University of Southern California Dissertations and Theses

Hybrid vat photopolymerization processes for viscous photocurable and non-photocurable materials

(USC Thesis Other)

Hybrid vat photopolymerization processes for viscous photocurable and non-photocurable materials

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

HYBRID VAT PHOTOPOLYMERIZATION PROCESSES FOR
VISCOUS PHOTOCURABLE AND NON-PHOTOCURABLE MATERIALS

by

Yang Xu

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the Requirements for the Degree
DOCTOR OF PHILOSOPHY
(INDUSTRIAL AND SYSTEMS ENGINEERING)

December 2021

Copyright 2021 Yang Xu
ii

Acknowledgments
Why do I wish to pursue a Ph.D.? For an answer, I have to go back to the 1990s. When I was
a little child, my grandfather told me many stories about Confucius (his family name is Kong, the
same as Confucius) day after day, and I believe in the stories. Since then, when asked what you
are going to be in the future, I will unhesitatingly reply that I want to be a professor. I think this is
the start of pursuing a Ph.D. On the last day of the year 2020, my grandfather, the one who helped
me set up this goal, passed away. Though I knew the day might come since I determined to study
abroad, I still feel sorrow for his death and his pity that he cannot see his grandson at the last
moment. I wish his soul rest in peace in heaven, and I can tell him that the dream you gave me
comes true one year later.
Do I ever think about giving it up? During the past five years, I have come up with this
question many times when I experience lonely, nervous, anxious, and furious feelings periodically.
Those feelings were intensified due to the time limit, shattering myself into pieces like a poisonous
snake. Even so, the answer is definitely not each time. Those negative motions were finally
overwhelmed by calm and invigorated feelings, for my parents always support me spiritually and
financially. They are the most powerful support to encourage me to follow my dreams. Especially
my mother, who is the greatest mother in the world, is the only one who thinks her son will succeed
even god doubts it.
Time flies. Five years just passed for my Ph.D. study. Doing a Ph.D. is truly a life-changing
experience for me, and it would not be possible without the support and guidance that I received
from many people.
iii

Firstly, I would like to express my sincere gratitude to my advisor, Professor Yong Chen, for
giving me a chance to chase after my dream and continuous support of my Ph.D. study. The weekly
individual meeting with him enabled me to work out all the challenges I have ever met in my
projects and help me form a scientific view of research. His guidance is valuable and will benefit
me all my life. In a word, my research life starts from here.
In addition to my advisor, I would like to thank the members of my Ph.D. qualifying exam
committee and Ph.D. defense committee: Prof. Qiang Huang, Prof. Satyandra K. Gupta, Prof.
Hangbo Zhao, Prof. Wei Wu, and Prof. John Carlsson, for their insightful comments and
encouragement.
I also want to express thanks to my fellows in the lab, Huachao Mao, Xiangjia Li, Jie Jin,
Yang Yang, YuenShan Leung, Han Xu, Laiming Jiang, Bin Yao, Yizhou Jiang, Yeowon Yoon,
Zhonghao Han, and Manqi Li. The research life has never been boring because of them.
I also gratefully acknowledge the assistance of my master and undergraduate assistants, Siyu
Gong, Ziqi Wang, Yizhen Zhu, Fangjie Qi, Xiangyun Gao, Yujie Shan, Jingwen Gong, Yun Zhou,
and Zhuangboyu Zhou. Without their efforts, I could not finish all the 3D printing experiments on
my own. Their help saved me a lot of time. I feel proud of them, for most of them got a Ph.D. offer
after graduation.
During this global COVID-19 pandemic, thanks should also be given to the medical staff on
the front lines, who are protecting this world day and night so that I can safely write this
dissertation at home.
Last but not least, I would like to thank myself and my future girlfriend. Thank you, Yang,
for your tenacious will and outstanding determination, for your genius can handle all the research
iv

problems I have met, for not ceasing to fight against all the difficulties. Thank you, my lovely
girlfriend. Even though I do not know where you are, you are out there, waving to me. With the
hope that you will be the best gift that God gives me after all the trials and tribulations, I wrote
down this dissertation.
v

TABLE OF CONTENTS
2.1 Reduction of separation force in vat polymerization process ................................. 10
2.2 Designs and optimization techniques of support structure ..................................... 15
2.3 Droplet dispensing techniques for viscous material ............................................... 18
2.4 3D printing processes to achieve microfluidic channels ......................................... 20
3.1 Statement of Problems ............................................................................................ 24
3.2 Hypotheses .............................................................................................................. 25
4.1 Reusable support for additive manufacturing ......................................................... 28
4.1.1 Reusable metal support design ........................................................................ 30
4.1.2 Layout optimization ......................................................................................... 37
4.1.3 Toolpath of residual printed support ............................................................... 44
4.1.4 Experimental validation ................................................................................... 46
4.1.5 Summary of this work ..................................................................................... 52
4.2 A vibration-assisted separation method for constrained-surface-based VPP ......... 53
4.2.1 Analytical model of the vibration-assisted separation method ........................ 54
4.2.2 Design of VPP system with VA separation mechanism .................................. 57
4.2.3 3D printing process with the VA separation method ...................................... 58
4.2.4 Experimental and Analysis Results ................................................................. 60
4.2.5 Process Settings and Comparison Study ......................................................... 72
4.2.6 Summary of this work ..................................................................................... 77
4.3 In-situ-Transfer vat photopolymerization for roof features .................................... 78
4.3.1 Light dose distribution ..................................................................................... 78
4.3.2 Experimental setup and process design ........................................................... 82
4.3.3 Microfluidic channels with 10 μm height ....................................................... 85
4.3.4 Multifunctional automation components for microfluidic devices ................. 87
4.3.5 3D-printed microparticle sorting device .......................................................... 90
vi

4.3.6 Summary of this work ..................................................................................... 93
4.4 Direct droplet writing of highly viscous materials ................................................. 94
4.4.1 Direct droplet writing process ......................................................................... 97
4.4.2 Direct droplet writing experimental setup design and analysis ..................... 101
4.4.3 Experimental study and analysis ................................................................... 106
4.4.4 Applications and Proof-of-concept for Multi-material Printing .................... 114
4.4.5 Summary of this work ................................................................................... 117
5.1 Answering the Research Questions/Testing Hypotheses ...................................... 119
5.2 Engineering Achievements and Scientific Contributions ..................................... 122
5.2.1 Engineering Achievements ............................................................................ 122
5.2.2 Scientific Contributions ................................................................................. 124
5.3 Future work ........................................................................................................... 124

vii

List of Tables
Table 1. Statistics of software running time. ................................................................................ 49
Table 2. Statistics of printing examples ........................................................................................ 52
Table 3. Parameter setting for test case 1. .................................................................................... 61
Table 4. Parameter setting for test case 2. .................................................................................... 61
Table 5. Parameter setting for test case 3. .................................................................................... 62
Table 6. Parameter setting for test case 4. .................................................................................... 62
Table 7. Parameter setting for test case 5. .................................................................................... 63
Table 8. Test results of the number of loading cycles to separation under different loading
conditions. ..................................................................................................................................... 64
Table 9. Test results of the number of loading cycles to separation for different area sizes. ....... 67
Table 10. Test results of the number of loading cycles to separation for different geometric
topologies. ..................................................................................................................................... 68
Table 11. Test results of the number of loading cycles to separation for different shapes. .......... 70
Table 12. Comparison of separation force between the direct pulling-up method and the VA-
separation-based method. .............................................................................................................. 73
Table 13. Mechanical properties of 304 stainless steel dispensing needle. ................................ 104
Table 14. Profiles of 304 stainless steel dispensing needle. ....................................................... 104
Table 15. Parameter settings for test cases 1-3. .......................................................................... 107
Table 16. Parameter settings for test cases 4-5. .......................................................................... 107
Table 17. Test results of droplet size under different punching speeds. ..................................... 108
Table 18. Test results of droplet size at various dispensing gaps. .............................................. 109
Table 19. Parameter settings for Trojan logo and Tai Chi pattern printing. ............................... 116
viii

List of Figures
Figure 1. Current popular additive manufacturing technologies and their corresponding
representatives................................................................................................................................. 1
Figure 2. Three main categories of the stereolithography process. ................................................ 2
Figure 3. An illustration of the cross-linking mechanism of photopolymers when exposed to UV
light. ................................................................................................................................................ 3
Figure 4. A DMD chipset with a micro-mirror array on it and the detailed structure of the
micromirrors. .................................................................................................................................. 4
Figure 5. An illustration of the classical MIP-VPP process. .......................................................... 5
Figure 6. A Gymnast test case ...................................................................................................... 30
Figure 7. The metal pin-like support and the 3-layer sheet structure design ................................ 31
Figure 8. The schematic diagram of an FFF system with the reusable metal support mechanism
....................................................................................................................................................... 32
Figure 9. The detailed structure of the first layer .......................................................................... 33
Figure 10. The detailed structure of the second and third layers. ................................................. 33
Figure 11. Sectional view of the 3D printing process with metal supports. ................................. 36
Figure 12. The height of a metal pin support ................................................................................ 36
Figure 13. Support planning generated by our slicing software. .................................................. 37
Figure 14. Comparison of 3D-printed support between the original layout and optimized layout.
....................................................................................................................................................... 39
Figure 15. Illustration of discontinuity in support volume during movement. ............................. 41
Figure 16. Discretization of metal support with the boundary region highlighted in blue. .......... 42
Figure 17. Creating Fermat spirals within a pin. .......................................................................... 44
ix

Figure 18. Partition of a level set. ................................................................................................. 45
Figure 19. Connecting the segmented curves. .............................................................................. 46
Figure 20. FFF 3D printer prototype with reusable metal supports. ............................................. 47
Figure 21. Printing platform. ........................................................................................................ 48
Figure 22. The second layer and the third layer. ........................................................................... 48
Figure 23. Wooden dummy. ......................................................................................................... 50
Figure 24. Roof. ............................................................................................................................ 50
Figure 25. Teapot. ......................................................................................................................... 50
Figure 26. Helmet. ........................................................................................................................ 50
Figure 27. An illustration of cyclic loading. ................................................................................. 56
Figure 28. The prototype system using the VA separation mechanism. ...................................... 57
Figure 29. 3D printing process with the VA separation-based method. ....................................... 59
Figure 30. Relationships between force amplitude and separation time related to frequency in the
current setup. ................................................................................................................................. 63
Figure 31. S-N curves and the corresponding separation time under different pre-stresses ......... 64
Figure 32. Equivalent S-N curve and corresponding separation time under different pre-stresses.
....................................................................................................................................................... 66
Figure 33. S-N curves and corresponding separation time for different pre-stresses and exposure
areas. ............................................................................................................................................. 67
Figure 34. Cyclic number (left) and corresponding separation time (right) for different geometric
topologies including single cylinder, two-cylinder array, and four-cylinder array, respectively. 69
Figure 35. Crack initiation and propagation of patterns with different geometric topologies ...... 70
Figure 36. Cyclic number (left) and corresponding separation time (right) for different shapes. 70
x

Figure 37. Crack initiation and propagation for fabricated different shapes. ............................... 71
Figure 38. Comparison of the separation forces of different areas between the direct pulling-up
and VA-separation-based methods. .............................................................................................. 73
Figure 39. The algorithm for the VA-VPP process. ..................................................................... 74
Figure 40. Printing result of the same hourglass using different separation methods. ................. 75
Figure 41. The measured separation force (left) and the measured separation time (right) of each
layer in the hourglass test model using the direct pulling-up and VA-separation-based methods.
....................................................................................................................................................... 77
Figure 42. Illustration of the effect of resin optical property and light dose distribution on VPP-
based microfluidic channels fabrication. ...................................................................................... 81
Figure 43. Principle of IsT-VPP process. ..................................................................................... 85
Figure 44. Fabrication of microfluidic channels through IsT-VPP. ............................................. 86
Figure 45. 3D-printed microfluidic valve and specimen platform for automation ....................... 89
Figure 46. 3D-printed microparticle sorting device ...................................................................... 92
Figure 47. Comparison of our approach (direct droplet writing) with material jetting and direct ink
writing ........................................................................................................................................... 95
Figure 48. Mechanism of direct droplet writing. .......................................................................... 99
Figure 49. DDW Printhead design. ............................................................................................. 101
Figure 50. Illustration of the self-alignment mechanism. ........................................................... 105
Figure 51. Effect of punching speed
p
v on droplet size
d
D and positional error  ..................... 107
Figure 52. Effect of dispensing gap
g
h on droplet size
d
D and positional error  ..................... 108
Figure 53. Schematic diagram on the influence of the dispensing gap and the contact line with a
substrate. ..................................................................................................................................... 110
xi

Figure 54. Relationship between droplet size
d
D and material viscosity  . .............................. 112
Figure 55. Study on 2D pattern printing ..................................................................................... 113
Figure 56. Layer thickness testing. ............................................................................................. 114
Figure 57. Demo of 3D printing structures. ................................................................................ 115

xii

Abstract
Vat photopolymerization (VPP) is an additive manufacturing (AM) process in which liquid
photopolymer in a vat is selectively cured by light-activated polymerization[1]. Through decades’
efforts, VPP has become one of the most popular AM processes to fabricate three-dimensional
(3D) objects. Compared with other polymer-based AM processes such as material extrusion, VPP
can build 3D objects with high accuracy, shape complexity and resolution, smooth surface finish,
and fast fabrication speed. However, despite these advantages of the VPP process, several
bottlenecks need to be addressed.
Firstly, due to the material contamination issue and time-consuming post-processing
operations when switching materials such as cleaning, it is challenging to achieve multi-material
fabrication in a single part efficiently, which is the critical technology for various applications.
Secondly, the light-induced fabrication mechanism limits the selection of available materials
used in the current VPP. Besides photocurable polymer resin, many more non-photocurable
materials are available in the market but cannot be 3D printed right now, such as heat cured
polymers. These materials span over a wide range of viscosity, and many of them are highly
viscous.
Thirdly, as the process uses liquid to form solid objects, there is no structural support from
the material during the printing phase. In this case, support structures will often need to be
added. The printed supports are totally a waste of materials since they need to be removed after
fabrication and cannot be reused.
Fourthly, it is essential to control the flatness and thickness of each layer in the layer-based
VPP since they will significantly affect the attachment between two adjacent layers and the
xiii

dimensional accuracy of the fabricated part. It has been a popular way to add a constrained surface
to ensure the flatness and thickness of the newly cured layer. However, for the constrained surface-
based VPP, the curing process leads to adhesive bonding between the newly cured and constrained
surfaces. To proceed with the fabrication process, a separation force is required to detach them.
Excessive separation force will cause damage to the built layers and the constrained surface.
Material jetting is an AM process that allows for full-color parts and enables multiple
materials in a single part easily. Photocurable resin or thermally molten materials are deposited
onto a build platform using a drop on demand (DOD) approach layer by layer to make parts. The
process benefits from its high accuracy and multi-material fabrication ability. However, current
material jetting is suffering from limit material options. Viscosity is the main determinant that
hinder the development of the process. Besides, a support structure is often required. Unlike VPP,
printing support structure via material jetting will waste material and slow the printing speed.
To solve the above issues, our methodology is to treat materials differently. Therefore, we
propose a novel hybrid VPP to address the limitations mentioned above. Four components work
coordinately and constitute the AM process.
Firstly, we adopt a top-down configuration for our hybrid VPP to print common photocurable
polymer resin. To achieve multi-material 3D printing, a new drop-on-demand method is developed
to deposit viscous material droplets onto the platform or built layers. For this purpose, we
developed a novel droplet-based process called direct droplet writing. Our printer head has two
working phases. Inspired by the shearing process blanking, a sealed metal capillary tube moves up
and down periodically to mechanically split a single material droplet from a continuum of liquid
material during each stroke. The deposition of material droplets hang on the tube tip is determined
xiv

by liquid bridge splitting. The novel printer head allows for firing precisely the right amount of
viscous material at precisely the right place. If the viscous material is photocurable, an ultraviolet
(UV) light source in VPP will be utilized to solidify the deposited material.
Secondly, water-soluble polymer or wax often works as sacrificial support material. However,
printing support will increase total fabrication time consumption and lead to material waste. In
order to reduce the amount of 3D printed external support, we developed an automatic and reusable
pin-like metal support for overhang features in part. After the fabrication of each layer, the metal
pins that work as support will move up one-layer thickness and stop at the specified height
automatically. The stop positions of the metal supports are determined by a combination of metal
tubes with pre-defined lengths and magnetic rings. 3D printing material can then be deposited on
the metal supports. Consequently, the supports that need to be printed will be dramatically reduced.
We demonstrate our idea on a filament fused fabrication (FFF) 3D printer incorporating our
automatic reusable metal support mechanism. The corresponding slicing software that can work
with our printer compatibly was developed. Several test cases are presented to validate the
effectiveness and efficiency of our approach.
Thirdly, for the roof features in a part design that needs an internal support structure, we are
enlightened by transfer printing and developed a double exposure method to solve this issue. The
roof feature is first fabricated on an extra build platform by mask image projection and then
connected to the built part with the second exposure of the mask image. By this approach, the
internal support structure is not required. Besides, we noted that the enrollment of an additional
build platform could largely reduce light penetration that may cure liquid resin inadvertently. With
elegant control over light energy distribution along the z-direction, we successfully applied this
method to microfluidic flow channels fabrication and printed 10 um heigh flow channels using
xv

transparent resin. The novel process is also verified by other test cases, including 3D microfluidic
channels, a particle sorting device, and a microfluid control valve. This work will significantly
advance the application of VPP in microfluidic device fabrication.
Lastly, a constrained surface was utilized to ensure z accuracy in the presented hybrid VPP.
As for the separation force problem, a new vibration-assisted (VA) separation design is presented.
To find the best way to use the VA separation method, experiments on the separation performance
under different parameters, including vibration frequency, pre-stress level, and exposure area, were
conducted. Based on the collected separation force data, an analytical model based on fatigue
fracture mechanics was built. The separation behaviors related to different shape sizes and
topologies were also studied and compared. The results showed that the separation force in VPP
was significantly reduced using the VA-separation method. Furthermore, the relationship between
the separation force and the separation time conforms to the stress-based fatigue model. This study
also provides insights into choosing process parameters by considering the trade-offs between
separation force and building efficiency.
The developed hardware prototypes and software algorithms provide a new systematic
strategy to address the issues we met and demonstrated our insights in the process study. This work
will significantly advance AM technologies in the future.

1

Introduction
Additive manufacturing (AM) [1]-[5], commonly known as three-dimensional (3D) printing
[6], involves the whole process of making 3D solid objects from Computer-aided Design (CAD)
files (e.g. a .STL file) [7]. According to ASTM F2792 standards, 3D printing can be classified into
seven main categories: vat photopolymerization [8]-[11], material extrusion [12]-[14], material
jetting [15]-[17], binder jetting [18][19], powder bed fusion [20]-[23], direct energy deposition
[24]-[26], and sheet lamination [27]. Figure 1 shows the current several popular 3D printing
technologies and their corresponding representatives.
Among all the 3D printing processes, the Vat Photopolymerization process [28]-[32] has been
recognized as one of the most accurate 3D printing technologies in the world so far. In the very
early 1980s, people first used this process as an affordable way to create prototypes for product

Figure 1. Current popular additive manufacturing technologies and their corresponding representatives.
development within industries. People usually call it Rapid Prototyping (RP) technologies [33].
The first patent of VPP -- Stereolithography Apparatus (SLA) can be traced back to 1986 when an
2

American inventor, Charles (Chuck) Hull, invented the first SLA 3D printer [6]. As the first
commercialized 3D printing technology, the stereolithography (SL) process has attracted more and
more attention as the current industry widely adopts it due to its high geometry accuracy and fast
speed. Still, it continues to grow in potential markets such as jewelry, dentistry, automobile, and
biomedical devices. SLA can be classified into three sub-categories according to the light source
used in the process: laser beam scanning based SLA [34][35] and digital light processing (DLP)
based SLA [36]-[45], and liquid crystal display (LCD) based SLA [46]-[48] as shown in Figure
2. All of them use light to induce polymerization of photocurable resins. But the way they go about
applying this principle is what sets them apart.

Figure 2. Three main categories of the stereolithography process.
Over the past four decades, VPP evolves a lot along with the development of technologies in
other fields. As a result, many research efforts have been made to improve VPP performance
regarding fabrication speed, building resolution, dimensional accuracy, post-processing, material
properties, etc.
The typical materials used in VPP are photopolymers. The corresponding chemical reaction
is called photopolymerization or photocuring. Photopolymers [49] are light-sensitive polymer
materials that change their properties when exposed to UV light. They can change their state from
water-like liquid to solid plastic-like substance. Only the area exposed to UV light hardens,
3

whereas unexposed parts will remain in the liquid state. The schematic representation of the
photopolymerization process is shown in Figure 3. Photopolymerization [50] is defined as a
synthesis of polymers by chain reactions initiated upon light absorption. Light serves only as an
initiating tool. It does not interfere with the propagation and termination stages of the chain process.

Figure 3. An illustration of the cross-linking mechanism of photopolymers when exposed to UV light.
The fundamental photo polymerization follows the Beer-Lambert law [51]. Given the light
intensity I , and light exposure time t , the curing depth
d
C is determined as [52]:
ln
dp
c
It
CD
E
 
=


(1)
p
D is penetration depth which is related to material property, and is derived as:

1
p
D
h
= (2)
Hence, the polymerization law can be rewritten as:

1
ln
d
c
It
C
hE
 
=


(3)
The recent development of Micro-Electro-Mechanical Systems (MEMS), such as Digital
Micromirror Device (DMD) [53], provides the capability of accurately and selectively controlling
4

the energy input of the target area. The ability makes it possible to develop a new Mask-Image-
Projection-based Stereolithography process (MIP-SL). As shown in Figure 4, a DMD chipset is a
reflective spatial light modulator that can dynamically control the direction of the reflected light
on each micromirror, thus turning on/off the light in a single pixel in a macro view.

Figure 4. A DMD chipset with a micro-mirror array on it and the detailed structure of the micromirrors.
A typical DMD chipset consists of a mirror array of up to 2 million individually controlled,
highly reflective aluminum micromirrors, the status of which can be switched at a very high
frequency (can be up to 16,000 Hz) [53]. Thus, the device enables users to program high-speed
light patterns. The design of a single unit of the micromirror is shown in Figure 4. The micromirror
can be tilted ± 10°, reflecting the exposed light either onto or away from the screen. The DMD’s
micro-mirror array is optically efficient from 350 nm – 2500 nm. The light intensity can be
controlled by Pulse Width Modulation (PWM), which is the principle to realize the grayscale for
a projection image.
Powered by the advanced DMD technology, the basic idea of the MIP-VPP process is to use
mask images sliced from a 3D CAD model to solidify the liquid photopolymer directly. An
illustration of the classical MIP-VPP process is shown in Figure 5. The classical MIP-VPP process
can be further classified into two main sub-categories: the top-down and bottom-up MIP-VPP
processes depending on where the newly cured layer is or where the projection direction is.
5

Figure 5. An illustration of the classical MIP-VPP process. a) a top-down MIP-VPP process; b) a bottom-up MIP-VPP
process.
When the building process begins, the DMD chipset driven by the DMD controller will project
the mask image of a sliced layer onto the liquid resin interface to selectively solidify it at the
desired regions. After one layer is finished, the building platform will move to the next position to
form a thin gap between the newly solidified surface and the liquid resin surface, where the next
layer will begin. The process repeats until the desired 3D object has been completed. Comparing
to the laser-based VPP process, the whole layer of the selected liquid resin can be solidified by
just one single exposure in the MIP-VPP process instead of scanning the tool path of the image
pattern by a laser beam. Hence, the entire building speed is relatively faster for a MIP-VPP than
the laser-based VPP.
Despite these advantages, there exist several bottlenecks that need to be addressed in VPP.
Firstly, it is relatively difficult to achieve multi-material fabrication due to material contamination
6

issues in a single part, which is the critical technology for various applications. Secondly, the light-
induced fabrication mechanism limits the selection of available materials used in the current VPP.
Besides photocurable polymer resin, many more non-photocurable materials are available in the
market but cannot be 3D printed right now, such as heat cured silicones. These materials span over
a wide range of viscosity, and many of them are highly viscous. Thirdly, due to the layer-based
fabrication process, it is essential to control the constant flatness of each layer since the layer
flatness will significantly affect the attachment between two adjacent layers and the dimensional
accuracy of the fabricated part. It has been a popular way to add a constrained surface (transparent
glass) to ensure the surface flatness in the newly cured layer, as shown in Figure 5. However, it
will be challenging to separate the newly cured layer from the constrained surface due to the near-
vacuum environment and chemical bonding of the two phases. In addition, the separation force
will increase with the increase of the contact area. Fourthly, the layer-based fabrication manner
means the building part can only grow in a single direction. Thus, some overhang structures will
occur in some layers due to the geometry complexity. Therefore, a significant amount of additional
support structures are required to be added for a CAD model. However, these printed supports are
a waste of materials since they need to be removed after fabrication and cannot be reused.
It is urgent and desirable to address these problems and explore the possibility of VPP in
future applications. This proposal aims to address the above issues by proposing a new hybrid VPP
process. The original method can work with photocurable and non-photocurable viscous materials.
Our methodology is to treat materials differently. Four components will work coordinately and
constitute the AM process.
Firstly, we adopt a top-down configuration and DLP as the light source for our hybrid VPP
to print general photocurable polymer resin. In order to achieve multi-material 3D printing, a new
7

drop-on-demand method is harnessed to deposit viscous material droplets onto the platform or
built layers. For this purpose, we developed a novel droplet-based process called material blanking.
Our material ejector has two working phases. Inspired by the shearing process blanking, a sealed
metal capillary tube moves up and down periodically to mechanically split a single material droplet
from a continuum of liquid material during each stroke. The detachment of material droplets hang
on the tube tip is determined by external pneumatic power. The novel ejector allows for firing
precisely the right amount of viscous material at precisely the right place. If the viscous material
is photocurable, a UV light source in VPP will be utilized to solidify the deposited material. Several
test cases of viscous self-solidifiable material and photocurable composite are presented to show
the effectiveness of the proposed method.
Secondly, water-soluble polymer or wax as sacrificial support material is deposited by the
material ejector mentioned above. However, printing those support structures will increase total
fabrication time consumption and lead to material waste. To reduce the amount of 3D printed
external support, we developed an automatic and reusable pin-like metal support for overhang
features in part. After the fabrication of each layer, the metal pins that work as support will move
up one-layer thickness and stop at the specified height automatically. The stop positions of the
metal supports are determined by a combination of metal tubes with pre-defined lengths and
magnetic rings. 3D printing material can then be deposited on the metal supports. Consequently,
the supports that need to be printed will be dramatically reduced. We demonstrate our method on
an FFF 3D printer incorporating our automatic reusable metal support mechanism. The
corresponding slicing software that can work with our printer compatibly was developed. Several
test cases are presented to validate the effectiveness and efficiency of our approach.
8

Thirdly, for the roof features in a part design that needs an internal support structure, we are
enlightened by transfer printing and developed a double exposure method to solve this issue. The
roof feature is first fabricated on an extra build platform by mask image projection and then
connected to the built part with the second exposure of the mask image. By this approach, the
internal support structure is not required. Besides, we noted that the enrollment of an additional
build platform could primarily reduce light penetration that may cure liquid resin inadvertently.
With elegant control over light energy distribution along the z-direction, we successfully applied
this method to microfluidic flow channels fabrication and printed 10 um- height channels using
transparent resin. The novel process is also verified by other test cases, including 3D microfluidic
channels, a particle sorting device, and a microfluid control valve. This work will significantly
advance the application of VPP in microfluidic device fabrication.
Lastly, a constrained surface was utilized to ensure z accuracy in the hybrid VPP. As for the
separation force problem, a new VA separation design is presented. To find the best way to use
the VA separation method, experiments on the separation performance under different parameters,
including vibration frequency, pre-stress level, and exposure area, were conducted. Based on the
collected separation force data, an analytical model based on fatigue fracture mechanics was built.
The separation behaviors related to different shape sizes and topology were also studied and
compared. The results showed that the separation force in VPP was significantly reduced using
the VA-separation method. Furthermore, the relationship between the separation force and the
separation time conforms to the stress-based fatigue model. This study also provides insights into
choosing process parameters by considering the trade-offs between separation force and building
efficiency.
9

The following sections are organized as follows. A literature review on the related work is
given in Section 2. Several current research problems will be stated in Section 3, while the
corresponding hypotheses to the questions are also provided. Section 4 shows the contributions of
this study. At last, Section 5 concludes the dissertation with some remaining future work.
The following sections are organized as follows. First, section 2 will briefly review the related
literature for the proposed research work. Then, several current research problems will be stated
in Section 3 and the corresponding hypotheses to the research problems aforementioned. Finally,
the research results and evaluations will be presented in Section 4, and Section 5 concludes the
dissertation with some remaining future work.

10

Literature Review
2.1 Reduction of separation force in vat polymerization process
In general, there are two approaches in photocuring liquid resin, i.e., the free surface method
and the constrained surface method[58]. In the free surface method, the liquid resin is exposed to
light and solidified with no constrained surface, and then a new layer of liquid resin is refilled by
lowering the building platform down[58]. In comparison, in the constrained surface method, there
are two kinds of configurations, top-down and bottom-up [54][55]. In both configurations, liquid
resin is sandwiched between the built layers and the constrained surface. The light penetrates
through the transparent constrained surface and cures the liquid resin. The constrained surface
method has several advantages over the free surface method[54]. For example, (1) a much smaller
layer thickness can be achieved because the gap size is only determined by the Z stage resolution
regardless of the fluidic properties of the liquid resin. (2) A flat layer of the new liquid polymer
could be refilled in a short time, even if the liquid is viscous [56]. (3) The building velocity is faster
since the oxygen inhibition effect in the free surface method is eliminated [57]. (4) Further, in the
bottom-up constrained surface method, the part height is no longer limited by the resin vat depth;
thus, a shallow vat can be used to reduce the required volume of the liquid resin.
Despite the aforementioned advantages of the constrained-surface-based VPP process, there
exists a bottleneck that needs to be addressed. That is, the curing process leads to adhesive bonding
between the newly cured layer and the constrained surface. In order to proceed with the fabrication
process, a separation force is required to detach them [58]. Most industrial bottom-up SLA uses
direct pulling-up as the separation method. However, during the pulling-up stage, the newly cured
layer of a large area may fail to be raised; instead, it may attach to the bottom surface, and the
11

building process will fail [59]. Moreover, the excessive separation force may damage the
constrained surface during the recursive separation process. Such a large separation force limits
the printable layer area and the reliability of the 3D printing process. So far, all the commercial
VPP machines with large building areas (e.g., 600 mm × 600 mm) used the free surface method.
Since the required separation force in the constrained surface method will exponentially increase
for a large building area, how to deal with the separation force issue is one of the critical challenges
to be addressed in developing future constrained-surface-based VPP machines with a large
building area.
Many attempts have been made to address this critical challenge before. Applying a suitable
coating medium (e.g., a Teflon, silicon, or PDMS film) to the bottom surface of the resin vat is a
widely used method to alleviate the pulling-up force [59][60]. Even with these coatings, the
separation force is still considerably large for the direct pulling-up method. Therefore, some
commercial systems (e.g., EnvisionTEC Ultra2 Printer) have adopted a slow pulling-up speed
since a large pulling-up velocity will significantly increase the separation force [61]. The reduced
moving-up speed will dramatically reduce the fabrication efficiency. Besides, the coated film has
to be constantly stretched, and the cyclic loading to the film in the Z direction will make the film
wrinkle after a certain time. Hence the life of the separation film is usually limited.
Different motions of the resin vat or the building platform have also been proposed to reduce
the separation force, such as tilting [61]. However, the tilting method performs poorly when the
part cross-section area is large. Furthermore, the additional tilting motion increases the fabrication
time, resulting in a reduction of productivity[61].
12

Applying the sliding motion based on a two-channel system has been presented by Zhou et
al.[43], which showed using the shearing force instead of directly pulling up can significantly
reduce the separation force. After the fabrication of one layer, the vat was slid in the horizontal
direction, so the separation between the built layers and the vat can be achieved with shear stress.
This two-channel system also incorporated an oxygen-aided inhibition mechanism discovered by
Dendukuri et al. [62]. However, the PDMS is sandwiched between the liquid resin and vat. The
isolated environment prevents the oxygen supply and the oxygen near the PDMS surface will be
consumed. Pan et al. [54] presented a micro-hole-based approach to provide a sufficient and
constant supply of oxygen to the PDMS medium. However, the required sliding motion should be
at least the size of the building platform. Hence a longer fabrication time due to the sliding motion
is required when the part cross-section area becomes larger.
Recently, Tumbleston et al. [63] proposed a continuous liquid interface production (CLIP)
approach by using an oxygen-permeable Teflon AF 2400 film that creates a “dead zone” to inhibit
photopolymerization. The thin inhibited layer allows the platform to move up continuously, which
results in 25-100 times faster printing speed than other layer-based SLA systems. Despite the
improvement, the oxygen permeable film will form a dome during the printing process because
there is no constrained plane to define the formed layer. The gap between the film and the
previously cured layers is not flat, which reduces the accuracy in the Z direction, especially for a
part with a larger cross-sectional area.
To solve the dome effect of the oxygen permeable film, Wang et al. developed an Active
Separation Bottom-Up Stereolithography (ASBUS) process [64]. An empty area called the active
zone is reserved between the Teflon film and the constrained plane. During the separation process,
water is pumped into and out of the active zone to raise and then peel the film off the newly cured
13

layer. The separation force is significantly reduced, and the dome effect is eliminated through this
method. However, the additional pumping action complicates the structure and reduces
manufacturing efficiency. Besides, this approach also faces the same disadvantage of the short-life
film mentioned earlier for the film-based separation method.
Instead of using films or motions, another mechanism of using vibration to reduce the
separation force was first proposed by Jin et al [58]. The cyclic loading applied on the constrained
surface will cause fatigue of the adhesive interface between the built part and the constrained
surface. After enough cycles, the built layers can be separated entirely from the constrained surface.
However, their vibration-assisted (VA) hardware design with the unrestricted reversed stress will
cause unexpected damage to small features of the part. Also, the vibration effect on the large
contact area and how the vibration parameters affect the separation performance were not studied.
An improved design and a theoretical model to address the vibration effect are further developed
in this paper.
Besides the various methods developed to solve the separation issue in constrained-surface-
based SL, researchers also examined the relationship between the separation force and the process
parameters. They developed different models to characterize the separation process. A better
understanding of the separation processes will significantly benefit the development of an
intelligent control method that can be used in SL.
Huang and Jiang [60] developed a separation force monitoring system for the bottom-up
constrained-surface SLA with a silicone film as the coating material. Kovalenko et al. [65]
experimentally verified the linear dependence between the pulling-up force and the built area on a
bottom-up constrained-surface VPP with a coated PDMS film. From the viewpoint of fracture
14

mechanics, they used the crack initiation toughness to estimate the separation process. Based on
the experimental results, they also found that the pulling-up force increased linearly with the
contact area. Besides, their study indicated that the projected images with different shapes of the
same area resulted in the same separation force[61]. The geometry only affected the shape of the
force-platform displacement profile.
Based on the measured force data, Zhou et al. [45] reported that the separation force was
affected by several factors, including exposure time, image area, and image geometry. The
exposure time effect was explained from the perspective of energy. The image geometry, unlike
the conclusion drawn by Huang and Jiang, was found to have impact on the peak force but not as
clear as the factors of the exposure time and image area.
Liravi et al. [66] characterized the separation of the cured layer from the PDMS coated resin
vat as the delamination of a laminated composite structure based on fracture mechanics. The
method for failure analysis of the laminated composite structure was developed to analyze the
pulling-up separation in the bottom-up based SL. The cohesive zone model (CZM) was adopted
to predict the separation behavior of different pulling-up velocities. The constitutive parameters
for the CZM were estimated from the experimentally measured pulling-up force. The finite
element (FE) simulation showed that the CZM was eligible for simulating the separation force
under different pulling-up velocities. It should be noted that the cohesive zone does not represent
any physical material. This zone can be perceived as a zone of vanishing thickness at the interface
between the two separate parts[66]. Hence CZM can be applied to different materials (e.g., the
photocurable polymer and the coating media) by changing the material parameters or the cohesive
law that defines the behavior of the cohesive zone.
15

From the fluid mechanics perspective, Pan et al. [54] used Navier–Stokes equations to study
how pulling-up velocity affects the pulling-up force under the circumstance that the oxygen-aided
inhibition effect is present. The result indicated the same conclusion that an increasing pulling-up
speed will lead to a larger separation force. Wu et al. [67] built a constrained-surface VPP based
on the tilting separation mechanism and applied the CZM in the analysis of the separation force.
Experimental studies on the effects of different building areas, part orientations, coating films, and
tilting velocity on the separation force were conducted by experimentally collecting the force data.
The study results can be used in choosing reasonable process parameters for the constrained-
surface based VPP process.
2.2 Designs and optimization techniques of support structure
The 3D printing process is a layer-based fabrication process. The printing material can only
be deposited on the top of an existing surface. This will be problematic for parts with overhang
structures. Such overhang features cannot be directly printed if there is no support surface
immediately beneath them. The 3D printing technologies such as FFF and VPP solve this problem
by creating additional supporting structures for the overhang features. The printed supports using
the same or different materials such as water-soluble material [68] will be discarded after
fabrication. This leads to longer fabrication time, more material waste, and extra post-processing
time. Therefore, support generation for 3D printing is a critical issue.
Most existing solutions to reduce support use for FFF are geometry-based approaches. These
methods can be divided into three categories:
For a given model, one type of approach is to select a suitable orientation of the CAD model
to cut down the support volume as presented in [69]. Besides support volume, several other aspects
16

like surface quality, building time, part accuracy, or contact area are taken into consideration in
these studies [70]-[76].
The second type of approach reduces the use of printed supports by modifying the model itself.
Mirzendehdel and Suresh [77] proposed a topology optimization method based on the constraint
of support volume. Another straightforward method is to divide 3D models into multiple small
pieces to reduce support use and printing time for large models which may exceed the size of the
printing tray [78]. Vanek et al. [79] converted the input 3D mesh into a shell before cutting it into
multiple segments. Then the small segments are tightly packed to minimize the support material
consumption and the whole bounding box volume. Since pyramidal shapes have no overhang
features, R. Hu et al. [80] decomposed the 3D models into a set of pyramidal segments, and no
supports are needed.
These two types of approaches require either changing the building orientation or the manual
assembly of multiple small pieces, which could be problematic for many applications. In our work,
we assume that the input CAD model will not be split, and a building orientation has been selected
by the user and given as the input.
The last category of methods tries to reduce printed support by presenting new geometrical
shapes for support structure and optimizing the layout of supports. The most common support
shape is the vertical solid wall-like structure that connects the surface facets with the tilting angle
large than a threshold value (e.g. commercial software such as MakerBot and Simplify3D). This
type of support ensures reliability at the expense of increased printing material and time. Instead
of vertical wall-like support, X. Huang et al. [81] presented a sloping wall-like support structure.
Basically, this method shrinks the size of the middle portion of vertical wall-like supports. To
17

reduce the support volume, Strano et al. adopted density-changeable cellular structures [82].
Unlike classical external supports that touch the printing platform, S. Cacace et al. [83] proposed
an algorithm that converts all the external supports to internal ones with both ends of the supports
connecting to the part itself. In this way, the support volume and printing time are reduced.
However, this method is only workable for chamfer features. For the long cantilever feature, the
material consumption of this method is larger than classical external support. Instead of generating
support for the whole overhang area, Autodesk company (Meshmixer [84]) and J. Vanke et al. [85]
proposed a kind of clever support that only touch the part at sparse points within the overhang area.
Similar to the natural tree, these supports will converge from the supporting points progressively
and form a set of brunches. Finally, vertical or non-vertical trunk-like pillars are generated to
support the branch-like supports. In this way, much less material will be used to print supports but
at the expense of reliability. The printing process may be unstable or even fail in some cases. This
is because printing slanted pillars is less reliable than printing vertical pillars due to smaller
bonding areas between layers and uneven warping during cool down as mentioned in [86].
Compared with vertical pillars, slanted pillars are more sensitive to printing parameters, filament
quality, and temperature. Besides, the weight of the branch-like pillars above and the torque
generated by the print head tend to bend the structure and cause deformation and lead to failure
eventually. J. Dumas et al. [86] presented a scaffolding-like support structure based on the property
of FFF that horizontal roof could be printed across small gaps. This method also inherits the high-
efficiency property that only supports certain sparse points instead of the whole overhang area.
Different from the long and slanted tree-like supports, J. Dumas uses wide horizontal roofs with
short slanted pillars to connect the points to be supported. Then sturdy vertical pillars are used to
support the roofs. The reliability is improved but the material consumption and printing time
18

accordingly increase too. E. Barnett and C. Gosselin [87] proposed two support strategies for FFF
using weak material such as foam and gel to lower the material cost and facilitate support removal.
One strategy is called shell technique. First, strong material is extruded to form an enclosed shell
and weak material (e.g. shaving foam) is deposited to fill the space between the shell and the part.
The shell technique is robust and compatible with any part geometry, but time-consuming due to
the large support volume. The other strategy is called film technique. Weak material works as a
thin film between the part and the rigid support to facilitate support removal. However, whether
this technique could work depends on part geometry. The surface quality of the part is not good
for both methods.
2.3 Droplet dispensing techniques for viscous material
Droplet-based 3D printing methods is capable of using multiple types of materials to fabricate
heterogeneous objects with the potential of defining their properties voxel by voxel. Objects
therefore can possess a range of material properties such as stiffness, strength, density,
transparency and color. A variety of applications [88]-[92] including functionally graded material
(FGM), full color parts and product performance optimization benefit from this voxel-based
fabrication approach. Despite these advantages of multi-material printing ability, high resolution,
contactless deposition, and high droplet generation rates, material jetting still suffers from limited
material options. Conventional material jetting is only suitable for low-viscosity inks [93][94].
This leads to objects produced with material jetting are usually weaker and is unsuitable for
functional applications.
To jet material of viscosity higher than 40 mP· s, researchers have to rely on different kinds
of external forces and proposed four kinds of technologies:
19

(1) Laser-induced jetting [95]-[98] uses pulsed laser beam to generate vapor bubbles. The
expansion of vapor bubbles propel ink suspending to the donor platform downward to the receiver
substrate which is close to the donor platform. The jetted ink thread contacts the receiver platform
and forms a liquid roof with the ink source. Then the liquid thread shrinks and finally breaks as
the vapor bubble fades off. Part of ink will remain on the receiver substrate. The deposited amount
of ink is strongly coupled with the gap thickness between the donor and receiver substrates and
the property of the substrate interface, which makes it challenging to control the droplet shape.
(2) Valve-based approach [99]-[101] extrudes ink out continuously which is similar to direct
ink write (DIW), meanwhile utilizes piezo-driven valve to cut off ink flow to realize DOD printing.
A liquid roof will also form between the nozzle and the build platform and lower the controllability
of jetting process. Extruded ink cannot form individual droplets and separate from nozzle because
only gravitational force G Vg  = (where  is ink density, V denotes droplet volume, g
represents gravitational acceleration) takes effect during this dripping process and it is smaller than
the maximal capillary force
m
FD  = , (where  is ink surface tension, D is nozzle diameter).
(3) Electrohydrodynamic jet printing [102] applied electric force to the extruded material to
achieve contactless droplet deposition with high resolution. However, this method is only
applicable to materials with electromagnetic properties.
(4) Acoustophoresis-based printing [103] is another non-contact droplet dispensing method.
In addition to gravity, external acoustic radiation force generated by a transducer also push down
pendant ink drops dangling at the end of a capillary tube against the capillary force. This method
can even work with highly viscous material like honey. However, the dispensing rate of high
viscous material is relatively slow. Large acoustic force required by viscous material will decrease
20

dispensing accuracy. Besides, this non-pulsed method may be difficult to print discrete pattern like
lattice structure efficiently due to a constant flow rate is required.
2.4 3D printing processes to achieve microfluidic channels
A microfluidic chip is intrinsically a set of connected microchannels or chambers within a
bulk material with inlets and outlets. The liquid fluid is directed, mixed, or split by microchannels’
network to achieve the desired applications in chemistry and biomedical fields such as
microreactors [114], fluid mixer [115]-[117], cell analysis/culture [118]-[121], drug assays [122]-
[125], and cell/particle sorting [126]-[128]. The miniaturized microfluidic devices allow for high-
throughput and fully automatic processing with reduced sample and reagent consumption.
Typically, for truly microfluidic devices, at least one channel dimension should be in the range of
10-100 μm (which is also the size range of cells) [129]. And transparency is always desired for
easy visualization in this application domain (i.e., fluorescently labeled samples).
In the past two decades, the majority of microfluidic devices were built with
Polydimethylsiloxane (PDMS) by soft lithography, a technique based on PDMS micro-molding
[130]. However, soft lithography is time-consuming, and the whole process involves substantial
human operations. The layered molding method limits the structural complexity of the
microchannels that can be produced. Aligning and bonding multiple PDMS layers to achieve
advanced functions is hugely challenging [130][131]. The high cost of instrumentation and
cleanroom required for soft lithography also hampered microfluidic devices to a broader audience
[132].
3D printing as a promising alternative method to fabricate microfluidic chips has attracted
considerable attention. Many researchers reported that with 3D printing, the fabrication time and
21

cost are dramatically reduced by circumventing mold-making procedures. It efficiently produces
quasi-arbitrary 3D microfluidic chips in one step without the need for the assembly process. Unlike
the physical format of soft lithographic masters or molds, the digital part design files for 3D
printing (i.e., STL files) can be shared effortlessly, facilitating microfluidic chips’ development.
Among different 3D printing technologies, VPP is a promising one because of its high resolution,
smooth surface quality, and affordability [133]. In the VPP process, an irradiation light source
such as a digital light processing (DLP) projector or a scanning laser beam is used to cure the vat’s
liquid resin [134]. By VPP, standard connectors such as barbed connectors for fluid and pressure
sources can be integrated into microdevice to achieve “plug-and-play”. Besides, flushing
unpolymerized resin in microvoids after VPP printing is much easier than removing solid
sacrificial support as required by other 3D printing processes such as material jetting and material
extrusion, which lead to extra post-processing time [135]-[137]. Therefore, self-supporting
structures for channel cross-sections such as triangles and circles are proposed based on max
printable threshold overhang angle to address the support issue in material extrusion
[114][138][139]. Multifunctional microfluidic channels can be fabricated by the material extrusion
method without requiring sacrificial materials. However, the channel cross-sectional profile
achieved is still in the large microfluidic regime (100-500 μm ) or at the sub-100 μm level at best
(Figure 42Aa). More importantly, the imposed cross-sectional profile constraint not only
complicate the design phase but may also limit its applications requiring precise laminar flow
control or structures of low aspect ratio such as perfusion system with concentration gradient
control [140][141], single-cell trapping devices [142]-[144], and particle sorting devices [126].
For material jetting, thin polycarbonate membranes are carefully laid onto open channels (located
on the surface of a printed part) to serve as channel roofs during fabrication [145]. The embedded
22

closed channels (reside inside a printed object) are converted into open channels with no need for
support material through this approach. Nevertheless, similar to soft lithography, this membrane-
sealing approach may also face alignment and bonding issues and impose limitations in 3D fluidic
networks.
Despite these advantages, VPP technologies for the fabrication of microfluidic chips also face
a severe limitation in resolution, especially in the z-direction. For the channel width dimension, 20
μm can be achieved with high-resolution DLP projectors (pixel resolution is 7.6 μm ) [146].
However, as for the channel height, most VPP 3D printers can only fabricate channels with more
than 200 μm height at best when printing transparent resin [147]-[150]( Figure 42Aa). The
practical limit of minimum channel height results from the over-curing issue. Irradiation of the
resin for building the channel roof can potentially photopolymerize the underlying channels that
are still filled with resin. The light will also penetrate the previously built layers to cure the residual
resin left inside the channel due to transparency and cause clogs. Researchers have investigated
different strategies to increase z-resolution. They can be divided into two categories. One type of
strategy is to fine-tune the printing parameters [151][152]. The second is to decrease light
penetration depth, a material optical property, by adding photosensitizing additives for the visible
blue light source (405 nm )[152][153] or shift the light source from visible blue light to ultraviolet
(UV) light (  385 nm ) [147] [148] or further adding UV absorbing dyes [121][146] [149][150].
However, all these methods will either result in channels larger than 100 μm or render the prints
colored. Specifically for VPP, the minimum channel heights
min
h and corresponding light
penetration depths
p
 of the customized resins in the literature are plotted in Figure 42Ab (if
provided in the literature). From Figure 42Ab, light penetration depth governs the z resolution of
23

microfluidic chips for previous approaches. That is, the minimum feasible channel height for a
given resin must be larger than 2.3 times its light penetration depth [146]. This rule dramatically
decreases the material options available for microfluidics.

24

Research Questions and Hypotheses
3.1 Statement of Problems
In this proposal, we try to develop a hybrid vat polymerization process for viscous
photocurable and non-photocurable materials to address the current limitations. They are support
structures for the overhang features and roof features, big separation force during breaking the
adhesive bonding in the constrained-surface-based VPP process, and limited materials available
for droplet-based 3D printing.
For the support structures in the VPP process, the support structures not only largely increase
the printing time (if using laser scanning method) and the material waste but also need substantial
post-processing time. Moreover, in some special cases, it is even impossible to reach and remove
the built support structures when the support structures are almost in a sealed empty chamber.
On the other hand, as the building area increases, the separation force will significantly
increase as well in a constrained-surface-based VPP process. It is very difficult to separate the
newly cured layer from the constrained surface. This difficulty results from the attaching force
developed between the newly cured layer and the constrained surface and the nearly vacuum
condition. Thus, a separation force is required to break the adhesive bonding and the vacuum state.
Such a separation force will significantly affect the printing speed, reliability of the printing
process, printable size, and life cycle of the constrained surface[58].
Furthermore, the light-induced fabrication mechanism in the VPP process limits the selection
of available materials that can be applied in the current VPP process. The current photopolymer
resin can produce parts with quite smooth surface finish and preserve the fine features. Despite the
impressive appearance and surface quality of the parts that can be fabricated by the traditional VPP
25

process, the printed parts may not be suitable for functional use due to its low strength and
relatively brittle mechanical property.
Last, the multi-material VPP process has been reported, but multiple resin vats are required
as well as cleaning and other post-processing operations when switching between materials[104].
The printing process has to be stopped for a certain period of time in the middle of the process for
material cleaning and change, making the entire production time extremely long. This problem
becomes even more apparent when a frequent material change is needed for hundreds of layers as
the material complexity of the part increases[105].
3.2 Hypotheses
The primary goal of this research is to address current limitations for the VPP process. It
will mainly focus on the following four aspects: 1) how can we easily reduce the 3D printed
support to reduce material waste and printing time? 2) how can we reduce the separation force in
the constrained-surface-based VPP process for the large-scale 3D printing? 3) how can we print
parts with roof feature via VPP without internal support? 4) how can we broaden material
selections available for VPP process?
In order to answer the above proposed questions, we have the following hypotheses:

To answer this question, the following hypotheses are investigated:
Q1: How can we easily reduce the 3D printed support to reduce material waste and printing time?
26

To answer this question, the following hypotheses are investigated:

To answer this question, the following hypotheses are investigated:

To answer this question, the following hypotheses are investigated:
Hypothesis 1.1: Low-cost and reusable metal pins can be used as external support for overhang
features of 3D objects.
Hypothesis 1.2: The reusable pin-like supports can move up automatically.
Q2: How can we reduce the separation force in the constrained-surface-based VPP process for
large-scale 3D printing?
Hypothesis 2.1: Less force but multiple attempts may lead to a more efficient result.
Hypothesis 2.2: The reduction of the separation force for a constrained-surface-based VPP
process can be achieved by introducing a small force to induce an initial crack in the interface
between the newly cured layer and the constrained surface.
Hypothesis 2.3: The initial crack in the interface between the newly cured layer and the
constrained surface can be quickly propagated to the entire interface.

Q3: How to print parts with roof feature via VPP without internal support?
Hypothesis 3.1: An in-situ-transfer method can be utilized in VPP to print roof feature so that
internal support is no longer required.
Q4: How can we achieve multi-material VPP efficiently without material contamination?
27

Hypothesis 4.1: Multi-material VPP can be achieved via incorporating a drop-on-demand method
so that there is no need to switch between resin vats as well as cleaning and other post-processing
operations.
Hypothesis 4.2: A drop-on-demand method can be realized by relying on mechanical force so
that highly viscous materials can be deposited.

28

Results and Evaluations
4.1 Reusable support for additive manufacturing
For 3D printing processes such as fused filament fabrication (FFF) and stereolithography,
supports are required in order to print parts with overhang structures successfully. These printed
supports using the same or different materials are a waste of materials since they need to be
removed after the printing process and cannot be reused. Support printing is also time-consuming
for FFF. To address the support-related challenges, we developed a new type of reusable metal
support to reduce the amount of 3D-printed supports. Instead of 3D printing material, we use metal
pins to support the overhang structure. After the fabrication of each layer, the metal pins that work
as support will move up one-layer thickness and stop at the specified height automatically. The
stop positions of the metal supports are determined by a combination of metal tubes with pre-
defined lengths and magnetic rings. 3D printing material can then be deposited on the metal
supports. Consequently, the supports need to be printed will be dramatically reduced. After the
printing process, the metal rods can be easily separated from the part and reused. A prototype of
the FFF 3D printer incorporating our automatic reusable metal support mechanism was built. The
corresponding slicing software that can work with our printer compatibly was developed. Several
test cases are presented to demonstrate the effectiveness and efficiency of our method.
In this work, we present a new type of automatic reusable metal support suitable for different
3D printing processes such as FFF and SL. We integrate our metal support mechanism with a
commercial FFF 3D printer since the FFF 3D printer is very common, inexpensive, and widely
used. One test case using our approach and the comparison with the traditional method is shown
in Figure 6. By comparing the tool paths of the same layer (178th layer indicated by the white
29

dash line in Figure 6(a)(b)(c)), we can see the tool path using our automatic support (Figure 6(d))
is much shorter than that of the traditional method (Figure 6(e)). Consequently, material
consumption (11mm VS 5mm) and printing time (37s VS 19s) are dramatically reduced. Another
advantage of our method is the enhancement of reliability. Long 3D-printed support (See the 3D-
printed pillar marked by the red rectangle in Figure 6(b)) may collapse during printing. In our
approach (Figure 6(a)), the pillar is truncated and replaced by our metal support (Figure 6(a)).
Our metal support can work with any given objects and any 3D-printed support structure
mentioned above as long as external supports are needed. Hence, all the previous support
techniques can work with our method compatibly. The metal supports can be easily removed after
the printing process. The residual 3D-printed supports are automatically generated by self-
developed software. The software will optimize the layout of a part on the printing tray to make
the utmost of our setup. We believe more reusable support techniques inspired by our work will
be developed in the future.
30

Figure 6. A Gymnast test case. (a) Printing result of a gymnast with reusable metal support. (b) Printing result of a gymnast
without reusable metal support. (c) Gymnast after cleanup. (d) Tool path of the 178th layer in (a) marked by the white dash line.
Printing time: 19s. Material consumption: 5mm. (e) Tool path of the 178
th
layer in (b) marked by the white dash line. Printing
time: 37 s. Material consumption: 11 mm
4.1.1 Reusable metal support design
There are many possible configurations of our support mechanism, according to different 3D
printing processes. One configuration for FFF is shown in this section. Other variations of our
support mechanism can be made based on the same design logic.
4.1.1.1 Basic principle
The schematic diagram of the key apparatus is shown in Figure 7. The reusable metal support
is composed of a set of pin-like metal supports and a 3-layer sheet structure. The pin-like metal
supports constitute a movable print platform on top of the first-layer sheet (See Figure 7(b)). The
metal rod will be lifted a small distance (i.e. the layer thickness used in the 3D printing process)
31

gradually by the third-layer sheet with magnetic discs on the bottom and, finally, stop at the desired
height to serve as the support as shown in Figure 8. The stop position of each metal rod is
determined by the inserted metal tube of pre-defined length with a magnetic ring on its tip (See
Figure 10). When the metal tube touches the second-layer metal sheet, the metal rod will no longer
be able to move up due to the constraint of the metal tube. Meanwhile, the metal tube will break
the attachment between the bottom magnetic disc and the third-layer metal sheet. After separating
from the third-layer sheet, the metal rod will be fixed to the second-layer metal sheet due to the
magnetic ring on its top and magnetic discs at the bottom of the rod. Hence the metal rod will
maintain at the same height thereafter even though the third-layer sheet continues moving up.
When the printing process is finished, the metal supports can be easily removed from the built
object. The lifted metal rods can be pulled down automatically and reset to the original state. The
metal support can be reused for different 3D objects by changing the metal tubes of different
lengths. Thus, the material cost and printing time will be significantly reduced. Note among all the
3D-printed supports, no change will be made for the internal supports that do not touch the building
platform at present.

Figure 7. The metal pin-like support and the 3-layer sheet structure design.
32

Figure 8. The schematic diagram of an FFF system with the reusable metal support mechanism.
The integrated FFF system incorporating the support mechanism consists of a general FFF
3D printer with a motion controller, pin-like metal supports, a 3-layer sheet structure, and an extra
motorized linear stage (See Figure 8). The extra linear stage only connects the third-layer metal
sheet. In our design, all the required metal support pins can be moved up during fabrication by
only one motor. At the same time, the metal pins can stop at the desired position. The structure of
each component is explained in more detail in the remainder of the section.
4.1.1.2 Pin-like support and three-layer sheet structure
The pin-like metal support is used to replace the 3D-printed support. Figure 9 (a) shows its
detailed structure. Each pin is composed of a metal rod, a square metal washer, a magnetic ring,
and a sleeve. These metal pins sitting on the fixed first-layer sheet form the printing platform as
shown in Figure 9 (b). Each pin will provide an individual small anchor surface on which the part
is printed. The magnetic ring beneath the washer makes the finished part easy to be taken off the
metal rods after fabrication. After peeling off the attached metal washers, the built part is totally
33

separated from the metal supports. The sleeve and metal rods hold the magnetic ring. The metal
washer adheres to the magnetic ring so that it is fixed on the metal rod.

Figure 9. The detailed structure of the first layer. (a) Construction of the pin-like metal support. (b) Printing platform
based on the reusable metal supports.

Figure 10. The detailed structure of the second and third layers.
The first layer is a fixed plate that holds all the metal pins at their original place to form the printing
platform. Raising metal supports is accomplished by using a motorized linear stage to move up the
third-layer metal sheet (Refer to Figure 10). The magnetic discs attaching to the third layer will
push the metal pins upwards until the related metal tubes touch the second-layer metal sheet. The
second-layer sheet limits the further movement of the metal supports when the combination of the
metal tube, magnetic ring, and magnetic disc touches the second layer and holds the metal supports
with the tube combination after they detach from the third layer. Note only the required metal
supports have magnetic discs and metal tubes. They are manually inserted before each printing job.
34

4.1.1.3 3D printing process based on reusable support
The reusable metal support works with the typical FFF process coordinately. The detailed
description of this process is given as follows:
1) Before the printing process, the metal tubes with magnetic rings on their tips and magnetic
discs on the bottom will only be added to the places where the metal supports are needed so that
all the other pins will stay at the original position on the platform during the printing process. The
first-layer sheet holds the metal supports. The magnetic discs at the bottom of the metal rods adhere
to the third layer. The metal tube combinations can be inserted from the bottom of the third-layer
sheet manually as shown in Figure 11 (a). The self-developed software selects the lengths of the
metal tubes and where to insert the tubes according to the given part geometry.
2) When the print job begins, the magnetic discs will push the metal rods as well as the metal
tube combinations up by one-layer thickness each time after finishing the fabrication of a layer
(refer to Figure 11 (b)).
3) The metal tube combinations will stop moving up once touching the second-layer sheet.
The tube will attach to the second-layer sheet due to the magnetic ring. The metal tube also
prevents its magnetic disc from moving up further. When the third-layer sheet continues moving
up, the magnetic disc at the bottom of a metal tube will detach from the third-layer sheet and keeps
the metal rod in its current position. This process is shown in Figure 11 (c) and Figure 11 (d).
4) After the magnetic disc detaches from the third-layer sheet, the corresponding metal rod is
held at the same position by the metal tube and the second layer. Residual 3D-printed support or
the part itself can then be printed on the metal washer of the pin. The third layer will continue
35

moving up until all the metal rods have reached their desired positions as shown in Figure 11 (e)
and Figure 11 (f).
5) After the printing process, the third-layer sheet moves down to the original height, which
is the reverse direction of the printing process. The third-layer sheet will pull down the tube
combinations together with the metal pin and detach them from the second layer-sheet. The metal
washers attaching to the built part can be separated and inserted back to the tip of the metal rods.
For different CAD models, the original tubes and magnetic discs are replaced with new ones
manually. The reusable metal support is ready for the next printing job.

36

Figure 11. Sectional view of the 3D printing process with metal supports: (a) The original state of the metal supports. (b)
One-layer movement of the third layer. (c) (d) The separation process due to the metal tubes. (e) (f) The final state of the metal
support.
The metal support height is related to the original distance between the second layer and
third-layer sheets, length of the metal tube, and thickness of the magnetic ring. Their relationship,
as shown in Figure 12, can be calculated as (4).

23 Layer Tube Magnet
H L L L
−
= − − (4)

Figure 12. The height of a metal pin support (Left half is the original state of a metal support and the right half is the final
state of the same metal support)
37

In our design, the metal pins are mounted on the FFF 3D printer. The metal tubes which could
provide support of specific lengths (such as 5 mm, 10 mm, 15 mm, and so on) are given to users.
Similar to Lego blocks, users can achieve different or longer standard length by assembling the
short metal tubes of standard lengths. For an arbitrary CAD model, we developed slicing software
to calculate the position of the metal supports and the related tube lengths that are required for the
desired support heights. For a CAD model that requires support at a different height from the
combination of the provided tubes, the nearest height to the standard length that is slightly shorter
will be selected. Accordingly, the metal pin will rise to the selected height. The remaining portion
of the support will be 3D printed on top of the metal pin. An example is given in Figure 13. For
the wooden dummy model, the heights of the four arms are 7.5 mm, 12.5mm, 17.5mm, and
22.5mm, respectively. Suppose the nearest standard lengths of the metal supports are 5 mm, 10
mm, 15mm, and 20mm, respectively. Therefore, an additional 2.5 mm long support will be
generated by our software and will be 3D printed on each metal pin.

Figure 13. Support planning generated by our slicing software.
4.1.2 Layout optimization
The material consumption for residual 3D-printed support is affected by three factors: (1) the
layout of the CAD model on the printing platform; (2) the height of each metal support; (3) the
structure of the 3D-printed supports. Many aforementioned 3D-printed support structures can be
directly applied to our setup. In this paper, we employed the vertical wall-like structure to print the
38

residual support. This structure will increase the robustness and reliability of our process. Each
metal support should reach its highest level while not intersecting with the CAD model to save the
printing material and printing time as much as possible. Therefore, the layout of the CAD model
is the only factor that affects the volume of the residual printed support. For traditional 3D printing
processes, given a model with a fixed orientation, the material consumption remains unchanged
whatever translation on the printing platform or rotation around the z axis is made to the model.
However, metal support in our process will undergo a remarkable change during the movement of
the model. Figure 14 shows two examples to explain this layout effect. A slight shift of the CAD
model could make a great difference in the demand for 3D-printed supports. This unique
phenomenon indicates that performing layout optimization before printing is necessary in order to
take full advantage of our metal support. In this paper, we will search the best position to set a
CAD model by translating the CAD model along the x and y axes, which refer to the printing
platform, and rotate it around the z axis, which is vertical to the printing platform. Here, we don’t
consider the rotation around the x and y axes because we hope the user has control over the
model’s orientation that affects the surface quality.
39

Figure 14. Comparison of 3D-printed support between the original layout and optimized layout. The red part represents
the extra 3D-printed supports. (a) Wooden dummy model before layout optimization. (b) Wooden dummy model after layout
optimization. (c) Roof model before layout optimization. (d) Roof model after layout optimization.
Due to the periodic shape of our printing platform, the translation range in , xy directions is
0, x y L     , where L is the size of each metal support. The variable range of the rotation
around the z axis is 02
Z
  . Besides, moving the part in z direction, which means printing
a base beneath the model, sometimes could save more material by moving up some metal pins to
its next level. The translation range in z direction is
0
0 zH    , where
0
H is the shortest
standard length of the metal support.
4.1.2.1 Optimization formulation
Only the external 3D-printed supports that touch the printing platform will be truncated and
replaced by the metal supports. These external 3D-printed supports are sandwiched between the
part itself and the printing platform in the traditional case. Given a CAD model with a certain
rotation angle around the z axis
Z
 , suppose the projection of the region to add supports on the
printing platform is represented by
Z
XY
S

. The support height of each point within the range
Z
XY
S


is ( , ) z x y . Then the maximal 3D-printed support volume is:
40

0
( , )
XY
Z
S
V z x y ds

=

(5)
Suppose the printing platform compose of mn  metal pins, where m is the number of rows
and n . is the number of columns. The range of each metal pin is
, rs
P and its corresponding height
is
, rs
H . Then the whole printing platform can be expressed as a set:

 
,
0 ,0
rs
P P r m s n =     (6)
Then the total material cost for the 3D-printed support ( , , , )
material Z
C x y z     after employing
the metal support is:

,
0 , ,
Z
rs
material r s XY r s
PP
C V P S H


=−

(7)
To fulfill the potential of our machine, the height of each metal pin should be:

,
, 0 0
( , )
min ( , ) /
rs
rs
x y P
H H z x y H

= 

(8)
The number of metal pins to be used should also be taken into account to cut down the
machine setup time. The number of metal pins is:

0
,
1
H
rs
pin
C

=

(9)
Since printing in the center of each metal pin is more preferred than the boundary for
reliability reasons, we should try to reduce printing at the boundary of the metal pin. Suppose
,, r s r s
EP  is the edge of each pin. Then the total area of the 3D-printed supports located at the
boundary region of the metal pins is:
41

,
,
0
Z
rs
edge r s XY
H
C E S


=

(10)
Therefore, the objective function for the layout optimization is:

, , ,
, , ,
0 , , ,
, , ,
00
min ( , , , ) min ( )
min ( 1 )
ZZ
r s r s r s
material pin edge
x y z
r s XY r s r s XY
x y z
P P H H
C x y z C C C
V P S H E S



  

   
   
  
    = + +
= − + +
  
(11)
subject to:
0
0 , ;0 ;0 2
Z
x y L z H          
4.1.2.2 Discretization and enumeration

Figure 15. Illustration of discontinuity in support volume during movement. (a) Original door model without any support.
(b) Layout resulting in no 3D-printed supports. (c) Layout resulting in maximal 3D-printed supports.
The door model case (refer to Figure 15) illuminates a common issue in solving the layout
optimization function. The material consumption to print supports is a step function with only two
values in this case though the variables are continuous. This is caused by discontinuity of metal
support number and the limited selection of the metal support height. It is difficult to guess a good
initial value to start a conventional iterative optimization method. Enumeration method, though
the running efficiency is usually low, can guarantee a good search result and much more robustness.
Therefore, we adopt an effective and efficient enumeration method in our scenario to satisfy the
requirement of a GUI application.
42

Figure 16. Discretization of metal support with the boundary region highlighted in blue.
First, a 3D CAD model is sliced to a set of polygons which indicates the shape of each layer
as many commercial slicers do before layout optimization. Then the layout optimization stage
begins. The
0
V in the objective function (8) is always a constant if not move the model along the
z-axis. Regarding the other three items in equation (8), computing
, rs
H costs most of the running
time in each iteration. To accelerate the search process, we approximate the variables by
discretizing the printing platform. The sub-platform provided by each metal pin is further divided
into
2
K ( 50 K = in our case) grids. Each grid is a sampling point recording the shortest distance
between the printing platform and the CAD model by layer serial number as shown in Figure 16.
This process can be viewed as shooting a ray from each sampling point and intersect with the
sliced polygons of the model. Only the lowest layer with intersection will be recorded. The grids
that do not intersect with the model are marked by + . Then the whole printing platform can be
further represented by a mK nK  matrix h which is called heightmap:

 
,
0 ,0
ij
h h i m K j n K =       (12)
43

The minimum layer serial number within each metal pin range
, rs
P determines the maximum
height of each metal support
, rs
H . This is a classical 2D RMQ (Range Minimum Query) problem
and can be processed quite fast by dynamic programming. The pre-processing stage takes
22
( (lg ) ) O K K time, and the query stage only takes (1) O time.
After digitizing the printing platform, the material consumption to print support
material
C is
converted to the sum of height at each sampling point that intersects with the CAD model:

,,
,,
( , )
()
r s r s
material i j r s
P P i j P
C h H

=−

(13)
Similarly, the total area of the 3D printed supports located at the boundary region (see the
blue region in Figure 16) of the lifted metal pins can be characterized by the number of grids:

,
,,
0
,( , )
1
rs
i j r s
edge
H
h i j E
C

 +  
=

(14)
At this point, the value of the objective function can be calculated. However, heightmap will
change in each iteration due to the translation and rotation of the model. The step size of x  and .
y  . in search is / LK . The step size of
z
 is 2/ N  ( 360 N = in our test cases). Instead of
transforming the model, the heightmap can be updated very fast by translating and rotating the
initial height map. Moreover, translation of the heightmap matrix can be achieved by re-querying
the elements of different indexes so that no need to perform pre-processing again after each
translation. Certainly, after each rotation, the heightmap needs to be pre-processed before querying.
Therefore, the total time complexity is
22
( (lg ) ) O K N K and is only dependent on the step size in
search. The number of triangles in the CAD model does not affect the processing speed.
44

As for translation in z direction, we can check and compare the value of the objective
function at last. The translation distance must be multiples of the layer thickness and smaller than
the minimum standard length of the metal support. Therefore, we only need to increase the
minimum layer serial number of each metal pin by the multiple and recalculate the height of each
metal support
, rs
H . After these operations, the layout of the part on the printing platform is fixed.
It usually takes about half a minute to finish the process.
4.1.3 Toolpath of residual printed support
Residual supports are required to print on the metal supports. The printed supports should be
allocated to each metal support because the metal supports may not stop at the same height. This
is achieved by grouping the grids marked by the layer serial numbers in the heightmap h into a
set of convex polygons within each metal pin (see Figure 17 (a)(b)(c)). By filling each convex
polygon with a certain pattern and connecting the neighboring polygons, the tool path can be
generated.

Figure 17. Creating Fermat spirals within a pin. (a) Grids allocated to a pin. (b) Grouping of neighboring grids. (c)
Convex polygon formed by each group of grids (d) Level set of the polygons (e) Connected Fermat spirals.
Zhao et al. developed a new kind of “space-filling” pattern - connected Fermat spirals. This
pattern can fill an arbitrary 2D polygon without a huge change of curvature. The spiral can start
and end in any given points of the polygon, which provides more freedom in determining the path
between different metal pins. In addition, this type of pattern is smoother and more stable
compared with traditional patterns such as zigzag and grid.
45

In order to fill each polygon with connected Fermat spirals, the level set of each polygon is
generated by shrinking the convex polygons equidistantly (see Figure 17 (d)). These level sets
form the basis for the connected Fermat spiral. By cutting the level sets and re-linking them, they
are converted to connected Fermat spirals as shown in Figure 17 (e).
The entry point and exit point of each polygon need to be specified before cutting and re-
linking the level set of each polygon. Besides, we need to determine the printing order of the
polygons to reduce the travel distance. We simply calculate the center of each polygon ( , )
ii
xy and
sort the sums of
i
x and
i
y . Suppose the origin is the top left corner of the printing platform. The
print always starts from the top left corner to the bottom right corner for one layer and reverses the
order for the next layer. Then the entry points and the exit points of the outermost polygon of each
level set can be determined by the minimum distance between two neighboring level sets according
to the printing order.

Figure 18. Partition of a level set. (a) The entry point and the exit point of the outermost polygon. (b) Entry points and exit
points of the inner polygons (c) Segmented curves.
Suppose the entry point and exit point of an outermost polygon are represented by
0
S and
0
E
respectively (refer to Figure 18 (a)). Then the entry points and exit points of the inner polygons
of the level set can be determined by the closest points on the next inner polygons from the outer
ones. Therefore, we can find the following entry point
1
S and exit point
1
E that are the nearest
points to
0
S and
0
E on the neighbor inner polygon. Repeat the process we will have
0 1 2
, , , S S S
46

and
0 1 2
, , , E E E (see Figure 18 (b)). For each
i
S and
i
E (excluding innermost ones), we set a
point  mm away anti-clockwise on the same polygon as
p
i
S and
p
i
E . Then we cut each polygon
except the innermost one into two parts ,
ii
ff
+−
by deleting the segments
p
ii
SS and
p
ii
EE (see
Figure 18 (b)(c)). For the innermost polygon, we only keep the shortest segment. Figure 19 shows
how to connect these separated curves.
0 1 2
, , , f f f
+ − +
form the clockwise portion of the connected
Fermat spiral and
0 1 2
, , , f f f
− + −
form the anti-clockwise portion. Finally, we can combine both
parts into one whole Fermat spiral.

Figure 19. Connecting the segmented curves. (a) The clockwise portion of the Fermat spiral. (b) The complete connected
Fermat spiral.
The filling toolpath can be smoother using the optimization method mentioned in (Traditional
filling patterns like zig-zag or grid, in contrast, are too difficult to have any improvement). Besides,
for the top layers of the residual printed supports, part of the Fermat spirals have to be removed
through Boolean operation to avoid intersection with the CAD model.
4.1.4 Experimental validation
4.1.4.1 Prototype of FFF 3D printer
The prototype of the FFF 3D printing system incorporating the reusable metal support is
developed to demonstrate the novel manufacturing process (Figure 20). The system is built on a
low-cost commercial FFF 3D printer-Orion Delta. Our reusable metal support is in the bottom half
47

of the printer and highlighted in the yellow box in Figure 20. The printing platform is composed
of 9 11  metal pins and an acrylic sheet beneath them (Figure 21). Undoubtedly, with smaller
washer used for each pin, we can save more material and printing time. Here, for laboratory
demonstration, the size of the wash is 12.7 mm  12.7 mm (0.5 in  0.5 in). We are confident there
is plenty of room for reducing the size of each pin. In addition, it is more reasonable to use square
or hex metal bars instead of metal rods to avoid the potential rotation of the metal support. However,
this never happens in our experiments due to the attraction of the magnets. The second and the
third-layer metal sheets are made of two acrylic sheets with metal washers inside as an interlayer
instead, which can achieve the same effect (refer to Figure 22). As for the heat bed, we plan to
enclose the 3D printer and heat the whole chamber as other commercial 3D printers do in the future.

Figure 20. FFF 3D printer prototype with reusable metal supports.
48

Figure 21. Printing platform.

Figure 22. The second layer and the third layer.
Compared to the general FFF 3D printing process, the reusable metal support operates after
the fabrication of each layer. The motorized linear stage used to push up the metal pins works as
the fourth axis and represented by A-axis. The command “G1 Axx” is added between the G-code
of two adjacent layers to control the movement of the metal supports, where “xx” is the travel
distance. The G-code exported from our slicing software is loaded into Repetier-Host 1.6.2, client
software running on PC to communicate with the modified Repetier firmware 0.91. When the
49

motion controller executes the self-defined command, the linear stage will move up the metal pins
by one-layer distance.
4.1.4.2 Support generation software
Our slicing software (written in C++) is developed based on a library libigl for mesh
processing and user interface. All the test cases (refer to Figure 6 and Figure 23-Figure 26) are
performed on a PC equipped with Intel Core i7 6770 3.4GHz and 8GB RAM. We measure the
execution time of the layout optimization and the whole process from the input of an STL file to
the output of the G-code. The collected time data are shown in Table 1. From Table 1, layout
optimization takes most of the execution time. The total execution time is still reasonable.
Table 1. Statistics of software running time.

4.1.4.3 Experimental results and analysis
Several objects (see Figure 6 and Figure 23-Figure 26) have been 3D printed on our
prototype using PLA. The printing speed is 30 mm/s. The layer thickness is 0.2mm. Two groups
of tests with and without the proposed metal supports are implemented. The pattern of the printed
supports is the same for the two groups.
50

Figure 23. Wooden dummy. (a) Printing result of a wooden dummy with metal supports. (b) Finished part.

Figure 24. Roof. (a) Printing result of a roof with metal supports. (b) Finished roof after cleanup.

Figure 25. Teapot. (a) Printing result of a teapot with metal supports. (b) Printing result of teapot without metal supports.
(c) Finished teapot after cleanup.

Figure 26. Helmet. (a) Printing result of a helmet with metal supports. (b) Printing result of the helmet without metal
supports. (c) Left: residual printed supports removed from (a); Right: printed supports removed from (b).
The standard length of metal support in our prototype is multiples of 5 mm. Since the heights
of the four arms of the wooden dummy in Figure 23 (a) are 15 mm, 30 mm, 45 mm, and 60 mm,
there is no need to print support. For other models, the printed support is shortened. This enhances
the reliability of the printing process because the torque resulting from the nozzle tends to topple
51

down the part and cause print failure. The torque becomes more significant as the printed support
becomes higher. This can be seen in Figure 6 (b). A very long printed pillar supports the right foot
of the gymnast (highlighted by the red box). The printed support collapses to some extent in the
upper portion though the print is successful. In contrast, the printed support in Figure 6 (a) is very
short with metal support. The print process is more stable, and no collapse happened. For the
internal supports like those inside the handle of the teapot in Figure 25, keep the original support
structure intact. In Figure 26 (a), nine metal pins rise inside the helmet. For the convenience of
comparison, the printed supports of two helmets are collected and shown in Figure 26 (c).
Table 2 summarizes statistics for all the test models, including printing time, material
consumption, and the percentage of reduction in printing time and material consumption. It is
reasonable to make comparison analyses with printed support filled by the same pattern. The
statistics show that our method brings significant superiority. The average saving on the material
is 64.7% (ranging from 22.6% to 100%) if compared with traditional printed support and 35.3%
(ranging from 14.0% to 59.1%) if compared with whole parts. The average saving on the printing
time is 63.0% (ranging from 20.9% to 100%) if compared with pure 3D printed support and 38.1%
(ranging from 14.1% to 64.3%) if compared with whole parts. The material and time consumption
is dramatically reduced while the reliability of print is increased.
52

Table 2. Statistics of printing examples

4.1.5 Summary of this work
This project presented a novel support technique for 3D printing - automatic reusable metal
support. We demonstrated our method on an FFF 3D printer and develop slicing software for the
new process. We use metal pins to replace the traditional printed support. The metal pins are raised
by only one motor during fabrication. The stop positions of the metal supports are determined by
manually inserted metal tubes with pre-defined lengths. Each metal pin works as a small build
platform. PLA is deposited on the top surface of metal pins to print the residual support or the part
itself. After the printing job, the metal pins can be easily detached from part and reused. The slicing
software generates connected Fermat spirals as toolpath for remaining 3D-printed support and
optimizes the layout of a part on the print platform to make the utmost of our setup. Sufficient test
cases have verified the capabilities of our reusable support in saving support material and printing
time.
53

4.2 A vibration-assisted separation method for constrained-surface-based
VPP
For bottom-up based stereolithography, a separation process is required to detach the newly
cured layer from the constrained surface in the fabrication process. Excessive separation force will
cause damage to the built layers and the constrained surface. Different surface coatings, platform
motions including tilting and sliding, and the utilization of oxygen-permeable films have been
developed to address the separation-related problems. Among these approaches, the vibration-
assisted (VA) separation method to reduce the separation force has limited study. The underlying
mechanism of the VA-separation-based method remains unexplored, and the best way to use VA
separation in the bottom-up based VPP process is still unclear. In this paper, a new VA separation
design for the VPP process is presented. A prototype system was built to study the VA separation
mechanism. Experiments on the separation performance under different parameters, including
vibration frequency, pre-stress level, and exposure area, were conducted. Based on the collected
separation force data, an analytical model based on the mechanics of fatigue fracture was built.
The separation behaviors related to different shape size and topology were also studied and
compared. The results showed that the separation force in VPP was significantly reduced using
the VA-separation-based method. Furthermore, the relationship between the separation force and
the separation time conforms to the stress-based fatigue model. This study also provides insights
on how to choose process parameters by considering the trade-offs between separation force and
building efficiency.
The main objective of this work is to present a new VA separation hardware design for the
bottom-up-based VPP process and to identify an analytical model based on the mechanics of
fatigue fracture. Accordingly, the influence of various factors, including layer area, layer shape,
54

and cyclic loading parameters on the VA separation process, will be evaluated. The obtained
knowledge from the study will help the development of the VA separation method for the
constrained-surface-based VPP process.
4.2.1 Analytical model of the vibration-assisted separation method
In the presented work, we incorporated the vibration fatigue mechanism in the pulling-up
separation process so that the separation force can be significantly reduced. According to Liravi et
al. [59][66] and Ye et al. [61], the direct pulling-up separation in the constrained-surface-based
VPP has been successfully demonstrated to be a fracture behavior between different materials.
After the fabrication of each layer, the newly cured layer and the coated film will form a laminated
composite structure. Therefore, the failure analysis of the laminated composite structure can be
analyzed using a method such as CZM. Similar to the brittle fracture that happened at the interface
between the cured part and the coating media during the direct pulling-up or the tilting methods,
vibration can also be used to break the interface. The main difference is that the pulling-up and
tilting process is the one-shot separation, while the VA separation results from progressively
accumulated physical damage under the cyclic loading.
From this perspective, the VA separation will be modeled using the concept of the vibration
fatigue failure process. There exist three major approaches that have been widely used to analyze
fatigue-related failures [106]-[111].
(1) The stress-based approach directly relates the fatigue life with the nominal (average) stress
and the stress amplitude in the affected region of the engineering component. The stress-fatigue
life (S-N) curve is the basis for the stress-based analysis. The effects of variables such as stress
level and geometry will be reflected in the trends or intercepts of the S-N curves. Generally, this
55

approach is mainly used to predict fatigue life. It is also applicable to complex geometry. Since we
care more about the separation force and separation time when evaluating a separation method and
the force data can be easily measured during the separation process, the stress-based approach is a
suitable candidate for our purpose.
(2) The strain-based approach is similar to the stress-based approach, but it explores strain
versus fatigue life. Instead of the nominal stress versus fatigue life curve, the strain versus fatigue
life curve is used as the basis in the analysis model. Based on the analysis of local yielding, the
strain-based method gives improved estimates for ductile materials under short fatigue cycles.
However, the method is not suitable for the detachment of the adhesive bonding between the
solidified resin and the coated film because little yielding phenomenon occurred.
(3) The crack-growth-based approach mainly focuses on the evolution of the crack growth
rate. Based on the stress intensity factor, the severity of initially existing defects, such as cracks
sustained during service, can be estimated. The estimation is important in setting periodic safety
inspection interval or damage-tolerant design, which requires structures to be able to survive even
in the presence of growing cracks. However, this method cannot predict crack initiation and can
only predict the propagation of the existing crack. Therefore, pre-existing crack and definite
propagation direction are required in this method [61]. However, no initial crack exists before
separation in the VA-separation-based SL. The crack initiation position and propagation direction
are stochastic in the separation process. Also, it is challenging to integrate the equipment that can
measure the crack growth rate into a VPP machine.
56

Figure 27. An illustration of cyclic loading. Cyclic stress amplitude 𝛥𝑆 represents the interval between the maximum and
minimal stress, and 𝑆 𝑚𝑖𝑛 represents the minimal stress.
In summary, a stress-based fatigue model will be applied to our VA separation system and
the related experiment design. Since the stress-based fatigue model was derived from the S-N
curve, two force sensors were mounted in our prototype system to measure and control the force
applied on the part-film separation portion throughout the whole fatigue life in real-time. In the
stress-based fatigue theory, cyclic stress amplitude S  shown in Figure 27 has a strong
relationship with fatigue life. The fatigue life was counted by the number of load cycles. High load
frequency can shorten the required time to accomplish the break of the component. Therefore,
compact equipment that can provide high-frequency load is desired in the VA-VPP system. For
this reason, two piezoelectric actuators were incorporated in the prototype system to provide cyclic
load at high frequency. Besides, the pre-stress
min
S applied to the component (see Figure 27) also
played a vital role in determining fatigue life. The pre-stress was achieved by stretching the part-
film interface when adding the vibration. The next two sections will give a detailed description of
the whole prototype system design and the related process investigation using the developed
prototype system.
57

4.2.2 Design of VPP system with VA separation mechanism
There are many possible configurations to achieve the VA separation mechanism design. One
configuration for the VA-VPP system is shown in Figure 28. Other variations of the VA
configuration can be made based on the presented design logic.

Figure 28. The prototype system using the VA separation mechanism. (Left) The hardware of the VA-separation-based
VPP with two piezoelectric actuators; (Right) the control and circuits required in the VA-separation-based VPP system.
The cyclic load was provided by two piezoelectric actuators (MTKK16S400F170R, Mechano
Transformer Corp., Tokyo, Japan). The transparent glass resin vat coated with Teflon film
(Shanghai Witlan Industry Corp., Shanghai, China) was connected with the two piezoelectric
actuators directly, and then mounted to a fixed aluminum frame. The piezoelectric actuators were
powered by the PiezoMaster power amplifier (VP7206-24K150, Viking Industrial Products,
Marlboro, Massachusetts, USA). The piezoelectric actuators only provided load downwards once
energized, keeping its top frame static. The cyclic loading was controlled by the DAC output of
an Arduino microcontroller. By varying the driving voltage, the piezoelectric actuator provided
different forces onto the newly cured layers. Two force sensors (FlexiForce A201 Sensor, Tekscan,
Inc., South Boston, MA, USA) were sandwiched between the fixed aluminum frame, and the
piezoelectric actuators were used to measure the separation force in real-time. Based on the
58

feedback of the force sensors, we controlled the force applied to the part-film interface. An Op-
amp circuit (FlexiForce Quickstart Board, Tekscan, Inc., South Boston, MA, USA) amplified and
converted the sensor signal to voltage. The Arduino microcontroller collected the voltage data
through its ADC inputs and then sent the readout to a computer. Since the resin vat was only fixed
at the top, the pulling force from the part was transferred to the sensors when the building platform
raised. Thus, the sum of the two measured forces is equal to the separation force between the newly
cured layers and the constrained vat surface. The resultant force data sequence reflected the
separation process and formed the basis of our analysis.
The linear stage to mount the building platform was driven by a KFLOP motion controller
and a SnapAmp amplifier (Dynomotion Inc., Calabasas, CA, USA). The 405 nm ultraviolet (UV)
digital light processing (DLP) projector served as the irradiation light source. A self-developed
software system synchronized the whole experimental process, including sending commands to
the microcontrollers and the DLP projector and acquiring data from the sensors.
4.2.3 3D printing process with the VA separation method
The VA separation mechanism works with the typical bottom-up SLA process. A detailed
description of the VA-VPP process is given as follows:
59

Figure 29. 3D printing process with the VA separation-based method. (a) Projecting a mask image to cure the liquid resin;
(b) moving up the building platform by a small distance 𝛥 𝑥 1
𝑚𝑚 to provide a pre-stress 𝑆 𝑚𝑖𝑛 to the separation region; (c)
applying cyclic load by driving the piezoelectric actuator with cyclic voltage 𝑉 ; (d) after the period of 𝛥𝑡 , the part-film interface
is separated; (e) moving up the building platform by 𝛥 𝑥 2
mm to facilitate resin refill; and (f) moving down the building platform
by 𝛥 𝑥 2
mm to form a one-layer-thickness gap (𝛥𝑙 ).
(1) When the fabrication of one layer begins, the DLP projector projects a mask image to
photocure a thin layer of liquid resin. After a certain exposure time, the liquid resin is solidified
between the previously cured layers and the coated film (using Teflon), as shown in Figure 29a.
(2) After the curing process, a computer begins to read and record the readouts of the two
force sensors until the end of the separation process (see Figure 29b-Figure 29e). Meanwhile, the
building platform moves up a small distance
1
x  mm to provide a pre-stress
min
S to the separation
region. Once the specified pre-stress
min
S has been achieved, the building platform will be held in
the position, as seen in Figure 29b.
(3) When the movement of the building platform is complete, the controller drives the
piezoelectric actuator with cyclic voltage cos
aa
V V t V  = − + . The interface between the newly
cured layers and the coated film will be under cyclic load. The peak voltage
a
V is determined by
the feedback of the force readout. The state will last for a time t  so that the fatigue failure occurs
within this period. After the period of t  , the piezoelectric actuator is disabled. This process can
60

be seen in Figure 29c to Figure 29d. The force data have been continuously recorded during this
period.
(4) After the piezoelectric actuator is disabled. The platform moves up a sufficient distance
2
x  to make sure the part-film interface is separated if the fatigue life is longer than t  and
facilitates the resin refilling, as shown in Figure 29e. At the end of the step, the process of
collecting the force data will be stopped. Till now, the whole separation process is complete.
(5) Finally, move down the building platform by
2
x  to form a one-layer-thickness ( l  ) gap,
as shown in Figure 29f. Now the setup is ready for the curing of the next layer (Figure 29a).
4.2.4 Experimental and Analysis Results
4.2.4.1 Experimental Design
For the VA separation method, many factors may affect the required separation time,
including loadings condition (frequency, cyclic stress amplitude, and pre-stress), exposure area
sizes, geometric topologies, and geometric shapes. In order to study the relationship between the
separation time and these factors, five test cases were designed. They are discussed as follows.
• Test case 1 – different loading frequency
To study the frequency effect, a cylinder was designed and fabricated using the VA
separation-based method with different loading frequencies. The driving voltage of the
piezoelectric actuator was fixed, while the frequency was varied. The parameter settings are listed
in Table 3. There were 8 parameter settings in total, and each experiment was repeated 10 times.
The first 20 layers were printed as a base. Data collected from the rest layers were used in the
analysis. The parameter sets were used alternately during the fabrication to avoid the errors
induced by the height.
61

Table 3. Parameter setting for test case 1.

• Test case 2 – different cyclic load amplitude and pre-stress
To compare the effect of the cyclic stress amplitude S  and the pre-stress
min
S on the
separation time, a cylinder was fabricated using the VA-separation-based method with different
process parameters. The frequency was fixed at 500 Hz, and the loading force was varied by
controlling the driving voltage of the piezoelectric actuator. The parameter settings are listed in
Table 4.
Table 4. Parameter setting for test case 2.

• Test case 3 – different exposure area
Following the same process described in test case 1, cylinders of various diameters were
printed to investigate the effect of exposure areas on the separation performance, including the
separation force and time. Under the same voltage, the stress will decrease with the increase of
exposure areas. Thus, the pre-stress was raised to shorten the fatigue life. The parameter settings
are shown in Table 5.
62

Table 5. Parameter setting for test case 3.

• Test case 4 – different geometric topologies of the same area size
The effect of geometric topologies on the VA separation behavior was studied by fabricating
a larger single cylinder and smaller cylinder arrays following the aforementioned process. Each
cylinder array is composed of a set of smaller cylinders of the same area. The sum of the cross-
sectional area of each cylinder array is equal to the area of the larger single cylinder. The parameter
settings are shown in Table 6.
Table 6. Parameter setting for test case 4.

• Test case 5 – different geometric shapes of the same area size
The geometric shape may also affect the VA separation time. Thus, various shapes of the
same area were designed to explore the influence of the geometric shapes, including circle, ring,
triangle, square, pentagon, and pentagram. The corresponding parameter settings are listed in
Table 7.
63

Table 7. Parameter setting for test case 5.

4.2.4.2 Result Analysis
4.2.4.2.1 Different Loading Frequency
To identify the optimal loading frequency, we printed the same part with a gradually ramped-
up frequency from 100 Hz to 800 Hz under the same driving voltage 150 V. The corresponding
force amplitude and separation time data of different frequencies are plotted in Figure 30.

Figure 30. Relationships between force amplitude and separation time related to frequency in the current setup.
From Figure 30, we can see the loading frequency and force amplitude are coupled in the
current VA-VPP system. The increase in the loading frequency from 100 Hz to 500 Hz leads to a
reduction in separation time. However, when the frequency is higher than 500 Hz, the separation
64

time starts to increase. The time increase is because the increase in frequency results in a dramatic
decrease in amplitude if the frequency is higher than 500 Hz. Therefore, after synthetically
considering the effect of frequency on the amplitude range and separation time, we decided to use
500 Hz as the loading frequency to fabricate all the rest test cases.
4.2.4.2.2 Different cyclic stress amplitude and pre-stress
The test results regarding loading amplitude and pre-stress effect on the number of cycles
required in the VA separation method are listed in Table 8 with a diagram shown in Figure 31.
Table 8. Test results of the number of loading cycles to separation under different loading conditions.

Figure 31. (a) S-N curves and the corresponding separation time under different pre-stresses. (b) Cyclic numbers and the
corresponding separation time under different pre-stress.
As shown in Figure 31a, the peak stress level
max
S of the same pre-stress
min
S and the
number of cycles
f
N approximate a straight line on a log-log plot. Therefore, Equation 15 can be
used to fit the obtained data. The bar chart Figure 31b reflects the number of cycles under different
cyclic stress amplitude S  . For the same pre-stress, larger cyclic stress amplitude can reduce the
65

number of cycles required to achieve total separation and therefore shorten the separation time
(Figure 31b); however, it also leads to larger peak stress and separation force (Figure 31a). For
the same cyclic stress, larger pre-stress can also reduce the separation time (Figure 31b). However,
the peak stress level will also be larger.

max
B
f
S AN = (15)
Both factors have a similar effect. However, for the same maximum stress, which is the sum
of the cyclic stress and the pre-stress, larger cyclic stress and lower pre-stress will reduce the
number of cycles. The reduction qualitatively indicates that the contribution of the cyclic stress on
the separation is more significant than the pre-stress, and the cyclic stress factor is more significant
than the pre-stress to the separation time.
Therefore, the Walker equation was utilized to consider the effect of the pre-stress
quantitatively.

11
min max
() S S S S S S
    −−
 = +   =  (16)
Where S  is the equivalent cyclic stress range as if the pre-stress equals 0, the value of
constant  is limited to the range [0,1].  can be viewed as a measurement of the sensitivity to
the pre-stress. Low values  correspond to high sensitivity, and values approaching 1 to low
sensitivity. If 0  = , then it indicates the pre-stress, and the cyclic stress takes equivalent effect.
1  = represents the pre-stress has no effect. Therefore,  value allows a quantitative estimation
of the two factors. Combine equation 15 and 16, and then solve for
f
N :

1
1
max
1
B
f
N S S
A
 −

=


(17)
66

Taking the logarithm to the base 10 of both sides will give

max
11
lg( ) lg( ) lg( ) lg( )
f
N S S A
B B B
 −
= +  − (18)
By doing multiple linear regressions with variables
max
lg( ) S , lg( ) S  and lg( )
f
N , we have:
0.5637  = (19)

Figure 32. Equivalent S-N curve and corresponding separation time under different pre-stresses.
Therefore, the equivalent cyclic stress amplitude can be obtained. The converted points from
different pre-stress form a straight line (Figure 32). With the constant γ, for the given circle mask
image, the separation time under different cyclic loading conditions can be estimated using
equation (20).

max
lg( ) lg( ) 0.93lg( ) 1.20lg( ) 12.06
f
N ft S S = = − −  + (20)
4.2.4.2.3 Different Area Sizes
The test results regarding the effect of the exposure area on the number of cycles required in
the VA separation method are listed in Table 9 and diagrammed in Figure 33.
67

Table 9. Test results of the number of loading cycles to separation for different area sizes.

Figure 33. (a) S-N curves and corresponding separation time for different pre-stresses and exposure areas. (b) Equivalent
S-N curve and the corresponding separation time for different pre-stresses and exposure areas.
As shown in Figure 33a, similar behavior as the ones in Figure 31a can be observed, except
a higher pre-stress has to be applied for a larger area to complete the separation within a reasonable
68

time. The peak stress level
max
S of the same pre-stress
min
S and the cycles to separation
f
N
approximate a straight line for different area sizes on a log-log plot. By applying the Walker
equation and multiple linear regression, we can get a straight line fitted by the points of zero stress
amplitude and the equivalent zero stress amplitude, as shown in Figure 33b. The  equals 0.5622,
which approaches the previous value calculated. The
2
R is 0.94, indicating a strong linear
relationship of the testing results. Therefore, for the individual circle pattern of different areas
under different loading conditions, the separation time can be estimated by the following linear
regression model.

max
lg( ) lg( ) 0.81lg( ) 1.04lg( ) 11.01
f
N ft S S = = − −  + (21)
4.2.4.2.4 Different Geometric Topologies
For a given cross-sectional area, the test result for different geometric topologies is presented
in Table 10 and Figure 34.
Table 10. Test results of the number of loading cycles to separation for different geometric topologies.

69

Figure 34. Cyclic number (left) and corresponding separation time (right) for different geometric topologies including
single cylinder, two-cylinder array, and four-cylinder array, respectively.
From Figure 34, the average separation time decreases as the number of cylinders increases,
which can be explained by the increase in the number of individual circles. The total fatigue life
can be divided into two stages—the crack formation and crack propagation [111]. According to
Liravi et al. [59], cracks initiate from the boundary of the specimen because of the local stress
concentration (see Figure 35). For the same shape and area, the pattern with more separate parts
forms a crack tip earlier under the same loading condition, which explains why the crack tip arises
first in the four-cylinder array followed by the two-cylinder array and the last is the single-cylinder
as indicated in Figure 35. Besides, more than one crack initiation was observed for the two and
four-cylinder arrays in Figure 35. More crack initiations also sped up the propagation process and
shortened the separation time.
70

Figure 35. Crack initiation and propagation of patterns with different geometric topologies, including single cylinder
(a)(b)(c), two-cylinder array (d)(e)(f), and four-cylinder array (g)(h)(i), respectively.
4.2.4.2.5 Different Shapes
The test result for different geometry is presented in Table 11 and Figure 36.
Table 11. Test results of the number of loading cycles to separation for different shapes.

Figure 36. Cyclic number (left) and corresponding separation time (right) for different shapes.
71

The separation time varied for different shapes when using the VA-separation-based method.
Among the different shapes, the circle took the longest time for separation. The test results
indicated that a longer separation time was required for the shapes with smooth borders to separate,
compared with the shapes with sharp corners or holes, which is compatible with the findings
reported in Liravi et al. [59] and Lin [112].

Figure 37. Crack initiation and propagation for fabricated different shapes, including circle (a)(b)(c), ring (d)(e)(f),
square (g)(h)(i), pentagon (j)(k)(l), triangle (m)(n)(o), and pentagram (p)(q)(r).
72

For the same boundary shape, the difference between circle and ring mainly stems from the
crack propagation stage [113]. The most important difference of the circle and ring in terms of
crack propagation is that, for the circular shape, the crack propagates toward the middle section;
in comparison, for the ring shape, the crack propagates from the two free openings and propagates
toward the inner portion individually (Figure 37). Hence the different boundary conditions lead
to different crack growth rates.
Shapes with sharp corners form crack tip more easily. The difference is especially obvious in
the case of pentagram and triangle (Figure 37). Therefore, the time for crack initiation was shorter
for pentagram and triangle than the other four shapes. Besides, multi-site crack initiation was also
observed in the pentagram, which reduced the time for crack propagation through the layer. Since
the total separation time is equal to the life of the crack formation and propagation, the pentagram
and triangle took the shortest time for the layer separation.
4.2.5 Process Settings and Comparison Study
4.2.5.1 Comparison of the separation forces in the direct pulling-up and VA separation methods
To compare the separation forces in the direct pulling-up and VA-separation-based methods,
three cylinders with different diameters were fabricated using the two separation methods. The
separation force data are list in Table 12 and plotted in Figure 38.
73

Table 12. Comparison of separation force between the direct pulling-up method and the VA-separation-based method.

Figure 38. Comparison of the separation forces of different areas between the direct pulling-up and VA-separation-based
methods.
To make the results from the two different separation methods comparable, the pulling-up
velocity is set to a slow speed so that their separation time are approximately the same. As shown
in Figure 38, with the VA-separation-based method, the separation force is reduced by around 75%
for different layer areas, which demonstrates the effectiveness of the VA separation method. The
smaller separation force will make the constrain-surface-based VPP process more reliable.
4.2.5.2 Algorithm on the process setting
To integrate the VA separation mechanism into a slicer for the VPP process, an algorithm on
the process settings for the VA-VPP process was developed (Figure 39), which aims to provide
different separation time options for users. For a model with a large layer area or features that can
74

be easily broken during fabrication, a longer separation time needs to be selected so the separation
force will be sufficiently small to increase the reliability of the fabrication process. In comparison,
for a model with a relatively small area, a short separation time can be applied to increase
fabrication efficiency.

Figure 39. The algorithm for the VA-VPP process.
In Figure 39, the total cross-sectional area can be calculated from the mask image of each
layer. If the cross-sectional area is smaller than the pre-defined area limit, the direct pulling-up
without vibration will be applied; otherwise, the VA separation method will be used. The total
cross-sectional area was converted to a circle with an equivalent area. Then the cyclic stress
amplitude was estimated based on the given separation time requirement and the optimal loading
frequency, which may vary for different setup using Equation (21) with the pre-stress setting.
Based on the above study, geometric patterns with disjoint shapes have a faster separation velocity
75

than a connected shape. Besides, the circle has the longest separation time among the six geometric
shapes that were tested. Therefore, the algorithm provides a conservative estimation. Similar to
the “safety factor” in predicting fatigue life, additional vibration time was pre-defined to guarantee
the separation of the part-film interface within the given time. If the max stress amplitude exceeds
the capability of the piezoelectric actuator, the pre-stress using Equation (21) was estimated with
the cyclic stress amplitude setting to the max value. The process was repeated until all the two-
dimensional (2D) layers of the model have been processed. In the end, the planned mask images
and the corresponding G-Code were generated for the given model.

Figure 40. Printing result of the same hourglass using different separation methods. (a) Finished part using the VA
separation method; (b) failed fabrication result using the tilting separation method; and (c) failed fabrication result using the
direct pulling-up separation method.
An example of an hourglass was used to demonstrate the effectiveness of our algorithm. In
the test case, the required separation time was set to be 30 s, among which 10 s was the additional
vibration time. The loading frequency was 500 Hz. The max cyclic loading force was set to 19.5
N. If the circle diameter is smaller than 10 mm, the direct pulling-up separation method was used.
Using the VA separation method, the hourglass was successfully printed (Figure 40a). For the
76

direct pulling-up separation method, the pulling up velocity is 0.5 mm/s. However, the hourglass
broke at the neck, which was the weakest portion of the model, and the built layers split into two
portions (Figure 40c-Bottom). One portion was still connected with the building platform, while
the other attached to the resin vat surface (Figure 40c-Top). The dashed circle indicated the layer
when the split happened. The same part was also tested on a commercial 3D printer MoonRay
(SprintRay, Los Angeles, CA), which adopted a tilting mechanism. However, the splitting of the
two neighboring layers was also observed in the printing result due to the large separation force
(Figure 40b).
The measured separation force and separation time of each layer are plotted in Figure 41.
Although the separation time was shorter for the direct pulling-up method, the related separation
force was much larger. During the fabrication of the second cone of the hourglass, the separation
force increased linearly with the increase of the layer area and finally leads to brittle fracture at
layer 413. In comparison, the peak force of the VA-separation-based method also varied
synchronously with the area of each layer; however, the peak force was much smaller than the one
recorded in the direct pulling-up method. The comparison of the pre-defined separation time limit
and the measured separation time has verified the developed process model to estimate the
separation time meets the process planning requirement of the constrain-surface-based VPP for
large building areas.
77

Figure 41. The measured separation force (left) and the measured separation time (right) of each layer in the hourglass
test model using the direct pulling-up and VA-separation-based methods.
4.2.6 Summary of this work
A vibration-assisted separation method based on the piezo-drive design has been presented
for the constrained-surface-based VPP process. The influence of various process parameters,
including layer area size, geometric shape, and cyclic loading parameters on the vibration-assisted
separation method has been studied based on the mechanics of fatigue fracture. A stress-based
model was utilized to estimate the separation time and separation peak force. Accordingly, an
algorithm to adjust the process parameters has been developed for the VA-separation-based VPP
process. The experiments based on several designed test cases were performed to demonstrate the
effectiveness of the developed prototype system. The comparison between the direct pulling-up
separation with and without the vibration-assisted method has verified the VA assisted separation
mechanism can significantly reduce the peak separation force in the constrained-surface-based
VPP process for layers with large cross-section areas.
78

4.3 In-situ-Transfer vat photopolymerization for roof features
In this work, we present a novel VPP process called In-situ-Transfer VPP to reliably produce
microfluidic channels of 10 μm height without compromising transparency, which is not reported,
to our best knowledge. The transparent resin used in this work has a light penetration depth of
179.1 μm (Figure 42B). Evidently, our approach breaks the light penetration limit on the
minimum channel height (Figure 42Ab). The key idea is to print the after-channel layer separately
via double exposure with different print platforms, respectively. An auxiliary print platform is
utilized to print the channel roof (the top portion that encloses the channel) to blocks the delivery
of light dose into the residual liquid resin inside the channel. The auxiliary print platform also
works as a constrained surface to ensure the accuracy and surface finish of the channel. The roof
feature is then in-situ transferred to the built part with the second exposure of a mask image.
According to the derivated analytical model, we can fabricate real microfluidic channels (10-100
μm ) by controlling the roof thickness and the layer thickness of the subsequent layers. The
algorithm to generate corresponding mask images is also given. Our printing process’s efficacy
and versatility are demonstrated by fabricating multifunctional devices, including 3D microfluidic
channels, microfluidic valves, and particle sorting devices. The results show that our approach is
universal and can be applied to commercially available transparent resins. This will significantly
broaden the range of available materials and challenge the dominant position of traditional
manufacturing methods in the microfluidic fabrication domain, such as soft lithography.
4.3.1 Light dose distribution
Fabrication of microfluidic chips emphasizes specifically the microchannels inside the bulk.
The resin trapped in the channels must remain liquid or gel state till the end of fabrication. The
79

curing of the subsequent layers determines the total light dose of the channel portion. Therefore,
the polymerization state of flow channels is governed by the following energy relationship.

1
( , , )
n
t i i i t c
i
D D I t z z D
=
= = 

(22)
Where
t
D is the accumulated light dose delivered to the channel top
t
z (layer height starting
from the build platform, as shown in Figure 42C) when curing the subsequent n layers.
c
D is the
critical dose the photocurable resin needs to solidify; ( , , )
i i i
D I t z represents the light dose
distribution along the building direction when curing layer i with exposure time
i
t given the light
intensity
i
I . Based on Beer-Lambert’s law, the light energy falls off exponentially from the
penetration surface. That is
( )/
( , , )
i t p
zz
i i i t i i
D I t z z t I e
 −−
== , where
p
 denotes light penetration
depth characterizing the material optical property. According to Jacob’s working curve
ln
d
dp
c
t
c
T


=


[152],
p
 for any photocurable resin can be experimentally measured by fitting the
paired values - curing depth
d
c and the corresponding exposure time
d
t (Figure 42B).
c
T denotes
the critical exposure time, which is only related to the given resin and light intensity. From the
fitted curve in Figure 42B, the light penetration depth is 179.1 μm . The slope of the fitted curve
indicates the sensitivity of the curing depth to the exposure time. The high sensitivity at smaller
curing depths increases the difficulty in the curing depth control. The error bar also indicates it is
difficult to control the channel roof’s thickness using only energy-related methods without a
constrained surface on the top.
As a contrast and to demonstrate the high geometric modeling accuracy of our IsT-VPP for
the assembly-free fabrication of 3D microfluidic devices that feature complex flow patterns, a
80

USC-shaped fluid router is fabricated using our proposed method and presented first (Figure 42D).
Figure 42Da gives the oblique view of the CAD model. Each letter is basically an individual
microchannel (
2
1400 30 μm  ) crisscrossing other channels in two different layers. Figure 42Db
shows the top view of the 3D-printed result. The integrated fluid ports were designed as 1/16 in.
barbed connectors. Opaque pigments (red, yellow, and blue) flow through the channels for easy
visualization. A partial cross-section of the channel indicated by the dashed line in Figure 42Db
is presented in Figure 42Dc using scanning electron microscopy (SEM). Both channel heights are
30 μm . And the inner channel surface is smooth and flat. There is a 1 mm spacing between the
two channels (the segments of the letter “S” and “C”) in the z-direction. After fabrication of the
first channel (the segment of the letter “S”), the subsequent 9 layers’ exposure tends to cure more
resin inside the channel and lead to obstructed channels. Besides, it is particularly challenging to
precisely control the total light dose starting from the channel top by accurately controlling each
layer’s curing depth. However, this scenario didn’t happen in our case. This benefits from utilizing
an additional build platform (Figure 43) to cure a 200- μm -height roof for the layer adjacent to
the channel. The auxiliary build platform blocks further transmission of light, thus lowers total
energy absorbed by the resin inside the channel. The corresponding normalized light dose
distribution of each projection and the accumulated light dose along the z-direction when printing
the portion in Figure 42Dc are shown in Figure 42Dd. The curing depth
d
c is set to slightly higher
than the layer thickness l ( 1.25
d
cl = ) to ensure good bonding between layers, meanwhile surface
quality. The origin 0 z = represents the channel bottom and 30 z = specifies the channel top in
Figure 42Dc, respectively. If the accumulated light dose in the channel portion ( 0 30 zm  = )
exceeds the threshold
c
D , that is, the normalized light dose /
tc
DD is greater than 1, it will lead
81

to occlusion of the channel. In contrast, our IsT-VPP process significantly reduces
t
D lower than
c
D (Figure 42Dd) and successfully fabricates the microfluidic chip. The details of IsT-VPP
bringing much lower energy penetration are illustrated in the following sections.

Figure 42. Illustration of the effect of resin optical property and light dose distribution on VPP-based microfluidic
channels fabrication. (A) Literature review on 3D-printed microfluidic channels. (a) Summarization of minimum printable
82

channel height
min
h and corresponding technical features. (b) Relationship between light penetration depth
p
 and minimum
printable channel height
min
h for VPP-based processes. (B) Measured results and fitted curve of curing depth
d
c and exposure
time
d
t for transparent resin. (C) Schematic diagram showing challenges in 3D printing transparent microfluidic channels. (D)
Crisscrossing USC-shaped fluid router fabricated by IsT-VPP. (a) Oblique view of the CAD model. (b) Top view of the fabricated
microchannels. (c) SEM image showing the cross-section of the channels indicated by the dashed line in Fig. 1Db. The result
demonstrates 30- μm -height channels. (d) Normalized light dose distribution of each projection and the accumulated light dose
along the z-direction when manufacturing the part in Fig. 1Dc via IsT-VPP.
4.3.2 Experimental setup and process design
To realize the above demonstrations of microfluidic channels, we created the In-situ-Transfer
Vat Photopolymerization Apparatus (IsT-VPPA), as shown in Figure 43A. Our setup has an
auxiliary print platform driven by two additional motorized linear stages compared with traditional
VPP. The extra print platform is coated with PDMS, while the resin vat is covered by fluorinated
ethylene propylene (FEP) film. The auxiliary platform will only participate in printing jobs when
curing the microfluidic channels’ roof to serve as a constrained surface, similar to top-down VPP
[134]. A simplified Y-junction fluidic mixer model (Figure 43C) illustrates our process (Figure
43D).
(1) The bottom of the model, a m -mm thick cube, is 3D printed first using only the main
build platform. The corresponding mask image is presented in Figure 43Da.
(2) For the h -mm high channel, the top () h  − -mm (10 30   ) thick portion is also
fabricated by the main platform (Figure 43Db). Before finishing the residual  -mm high channel,
the fabrication of the key layer containing the channel roof, next to the flow channel, will be
accomplished in advance in the following two steps.
(3) The aux platform will participate in printing the channel roof portion to enclose the
channel (Figure 43Dc). Therefore, the main platform is lifted to make room for the aux platform.
The aux platform then moves leftward and downward to form a
r
l -mm gap with the resin vat’s
83

FEP film. A grayscale mask image is projected, also shown in Figure 43Dc. The roof part’s
grayscale level is set to the maximum value of 255 to cure the roof thoroughly. The channel roof
extends  -mm ( 0 0.1   ) on both sides to be a little wider than the channel width w . The aux
platform constrains the top surface of the channel roof. By taking these measures, the channel
accuracy and surface quality can be guaranteed. The rest part will be semi-cured using a lower
grayscale level (50 ≤ grayscale level ≤ 200). This will increase the roof feature’s bonding force
difference between the resin vat and aux platform, which is crucial in our process and explained
in the next step.
(4) When the aux platform moves back to the original idle position to leave space for the main
platform, the printed roof will remain stationary at the resin vat surface rather than following the
aux platform’s movement (Figure 43Dd). This requires the roof’s bonding force is more
significant with the resin vat than with the aux platform. We integrated three mechanisms to realize
this. First, the resin vat is coated with FEP film, while the aux platform is covered with PDMS. By
investigating the bonding force of different material interfaces under different exposure areas, we
found that a larger force is always needed to break the polymer-FEP film interface than the
Polymer-PDMS interface given the same contact area (Figure 43B). This ensures the split happens
between the channel roof and the aux platform rather than the resin vat when the aux platform
removes. Second, in our implementation, the roof and aux platform’s contact area is always smaller
than between roof and resin vat (Figure 43Dc). This further helps intensify the bonding force
difference as the bonding force will rise with the increase of the contact area for both coating media
(Figure 43B). Third, the adhesive bonding strength between the part and the coating media will
increase with more exposure energy. Therefore, the adhesion between the roof and the resin vat is
naturally stronger as the roof bottom always receives more energy than the top when printing the
84

roof. After the aux platform separates from the channel roof and returns to the original position,
the main platform moves down to form a  -mm gap with the channel roof (Figure 43Dd).
(5) The mask image shown in Figure 43De is projected to fully cure the layer containing the
channel roof and in-situ transfer it to the previously built part in the z-direction. The polymer-
polymer interface and polymer-glass interface’s bonding force is much stronger than the polymer-
FEP film interface (Figure 43B). Therefore, the two previously separated parts will combine as
one and attach firmly to the main platform during the release process. At this point, the target
channel is created, and its height is only determined by the mechanical motion system, which is
reliable and accurate. The resins trapped inside the channel remain unpolymerized state as they
receive no energy completely.
(6) The rest n -mm thick cube is printed by the main platform with layer thickness l mm until
the end. The channel roof will exponentially cut down the light dose delivered into the channel
when curing the rest part. The energy caused by curing the layers far from the channel is negligible.
By synthetically controlling the roof layer thickness
r
l and the subsequent layer thickness l
according to our analytical model, we can ensure that the accumulated light dose at the channel
top is no more than the critical dose
c
D as shown in the previous result (Figure 42D). Therefore,
the over-curing issue is resolved.
If the channel height h is equal to or smaller than the predefined  , step 2 of the printing
process (Figure 43Db) can be skipped. Meanwhile, in step 4 (Figure 43Dd), let h  = so that the
channel is fabricated when combining the roof and the built part.
85

Figure 43. Principle of IsT-VPP process. (A) The detailed structure of the IsT-VPP apparatus. (B) Measured separation
forces of different contact areas for different material interfaces during VPP printing. Given the same contact area, the
separation force between part and aux build platform corresponding to the polymer-PDMS interface is the smallest. The
separation force between part and vat surface represented by the polymer-FEP film interface is larger than the polymer-PDMS
interface. A glass sheet works as the main build platform because the polymer-glass interface has stronger bonding than the
previous two interfaces and weaker than the polymer-polymer interface. The bonding force between neighboring layers of the
printed part is derived from the resin’s ultimate strength. (C) A simplified Y-junction fluidic mixer model used to illustrate the
IsT-VPP process and its three orthographic views. (D) Illustration of IsT-VPP process. (a) (b) Fabrication of the bottom portion
and partial channel using only the main build platform and the corresponding projection images. (c) Fabrication of channel roof
using the aux build platform and corresponding grayscale mask image. (d) (e) In-situ transfer of the channel roof to the
previously built part with bonding connection in z-direction via the second exposure and corresponding mask image. (f)
Fabrication of the rest of the part using the main build platform.
4.3.3 Microfluidic channels with 10 μm height
Microchannel resolution is a significant issue for all 3D printing processes [133]. Large 3D-
printed internal microfluidic features have dimensions between 100 and 500 μm , making
available additional analysis capabilities but still do not allow the full range of microfluidic
applications [129]. To figure out the minimum channel height that we can print with the above
86

process, we first designed a test case containing one layer of embedded channels of 10, 20, 30, 40,
50, 60 μm height (Figure 44Aa). The channel width is 75 μm (4 pixels size of the DLP). The
SEM image shown in Figure 44Ab gives a detailed cross-sectional profile of the printed channels.
The channel height is equal to the gap size  =10 μm for the 10 μm-height channel (circled by
green). Therefore, the step shown in Figure 43Db is skipped, and the channel is fabricated when
connecting the roof with the previously built part corresponding to Figure 43Dd. For the other
channels (circled by red) with heights of more than 10 μm, partial channels are fabricated using
the main build platform (Figure 43Db) except for the last 10 μm layer. It shows that all the
channels are accurate with a good surface finish. To further verify the feasibility of the proposed
method to fabricate multi-layer microchannels, we consider the fabrication of 24 channels with the
height ranging from 10-60 μm as shown in Figure 44Ba. The SEM results (Figure 44Bb) show
our method is accurate and reliable. It turns out that we can print microchannels with only 10 μm
in height. This is the first time that 10 μm transparent channels can be 3D printed to the best of
our knowledge. We envision the manufacturing ability will help boost 3D-printed microfluidic
chips’ capabilities.

Figure 44. Fabrication of microfluidic channels through IsT-VPP. (A) Fabrication of single-layer microfluidic channel
with the height ranging from 10 to 60 m  . (a) CAD model of part with single-layer microfluidic channels embedded. (b) SEM
images of microfluidic channel cross-section. The channel circled by green has 10 m  in height. The fabrication of such a
channel is accomplished when connected with the previously built part corresponding to Fig. 2Dd. The channels circled by red
with heights of more than 10 μm are fabricated using the main build platform corresponding to Fig. 2Db except for the last 10
μm layer. (B) Fabrication of multiple layer microfluidic channels with the height ranging from 10 to 60 m  . (a) CAD model of
part with multiple layer microfluidic channels. (b) SEM image of multi-layer microfluidic channel cross-section.
87

4.3.4 Multifunctional automation components for microfluidic devices
Next, we demonstrate our process can benefit microfluidic automation components in terms
of miniaturization, material options, and function extension. Automation components such as
microfluidic valves find widespread use in controlling the liquid environment and actuation of
microfluidic devices. Current 3D-printed microfluidic valves either occupy a large space [147] or
have constraints in geometric design [149] and material transparency [132]. Using the proposed
method, we can 3D-print transparent Quake-style microfluidic valves with any gap size between
the membrane and the bottom seat and no longer need to worry about clogging issue. This
facilitates microfluidic valves design when integrated with channels of various heights and enables
new applications.
The microfluidic valve consists of a flow channel that allows the liquid to pass through and a
control channel overlaid orthogonally atop the flow channel (Figure 45A). The control channel’s
vent is used for clearing uncured resin after printing and is sealed before use. The membrane-like
suspended region ( 700 μm 800 μm 25 μm  ) between the control channel and the flow channel is
the valve switch (Figure 45B, a and b). The valve is open at rest so that blue-dyed deionized (DI)
water can traverse (Figure 45Bc). When applying sufficient pressure via the control channel, the
membrane will deflect and contact the half-pipe ramp below (Figure 45C, a and b). The maximum
vertical gap from the ramp top surface to the membrane bottom is 40 μm (Figure 45Bb). The DI
water between the ramp and membrane is displaced, and the valve is closed (Figure 45Cc). This
process can be modeled as a rectangular plate with all edges fixed and under uniformly distributed
loading. The maximum deflection can be approximated by [154][155]

4
max 3
Pb
y
Et

= (23)
88

where P is the pressure applied, b and t are the smaller side and the thickness of the
membrane, respectively, E is Young’s modulus, and  is coefficient ( 0.0138   in our case).
For Anycubic clear resin ( E is estimated to be around 164MPa ), Eq. 23 predicts the membrane
will deflect by 40 μm at a pressure of about 4.5 psi. Finite element analysis (FEA) predicts a 4.4%
higher pressure required to achieve the deflection of 40 μm . The loading stress simulations
indicate that the membrane (shown at rest in Figure 45Da) is maximally deformed at the center
point and begins to contact the ramp at ~4.7 psi (Figure 45Db) and nearly conform to the ramp’s
curved surface at ~5.6 psi (Figure 45Dc). The fabrication result is given in Figure 45E. We tested
the valve’s closing pressure by measuring the flow rate under different driving pressure applied to
the flow channel. The flow channel inlet is connected to a bottle containing blue-dyed DI water.
An electronically-controlled air pressure source pressurizes the bottle to drive the DI water into
the flow channel. The flow rate was measured by the mass of the DI water with an accurate balance.
Generally, a higher fluid driving pressure requires a correspondingly higher closing pressure to
stop the flow (Figure 45F). We found that by applying pressure 4 psi larger than the flow channel’s
driving pressure, the flow rate will approach ~ 0 μl/s but not fully stopped. Therefore, we determine
to set the closing pressure 5 psi higher than the fluid driving pressure due to our setup resolution
(e.g., 3 psi for fluid driving pressure and 8 psi for control pressure). In this case, the valve can
always ensure a successful closure. The results agree reasonably well with the FEA simulation.
In addition to microfluidic valves, the printed fluid automation device can also function as a
specimen platform to replace costly robotic pipettors or tedious manual pipetting with the accurate
control of the gap between the membrane and the ramp (Figure 45G). By placing the device on a
microscope stage, the flow channel works as a microscope slide; meanwhile, the membrane
functions as a coverslip. Particle/cell samples suspension are pressurized into the device. Once
89

inflated, the membrane will deflect and capture the particle/cell samples for observation and
measurement in real time. The gap size ( 40 μm  ) ensures that only one layer of particles/cells
( 20-27 μm ) can pass through the microscope’s field of view so that no overlap will occur. The
device can further switch samples for convenient and fast measurement if integrated with multiple
valves and sample suspension sources.

Figure 45. 3D-printed microfluidic valve and specimen platform for automation. (A) CAD model showing the
configuration of the 3D-printed microfluidic valve. The vent at the end of the control channel will be sealed via construction
glass glue after clearing uncured resin. The inset gives the detailed structure of the membrane and the ramp. (B) The open state
of the valve filled with blue-dyed DI water. (a) Schematic diagram showing the front sectional view of the open valve. (b)
Schematic diagram showing the side sectional view of the closed valve. (c) Microscope image showing the top view of the valve
in the open state. (C) The closed state of the 3D-printed microfluidic valve. (a) Schematic diagram showing the front sectional
view of the valve in the closed state. The liquid cannot pass through. (b) Schematic diagram showing the side sectional view of
the valve in the closed state. The membrane is fully conformal to the curved ramp surface under pressure. (c) Microscope image
showing the top view of the valve in the closed state. The liquid between the ramp and the membrane is displaced. (D) FEA
simulation of the valve membrane deflection at different pressures. (a) 0 psi. (b) 4.7 psi. (c) 5.6 psi. The color heat map shows the
von Mises stress on the membrane at these pressures. (E) Fabrication result of the microfluidic valve (isometric view). (F)
Closing pressure test of the fabricated valve under varying fluid driving pressures. (G) Specimen platform for particle samples
observation and measurement. The deflected membrane captures the green particles.
90

4.3.5 3D-printed microparticle sorting device
Developments in cell-based clinical diagnosis, cell transplantation therapy, and stem cell
engineering have significantly increased the demand for efficient techniques to purify micrometer-
sized particles, including living cells [126]. Microfluidic systems have been recognized as simple,
effective, and precise platforms for microfiltration. However, nearly all the microfiltration devices
were fabricated utilizing standard soft lithography and replica molding techniques which are labor-
intensive. Most 3D-printed interior channels are currently in the sub-millifluidic range (0.5 –1.0
mm), limiting many analytical applications. To demonstrate our process’s practicability, we,
therefore, constructed a weir-type microfilter system.
The isometric view of the 3D-printed microfilter chip and the interior structure’s schematic
are shown in Figure 46A. The device’s fundamental structure is a microchannel with a sudden
decrease in the cross-section working as a barrier in the flow direction. The three weir-like, equally
spaced barriers form gaps with the channel roof. Only particles smaller than the gap are permitted
to pass over the obstruction. Larger particles will be captured and accommodated by the
corresponding reservoir and collected later. The whole microparticle sorting system is given in
Figure 46B. The system consists of three inlets for particle suspension (inlet 1), DI water (inlet 2
& 3), and four outlets for red particles (outlet 1), blue particles (outlet 2), green particles (outlet 3),
and DI water (outlet 4). In phase 1, a certain amount of particle mixture is introduced into the chip
via inlet 1 with outlet 4 open. In phase 2, DI water flows in and out of the chip through inlet 2 and
outlet 4, carrying particles to each corresponding reservoir. In phases 3 – 5, green, blue, and red
particles are flushed by DI water coming from inlet 2 and 3 and recovered from outlets 3, 2, and 1
in order.
91

To validate the fabricated microfluidic device’s effectiveness, we collected particles from
outlets 1-3 and examined the purity of the particles from each outlet by measuring their relative
ratios under different doses – volume processed in one cycle, as shown in Figure 46, C to E. The
purity of green particles from outlet 3 can always reach 100% at any dose, which benefits from
precise channel height control and the sized-based approach (Figure 46C, a and d, Figure 46D,
a and d, and Figure 46E, a and d). We observed a dominant portion of blue particles (75.2%) in
reservoir 2 and consequently drained from outlet 2 while with some green particles (24.8%) mixed
in, which reduces the purity (Figure 46Cc). The phenomenon is more obvious in outlet 1 (Figure
46Cb). The earlier-arrived large red particles will block the passway so that the following middle-
size blue particles cannot squeeze through the gap into the next reservoir. For the small green
particles, the gap is still big enough to pass through. So the ratio of green particles collected from
outlet 1 is much less than blue ones. The particle sorting behaviors were similar if the dose
decreased to 0.33 mL except for more red particles and blue particles recovered from outlets 1 and
2, respectively. At the amount of 0.165 mL, particles purity from all three outlets can reach above
92% simultaneously that have confirmed the device can work as designed.
The purity can be further enhanced by cascading multiple microfilters or adding external
forces such as acoustic force [156] to introduce shuffling at the barriers. Besides, the commonly
used parallelization strategies can also be applied here to multiply the throughput [157][158]. Since
the last gap size is 10 μm, particles smaller than 10 μm can also be sorted if needed. Therefore, the
presented IsT-VPP proves to be a valuable manufacturing method that is universally applicable to
general microfluidic particle/cell sorters and may facilitate the actual application of microfluidic
systems in biological studies and clinical diagnosis.
92

Figure 46. 3D-printed microparticle sorting device. (A) Illustration of the structure of the microfilter chip. (a) Fabrication
result of the microfilter chip (isometric view). (b) Schematic drawing illustrating the interior of the microfilter chip (not to scale,
side sectional view). (B) Illustration of the working principle of the microparticle sorting system (top view). (C-E) Particle
sorting behaviors with different doses. (a) Purity of the particles from each outlet. (b-d) Microscope images showing the sorted
microparticles in each microfilter reservoir
93

4.3.6 Summary of this work
This work presented an original VPP-based 3D printing methodology for truly microfluidic
channels with z-resolution as high as 10 m. An aux build platform was integrated to manufacture
channel roof, which dramatically cut down the energy penetration. Thus, the light penetration
depth constraint on minimum printable channel height was broken. More material options were
available for microfluidics with the IsT-VPP process. Material scientists were able to create new
resins or fine-tune existing ones solely from users’ needs, such as elasticity and biocompatibility
rather than printability. Separation force study for different material interfaces validated the
feasibility and reliability of the presented approach.
Multiple microfluidic devices were realized by software-generated mask images for the main
and aux build platforms, including fluid routers, microvalves, and particle sorting chips. The USC-
shaped fluid router demonstrated the ability to 3D print serpentine channels. With the aux build
platform, the transparent microvalve’s gap can reach 40 m breaking the design constraint and
facilitating functional extension, miniaturization, and integration with different channel heights. A
microfilter chip was fabricated with accurately controlled gap size, successfully realizing various
microparticles sorting with high purity. SEM images, FEA simulation as well as experimental
videos verified the effectiveness of our fabrication results. Notably, the aux build platform working
as the constrained surface also contributed to high channel surface quality, which could be good
for applications involving issues such as cell adhesion. We envision that this methodology
involving an additional build platform could open a new door for 3D-printed microfluidic devices
with further generalization.
In future work, the performance and functionality of our IsT-VPP printed microfluidic devices
will be further enhanced and expanded. For instance, (i) higher channel resolution like 1 μm could
94

be achieved by utilizing mechanical components with higher resolution; (ii) optical approaches
used in [146][150][152] could be incorporated to handle transparent resins of much higher light
penetration depth without comprising channel resolution; (iii) projectors with smaller pixel size
could be applied to achieve truly micro-resolution in all X/Y/Z directions; (iv) biocompatible
photocurable resins such as Poly(ethylene glycol) diacrylate (PEGDA)-based resins can be used
to create microfluidic devices suitable for living cells; (v) the microvalves’ size can be further
scaled down using photocurable resins with smaller Young’s modulus; and (vi) large-scale
parallelization and acoustic force could be performed for particle sorting devices for higher
efficiency and purity. In summary, the presented IsT-VPP process has been demonstrated as a
highly effective method in controlling channel height for 3D-printed microfluidic channels with
the best resolution.
4.4 Direct droplet writing of highly viscous materials
New methods of depositing materials will advance various applications in medical science,
industry, and daily life, such as prosthesis [159][160], electronics [161], foods [162], robotics
[163]-[165], and batteries [166][167]. To date, material jetting is one of the most popular
technologies capable of generating and depositing controlled droplets. Figure 47a illustrates the
principle of the material jetting process. Rather than a continuous flow of liquid material, a
sequence of discrete droplets on demand is ejected out of small nozzles by micro piezo (Figure
47a vs. Figure 47c). The ejected droplets, after separating from the nozzles, land in places that
can be precisely controlled. The deposited materials usually become solid by a solvent’s
evaporation or photocured by an additional ultraviolet lamp. A unique feature of this DOD
approach is its capability of controlling an object’s property voxel by voxel. Such ability to create
“digital materials” has been demonstrated by manufacturing objects with varying colors and
95

transparency [168][90], metamaterials [169], and functionally graded materials [170][89][91].
Despite the superiority in droplet control, the material jetting process strongly suffers from low-
viscosity restriction, which dramatically limits the material options. For example, only the
materials with a viscosity lower than 40 mPa· s can be successfully ejected by the commercial
inkjet printheads [171]. Besides, the contactless jetting mode with a certain gap between the jetting
nozzles and the build platform also brings troubles. Satellite drops between primary droplets and
splashing during droplet impact often occur without optimizing material surface tension and
rheological properties and fine-tuning process parameters, including pulse waveform and stand-
off distance [172].
(a) Material jetting (b) Direct droplet writing (c) Direct ink writing

Figure 47. Comparison of our approach (direct droplet writing) with material jetting and direct ink writing. (a) A
schematic to illustrate the principle of material jetting; (b) a schematic to illustrate the principle of direct droplet writing; and (c)
a schematic to illustrate the principle of direct ink writing.
On the other side, the syringe-based DIW method is a popular 3D printing method that enables
materials over a wide range of viscosity (from 1 to over 1,000,000 mPa· s) [172]. It can fabricate
patterns using various materials, including high viscosity such as highly concentrated colloidal
suspensions of ceramic powders. The filament of paste-like material is extruded from a small
nozzle continuously while the nozzle is moved across a build platform (See Figure 47c). The
96

continuous material filament keeps contact with the build platform or the previously built layers.
Hence the process is more robust than contactless material jetting. Multiple nozzles can be
involved in the printing process to achieve multi-material 3D printing capability. However, this
continuous material deposition approach loses the ability to define an object’s property voxel-by-
voxel [88]. It also restricts shape complexity and limits its use in many applications such as lattice
structures [173] and full-color parts [90]. Besides, highly viscous solutions used in DIW require
high pressures that may damage the dispensing tip [172].
Overall, the DOD method, such as material jetting, is more promising in terms of the digital
droplet control at each voxel. In contrast, the DIW method has advantages in handling viscous
materials and robustness. To inherit both sides’ merits, we developed a novel contact-based and
digital droplet-on-demand approach for highly viscous materials in this paper. The newly
developed process is called direct droplet writing (DDW). The schematic of our method is shown
in Figure 47b. Inspired by the metal sheet shearing process, such as punching and blanking, we
utilize a top-sealed capillary tube to punch a tiny amount of ink from a small nozzle onto the build
platform. Then the ink is deposited on the build platform or the previously built layers by capillary
bridge splitting. Only one single droplet will be precisely deposited in each deposition cycle with
no satellites nor droplet slashing. At the same time, the mechanical punching force is big enough
to work with highly viscous materials (over 190,000 mPa· s, See Figure 57A).
In this work, we present a low-cost DDW process to realize the droplet-on-command 3D
printing using highly viscous materials (up to 190,000 mPa· s, Figure 57A). It also has no adverse
effects from satellite droplets, droplet splashing, and deflection. In the DDW process, only one
single droplet will be precisely deposited in each dispensing cycle at the point where the tool
touches the substrate. Also, the DDW process is robust and reliable since the printing parameters,
97

and the change of material viscosity will not affect material printability. We demonstrate the
droplet size can be varied by adjusting printing parameters, including the punching speed and
dispensing gap. This digital droplet-on-command method allows printing 3D complex structures
on various substrates with different shapes and textures. Finally, since this work focuses on
addressing the DOD’s dilemma regarding material viscosity, satellite drops, and splashing issues,
we present a set of test cases using highly viscous photocurable tricalcium phosphate (TCP) ink
and self-solidifiable polyurethane (PU) leather ink to demonstrate the DDW method’s efficacy and
versatility. The printing results show that the DDW process will drastically broaden the range of
3D printable materials and significantly advance the droplet-based 3D printing method in
fabricating functional structures in the future.
4.4.1 Direct droplet writing process
4.4.1.1 Droplet punching
To print materials with high viscosity, shear stress that is orders of magnitude larger than that
of current methods, such as electric field, acoustic force, and valve pressure, is required. Inspired
by the metal sheet shearing process, such as punching and blanking, we directly use mechanical
force in the shearing direction in the DDW process so a large force can be directly applied to a
small portion of highly viscous liquids to make it more flowable. Due to the smaller area for a
large punching force and the well-known shear-thinning phenomenon (See Figure 57Ag), our
method based on droplet punching can handle the printing materials with a viscosity at 190,000
mPa· s (See Figure 57A).
Each cycle of our dispensing approach involves two main stages. In the first stage, as shown
in Figure 49A, a top-sealed capillary tube is driven by a mechanical force to split a tiny amount
of ink from the ink stored in the cartridge. The capillary tube working as a hollow “puncher” moves
98

downward at a constant speed v and passes through the opening nozzle (Figure 49Ab). Therefore,
the capillary tube and the nozzle hole are clearance fit. During this process, a small fraction of ink
rushes in the capillary tube and is captured by the tube channel because of surface tension and the
pressure rise induced by the tube movement, while most of the ink blocking the tube is pushed
away (See Figure 49Ab). The following force balance governs the volume of the trapped ink
inside the tube.

ct t a t
F F F G + = + (24)
Where the force resulting from a capillary rise cos
ct t ct
Fd    = , where
t
d is the inner
diameter (ID) of the tube;
ct
 is the contact angle between ink and tube surface; and  is the ink
surface tension. The pressure force
22
0.25 ( 0.5 )
t t t p
F d gh v    =+ , including static pressure
t
gh 
and dynamic pressure
2
0.5
p
v  applied on the trapped ink surface
2
0.25
t
d  , where  is ink density;
t
h is tube length immersed in ink; and
p
v is punching velocity.
t
F monotonically increases with
increased tube depth
t
h and punching velocity
p
v . The two forces (
ct
F and
t
F ) are balanced by
compressed air pressure force
a
F and corresponding gravitational force.
2 00
0
4( )
at
t
PV
Fd
VV
 =
−
, where
0
P is the standard atmospheric pressure;
0
V is the internal volume of the capillary tube; and
t
V is
the volume of ink trapped inside the capillary tube.
tt
G V g  = , where g is the gravitational
acceleration.
99

Figure 48. Mechanism of direct droplet writing. (A) Schematic diagram of the first stage of DDW– droplet punching. (a) A tiny
amount of ink captured by the capillary tube moving downward; and (b) the punched-out ink droplet hanging on the capillary tube tip. (B)
Schematic diagram of the second stage of DDW– capillary splitting. (a) Liquid bridge formation; and (b) splitting of liquid bridge. (C)
Photographs showing the second stage of DDW. (a) (b) The capillary tube with extruded ink droplet continues moving downward and
approaches the build platform; (c) (d) a capillary bridge forms between the build platform and the tube end once the pendant ink contacts with
the build platform; and (e) (f) the capillary bridge undergoes thinning and breakage, leaving a portion of ink on the build platform during the
retraction of the capillary tube.
Though the ink cartridge is not sealed, the ink will not leak from the nozzle opening during
the capillary tube’s upward movement due to the liquid’s surface tension. The volume trapped
inside the nozzle orifice is:

2
=0.25
n n n
V D h  (25)
The extruded ink droplet volume is the sum of two parts – the ink trapped inside the capillary
tube
t
V characterized by Eq. 24 and the ink residing in the nozzle opening
2
=0.25
n t n
V D h 


(around 125.6 nL).

1 tn
V V V

=+ (26)
The extruded droplet volume
1
V in the stage of droplet punching will increase as punch
velocity v increases, as shown in Figure 51a and Figure 51b. The ejected ink droplet attaches to
the bottom of the capillary tube since the maximal capillary force
d
F is much larger than the
gravitational force
d
G (See Figure 48Ab).
100

4.4.1.2 Capillary splitting
In the second stage of the DDW process, the capillary tube continues moving downward and
approaches the build platform (See Figure 48Ca and Figure 48Cb). When the pendant ink touches
the build platform, a liquid bridge will form between the build platform and the tube end, as shown
in Figure 48Cc. When the tube arrives at the lowest position (
g
h higher than the build platform),
the contact area between the droplet and the build platform reaches its maximum value
2
0.25
d
D 
(See Figure 48Cd and Figure 48Ba). When the capillary tube retracts from the build platform,
the capillary bridge will undergo thinning and finally breakage, leaving a portion of ink on the
build platform (See Figure 48Ce and Figure 48Cf and Figure 48Bb). This stage is a liquid
transfer process in the dynamic regime. In this case, the droplet contact diameter with both the
donor surface (capillary tube in our case) and the receiver surface (i.e., the build platform or the
previously built layers) will be pinned during almost the entire liquid transfer process [176].
We found that the dispensing gap
g
h , i.e., the distance between the lowest tube bottom
position and the substrate, will affect the deposited droplet diameter
d
D and further affect the
splitting ratio r , given the wettability of both contact surfaces (See Figure 53). Therefore, the
deposited ink volume on the substrate in the second stage
2
V can be expressed as

21
()
g
V V r h = (27)
Accordingly, we can use a single nozzle to dispense viscous liquid into droplets with varying
droplet sizes based on these mechanisms.
101

4.4.2 Direct droplet writing experimental setup design and analysis
4.4.2.1 Experimental setup design
To realize the accurate dispensing of viscous material into droplets at right positions, we
designed a printhead based on the DDW process and build a prototype 3D printing system. We
designed the printhead as an assembly of several separate parts to demonstrate the main functions
required by DDW. The detailed structure is shown in Figure 49. Various variations based on the
same DDW principle can be designed for better performance and improved manufacturability in
the future.

Figure 49. DDW Printhead design.
The capillary tube that we selected is a 27 Gauge stainless steel blunt dispensing needle with
a male Luer lock connector (ID = 200 μm , outside diameter (OD) = 400 μm ) bought from
McMaster-Carr. The needle length is 0.5 in (12.7 mm). The dispensing needle is connected to a
pin with a female Luer lock connector fabricated by an SLA 3D Printer (Anycubic Photon). A
cam-spring mechanism drives the combination of the pin and the dispensing needle. The cam is
3D printed via a FFF 3D printer (3DWOX1, Sindoh Corp.) and coated by XTC-3D™ (Smooth-
On, Inc.) to increase its surface smoothness, impact, and abrasion resistance. 3-IN-ONE multi-
102

purpose oil (bought from True Value) is also applied to lubricate the pin top and the cam surface.
A DC motor integrated with a hall sensor controls the cam’s rotation. The maximum speed of the
cam is 1,000 RPM. The resolution of the sensor is 640 counts per round. The pin keeps making
sinusoidal reciprocating motion vertically under the linear bearing restriction. The linear bearing
is retained in the ink cartridge (fabricated by FFF) by a light press fit and a clamp ring structure.
The nozzle contains a 45-degree slope followed by a round corner and a ~1 mm long channel (See
Figure 48A(a), 1 mm
n
h = ). The nozzle orifice’s inner diameter is around 500 μm , which is in
clearance fit with the dispensing needle. The cartridge is not sealed in our design, so the volume
change inside the cartridge chamber caused by the pin movement will not affect the air pressure
and the punching process.
Ink is held in the cartridge and will not leak from the bottom nozzle opening due to its surface
tension. A self-built syringe (10 mL, ID = 14.5 mm) pump driven by a stepper motor is utilized to
refill ink from the side inlet, which has a hole of 2 mm in diameter. The motor speed is set to 0.01
mm/s, so the torque required to push viscous material is as small as possible. The ink surface level
in our setup is maintained at ~20 mm.
During the actual operation, the ink refilling is conducted recursively after depositing 500
droplets. We calculated the refilling volume by measuring the mass of 500 droplets and dividing
it by its density. Take the PU leather ink shown in Figure 57 as an example. The ink density is
1.05 g/mL. The mass of 500 PU droplets is measured as
0.001
0.001
0.010
+
−
g. The measurement is repeated
five times. Given the cartridge inner diameter (16 mm), we can estimate the maximum variance of
the ink surface level in our case is
33
2
0.01 10 10
47 μm ( )
1.05 16 0.25 

=
  
, which is only 0.235% of the original
ink height (20 mm). With the cartridge connected to the open-air, combining the 2-mm wide inlet
103

and 0.01 mm/s refilling motor speed, the pressure caused by ink refilling will not affect the printing
process. The experimental study also verifies good repeatability of droplet size.
The printhead is mounted on a 3-axis motion system. The printhead moves horizontally above
the build platform, depositing droplets of material onto the surface where material deposition is
required. After dispensing materials for one layer, the whole layer can be cured by raising
temperature if the material is thermally cured or by UV light if the material is photocurable. Then
the build platform moves downward by the layer thickness vertically after each new layer is
fabricated. An absolute indicator (iGaging, 35-700-25, San Clemente, CA) was applied to ensure
the printhead was parallel to the build platform. The resolution of the indicator was 1 µ m. The
height variances between multiple printheads were compensated by moving the build platform in
the Z axis based on their calibrated heights. The resolution of the motion system used in our
prototype was 2 µ m.
4.4.2.2 Self-alignment mechanism of the dispensing needle
The designed printhead for the lab demonstration was assembled from several off-the-shelf
components using self-fabricated fixing components. Even with limited manufacturing and
mounting accuracy, the short and thin stainless-steel dispensing needle can pass through the small
nozzle with the help of the linear bearing and the internal slop of the nozzle. The self-alignment
mechanism between the dispensing needle and the nozzle is analyzed as follows.
Suppose the pin moves down at a maximum speed of 1000 RPM (16.7 Hz) (See Figure 50a).
The stroke s is 5 mm. The full punching speed of the needle is 0.167 m/s
p
v = . Suppose our
manufacturing and assembling tolerance’s misalignment (i.e., the distance  between the needle
and the nozzle orifice) is larger than 100 μm (Figure 50b). Hence the dispensing needle will
104

contact the 45-degree nozzle slope. The mechanical properties and geometric parameters of the
dispensing needle used in our prototype system are listed in Table 13 and Table 14.
When impacting the nozzle slope, the dispensing needle will gain a horizontal speed
hp
vv =
and the max deflection
max
 . The impact force F applied to the dispensing tip can be calculated
by Equation (5).

2 3 2 3
2
max
1
6 2 3
h
F l F l
mv F
EI EI
+ =  = (28)
Table 13. Mechanical properties of 304 stainless steel dispensing needle.
Mechanical properties Value Unit
Young’s modulus E
11
1.9 10  Pa
Moment of inertia I
15
1.1775 10
−

4
m
Section modulus W
12
5.8875 10
−

3
m
Shear strength [] 
8
5.17 10  Pa
Yield tensile strength [
Y
 ]
8
2.15 10  Pa

Table 14. Profiles of 304 stainless steel dispensing needle.
Profile properties Value Unit
Needle length, l
3
12.7 10
−
 m
Inner diameter,
t
d
3
0.2 10
−
 m
Outer diameter,
t
D
3
0.4 10
−
 m
Cross-section area, A
8
9.42 10
−

2
m
Mass, m
3
0.02 10
−
 Kg

105

Figure 50. Illustration of the self-alignment mechanism. (a) The dispensing needle moves down at full speed
p
v ; (b) the
downward dispensing needle impacts with the nozzle slop; and (c) the dispensing needle is led into the nozzle opening hole
through self-alignment.
Therefore, the impact force 0.014N F = . The impact force is far less than the buckling force
limit, which is

2
2
3.42N
(2 )
cr
EI
F
l

== (29)
The maximum normal stress caused by the impact is

7
max
3.02 10 Pa [ ]
Y
Fl
W
 = =   (30)
The maximum shear stress caused by the impact is

22
5
max 22
4
( ) 2.77 10 Pa [ ]
3
t t t t
tt
r R r R F
A r R

++
= =  
+
(31)
Both stresses are much smaller than the material’s stress limits - shear strength
8
5.17 10 Pa 
and yield strength
8
2.15 10 Pa  . The maximum deflection
max
 is tiny at 41.26 μm . Therefore,
the dispensing needle will bend itself and slip along the nozzle slop. Finally, it will enter the nozzle
106

opening hole during the downward movement (See Figure 50c). The analysis results were also
verified in our experimental results, which will be discussed in the next section.
4.4.3 Experimental study and analysis
The process parameters of the DDW process, including punching speed v , dispensing gap
g
h , and material viscosity  , may affect the printing performance, including resolution (droplet
size) and accuracy (positional error). To study their relationship, we designed test cases 1-3 for
different punching speeds, dispensing gaps, and material viscosity, respectively. Based on them,
we performed test cases 4 and 5 for 2D patterns and 3D structures, respectively, to demonstrate
2D/3D printing capability and, more importantly, how to calibrate process parameters for other
dispensing materials. The performed test cases are discussed in detail as follows.
4.4.3.1 Test case 1 – Different punching speed
We fabricated rows of ink dots on a glass slide using the DDW method to study the effect of
punching speed on droplet size and positional error. The dispensing gap was fixed, while the
punching speed, namely the motor rotation speed, was varied. The parameter settings are listed in
Table 15. The motor rotation speed was converted into dispensing needle’s average punching
speed by v
30
m
p
sv
= and listed in Table 15. There were four parameter settings in total, and each
experiment was repeated ten times. In each experiment, ten droplets were generated. The ink used
in Test case 1 was photocurable resin (SI500, EnvisionTEC) mixed with 40 wt % tricalcium
phosphate (TCP) powder whose average particle size is around 4 µ m (purchased from Sigma-
Aldrich). The viscosity of the TCP and resin slurry is shown in Table 15 and Figure 54b.
107

Table 15. Parameter settings for test cases 1-3.
Test case 1 (different
punching speed)
Test case 2 (different
dispensing gap)
Test case 3 (different
material viscosity)
Stroke 𝑠 ( mm ) 5 5 5
Dispensing gap ℎ
𝑔 ( μm ) 180 100, 180, 220 220
Motor rotation speed 𝑣 𝑚 ( RPM ) 240, 480, 720, 960 960 960
Calculated punching speed 𝑣 𝑝 ( mm/s ) 40, 80, 120, 160 160 160
TCP concentration 𝑐 (wt%) 40 40 20, 30, 35, 40
Measured viscosity 𝜇 (mPa ⋅ 𝑠 ) 58004 58004 1871, 6237, 22765, 58004

Table 16. Parameter settings for test cases 4-5.
Test case 4 (line) Test case 4 (surface) Test case 5 (multi-layer structure)
Stroke 𝑠 ( mm ) 5 5 5
Motor rotation speed 𝑣 𝑚 ( RPM ) 960 960 960
Calculated punching speed 𝑣 𝑝 ( mm/s ) 160 160 160
Dispensing gap ℎ
𝑔 ( μm ) 180 180 180
Corresponding droplet size 𝐷 𝑑 (μm) 398.8 398.8 398.8
X/Y-axis motion speed 𝑓 (mm/s) 3.8, 4.0, 4.2 4.0 4.0
Calculated droplet spacing 𝑤 𝑑 (μm) 253, 267, 280 267 267
Corresponding line width 𝑤 𝑙 (μm) 350 350
Tool path spacing 𝑤 𝑡 (mm) 0.3, 0.35, 0.4 0.3
Square length for each layer a (mm) 30, 25, 20, 15, 10, 5
Square frustum height in number of layers 𝑛 1, 2, 3, 4, 5, 6

Figure 51. Effect of punching speed
p
v on droplet size
d
D and positional error  . (a) Photographs of fabrication result
using different punching speeds; (b) plot of droplet size
d
D over the range of varying punching speeds
p
v ; (c) schematic view of
droplet deposition (top) illustrating deflection errors along the printing direction  ; and (d) plot of positional errors  of
droplets deposited with different punching speeds
p
v .
108

Figure 52. Effect of dispensing gap
g
h on droplet size
d
D and positional error  . (a) Photographs of fabrication result
using different dispensing gaps; (b) plot of droplet size
d
D over the range of varying dispensing gap
g
h explored; and (c) plot of
positional errors  of droplets deposited at various dispensing gaps
g
h .
The deposited ink dots for different punching speeds are shown in Figure 51a. The
corresponding droplet sizes are summarized in Table 17 and plotted in Figure 51b. The droplet
deposition process is illustrated in Figure 51c. From Figure 51b, it can be observed that the
increase of punching speed from 40 mm/s to 160 mm/s leads to droplets with larger diameter
sizes. This is because more ink is trapped inside the capillary tube and punched out in stage one,
as shown in Eq. 24. However, when the punching speed is faster than 80 mm/s , the droplet size
stabilizes and reaches a constant value. This is due to the compressed air pressure inside the top-
sealed capillary tube grows faster than the dynamic pressure caused by the punching movement,
and ultimately both pressures reach an equilibrium. The droplet size variation is within 5%, which
shows good repeatability of the DDW process.
Table 17. Test results of droplet size under different punching speeds.
Punching speed
p
v ( mm/s ) Droplet size
d
D ( μm )
40
14.4
12.6
327.6
+
−

80
16.5
18.5
401.5
+
−

120
19.9
17.1
405.1
+
−

160
18.4
15.6
410.6
+
−

109

Positional accuracy is crucial in printing delicate structures. The positional error  is
characterized by the deflection distance perpendicular to the printing direction (See Figure 51d,
the average positional error is almost zero. Unlike other droplet dispensing approaches, the
positional accuracy is not sensitive to dispensing speed in our method. The high positional
accuracy is attributed to the contact-based process and the small dispensing gap.
4.4.3.2 Test case 2 – Different dispensing gap
Following the same procedure described in Test case 1, rows of ink dots were fabricated on
the glass slide to study the effect of dispensing gap size on droplet size and positional accuracy.
The motor speed was fixed at 960 RPM, considering printing efficiency. The corresponding
punching speed was 160 mm/s . The dispensing gap was varied. The ink for Test case 2 is the same
photocurable slurry that was used in Test case 1. The parameter settings are listed in Table 3. If
the dispensing gap reaches above 0.24 mm, the punched-out droplet cannot touch the build
platform. No ink droplet will be deposited on the build platform, and all the punched-out inks will
be returned to the cartridge. Therefore, we set the maximum dispensing gap at 0.22 mm in the test.
The test results regarding the dispensing gap effect on the droplet size and positional error are
listed in Table 18 and shown with a diagram in Figure 52.
Table 18. Test results of droplet size at various dispensing gaps.
Dispensing gap
g
h ( μm ) Droplet size
d
D ( μm )
100
23
15
490.0
+
−

180
19.1
19.8
398.8
+
−

220
12.1
6.9
273.9
+
−

As shown in Figure 52a and Figure 52b, enlarging the dispensing gap
g
h will lead to a
smaller droplet diameter
d
D . This is because, according to [176], the transfer ratio in the dynamic
liquid transfer process is dominated by the droplet contact diameter
c
D on both surfaces (See
110

Figure 53a). Besides, the contact diameter
c
D will be pinned during the entire liquid transfer
process. By controlling the dispensing gap
g
h , we can achieve different contact lines between the
droplet and the substrate and further change the droplet diameter
d
D since they are positively
correlated, as shown in Figure 53a.

Figure 53. (a) Schematic diagram on the influence of the dispensing gap and the contact line with a substrate; and (b)
schematic diagram of droplet splitting.
The capillary pressure inside the deposited droplet and the residual droplet at the receding end
are equal (See Figure 53b). Taking the Laplace Equation into account [177], the relationship is
represented as

22
1
21
( / 2)
d
h
D h R
=
+
(32)
In this way, the transfer ratio ()
g
rh can be varied to get different droplet volumes, as shown
in Eq. 27.
The maximum droplet size variation is about 5%, which shows our approach’s good printing
repeatability. For the positional error, similar behavior as Test case 1 can be observed from Figure
52c. The average positional error is almost zero. Unlike the other droplet dispensing methods, the
111

DDW’s positional accuracy is not sensitive to the dispensing gap. The high positional accuracy is
attributed to the contact-based droplet splitting process.
4.4.3.3 Test case 3 – Different material viscosity
Material viscosity plays a significant role in determining material printability in the
aforementioned droplet-based approaches. Thus, for a given punching speed and dispensing gap,
materials with various viscosities were used for the DDW-based droplet deposition to explore the
influence of ink viscosity. The material viscosity was varied by mixing photocurable resin with
TCP powders using different mass ratios. The parameter settings and the measured viscosity of
the mixed slurry for each mass ratio are given in Table 15.
As shown in Figure 54, for the slurry with a broad range of viscosity (1,800 ~ 58,000 mPa· s),
the effect of material viscosity on the deposited droplet size is negligible given the same process
parameters. This behavior is consistent with the model presented above to describe the depositing
process because the material viscosity does not appear in the model. The minor variations of the
droplet sizes come only from the difference in ink density and surface tension. The DDW method’s
insensitivity to material viscosity brings another advantage compared with other DOD and DIW
methods, such as the valve-based and laser-induced jetting approaches. Consequently, we do not
need to adjust printing parameters frequently to make materials jettable to accommodate the
potential viscosity changes due to the changing temperature or the evaporation of the solvent in
ink. In comparison, such adjustments are required for most DOD and DIW methods.
112

Figure 54. Relationship between droplet size
d
D and material viscosity  . (a) Photographs of fabrication result using
inks of different viscosity; and (b) plot of droplet size
d
D with varying material viscosity  .
4.4.3.4 Test case 4 – Fabrication of 2D patterns
After studying the deposition of individual droplets using the DDW process, we tested the
printing of lines and surfaces to find the suitable XY-axis motion speed and the line spacing in the
tool paths to generate continuous features that are as smooth as possible. The punching speed and
the dispensing gap were fixed in the study, so the droplet size remained unchanged. The ink for
the test case was the same as Test case 1. The parameter settings for the line printing are listed in
Table 16.
Ideally, the XY linear stage’s motion speed should coordinate with the droplet punching speed
to get a continuous and smooth line, as shown in Figure 55Aa and Figure 55Ad. Otherwise, an
unconnected or rugged line will be produced when the stage motion speed is faster than the droplet
generation speed (Figure 55Ac). In contrast, a thicker line will be generated if the stage motion
speed is slower than the droplet punching speed (Figure 55Ab). An interesting phenomenon is the
fabricated line width
l
w (~350 µ m) is smaller than the droplet size
d
D (~400 µ m). This is because
the droplet is stretched by the pin’s horizontal movement during the contact period, leading to a
slightly thinner line.
113

Following the same logic, the tool path spacing in printing a surface is also essential to ensure
no internal gaps (See Figure 55Ba). The parameter settings for the surface test are listed in Table
16. Overly large spacing will result in inner holes or gaps (See Figure 55Bb and Figure 55Bc).

Figure 55. Study on 2D pattern printing. (A) Photographs of lines printed using different motion speeds. (a) Lines printed
with motion speed 4.0 mm/s; (b) microscopic image of the line printed with motion speed 3.8 mm/s; (c) microscopic image of the
line printed with motion speed 4.2 mm/s; and (d) microscopic image of the line printed with motion speed 4.0 mm/s. (B)
Photographs of surfaces printed using different tool path spacing. (a) Surface printed with tool path spacing 0.3 mm; (b) surface
printed with tool path spacing 0.35 mm; and (c) surface printed with tool path spacing 0.4 mm.
4.4.3.5 Test case 5 – Fabrication of multi-layer structure
To study the layer thickness, we printed a square frustum with six layers and measured its
height. The ink for the test case is the same as Test case 1. The corresponding parameter settings
are listed in Table 16. The CAD model shown in Figure 56a was used for layer thickness testing.
Each layer was a square with a different edge length a, as shown in Table 16.
The corresponding height was measured and plotted in Figure 56b. The layer thickness was
obtained by applying a linear fit to the height of each layer of the square frustum. Based on the
linear fitting result, 77 μm was the calibrated layer thickness, which will be used as the layer
thickness in 3D printing. Note the fitted line does not pass through the origin. This is because the
114

ink has different wettability on various substrates, i.e., when printing the first layer on the build
platform and the subsequent layers on the previously printed layers.

Figure 56. Layer thickness testing. (a) CAD model used for layer thickness testing; and (b) linear fitting of the height on
multi-layered square frustum.
4.4.4 Applications and Proof-of-concept for Multi-material Printing
To demonstrate the DDW’s 3D printing ability and versatility, we designed and fabricated
two 3D objects, as discussed in this section.
(1) Figure 57Ac and Figure 57Ad show a 3D USC Trojan logo (30 mm × 37 mm × 2.3 mm,
26 layers) printed on a piece of fabric using a self-solidifiable PU leather ink. The PU leather ink
will solidify within 3 minutes at room temperature after its deposition. The rheological property
of the ink is shown in Figure 57Ag. The rheological behaviors were tested using a dial reading
viscometer (Brookfield RVT, AMETEK. Inc.). The CAD model (See Figure 57Aa) and the tool
path of the printed 26 layers (See Figure 57Ab) were generated using Autodesk ArtCAM Premium
2018. The corresponding parameter settings are listed in Table 19.
Note that, although the build platform used in this demo was a shoe fabric, which was not flat
and had many small holes (around 50-100 μm ), our DDW process can make sure the ink soak into
the fiber and fill the gaps to generate a solid base for the following layers after printing the first
layer. Simultaneously, the DDW process will not be affected by the needle’s collision with the
115

fabric surface. In comparison, such a collision issue would be a disaster for other DIW methods.
Figure 57Ae and Figure 57Af show the details of some sharp edges of the printed part. The DDW
process enables various types of inks and can be beneficial for applications such as wearable
electronics and 3D printed shoes in the future.
(2) Finally, to validate the multi-material 3D printing ability of the DDW process, we
fabricated a multi-layer stacked Tai Chi pattern (23 mm × 2mm, 20 layers for the black portion
and 25 layers for the white portion) using two printheads filled with two kinds of materials,
respectively. As shown in Figure 57Bb, the black part was printed using PU leather ink, and the
white part was printed using photocurable TCP ink. The tool path of one layer is shown in Figure
57Ba. For the designed pattern, the tool path was generated using Autodesk ArtCAM Premium

Figure 57. Demo of 3D printing structures. (A) A USC Trojan logo printed on a piece of fabric. (a) CAD model; (b) tool
path of the whole part (26 layers); (c)(d) the fabrication result; (e)(f) microscopic images showing the details; and (g)
116

rheological property of the used PU leather ink. (B) Multi-layer stacked Tai Chi pattern printed using multiple materials. (a) The
tool path of one layer, and (b) the fabrication result.
2018. The corresponding parameter settings are listed in Table 19. We achieved the same droplet
diameter and line width for both inks by adjusting the dispensing gap. The difference in layer
heights was compensated when printing different layers.
It can be observed that the boundary surface of the Tai Chi model is curved. This curved
surface is common in DIW and other droplet-based approaches. The curvature is governed by the
capillary forces acting on the triple line defined by air, solid platform, and liquid ink. As described
by Young’s relation,
cos
LG E SG SL
    =− (33)
where 𝛾 are the interface tensions between the three phases (solid/liquid/gas), and 𝜃 𝐸 is the
equilibrium contact angle. This kind of curved surface due to partial wetting will affect printing
the following layers and ultimately leading to fillet edges and expanded boundaries, as shown in
Figure 57Bb. Therefore, each component’s boundary defined in the Tai Chi pattern was offset by
0.35 mm in both directions (See Figure 57Ba, 0.7 mm boundary spacing in total) when building
the pattern to avoid material contamination. The relationship between the part height and the
expanded fillet edge should be considered when developing a tool path planning software for the
DDW process in the future. Alternatively, additional supports can be added to the layer boundary
to limit the expanded borders.
Table 19. Parameter settings for Trojan logo and Tai Chi pattern printing.
PU leather ink TCP ink
Motor rotation speed 𝑣 𝑚 ( RPM ) 960 960
Calculated punching speed 𝑣 𝑝 ( mm/s ) 160 160
Dispensing gap ℎ
𝑔 ( μm ) 160 180
Corresponding droplet size 𝐷 𝑑 ( μm ) ~400 ~400
117

X/Y-axis motion speed 𝑓 ( mm/s ) 4.0 4.0
Calculated droplet spacing 𝑤 𝑑 ( μm ) 267 267
Corresponding line width 𝑤 𝑙 ( μm ) ~350 ~350
Tool path spacing 𝑤 𝑡 ( mm ) 0.3 0.3
Layer height 𝑙 ℎ
( μm ) 95 77
Bottom layer count for USC logo 𝑛 𝑏𝑓
2
Total layer count for USC logo 𝑛 𝑈𝑆𝐶 26
Bottom layer height on glass 𝑙 ℎ𝑔 ( μm ) 151 125
Bottom layer count for Tai Chi 𝑛 𝑏𝑔
1 1
Total layer count for Tai Chi 𝑛 𝑇𝑎𝑖𝐶 ℎ𝑖 20 25
4.4.5 Summary of this work
A novel direct droplet writing process using the droplet-punching and capillary-splitting
principles have been developed for highly viscous materials. The newly developed DDW-based
3D printing method has been demonstrated using liquid materials with viscosity as high as 190,000
mPa·s. A capillary tube driven by a servo motor is used to punch viscous materials out from the
nozzle in our design. The punched-out material was then dispensed on the build platform through
the liquid-bridge splitting mechanism. In addition to overcoming the material viscosity limitation,
our contact-based DDW method also avoids common jetting issues caused by satellite droplets,
droplet splashing, and deflection. In each dispensing cycle, only one single droplet will precisely
and stably form at the position where the needle touches the substrate. We did a comprehensive
process study and found the DDW process can achieve various droplet sizes by adjusting process
parameters, including punching speed and dispensing gap. The positional accuracy will not be
affected by the printing speed and dispensing gap. Another advantage of DDW is its robustness.
The material viscosity change will not affect the material printability and printing results for the
same parameter settings. Finally, the results of several test cases have been presented to
demonstrate the DDW process’s effectiveness. Our work will drastically broaden the range of
printable materials for the droplet-based 3D printing methods and significantly advance using such
3D printing methods in fabricating multi-material objects for various functional applications.
118

In the future, we would like to explore the tool path planning for the DDW process, including
the curved boundary effect. We will also reduce the capillary tube size to investigate how to reduce
the droplet size further. Besides the printing resolution, we will also increase the printing speed by
using a DC servo motor with a higher rotational speed or involving reusable support for overhang
features [135], or integrating DDW with the projection-based Stereolithography process [134].
Finally, we will investigate using a multi-axis DDW process to print materials on any curved
substrate. Therefore, the DDW method could provide a new direction in developing multi-material
and multi-directional 3D printing methods in the future.

119

Conclusion and Recommendation for Future Research
5.1 Answering the Research Questions/Testing Hypotheses
As stated in Section 3, we have the following four research questions to answer in this
dissertation to address current limitations in VPP and push it forward:

To answer the above-proposed questions, we have the following hypotheses:

Q1: How can we easily reduce the 3D printed support to reduce material waste and printing
time?
Q2: How can we reduce the separation force in the constrained-surface-based VPP process
for large-scale 3D printing?
Q3: How to print parts with roof feature via VPP without internal support?
Q4: How can we achieve multi-material VPP efficiently without material contamination?
Hypothesis 1.1: Low-cost and reusable metal pins can be used as external support for
overhang features of 3D objects.
Hypothesis 1.2: The reusable pin-like supports can move up automatically.
120

Hypothesis 1.1 and Hypothesis 1.2 are tested in Section 4.1, in which we present a novel
type of automatic reusable support suitable for different 3D printing processes such as FFF and
VPP. Consequently, material consumption and printing time are dramatically reduced. Another
advantage of our method is the enhancement of reliability. Long 3D-printed support may collapse
during printing. In our approach, the pillar is truncated and replaced by our metal support. Our
metal support can work with any given objects and any 3D-printed support structure as long as
external supports are needed. Hence, all the previous support techniques can work with our method
Hypothesis 2.1: Less force but multiple attempts may lead to a more efficient result.
Hypothesis 2.2: The reduction of the separation force for a constrained-surface-based VPP
process can be achieved by introducing a small force to induce an initial crack in the interface
between the newly cured layer and the constrained surface.
Hypothesis 2.3: The initial crack in the interface between the newly cured layer and the
constrained surface can be quickly propagated to the entire interface.
Hypothesis 3.1: An in-situ-transfer method can be utilized in VPP to print roof feature so that
internal support is no longer required.

Hypothesis 4.1: Multi-material VPP can be achieved via incorporating a drop-on-demand
method so that there is no need to switch between resin vats as well as cleaning and other
post-processing operations.
Hypothesis 4.2: A drop-on-demand method can be realized by relying on mechanical force
so that highly viscous materials can be deposited.

Hypothesis 4.1: Besides photocurable polymer resin, there are many non-photocurable
materials available in the market but cannot be 3D printed right now, such as mold making
materials from Smooth-On Inc. These materials span over a wide range of viscosity, and many
of them are highly viscous. With a novel drop-on-demand approach suitable for viscous
material and specifically designed toolpath, the materials can be 3D printable.
121

compatibly. The metal supports can be easily removed after the printing process. The residual 3D-
printed supports are automatically generated by self-developed software. The software will
optimize the layout of a part on the printing tray to make the utmost of our setup. We believe more
reusable support techniques inspired by our work will be developed in the future.
Hypothesis 2.1, Hypothesis 2.2, and Hypothesis 2.3 are tested in Section 4.2. We present
a new VA separation hardware design for the bottom-up-based VPP process and identify an
analytical model based on fatigue fracture mechanics. Accordingly, the influence of various factors,
including layer area, layer shape, and cyclic loading parameters on the VA separation process, will
be evaluated. The obtained knowledge from the study will help the development of the VA
separation method for the constrained-surface-based VPP process.
Hypothesis 3.1 is tested in Section 4.3. We present a novel VPP process called In-situ-
Transfer VPP to reliably produce microfluidic channels of 10 μm height without compromising
transparency, which is not reported, to our best knowledge. The transparent resin used in this work
has a light penetration depth of 179.1 μm . Our approach breaks the light penetration limit on the
minimum channel height. The key idea is to print the after-channel layer separately via double
exposure with different print platforms, respectively. An auxiliary print platform is utilized to print
the channel roof (the top portion that encloses the channel) to blocks the delivery of light dose into
the residual liquid resin inside the channel. The auxiliary print platform also works as a constrained
surface to ensure the accuracy and surface finish of the channel. The roof feature is then in-situ
transferred to the built part with the second exposure of a mask image. According to the derivated
analytical model, we can fabricate real microfluidic channels (10-100 μm ) by controlling the roof
thickness and the layer thickness of the subsequent layers. The algorithm to generate
122

corresponding mask images is also given. Our printing process’s efficacy and versatility are
demonstrated by fabricating multifunctional devices, including 3D microfluidic channels,
microfluidic valves, and particle sorting devices. The results show that our approach is universal
and can be applied to commercially available transparent resins. This will significantly broaden
the range of available materials and challenge the dominant position of traditional manufacturing
methods in the microfluidic fabrication domain, such as soft lithography.
Hypothesis 4.1 and Hypothesis 4.2 are tested in Section 4.4. We present a novel low-cost
DDW process to realize the droplet-on-command 3D printing of highly viscous materials (up to
190,000 mPa· s). It also has no adverse effects from satellite droplets, droplet splashing, and
deflection. In the DDW process, only one single droplet will be precisely deposited in each
dispensing cycle at the point where the tool touches the substrate. Also, the DDW process is robust
and reliable since the printing parameters, and material viscosity change will not affect material
printability. We demonstrate the droplet size can be varied by adjusting printing parameters,
including the punching speed and dispensing gap. This digital droplet-on-command method allows
printing 3D complex structures on various substrates with different shapes and textures. The
printing results show that the DDW process will drastically broaden the range of 3D printable
materials and significantly advance the droplet-based 3D printing method in fabricating functional
structures in the future.
5.2 Engineering Achievements and Scientific Contributions
5.2.1 Engineering Achievements
The proposed hybrid VPP processes can address the limitations of the current VPP process,
including novel DOD for highly viscous materials, reusable support for overhang features, an in-
situ-transfer method for roof feature, and a vibration-assisted approach to reduce the separation
123

force in the constrained-surface-based VPP process. The engineering achievements of this
dissertation are listed as follow:
1. To reduce the external support for the overhang feature, we present a new type of
automatic reusable support suitable for different 3D printing processes such as FFF and
SL. The reusable support strategy in the developed process shows an average of ~40%
saving on the printing time and material with increased reliability and robustness,
compared with all current support designs;
2. A vibration-assisted separation method based on the piezo-drive design has been
presented for the constrained-surface-based VPP process. The experiments based on
several designed test cases were performed to demonstrate the effectiveness of the
developed prototype system. The comparison between the direct pulling-up separation
with and without the vibration-assisted method has verified the VA assisted separation
mechanism can significantly reduce the peak separation force in the constrained-surface-
based VPP process for layers with large cross-sectional areas;
3. An original in-situ-transfer method was integrated into the VPP process to avoid the
internal support required for roof feature. With this innovation, the z-accuracy of roof
features can be reliably guaranteed. Microfluidic channels fabricated using the developed
approach can achieve 10 μm in height, which is the highest z-resolution in current 3D-
printed microfluidic devices. We envision that this methodology involving an additional
build platform could open a new door for 3D-printed microfluidic devices with further
generalization;
4. To address the material viscosity limit in the DOD approach, an innovative printer head
based on droplet-punching and capillary-splitting principles enable the printing materials
124

to have a viscosity up to 190,000 mPa· s, which is the highest record in current droplet-
based approaches.
5.2.2 Scientific Contributions
When seeking answers to the abovementioned four questions, new knowledge is discovered
and listed as below:
1. Modeling of the vibration-assisted separation process. The influence of various process
parameters, including layer area size, geometric shape, and cyclic loading parameters on
the vibration-assisted separation method, has been studied based on fatigue fracture
mechanics. In addition, a stress-based model was utilized to estimate the separation time
and separation peak force.
2. Modeling of the light distribution in IsT-VPP. The analytical model based on Beer-
Lambert’s law described the relationship between the accumulated light penetration
energy and process parameters, including roof thickness and the layer thickness of the
subsequent layers.
3. Modeling of Analytical models to characterize the DDW process. A set of test cases have
been conducted using the in-house developed prototype system to describe the
relationship between droplet size and process parameters such as droplet punching speed
and dispensing gap.
5.3 Future work
The remaining work can be classified into two directions:
1. Refining the proposed fabrication processes to improve the performance, including:
125

1) For the material extrusion processes such as direct ink writing (DIW), the printing
materials, especially those used in the bio-related study, could be expensive. We will
reduce the metal pin size for such processes to increase the material saving further;
2) The performance and functionality of our IsT-VPP printed microfluidic devices will
be further enhanced and expanded. For instance, (i) higher channel resolution like 1
μm could be achieved by utilizing mechanical components with higher resolution;
(ii) optical approaches could be incorporated to handle transparent resins of much
higher light penetration depth without comprising channel resolution; (iii) projectors
with smaller pixel size could be applied to achieve truly micro-resolution in all
X/Y/Z directions;
3) Reducing the capillary tube size to reduce the droplet size further. Besides the
printing resolution, we will also increase the printing speed by using a DC servo
motor with a higher rotational speed.
2. Exploring potential applications:
1) We will investigate the extension of the reusable support method to other AM
processes such as projection-based stereolithography and powder-based selective
laser melting. Such AM processes are all facing the common support generation
challenges, i.e., sufficient support needs to be added for the printing process, and, at
the same time, the added support needs to be easily removed after the printing process.
The reusable support could provide a new direction to address such a dilemma in
support generation in the future.
2) For the In-situ-Transfer process, with biocompatible photocurable resins such as
Poly(ethylene glycol) diacrylate (PEGDA)-based resins, we can create microfluidic
126

devices suitable for living cells; microvalves’ size can be further scaled down using
photocurable resins with smaller Young’s modulus; large-scale parallelization and
acoustic force could be performed for particle sorting devices for higher efficiency
and purity.
3) The proposed DDW 3D printing process can fabricate more complex composite
materials, such as wearable electronics, soft robotics, and biomedical devices. The
advance of the proposed method is to selectively insert various materials, even the
materials with high viscosity, in the photocurable matrix material layer by layer to
form a multi-functional object. The potential functional materials can be liquid metal,
magnetized particles, bio-inks, etc.

127

Bibliography
[1] Ligon, Samuel Clark, Robert Liska, Jurgen Stampfl, Matthias Gurr, and Rolf Mü lhaupt.
"Polymers for 3D printing and customized additive manufacturing." Chemical reviews 117,
no. 15 (2017): 10212-10290.
[2] Chen, Y. H., C. T. Ng, and Y. Z. Wang. "Generation of an STL file from 3D measurement
data with user-controlled data reduction." The international journal of advanced
manufacturing technology 15, no. 2 (1999): 127-131.
[3] Gibson, Ian, David Rosen, Brent Stucker, and Mahyar Khorasani. Additive manufacturing
technologies. Vol. 17. New York: Springer, 2014.
[4] Guo, Nannan, and Ming C. Leu. "Additive manufacturing: technology, applications and
research needs." Frontiers of Mechanical Engineering 8, no. 3 (2013): 215-243.
[5] Gao, Wei, Yunbo Zhang, Devarajan Ramanujan, Karthik Ramani, Yong Chen, Christopher
B. Williams, Charlie CL Wang, Yung C. Shin, Song Zhang, and Pablo D. Zavattieri. "The
status, challenges, and future of additive manufacturing in engineering." Computer-Aided
Design 69 (2015): 65-89.
[6] C. W. Hull, "Apparatus for production of three-dimensional objects by stereolithography.".
U.S. Patent 4,575,330, 11 March 1986.
[7] Szilvśi-Nagy, Má rta, and G. Y. Matyasi. "Analysis of STL files." Mathematical and
computer modeling 38, no. 7-9 (2003): 945-960.

128

[8] Peterson, Gregory I., Johanna J. Schwartz, Di Zhang, Benjamin M. Weiss, Mark A. Ganter,
Duane W. Storti, and Andrew J. Boydston. "Production of materials with spatially-
controlled cross-link density via vat photopolymerization." ACS applied materials &
interfaces 8, no. 42 (2016): 29037-29043.
[9] Griffith, Michelle L., and John W. Halloran. "Freeform fabrication of ceramics via
stereolithography." Journal of the American Ceramic Society 79, no. 10 (1996): 2601-2608.
[10] Stampfl, J., S. Baudis, C. Heller, R. Liska, A. Neumeister, R. Kling, A. Ostendorf, and M.
Spitzbart. "Photopolymers with tunable mechanical properties processed by laser-based
high-resolution stereolithography." Journal of Micromechanics and Microengineering 18,
no. 12 (2008): 125014.
[11] Dhariwala, Busaina, Elaine Hunt, and Thomas Boland. "Rapid prototyping of tissue-
engineering constructs, using photopolymerizable hydrogels and
stereolithography." Tissue engineering 10, no. 9-10 (2004): 1316-1322.
[12] Weng, Zixiang, Jianlei Wang, T. Senthil, and Lixin Wu. "Mechanical and thermal
properties of ABS/montmorillonite nanocomposites for fused deposition modeling 3D
printing." Materials & Design 102 (2016): 276-283.
[13] Siqueira, Gilberto, Dimitri Kokkinis, Rafael Libanori, Michael K. Hausmann, Amelia
Sydney Gladman, Antonia Neels, Philippe Tingaut, Tanja Zimmermann, Jennifer A. Lewis,
and André R. Studart. "Cellulose nanocrystal inks for 3D printing of textured cellular
architectures." Advanced Functional Materials 27, no. 12 (2017): 1604619.
129

[14] Vaezi, Mohammad, and Shoufeng Yang. "Extrusion-based additive manufacturing of
PEEK for biomedical applications." Virtual and Physical Prototyping 10, no. 3 (2015):
123-135.
[15] Yap, Yee Ling, Chengcheng Wang, Swee Leong Sing, Vishwesh Dikshit, Wai Yee Yeong,
and Jun Wei. "Material jetting additive manufacturing: An experimental study using
designed metrological benchmarks." Precision engineering 50 (2017): 275-285.
[16] Yang, Hongyi, Jingying Charlotte Lim, Yuchan Liu, Xiaoying Qi, Yee Ling Yap,
Vishwesh Dikshit, Wai Yee Yeong, and Jun Wei. "Performance evaluation of projet multi-
material jetting 3D printer." Virtual and physical prototyping 12, no. 1 (2017): 95-103.
[17] Sugitani, Hiroshi, Hiroto Matsuda, and Masami Ikeda. "Liquid-jetting head." U.S. Patent
4,611,219, issued September 9, 1986.
[18] Meteyer, Simon, Xin Xu, Nicolas Perry, and Yaoyao Fiona Zhao. "Energy and material
flow analysis of binder-jetting additive manufacturing processes." Procedia Cirp 15
(2014): 19-25.
[19] Deckers, Jan, Jef Vleugels, and Jean-Pierre Kruth. "Additive manufacturing of ceramics:
a review." Journal of Ceramic Science and Technology 5, no. 4 (2014): 245-260.
[20] Fina, Fabrizio, Alvaro Goyanes, Simon Gaisford, and Abdul W. Basit. "Selective laser
sintering (SLS) 3D printing of medicines." International journal of pharmaceutics 529, no.
1-2 (2017): 285-293.

130

[21] Murr, Lawrence E., Edwin Martinez, Krista N. Amato, Sara M. Gaytan, Jennifer
Hernandez, Diana A. Ramirez, Patrick W. Shindo, Frank Medina, and Ryan B. Wicker.
"Fabrication of metal and alloy components by additive manufacturing: examples of 3D
materials science." Journal of Materials Research and technology 1, no. 1 (2012): 42-54.
[22] Yadroitsev, I., I. Shishkovsky, Ph Bertrand, and I. Smurov. "Manufacturing of fine-
structured 3D porous filter elements by selective laser melting." Applied Surface
Science 255, no. 10 (2009): 5523-5527.
[23] Dawood, Andrew, B. Marti Marti, V. Sauret-Jackson, and Alastair Darwood. "3D printing
in dentistry." British dental journal 219, no. 11 (2015): 521-529.
[24] Thompson, Scott M., Linkan Bian, Nima Shamsaei, and Aref Yadollahi. "An overview of
Direct Laser Deposition for additive manufacturing; Part I: Transport phenomena,
modeling and diagnostics." Additive Manufacturing 8 (2015): 36-62.
[25] Wilson, J. Michael, Cecil Piya, Yung C. Shin, Fu Zhao, and Karthik Ramani.
"Remanufacturing of turbine blades by laser direct deposition with its energy and
environmental impact analysis." Journal of Cleaner Production 80 (2014): 170-178.
[26] Shamsaei, Nima, Aref Yadollahi, Linkan Bian, and Scott M. Thompson. "An overview of
Direct Laser Deposition for additive manufacturing; Part II: Mechanical behavior, process
parameter optimization and control." Additive Manufacturing 8 (2015): 12-35.

131

[27] Vaezi, Mohammad, Srisit Chianrabutra, Brian Mellor, and Shoufeng Yang. "Multiple
material additive manufacturing–Part 1: a review: this review paper covers a decade of
research on multiple material additive manufacturing technologies which can produce
complex geometry parts with different materials." Virtual and Physical Prototyping 8, no.
1 (2013): 19-50.
[28] Gauvin, Robert, Ying-Chieh Chen, Jin Woo Lee, Pranav Soman, Pinar Zorlutuna, Jason
W. Nichol, Hojae Bae, Shaochen Chen, and Ali Khademhosseini. "Microfabrication of
complex porous tissue engineering scaffolds using 3D projection
stereolithography." Biomaterials 33, no. 15 (2012): 3824-3834.
[29] Lee, Michael P., Geoffrey JT Cooper, Trevor Hinkley, Graham M. Gibson, Miles J.
Padgett, and Leroy Cronin. "Development of a 3D printer using scanning projection
stereolithography." Scientific reports 5, no. 1 (2015): 1-5.
[30] Wang, Zongjie, Raafa Abdulla, Benjamin Parker, Roya Samanipour, Sanjoy Ghosh, and
Keekyoung Kim. "A simple and high-resolution stereolithography-based 3D bioprinting
system using visible light crosslinkable bioinks." Biofabrication 7, no. 4 (2015): 045009.
[31] Zhang, A. Ping, Xin Qu, Pranav Soman, Kolin C. Hribar, Jin W. Lee, Shaochen Chen,
and Sailing He. "Rapid fabrication of complex 3D extracellular microenvironments by
dynamic optical projection stereolithography." Advanced materials 24, no. 31 (2012):
4266-4270.

132

[32] Cooke, Malcolm N., John P. Fisher, David Dean, Clare Rimnac, and Antonios G. Mikos.
"Use of stereolithography to manufacture critical‐sized 3D biodegradable scaffolds for
bone ingrowth." Journal of Biomedical Materials Research Part B: Applied Biomaterials:
An Official Journal of The Society for Biomaterials, The Japanese Society for Biomaterials,
and The Australian Society for Biomaterials and the Korean Society for Biomaterials 64,
no. 2 (2003): 65-69.
[33] Kruth, J-P., Ming-Chuan Leu, and Terunaga Nakagawa. "Progress in additive
manufacturing and rapid prototyping." Cirp Annals 47, no. 2 (1998): 525-540.
[34] Weng, Zixiang, Yu Zhou, Wenxiong Lin, T. Senthil, and Lixin Wu. "Structure-property
relationship of nano enhanced stereolithography resin for desktop SLA 3D
printer." Composites Part A: Applied Science and Manufacturing 88 (2016): 234-242.
[35] Xing, Hongyu, Bin Zou, Shasha Li, and Xiangsong Fu. "Study on surface quality,
precision and mechanical properties of 3D printed ZrO2 ceramic components by laser
scanning stereolithography." Ceramics International 43, no. 18 (2017): 16340-16347.
[36] Raman, Ritu, Basanta Bhaduri, Mustafa Mir, Artem Shkumatov, Min Kyung Lee, Gabriel
Popescu, Hyunjoon Kong, and Rashid Bashir. "High‐resolution projection
microstereolithography for patterning of neovasculature." Advanced healthcare
materials 5, no. 5 (2016): 610-619.
[37] Limaye, A. S., and David W. Rosen. "Process planning method for mask projection
micro‐stereolithography." Rapid Prototyping Journal (2007).
133

[38] Song, Xuan, Yong Chen, Tae Woo Lee, Shanghua Wu, and Lixia Cheng. "Ceramic
fabrication using mask-image-projection-based stereolithography integrated with tape-
casting." Journal of Manufacturing Processes 20 (2015): 456-464.
[39] Zhou, Chi, and Yong Chen. "Additive manufacturing based on optimized mask video
projection for improved accuracy and resolution." Journal of manufacturing processes 14,
no. 2 (2012): 107-118.
[40] Xu, Kai, and Yong Chen. "Mask image planning for deformation control in projection-
based stereolithography process." Journal of Manufacturing Science and Engineering 137,
no. 3 (2015).
[41] Pan, Yayue, Xuejin Zhao, Chi Zhou, and Yong Chen. "Smooth surface fabrication in mask
projection based stereolithography." Journal of Manufacturing Processes 14, no. 4 (2012):
460-470.
[42] Pan, Yayue, Chi Zhou, and Yong Chen. "A fast mask projection stereolithography process
for fabricating digital models in minutes." Journal of Manufacturing Science and
Engineering 134, no. 5 (2012).
[43] Zhou, Chi, Yong Chen, Zhigang Yang, and Behrokh Khoshnevis. "Digital material
fabrication using mask‐image‐projection‐based stereolithography." Rapid Prototyping
Journal (2013).

134

[44] Pan, Yayue, Chi Zhou, and Yong Chen. "Rapid manufacturing in minutes: The
development of a mask projection stereolithography process for high-speed fabrication."
In International Manufacturing Science and Engineering Conference, vol. 54990, pp. 405-
414. American Society of Mechanical Engineers, 2012.
[45] Zhou, Chi, Hang Ye, and Feng Zhang. "A Novel Low-Cost Stereolithography Process
Based on Vector Scanning and Mask Projection for High-Accuracy, High-Speed, High-
Throughput and Large-Area Fabrication." In International Design Engineering Technical
Conferences and Computers and Information in Engineering Conference, vol. 46285, p.
V01AT02A068. American Society of Mechanical Engineers, 2014.
[46] Luo, Ren C., Jyh Hwa Tzou, and Wei Zen Lee. "The development of LCD panel display
based rapid prototyping system for advanced manufacturing." In Proceedings 2000 ICRA.
Millennium Conference. IEEE International Conference on Robotics and Automation.
Symposia Proceedings (Cat. No. 00CH37065), vol. 4, pp. 3083-3088. IEEE, 2000.
[47] Wang, Zongjie, Nicholas Martin, Delator Hini, Barry Mills, and Keekyoung Kim. "Rapid
fabrication of multilayer microfluidic devices using the liquid crystal display-based
stereolithography 3D printing system." 3D Printing and Additive Manufacturing 4, no. 3
(2017): 156-164.
[48] Luo, Ren C., and Jyh Hwa Tzou. "Development of an LCD-photomask-based desktop
manufacturing system." IEEE Transactions on Industrial Electronics 55, no. 10 (2008):
3715-3723.
135

[49] Liska, R., M. Schuster, R. Infü hr, C. Turecek, C. Fritscher, B. Seidl, V. Schmidt et al.
"Photopolymers for rapid prototyping." Journal of Coatings Technology and Research 4,
no. 4 (2007): 505-510.
[50] Andrzejewska, Ewa. "Photopolymerization kinetics of multifunctional
monomers." Progress in polymer science 26, no. 4 (2001): 605-665.
[51] Swinehart, Donald F. "The beer-lambert law." Journal of chemical education 39, no. 7
(1962): 333.
[52] Lee, Jim H., Robert K. Prud’homme, and Ilhan A. Aksay. "Cure depth in
photopolymerization: Experiments and theory." Journal of Materials Research 16, no. 12
(2001): 3536-3544.
[53] "DMD chip," Texas Instruments, [Online]. Available: http://www.ti.com/dlp-
chip/overview.html. [Accessed June 2021].
[54] Pan, Yayue, Haiyang He, Jie Xu, and Alan Feinerman. "Study of separation force in
constrained surface projection stereolithography." Rapid Prototyping Journal (2017).
[55] Li, Xiangjia, and Yong Chen. "Micro-scale feature fabrication using immersed surface
accumulation." Journal of Manufacturing Processes 28 (2017): 531-540.
[56] Song, Xuan, Zhuofeng Zhang, Zeyu Chen, and Yong Chen. "Porous structure fabrication
using a stereolithography-based sugar foaming method." Journal of Manufacturing
Science and Engineering 139, no. 3 (2017).

136

[57] Li, Xiangjia, Huachao Mao, Yayue Pan, and Yong Chen. "Mask video projection based
stereolithography with continuous resin flow to build digital models in minutes."
In International Manufacturing Science and Engineering Conference, vol. 51357, p.
V001T01A021. American Society of Mechanical Engineers, 2018.
[58] Jin, Jie, Jingfan Yang, Huachao Mao, and Yong Chen. "A vibration-assisted method to
reduce separation force for stereolithography." Journal of Manufacturing Processes 34
(2018): 793-801.
[59] Liravi, Farzad, Sonjoy Das, and Chi Zhou. "Separation force analysis based on cohesive
delamination model for bottom-up stereolithography using finite element analysis."
In Solid Freeform Fabrication Symposium, vol. 1, pp. 1432-1451. 2014.
[60] Huang, You-Min, and Cho-Pei Jiang. "On-line force monitoring of platform ascending
rapid prototyping system." Journal of materials processing technology 159, no. 2 (2005):
257-264.
[61] Ye, Hang, Abhishek Venketeswaran, Sonjoy Das, and Chi Zhou. "Investigation of
separation force for constrained-surface stereolithography process from mechanics
perspective." Rapid Prototyping Journal (2017).
[62] Dendukuri, Dhananjay, Daniel C. Pregibon, Jesse Collins, T. Alan Hatton, and Patrick S.
Doyle. "Continuous-flow lithography for high-throughput microparticle synthesis." Nature
materials 5, no. 5 (2006): 365-369.
137

[63] Tumbleston, John R., David Shirvanyants, Nikita Ermoshkin, Rima Janusziewicz, Ashley
R. Johnson, David Kelly, Kai Chen et al. "Continuous liquid interface production of 3D
objects." Science 347, no. 6228 (2015): 1349-1352.
[64] Wang, Jia-Chang, Manuel Ruilova, and Yen-Hsun Lin. "The development of an active
separation method for bottom-up stereolithography system." In 2017 IEEE/SICE
International Symposium on System Integration (SII), pp. 108-114. IEEE, 2017.
[65] Kovalenko, Iaroslav, Maryna Garan, Andrii Shynkarenko, Petr Zelený, and Jiří Šafka.
"Examining the relationship between forces during stereolithography 3D printing and
geometric parameters of the model." In MATEC Web of Conferences, vol. 40, p. 02005.
EDP Sciences, 2016.
[66] Liravi, Farzad, Sonjoy Das, and Chi Zhou. "Separation force analysis and prediction based
on cohesive element model for constrained-surface Stereolithography
processes." Computer-Aided Design 69 (2015): 134-142.
[67] Wu, Xiangquan, Qin Lian, Dichen Li, and Zhongmin Jin. "Tilting separation analysis of
bottom-up mask projection stereolithography based on cohesive zone model." Journal of
Materials Processing Technology 243 (2017): 184-196.
[68] J. L. Lombardi, D. Popovich, G. J. Artz, Water soluble rapid prototyping support and
mold material, US Patent 6,070,107, 2010.
[69] Ezair, Ben, Fady Massarwi, and Gershon Elber. "Orientation analysis of 3D objects
toward minimal support volume in 3D-printing." Computers & Graphics 51 (2015): 117-
124.
138

[70] Hong, Jun, Wei Wang, and Yiping Tang. "Part building orientation optimization method
in stereolithography." Chinese Journal of Mechanical Engineering(English Edition) 19, no.
1 (2006): 14-18.
[71] Frank, Dietmar, and Georges Fadel. "Expert system-based selection of the preferred
direction of build for rapid prototyping processes." Journal of Intelligent Manufacturing 6,
no. 5 (1995): 339-345.
[72] Alexander, Paul, Seth Allen, and Debasish Dutta. "Part orientation and build cost
determination in layered manufacturing." Computer-Aided Design 30, no. 5 (1998): 343-
356.
[73] Majhi, Jayanth, Ravi Janardan, Michiel Smid, and Prosenjit Gupta. "On some geometric
optimization problems in layered manufacturing." In Workshop on Algorithms and Data
Structures, pp. 136-149. Springer, Berlin, Heidelberg, 1997.
[74] Pandey, Pulak M., K. Thrimurthulu, and N. Venkata Reddy*. "Optimal part deposition
orientation in FDM by using a multicriteria genetic algorithm." International Journal of
Production Research 42, no. 19 (2004): 4069-4089.
[75] Paul, Ratnadeep, and Sam Anand. "Optimization of layered manufacturing process for
reducing form errors with minimal support structures." Journal of Manufacturing
Systems 36 (2015): 231-243.
[76] Das, Paramita, Ramya Chandran, Rutuja Samant, and Sam Anand. "Optimum part build
orientation in additive manufacturing for minimizing part errors and support
structures." Procedia Manufacturing 1 (2015): 343-354.
139

[77] Mirzendehdel, Amir M., and Krishnan Suresh. "Support structure constrained topology
optimization for additive manufacturing." Computer-Aided Design 81 (2016): 1-13.
[78] Yao, Miaojun, Zhili Chen, Linjie Luo, Rui Wang, and Huamin Wang. "Level-set-based
partitioning and packing optimization of a printable model." ACM Transactions on
Graphics (TOG) 34, no. 6 (2015): 1-11.
[79] Vanek, Juraj, JA Garcia Galicia, Bedrich Benes, R. Měch, N. Carr, Ondrej Stava, and G.
S. Miller. "Packmerger: A 3d print volume optimizer." In Computer Graphics Forum, vol.
33, no. 6, pp. 322-332. 2014.
[80] Hu, Ruizhen, Honghua Li, Hao Zhang, and Daniel Cohen-Or. "Approximate pyramidal
shape decomposition." ACM Trans. Graph. 33, no. 6 (2014): 213-1.
[81] Huang, Xiaomao, Chunsheng Ye, Siyu Wu, Kaibo Guo, and Jianhua Mo. "Sloping wall
structure support generation for fused deposition modeling." The International Journal of
Advanced Manufacturing Technology 42, no. 11-12 (2009): 1074.
[82] Strano, Giorgio, L. Hao, R. M. Everson, and K. E. Evans. "A new approach to the design
and optimisation of support structures in additive manufacturing." The International
Journal of Advanced Manufacturing Technology 66, no. 9-12 (2013): 1247-1254.
[83] Cacace, Simone, Emiliano Cristiani, and Leonardo Rocchi. "A level set based method for
fixing overhangs in 3D printing." Applied Mathematical Modelling 44 (2017): 446-455.
[84] Schmidt, Ryan Michael. "Generating support material for three-dimensional printing."
U.S. Patent 10,226,895, issued March 12, 2019.
140

[85] Vanek, Juraj, Jorge A. Garcia Galicia, and Bedrich Benes. "Clever support: Efficient
support structure generation for digital fabrication." In Computer graphics forum, vol. 33,
no. 5, pp. 117-125. 2014.
[86] Dumas, Jé ré mie, Jean Hergel, and Sylvain Lefebvre. "Bridging the gap: Automated steady
scaffoldings for 3d printing." ACM Transactions on Graphics (TOG) 33, no. 4 (2014): 1-
10.
[87] Barnett, Eric, and Clé ment Gosselin. "Weak support material techniques for alternative
additive manufacturing materials." Additive Manufacturing 8 (2015): 95-104.
[88] Leung, Yuen-Shan, Tsz-Ho Kwok, Huachao Mao, and Yong Chen. "Digital material
design using tensor-based error diffusion for additive manufacturing." Computer-Aided
Design 114 (2019): 224-235.
[89] Zhu, Bo, Mé lina Skouras, Desai Chen, and Wojciech Matusik. "Two-scale topology
optimization with microstructures." ACM Transactions on Graphics (TOG) 36, no. 4
(2017): 1.
[90] Morovič, Peter, Ján Morovič, Ingeborg Tastl, Melanie Gottwals, and Gary Dispoto.
"HANS3D: a multi-material, volumetric, voxel-by-voxel content processing pipeline for
color and beyond." In Color and Imaging Conference, vol. 2017, no. 25, pp. 219-225.
Society for Imaging Science and Technology, 2017.
[91] Ituarte, Inigo Flores, Narasimha Boddeti, Vahid Hassani, Martin L. Dunn, and David W.
Rosen. "Design and additive manufacture of functionally graded structures based on digital
materials." Additive Manufacturing 30 (2019): 100839.
141

[92] Doubrovski, Eugeni L., Elizabeth Yinling Tsai, Daniel Dikovsky, Jo MP Geraedts, Hugh
Herr, and Neri Oxman. "Voxel-based fabrication through material property mapping: A
design method for bitmap printing." Computer-Aided Design 60 (2015): 3-13.
[93] Kyle, Stuart, Zita M. Jessop, Ayesha Al‐Sabah, and Iain S. Whitaker. "‘Printability'of
candidate biomaterials for extrusion based 3D printing: state‐of‐the‐art." Advanced
healthcare materials 6, no. 16 (2017): 1700264.
[94] Tsai, Ho-Lin, Weng-Sing Hwang, Jhih-Kai Wang, Wen-Chih Peng, and Shin-Hau Chen.
"Fabrication of microdots using piezoelectric dispensing technique for viscous
fluids." Materials 8, no. 10 (2015): 7006-7016.
[95] Guillotin, Bertrand, Muhammad Ali, Alexandre Ducom, Sylvain Catros, Virginie
Keriquel, Agnè s Souquet, Murielle Remy, Jean-Christophe Fricain, and Fabien Guillemot.
"Laser-assisted bioprinting for tissue engineering." In Biofabrication, pp. 95-118. William
Andrew Publishing, 2013.
[96] Gruene, Martin, Claudia Unger, Lothar Koch, Andrea Deiwick, and Boris Chichkov.
"Dispensing pico to nanolitre of a natural hydrogel by laser-assisted
bioprinting." Biomedical engineering online 10, no. 1 (2011): 1-11.
[97] Brown, Matthew S., C. Frederik Brasz, Yiannis Ventikos, and Craig B. Arnold.
"Impulsively actuated jets from thin liquid films for high-resolution printing
applications." Journal of Fluid Mechanics 709 (2012): 341.
142

[98] Zhang, Zhengyi, Ruitong Xiong, Renwei Mei, Yong Huang, and Douglas B. Chrisey.
"Time-resolved imaging study of jetting dynamics during laser printing of viscoelastic
alginate solutions." Langmuir 31, no. 23 (2015): 6447-6456.
[99] Yang, H., Y. He, C. Tuck, R. Wildman, and R. Hague. "High viscosity jetting system for
3D reactive inkjet printing." In Twenty Forth Annual International Solid Freeform
Fabrication Symposium—An Additive Manufacturing Conference, pp. 505-513. 2013.
[100] Lu, Shizhou, Xi Chen, Hai Zheng, Yupei Zhao, and Yizhou Long. "Simulation and
Experiment on Droplet Volume for the Needle-Type Piezoelectric Jetting
Dispenser." Micromachines 10, no. 9 (2019): 623.
[101] Yang, Yang, Shoudong Gu, Jianfang Liu, Hongyu Tian, and Qingqing Lv. "Research
and Development of a 3D Jet Printer for High-Viscosity Molten
Liquids." Micromachines 9, no. 11 (2018): 554.
[102] Park, Jang-Ung, Matt Hardy, Seong Jun Kang, Kira Barton, Kurt Adair, Deep kishore
Mukhopadhyay, Chang Young Lee et al. "High-resolution electrohydrodynamic jet
printing." Nature materials 6, no. 10 (2007): 782-789.
[103] Foresti, Daniele, Katharina T. Kroll, Robert Amissah, Francesco Sillani, Kimberly A.
Homan, Dimos Poulikakos, and Jennifer A. Lewis. "Acoustophoretic printing." Science
advances 4, no. 8 (2018): eaat1659.

143

[104] Roach, Devin J., Craig M. Hamel, Conner K. Dunn, Marshall V. Johnson, Xiao Kuang,
and H. Jerry Qi. "The m4 3D printer: A multi-material multi-method additive
manufacturing platform for future 3D printed structures." Additive Manufacturing 29
(2019): 100819.
[105] Han, Daehoon, Chen Yang, Nicholas X. Fang, and Howon Lee. "Rapid multi-material
3D printing with projection micro-stereolithography using dynamic fluidic
control." Additive Manufacturing 27 (2019): 606-615.
[106] Dowling, Norman E. Mechanical behavior of materials: engineering methods for
deformation, fracture, and fatigue. Pearson, 2012.
[107] Pascoe, J. A., R. C. Alderliesten, and R. Benedictus. "Methods for the prediction of
fatigue delamination growth in composites and adhesive bonds–a critical
review." Engineering Fracture Mechanics 112 (2013): 72-96.
[108] Duggan, Terance V., and James Byrne. Fatigue as a design criterion. Macmillan
International Higher Education, 1977.
[109] Schwalbe, Karl-Heinz. "Comparison of several fatigue crack propagation laws with
experimental results." Engineering Fracture Mechanics 6, no. 2 (1974): 325-341.
[110] Walker, K. "The effect of stress ratio during crack propagation and fatigue for 2024-T3
and 7075-T6 aluminum." In Effects of environment and complex load history on fatigue
life. ASTM International, 1970.
[111] Xin, Qianfan. "Durability and reliability in diesel engine system design." Diesel Engine
System Design (2013): 113-202.
144

[112] Lin, Yu-Sheng, and Cheng-Jung Yang. "Spring assisting mechanism for enhancing the
separation performance of digital light process 3D printers." IEEE Access 7 (2019): 71718-
71729.
[113] Bae, Keun-Ho, and Soon-Bok Lee. "The effect of specimen geometry on the low cycle
fatigue life of metallic materials." Materials at high temperatures 28, no. 1 (2011): 33-39.
[114] Gal-Or, Eran, Yaniv Gershoni, Gianmario Scotti, Sofia ME Nilsson, Jukka Saarinen,
Ville Jokinen, Clare J. Strachan, Gustav Boije af Gennä s, Jari Yli-Kauhaluoma, and Tapio
Kotiaho. "Chemical analysis using 3D printed glass microfluidics." Analytical Methods 11,
no. 13 (2019): 1802-1810.
[115] Enders, Anton, Ina G. Siller, Katharina Urmann, Michael R. Hoffmann, and Janina
Bahnemann. "3D printed microfluidic mixers—A comparative study on mixing unit
performances." Small 15, no. 2 (2019): 1804326.
[116] Williams, Manda S., Kenneth J. Longmuir, and Paul Yager. "A practical guide to the
staggered herringbone mixer." Lab on a Chip 8, no. 7 (2008): 1121-1129.
[117] Therriault, Daniel, Scott R. White, and Jennifer A. Lewis. "Chaotic mixing in three-
dimensional microvascular networks fabricated by direct-write assembly." Nature
materials 2, no. 4 (2003): 265-271.
[118] Kim, Hyun Soo, Timothy P. Devarenne, and Arum Han. "A high-throughput
microfluidic single-cell screening platform capable of selective cell extraction." Lab on a
Chip 15, no. 11 (2015): 2467-2475.
145

[119] Wang, Minjiao, Zhihua Wang, Mingkan Zhang, Wei Guo, Ning Li, Yuliang Deng, and
Qihui Shi. "A microfluidic chip with double-sided herringbone microstructures for
enhanced capture of rare tumor cells." Journal of Materials Chemistry B 5, no. 46 (2017):
9114-9120.
[120] Lin, Ching-Hui, Yi-Hsing Hsiao, Hao-Chen Chang, Chuan-Feng Yeh, Cheng-Kun He,
Eric M. Salm, Chihchen Chen, Ming Chiu, and Chia-Hsien Hsu. "A microfluidic dual-well
device for high-throughput single-cell capture and culture." Lab on a Chip 15, no. 14
(2015): 2928-2938.
[121] Zhang, Rujing, and Niels B. Larsen. "Stereolithographic hydrogel printing of 3D culture
chips with biofunctionalized complex 3D perfusion networks." Lab on a Chip 17, no. 24
(2017): 4273-4282.
[122] Li, Yang, Ting Zhang, Yuan Pang, Ling Li, Zhi-Nan Chen, and Wei Sun. "3D bioprinting
of hepatoma cells and application with microfluidics for pharmacodynamic test of
Metuzumab." Biofabrication 11, no. 3 (2019): 034102.
[123] Liu, Xiaoyan, Wenfu Zheng, and Xingyu Jiang. "Cell-based assays on microfluidics for
drug screening." ACS sensors 4, no. 6 (2019): 1465-1475.
[124] Santana, Harrson S., Mauri SA Palma, Mariana GM Lopes, Johmar Souza, Giovanni AS
Lima, Osvaldir P. Taranto, and Joã o Lameu Silva Jr. "Microfluidic devices and 3D printing
for synthesis and screening of drugs and tissue engineering." Industrial & Engineering
Chemistry Research 59, no. 9 (2019): 3794-3810.
146

[125] Dhiman, Nandini, Peter Kingshott, Huseyin Sumer, Chandra S. Sharma, and Subha
Narayan Rath. "On-chip anticancer drug screening–recent progress in microfluidic
platforms to address challenges in chemotherapy." Biosensors and Bioelectronics 137
(2019): 236-254.
[126] Ozawa, Ryoken, Hideki Iwadate, Hajime Toyoda, Masumi Yamada, and Minoru Seki.
"A numbering-up strategy of hydrodynamic microfluidic filters for continuous-flow high-
throughput cell sorting." Lab on a Chip 19, no. 10 (2019): 1828-1837.
[127] Nasiri, Rohollah, Amir Shamloo, Samad Ahadian, Leyla Amirifar, Javad Akbari,
Marcus J. Goudie, KangJu Lee et al. "Microfluidic‐Based Approaches in Targeted
Cell/Particle Separation Based on Physical Properties: Fundamentals and
Applications." Small 16, no. 29 (2020): 2000171.
[128] Shen, Yigang, Yaxiaer Yalikun, and Yo Tanaka. "Recent advances in microfluidic cell
sorting systems." Sensors and Actuators B: Chemical 282 (2019): 268-281.
[129] Beauchamp, Michael J., Gregory P. Nordin, and Adam T. Woolley. "Moving from
millifluidic to truly microfluidic sub-100-μm cross-section 3D printed devices." Analytical
and bioanalytical chemistry 409, no. 18 (2017): 4311-4319.
[130] Bhattacharjee, Nirveek, Arturo Urrios, Shawn Kang, and Albert Folch. "The upcoming
3D-printing revolution in microfluidics." Lab on a Chip 16, no. 10 (2016): 1720-1742.
[131] Ho, Chee Meng Benjamin, Sum Huan Ng, King Ho Holden Li, and Yong-Jin Yoon. "3D
printed microfluidics for biological applications." Lab on a Chip 15, no. 18 (2015): 3627-
3637.
147

[132] Gong, Hua, Adam T. Woolley, and Gregory P. Nordin. "High density 3D printed
microfluidic valves, pumps, and multiplexers." Lab on a Chip 16, no. 13 (2016): 2450-
2458.
[133] Macdonald, Niall P., Joan M. Cabot, Petr Smejkal, Rosanne M. Guijt, Brett Paull, and
Michael C. Breadmore. "Comparing microfluidic performance of three-dimensional (3D)
printing platforms." Analytical chemistry 89, no. 7 (2017): 3858-3866.
[134] Xu, Yang, Yizhen Zhu, Yifeng Sun, Jie Jin, and Yong Chen. "A Vibration-Assisted
Separation Method for Constrained-Surface-Based Stereolithography." Journal of
Manufacturing Science and Engineering 143, no. 5 (2021).
[135] Xu, Yang, Ziqi Wang, Siyu Gong, and Yong Chen. "Reusable support for additive
manufacturing." Additive Manufacturing 39 (2021): 101840.
[136] Walczak, Rafał, and Krzysztof Adamski. "Inkjet 3D printing of microfluidic
structures—on the selection of the printer towards printing your own microfluidic
chips." Journal of Micromechanics and Microengineering 25, no. 8 (2015): 085013.
[137] Donvito, Lidia, Laura Galluccio, Alfio Lombardo, Giacomo Morabito, Alfio Nicolosi,
and Marco Reno. "Experimental validation of a simple, low-cost, T-junction droplet
generator fabricated through 3D printing." Journal of Micromechanics and
Microengineering 25, no. 3 (2015): 035013.
[138] Su, Ruitao, Jiaxuan Wen, Qun Su, Michael S. Wiederoder, Steven J. Koester, Joshua R.
Uzarski, and Michael C. McAlpine. "3D printed self-supporting elastomeric structures for
multifunctional microfluidics." Science advances 6, no. 41 (2020): eabc9846.
148

[139] Nelson, Matt D., Nirupama Ramkumar, and Bruce K. Gale. "Flexible, transparent, sub-
100 µ m microfluidic channels with fused deposition modeling 3D-printed thermoplastic
polyurethane." Journal of Micromechanics and Microengineering 29, no. 9 (2019):
095010.
[140] Cooksey, Gregory A., Christopher G. Sip, and Albert Folch. "A multi-purpose
microfluidic perfusion system with combinatorial choice of inputs, mixtures, gradient
patterns, and flow rates." Lab on a Chip 9, no. 3 (2009): 417-426.
[141] Ramanathan, Nirupama, Oleg Simakov, Christoph A. Merten, and Detlev Arendt.
"Quantifying preferences and responsiveness of marine zooplankton to changing
environmental conditions using microfluidics." PLoS One 10, no. 10 (2015): e0140553.
[142] Faley, Shannon L., Mhairi Copland, Julien Reboud, and Jonathan M. Cooper. "Cell chip
array for microfluidic proteomics enabling rapid in situ assessment of intracellular protein
phosphorylation." Biomicrofluidics 5, no. 2 (2011): 024106.
[143] Dura, Burak, Mariah M. Servos, Rachel M. Barry, Hidde L. Ploegh, Stephanie K.
Dougan, and Joel Voldman. "Longitudinal multiparameter assay of lymphocyte
interactions from onset by microfluidic cell pairing and culture." Proceedings of the
National Academy of Sciences 113, no. 26 (2016): E3599-E3608.
[144] Dura, Burak, Stephanie K. Dougan, Marta Barisa, Melanie M. Hoehl, Catherine T. Lo,
Hidde L. Ploegh, and Joel Voldman. "Profiling lymphocyte interactions at the single-cell
level by microfluidic cell pairing." Nature communications 6, no. 1 (2015): 1-13.
149

[145] Castiaux, Andre D., Cody W. Pinger, Elizabeth A. Hayter, Marcus E. Bunn, R. Scott
Martin, and Dana M. Spence. "PolyJet 3D-printed enclosed microfluidic channels without
photocurable supports." Analytical chemistry 91, no. 10 (2019): 6910-6917.
[146] Gong, Hua, Bryce P. Bickham, Adam T. Woolley, and Gregory P. Nordin. "Custom 3D
printer and resin for 18 μm× 20 μm microfluidic flow channels." Lab on a Chip 17, no. 17
(2017): 2899-2909.
[147] Au, Anthony K., Nirveek Bhattacharjee, Lisa F. Horowitz, Tim C. Chang, and Albert
Folch. "3D-printed microfluidic automation." Lab on a Chip 15, no. 8 (2015): 1934-1941.
[148] Urrios, Arturo, Cesar Parra-Cabrera, Nirveek Bhattacharjee, Alan M. Gonzalez-Suarez,
Luis G. Rigat-Brugarolas, Umashree Nallapatti, Josep Samitier et al. "3D-printing of
transparent bio-microfluidic devices in PEG-DA." Lab on a Chip 16, no. 12 (2016): 2287-
2294.
[149] Lee, Yuan-Sheng, Nirveek Bhattacharjee, and Albert Folch. "3D-printed Quake-style
microvalves and micropumps." Lab on a Chip 18, no. 8 (2018): 1207-1214.
[150] Kuo, Alexandra P., Nirveek Bhattacharjee, Yuan‐Sheng Lee, Kurt Castro, Yong Tae
Kim, and Albert Folch. "High‐Precision Stereolithography of Biomicrofluidic
Devices." Advanced materials technologies 4, no. 6 (2019): 1800395.
[151] Mä nnel, Max J., Lukas Selzer, Ricardo Bernhardt, and Julian Thiele. "Optimizing
Process Parameters in Commercial Micro‐Stereolithography for Forming Emulsions and
Polymer Microparticles in Nonplanar Microfluidic Devices." Advanced Materials
Technologies 4, no. 1 (2019): 1800408.
150

[152] Gong, Hua, Michael Beauchamp, Steven Perry, Adam T. Woolley, and Gregory P.
Nordin. "Optical approach to resin formulation for 3D printed microfluidics." RSC
advances 5, no. 129 (2015): 106621-106632.
[153] Rogers, Chad I., Kamran Qaderi, Adam T. Woolley, and Gregory P. Nordin. "3D printed
microfluidic devices with integrated valves." Biomicrofluidics 9, no. 1 (2015): 016501.
[154] Young, Warren C., Richard G. Budynas, and Ali M. Sadegh. Roark's formulas for stress
and strain. McGraw-Hill Education, 2012.
[155] Rosalia, Luca, Benjamin Wee-Keong Ang, and Raye Chen-Hua Yeow. "Geometry-
based customization of bending modalities for 3D-printed soft pneumatic actuators." IEEE
Robotics and Automation Letters 3, no. 4 (2018): 3489-3496.
[156] Li, Peng, Zhangming Mao, Zhangli Peng, Lanlan Zhou, Yuchao Chen, Po-Hsun Huang,
Cristina I. Truica et al. "Acoustic separation of circulating tumor cells." Proceedings of the
National Academy of Sciences 112, no. 16 (2015): 4970-4975.
[157] Zhang, Jun, Sheng Yan, Weihua Li, Gursel Alici, and Nam-Trung Nguyen. "High
throughput extraction of plasma using a secondary flow-aided inertial microfluidic
device." RSC advances 4, no. 63 (2014): 33149-33159.
[158] Martel, Joseph M., Kyle C. Smith, Mcolisi Dlamini, Kendall Pletcher, Jennifer Yang,
Murat Karabacak, Daniel A. Haber, Ravi Kapur, and Mehmet Toner. "Continuous flow
microfluidic bioparticle concentrator." Scientific reports 5, no. 1 (2015): 1-12.

151

[159] Bijadi, Sachin, Erik de Bruijn, Erik Y. Tempelman, and Jos Oberdorf. "Application of
Multi-Material 3D Printing for Improved Functionality and Modularity of Open Source
Low-Cost Prosthetics: A Case Study." In 2017 Design of Medical Devices Conference.
American Society of Mechanical Engineers Digital Collection, 2017.
[160] Mohammed, Mazher Iqbal, Joseph Tatineni, Brenton Cadd, Greg Peart, and Ian Gibson.
"Advanced auricular prosthesis development by 3D modelling and multi-material
printing." In DesTech 2016: Proceedings of the International Conference on Design and
Technology, pp. 37-43. Knowledge E, 2017.
[161] Macdonald, Eric, Rudy Salas, David Espalin, Mireya Perez, Efrain Aguilera, Dan Muse,
and Ryan B. Wicker. "3D printing for the rapid prototyping of structural electronics." IEEE
access 2 (2014): 234-242.
[162] Liu, Zhenbin, Min Zhang, Bhesh Bhandari, and Yuchuan Wang. "3D printing: Printing
precision and application in food sector." Trends in Food Science & Technology 69 (2017):
83-94.
[163] Zhu, Mingzhu, Yoshiki Mori, Tatsuhiro Wakayama, Akira Wada, and Sadao Kawamura.
"A fully multi-material three-dimensional printed soft gripper with variable stiffness for
robust grasping." Soft robotics 6, no. 4 (2019): 507-519.
[164] Zhu, Mingzhu, Mengying Xie, Xuanming Lu, Shima Okada, and Sadao Kawamura. "A
soft robotic finger with self-powered triboelectric curvature sensor based on multi-material
3D printing." Nano Energy 73 (2020): 104772.
152

[165] Vogt, Daniel M., Kaitlyn P. Becker, Brennan T. Phillips, Moritz A. Graule, Randi D.
Rotjan, Timothy M. Shank, Erik E. Cordes, Robert J. Wood, and David F. Gruber.
"Shipboard design and fabrication of custom 3D-printed soft robotic manipulators for the
investigation of delicate deep-sea organisms." PloS one 13, no. 8 (2018): e0200386.
[166] Sun, Ke, Teng‐Sing Wei, Bok Yeop Ahn, Jung Yoon Seo, Shen J. Dillon, and Jennifer
A. Lewis. "3D printing of interdigitated Li‐Ion microbattery architectures." Advanced
materials 25, no. 33 (2013): 4539-4543.
[167] Fu, Kun, Yibo Wang, Chaoyi Yan, Yonggang Yao, Yanan Chen, Jiaqi Dai, Steven Lacey
et al. "Graphene oxide‐based electrode inks for 3D‐printed lithium‐ion
batteries." Advanced Materials 28, no. 13 (2016): 2587-2594.
[168] Brunton, Alan, Can Ates Arikan, Tejas Madan Tanksale, and Philipp Urban. "3D
printing spatially varying color and translucency." ACM Transactions on Graphics
(TOG) 37, no. 4 (2018): 1-13.
[169] Mirzaali, M. J., A. Caracciolo, H. Pahlavani, S. Janbaz, L. Vergani, and A. A. Zadpoor.
"Multi-material 3D printed mechanical metamaterials: Rational design of elastic properties
through spatial distribution of hard and soft phases." Applied Physics Letters 113, no. 24
(2018): 241903.
[170] Sundaram, Subramanian, Melina Skouras, David S. Kim, Louise van den Heuvel, and
Wojciech Matusik. "Topology optimization and 3D printing of multimaterial magnetic
actuators and displays." Science advances 5, no. 7 (2019): eaaw1160.
153

[171] Han, Daehoon, and Howon Lee. "Recent advances in multi-material additive
manufacturing: methods and applications." Current Opinion in Chemical Engineering 28
(2020): 158-166.
[172] Hon, K. K. B., L. Li, and I. M. Hutchings. "Direct writing technology—Advances and
developments." CIRP annals 57, no. 2 (2008): 601-620.
[173] Kowsari, Kavin, Saeed Akbari, Dong Wang, Nicholas X. Fang, and Qi Ge. "High-
efficiency high-resolution multimaterial fabrication for digital light processing-based
three-dimensional printing." 3D Printing and Additive Manufacturing 5, no. 3 (2018): 185-
193.
[174] Galliker, Patrick, Julian Schneider, Hadi Eghlidi, Stefan Kress, Vahid Sandoghdar, and
Dimos Poulikakos. "Direct printing of nanostructures by electrostatic autofocussing of ink
nanodroplets." Nature communications 3, no. 1 (2012): 1-9.
[175] Ledesma-Fernandez, J., C. Tuck, and R. Hague. "High viscosity jetting of conductive
and dielectric pastes for printed electronics." In Proceedings of the International Solid
Freeform Fabrication Symposium, pp. 40-55. 2015.
[176] Chen, H., T. Tang, and A. Amirfazli. "Effects of surface wettability on fast liquid
transfer." Physics of Fluids 27, no. 11 (2015): 112102.
[177] Li, Huizeng, Qiang Yang, Guannan Li, Mingzhu Li, Shutao Wang, and Yanlin Song.
"Splitting a droplet for femtoliter liquid patterns and single cell isolation." ACS applied
materials & interfaces 7, no. 17 (2015): 9060-9065.

Abstract (if available)

Abstract Vat photopolymerization (VPP) is an additive manufacturing (AM) process in which liquid photopolymer in a vat is selectively cured by light-activated polymerization. Through decades’ efforts, VPP has become one of the most popular AM processes to fabricate three-dimensional (3D) objects. Compared with other polymer-based AM processes such as material extrusion, VPP can build 3D objects with high accuracy, shape complexity and resolution, smooth surface finish, and fast fabrication speed. However, despite these advantages of the VPP process, several bottlenecks need to be addressed. ❧ Firstly, due to the material contamination issue and time-consuming post-processing operations when switching materials such as cleaning, it is challenging to achieve multi-material fabrication in a single part efficiently, which is the critical technology for various applications. ❧ Secondly, the light-induced fabrication mechanism limits the selection of available materials used in the current VPP. Besides photocurable polymer resin, many more non-photocurable materials are available in the market but cannot be 3D printed right now, such as heat cured polymers. These materials span over a wide range of viscosity, and many of them are highly viscous. ❧ Thirdly, as the process uses liquid to form solid objects, there is no structural support from the material during the printing phase. In this case, support structures will often need to be added. The printed supports are totally a waste of materials since they need to be removed after fabrication and cannot be reused. ❧ Fourthly, it is essential to control the flatness and thickness of each layer in the layer-based VPP since they will significantly affect the attachment between two adjacent layers and the dimensional accuracy of the fabricated part. It has been a popular way to add a constrained surface to ensure the flatness and thickness of the newly cured layer. However, for the constrained surface-based VPP, the curing process leads to adhesive bonding between the newly cured and constrained surfaces. To proceed with the fabrication process, a separation force is required to detach them. Excessive separation force will cause damage to the built layers and the constrained surface. ❧ Material jetting is an AM process that allows for full-color parts and enables multiple materials in a single part easily. Photocurable resin or thermally molten materials are deposited onto a build platform using a drop on demand (DOD) approach layer by layer to make parts. The process benefits from its high accuracy and multi-material fabrication ability. However, current material jetting is suffering from limit material options. Viscosity is the main determinant that hinders the development of the process. Besides, a support structure is often required. Unlike VPP, printing support structure via material jetting will waste material and slow the printing speed. ❧ To solve the above issues, my methodology is to treat materials differently. Therefore, I propose a novel hybrid VPP to address the limitations mentioned above. Four components work coordinately and constitute the AM process. ❧ Firstly, I adopt a top-down configuration for the hybrid VPP to print common photocurable polymer resin. To achieve multi-material 3D printing, a new drop-on-demand method is developed to deposit viscous material droplets onto the platform or built layers. For this purpose, I developed a novel droplet-based process called direct droplet writing. The printer head has two working phases. Inspired by the shearing process blanking, a sealed metal capillary tube moves up and down periodically to mechanically split a single material droplet from a continuum of liquid material during each stroke. The deposition of material droplets hang on the tube tip is determined by liquid bridge splitting. The novel printer head allows for firing precisely the right amount of viscous material at precisely the right place. If the viscous material is photocurable, an ultraviolet (UV) light source in VPP will be utilized to solidify the deposited material. ❧ Secondly, water-soluble polymer or wax often works as sacrificial support material. However, printing support will increase total fabrication time consumption and lead to material waste. In order to reduce the amount of 3D printed external support, I developed an automatic and reusable pin-like metal support for overhang features in part. After the fabrication of each layer, the metal pins that work as support will move up one-layer thickness and stop at the specified height automatically. The stop positions of the metal supports are determined by a combination of metal tubes with pre-defined lengths and magnetic rings. 3D printing material can then be deposited on the metal supports. Consequently, the supports that need to be printed will be dramatically reduced. I demonstrate the idea using a filament fused fabrication (FFF) 3D printer incorporating the automatic reusable metal support mechanism. The corresponding slicing software that can work with the printer compatibly was developed. Several test cases are presented to validate the effectiveness and efficiency of my approach. ❧ Thirdly, for the roof features in a part design that needs an internal support structure, I was enlightened by transfer printing and developed a double exposure method to solve this issue. The roof feature is first fabricated on an extra build platform by mask image projection and then connected to the built part with the second exposure of the mask image. By this approach, the internal support structure is not required. Besides, I noted that the enrollment of an additional build platform could largely reduce light penetration that may cure liquid resin inadvertently. With elegant control over light energy distribution along the z-direction, I successfully applied this method to microfluidic flow channels fabrication and printed 10 um heigh flow channels using transparent resin. The novel process is also verified by other test cases, including 3D microfluidic channels, a particle sorting device, and a microfluidic valve. This work will significantly advance the application of VPP in microfluidic device fabrication. ❧ Lastly, a constrained surface was utilized to ensure z accuracy in the presented hybrid VPP. As for the separation force problem, a new vibration-assisted (VA) separation design is presented. To find the best way to use the VA separation method, experiments on the separation performance under different parameters, including vibration frequency, pre-stress level, and exposure area, were conducted. Based on the collected separation force data, an analytical model based on fatigue fracture mechanics was built. The separation behaviors related to different shape sizes and topologies were also studied and compared. The results showed that the separation force in VPP was significantly reduced using the VA-separation method. Furthermore, the relationship between the separation force and the separation time conforms to the stress-based fatigue model. This study also provides insights into choosing process parameters by considering the trade-offs between separation force and building efficiency. ❧ The developed hardware prototypes and software algorithms provide a new systematic strategy to address the issues mentioned above and demonstrated the insights in AM processes study. This work will significantly advance AM technologies in the future.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Hybrid vat photopolymerization: methods and systems

PDF

Motion-assisted vat photopolymerization: an approach to high-resolution additive manufacturing

PDF

Deformation control for mask image projection based stereolithography process

PDF

Slurry based stereolithography: a solid freeform fabrication method of ceramics and composites

PDF

Multi-scale biomimetic structure fabrication based on immersed surface accumulation

PDF

Fabrication-aware machine learning for accuracy control in additive manufacturing

PDF

3D printing of polymeric parts using Selective Separation Shaping (SSS)

PDF

Enhancing the surface quality and dimensional accuracy of SIS-metal parts with application to high temperature alloys

PDF

Energy control and material deposition methods for fast fabrication with high surface quality in additive manufacturing using photo-polymerization

PDF

Scalable polymerization additive manufacturing: principle and optimization

PDF

Selective separation shaping: an additive manufacturing method for metals and ceramics

PDF

Selective Separation Shaping (SSS): large scale cementitious fabrication potentials

PDF

Statistical modeling and machine learning for shape accuracy control in additive manufacturing

PDF

Machine learning-driven deformation prediction and compensation for additive manufacturing

PDF

Nanostructure interaction modeling and estimation for scalable nanomanufacturing

PDF

The extension of selective inhibition sintering (SIS) to high temperature alloys

PDF

Modeling and analysis of nanostructure growth process kinetics and variations for scalable nanomanufacturing

PDF

Real-time flow control of viscous fluids using 3D image processing

PDF

Material and process development and optimization for efficient manufacturing of polymer composites

PDF

Metallic part fabrication wiht selective inhibition sintering (SIS) based on microscopic mechanical inhibition

Asset Metadata

Creator Xu, Yang (author)

Core Title Hybrid vat photopolymerization processes for viscous photocurable and non-photocurable materials

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Industrial and Systems Engineering

Degree Conferral Date 2021-12

Publication Date 09/14/2021

Defense Date 07/15/2021

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag additive manufacturing,hybrid manufacturing processes,multi-material 3D printing,OAI-PMH Harvest,vat photopolymerization,viscous materials

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Chen, Yong (committee chair), Huang, Qiang (committee member), Zhao, Hangbo (committee member)

Creator Email yxu195@alumni.usc.edu,yxu195@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC15916712

Unique identifier UC15916712

Legacy Identifier etd-XuYang-10060

Document Type Dissertation

Format application/pdf (imt)

Rights Xu, Yang

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu