Alternative methods and materials for patterning organic thin film electronics

Alternative Methods and Materials for Patterning

Organic Thin Film Electronics

by

Matthias Erhard Bahlke

B.A. Physics (2009)

Bard College

B.S. Electrical Engineering (2009)

Columbia University

M.S. Electrical Engineering and Computer Science (2011)

Massachusetts Institute of Technology

Submitted to the Department of Electrical Engineering and Computer

Science

in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

at the

MASSACHUSETTS INSTITUTE OF TECHNOLOGY

June 2014

c

Massachusetts Institute of Technology 2014. All rights reserved.

Author . . . .

Department of Electrical Engineering and Computer Science

May 14, 2014

Certified by . . . .

Marc A. Baldo

Professor of Electrical Engineering

Thesis Supervisor

Accepted by . . . .

Leslie A. Kolodziejski

Chair, Department Committee on Graduate Students

Alternative Methods and Materials for Patterning Organic

Thin Film Electronics

by

Matthias Erhard Bahlke

Submitted to the Department of Electrical Engineering and Computer Science on May 14, 2014, in partial fulfillment of the

requirements for the degree of Doctor of Philosophy

Abstract

Photolithography’s accuracy and scalability have made it the method for sub-micron-scale definition of single-crystal semiconductor devices for over half a century. The ultimate goal for OLED manufacturing, however, is to replicate the widespread suc-cess of photoresist lithography without the use of the types of resists, solvents, and etchants traditionally used—organic small molecules are simply not compatible with these tools. Hence, there is motivation for a renewed examination of variants of this inherently parallel, high speed approach. This work investigates the use of chemically inert resists that rely on clearance mechanisms not found in traditional lithography. These primarily include employing phase changes for lift-off patterning thin films of organic semiconductors and metals, and also propose and discuss the use of com-bustible and magnetic materials.

Thesis Supervisor: Marc A. Baldo

Acknowledgments

First and foremost I’d like to thank my research advisor, Marc, for his ideas, interest and encouragement—I’d have been unable to complete this work without his help. I’d also like to thank Hiroshi Mendoza for his sustained optimism and undying motivation and work ethic. Phil Reusswig is always thinking of new ideas and would lead off even his greatest ground-breaking ideas with “this is probably not very good, but...” He’s been an amazing colleague, roommate, and, most of all, friend. I learned so much about modern OLED materials, devices and their characterization, bicycles and the best bread in the world from Sebastian Reineke. Working with him and Markus was always exciting even when results were less than perfect. Mingjuan Su and I put a great deal of work into some interesting material that did not fit into the scope of this thesis, but she’s a brilliant and enthusiastic synthetic chemist and friend. Nicholas Thompson has been very helpful with his continued interest, feedback and suggestions. It will be very exciting to see televisions with the materials and devices Nick and Mingjuan develop at UDC on the shelves of stores and someday in our homes. Apoorva Murarka saved us a lot of time donating one of his old MEMS stamps and assisting with making the first of my own. Jiye Lee was an amazingly self-motivated driving force in the group and was the person that got me started in the group. Priya Jadhav provided necessary and much-appreciated assistance with experimental design. Jason Sussman always has new ideas to discuss and has a great historical knowledge of the physicists that laid the foundation we built on. Many thanks to Jon Mapel for his assistance with scaling and market evaluation. I’m very grateful to Cathy Bourgeois for being on top of everything and helping out every step of the way. I’d like to acknowledge the rest of The Soft Semiconductor Group that I’ve worked along side of: Paul Azunre, Daniel Congreve, Jean Anne Currivan, Sumit Dutta, Markus Einzinger, Tim Heidel, Brian Modtland, Carlijn Mulder, David

Ciudad R´ıo-P´erez, Carmel Rotschild, Saimi Siddiqui, Amador Vel´azquez, Mengfei

Wu, Tony Wu, and Allen Yin for assistance and helpful discussions along the way. I’m indebted to John Kymissis and the Columbia Laboratory for Unconventional Electronics for helping me realize that graduate school in engineering was right for me. Marshall Cox, Eddy Hsu, Zhang Jia, Vincent Lee, Nadia Pervez, Samuel Sabbarao, and John Sarik are amazing teachers, friends and colleagues. Jon Beck and I worked on projects together as undergraduates and, even though I left CU for MIT, we worked concurrently on our PhD’s sharing ideas, advice, and feedback. Security guru Michael Halsall held a continued interest in my work and inadvertently sprouted my great interest in his. I appreciate Leslie’s patience in the final months of this work when my free time dwindled and my mind was never off the thesis and projects. Outside of those that assisted with theory and experimental work, I’d like to thank my mother and father for their support as well as Marlene, George, Sarah and Cora.

Contents

1 Introduction 17

1.1 Organic Light-Emitting Diode Displays . . . 18

1.1.1 Current Commercial Status . . . 19

1.1.2 Structure of an OLED . . . 20

1.2 Motivations for an Alternative Patterning Process . . . 21

1.2.1 Fine-Metal Masks . . . 22

1.2.2 Inkjet Printing . . . 23

1.2.3 Photolithographic Processes Involving Solvents . . . 25

1.2.4 Other Ways to Pattern Organic Thin Films . . . 25

2 Physics of Organic Materials and Devices 27 2.1 Electrons, Excitation, and Spin . . . 27

2.2 The OLED: Turning Charge into Light . . . 29

3 The Sublimable Mask Lithographic Process 33 3.1 Theory of Practice . . . 34

3.2 Process Flow . . . 35

4 Sublimable Mask Patterning 39 4.1 Considerations for Demonstration . . . 39

4.2 Experimental Setup . . . 40

4.2.1 Patterning the Resist with Resistive Heating . . . 44

4.3 OLEDs in a Cold Environment . . . 62

4.4 Patterning the Resist Photolithographically . . . 64

5 Future Directions 67

5.1 Combustion Lithography . . . 67

5.2 Magnetic Resists . . . 68

6 Conclusion 71

A Selected Molecules Used in this Work 85

B Triple Points of Some Gases 89

List of Figures

1-1 Dimensions of subpixels making up each of the over 8 million pixels

found in a 55-inch UHDTV. The area obstructing the top of the pixel is from the driving transistors. It is worth noting that this is an over-simplification as the variation in performance of materials for different

colors in an OLED display necessitates different sizes for each color. . 19

1-2 A fine-metal mask used for arrays of OLED subpixels[131] (top) and

some of the masks we use in our group(bottom). . . 22

1-3 Inkjet printing deposits droplets of dissolved organic compounds that

leave a solid thin film after the solvent evaporates. . . 24

1-4 Kateeva’s YIELDjetTM _{Gen-8 inkjet manufacturing equipment[62]. . . 25}

2-1 Figure from [5] showing results of density function theory(DFT)

cal-culations of pentacene and fullerene C60 exciton and charge-transfer

states. . . 28

2-2 Organic transport level diagram showing the electroluminescence

pro-cess. The hole and electron are injected from their corresponding elec-trode and move through their transport layers to meet and form an exciton. The exciton recombines dissipating its energy through a

pho-ton. . . 30

2-3 (a) Thermally activated delayed fluorescence (TADF) energy diagram.

(b) Some of the molecules that Uoyama et al. presented in their sem-inal TADF paper that this figure is borrowed from (Reprinted with

3-1 An SEM image of The Harvard Nanopore Group’s demonstrated pat-terning of chromium using water ice resist. Reprinted with permission

from [65]. Copyright 2005 American Chemical Society. . . 34

3-2 Inside cover of December 2012 issue of Advanced Materials featuring

sublimation lithography[7] . . . 35

3-3 General phase diagram. The arrow shows the ideal region of operation

with respect to sublimation lithography. Deposition of the mask takes place along the low temperature region of the arrow and sublimation and lift-off take place along the higher temperature region of the arrow. 36

3-4 Phase diagram reflecting full process flow parameters. The defining of

the mask pattern is performed immediately prior to the deposition of

the thin film. . . 37

3-5 Simplified process flow for sublimation lithography. (not to scale) (a)

Begin with a cooled substrate to facilitate resist deposition. (b) Deposit resist. (c) Selectively pattern resist. (d) Deposit desired thin film. (e) Lift-off resist leaving patterned thin film. (f) Repeat as necessary to

complete device. . . 37

4-1 The side of the evaporation chamber showing the feedthroughs and

liquid nitrogen reservoir. . . 41

4-2 Flange making up the bottom of the liquid nitrogen reservoir. The

substrate holding plate is bolted in with a glass sample tightly secured. The washer-mounted thermocouple can be seen towards the top of the

image mounted with a bolt. . . 42

4-3 Phase diagrams of CO2and H2O extrapolated (broken lines) from data

(solid lines) in references [83] and [120]. The “×” represents the process

operating point of 77 K at 106 _{Torr. The curves are extrapolated using}

the Clapeyron equation[112]. Figure from [2] . . . 43

4-4 The pattern of the ITO used in resistive heating experiments. The

4-5 A mask of CO2 patterned for a 100 µm-wide line. The substrate holder

is rigged with probe-tipped clips for electrical contact. . . 47

4-6 In situ photograph and schematic representation of the resistive

heat-ing method of patternheat-ing the frozen mask. Electrical contacts on the

substrate allow current to be driven at a current density of 625 kA cm2

through a 160 nm-thick strip of ITO on the substrate. The red arrows in the schematic drawing indicate the flow of current through the ITO

enabled by affixed copper foil tape[2]. . . 48

4-7 An exmaple of photoluminescence from a 100 µm-wide line of

DCJTB-doped Alq3(top), and photoluminescence(bottom-left ) and a

micrograph(bottom-right ) from a 100 µm-wide line of Alq3 patterned by resistive heating

of ITO. . . 49

4-8 A micrograph from a 100 µm-wide line of silver patterned by resistive

heating of ITO . . . 50

4-9 Simplified process flow for sublimation lithography using a frozen CO2

resist (not to scale). (a) Cool substrate below 100 K. (b) Freeze on

thin film of CO2. (c,d) Pattern CO2 film by heating selectively. Here

a stamp is pressed against the resist to remove some areas. (e) De-posit desired organic or metal thin film by thermal evaporation. (f)

Warm substrate to sublime CO2, thus lifting off unwanted material.

(g) Repeat steps (a-f) as necessary to complete the device[2]. . . 51

4-10 2D simulation of thermal diffusivity in the stamping process. . . 53

4-11 A micrograph of the 115 µm-tall pillar SU-8 stamp(top) used to pattern

4-12 Schematic of the updated experimental stamping setup showing the process-critical components as described in the main text. The en-tire setup is located within the standard thermal evaporation chamber (Angstrom Engineering). The arrows by the stages indicate their re-spective direction and range of motion. Both the stamp and the mi-croscope used for positioning and stamping evaluation are located on

the smaller stage as indicated[2]. . . 55

4-13 Micrograph of a region of the SU-8 stamp used to selectively sublime

regions of the CO2 resist for 78 µm-pitch patterns. In the center of the

image is a pillar that has been knocked over displaying the profile of the pillar[2]. . . 57

4-14 Optical micrograph (a) of a 78 µm pitch-patterned mask of CO2defined

via contact with a stamp. The corresponding inset (b) shows the Alq3

thin film after deposition and lift-off. Photoluminescence micrograph (c) from the same film[2]. A longpass filter is used to lessen pump

light’s presence in the image. . . 58

4-15 A false-color topography obtained by an optical interferometer (top) and a cross section of the same (bottom) detailing the profile of the

patterned organic pixels of figure 4-14[2]. . . 59

4-16 Examples of red and green subpixels side-by-side. The left two images show actual red and green subpixels next to each other whereas the right image exemplifies some of the complications that arose in the

attempts. . . 61

4-17 Photoluminescence micrograph of a free-floating 50 nm-thick film of

Alq3 folding away from the substrate yet remaining intact. It measures

about 150 µm on a side. . . 62

4-18 External quantum efficiency versus current density of OLEDs grown at T = 112 ± 24 K and room temperature (a). The normalized electrolu-minescence spectrum is indistinguishable from the room temperature

5-1 SEM micrograph of nitrocellulose on silicon patterned via an argon ion beam by Geis et al. in 1983. Reprinted with permission from [41].

Copyright 1983 American Vacuum Society. . . 68

5-2 Suggested process flow for using magnetic materials as dry lift-off

re-sists. (not to scale) (a) Begin with clean substrate (b) Apply magnetic resist where subsequent thin film is not desired (c) Deposit thin film (d) Remove resist and unwanted thin film areas with a magnet or other strong magnetic field (e) Repeat as necessary to complete all layers of

desired device. . . 69

5-3 Jacobs et al. stamped elecrolets on PMMA to position the above

mate-rials. Reprinted with permission from [58]. Copyright 2001 American

Association for the Advancement of Science. . . 70

5-4 Magnet domain arrangement of a flexible vinyl magnet and the

result-ing field lines. Image from [123] . . . 70

A-1 Aluminium tris(quinolin-8-olate) (Alq3)[125] . . . 85

A-2 Carbon dioxide (CO2) . . . 85

A-3

4-(Dicyanomethylene)-2-tert-butyl-6-(1,1,7,7-tetramethyljulolidin-4-yl-vinyl)-4H-pyran (DCJTB)[125] . . . 86

A-4 SU-8 Photoresist[125] . . . 86

A-5 Tris[2-phenylpyridinato-C2_{,N ]iridium(III)(Ir(ppy)}

List of Tables

Chapter 1

Introduction

O

up to their full potential as the basis of cheap optoelectronic devices. The problem is not one of material costs, but of obstacles to patterning for large-scale fabrication. This work provides an alternative manufacturing technique employing phase-change resists that may allow for the retirement of the fine-metal-masking methods currently used.

This chapter introduces the physics, working principles, commercial status and the as-of-yet untapped potential of organic light-emitting diodes. It also explains how a dry lithographic process could overcome present challenges in scaling up production of OLEDs on larger sizes of mother glass.

Chapter 2 provides a brief introduction of organic materials and the physics that enable OLEDs. Modern emitter concepts are discussed.

Chapter 3 introduces the sublimable mask patterning idea and provides a step-by-step overview of the process.

Chapter 4 describes the approach and design of a proof-of-concept implementation using carbon dioxide ice as the resist.

Chapter 5 discusses future directions of dry patterning of organic thin films. Chapter 6 summarizes the results and describes options for large-scale patterning of the resists and the under- and overlying thin films. Limitations of techniques and the feasibility of such processes are briefly discussed.

Table 1.1: Typical display specifications of current competing mobile displays[3][4][106][55].

Device Display Size (in) Resolution Technology ppi1

Apple iPhone 5S Retina 4 1136x640 IPS LCD 326

Apple iPad Retina Mini 7.9 2048x1536 IPS LCD 326

Samsung Galaxy S5 5.1 1920x1080 AMOLED 432

HTC One M8 4.7 1920x1080 LCD 441

1.1

Organic Light-Emitting Diode Displays

A 32-inch full high-definition television (HDTV) requires a pixel size of a 370 µm-per side square further broken up into 3 subpixels: one for red, one for green and one for blue to reproduce the necessary colors required for full color. Having features with dimensions of about 120 µm necessitates patterning accuracies on the order of 10 µm or less. For larger displays of the same resolution the pixel dimensions are larger, but for smaller displays the constraints are tighter. The move to even higher resolutions brings these patterning accuracy requirements tighter still. Ultra high definition resolution in a 55-inch display, for example, means pixels 320 µm on a side with subpixel widths of around 100 µm as shown in figure 1-1.

The highest resolution displays are found in mobile electronics like smartphones. They must reproduce great amounts information clearly on a screen of about 4 to 5 inches. The size, resolutions and pixel-per-inch (ppi) specifications of some of today’s best displays are shown in table 1.1. 450 ppi corresponds to a pixel pitch of about 56 µm; 350 ppi corresponds to a pixel pitch of about 73 µm; 250 ppi is more like 101 µm. OLED manufacturers are at present unable to realize pixel pitches much less than 60 µm due to the problems laid out in section 1.2.

1_{I’d like to note that beyond ∼326 ppi the human eye can no longer discern individual}

R G B

320μm

320

μ

m

100μm

Figure 1-1: Dimensions of subpixels making up each of the over 8 million pixels found in a 55-inch UHDTV. The area obstructing the top of the pixel is from the driving transistors. It is worth noting that this is an oversimplification as the variation in performance of materials for different colors in an OLED display necessitates different sizes for each color.

1.1.1

Current Commercial Status

Up until 2013 OLED televisions were only available in 11-inch and 15-inch sizes. The resolutions of these displays were sub-full high-definition. Following these, 55” full-HD units were introduced first by LG Electronics Display and soon after Samsung, although they came at a premium; a customer taking one of these home would be very lucky to pay less than $10,000. As production yields and fabrication techniques have improved these prices have slowly begun to creep closer to about half of their introductory prices. Panels as large as 77 inches have been demoed at Ultra High Definition resolutions by LG[29] and Panasonic has shown its own 56-inch 4K televi-sions using inkjet printing rather than thermal evaporation for its material deposition and patterning[93].

Samsung is the OLED industry leader and opened a new $2.1 billion 5.5th-generation OLED factory in the summer of 2011 to produce larger active-matrix OLED HDTVs[20][74] and has a 6th-generation fabrication facility on order[13]. Sam-sung’s yearly investment in OLED technology in 2011 was about $4.8 billion, matching its investment in liquid crystal display technology[75].

In 2010 Samsung Mobile Display announced that after years of investment, the OLED division became profitable[64]. It is of interest here to note that the market contribution from televisions is negligible at this time, so there is much room for growth. Reducing the cost of OLED displays by solving yield and scaling complica-tions is key. The current organic electronics market stands at an estimated $11-12 billion[31][107] and is expected to increase to $44 billion by 2019[91]. It is expected that by 2015, OLED HDTVs alone will be a $2 billion industry[23][51].

1.1.2

Structure of an OLED

OLEDs have a simple general structure and principle of operation. The first OLED demonstrated by C.W. Tang and S.A. VanSlyke in 1987[117] consisted of a monopolar transport layer and an emissive layer sandwiched between a transparent anode and an alloyed metal cathode. Today’s OLEDs are a little more complicated: they make use of both singlet and triplet excitons for light generation[8, 34, 88, 101, 119, 124, 132, 135], quadrupling efficiencies relative to those devices relying on fluorescence alone[9]. Modern devices consist of an anode, hole-injection and hole-transport lay-ers (HIL and HTL, respectively), host and dopant materials, electron-transport and electron-injection layers (ETL and EIL, respectively), blocking layers and a cathode. The anode or cathode must be transparent to allow light emission, or both for a transparent display[44, 71, 76, 95, 97, 128, 136].

The reason for so many layers is to boost efficiency by increasing the probability of charges recombining and, in turn, generating light. When an external voltage is applied across the layers of an OLED, holes are injected from the anode and electrons are injected from the cathode. The positive and negative charges meet at a charge recombination zone (the emission layer) in the middle where light emission occurs. This is covered in greater detail in chapter 2.

1.2

Motivations for an Alternative Patterning

Pro-cess

Silicon is undoubtedly the most widespread semiconductor and technology would be nowhere close to where it is now without it. It’s manufacturability[92, 127], mobility, and cost have enabled everything from small photodetectors, thermal diodes, and microelectricalmechanical systems (MEMS) to processors for supercomputer nodes containing billions of transistors. However silicon is mostly limited to rigid integrated circuits and small devices. The class of high-performance crystalline semiconductors can make for highly efficient LEDs; to make an LED TV, on the other hand, costs would be prohibitively high in the case of individually wired devices, not to mention a single crystal substrate a square meter in area with millions of pixels.

Organic materials, on the other hand, are generally less expensive, but have the problem that most are sensitive to solvents. Every step of the lithography process

must be completely dry once any organic thin films are introduced2. This means that

a preliminary solution-processed step is allowable, but if you are trying to pattern multiple successive layers (like in the case of separate red, green, and blue subpixels), it can’t be a defining part of the patterning process. In the case of color-by-white, as developed by Kodak and Sanyo[122, 130] and used by LG[80] and eMagin[33], only one patterning step is necessary if at all.

The only OLED displays that have reached profitability are in the small and mobile display markets. While larger displays are available, quantities and choice of products are limited and prohibitively expensive. The prices of larger displays would come down if the generation of display production (which corresponds directly to the size of the mother glass that the displays are grown on) were able to increase. This is currently limited by manufacturing processes.

The main competitor to OLEDs is liquid crystal displays (LCDs)(which are made in production plants rated about five generations higher than those of OLEDs). While an inherently more complicated design (involving polarizers, color filters, spacers, and

liquid reservoirs[90]), their photolithography-compatible materials allow for simple scaling up of pixel electrode, transistor and color filter definition, thereby allowing larger mother glass production and consequently greater throughput[22, 63, 82]. The section that follows describes the current methods used to pattern organic electronic thin films.

1.2.1

Fine-Metal Masks

Current OLED patterning technology uses thin sheets of steel (see figure 1-2) that are used to define pixels during the thermal evaporation of organic semiconductors and metals. Placing these fine-metal masks (FMM) across sheets of glass like stencils to produce features on the order of 10 µm introduces complications.

Figure 1-2: A fine-metal mask used for arrays of OLED subpixels[131] (top) and some of the masks we use in our group(bottom).

FMMs are fragile and, due to the nature of the masking process, must be cleaned after a number of growths to prevent defects caused by debris. This puts stress on the features of the mask and can result in tears and bending. They cost on the order of $200,000 and need to be replaced every one to two months. Moreover, temperature deviations during the deposition can cause thermal expansion and contraction limit-ing the feature size[114]. Because the mask is not projected, but instead renders a 1:1 reproduction, incorporating multiple masks in a fabrication line requires tedious alignment[39]. This puts a bottleneck on throughput that, if eliminated or improved upon, would greatly reduce costs.

One temporary measure introduced to reduce the complications of a FMM that spans an entire mother glass is to step a smaller area mask across for multiple material depositions. This is termed “Small Mask Scanning technology” by Samsung[108]. The mask-area-limiting factor becomes only the size of an individual display, but the takt time is increased by at least the number of displays per mother glass over a mother glass-spanning mask. In addition to this increased takt time, the accumulated layers would cause a mask to wear out sooner due to the higher exposure of incident metal and semiconductor sublimation per mother glass.

1.2.2

Inkjet Printing

Inkjet printing is another potential method for full-scale fabrication of organic elec-tronics. The oft-touted benefits include production without vacuum equipment, low operating costs and high material-use efficiency[52]. An inket head operating on an XY-stage may also allow for simple pattern programming.

Unfortunately, inkjet printing has its own limitations. Such an inherently se-rial process requires an enormous number of printing heads to allow for reasonable throughput[10, 113]. Even then, the organics in such a system must usually be put in solution before being printed requiring the replacement of many industry-leading

small molecule organics with molecules with poorer performance3_{. Printing solutions}

3_{It should be noted that this material limitation is not the case for molecular jet printing}

Figure 1-3: Inkjet printing deposits droplets of dissolved organic compounds that leave a solid thin film after the solvent evaporates.

also have issues with drop uniformity due to surface tension during the evaporation of the solvent[96][27][39], but this is actively being researched.

In late 2013, Kateeva announced their YIELDjetTM _{inkjet manufacturing}

equip-ment specifically for OLED processing.[62] YIELDjetTM is the culmination of years

of research focused on material and environmental purity, film uniformity, and device and processing reliability. The intention is compatibilty with Gen-8 manufacturing lines and Kateeva claims that their current Gen-8 unit, shown in figure 1-4, is capable of covering a full mother glass with appropriately patterned organics in five minutes’ time. At the time of writing, no consumer products have been released, but the technology shows much promise.

Figure 1-4: Kateeva’s YIELDjetTM _{Gen-8 inkjet manufacturing equipment[62].}

1.2.3

Photolithographic Processes Involving Solvents

Photolithography is widely used for patterning in the semiconductor industry, but is difficult to apply to the production of organic semiconductor devices. While solution-processable polymer devices have been photolithographically patterned[72, 133, 134], the resists, solvents and etchants are chemically incompatible with all small-molecule

organic semiconductors; in our group’s experience, some common compounds (Alq3,

TPBi, and BPhen to name a few) have proven soluble in the fluorous solvents used in orthogonal photolithography employing hydrofluroethers. More general photolitho-graphic processes have required an intermediary polymer barrier [28] or the use of

super-critical CO2 [56]. The work proposed here investigates the use of a few dry

chemically inert resists that can be deposited and patterned in situ.

1.2.4

Other Ways to Pattern Organic Thin Films

Many other patterning methods have been developed and tested in universities and industry. Some, like in the case of laser-induced thermal imaging (LITI)[61, 73, 77, 78], have been seriously considered for use in full production lines. For more details

and more general information on patterning and manufacturing OLEDs, I recommend Martin B. Wolk’s chapter Patterning of OLED Device Materials in [60].

Chapter 2

Physics of Organic Materials and

Devices

S

weight compounds consisting mostly of carbon and hydrogen with nitrogen, oxygen, sulfur, and/or some heavy metal or metals. Because of the strength of the intermolecular bonds, they tend to form amorphous morphologies as opposed to crystalline ones. This weak bonding also results in the electronic and photophysical properties for a solid being similar to those of the single molecule. This also results in low evaporation points. These two characteristics allow one to rather easily make heterojunctions of multiple different materials that exhibit properties of the individual materials.

2.1

Electrons, Excitation, and Spin

Electrons are not arranged the same way on molecules as they are in atoms because molecules are made up of arrangements of multiple atoms of various elements in relatively complicated geometries. Instead, electrons in a molecule are considered to be in molecule orbitals[85] that describe the locations around a molecule that help one determine overlap of orbitals between adjacent molecules, approximate energy levels of a molecule, and determine what kinds of interactions are possible. There are a

number of mathematical formulations that allow predictions of values and properties of a molecule like density functional theory (DFT)[54, 68, 94] and the Hartree-Fock methods[35, 36, 50, 115] for example, the details of which are outside the scope of this work and make up an entire field of research on its own. While singlet fission and charge dissociation are not the focus of this work, figure 2-1 exemplifies the value of these calculations.

Figure 2-1: Figure from [5] showing results of density function theory(DFT)

calcula-tions of pentacene and fullerene C60 exciton and charge-transfer states.

When an incident photon of adequate energy is absorbed by an organic molecule, an electron is excited creating an electron-hole pair called an exciton. This is in contrast to what happens in common inorganic semiconducting materials like silicon where you’d almost immediately have a free charge. This is because of the relative

permittivity r being so much lower in organics. The high dielectric constant of

inorganic materials makes for binding energies so small that room temperature phonon vibrations are sufficient to dissociate the excitons formed therein. The reverse of dissociation is also possible, in which case an electron and hole combine to form an exciton.

Excitons

The wavefunction describing two electrons must be antisymmetric. Thus, if the spins of those two states are symmetric, the spatial factor of the wavefunction must be antisymmetric. The greater distance associated with this spatial portion means the electons shield eachother less and are as a result lower energy, more tightly-bound states. These are known as triplet excitons as there exist three unique possible spin

pairings as opposed to singlet excitons that have only one[98]. Singlet radiative

relaxation is known as fluorescence whereas triplets quantum-mechanically forbidden relaxation is known as phosphorescence[79, 98]. These states importance in OLEDs is discussed further in the next section.

When considering a molecule’s interaction within a device or system, the two orbitals at the cusp of the ground state and the excited state are considered: the highest occupied molecule orbital (HOMO) and lowest unoccupied molecular orbital

(LUMO). They serve as a guide to electron and hole transport1 _{in a molecule and}

device as well as the excitation energy of the molecule. The difference between the HOMO and LUMO is often referred to as the band gap and this level is the sum of the optical gap and binding energy (which, itself, is made up of the electron Coulomb interaction and exchange energy)[67, 98].

2.2

The OLED: Turning Charge into Light

As touched on in the introduction, an OLED is a suitably arranged stack of metallic and organic thin films enabling the injection of charges into an organic layer that later radiatively relax, i.e., photons are emitted from these induced excited states. Which particular layers are used is determined by examining the HOMO and LUMO of the emissive material. In order to get electrons and holes into this material, it is necessary to use charge transport materials with a low enough LUMO and a high enough HOMO (both in terms of energy), respectively, paired with appropriate work-1_{The HOMO and LUMO levels are for this reason sometimes referred to as organic transport}

function electrodes for injection of each corresponding type of carrier[8, 117]. It isn’t as simple as that, though, as a low-energy transport material might serve as a loss mechanism if excitons would find lower energy states in it or if the transport material isn’t transparent to the EML’s emitted light. It is for this reason that wide-gap blocking layers are sometimes used[9] .

In essence, it is important to confine excited states in the emissive material for an OLED to be efficient[117]. Figure 2-2 shows a generalized energy level diagram of an OLED with the path of each carrier from electrode to emission. The electron and hole transport layers are often abbreviated as ETL and HTL, respectively, and most modern emitter materials are grown dilute in a wide gap host to prevent annihilation effects stemming from aggregation[102, 103, 104]. The emissive layer, being either a single material or an emitter doped into a wide-gap host, is abbreviated as the EML.

E

n

er

gy

hv

+

HTL

ETL

Anode

_

EML

*

Cathode

Host

Figure 2-2: Organic transport level diagram showing the electroluminescence pro-cess. The hole and electron are injected from their corresponding electrode and move through their transport layers to meet and form an exciton. The exciton recombines dissipating its energy through a photon.

Excitons formed from injected charges follow the statistics of forming three triplets for every singlet. This limits the internal quantum efficiency of fluorescent emitters

to ∼25%[8] because the triplet states don’t emit. In 1998 Baldo et al. showed that this limit could be raised to 100% by making use of phosphorescent emitters that emit from triplet states[8, 9]. These highly efficient phosphorescent emitters are the

type that you’ll find in modern red and green subpixels of OLED displays2_{. Almost}

fifteen years later, Uoyama et al. showed that similar efficiencies can be realized in fluorescent emitters that reverse-intersystem-cross triplets back to the singlet excited

state enabled by a small singlet-triplet gap ∆EST[119, 135]. This mechanism, termed

thermally activated delayed fluorescence (TADF), is shown schematically in figure 2-3 along with some of the molecules that exhibit it.

Figure 2-3: (a) Thermally activated delayed fluorescence (TADF) energy diagram. (b) Some of the molecules that Uoyama et al. presented in their seminal TADF paper that this figure is borrowed from (Reprinted with permission from [119]. Copyright 2012 Nature Publishing Group.).

To achieve such low ∆EST (≤ 100 meV), the single molecule is made up of both

donor and acceptor groups. While not necessarily faster or more efficient than phos-2_{Blue phosphorescent materials are, as of the time of writing, not yet efficient or stable}

phorescent materials, TADF materials have a commercial advantage in that they don’t make use of rare and expensive high spin–orbit coupling heavy metal cores like the iridium found in Ir(ppy)33.

3_{The molecular structure of Ir(ppy)}

Chapter 3

The Sublimable Mask Lithographic

Process

T

detail. While the idea and work here were developed independently, it is important to mention that similar methods were investigated by IBM in the late seventies[59] and early nineties[25] and by the Harvard Nanopore Group at Harvard University more recently[65][14][48]. The work described here is the first time that the dry nature of the method has been explored for use in patterning organic semiconductor devices.

Each of the aforementioned works suggest or make use of a condensed gas as a resist that is patterned, deposited on, and subsequently heated to induce lift-off. The Harvard Nanopore Group used water as their resist material as it was inexpensive and has a reasonably attainable freezing point. They demonstrated nanometer-scale definition as shown in Figure 3-1. Unfortunately, water is one of the main materials responsible for OLED degradation[16, 57, 110, 129] and an alternative sublimable resist material is required.

Figure 3-1: An SEM image of The Harvard Nanopore Group’s demonstrated pat-terning of chromium using water ice resist. Reprinted with permission from [65]. Copyright 2005 American Chemical Society.

3.1

Theory of Practice

Below a certain pressure and above a certain temperature, materials sublime: their thermodynamically distinct phase transitions directly from solid to vapor without passing through a liquid phase. Deposition is the reverse transition. A phase diagram showing this region is seen in figure 3-3. As previously mentioned, a dry process is necessary for the production of organic semiconductors due to their sensitivity to traditional solvents. Depositing a material on a substrate whose phase can be easily changed between solid and vapor would make a versatile, clean and dry mask suitable for patterning.

The heat delivery requirements Q to sublime a solid of mass m, with specific heat

capacity cV, and heat of sublimation hs are

Q = m Z Tsub

Ti

cVdT + mhS[38] (3.1)

= m(cV∆T + hS) (3.2)

Because the change in temperature ∆T (the change from the initial

tempera-ture Ti to the sublimation point Tsub required to bring the mask to the sublimation

point is no more than 30◦C, cV is much less than its room temperature value of 37.1

J·mol−1K−1[81] for the temperature regions of interest here and hS is 26.1 kJ·mol−1[1]

Vol. 24 • No. 46 • December 4 • 2012

www.advmat.de

D10488

Figure 3-2: Inside cover of December 2012 issue of Advanced Materials featuring sublimation lithography[7]

rather than temperature increase, so

Q = m(cv∆T + hS) ≈ mhS (3.3)

3.2

Process Flow

In practice, there are other necessary considerations. Heat exchange with gas molecules at higher pressures causes the temperature to significantly fluctuate. The mask ma-terial would be very ineffective at the sublimation point due to the heat capacity and heat of deposition. The system temperature must therefore be kept significantly below this critical point for the resist material to be perform as intended. To raise the temperature of the sublimation point, the mask is deposited at a higher pressure to ensure that a higher portion of the gas is solidified. The pressure is later reduced for the deposition of the thin film. Figure 3-4 shows an updated phase diagram to reflect these details.

Temperature

P

re

ss

u

re

Solid

Liquid

Gas

Figure 3-3: General phase diagram. The arrow shows the ideal region of operation with respect to sublimation lithography. Deposition of the mask takes place along the low temperature region of the arrow and sublimation and lift-off take place along the higher temperature region of the arrow.

With the substrate at higher pressures and low temperature (so as to be well inside the solid portion of the phase diagram), the resist gas flows at and across the substrate where much of it deposits on the surface (part (a) and (b) of figure 3-5. Next, the mask is selectively patterned exposing the substrate to allow for the subsequently deposited neat films of the desired material to remain as intended post lift-off (part (c)). The thin film is then grown over the mask at high-vacuum (part(d)). Following this, the substrate temperature is brought up to sublime away the mask which carries the undesired regions of thin film along with it (part(e)). These steps can be performed again and again as needed to build devices that require multiple patterned layers.

Temperature

P

re

ss

u

re

Solid

Liquid

Gas

Figure 3-4: Phase diagram reflecting full process flow parameters. The defining of the mask pattern is performed immediately prior to the deposition of the thin film.

Cooled

(a) (b) (c)

(d) (e) (f)

Glass Phase-change resist Organic 1 Organic 2 Metal

Figure 3-5: Simplified process flow for sublimation lithography. (not to scale) (a) Begin with a cooled substrate to facilitate resist deposition. (b) Deposit resist. (c) Selectively pattern resist. (d) Deposit desired thin film. (e) Lift-off resist leaving patterned thin film. (f) Repeat as necessary to complete device.

Chapter 4

Sublimable Mask Patterning

T

stamping. Photolithography is a third seemingly more elegant method and is considered, but found to be impractical for the proof-of-concept material used.

4.1

Considerations for Demonstration

A small research-scale investigation is necessary to gauge the practicality of a full-scale sublimation patterning implementation. The basic principals of operation must be demonstrated and fundamental limitations must be determined. If realizable, the accuracy, repeatability, and resolution must be evaluated. Commercialization would require evidence of 10 µm or smaller edge definition, as mentioned previously, as well as some proof or arguments for yield and high throughput.

This work is a proof of principle for the fabrication of small-molecule organic semiconductor devices. An existing thermal evaporator served as a testbed for the original concept and all deviations therefrom. It was modified and refashioned as necessary to accommodate improvements upon steps of the process or to approach parts of the process from different angles. Full proof of commercializability would require a full-scale evaporator built from the ground up with sublimable mask lift-off patterning in mind, rather than a small research-scale system.

4.2

Experimental Setup

The experimental setup had to allow sufficient control of the process parameters outlined in section 3.2. We have a thermal evaporator attached to a glovebox, designed and built by Angstrom Engineering, that was easily modified to meet the process’ needs. The dry and inert nitrogen environment of the glovebox is crucial to avoid water condensation on the substrate.

No substrate cooling was in place, so the rotating substrate holder was initially replaced with a liquid nitrogen reservoir. A flange with tapped holes allowing for a mounted substrate holder made up the bottom of the reservoir so the holder could

be in nearly direct contact with the liquid nitrogen. Temperatures as low as -165◦C

were measured at pressures of 10−6 Torr with a type K thermocouple cemented to the

surface of a reference substrate. The difference between the surface temperature and

that of boiling liquid nitrogen (-196◦C) is attributed to the thermal conductivities of

glass and steel. Indium foil inserted between each mating surface improves thermal conductivity. More traditional cryogenic thermal greases were abandoned because of their high photoluminescence that made it difficult to characterize patterned films and the grease’s difficulty to cleanly work with in the glovebox.

Four feedthroughs (figure 4-1) allowed for observation and finer control of the frozen mask: a feedthrough for the aforementioned thermocouple, a four-pin elec-trical feedthrough, a 1/4" steel tube feedthrough, and a universal serial bus (USB) feedthrough. These allowed for temperature measurement, current sourcing (which will be elaborated on in 4.2.1), gas flow for resist deposition, and use of USB periph-erals, respectively. A USB hub with an attached light and webcam allowed for in situ optical observation.

A substrate holder was machined out of aluminum. Holes in the corners allowed for direct mounting to the bottom flange of the reservoir, while clips for securing the substrate and any electrical probes were located along the perimeter, as seen in figures 4-2 and 4-5. The center, though, was left in tact and smooth to minimize thermal resistance.

Figure 4-1: The side of the evaporation chamber showing the feedthroughs and liquid nitrogen reservoir.

Just below the holder, a copper tube is connected to the 1/4" steel tube feedthrough

and aimed at the substrate for CO2 gas flow. The camera is clipped to this tube out

of the way of organic material sources and thin film thickness sensors. A mass flow controller is connected on the outside end of this tube for gas flow rate control.

Carbon Dioxide as a Resist

Carbon dioxide seemed like an obvious choice for a sublimation resist. It is chemi-cally inert, and its sublimation point is easily accessible with a thermal evaporator equipped with liquid nitrogen cooling and simple heaters. As phase data was not readily available for the low pressure and temperature regime of this study, we ex-trapolated existing data with a Clapeyron equation-based thermodynamic fit to better

understand how CO2 would behave in our setup[112]. CO2’s phase diagram and our

extrapolation with pressure and temperature extremes of the experimental setup is shown in figure 4-3.

Figure 4-2: Flange making up the bottom of the liquid nitrogen reservoir. The substrate holding plate is bolted in with a glass sample tightly secured. The washer-mounted thermocouple can be seen towards the top of the image washer-mounted with a bolt.

The pressure of the chamber and temperature of the reservoir aren’t the sole

factors determining the phase of the CO2 resist. When a thin film is being deposited

onto the surface of the resist it must come to thermal equilibrium. Some of the resist will warm up and sublime during this process. Based on the development in section 3.1, the thickness of resist required to survive the deposition of a particular thickness

of a thin film with enthalpy of fusion hf and vaporization hv can be approximated:

mCO2(cv∆T + hS)CO2 = mf ilm(cv∆T + hv + hf)f ilm (4.1)

Instead of using mass m, the density ρ can be used with the volume being made up of area A and thickness t of the film to provide the thickness per unit area:

ρ = m

V → m = ρAt (4.2)

0

50

100 150 200

Temperature / K

P

ressu

re

/T

or

r

10

0

250 300 350

CO

₂

Solid

Vapor

Vapor

Solid

Liquid

H

₂

O

10

-5

10

-10

10

5

Figure 4-3: Phase diagrams of CO2 and H2O extrapolated (broken lines) from data

(solid lines) in references [83] and [120]. The “×” represents the process

operat-ing point of 77 K at 106 _{Torr. The curves are extrapolated using the Clapeyron}

equation[112]. Figure from [2]

Then the thickness of CO2 given a desired thin film thickness is:

tCO2 =

ρf ilmtf ilm(cv∆T + hv+ hf)f ilm

ρCO2(cv∆T + hS)CO2

(4.4)

For 100 nm of silver, this is ∼3 µm and for 100 nm Alq3, this is ∼400 nm (Alq3’s

hv, hf, and cv approximated with water’s values as thermal properties are not

read-ily available). This approximation does not include the active cooling of the liquid nitrogen reservoir or the heating from the evaporation source’s blackbody radiation. The density of the resist depends on pressure, temperature[109], and gas flow rate and, as mentioned in reference [46], a more amorphous resist avoids inhomogeneity at the length scales of the crystalline domains and is preferred for greater resolution.

see the Supporting Information of [2] for Allen Yin’s description of the interferometric technique employed to measure density and film growth rates[32, 118].

Once the CO2resist is patterned by selective sublimation, it is important to control

the partial pressure of CO2 in the chamber to prevent unwanted re-condensation of

CO2 vapor on patterned regions of the substrate. It is also possible to freeze other

impurity gases onto the substrate, notably H2O, whose phase diagram is also shown

alongside that of CO2 in figure 4-3[120]. In previous studies of frozen CO2 films at

10−7Torr, Gerakines et al. measured a water deposition rate of 2 nm h−1[43]. At these

rates re-deposition must be considered in our experiments, but should ultimately be of little consequence in high throughput manufacturing since the acceptable background

pressures of CO2 and H2O increase with reduced takt time.

4.2.1

Patterning the Resist with Resistive Heating

We selectively patterned the CO2 mask by resistively heating a photolithographically

predefined conductor on the substrate surface. OLEDs always feature a bottom elec-trode layer such as the transparent conductor indium tin oxide (ITO), a thin metal, or a transistor, so we considered ITO a realistic testbed.

The power P through any circuit element is given by

P = IV (4.5)

where I is the current passing through the element and V is the voltage across that element. By Ohm’s law,

V = IR (4.6)

where R is the resistance of the circuit element. The resistance depends on the resistivity ρ, the area A and the length l by the relation

R = ρl

A (4.7)

of CO2 of mass m is Q ≈ mhS = Z t0 0 I(t)V (t)dt = Z t0 0 I2(t)Rdt = Z t0 0 I2(t)ρ l Adt (4.8)

As the ITO is essentially uniform in thickness D across the substrate and with pre-defined lines equal in width, this can be rewritten in terms of heat for a constant thickness per unit width:

DQ w ≈ D mhS w = ρl Z t0 0 I2(t)dt (4.9)

This leaves the current I(t) as the only active process parameter, which can be run constant or pulsed accordingly to heat up and sublime away a desired mask region.

The ITO pattern used to test this selective sublimation method is displayed in figure 4-4. Three 100 µm-wide lines allow for three 100 µm-per-side square subpixels

given a 100 µm-wide horizontally1 _{aligned top electrode. The six contact pads are}

wider, both to facilitate electrical contact and so current passed through a line will mostly heat the regions of highest resistance. The horizontal pad and center-right island allow for contact with a cathode laid across the top of all three subpixels. In this work, electrical contact in this work was made from the electrical feedthrough to the substrate’s ITO pads in one of two ways: copper foil tape or small probe clips (as seen in figure 4-5).

After a blanket deposition of CO2 at 100 Torr and ∼-165◦C the chamber was

pumped down to ∼10−5 Torr. At this temperature, the resistance of a single 100

µm-wide line of the ITO used was about 500 Ω. 100 mA (current density of about

625 kA cm−2) was pulsed through the line with a Keithley 2400 sourcemeter while

the mask’s selective sublimation was observed via the camera. 20 nm of either

tris(8-hydroxyquinolinato)aluminium (Alq3),

4-(Dicyanomethylene)-2-tert-butyl-6-(1,1,7,7-tetramethyljulolidin-4-yl-vinyl)-4H-pyran (DCJTB) (see Appendix A) or silver was then thermally evaporated onto to the substrate (the choice of materials is explained below). We then vented the chamber to atmospheric pressure, and heated the sample

Figure 4-4: The pattern of the ITO used in resistive heating experiments. The black regions are ITO; the white regions are bare glass.

with a thin ∼100 W kapton heater (Omega Engineering) sandwiched between the

substrate holder and liquid nitrogen reservoir. This forced the lift off of the CO2

mask and the undesired regions of the thin film. This was observed by video to confirm complete sublimation. The pressure and temperature changes here reflect those in figure 3-4.

Alq3 was chosen because it is a very well known OLED material (it was part

of Tang and VanSlyke’s first devices[117]). It is also has the advantage of obvious photoluminescence. DCJTB is another highly photoluminescent material that has

been used as a dopant with Alq3 to produce red OLEDs[18]. An ultraviolet lamp

highlights the pattern resulting from the experimental procedure above, as in figure 4-7. Because the copper foil tape used to make electrical contact floats, it is poorly

cooled, so the adjacent regions show signs of CO2 mask degradation. Figure 4-8 shows

Figure 4-5: A mask of CO2 patterned for a 100 µm-wide line. The substrate holder

is rigged with probe-tipped clips for electrical contact.

fuzzy and the lines are not exactly 100 µm wide. In large part this is because of the difficulties in fine calibration of power delivery by camera-mediated observation. A specially designed system would ameliorate the problem and that is what led to an

160nm

Glass Substrate

ITO

CO

Copper

foil tape

Figure 4-6: In situ photograph and schematic representation of the resistive heating method of patterning the frozen mask. Electrical contacts on the substrate allow

current to be driven at a current density of 625 kA cm2 _{through a 160 nm-thick strip}

of ITO on the substrate. The red arrows in the schematic drawing indicate the flow of current through the ITO enabled by affixed copper foil tape[2].

Figure 4-7: An exmaple of photoluminescence from a 100 µm-wide line of

DCJTB-doped Alq3(top), and photoluminescence(bottom-left ) and a

Figure 4-8: A micrograph from a 100 µm-wide line of silver patterned by resistive heating of ITO

4.2.2

Patterning the Resist with a Stamp

We next investigated selective sublimation through direct stamping. A

micron-featured stamp could heat specific regions of the mask through contact thereby pat-terning the mask. A stamp-specific version of the process flow is shown in figure 4-9. We chose the epoxy-based SU-8 photoresists (inspired by [86]) to take advantage of their versatility and rapid-prototyping capabilities: structures with thicknesses rang-ing from 1-600 µm with aspect ratios as high as 20:1[116] that can be made in a matter of hours. Cooled (a) (b) (c) (d) (e) (f) Organic Materials Glass Frozen CO Metal 2 SU-8 2150 (g)

Figure 4-9: Simplified process flow for sublimation lithography using a frozen CO2

resist (not to scale). (a) Cool substrate below 100 K. (b) Freeze on thin film of CO2.

(c,d) Pattern CO2 film by heating selectively. Here a stamp is pressed against the

resist to remove some areas. (e) Deposit desired organic or metal thin film by thermal

evaporation. (f) Warm substrate to sublime CO2, thus lifting off unwanted material.

(g) Repeat steps (a-f) as necessary to complete the device[2].

With stamping, the thermal energy required to sublime the CO2 mask is, from

equation 3.3,

Q = m(cV∆Tstamp+ hS) ≈ mhS = mstampcV stamp∆Tstamp (4.10)

Here ∆Tstamp does not reflect the difference of initial temperatures between the mask

and stamp, but rather the temperature change of the stamp after transferring energy to sublime away the mask. The heat transfer is a function of the stamping pressure and time, and if it is too great will sublime away necessary parts of the mask. To

develop this further, we could use Fourier’s law of conduction[70]

q = −k∇T (4.11)

with the relation for thermal diffusivity[70]

α = k

ρcV

, (4.12)

where k is the thermal conductivity and ρ is the density. But until we can grow a consistent resist layer and more accurately measure the thickness of the resist—and thereby estimate the surface area—this exercise is not illuminative. An investigation into the thickness of these resists under similar conditions can be found in Allen Yin’s related work in the supplementary materials of the publication in Advanced Materials[2].

A 2D simulation of the thermal diffusivity is shown in figure 4-10 and provides a guide as to how the heat from the stamp moves once contact is made. The sim-ulation expands on a unitless script written in MATLAB[99] to take into account physical units, to match the diffusivity of the glass substrate as well as the length, and time scales of interest. The method is a finite-difference approximation of the heat diffusivity equation:

∂T

∂t = α∇

2_{T (4.13)}

where α is as in equation 4.12, T is temperature, and t is time.

A few assumptions were made to reduce computation complexity and appropriate boundary conditions were set. The bottom edge is set at liquid nitrogen’s boiling point of 77 K, and all other edges have effectively pure vacuum insulation. The top middle surface has a length held at 300 K to act as the stamp. The stamp temperature does not drop and the thickness of the substrate is much less than it is in the experiments, but should provide a sense of the movement of heat in the stamping progress. The enthalpy of sublimation would also dissipate a great deal of heat from the stamp as shown in equation 4.10. The iterations stop once the temperature no longer changes

by more than 1% of the maximum temperature.

Tempe

rature

/ K

Figure 4-10: 2D simulation of thermal diffusivity in the stamping process.

We used SU-8 2150 obtained from Microchem to take advantage of its high thick-ness and aspect ratio potential—if the height of the features are not sufficiently taller than the resist thickness, we’d risk subliming away the entire mask. The resist was spun on solvent- and oxygen-plasma-cleaned silicon at 3000 rpm. Contact pho-tolithography resolved ∼115 µm-tall pillars that tapered slightly after development in propylene glycol methyl ether acetate (PGMEA). For these experiments, the tapering is not so severe as to interfere with patterning as the resist thickness is on the order of 50 µm. A micrograph of these features can be seen in figure 4-11. After growing a

blanket of CO2 on the cold substrate, the stamp is pressed into the frozen CO2 either

by hand or by a linear solenoid mounted on a rotational feedthrough. After stamping,

the thin film is deposited at pressures of ∼10−5 Torr. The mask is then lifted off,

ideally taking the undesired regions along with it. The results of these experiments demonstrate the need for some optimization, as can be seen in figure 4-11.

Figure 4-11 suggests that either the heating in this configuration fails to fully lift-off or—more likely— the stamping methods employed lack adequate feedback to control the definition. Our hands may slip, and our solenoid-based mechanical system

Figure 4-11: A micrograph of the 115 µm-tall pillar SU-8 stamp(top) used to pattern circles of Alq3(bottom)

is not sufficient so as to prevent twisting and shifting that would thin or obfuscate the mask. The particular stamp’s features may also have been too short to make contact with the surface of the substrate without touching the rest of the mask.

Colder and with finer control

The next step was to gain better control of temperature, observation, and stamping. This redesigned setup is shown schematically in figure 4-12. A cryogenic pump was repurposed for use as a cooling source and all components are mounted onto it via an

oxygen-free high-conductivity (OFHC) copper rod. This gave us not only consistent and reliable temperatures without having to refill boiling liquid nitrogen, but also added temperature control in the previously unexplored region from around 10 K to 80 K. All cold parts are machined out of OFHC copper and indium foil is sandwiched

between all temperature-critical interfaces. A kapton encapsulated heater placed

in between the substrate and substrate holder provides adequate local heating for encouraging lift-off without adding too much heat to the bulk thermal mass of the

apparatus. Below 80 K, a type K thermocouple is no longer of use so a silicon

thermal diode was attached to the copper substrate holder to approximately monitor the temperature of the sample and a cryogenic temperature controller (Lakeshore Cryotronics) was employed to manage operating temperature. To greatly improve observation of our patterning as well as to investigate the yield of patterned thin films, we added a digital microscope mounted on the stamp actuator. The microscope had both white and 405 nm LEDs so photoluminescence imaging was attainable in situ and we could also now record videos of lift-off in process.

CO2

Active

cooling

Stamp &

microscope

x & z stages

Cooled substrate

with underlying

heater

Thermal diode

high-vacuum

Evaporation source

Figure 4-12: Schematic of the updated experimental stamping setup showing the process-critical components as described in the main text. The entire setup is located

within the standard thermal evaporation chamber (Angstrom Engineering). The

arrows by the stages indicate their respective direction and range of motion. Both the stamp and the microscope used for positioning and stamping evaluation are located on the smaller stage as indicated[2].

As seen in figure 4-12, the camera and stamp were mounted to two stages provid-ing fine in situ control of motion. Stampprovid-ing location and efficacy could be evaluated in a matter of seconds and retried if insufficient stamping depth was observed. The repurposed cryogenic pump’s compressor and cold head made for a lot of vibration that made stamping during their operation impossible. Thus the compressor was briefly turned off during the actual stamping and fine observations so that the vibra-tions did not interfere while the stamp and resist make contact or while recording images. Because the cryopump’s cooling mechanism is closed-cycle rather than evap-orative like the liquid nitrogen cooling it was replacing, longer cooling cycles reduced throughput.

We also decided to redesign the stamp geometry in an attempt to demonstrate features found in the current state of the art. Figure 4-13 shows this design meant to meet the 325 pixels-per-inch in the tightest pitch commercially available product of the time—the Apple iPhone Retina display[2]. The material and process for fabricating the stamp remains the same.

Cooling down from room temperature takes a couple of hours depending on the target temperature, but with effectively two cryogenic pumps in operation lower

op-erating pressures as low as 10−7 Torr were attained. This isn’t necessarily good for

cleanliness because impurities, most importantly H2O, could be sticking directly to

the substrate rather than just the main cryogenic pump as it can become just as cold as the coldest object in the chamber. The steps that follow are analogous to those from the earlier stamping discussion with the added complexity of controlling the motorized linear stages. The X stage (150 mm travel) is moved to change between stamping and observation mode whereas the Z stage (30 mm travel) controls stamp-ing height or focus, respectively. Both of these components are controlled via USB using proprietary software for each.

After flowing the CO2 onto the cooled substrate and carefully incrementing the

displacement of the stamp up to the resist the sublimable mask is formed, displayed

in part a of figure 4-14. As before, the organic film, Alq3 here, is sublimed as a

Figure 4-13: Micrograph of a region of the SU-8 stamp used to selectively sublime

regions of the CO2 resist for 78 µm-pitch patterns. In the center of the image is a

pillar that has been knocked over displaying the profile of the pillar[2].

kapton heater to overcome the cooling of the repurposed cryopump rapidly induces sublimation of the resist layer. The subsequently deposited organic layer is lifted off revealing the patterned film as in figure 4-14b. Figure 4-14c is a photoluminescence micrograph of the obtained 325 pixel-per-inch density using a 380 nm LED with an optical microscope (Carl Zeiss AG. Axioskop). For clarity, a longpass filter (Thorlabs, Inc. FEL0450) was used to remove the pump from the image.

Figure 4-14: Optical micrograph (a) of a 78 µm pitch-patterned mask of CO2 defined

via contact with a stamp. The corresponding inset (b) shows the Alq3 thin film after

deposition and lift-off. Photoluminescence micrograph (c) from the same film[2]. A longpass filter is used to lessen pump light’s presence in the image.

The edge definition and profile is of key importance ensuring that the processing is suitable for device fabrication and that no roughness or debris is introduced by this process. A false-color surface topography obtained using a Veeco Instruments Inc optical profiler showing multiple pixels’ finer definition is displayed in figure 4-15(this was also measured and confirmed on a Veeco-Tencor contact surface profilometer). The bottom portion of the figure shows the cross section of a single pixel to examine the smoothness of the remaining structured film. The observed topography shows no problematic obstructions, roughness, or irregularities.

0

50

Height / nm

Distance / µm

0

50

200µm

230µm

Figure 4-15: A false-color topography obtained by an optical interferometer (top) and a cross section of the same (bottom) detailing the profile of the patterned organic pixels of figure 4-14[2].

It’s observed that the relative heat capacity of a stamp maintained at room tem-perature is more than sufficient to rapidly remove the frozen resist, taking mere seconds to clear. To prevent abrasion and dust formation, the surface of the stamp need not make contact with the hard substrate surface if a universal burn-off step is performed to uniformly ‘etch’ the residual resist[48]. In this step, the temperature of the sample and subsequently the resist is warmed up. A heater below the substrate

or one embedded in the substrate holder is used to combat the cooling mechanism. This uniformly removes a suitable depth of all the resist such that none remains in the areas where the desired thermally deposited film is to remain.

With low repeatability and yield from the experimental setup, it was not possible to generate more than a few subpixels of successive layers at a time. Despite this, the photoluminescence micrograph shown in figure 4-16 demonstrates the physical capability of patterning more than one organic layer of different compounds one after

the other. The intended result was an array of Alq3 subpixels as shown in figure

4-14 followed by an array of Alq3 doped with the red dye DCJTB shifted 26 µm to

the side as would be seen in the first two color subpixel arrays of an RGB display. Complete lift-off would not occur within the same area of the substrate for each color film and we attributed this to insufficient flat levelling of the stamp with respect to the substrate. The repeated red-green side-by-side subpixels seen only on the edge of the substrate as seen in figure 4-16 support this idea because the edge would be the only axis to make contact if the two interfaces do not share a normal. This is not inherently a problem with using sublimable masks, but rather that the stamping tools used in this study were not optimized. This is an essentially solved problem as is seen by the success of soft lithography, nanoimprinting, and sophisticated contact alignment[5, 15, 21, 45, 84, 105, 126].

Figure 4-16: Examples of red and green subpixels side-by-side. The left two images show actual red and green subpixels next to each other whereas the right image exemplifies some of the complications that arose in the attempts.

Free-floating thin film tensile strength

A repeatable yet quite unexpected phenomenon that was a common observation in both successful and failed lift-off attempts was the ability of the organic thin film to remain as a flexible continuous film. The organic film held together as it floated freely folding away from the substrate with only a small portion of it still attached to the substrate. We have recorded many instances of this in both white and UV light

micrographs and videos. Figure 4-17 shows a 50 nm-thick film of Alq3 lifting away

from the substrate yet remaining intact. It measures about 150 µm on a side. As noted in chapter 2, it is a bit surprising at this scale that the weak Van der Waals forces that bind organic molecules can hold the film together.