the substrate. IndLis thus capable of greater applied loads than is SPL, whichrelies on a cantilever, and is limited to patterning relativelysoft materials (e.g., polymers). (ii) IndL can generate struc-tures with a greater range of different widths and depthsthan can SPL, because IndL is capable of applying loads overseveral orders of magnitudes (100 nN - 1 N). AFM lithog-raphy, for example, can only produce shallow structures,since it can only apply relatively small forces (typically 1 pN

- 10 µN). Higher loading forces could damage or reduce thelifetime of the cantilever.21 (iii) IndL has a large workingwindow (10 cm ×10 cm) over which it can pattern withoutstitching. (iv) IndL can produce three-dimensional indentedfeatures by using probes with different shapes.

IndL employed in this work has lower resolution and lessaccuracy (∼100 nm) in positioning than SPL (∼1 nm). Thisaccuracy, however, does not depend on fundamental limita-tions, but on the accuracy of the system built for metrology,which does not require single-nanometer accuracy. Weexpect that a system for nanoindentation designed forlithography could be engineered to approach the accuracyof SPL (e.g., use of a higher-resolution sample displacementstage).

Experimental Design. Nanoindentation Systems. Weused a CSM Instruments Open Platform equipped withtwo dedicated modules: a CSM Instruments Ultra Nanoin-dentation Tester (UNHT) mounted with a Berkovich in-denter (symmetrical pyramid, total included angle 142.35°)was used for making 2D arrays of nanoindentations; aCSM Instruments Nano Scratch Tester (NST) mounted witha 1 µm radius conical diamond indenter was used forcontinuous scratching of the surface. Accurate sample

FIGURE 1. (a) Diagram summarizing the method for patterningsubstrates using Indentation Lithography. (i) The diamond indentergenerates patterns by either scratching or pressing into the surface.(ii) These patterns are replicated by pouring an elastomeric pre-polymer over the patterned substrate. Curing and separating theelastomer generates a stamp bearing relief features; this stamp isan inverse replica of the original pattern. (iii) The elastomeric stamp,in turn, can be used to template replicas of the original pattern inother materials (e.g., epoxy). (b) Schematic diagram of the procedurefor template stripping of metal films using the composite of glass/epoxy as a mechanical support.

© XXXX American Chemical Society B DOI: 10.1021/nl101675s | Nano Lett. XXXX, xxx, 000-–000

241

displacement was achieved with a Physik Instrumente (PI)P-733.2CD stage to pattern furrows and split rings. Wechose these indentation/scratch systems because they arecontrolled by software that allows easy programming ofpatterns of indentations and furrows. Berkovich andconical probes are two of the most common, com-mercially available diamond tips (Figure 2). An integratedAFM system allowed immediate imaging of the nanoin-dentations or scratches without having to search for themin a separate system.

Substrates. We used three substrates, a 1 µm thicksilica film thermally grown on a Si(100) wafer, a Si(100)wafer bearing a native layer of SiO2 (∼2.5 nm), and a flatepoxy substrate generated by puddle-casting an epoxyprepolymer against the polished surface of a Si(100)wafer. The mechanical properties (e.g., hardness, elasticmodulus) of the materials determine the extent of “sink-in” or “pile-up” around the peripheries of the indenta-tions.13 We chose thermally grown SiO2 because it pro-duced negligible pile-up.

Patterns. We chose three patterns, (i) a square array ofindentations with either uniform size (as a matter of replica-tion for SERS substrates) or two different layers of relief,“large” (deep and wide) and “small” (shallow and narrow),in the shape of inverted trigonal pyramids; (ii) straightparallel furrows connected at the ends by right angles todemonstrate the fabrication of channels with constant depth(with a potential application of making microfluidic devices);and (iii) a rectangular array of split rings to demonstrate theability to make curved features. We chose 2D arrays of splitrings to demonstrate the ability to produce curved lines, andbecause concentric split ring structures have interesting anduseful plasmonic properties.22

Molding. We replicated the topographic patterns pro-duced by indentation lithography in PDMS, gold, and silver.The choice of PDMS allowed us to demonstrate the abilityto produce stamps whose features retained the dimensionsand sharpness of the template. We also replicated a squarearray of indented features using template stripping14 to

obtain ultrasmooth pure metallic films with pyramidal struc-tures that served as substrates for surface-enhanced Ramanscattering.

Results and Discussion. Fabrication of a 2D Array ofIndentations. IndL makes it relatively easy to producepatterns with multiple levels of relief, by using differentapplied loads on the same substrate to change the depth ofpenetration. Figure 3a shows a schematic drawing of apattern of trigonal pyramidal indentations that were pro-duced in a 1 µm thick film of SiO2. We generated two sizesof indentations with the same shape. These indentationsdiffered by an order of magnitude in width and depth. Figure3b is a three-dimensional AFM image of a 3 × 3 array ofindentations in a SiO2 film that alternated between large andsmall indentations. Figure 3c shows a close-up 2D AFMimage of a 2 × 2 array of indentations. We obtained a profileof the depths of a large and a small indentation in the pattern(Figure 3d) across the red line shown in Figure 3c. Themeasurements showed the reproducibility of the features.Large indentations were 920 ( 13 nm (N ) 5) in width and110 ( 3 nm in depth; small indentations were 190 ( 3 nm(N ) 4) in width and 11 ( 0.5 nm in depth. The spacingbetween indentations was 3.5 µm. The resolution of thedisplacement stage of the system determined the minimumachievable distance between indentations (∼100 nm).

Fabrication of Channels. Figure 4a is a schematic dia-gram of a pattern of channels we scratched into epoxy. Thepattern consisted of parallel channels with length of 70 µmand spacing of 5 µm, connected at the ends to form a single,unbroken channel. Figure 4 panels b and c are AFM imagesof portions of the channels in epoxy. These channels ac-curately reproduced the geometry of the desired pattern. The1 µm radius conical diamond indenter produced grooveswith shallow V-shaped profiles (Figures 4d). Line scansperpendicular to the channels yielded an average depth of77 ( 12 nm (N ) 8). The average height of pile-up was 40 (6 nm (N ) 8). The shape, depth, and width of channelsdepended on the loading force and the shape of the indenterprobe. We did not observe any elastic recovery of theindentations, suggesting that the diamond indenter de-formed the substrate permanently. The same loading force(250 µN) left a shallower scratch (not shown) in Si(100) (∼3nm) than it did in epoxy (∼ 75 nm), because silicon is muchharder than epoxy.

The line width of a scratch is a function of both thecontact force and the sharpness of the indenter. The instru-ments we used were capable of extremely low contact forces(g10 µN), which are generally not high enough to perma-nently deform hard substrates (e.g., SiO2, TiO2, and car-bides). The minimum line width, therefore, dependedstrongly on the sharpness of the probe and on the hardnessof the material.

Fabrication of 2D Array of Split Rings. We designedthe split rings shown in Figure 5a in order to demonstratethat IndL can be used to generate curved structures. The

FIGURE 2. Schematic illustration of several types of standardindenter probes. These probes can be selected based on require-ments of the application.

© XXXX American Chemical Society C DOI: 10.1021/nl101675s | Nano Lett. XXXX, xxx, 000-–000

242

rings had an outer diameter of 10 µm, a line width of 350nm, and an opening of 60°. The pitch of these rings withinthe array was 20 µm. Figure 5 panels b and c are optical andAFM images of split rings scratched into a silicon wafer. Thispattern reproduced the geometry that we input into thesoftware that controlled the stage. Depth profiles from the

FIGURE 3. (a) Schematic drawing of a 2D array with alternating largeand small indentations separated by 3.5 µm and produced byapplying two different loading forces. (b) Three-dimensional AFMimage of the 3 × 3 array of indentations patterned on a silica film(1 µm thick) grown on Si(100). (c) An expanded 2D AFM image of 2× 2 array of indentations. (d) Two-dimensional line profile alongx-direction marked with the red line in image (c). The inset is a high-resolution AFM image of a small indentation. The loading forcesused for large and small indentations were 0.5 and 10 mN, respec-tively. The width of a large indentation is 920 ( 13 nm (N ) 5) andthe depth is 110 ( 3 nm. The small indentations are 190 ( 3 nm (N) 4) in width and 11 ( 0.5 nm in depth.

FIGURE 4. (a) Schematic drawing of channels with 70µm length and5 µm between each channel. (b) The 3D AFM images of a portion ofthe channels on epoxy. (c) A close-up 2D AFM image of the ends ofthe channels, which are connected by orthogonal line segments, onepoxy. (d) Two-dimensional line profile (note that x and y axes arenot in the same scale) along the red line drawn on image (b). Theinset is a profile of a single channel with the x- and y-axes rescaledto illustrate the shape of the profile.

© XXXX American Chemical Society D DOI: 10.1021/nl101675s | Nano Lett. XXXX, xxx, 000-–000

243

AFM image (Figure 5c) indicated that the depth of these ringswas 2.0 ( 0.4 nm. The average height of pile-up was 2.3nm. We also successfully fabricated a 2D array of split ringson an epoxy substrate; Figure 5d is a three-dimensional AFMimage of a single feature.

Pile-Up. Indentation of some materials (e.g., epoxy andSi, but not a 1 µm thick layer of thermally grown SiO2 on a

Si(110) substrate) produced “pile-up” of material displacedby plastic deformation around the perimeter of the indenta-tions. We were unable to remove piled-up material byblowing with a stream of compressed nitrogen, stripping itwith adhesive tape, or sonicating it in acetone. The strengthof adhesion of the pile-up to the substrate suggested that itwas either plastically deformed material or fragmentedparticles attached strongly to the substrate. Generation ofpile-up depends on intrinsic material properties, such as theratio of the effective modulus to the yield stress (Eeff/σy), andthe work-hardening behavior.23 In general, pile-up is signifi-cant in materials with large values of Eeff/σy (e.g., materialsthat show rigid-plastic behavior) and little or no capacity forwork hardening (i.e., “soft” metals that have been cold-worked prior to indentation). The capacity for work harden-ing inhibits pile-up because, as material at the surfaceadjacent to the indenter hardens during deformation, itconstrains the upward flow of material to the surface. Forexample, ceramics, silica, and cold-worked hard metalsgenerate negligible pile-up.23

Profile of Structures. In IndL, the profile (e.g., shape andaspect ratio of depth and width) of structures is predefinedby the geometry of the indenter probe. There are severaltypes of standard probes, which can be divided into thefollowing three groups: three-sided, four-sided, and conical(Figure 2). The probe can, also, be customized based on theapplication.

Generation of Elastomeric Stamps from PatternedSubstrates. Elastomeric stamps can reproduce nanoscalefeatures with high fidelity. For example, Xu at al. showedthat it was possible to replicate a crack in a silicon wafer withPDMS with resolution of 0.4 nm.24 To show that patterns ofsilicon generated in the SiO2 film could be transferred toPDMS, we poured a PDMS prepolymer over the SiO2 sub-strate. Thermal curing provided an inverse replica with relieffeatures whose dimensions were indistinguishable fromthose on the substrate. Figure 6a is a three-dimensional AFMimage of the PDMS replica bearing a 3 × 3 grid of large andsmall pyramids. We made these structures by casting thePDMS prepolymer against the recessed features indented inSiO2 shown in Figure 3. Figure 6b is a close-up topographicimage of a small indentation marked by a green oval inFigure 6a. Figure 6c is a line profile of a large and smallindentation corresponding to the red line in Figure 6a.Comparison of the line profiles in Figure 3d and Figure 6cindicates that the heights and widths of the raised featuresare the same as the depths and widths of the recessedfeatures of the SiO2 master.

Generation of Ultrasmooth Pure Metal Films withPyramidal Structures. After the wafer was patterned withindentation lithography, we coated it with a thin film of silveror gold using e-beam evaporation, followed by a layer ofepoxy, by puddle-casting and UV-curing a prepolymer.14,25

We then peeled off the epoxy-metal bilayer to reveal apatterned metallic film whose surface roughness between

FIGURE 5. (a) Schematic drawing of a 2D array of split rings with anouter diameter of 10 µm, a line width of 350 nm, and a 60° opening.The center-to-center distance between rings is 20 µm. (b) Opticaland (c) 3D AFM images of 2D array of split rings on silicon. (d) 3DAFM image of a split ring on epoxy.

© XXXX American Chemical Society E DOI: 10.1021/nl101675s | Nano Lett. XXXX, xxx, 000-–000

244

features was determined by the wafer template (Figure 1b).Figure 7a shows a SiO2/Si(100) (1 µm thick silica filmthermally grown in Si(100)) substrate patterned with 5 × 5array of indentations (using a loading force of 20 mN). Weevaporated 100 nm of silver on this substrate, added epoxy,and peeled off the bilayer. Figure 7b,c shows the smoothsilver film with raised pyramidal structures that was peeledoff from this wafer. Comparison of Figure 7 panels a and bsuggests the high quality of the process.

Pyramidal Structures for Surface-Enhanced RamanScattering. Gold and silver films have been investigated forenhanced molecular and biological sensing due to theconcentrated electric fields near patterned metallic nano-structures.26,27 The ability to generate enhancements withfilms that are easily and reproducibly fabricated is a currentchallenge in fabrication for plasmonics.25 We formed a self-assembled monolayer (SAM) of 4-methylbenzenethiol to thissilver surface (shown in Figure 7b,c) and used scanningconfocal Raman microscopy to collect the SERS signal as afunction of position. The resulting image for scans revealed

FIGURE 6. AFM images of PDMS stamps templated by the SiO2substrate shown in Figure 3. (a) A 3D image of an inverse replica ofthe array. (b) A close-up 3D image of a small indentation markedby a green oval in (a). (c) One-dimensional line profile of a largeand small indentation corresponding to the red line in (a).

FIGURE 7. (a) AFM image of a 100 µm thick film of thermally grownSiO2 on a Si(100) wafer patterned with a 5 × 5 array of indentationsseparated by 5 µm with a loading force of 20 mN. The indentationsare 2000 ( 6 nm (N ) 9) in width and 170 ( 3 nm (N ) 9) in depth.(b) A silver film (100 nm thick) with a 5× 5 array of pyramids formedfrom the wafer in (a) upon template stripping. (c) A close-up high-resolution AFM image of a single pyramid on silver surface. (d)Raman scattering spectra for the same 4-methylbenzenethiol-coatedpyramidal structure (at the tip) shown in (c) (green curve) and forneat 4-methylbenzenethiol (blue curve). The inset is an imageacquired using a confocal Raman microscope with a wavelength ofexcitation of 633 nm. The image is the intensity of the signal at 1077cm-1 (C-C symmetric stretching and C-S stretching).

© XXXX American Chemical Society F DOI: 10.1021/nl101675s | Nano Lett. XXXX, xxx, 000-–000

245

uniformly enhanced signals near the tip of a silver pyramid(inset of Figure 7d). Using a protocol reported previously,28

we quantified the SERS enhancement by comparing theRaman signal from neat 4-methylbenzenethiol with thatfrom our monolayer-coated pyramidal feature (Figure 7d).After correcting for the number of molecules, we determinedan enhancement factor of 107-108 at the tip of the pyrami-dal structure.

Conclusions. Indentation lithography provides an inex-pensive method for fabricating templates with arbitrarypatterns of nano- and microscale features on a wide rangeof materials. Commercial indenter systems are available inmaterials science and engineering laboratories of manyresearch universities. Rudimentary nanoindentation sys-tems are sold by some manufacturers as “add-ons” to AFMsystems. Indentation is, thus, becoming a ubiquitous tool.

Indentation lithography is part of a group of techniquesthat exploits the control and precision of analytical instru-ments for micro- and nanofabrication. In this sense, IndL isrelated to dip-pen nanolithography,16 which uses an AFMto fabricate nanostructures, and nanoskiving,29 which usesthe diamond knife of an ultramicrotome. These techniquesare capable of producing structures that are complementaryto those produced by conventional methods of fabrication.As a purely mechanical technique, capable of applying verylarge or very small forces, IndL is unique in its insensitivityto the chemical properties of the substrate. It could, thus,be uniquely suited to mechanical patterning/machining ofmultilayered materials, delicate thin films, such as SAMs,and biological materials.

As with any technique, indentation lithography has limi-tations. Like conventional mastering techniques such as EBL,indentation lithography is serial and therefore its mostpromising applications are in mastering, rather than inreplication, and in making unique structures for fundamentalphysical measurements. IndL tends to produce wide andshallow features. The exploration of custom-made indentertips could enable the generation of high-aspect-ratio features.IndL is also slow, because the use of nanoindentation as atechnique of measurement requires high sensitivity, ratherthan high speed. It might, however, be possible to fabricatearrays of diamond-tipped indenters for high-throughputpatterning, in the same way that parallel dip-pen nanolithog-raphy30 or IBM’s Millipede19 enables large-area patterningor data coding.

Acknowledgment. This work was supported by the Na-tional Science Foundation under award CHE-0518055. Theauthors used the shared facilities supported by the NSFunder NSEC (PHY-0117795 and PHY-0646094) and MRSEC(DMR-0213805 and DMR-0820484). This work was per-formed in part using the facilities of the Center for NanoscaleSystems (CNS), a member of the National NanotechnologyInfrastructure Network (NNIN), which is supported by theNational Science Foundation under NSF Award No. ECS-0335765. CNS is part of the Faculty of Arts and Sciences at

Harvard University. D.J.L. acknowledges a graduate fellow-ship from the American Chemical Society, Division of Or-ganic Chemistry, sponsored by Novartis. Z.H.N. acknowl-edges a postdoctoral fellowship from the Natural Science andEngineering Research Council of Canada. N.X.R. acknowl-edges Jim Gareau from Physik Instrumente for kindly loaninga displacement stage and controller.

Supporting Information Available. Details of experimen-tal procedures, fabrication processes, and calculation ofenhancement factors. This material is available free ofcharge via the Internet at http://pubs.acs.org.

REFERENCES AND NOTES(1) Shir, D. J.; Jeon, S.; Liao, H.; Highland, M.; Cahill, D. G.; Su, M. F.;

El-Kady, I. F.; Christodoulou, C. G.; Bogart, G. R.; Hamza, A. V.;Rogers, J. A. J. Phys. Chem. B 2007, 111, 12945–12958.

(2) Fischer-Cripps, A. C. Nanoindentation; Springer-Verlag: New York,2002.

(3) Rosa, L. G.; Liang, J. J. Phys.: Condens. Matter 2009, 21, 18.(4) Gates, B. D.; Xu, Q. B.; Stewart, M.; Ryan, D.; Willson, C. G.;

Whitesides, G. M. Chem. Rev. 2005, 105, 1171–1196.(5) Quake, S. R.; Scherer, A. Science 2000, 290, 1536–1540.(6) Xia, Y. N.; Whitesides, G. M.Annu. Rev. Mater. Sci. 1998, 28, 153–

184.(7) Ruiz, S. A.; Chen, C. S. Soft Matter 2007, 3, 168–177.(8) Bowen, A. M.; Nuzzo, R. G. Adv. Funct. Mater. 2009, 19, 3243–

3253.(9) Stewart, M.; Motala, M.; Yao, J.; Thompson, L.; Nuzzo, R. Proc.

Inst. Mech. Eng., IMechE Conf. 2006, 220, 81–138.(10) Xia, Y. N.; Tien, J.; Qin, D.; Whitesides, G. M. Langmuir 1996, 12,

4033–4038.(11) Linder, V.; Wu, H. K.; Jiang, X. Y.; Whitesides, G. M. Anal. Chem.

2003, 75, 2522–2527.(12) Craighead, H. G. Science 2000, 290, 1532–1535.(13) Oliver, W. C.; Pharr, G. M. J. Mater. Res. 1992, 7, 1564–1583.(14) Weiss, E. A.; Kaufman, G. K.; Kriebel, J. K.; Li, Z.; Schalek, R.;

Whitesides, G. M. Langmuir 2007, 23, 9686–9694.(15) Dinelli, F.; Menozzi, C.; Baschieri, P.; Facci, P.; Pingue, P. Nano-

technology 2010, 21, No. 075305.(16) Piner, R. D.; Zhu, J.; Xu, F.; Hong, S. H.; Mirkin, C. A.Science 1999,

283, 661–663.(17) Liu, G.-Y.; Xu, S.; Qian, Y. Acc. Chem. Res. 2000, 33, 457–466.(18) Pires, D.; Hedrick, J. L.; De Silva, A.; Frommer, J.; Gotsmann, B.;

Wolf, H.; Despont, M.; Duerig, U.; Knoll, A. W.Science 2010, 328,732–735.

(19) Vettiger, P.; Cross, G.; Despont, M.; Drechsler, U.; Durig, U.;Gotsmann, B.; Haberle, W.; Lantz, M. A.; Rothuizen, H. E.; Stutz,R.; Binnig, G. K. IEEE Trans. Nanotechnol. 2002, 1, 39–55.

(20) Smith, R. K.; Lewis, P. A.; Weiss, P. S. Prog. Surf. Sci. 2004, 75,1–68.

(21) Tseng, A. A.; Notargiacomo, A.; Chen, T. P. J. Vac. Sci. Technol., B2005, 23, 877–894.

(22) Gwinner, M. C.; Koroknay, E.; Fu, L. W.; Patoka, P.; Kandulski,W.; Giersig, M.; Giessen, H. Small 2009, 5, 400–406.

(23) Oliver, W. C.; Pharr, G. M. J. Mater. Res. 2004, 19, 3–20.(24) Xu, Q.; Mayers, B. T.; Lahav, M.; Vezenov, D. V.; Whitesides, G. M.

J. Am. Chem. Soc. 2004, 127, 854–855.(25) Nagpal, P.; Lindquist, N. C.; Oh, S.-H.; Norris, D. J. Science 2009,

325, 594–597.(26) Haes, A. J.; Haynes, C. L.; McFarland, A. D.; Schatz, G. C.; Van

Duyne, R. R.; Zou, S. L. MRS Bull. 2005, 30, 368–375.(27) Campion, A.; Kambhampati, P. Chem. Soc. Rev. 1998, 27, 241–

250.(28) Bantz, K. C.; Haynes, C. L. Langmuir 2008, 24, 5862–5867.(29) Xu, Q. B.; Rioux, R. M.; Dickey, M. D.; Whitesides, G. M. Acc.

Chem. Res. 2008, 41, 1566–1577.(30) Salaita, K.; Wang, Y. H.; Fragala, J.; Vega, R. A.; Liu, C.; Mirkin,

C. A. Angew. Chem., Int. Ed. 2006, 45, 7220–7223.

© XXXX American Chemical Society G DOI: 10.1021/nl101675s | Nano Lett. XXXX, xxx, 000-–000

246

Supporting Information

Micro- and Nano-Patterning of Inorganic and Polymeric Substrates

by Indentation Lithography

Jinlong Gong,1 Darren J. Lipomi,1 Jiangdong Deng,2 Zhihong Nie,1 Xin Chen,1 Nicholas X.

Randall,3 Rahul Nair,3 and George M. Whitesides1,*

1 Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street,

Cambridge, MA 02138

2 Center for Nanoscale Systems, Harvard University, 9 Oxford Street, Cambridge, MA 02138

3 CSM Instruments Inc., 197 1st Avenue, Suite 120, Needham, MA 02494

*corresponding author: gwhitesides@gmwgroup.harvard.edu

247

Experimental Preparation of Epoxy Substrates. We first exposed a polished surface of a Si(100)

wafer to a vapor of (tridecafluoro-1,1,2,2-tetrahydrooctyl)-1-trichlorosilane in a vacuum

desiccator for ~12 h to passivate the surface and to inhibit adhesion. We obtained epoxy (hard

and brittle) substrates by puddle-casting a mixed and degassed epoxy prepolymer (Epo-Fix,

Electron Microscopy Sciences) against the passivated silicon wafer. We cured the epoxy at 60 ºC

for 2 h, and then detached the epoxy from the wafer using a razor blade.

Array of Pyramidal Indentations. We produced arrays of indentations on the 1-!m-

thick silica film with a CSM Instruments Ultra Nanoindentation Tester (UNHT) equipped with a

diamond Berkovich indenter. The UNHT measurement head was mounted on an Open Platform

together with an integrated optical video microscope and an Atomic Force Microscope (AFM)

system, thus allowing direct imaging of the indentation array. We patterned the substrates using

an automated sample displacement stage to place indentations in a 5 × 5 array with uniform size

and a 3 × 3 array with two levels of relief on the silica film. For the 5 × 5 array, the indentations

were produced with a maximum applied load of 20 mN. For the 3 × 3 array, the large

indentations were produced with a maximum applied load of 10 mN, and the small indentations

were produced with a maximum applied load of 0.5 mN. The loading function for the

248

indentations consisted of a 5-second loading to peak force, followed by a 2-second hold, then by

a 5-second unloading.

Fabrication of Channels. We used a CSM Instruments Nanoscratch Tester (NST)

mounted on an Open Platform to write channels into the epoxy and silicon surfaces. The sample

was mounted on a PI P-733.2 Piezo Nanopositioning Stage. The spherical diamond indenter had

a tip radius of 1 µm and cone angle of 90 degrees. We applied a constant feedback-controlled

load sufficient to scratch the epoxy surface (250 !N), and the x and y stages were displaced to

replicate the desired pattern. The writing speed was 250 nm/s.

Fabrication of Split Rings. We used the same CSM Instruments NST system to produce

an array of split rings. We patterned the substrates by applying a constant feedback-controlled

applied load of 800 µN (silicon) and 180 µN (epoxy), while displacing the PI stage in the

preprogrammed pattern. Patterns were then immediately imaged using a CSM Instruments AFM

integrated on the same platform as the NST.

Generation of PDMS Stamps from Indented Substrates. We transferred patterns

produced in SiO2, silicon, and epoxy substrates by standard soft lithographic procedures. We

cleaned the patterned silicon or SiO2 substrates with a 5-min air plasma and passivated the

surface by exposing it to a vapor of (tridecafluoro-1,1,2,2-tetrahydrooctyl)-1-trichlorosilane in a

vacuum desiccator for ~12 h. We left the epoxy substrates untreated. To make an elastomeric

249

stamp, we mixed and degassed a PDMS prepolymer in a ratio of 10:1 of base to hardener

(Sylgard 184), poured it over the patterned Si or epoxy template, and cured it at 60 °C for 2 h.

Template Stripping of Patterned Metal Films. The patterned structures on a

SiO2/Si(110) wafer were coated with the desired metal (silver or gold) using electron-beam

evaporation with a base pressure of 2 × 10-6 torr and typical rates of deposition of 2-4 Å/second.

We placed the substrate directly over the source of metal atoms to obtain uniform filling of the

indentations and clean edges of the features. We stripped the metallic film from the substrate by

placing a drop of UV-curable epoxy prepolymer (UVO-114) on the film, placing a 1-cm2 glass

slide on the prepolymer as a mechanical support, curing the prepolymer, and stripping off the

three-component structure with a razor blade (Figure 1b). This action exposed the ultra-flat

surface of the metallic film, as well as metallic features that were inverse replicas of the patterns

of indentations.

AFM Images. We obtained AFM images using an Asylum MFP-3D (Asylum Research).

The scanning was performed in tapping mode using a silicon tip (Veeco RTESPW). The

cantilever spring constant was ~ 30 N/m and its resonant frequency was 260 kHz.

Confocal Raman Microscopy. We prepared the sample for SERS measurements by

immersing it in a 1 mM 4-methylbenzenethiol (4-MBT) solution (98%, Sigma Aldrich) in

ethanol (ACS grade, 99.98%, Pharmco-Aaper) at ambient conditions. After 2 h, we removed it

250

from the solution, rinsed it with excess ethanol (removing any 4-MBT not absorbed to the

surface), and gently dried it with nitrogen. Confocal Raman scans were then performed on a

Witec CRM300 system equipped with a HeNe laser (! = 632.8 nm, 25 mW). The spectral range

of the scan was between 0 and 4000 cm-1 with a resolution of 1 cm-1. The spatial resolution of the

confocal system was ~450 nm in the scanning plane and ~500 nm in the perpendicular direction.

Calculation of Enhancements Factors. We calculated enhancement factors (EFs) using

the intense ring-breathing stretch at 1077 cm-1 from both liquid and surface-adsorbed 4-MBT and

the equation:

volsurf

surfvol

IN

INEF ! (1)

where Nvol is the number of 4-MBT molecules contributing to the normal Raman scattering

signal, Nsurf is the number of 4-MBT molecules contributing to the SERS signal, and Isurf and Ivol

are the intensities of the scattering band at1077 cm-1 in the SERS and normal Raman scattering

spectra, respectively. Note that the silver film itself is an excellent SERS substrate; the EF of our

flat Ag film was calculated to be around 105-106. The EF at the tip of a pyramid is on an order of

107-108.