University of Southern California Dissertations and Theses

Expanded functionality and scalability of modular fluidic and instrumentation components

(USC Thesis Other)

Expanded functionality and scalability of modular fluidic and instrumentation components

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Expanded functionality and scalability of Modular Fluidic
and Instrumentation Components.
Bryant Thompson
December 2018
Doctor of Philosophy (BIOMEDICAL ENGINEERING)
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
1
1 Abstract
In this report, a novel micro
uidic platform is introduced. By leveraging the benets of advanced
stereolithography over traditional methods of fabricating micro
uidic devices, a library of stan-
dardized, Modular Fluidic and Instrumentation Components (MFIC's) is designed, manufactured,
and tested. This manufacturing approach allows for the large-scale production of 3D micro
uidic
components that were previously dicult, if not impossible, to generate. In this report we apply
MFIC assembled systems towards high-precision mixing, the control of multi-phase
ows, and
real-time spectrophotometry. A statistical approach to analyze the performance and predictabil-
ity of micro
uidic mixing in the context of mass manufacturing is demonstrated. This study
serves as a basis for considering massively parallelized MFIC systems that are able to execute
high-precision mixing for general bench top applications. Furthermore, the surface chemistry of
MFIC's are modied via initiated chemical vapor deposition (iCVD). Surface channel modication
via iCVD further enables this platform to t the demands of mass customization and manufac-
turing in the context of scalable micro
uidic design. Lastly, an MFIC enabling real-time optical
sensing is introduced to expand the functionality of the MFIC library. This component is scalable
and incorporates o-the-shelf optical detection for absorbance-based sensing. This novel MFIC
is integrated into a "Multi-Stage Stopped Flow" system. This system has been developed to col-
lapse the bench top work
ow necessary to execute Gold Nanoparticle (AuNP) based aggregation
assays commonly used to monitor enzyme activity. Here, the activity of a urinary biomarker for
bladder cancer, Hyaluronidase, is monitored.
2
Contents
1 Abstract 2
2 Introduction 5
2.1 The rise of micro
uidic technology . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Early attempts to modularize micro
uidic device fabrication . . . . . . . . . . . 8
2.3 Fluid
ow in the context of micro
uidics . . . . . . . . . . . . . . . . . . . . . 15
2.3.1 Navier-Stokes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.2 Reynolds Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3 Discrete Elements for 3D Micro
uidics 16
3.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Hydraulic Analogy to Electronics . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3 Design Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Resistance Validation via Tunable Mixing . . . . . . . . . . . . . . . . . . . . . 22
3.5 Recongurable Droplet Generator . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4 Predictable Mixing using Discrete Element Micro
uidics 36
4.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Resistor-Based MFIC Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3 Rened Resistor Tolerance Approximation . . . . . . . . . . . . . . . . . . . . . 40
4.4 Parallel and Series Mixing Circuit Topologies . . . . . . . . . . . . . . . . . . . 42
4.5 Network Analysis of Circuit Topologies . . . . . . . . . . . . . . . . . . . . . . 44
4.5.1 Analysis of 2-input Fork Topology . . . . . . . . . . . . . . . . . . . . . 44
4.5.2 Analysis of 3-input Fork Topology . . . . . . . . . . . . . . . . . . . . . 45
4.5.3 Analysis of 3-input Ladder Topology . . . . . . . . . . . . . . . . . . . 46
4.6 Statistical Determination of Operational Tolerance . . . . . . . . . . . . . . . . 48
4.7 Mixing Law Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.7.1 Deriving Volume Fraction from 2-input Fork Experimentation . . . . . . 58
4.7.2 Deriving Volume Fraction from 3-input Fork Experimentation . . . . . . 59
3
4.7.3 Deriving Volume Fraction from 3-input Ladder Experimentation . . . . . 59
4.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5 Engineered hydrophobicity of Modular Fluidic and Instrumentation Components 61
5.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.2 Device Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3 Surface Modication via iCVD . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.4 Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6 Spectrophotometry in modular micro
uidic architectures 73
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.1.1 Multi-stage Stopped-Flow System (MSSF) for Quantitative Enzyme Biomarker
Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.1.2 Spectrophotometer MFIC . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.1.3 Urinary Enzyme Biomarker for Bladder Cancer . . . . . . . . . . . . . . 75
6.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.2.1 MSSF System Principle of Design and Operation . . . . . . . . . . . . . 76
6.2.2 Design of a Spectrophotometer MFIC . . . . . . . . . . . . . . . . . . . 79
6.2.3 MSSF Benchmarking on HAase Assay . . . . . . . . . . . . . . . . . . . 81
6.3 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.3.1 Fabrication of MFICs . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.3.2 MSSF System Architecture and Maintenance . . . . . . . . . . . . . . . 87
6.3.3 HAase Activity Assay Parameters . . . . . . . . . . . . . . . . . . . . . 87
6.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
7 Conclusions 91
8 Acknowledgments 92
References 93
4
2 Introduction
2.1 The rise of micro
uidic technology
In 1979, researchers at Stanford Electronic Laboratories at Stanford University set their sights
on developing a high-precision method for quantifying gases in a mixture. In order to benet from
the scaling-laws inherent to micron-scale architectures, manufacturing techniques from semicon-
ductor processing were adopted to create a miniaturized gas chromatography (GC) system [1].
Researchers quickly began to note the advantages in adopting classic electronics manufacturing
processes for the development and application of micro analysis systems. A decade later, the
emergence of the term micro
uidic was rst brought to light under the Swedish Pharma com-
pany Pharmacia Biosensor AB. Researchers in this group applied photolithographic techniques to
fabricate a miniaturized sensor capable of quantifying the kinetics of monoclonal antibody-antigen
reactions [2]. The technological breakthrough of being able to manipulate such low volumes of

uids (10
9
to 10
18
liters) within channels in the range of ten to hundred of microns in dimen-
sion [3] caught on rampantly over the next three decades. Today, micro
uidic technology has
advanced to several sectors including molecular analysis, analytical chemistry, materials synthesis,
and biomedical research [3{7].
As part of its advancement to new elds, micro
uidic device fabrication has experienced several
generations of ingenuity in the past few decades. In the late 90's researchers aimed to redirect
the standard method by which micro
uidic devices were fabricated [8]. Adoption of micro
uidic
technology grew quickly with the inception of rapid prototyping of micro
uidic systems through
the use of a transparent elastomeric material known as poly(dimethylsiloxane), or PDMS [9, 10].
In 1998, work described by Duy et al. demonstrated a method of designing and fabricating a
micro
uidic system out of PDMS in less than 24 hours. Since then, much of modern micro
u-
idic work in the academic setting has adapted to fabrication via soft-lithography to manufacture
devices. During this time, the current standard of device fabrication was dependent on microma-
chining. To eliminate the need for specialized facilities that lead to expensive and time consuming
processes, researchers at the Whitesides Research Group at Harvard University introduced PDMS
as a polymer to reduce the time, complexity, and cost of prototyping and manufacturing [11]. The
5
2002 work by McDonald and Whitesides breaks down the steps towards generating devices out
of PDMS for rapid prototyping. The process begins by developing a high-resolution transparency
as a photomask to generate a master on a silicon wafer through photolithography (Fig. 1). Once
the master has been generated, PDMS is poured over the master and left to cure at 70

C for
1 hour. PDMS is commonly mixed in a 10:1 ratio ((v/v) base:curing agent), however, this is
typically tuned depending on the desired substrate stiness (e.g. culturing of dierent cell lines
will require dierent PDMS stiness). The inverse mold is then peeled o and sealed against a

at surface, such as on glass or another slab of PDMS that has been cured on a blank wafer.
6
Figure 1: Scheme describing the rapid prototyping of a mi-
cro
uidic device via PDMS, named Soft Lithography. Prop-
erty of [11].
Soft lithography brought us closer to realizing the large scale automation of chemical and bio-
logical processes in a cost eective manner. Nonetheless, this approach requires a clean-room
setting to generate a master mold, which has a nite life span.
Further, existing methods of device fabrication may obscure the ability to perform rapid network
analysis and system characterization. Classic device fabrication will place most of the design eort
7
on device functionality and little emphasis placed on spatial layout. While eective, this method
continues to be a highly time consuming, labor intensive, and costly process that is limited to
yielding devices on a 2D plane [12]. Approaches for rapid, facile, and low-cost construction of
micro
uidic devices continues to be explored.
2.2 Early attempts to modularize micro
uidic device fabrication
In the technology sector, competition and a drive for 'Mass Customization' has pushed modu-
larity to be a key factor in the development of a product architecture that is eective in several
respects: product cost, manufacturing cycle, product
exibility, serviceability and multi-generation
product platform planning [13]. This has been evident in products such as computers, telecom-
munication devices and related technologies. As Ishii and Lee note, the short technology life-cycle
of the functionality of these example products on top of the growing demand for more features
drives designers to optimize the modularity of components [14]. The same necessities can be
experienced by a micro
uidic device designer wherein modularization can help expand device
features, expedite the manufacturing cycle and reduce product cost. As such, several pieces of
work have aimed to develop modularized micro
uidic platforms. The following is a brief survey
of studies which have demonstrated the modularization of micro
uidic systems.
In the work by Shaikh et al., researchers developed a modular micro
uidic platform whose ar-
chitecture consists of single-chip modules fabricated via soft lithography that includes passive
components (e.g. chemical reactors, micro channels) and a micro
uidic breadboard (FBB) that
contains active components (e.g.
ow rate, temperature, and pressure sensors), as seen in Figure
2. In this platform, active and passive structures are built separately on dierent substrates. Fab-
rication of the FBB device was accomplished by drilling 50m through-holes via deep reactive ion
etching on an oxidized silicon chip that is reversibly bonded to a PDMS layer. This layer contains
valves accomplished through multilayer soft lithography. A visual representation describing the
bonding of a passive PDMS chip to the FBB is shown in Figure 3 [15]. This approach expands
device features by incorporating both passive and active components, but does little to reduce
device cost as it continues to heavily depend on classic device fabrication methodology. More-
8
over, it does not outright reduce the manufacturing cycle and a designer may end up needing
to prototype several 'passive layer' chips to reach a design freeze, where each iteration is still
limited by the underlying fabrication method. Lastly, inherent to its multi-layer architecture, the
platform realizes an increase in dead volume and total channel path length.
9
Figure 2: Overview of the micro
uidic breadboard architec-
ture. (A) Schematic of the micro
uidic breadboard (FBB).
(B) Schematic demonstrating assembly of an FBB that can
be used to create a complete lab on a chip (LOC) by bond-
ing a passive layer. (C) Schematic demonstrating dierent
routing that can be achieved by bonding a passive chip.
(D) Schematic demonstrating the assembly of a multiple-
chip system by interconnecting multiple chips. Reproduced
from Reference [15]. Copyright 2005 National Academy of
Sciences, USA.
10
Figure 3: First generation FBB chip demonstrated by [15].
(A) System assembly using a passive PDMS layer reversibly
bonded to an FBB chip, with an oxidized silicon wafer with
through-waver holes in between. (B) Formation of pneu-
matically actuated valves at the crossing of pneumatically
trigged channels (red) and
uid channels (blue). (C) Mi-
crograph demonstrating a single LOC. (D) A cross-sectional
view of the LOC stack demonstrating activation and func-
tion of pneumatic channels to deform thin membrane for

uid
ow manipulation .Reproduced from Reference [15].
Copyright 2005 National Academy of Sciences, USA.
11
In 2008, work by P.K. Yuen, M. Rhee and M.A. Burns, and Sun et al. looked to provide solutions
to modularization in micro
uidic device assembly. Yuen et al. introduced a modular micro
uidic
platform analogous to the LEGO concept, which enabled 'Plug-n-Play' components fabricated
via stereolithography [16]. Here, a motherboard (similar to the work introduced by Shaikh et
al.), tting components, microchannel inserts and microchips with dierent functions comprise
the platform. Systems are assembled by snapping microchips into place on the motherboard and
connections are made with custom H-shaped microchannel inserts (Figure 4). The 2008 work by
Rhee and Burns introduces modular micro
uidic assembly blocks (MABs) fabricated out of PDMS
via soft lithography (Figure 5). The MAB's contain
uidic routing with geometries constrained
to a 2D plane. To assemble a system, MAB's are laid down side by side, in a fashion similar
to jigsaw pieces, such that channels are aligned to create continuous
ow
uidic pathways [17].
Lastly, Sun et al. introduces modular
uidic components made of PDMS via soft lithography
techniques that interconnect with plastic tubes (Figure 6). The renement here over earlier
pieces of work is the characterization of components as resistors, done so by lumping together
parameters of individual PDMS components [18]. This analysis is not necessarily an advantage as
components do not follow a standardized footprint and modeling isn't reduced to an element level.
Figure 4: (left) CAD representation of system motherboard.
(right) Micro
uidic components and microchannel inserts
used to construct micro
uidic assembly. Reproduced from
Reference [16]. Copyright 2008 Royal Society of Chemistry
12
Figure 5: (a) photo of SU-8 mold used to create PDMS
MABs. (b) Free-standing MABs, pre-assembly. (c) Assem-
bly of MABs used to assemble a U-turn channel, bonded
onto a glass slide. (d) MAB alignment, with no post-
treatment. (e) MAB alignment with post-treatment. Re-
produced from Reference [17]. Copyright Royal Society of
Chemistry.
Figure 6: (a-e) Illustrations demonstrating example mod-
ules employed in the system introduced by [18], with respec-
tive
ow resistances (f-j). Reproduced from Reference [18].
Copyright 2008 Royal Society of Chemistry
In this report, a platform consisting of Modular Fluidic and Instrumentation Components fabri-
cated via stereolithography is demonstrated. Connections between components are self-registered
and reversible allowing for the recycling of components. Assemblies and channel routing are not
limited to a 2D plane. Beyond this, the system performance for scaling in the context of mass
13
manufacturing is explored and validated through experimental and statistical analysis. The ability
to modify the surface chemistry of discrete micro
uidic elements at scale is also demonstrated.
Lastly, a scalable method for incorporating o-the-shelf optical detection for real-time absorbance-
based sensing is introduced in a novel MFIC. This MFIC is then integrated into a "Multi-Stage
Stopped Flow" system that collapses the bench top work
ow necessary to execute Gold Nanopar-
ticle (AuNP) based aggregation assays to monitor enzyme activity. This system is non-disposable,
self calibrating, and utilizes minimal reagent volumes. Here, we monitor Hyaluronidase, a urinary
biomarker for bladder cancer.
14
2.3 Fluid
ow in the context of micro
uidics
2.3.1 Navier-Stokes
Fluid
ow in the nano- and picoliter range behave dierently than
ow in larger scale systems [6].
Key physical properties that dene
uid
ow in micro
uidic devices are brie
y surveyed here, with
which a foundation is established for the bulk of studies to follow.
Velocity elds of incompressible
uids (i.e. constant density where the
ow velocity is much
smaller than the sound velocity in the liquid) in motion can be expressed by the incompressible
Navier-Stokes (N-S) equation, Newton's second law for
uid particles. For a uniformly viscous
Newtonian
uid, the N-S equation is described in terms that account for inertial forces, internal
pressure forces, and viscous forces, respectively:

@~ u
@t
+uru

=rp~ ur~ u +r
2
~ u (1)
This can then be reduced to a simplied Stokes equation, where convective forces are negligible
(explanation on this assumption provided in the the section titled Reynolds Number):

@~ u
@t
=rp~ ur~ u +r
2
~ u (2)
Here, ~ u, m s
1
, is dened as ~ u = ~ u(~ r;t), the velocity eld of a
uid at a particular time, t,
in space, ~ r. Moreover, the
uid density, , is described in units of kg m
3
, the viscosity, , is
described in units ofPas, andp is the pressure characterized by the unitPa. If we now consider
the aspect ratio (L x W) of a micro
uidic channel, the channel length dwarfs the cross-sectional
width (e.g. 20um x 5000um). With this in mind, we consider a system with an innitely long
cylindrical channel where
ow has converged to steady state The inertial and internal forces from
Eq. 1 are then reduced to zero, and Eq. 2 becomes:
rp =r
2
~ u (3)
which shows a balanced relationship between the net pressure of a system and the net viscous
force attributed to it. Flow within a pressure-driven system is termed Poiseuille
ow, which follows
a parabolic velocity prole across the cross section of a channel, such that u = 0 at r =R [19]
gives
ru =
R
2
r
2
4

dp
dx

=u
max

1
r
2
R
2

(4)
15
The Pousielle Flow, Eq. 4, is spatially integrated such that Hagen-Poiseuilles law is described as
Q =
R
4
8
p
L
(5)
where the volumetric
ow rate, Q (m
3
s
1
) , in a circular channel of length L (m), is described
in terms of the pressure loss, p, the radial cross section, R (m), and the dynamic viscosity
(Pa s).
2.3.2 Reynolds Number
Perhaps the most recognized dimensionless number in micro
uidics is the Reynolds number (Re),
an expression that characterizes the relationship between inertial forces to viscous forces in a

uidic system (Eq. 6). Inertial force is force from the momentum of a
uid. Denser
uids with
high velocity with have more momentum, or inertia, than
uids with low density traveling at low
velocity. Viscous forces arise from frictional shear forces from the dierent layers in a
owing

uid.
Re =
f
intertial
f
viscous
=
U
0
H
0

(6)
Here, is the density of the
uid,U
0
the velocity,H
0
the channel dimension, and the dynamic
viscosity in terms of Pa s. In the realm of micro
uidics, systems experience
uid
ow with low
Re values, which comes as a result of viscous forces dominating the inertial forces of
uid
ow,
causing the characteristic laminar
ow experienced by micro
uidic
uid
ow. Low Re values leads
to a simplication of the Navier-Stokes equation by removing the nonlinear terms of the equation
leading to the Stokes equation (Eq. 2).
3 Discrete Elements for 3D Micro
uidics
K.C. Bhargava, B. Thompson, N. Malmstadt. (2014) "Discrete elements for 3D micro
uidics".
Proceedings of the National Academy of Sciences, 111(42), pp.15013-15018.
3.1 Background and Motivation
Most advancements in the micro
uidics space have been accomplished on devices limited to 2D
architectures. The planar orientation inherent to integrated circuit design does not necessarily
16
translate to hydraulic micro systems that increasingly tend to require a level of intricacy not
tenable by planar geometries. With this restriction in place, increased system complexity leads to
increased spatial requirements, which leads to in
ated costs and increased manufacturing di-
culty. In order to address these limitations, a platform is herein introduced in which self aligning
MFIC's can be quickly connected in a reversible manner. MFIC's can be connected to create
fully customized micro
uidic systems free of clean-room processing. This work addresses the
increasing need for mass customization with rapid device turnaround and robust functionality. A
library of components was developed to give users the freedom of building systems analogous
to electronic elements that invoke a standardized interconnect footprint. MFIC systems exhibit
hydraulic characteristics analogous to circuit theory that allow for network analysis. Work
ow
for device fabrication is minimized to simple modications in computer-aided design (CAD) that
are translated to operational components with distinct hydraulic characteristics via additive man-
ufacturing.
Several eorts have been made to redirect the current standard of micro
uidic device assembly to-
wards a work
ow that is more cost ecient, placing more focus on customization and turnaround.
In general, three strategies have been proposed: interconnecting monolithic chips [18,20,21], plat-
forms based on a micro
uidic breadboard to interconnect devices [15, 16] and micro
uidic chips
resembling jigsaw pieces with self-aligning channels that assemble in a chip-like fashion [17, 22].
These platforms oer customization of passive components, and in some cases active components,
but are not immediately scalable as they continue to ultimately rely on clean-room fabrication
processes. In this chapter, an entirely new micro
uidic platform is developed from a library
of standardized micro
uidic components fabricated using high-precision additive manufacturing
techniques.
3.2 Hydraulic Analogy to Electronics
The highly complex and often times unstructured nature of micro
uidic design today has taken a
step away from the ability to perform network analysis in a discrete manner. While computational
17

uid dynamic (CFD) analysis provides methods of resolving the issue, it remains the case that CFD
approaches require expensive proprietary software that demands knowledge in theoretical
uid
dynamics [23]. Though several academic groups have adopted network analysis methods from
circuit theory in their platforms [18,24,25], these techniques have not been particularly attractive
as they are built on platforms necessitating clean room fabrication. Here, additive manufacturing
is explored to generate MFIC's liberated from limitations common to micromachining techniques,
that have been previously explored and validated [26{28].
The platform of discrete micro
uidic components introduced here follows the
uidic analogy to
Ohms law from circuit theory, Hagen-Poisueille's law, that holds true for a
ow with low Re by
an incompressible
uid that is pressure-driven [29]. As stated earlier, the Hagen-Poiseuille law
(Eq. 5) distinguishes the relationship between pressure,
ow-rate, and resistance of a system.
Intuitively, the law can be reduced to:
p =QR
h
(7)
such that the expression is analogous to Ohms law
V =IR (8)
where the change in pressure, p, is equivalent to voltage drop,
ow rate,Q, follows the behavior
of current in electronics and the hydraulic resistance, R
hyd
(Pa s
3
m
1
), is simply:
R
hyd;cylinder
=
8L
R
4
(9)
for a straight cylindrical channel. The channels implemented here are square, not cylindrical, such
that the resistance is expressed as:
R
hyd;square
= 28:4
L
h
4
(10)
where resistance is dominated by h, the square channel side length (m), as demonstrated in
simulation (Figure 7). This allows for an overall system development strategy that is more closely
based on the methodology behind circuit element production and electronic circuit design [30].
18
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
10
20
30
40
50
60
70
80
R (mm)
Hydraulic Resistance (GPa s m
3
)

Channel Length = 24.4mm
Channel Length = 44.8mm
200 400 600 800 1000 1200
0
50
100
150
h (um)
Hydraulic Resistance (GPa s m
3
)

Channel Length = 6mm
Channel Length = 8mm
A B
Figure 7: (A) Hydraulic resistance for cylindrical channel simulated at two dierent
lengths: 24.4mm and 44.8mm. For the case of cylindrical channels, hydraulic resistance
is dominated by the channel radius, simulated here from 0.2- to 1-mm radii, as a steep
decay is shown for increased channel radii. (B) Hydraulic resistance for square channel
simulated at two dierent path lengths, 6mm and 8mm, in this case the path length of
a straight pass component and a connector, respectively. In this case, the resistance
is dominated by the cross sectional side length of the channel. Simulation is shown
for cross sectional side lengths ranging from 200-um to 1200-um to demonstrate quick
drop around 300-um.
3.3 Design Concept
The micro
uidic platform introduced here is built around individual 'block' components, designed
and fabricated to follow a standard 1x1x1 cm cubic geometry. This architecture enables simple
to complex system assembly having to pay no concern to the spatial distribution of components.
Individual components are characterized by the
uid manipulating element they contain, such as t-
junctions, mixers, connectors, and port components. Port elements (i.e inlet/outlet components)
were designed to t standardized 1/16" OD tubing, building towards a platform that does not
require proprietary interconnect solutions.
Plug-plug ended connectors were designed to create reversible seals between plug ends and socket
ports on components. Self aligning plug ends insure channel continuity between the 1 cm side
19
length connector channel to the 500 m or 750 m channel side length of the components
(Figure 8). Connectors channel side length was made larger than component channel side length
to reduce the eects of hydraulic resistance while assuring low Reynolds numbers. Table 1
summarizes the library of components alongside their hydraulic resistances, calculated with the
use of Eq. 10. All components were constructed with a square channel geometry such that net
resistance was varied by modulating channel side length and/or channel path length (Table 2).
All components were built using the stereolithography additive manufacturing services of Fineline,
a protolabs company (now known as proto labs). The material chosen, Somos
R
WaterShed XC
11122, is a photopolymer material that generates ABS like components upon print. In particular,
the material generates components that are optically clear, allowing for visual inspection to
debug or identify root cause faults. Past work has validated this material as a viable choice for
applications requiring biocompatibility [27].
Connector
Component
Cue
Element
Spacer
a b
c
10mm
B A
C
Figure 8: (A) CAD assembly of a male-male connector con-
nected to a component with female-type port and a 750-um
straight-pass element. (B) Actual image of two components
connected together via a connector. Flat surfaces on com-
ponents allow for easy visual inspection to assure correct
seal has been made between connect pin and component
port. Cylindrical spacer on connector shown to magnify
contents passing through it. (C) CAD example of chip-
to-world interface accomplished by connecting standardized
1/16" PEEK tubing to interface tting.
20
Table 1: A compiled list of the constructed components with their designed
cross-sectional length, h, and resulting designed resistance, R
Des
. Expected
resistance, R
Exp
, was determined by resolving the square channel resistance
equation, Eq. 10, with measuredh values, and presented here with percentile
SD.
21
Table 2: List of components with their respective names,
CAD drawings, physical cues, and equivalent circuit dia-
grams.
3.4 Resistance Validation via Tunable Mixing
The hydraulic resistance of discrete elements was quantied experimentally through the construc-
tion and use of a 2-input 1-output circuit that allows for comparison of
ow rate through each
branch, which is associated to the respective branch resistance. The system, shown in Figure
22
9, is set up with two parallel branches with resistances R and R
s
each of which is grounded by
a Milli-Q water solution containing colored dye at the inlets. The outlet, R
o
, is also grounded
to Milli-Q water. One branch remains as a reference branch with constant resistance, R, and
parallel to it a branch, R
s
, that can be tuned by modulating a single component, R
select
, while
structural components are maintained identical on both branches. In order to determine the
ow
rate division attributed to changes by R
select
,
ow rates of each branch, originally generalized by
Eq. 7, were determined (Eq.11 -13) and a ratio of the two
ow rates (Eq.14 -15) produced the
mixing ratio, m
o
, described by Eq. 16.
R =R
I
+R
TJ;750
+ 3R
C;1000
+R
L;750
+R
SP;750
=R
struct
+R
ref
(11)
R
s
=R
I
+R
TJ;750
+ 3R
C;1000
+R
L;750
+R
select
=R
struct
+R
select
(12)
R
o
=R
TJ;750
+R
C;1000
+R
I;750
(13)
Q
y
=
0P
R
s
(14)
Q
b
=
0P
R
(15)
m
o
=
Q
y
Q
b
=
R
struct
+R
ref
R
struct
+R
select
(16)
The component dening R
select
was then swapped with components characterized by dierent
hydraulic resistances to modulate the mixing ratio. The fraction of resident widths of each dye
steam entering the junction was optically measured and compared to the theoretical fraction
determined by Eq. 16. As suggested by [31], if both input streams share equal dynamic viscosity,
then the ratio of resident widths can accurately represent the
ow rate ratio. The comparison
between theoretical and experimental mixing ratio is given in Figure 10, and is shown to have
strong agreement, validating the premise of applying resistance components to modulate mixing
23
ratio.
This concept can be expanded to systems containing several outlets, as referenced in Figure 11.
With classic micromachining techniques, this sort of assembly would be highly dicult if not
impossible to recreate. Figure 11 takes the simple comparator sub circuit and expands it to a 2-,
3-, and 4-outlet system with the integration of a TJ, X, or XT component to expand the number
of sub-circuits. A constant pressure source driven system then allows for the analysis of each
sub-circuit outlet where output is a resistance ratio driven system. Equivalent circuit diagrams
are shown in Figure 12.
Q
b
R
s
R Q
y

P
Q
o

R
o

a b
R
select

R
ref

A B
Figure 9: (A) CAD assembly for the 2-input, 1-output system in which
R
select
is varied and concentration changes respectively with R
ref
as a
constant reference resistance. (B) Circuit diagram equivalent to CAD
assembly where R is the total branch resistance of the reference branch
andR
s
is the total branch resistance of the tuning branch, with R
select
included in calculation along with interface, connector, and T-junction
components.
24
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.5 1 1.5 2 2.5 3 3.5
m
0

R
ref
/R
select

Experimental
Model
Figure 10: Validation of experimental mixing ratio model in
comparison to resident width ratio for reference branch to the
resistance modulated branch, R
s
. Several R
select
were chosen
and error bars represent the standard deviation over 12 optical
measurements. Inset micrographs depict sample images used
to determine the width of collinear streams
owing into the
junction for particular R
select
components .
25
a
b
c
d
e
f
B D F
A C E
Figure 11: The standardized library footprint allows for the expansion of single-
outlet sub circuits to be parallelized in construction in order to create tunable mixers
with (A/B) two, (C/D) three, and (E/F) four outlets. As in the simpler 2-input
1-output scenario, each sub circuit in the expanded constructs can be tuned to
perform particular mixing ratios by selecting a resistance component to modulate
the branch resistance to the desired mixture.
26
C
B
A
Figure 12: Circuit equivalents for (A) 2-, (B) 3-, and (C) 4-
outlet tunable mixer system. Each sub-circuit produces an
output solution,m
o;n
, that is driven by a negative displace-
ment pressure source, which allows for parallel sub circuits
to be treated as equivalent parallel resistances. Moreover,
the symmetrical construct with respect to each output al-
lows the select and reference resistors to be the only fac-
tors, apart from structural resistance, that play a role in
output mixture.
27
3.5 Recongurable Droplet Generator
One of the key features driving the need for modularity is the ability to give a designer the
capability of mass customization. In this section, two quickly recongurable droplet generating
systems based on dierent principles of operation are demonstrated: T-junction and Flow-Focus
droplet generators. The T-Junction formatted droplet generator, represented in Figure 13,
utilizes an individual syringe pump to drive two dye-bearing
uid aqueous streams, which are
mixed and sheared by a carrier oil stream at the junction to generate micro-droplets. The carrier
oil stream was held constant at 1 ml h
1
, while the aqueous phase
ow rate was varied to
show how modulating the introduction of the aqueous phase to the carrier phase aects the size
distribution of droplets (Figure 14A). The T-Junction droplet generator device is expanded to
a four outlet system, as shown in Figure 15. This serves as a demonstration in the platform's
ability to quickly customize and expand devices to complex geometries with high throughput.
Here, the input carrier phase is split into four streams where each stream intersects an input
aqueous phase stream, eectively generating micro droplets in each of the four output branches.
As a further demonstration of recongurability, a Flow-Focusing droplet generator is assembled to
produce smaller droplets. In this template, a carrier oil phase is focused around an aqueous phase
that
ows perpendicular to it (Figure 16). To go from the T-Junction system to the Flow-Focus
system, an X-junction component replaces the T-Junction component of the previous system in
order for two carrier phase streams to focus around the aqueous phase and pinch the stream,
creating droplets. The carrier phase was kept at a constant
ow rate of 5 ml hr
1
to prevent
coalescing of droplets near the outlet. Droplet lengths were determined for the Flow-Focus
system, verifying the ability to generate droplets whose size depend on the input aqueous phase

ow rate (Figure 14B).
28
a b
c d
Carrier Phase
Aqueous
Phase
Aqueous Phase
A
C
B
D
Figure 13: (A) CAD assembly and (B/C) actual T-junction
droplet generation device. (C) Collinear dye-bearing water
streams passing through a 3-D helical mixer and outputting
a mixed solution that is then (D) used to create droplets
from a shearing carrier oil phase. In this assembly all con-
nectors have a cross sectional side length of 1 mm and all
other elements are 750 m.
29
0
100
200
300
400
500
600
700
800
900
0 500 1000 1500 2000 2500
Droplet Length (µm)
Aqueous Flow Rate (µL/hr)
Carrier Flow Rate = 5000 µL/hr
0
500
1000
1500
2000
2500
3000
3500
0 200 400 600 800 1000 1200
Droplet Length (µm)
Aqueous Flow Rate (µL/hr)
Carrier Flow Rate = 1000 µL/hr
a
b
B
A
Figure 14: Droplet length measurement distribution deter-
mined by taking micrographs at the center axis of the exit
tubing, done for (A) T-Junction and (B) Flow-Focus sys-
tems. Error bars denote SD of 12 length measurements.
We note the smaller achievable droplet sizes for the Flow-
Focus device in comparison to the droplet distribution of
the T-Junction system. For both systems the components
contain a 750-um channels and all connectors have a 1-
mm channel side length.
30
a b
c
d
Carrier Phase
Aqueous Phase
A B
C
D
Figure 15: (A/B) Actual and (C/D) CAD assembly of a
3-D T-Junction assembly to create droplets. Here, cross-
sectional side length for all connectors is 1 mm and all
other components are 750 m.
31
a
Carrier Phase
Aqueous Phase
Carrier Phase
b
c
A B
C
Figure 16: (A) CAD schematic and (B/C) actual
ow-
focus construct. The terminal end of the T-Junction
schematic is replaced with an X-junction component for
which two carrier phase streams focus around an aque-
ous phase
owing through it. The system uses 750 um
components and 1 mm connectors.
As an initial demonstration of integrating o-the-shelf electronics to MFIC's, an optical detection
system in the form of an infrared (IR) sensor pair is explored. Figure 17A demonstrates the
incorporation of a near-infrared emitter-receiver pair housed into a custom IR sensor MFIC. Cus-
tomized features were designed to allow both diodes to sit rmly in place opposite one another
such that the infrared beam intersects a
uidic bearing channel. Here, the channel side length
follows the previously introduced standard resistor class architecture at 642.5um and features a
standard 6 mm path length.
The IR sensor pair was placed downstream of a T-Junction droplet generator to determine the
32
frequency and size distribution of droplets intersecting the IR beam path (Figure 17B). A
uoro-
carbon oil (Halocarbon 4.2) was applied as the carrier phase, which intersected an aqueous phase
stream to create water-in-oil emulsions. Water droplets that cross the infrared beam path will
absorb the infrared light far more than the carrier oil, allowing for acquisition of an analog signal
that dierentiates between water and oil based on a pre-calibrated threshold value (Figure 17C).
After communicating across a micro controller and converting the analog signal to a digital signal,
the droplet length is calculated by utilizing the average
ow velocity in the channel and temporal
duration of a droplet across the beam path, which was then directly compared to measurements
of the microdroplets taken by optical micrographs (Figure 17D). Results are in good agreement
with each other as they fall within deviation of one another, leading to the conclusion that inte-
gration of active sensing and feedback components to micro
uidic architectures is immediately
realizable.
33
Figure 17: (A) CAD assembly of a straight pass channel of 642.5um cross-sectional
side length that runs perpendicular to the beam of a near-infrared (NIR) diode
emitter and phototransistor receiver. (B) The IR sensor is placed down one block
length downstream from the droplet generating t-junction such that the water-in-
oil emulsions absorb the beam as the path is intersected, in turn (C) generating a
periodic signal corresponding to droplet frequency and size. (D) Duration of travel
across beam path and corresponding
ow rate of droplet train was used to calculate
droplet lengths, which were then compared to optical micrograph measurements.
Flow rates for the carrier and aqueous phases were held constant at 5 ml h
1
and
2mlh
1
, respectively. Droplet length average was determined to be 421.22um
27.54um and 416.90um 16.36um by the NIR sensor and optical micrographs,
respectively.
34
3.6 Discussion
In this chapter, we have demonstrated a novel micro
uidic platform enabling the rapid benchtop
assembly of MFIC's. Leveraging the high-resolution of SLA manufacturing and availability of clear
photoresins, designers can assemble and debug complex 3D devices with ease. Furthermore, the
standardized framework on which our library is constructed around allows for hierarchal system
analysis with application of the hydraulic analogy to electronic circuit design, as described earlier.
We also introduce an initial demonstration of the capability to integrate o-the-shelf optical
components into MFIC's with relative ease.
35
4 Predictable Mixing using Discrete Element Micro
uidics
K.C. Bhargava*, B. Thompson*, D. Iqbal, N. Malmstadt (2015) "Predicting the behavior of
micro
uidics circuits made from discrete elements". Scientic Reports, 5, 15609.* Equal Contri-
bution
4.1 Background and Motivation
As a method of elucidating naturally occurring concentration gradients in nature, micro
uidic
technology has been used for their ability to accurately mimic biological events. Biological con-
centration gradients have been noted to govern the development of metastatic cancer [32], mi-
gration of immune cells [33, 34] and cellular development [35]. More over, micro
uidic gradient
generation has served as a powerful tool to characterize cell toxicity of varying drug concentra-
tions [36].
Mixing of reagents and
uids is classically accomplished by the use of syringes, pipettes, burettes,
and other expensive tools that require manual and procedural methodologies. In these systems,
precision tends to widely vary with attributed error highly dependent on volume and method of
operation. As adverse eects on nal desired mixing concentration continue to become an issue
in a world demanding high-precision, a need arises for a simple to use hand held tool with precise
and predictable mixing of low volumes, invariant to operator variability.
In vitro methods of generating gradients have included the use of biological hydrogels [37{39] and
micropipette-generated gradients [40], which are limited to results that lack precision, controlla-
bility, and predictability. More modern advancements in micro
uidic technology have allowed for
the development of devices capable of serial dilution [29, 41, 42], parallel mixing [42], two-layer
linear dilutions [43], 3-D combinatorial mixing [44], and logarithmic concentration gradients [41].
As a limitation of their fabrication process, these systems lack the ability to be rapidly assem-
bled into systems capable of generating dilution gradients to meet user specications. In other
words, these devices are not customizable after fabrication and serve xed applicability in settings
requiring
exibility. Previously, a micro
uidic platform of MFIC's capable of self-aligning and
connecting in a reversible manner was introduced [45]. Here, the same platform is rened with a
library of discrete components dened by their internal hydraulic resistance. This more directly
36
allows micro
uidic components to translate to electronic components and enables the ability to
perform circuit analysis to identify terminal characteristics. A platform of this nature frees users
from focusing their eorts on predetermining spatial distribution required by standard 2D silicon
substrates and redirects attention to the simple selection of components to construct micro
uidic
dilution systems unhindered by complexity. As such, a systematic 3-step device assembly strategy
is presented, which includes: (A) selection of resistance based components, (B) selection of cir-
cuit topology to achieve desired mixture, and (C) initial modeling of each micro
uidic circuit with
expected manufacturing error to simulate operating space of the terminal characteristics. The
rst step is select components with specic hydraulic resistance to achieve the desired output
mixture. Systems assembled from these MFIC's are free from user-to-user error that is typical of
classical mixing methods applied in the laboratory setting. Next, the style of mixing can be rapidly
manipulated with reassembly of components to achieve customized circuit topologies. Lastly, a
simple to follow statistical method is used to determine the operating range of the chosen circuit
topology, taking into account the selected components, their hydraulic resistance and associated
manufacturing error. In this chapter, this work
ow is validated through the construction of two
circuit schematics wherein the output concentration of a single branch in each system is probed
and validated against physical simulation.
4.2 Resistor-Based MFIC Library
As previously described in Discrete Elements for 3D Micro
uidics, the hydraulic resistance of
MFIC's can be tuned by modifying the channel path length or channel side length of an MFIC.
For the purpose of these studies, the channel side length is held constant across the MFIC
library presented in this chapter (Table 3). This library allows designers to select components in
the manner an engineer would select resistors for an electrical assembly. To keep in the spirit of
creating a direct analogy to electronics, 'wires' used in the electronics community make there way
into this MFIC library in the form of wire components. Tese components are designed to have a
hydraulic resistance two orders of magnitude below the least resistive component of the resistor
class components. As previously described by Equation 10, resistance of a square channel can be
modulated by path lengthL or the channel side lengthh, assuming a constant dynamic viscosity,
37
, of 1mPas. A standard resistance unit of 1GPasm
3
is denoted as 1G and constitutes the
basic resistance unit of the resistance based library presented here, dened by a 6mm path length
and 642.5 m channel side length. Resistors above 1G resistance were fabricated by snaking or
coiling path lengths within the blocks to satisfy the square channel resistance equation. In order
for resistor class components to be the only contributers to network resistance, the wire class
components are fabricated to have a 0.01G resistance. To achieve negligible resistance, the cross
sectional side length is extended to 2.0317 mm, which still allows
uid
ow to remain in the
laminar
ow regime at reasonable
ow rates (Re roughly 0.1 at 200 mL/hr) in a 6 mm path
length component. Furthermore, port components used for interfacing to out-of-world systems
(e.g. pumps or syringes) were fabricated with a cross sectional side channel length of 1.1425mm,
creating a resistance of 0.05G. Wire-like components are dened as having parasitic resistance,
similar to the parasitic resistance in electronics that is neglected due to its considerably low
resistance dened as being 2-3 orders of magnitude lower than resistance components.
38
Class Name CAD Model Nomenclature Designed R (G) Expected R (G) Expected Error (%)
Port
(h = 1142.5 um)
Port
R
P
0.05 - -
R
in
3 3.018 10.797 %
Wire
(h = 2031.7 um)
Connector W
C
0.01 - -
Straight Pass W
SP
0.01 - -
L-Joint W
LJ
0.01 - -
T-Junction W
TJ
0.01 - -
Resistor
(h = 642.5 um)
Straight Pass R
1
1 2.304 7.662%
Snaked R
2:5
2.5 0.920 3.305 %
Helix
R
5
5 4.582 2.592 %
R
10
10 9.139 2.244 %
R
25
25 22.824 1.852 %
Table 3: Library of the fabricated MFIC components with their designed hydraulic resistance. Expected Resistance and Error
were determined through Monte Carlo Simulation.
39
4.3 Rened Resistor Tolerance Approximation
Additive manufacturing techniques have an inherent limitation on print resolution, particular to
the mechanics of optics that dene xy and z resolution. Therefore, a certain amount of error is to
be expected in the width,w, length,L, and height,h, of component channels. The resistance of
channels is thus better approximated as a rectangular channel, with error accounted for in the two
print axis, xy and z, as described for by Eq. 17, the hydraulic resistance of rectangular channels
derived by solving the earlier presented Navier-Stokes equation with a Fourier Series method [30].
R
hyd;rectangle
=
12L
h
3
w
"
1
1
X
n;odd
192h
n
5

5
w
tanh

nw
2h

#
1
(17)
Plane tolerance in the xy and z direction were accounted for by fabricating a large quantity of
MFIC's and optically measuring their channel side lengths (Figure 18). Tolerances were then used
in a Monte Carlo simulation to more accurately predict the standard deviation, , of hydraulic
resistance attributed to each element in the resistance-based library. More explicitly, the resistance
of each channel in the xy and z planes for every component was determined by drawing parameters
w,L, andh from a psuedorandom distribution characterized by the measured tolerances for both
orientations. Each channel segment within a component contributing to the resistance was then
summed to give the expected resistance (Figure 19), along with the expected error that is dened
as 2 for 5000 parameter drawings (Table 3). The deviation for every resistor class component,
termed here as the manufacturing tolerance, assures that 95% of the fabricated components fall
within the listed tolerance.
40
B
A
Figure 18: Measured cross-sectional side length of the (A)
xy and (B) z directions for resistance-based compo-
nents. Optical micrograph measurements determined xy
and z to be 659:98 12:47m and 652:49 4:24m,
respectively, for over 70 measurements in each direction.
41
B A
Δxy
Δz
Δxy
Δz
Figure 19: (A) Port opening with centered channel for a 5G component with
642.5m cross-sectional side length. Due to the mechanics of stereolithography,
precision in the xy and z direction is expected to vary. (B) Determined cross-
sectional side length for the xy and z direction, also dened as the print plane
and print axis, respectively, used to create normal distributions for which a Monte
Carlo simulation could draw pseudorandom values from and input into resistance
equation. Total component resistance was then approximated by determining the
resistance of segments respective to xy (green lines) and z (red lines), by
drawing parameter values from their respective distribution and then adding all
calculated segment resistances together.
4.4 Parallel and Series Mixing Circuit Topologies
Two circuit topologies are introduced to study and validate the Hagen-Poisueille law with the re-
ned discrete element library: the Fork topology (Figure 20) and the Ladder topology (Figure 21).
Branches of the Fork topology feature a distinct resistive component. Streams infused through
individual branches meet at a junction. The system is driven by a negative-pressure source, re-
sulting in every branch experiencing an equivalent pressure drop from the point of infusion to
the point of mixing. This makes the concentration of each input at the output very simple to
determine. Moreover, a system of this nature is source-invariant, as the particular volume fraction
42
of each input, in reference to the output solution, is determined solely by the resistive component
in each branch, and not the external source driving the system. This becomes more evident in the
Mixing Law derivation, particular to the topology being investigated, where the resultant volume
fraction of each input, for all topologies explored, is characterized by the resistive components of
the topology.
In the Ladder topology, each input branch, starting from the branch furthest from the negative-
pressure source, acts in series with the mixing resistor. This resistance then functions in parallel
to the adjacent branch. Clumping these resistances together to form a single branch facilitates
analysis, such that this clumped branch acts in series with the proceeding mixer and in parallel
with the adjacent branch. The source-invariant nature of the system allows us to determine the
input volume fraction in reference to the output solution in a facile manner. In this study, a 2-
and 3-input Fork topology is reviewed as well as a 3-input Ladder topology.
. . .
R
3
R
2
R
1
R
N-1
R
N

Figure 20: Generalized Fork topology where each branch
resistance experiences the same pressure drop across each
branch. Mixture between inlets thus occurs in parallel be-
tween branches.
43
R
M
R
M
R
M

R
1
R
2
R
N-2
R
N-1
R
N

. . .
Figure 21: Generalized Ladder topology where mixing
occurs in a serial manner, such that adjacent parallel
branches serially reduce to what is eectively a 2-inlet fork
topology, where R
N
constitutes one branch and all other
branches constitute the other branch (derivation shown in
text).
4.5 Network Analysis of Circuit Topologies
Analogous to practices performed in circuit analysis, the circuit topologies explored here were
analyzed through nodal analysis, such that an expression can be formed for the output concen-
tration of each input with respect to other input lines, denoted here as the volume fraction,

.
We begin by analyzing the 2-input Fork topology and generate a resistor based expression for
quick volume fraction calculation.
4.5.1 Analysis of 2-input Fork Topology
Analysis of a 2-input Fork Topology is accomplished by deriving the volume fraction of each inlet
substance, in reference to the output. The volume fraction of each input stream is expressed as
the ratio between an inlet
ow rate, Q, to the sum total of all input stream
ow rates (Eq. 18 -
19), such that:

1
=
Q
1
Q
1
+Q
2
(18)

2
=
Q
2
Q
1
+Q
2
(19)
44
The
ow-rate across each branch is then expressed by the Hagen-Poiseuille equation, (P =
QR). Here, we characterize these terms by the dierence in pressure from the point of infusion,
P
0
, to the point of at which the two streams converge at the junction, P
x
, and the resistance of
the respective branch (Eq. 20 - 21).
Q
1
=
P
0
P
x
R
1
=
P
x
R
1
(20)
Q
2
=
P
0
P
x
R
2
=
P
x
R
2
(21)
By inserting Eq. 18 - 19 into Eq. 20 - 21, the volume fraction of each inlet substance is reduced
to a set of equations dened by the selected resistance components, eectively the mixing laws
for the 2-input topology (Eq. 22 - 23).

1
=
R
2
R
1
+R
2
(22)

2
=
R
1
R
1
+R
2
(23)
4.5.2 Analysis of 3-input Fork Topology
As in the case for the 2-input Fork topology, the 3-input Fork topology follows the same equiv-
alent pressure paradigm, which allows for a simple reduction to resistor-based volume fraction
expressions. Again, we express

as a ratio of branch
ow-rate to the total
ow-rate (Eq. 24 -
26).

1
=
Q
1
Q
1
+Q
2
+Q
3
(24)

2
=
Q
2
Q
1
+Q
2
+Q
3
(25)

3
=
Q
3
Q
1
+Q
2
+Q
3
(26)
45
The Hagen-Poiseuille Law is then applied to each branch in order to reduce the
ow-rates to an
expression of driving pressure and resistance:
Q
1
=
P
0
P
x
R
1
=
P
x
R
1
(27)
Q
2
=
P
0
P
x
R
2
=
P
x
R
2
(28)
Q
3
=
P
0
P
x
R
3
=
P
x
R
3
(29)
Lastly, the
ow-rate expressions (Eq. 27 - 29) are plugged into (Eq. 24 - 26) to get the volume
fraction equations for each inlet of the 3-input Fork topology. Again, we note that volume fraction
is invariant to the source driving the system:

1
=
R
1
R
1
+R
2
+R
3
(30)

2
=
R
2
R
1
+R
2
+R
3
(31)

3
=
R
3
R
1
+R
2
+R
3
(32)
4.5.3 Analysis of 3-input Ladder Topology
The Ladder topology is examined in the same manner as the previous topologies explored. Again,

is expressed as the ratio between a branch
ow rate to the sum total of
ow rates in the system.

1
=
Q
1
Q
1
+Q
2
+Q
3
(33)

2
=
Q
2
Q
1
+Q
2
+Q
3
(34)
46

3
=
Q
3
Q
1
+Q
2
+Q
3
(35)
Again, the

of each input is reduced to expressions of resistances by application of the Hagen-
Poiseuille Law:

1
=

P
2
R
1

P
2
R
1

P
2
R
2

P
3
R
3
=
1
R
1
1
R
1
+
1
R
2
+

R
3
(36)

2
=

P
2
R
2

P
2
R
1

P
2
R
2

P
3
R
3
=
1
R
2
1
R
1
+
1
R
2
+

R
3
(37)

3
=

P
3
R
3

P
2
R
1

P
2
R
2

P
3
R
3
=

R
3
1
R
1
+
1
R
2
+

R
3
(38)
Here, =
P
3
P
2
. A relative pressure is found through further analysis, where two auxiliary equations
are used to describe the conservation of
ow in the node most near the outlet:
P
2
P
3
R
M

P
3
R
3
=Q (39)
P
2
R
1

P
2
R
2

P
3
R
3
(40)
so that is reexpressed as:
= 1 +R
M

1
R
1
+
1
R
2

(41)
Collectively, this gives an expression for the volume fraction of each input solution:

1
=
R
2
R
3
R
2
R
3
+R
1
R
3
+R
1
R
2
+R
M
(R
1
+R
2
)
(42)

2
=
R
1
R
3
R
2
R
3
+R
1
R
3
+R
1
R
2
+R
M
(R
1
+R
2
)
(43)
47

3
=
R
1
R
2
+R
M
(R
1
+R
2
)
R
2
R
3
+R
1
R
3
+R
1
R
2
+R
M
(R
1
+R
2
)
(44)
In this study, the rst branch of every topology runs a stock 0.34 M NaCl solution which is
diluted down in dierent forms, depending on the topology and set of resistances chosen. All
other branches run DI water. For this reason the focus of the volume fractions is particular to
the R
1
branch of every topology, whose mixing laws is summarized by Table 4.
Circuit Topology Mixing Law
2-Inlet Fork Mixer

=
R
2
R
1
+R
2
3-Inlet Fork Mixer

=
R
2
R
3
R
1
+R
2
+R
3
3-Inlet Ladder Mixer

=
R
2
R
3
R
2
R
3
+R
1
R
3
+R
1
R
2
+R
M
(R
1
+R
2
)
Table 4: General mixing rules for the 2- and 3-inlet Fork
topology, as well as the 3-input Ladder topology. The
nal dilution ratio,

, is then easily calculated by input-
ing resistor component values, and deviations, from the
library into the mixing law for the particular topology.
4.6 Statistical Determination of Operational Tolerance
The tolerance in channel side length is expected to cause variations in system operation, which
can be approximated using simple error analysis techniques. However, for increasingly complex
systems, by-hand error analysis computation becomes increasingly dicult as systems grow larger.
Here, the Monte Carlo analysis method is again employed (Figure 22) to closely approximate the
operating range of each topology by taking into consideration error propagation attributed to
fabrication. Monte Carlo analysis is applied to generate a simulated resistor bin, with each bin
48
having a respective distribution based on the initial distribution of channel side length measured
for the xy and z plane that was input into simulation. Again, Monte Carlo is applied by pseudo-
randomly selecting resistors from the simulated resistor bins and inputing them into the mixing
laws dened per topology. The resultant output is a distribution of volume fractions for each
topology, allowing users to better understand the range of the expected output concentration.
49
Length
Probability
Δxy
Length
Probability
Δz
Resistance
Probability
R
1

Resistance
Probability
R
2.5

Resistance
Probability
R
25

.
.
.
Volume
Fraction
Probability
2-1 Fork
R
hyd,rectangle
Mixing Laws
Simulated
Resistance
Measured Channel
Tolerance
Expected Topology
Operating Range
Volume
Fraction
Probability
3-1 Fork
Volume
Fraction
Probability
3-1 Ladder
Figure 22: A general schematic for the Monte Carlo approach to determine the

range for input branch 1 of each topology, which carries the stock NaCl solution to be
diluted. Initially, a large batch of resistance components is fabricated for which the cross
sectional channel side length of component channels is measured optically to create a
distribution for the xy and z printing planes. Values from the tolerance distributions
(normally distributed) are then pseudo-randomly selected as input parameters for the
hydraulic resistance equation respective to rectangular channels. A simulated resistor
bin is then created, such that each resistor has a respective distribution with tolerance
set as the deviation from the mean simulated resistance value. Again, resistance values
are pseudo-randomly drawn and input as parameters to the mixing laws, respective
to the circuit topology being investigated, resulting in a distribution of

values that
eectively stand as the expected operating range of the particular system.
50
4.7 Mixing Law Validation
The three circuit topologies investigated were constructed with dierent resistor combinations,
as described in Table 5 - 7, in order to validate the simulated operational tolerance set fourth
by Monte Carlo analysis (Fig. 23 - 25). Each topology was constructed such that inlet PEEK
tubing at every inlet was cut to 24.4mm in length, branch R
1
ran an NaCl solution (0.34M),
and the remaining inlet branches ran Milli-Q water. Systems were driven by connecting a syringe
on the output end, and manually withdrawing solutions slowly so as to insure a suciently low
Reynolds number. Once channels were primed, the syringe barrel was interchanged with a clean
barrel to collect roughly 0.5 - 1 mL of mixed solution. Osmolality of the collected NaCl mixture
was measured with an Osmomat 3000 osmometer, which was then used to determine the volume
fraction for each resistance combination, for all topologies (shown at the end of this section for
each topology). Once a system was constructed, data was collected by this manner in triplicate.
Figure 26 is arranged to show the simulated and experimental volume fraction data as deviations
from the designed volume fraction, better dened as the volume fraction calculated by inputting
perfect resistors (assuming no variation from designed CAD model) into the mixing laws, repeated
for every resistor combination attributed to each topology. The plots conrm the experimentally
tested volume fraction of NaCl in output solution fall within the simulated operating range for
each topology, and each resistor combination.
51
R
1
R
2

R
M

R
M

R
P

R
P

R
in

R
2
R
1

R
P

W
LJ
W
TJ

R
in

B C A
Figure 23: (A) Circuit diagram of a 2-input Fork topology
whereR
1
andR
2
are selectively chosen for desired output
mixtures. (B) Equivalent hydraulic circuit where only R
1
and R
2
the selected resistance components, contribute to
the mixing ratio and wire components, denoted as parasitic
resistance, have negligible eect on resistance. Note that
the current source symbol in (A) is represented by a syringe
withdrawing solution through the outlet in (C).
52
R
3

R
M

R
2
R
1

R
M

R
P

R
in

R
1
R
3

W
LJ

R
in
R
in

R
P

R
P

W
TJ

W
TJ

R
2

R
P

B
A
C
Figure 24: (A) Circuit diagram of a 3-input Fork topology
where R
1
, R
2
, and R
3
are selectively chosen for desired
output mixtures. (B) Equivalent hydraulic circuit where
onlyR
1
,R
2
, andR
3
, the selected resistance components,
contribute to the mixing ratio and wire components, de-
noted as parasitic resistance, have negligible eect on re-
sistance. Note that the current source symbol in (A) is
represented by a syringe withdrawing solution through the
outlet in (C).
53
R
M
R
M

R
1
R
2
R
3

R
in

R
in

R
in

R
M

W
LJ
W
TJ

R
2
R
1
R
3

R
P

R
P
R
P

R
M

W
TJ

R
P

A
B
C
Figure 25: (A) Circuit diagram of a 3-input Ladder topol-
ogy whereR
1
,R
2
, andR
3
are selectively chosen for desired
output mixtures. (B) Equivalent hydraulic circuit where
onlyR
1
,R
2
,R
3
, the selected resistance components, con-
tribute to the mixing ratio and wire components, denoted
as parasitic resistance, have negligible eect on resistance.
Note that the current source symbol in (A) is represented
by a syringe withdrawing solution through the outlet in
(C).
54
2-Inlet Fork Mixer
R
1
(G) R
2
(G) R
3
(G) Designed

Expected

Expected Error (%)
R
1
R
1
- 0.500 0.499 11.859 %
R
1
R
2:5
- 0.421 0.426 11.998 %
R
1
R
5
- 0.333 0.341 12.511 %
R
1
R
10
- 0.235 0.245 13.327 %
R
1
R
25
- 0.125 0.132 14.400 %
R
2:5
R
1
- 0.579 0.575 8.711 %
R
5
R
1
- 0.667 0.659 6.431 %
R
10
R
1
- 0.765 0.756 4.318 %
R
25
R
1
- 0.875 0.868 2.247 %
Table 5: Resistor combination for the 2-inlet Fork Topology, where the R
1
branch runs
a 0.34 M solution, and remaining branches run Milli-Q water, which are then withdrawn
into a syringe after being mixed in the system. The designed mixing ratio using as-
designed resistance values, Designed

, and expected mixing ratio from Monte Carlo
simulation that takes into account build error, Expected

, are calculated by using the
respective mixing law shown in Table 4. The Expected Error presented is two standard
deviation, 2, from the expected mixing ratio, the average value, initially determined
by manufacturing tolerance.
55
3-Inlet Fork Mixer
R
1
(G) R
2
(G) R
3
(G) Designed

Expected

Expected Error (%)
R
1
R
1
R
1
0.333 0.333 13.587 %
R
1
R
5
R
1
0.200 0.206 11.569 %
R
5
R
1
R
1
0.400 0.397 12.169 %
R
1
R
10
R
1
0.133 0.139 11.418%
R
10
R
1
R
1
0.433 0.430 11.930 %
R
1
R
2:5
R
1
0.366 0.365 12.636 %
R
2:5
R
1
R
1
0.266 0.270 12.754 %
Table 6: Resistor combination for the 3-inlet Fork Topology, where the R
1
branch runs
a 0.34 M solution, and remaining branches run Milli-Q water, which are then withdrawn
into a syringe after being mixed in the system. The designed mixing ratio using as-
designed resistance values, Designed

, and expected mixing ratio from Monte Carlo
simulation that takes into account build error, Expected

, are calculated by using the
respective mixing law shown in Table 4. The Expected Error presented is two standard
deviation, 2, from the expected mixing ratio, the average value, initially determined
by manufacturing tolerance.
56
3-Inlet Ladder Mixer
R
1
(G) R
2
(G) R
3
(G) Designed

Expected

Expected Error (%)
R
1
R
1
R
1
0.286 0.289 14.241 %
R
1
R
1
R
5
0.174 0.181 12.618 %
R
1
R
10
R
1
0.380 0.380 12.253 %
R
5
R
1
R
1
0.364 0.362 12.371 %
Table 7: Resistor combination for the 3-inlet Ladder Topology, where the R
1
branch
runs a 0.34 M solution, and remaining branches run Milli-Q water, which are then
withdrawn into a syringe after being mixed in the system. The designed mixing ratio
using as-designed resistance values, Designed

, and expected mixing ratio from Monte
Carlo simulation that takes into account build error, Expected

, are calculated by
using the respective mixing law shown in Table 4. The Expected Error presented is
two standard deviation, 2, from the expected mixing ratio, the average value, initially
determined by manufacturing tolerance.
57
Measured
Expected Upper Bound
Expected Lower Bound
Avg. Expected
c b a
Figure 26: Comparison of experimental mixing ratio deviation from designed mixing ratio in com-
parison to simulated mixing ratio deviation to designed mixing ratio. For each graph, the upper
and lower bound describe a 2 deviation from the expected mixing ratio, such that the shaded
region, eectively the simulated operating space, is established by the manufacture tolerance that
suggests 95% of the fabricated resistor elements fall within specication. Experimental data lie
within the simulated operational working space for the (A) 2-inlet Fork Topology, (B) 3-inlet Fork
topology, and (C) 3-inlet Ladder Topology.
4.7.1 Deriving Volume Fraction from 2-input Fork Experimentation
We consider the output concentration and its corresponding
ow rate to be equivalent to the
ow
rate and the concentration it carries across each branch, such that for the 2-input Fork topology
the following expression is formulated:
c
out
Q
3
=c
1
Q
1
+c
2
Q
2
(45)
which can be further reduced to:
c
out
=c
1

1
+c
2

2
(46)
Here, we assume the concentration c
2
to be equivalent to zero as, in all topologies, only the R
1
branch carries concentrate solution and all others run Milli-Q water. With this, an expression for
the volume fraction of branch R
1
can be written as:
58

1
=
c
out
c
1
(47)
where c
1
is the stock NaCl concentration.
4.7.2 Deriving Volume Fraction from 3-input Fork Experimentation
As in the case of the 2-Input Fork topology, the output concentration is again expressed as a
weighted sum of the concentration and the
ow rate that drives it:
c
out
Q
out
=c
1
Q
1
+c
2
Q
2
+c
3
Q
3
(48)
c
out
=c
1

1
+c
2

2
+c
3

3
(49)
In this case, c
2
and c
3
are equivalent to zero, so that

1
=
c
out
c
1
(50)
4.7.3 Deriving Volume Fraction from 3-input Ladder Experimentation
Lastly, the 3-input Ladder topology again reduces the volume fraction,

, to a ratio of the output
concentration to the stock concentration:
c
out
Q
out
=c
1
Q
1
+c
2
Q
2
+c
3
Q
3
(51)
c
out
=c
1

1
+c
2

2
+c
3

3
(52)
where c
2
and c
3
are equivalent to zero, so that

1
=
c
out
c
1
(53)
59
4.8 Discussion
In this chapter, a method to determine the performance of micro
uidic devices assembled using
MFIC's is demonstrated using a Monte Carlo technique in the context of mass manufacturing.
Error propagation generated through the fabrication process in relation to varying network as-
semblies was simulated using empirically gathered parameters with a mechanistic understanding
of the stereolithgraphy process. Experimentally, the devices demonstrated here were useful as a
hand-held laboratory tool, capable of rapid reconguration to create mixtures with a high level
of precision.
60
5 Engineered hydrophobicity of Modular Fluidic and Instru-
mentation Components
B. Thompson, C.T. Riche, N. Movsesian, K.C. Bhargava, M. Gupta, N. Malmstadt. (2016)
"Engineered hydrophobicity of discrete micro
uidic elements for double emulsion generation".
Micro
uidics and Nano
uidics, 20(5), pp.1-5.
B. Thompson, N. Movsesian, C. Cheng, P. Karandikar, M. Gupta, N. Malmstadt. (2018) "Mod-
ular Micro
uidics for Double Emulsion Generation". SUBMITTED.
5.1 Background and Motivation
Having the ability to control the channel surface wetting properties of micro
uidic devices is
advantageous for a broad set of engineering applications. The ability to customize this feature
enables creating and manipulating microdroplets [46, 47], generating single and double emul-
sions [48], and passive pumping through surface energy gradients [49]. Several pieces of work
have looked to enable designers with this feature through sol-gel coating of channel surfaces [50],
UV-induced surface polymerization [51], and laminar
ow patterning via surface adsorption and
reaction phenomena [52{54]. These methods are not well suited to address modication of a
bulk set of devices at the same time. Moreover, photolithographic approaches are limited to
photopolymers and constrained to the geometric complexity they can address. Flow-based pat-
terning methods can only create patterns that follow the routing of micro
uidic channels [55].
This is particularly limiting in devices fabricated on planar substrates that may require surfaces
with step-like wettability patterns, such as in the case of devices used to generate double emul-
sions [56, 57].
Double emulsions have a variety of applications in food chemistry [58, 59], personal cosmet-
ics [60, 61] and pharmaceutics [62]. In more recent biological applications, double emulsions
have been applied in cell-handling [63], culturing [64, 65], and analysis [66]. In general, advance-
ments in double emulsion creation through micro
uidic technology has been made largely in glass
61
microcapillary-based devices. The nature of these devices rely heavily on the skill of the assembler
to align components by hand, are inherently delicate, and are generally unsuitable for paralleliza-
tion or scalability [50, 67{69]. Alternatively, double emulsion devices that have been fabricated
from polydimethylsiloxane (PDMS) require a method of creating hydrophilic surfaces in order to
meet the appropriate
ow conditions. Past methods have modied channel surface chemistry via
oxidation [56], labor intensive photografting [70], and modication via sol-gel methods [50, 57].
These methods, while ultimately eective towards producing hydrophobic gradients, are inher-
ently prohibitive to modifying large volumes of devices as their involved nature slows down the
manufacturing process at the device level.
Previously, initiated chemical vapor deposition (iCVD) was demonstrated as a useful process for
coating discrete micro
uidic elements fabricated using stereolithography (SLA) [45] as well as
traditional micro
uidic devices fabricated via soft lithography [71]. Vapor-phase coating has lim-
itations with post-assembly process work
ows due to the distance that vapor can diuse into
microchannels; this inherently limits the size and complexity of devices to which this technique
can be applied. Vapor-phase deposition is well suited to coat MFIC's, which contain short chan-
nel length, ensuring successful diusions and continuous coating through the length of channels.
Once modied, MFIC's are used to assemble systems with complex geometries and long track
length. By creating a library of MFIC's with dierent surface chemistries, we introduce surface
functionality as a selectable control parameter for device construction. In this chapter, surface
chemistry modied MFIC's are assembled to create a micro
uidic system with abrupt changes in
surface wettability. Discrete MFIC's are inherently easy to assemble by hand on the bench-top,
are reusable, and not limited to a single application after assembly. This library of MFIC's is
applied towards assembling a circuit capable of creating water-oil-water (W/O/W) double emul-
sions.
In earlier work, the hydrophobicity of channel surfaces was modied to fabricate surfaces with
lower surface energy. Figure 28 is an example of an MFIC that was coated via iCVD with
poly(1H,1H,2H,2H-per
uorodecyl acrylate-co-ethylene glycol diacrylate), or poly(PFDA-co-EGDA),
which eectively created a low surface energy lm that increased hydrophobicity. The contact
angle of a water droplet surrounded by oil in the channel increased from 67.9

to 138.3

. Al-
ternatively, 2-hydroxyethyl methacrylate (HEMA) monomer could be cross linked with EGDA to
62
deposit a thin lm layer of poly(HEMA-co-EGDA) that would increase hydrophilicity of channel
surfaces. Modularized micro
uidic components enable the coating of batches of components at a
time, highly increasing the throughput of surface modied component channels. Here, a system
of interchangeable discrete components with variable surface modications for double emulsions
generation is introduced.
5.2 Device Overview
The MFIC's described here were fabricated using SLA services from Proto Labs Inc., and designed
to feature 500 m channel side lengths. The library of the modied micro
uidic elements is
provided in Table 29. Several elements were coated to have hydrophobic and hydrophilic surfaces
using iCVD. A new connector and t-junction component are here introduced, fabricated to allow
the facile assembly of junctions requiring step-like changes in hydrophobicity (Figure 30). Figure
30 demonstrates the ability to create a junction where the wettability at the interface between a
straight pass channel (orange component) and a junction channel (purple component) changes
abruptly. These MFIC's are assembled to create a system that produces water-oil-water (W/O/W)
emulsions (Figure 31). Figure 31a is a schematic describing the underlying approach taken here
to generate double emulsions. In this system, an inner aqueous phase
ow rate,Q
i
, is dispersed in
a middle oil phase
ow rate,Q
m
, generating a monodisperse emulsion. Droplets in this emulsion
are then dispersed into an outer aqueous phase
ow rate,Q
o
. By controlling the hydrophobicity of
each channel segment, stable double emulsions can be generated by preventing inner and middle
phases from wetting the channel walls, Figure 31b is a CAD diagram of the double T-junction
device where MFIC's colored purple have been modied to express a high-surface energy channel
surface, components colored orange have been modied to express a low-surface energy channel
surface, and components colored blue have not been altered. Figure 31c is a photograph of the
actual device, assembled to specication as per Figure 31b.
5.3 Surface Modication via iCVD
MFIC surface chemistry modications were accomplished through initiated chemical vapor depo-
sition (iCVD). iCVD is a solventless vapor-phase polymerization technique that can be used to
63
apply thin lms on substrate surfaces [71{73]. This process eliminates issues with solvent compat-
ibility and monomer solubility, which are common to liquid-phase polymerization [74,75]. Brie
y,
components were loaded into a custom iCVD reactor (GVD Corporation) on a temperature-
controlled stage held at 30

C and oriented such that MFIC channels were open to gas diusion
at all entry ports (Figure 27). Monomer and initiator vapors were introduced to the reaction
chamber at 80 mTorr. An array of nichrome wire suspended above the micro
uidic components
was heated to 200

C. The initiator, di-tert-butyl peroxide (DTBP 98%, Sigma), is introduced
into the reactor and subsequently thermally decomposed into free radicals which then react
with monomer adsorbed to the surfaces of the components to form a continuous polymer lm.
For low-surface-energy coatings, 1H,1H,2H,2H-per
uorodecyl acrylate (PFDA 97%, Sigma) was
used as a monomer. For high-surface-energy coatings, 2-hydroxyethyl methacrylate (HEMA 97%,
Sigma) was used as a monomer. In both cases ethylene glycol diacrylate (EGDA 90%, Sigma)
was included to cross-link the polymer lms.
Monomer
Initiator
Vacuum Line
Quartz Plate
Cooled Stage
Filament Array
Figure 27: A cross-sectional view of the iCVD reaction
chamber with Modular Fluidic and Instrumentation Com-
ponents placed on the cooling stage.
A Ram-Hart Model 290-F1 goniometer was used to determine the contact angle of non-coated
elements as well as hydrophobic and hydrophilic surface coated components. Measurements were
completed on element surfaces of treated material and on an uncoated slab of SLA material.
Contact angle images are provided in Figure (Fig. 34). Non-coated elements had a contact
angle of 80 3

. MFIC's coated with a continuous poly(PFDA-co-EGDA) lm, the hydrophobic
64
polymer, were determined to exhibit contact angle of 125 5

. MFIC's coated with a continuous
poly(HEMA-co-EGDA), the hydrophilic polymer, exhibited a contact angle of 56 7

.
5.4 Experimental Design
In all experiments, 10% wt/wt glycerol (99.5%, Sigma) in Milli-Q water was used as the inner
aqueous phase and 5% wt/wt poly(vinyl alcohol) (PVA MW 13,000-23,000, Sigma) in Milli-Q
water was used as an outer aqueous phase. Light mineral oil (Avantor) with 3% wt/wt SPAN-80
surfactant was used as a middle phase. Syringe pumps (Harvard Apparatus 22) were used to drive

uids through the circuit shown in Figure (Figure 31). A sample micrograph of double emulsions
generated by the modular device is shown in Figure (Figure 32). Two sets of experiments were
conducted to characterize the eects of
ow rate variation on double emulsion size distribution.
In the rst set of experiments, the inner and outer phase
ow rate were maintained constant
at 0.9 mLhr

1 and 1.5 mLhr

1, respectively, while the
ow rate of the middle phase was
modulated between 1.0 and 2.3 mLhr

1. This allows control over the volume of the oil shell
and water core. In the second set of experiments, the inner and middle phase
ow rates were
held constant at 0.9 mLhr

1 and 1.9 mLhr

1, respectively, while the
ow rate of the outer
phase was varied between 1.0 and 2.25 mLhr

1. These
ow rates are well within the operable
bounds towards preventing leakage problems, as established by previous assessment of pressure
limitations. This eectively controls the volume of the oil shell. Figure (Figure 33) shows the
change in diameter of oil shells (Figure 33a) and water cores (Figure 33b) versus Q
O
/Q
SUM
, where Q
O
is the outer
ow rate and Q
SUM
is the sum of the aqueous and carrier oil phase
Q
i
+Q
m
. In both experiments, the oil shell diameter tends to decrease as the ratio between
Q
O
and Q
SUM
increases since the volume of oil entering the second T-junction is reduced with
increasing outer phase
ow rates or lower middle phase
ow rates. The water core diameter did
not largely vary with middle or outer
ow rate modulation. An increase in middle
ow rate tends
to slightly decrease the water core diameter, however this decrease is limited by the geometry of
the micro
uidic channels, which have a designed channel side length of 500m. Outer
ow rate
modulation did not signicantly alter the water core diameter as Q
O
shears the oil phase, not
the inner aqueous phase that forms into the water core upstream of the device. For all cases, the
65
assembly of discrete elements produced double emulsions with a coecient of variation (CV) of
8.5% or less for oil shell and water core diameters, as shown in Table 35.
Figure 28: Contact angle is here measured between a wa-
ter droplet in two components: Water droplet surrounded
by an oil carrier stream in an (A) uncoated and (B) coated
component. Surface modication eectively shows the in-
creased hydrophobicity of channel walls via initiated chem-
ical vapor deposition.
66
Figure 29: MFICs used to assemble double emulsion device. A modied connector and
corresponding T-junction were fabricated to allow the oil carrier stream entering the
second T-junction to feel the eects of hydrophilic surfaces only at the junction and
not at the short straight pass present in the unmodied T-junction element.
67
Figure 30: A specially designed connector with extended male pin (orange element) that
inserts into a modied T-junction (purple element). This type of connection allows for a
step-like hydrophobic gradient to occur between elements with dierent surface energies.
68
Figure 31: (A) General schematic of a double emulsion device such thatQ
O
is a carrier
oil stream that comprises the middle layer of a double emulsion, Q
W;IL
comprises the
inner aqueous core of a double emulsion andQ
W;OL
is an outer aqueous carrier stream.
(B) An equivalent CAD diagram of the double emulsion device where droplets in the
carrier oil stream
ow through surface modied channels with increased hydrophobicity
(orange components), which then enter hydrophilic channels as the outer carrier phase
shears the incoming droplet containing oil stream (purple components). (C) Actual
assembly to generate double emulsions.
69
Figure 32: Double emulsions generated from modular micro
uidic device with inner,
middle, and outer
ow rates set to 0.9, 1.5, and 1.5 mLhr

1, respectively. Inset is
a bright eld image with inner, middle, and outer
ow rates set at 0.9, 1.9, and 2.25
mLhr

1, respectively.
70
Figure 33: Size distribution of (a) oil shell and (b) water core diameter of double-
emulsion droplets vs. the ratio of outer phase
ow rate (Q
O
) to the sum of the inner
and middle phase
ow rates (Q
SUM
). Two experiments were conducted such that
outer phase
ow rate (red points) and the middle phase
ow rate (blue points) were
modulated.
Figure 34: Contact angle images taken with a Ram-Hart Model 290-F1 goniometer
for (A) uncoated element, (B) hydrophobic surface coating and (C) hydrophilic surface
coating. Measurements were taken on the surface of an uncoated slab of SLA material
for (A) and on element surfaces for surface modied components, (B) and (C).
71
Figure 35: Coecient of Variance (CV) for oil shell and water core diameters experienc-
ing middle phase and outer phase
ow rate modulation determined by measuring 50-100
droplets using NIH ImageJ image processing software. Inner and outer phase
ow rate
was held constant at 0.9 and 1.5mLhr

1 for middle phase
ow rate modulation. Inner
and middle
ow rate was held constant at 0.9 and 1.9 mLhr

1 for outer phase
ow
rate modulation.
5.5 Discussion
Modular Fluidic and Instrumentation Components enable the development of complex, three
dimensional micro
uidic circuit designs. Control over their surface chemistry enhances their ver-
satility in the management of multi-phase
ows, where multiple surface wetting properties may
be critical for optimal system performance. A modular approach also inherently lends itself to
creating circuit features that are not easily fabricated through conventional micromachining ap-
proaches, such as the abrupt surface energy gradients demonstrated in this report. iCVD coating
is uniquely suitable for coating discrete micro
uidic elements in that their inherent compartmen-
talization allows for vapor-phase reagents to uniformly adsorb to channel surfaces. Furthermore,
the materials employed in additive manufacturing (e.g. stereolithography) are typically propri-
etary formulations, limiting their direct application in systems where surface chemistry control
is important (such as in the control of biological solutions). Application of iCVD coating to
proprietary materials alleviates this constraint, and is able to provide a sucient level of control
at the element manufacturing level. This micro
uidic platform can therefore enable engineers to
rapidly construct complex circuits with patterned surface chemistry in three dimensions within
days, without the need of complex tooling for sophistical processing.
72
6 Spectrophotometry in modular micro
uidic architectures
B. Thompson, K.C. Bhargava, B. Pan, B.T. Samuelsen, N. Malmstadt. (2018) "Spectrophotom-
etry in modular micro
uidic architectures". SUBMITTED.
6.1 Introduction
In the developing world and many rural areas, patients and caregivers often do not have the re-
sources needed to provide data-driven, personalized medical treatment, contributing to increased
mortality and lower life expectancy [76]. For example, many diagnostic testing protocols require
specialized instruments, laboratory facilities, and highly trained operators. Automation has done
little to improve access to these tests: sample-to-answer robotic systems that produce quantita-
tive biomarker data in central and hospital labs are not inherently portable or cost eective to
manufacture at scale [77]. Non-robotic, more compact systems traditionally require micro
uidic
disposables that are limited in their application scope and dicult to cost-eectively manufacture
at scale. Hence, in low-resource environments there is an unmet need for compact, turn-key
solutions that deliver quantitative biomarker data without the use of specialized disposables.
We previously introduced a system of 3-D printed, discrete micro
uidic elements referred to as
Modular Fluidic and Instrumentation Components (MFICs) [78{81]. Brie
y, the MFIC technol-
ogy is designed to automate complex wet lab procedures in a micro
uidic environment while
alleviating the scaling challenges presented by legacy micro
uidic manufacturing techniques. 3-D
printing enables each MFIC with a wide array of
uid handling, sensing, and actuation functions.
MFIC's can be assembled into multi-plane, functionally sophisticated networks that handle an
unrestricted number of reagents and chemical processes. In this report, we demonstrate the
use of MFIC technology to realize a turn-key, compact system for performing quantitative en-
zyme biomarker tests that would normally require a central or hospital laboratory to execute.
This Multi-Stage Stopped Flow (MSSF) system does not require specialized disposables, requires
minimal skill to operate, uses minimal reagent and sample material, and produces data quality
equivalent to that produced by legacy microtiter-plate based techniques.
73
6.1.1 Multi-stage Stopped-Flow System (MSSF) for Quantitative Enzyme Biomarker
Measurements
Stopped
ow is a laboratory technique commonly used to quantify enzyme activity. Modern
stopped-
ow technology can generate high-resolution reaction kinetics data for a wide variety
of chemical reactions relevant to life science [82]. For example, stopped
ow has been used
to characterize protein-protein interactions9 and protein folding [83]. A stopped-
ow apparatus
traditionally comprises a number of syringe pumps, a mixer, an optical detection system, and a
data acquisition system. In the case of commercial instruments, reactants are combined in the
mixer at a very high
ow rate (e.g. a Reynolds Number greater than 500) [84]. The
ow is
then abruptly stopped and the development of an optically-detectable reaction product in the
mixture solution is monitored over time. Traditional stopped-
ow systems require a relatively
high volume of potentially costly or rare sample and reagent. Micro
uidics based stopped-
ow
devices alleviate this issue by drastically lowering the total amount of
uid resident in the
ow
path of the apparatus [85]. Consequently, alternative mixing and detection strategies must be
used to ensure quality results [86].
The MSSF system introduced in this report is inspired by the stopped-
ow technique: two sequen-
tial stopped-
ow reactions are executed within a micro
uidic network. The network comprises
MFICs that have been selected based on their resident volume,
ow-rate-dependent mixing e-
ciency, and detection functionality. Combining MFIC technology and advanced controls methods
in turn eliminates the traditional limitations of stopped-
ow systems, such as high reagent volume
requirements, limited application scope, and large instrument footprints. This new MSSF system
is demonstrated in an analytical application: quantifying an enzyme biomarker of bladder cancer.
6.1.2 Spectrophotometer MFIC
Detection in the MSSF system is facilitated by a novel MFIC for spectrophotometry. This
uidic
element is implemented in the MSSF system to enable real-time detection of reaction products.
Spectrophotometry is a critical method of detection for many bioassays. Several research groups
have explored integrating spectrophotometry into microfabricated lab-chip devices. Strategies
include fabricating o-chip optical detection systems [87] or optical waveguide structures [88] to
74
direct light in and out of monolithic micro
uidic devices. Detection sensitivity can be limited in
these devices. This is due in part to the shortened light path intrinsic to micro
uidic channels
fabricated in a single plane using microfabrication techniques [87]. Other approaches overcome
this hurdle by using high-powered microscope and camera systems [89]. This is at a substantial
cost to portability and manufacturing scalability.
The integration of optical waveguides coplanar to micro
uidic channels enables the measurement
of optical density in a manner directly analogous to that of a classic spectrophotometer and
without the use of a microscope. However, this adds substantially to the complexity and cost
associated with microfabricating such multi-material devices. In addition, the need for external
free-space optics to couple light emitters and detectors with the waveguides also presents a sub-
stantial manufacturing challenge and limits portability. The spectrophotometer MFIC presented
in this report does not require microscopes or other complex optics and can be used at an arbitrary
number of nodes in a micro
uidic network. In addition, it is designed to be insensitive to ambi-
ent light levels, readily integrate into the MFIC interconnect system, and directly interface with
industry-standard optical ber cables. Moreover, limitations on sensitivity due to short light beam
path are removed with out-of-plane fabrication enabled by additive manufacturing techniques.
6.1.3 Urinary Enzyme Biomarker for Bladder Cancer
Cancers are particularly dicult to manage in low-resource environments due to the need for con-
tinuous monitoring and high levels of intervention. For example, the standard of care for bladder
cancers relies heavily on cystoscopy, a procedure that intrinsically requires substantial clinical re-
sources. Furthermore, cystoscopies can cause incontinence in patients, increase their chances of
an infection, require substantial recovery time, and can cause anxiety [90,91]. Long-term survival
in face of high recurrence rates requires that patients be routinely monitored every three to six
months [92]. In the US, where resources are abundant, these procedures cost the average patient
between $96,000 to $187,000 through the course of their lifetime [92]. Less invasive approaches
such as voided urine cytology and quantitative measurement of urinary biomarkers present an
opportunity to improve patient quality of life while reducing dependence on cystoscopy.
Over the last decade hyaluronic acid (HA) has emerged as a compelling biomarker for in prostate
[93], breast [94], and bladder cancer [95]. HA is a glycosaminoglycan that promotes cell adhesion
75
and migration. HA is known to promote angiogenesis [95] when broken down to small fragments.
Hence, hyaluronidase (HAase), an enzyme that degrades HA, has been shown to be elevated in
urine when high-grade bladder (G2/G3) tumors are present. Moreover, researchers have shown
the use of HAase as a highly specic and sensitive diagnostic urinary biomarker through a com-
plex ELISA-like assay [96]. Assaying HAase concentrations therefore has the potential to reduce
clinical dependence on invasive cystoscopies.
HAase activity has been detected through a homogenous assay that relies on a plasmonic nanopar-
ticles that interact electrostatically with HA [97]. This approach enables the monitoring of blad-
der cancers non-invasively without the long turnaround times and resource costs associated with
ELISA-like assays. However, adoption of this technique is still hindered due to the dependency
on central or hospital lab settings, where disposables, specialized analytical tools, and technical
operators are accessible. In this report, we demonstrate the use of the MSSF system to automate
a plasmonic nanoparticle assay for HAase.
6.2 Results and Discussion
6.2.1 MSSF System Principle of Design and Operation
Gold nanoparticle (AuNP) aggregation assays can be used to monitor enzymatic activity through
a spectral shift in their peak absorbance induced by their aggregation. In a negative test case
(i.e. no enzyme present) aggregation of AuNPs with a certain surface charge is induced by a
biomolecular binder with an opposite charge through electrostatic interactions. The increased
proximity between AuNPs will cause an observable color change of the AuNP solution. Often,
solutions containing AuNPs with diameters on the order of tens of nanometers will visibly appear
to change from red to blue. In a positive test case, presence of an enzyme will degrade the
biomolecular binder, reducing its ability to aggregate AuNPs. While eective, this multi-step
assay is challenging in low-resource settings. There is a strong dependency on precision
uid
manipulation steps, controlled incubation periods, instruments for measuring optical absorbance
rapidly, and other specialized resources and skills.
The multi-stage stopped
ow (MSSF) system introduced here comprises of multiple injection
port, mixer, transport, and spectrophotometer MFICs (Figure 36). To initiate an assay, a user
76
installs reagent-loaded syringes into syringe pumps and executes a program that automates the
assay protocol and collects sensor data (Figure 37). Referring to simplied network diagram in
Figure 36B, the system infuses an aliquot of enzyme and substrate solutions simultaneously at
preset mixing ratios through Inlets A1 and A2. The two solutions mix through Mixer A1 and are
stored in Mixer A2, where they incubate. The resulting partially or completely reacted solution
of enzyme and substrate is henceforth referred to as Solution B1, which develops over the course
of the incubation period. Next, aliquots of Solution B1 and a solution of AuNPs originating
from Inlet B2 are injected through Mixers B1 and B2. The resulting mixture, Solution C, is
incubated in the detector, a spectrophotometer MFIC. The aggregation of AuNPs to substrate
is monitored in Solution C using the detector. All mixers were selected based on their optimal
operating
owrate to guarantee rapid results [98]. A generalized system diagram and
ow of
control diagram are presented in Figure 37 and Figure 38 respectively.
A B
Figure 36: (A) A computer aided design (CAD) model of the multi-stage stopped
ow
(MSSF) system. Inlets and outlet are indicated by black arrows. (B) A simplied
network diagram of the MSSF system. Mixer A1, B1, and B2 are used to combine and
mix solutions inject at Inlets A1, A2, and B2, as well as from node B1. Mixer A2 serve
as a reservoir incubation as well as a mixer for solutions infused from Inlets A1 and A2.
77
Spectrometer
Optical Fiber
Waste
Syringe Pump 3
Syringe Pump 2
Syringe Pump 1
Lamp
MSSF Reactor
Figure 37: MSSF system diagram highlighting system syringe pumps and spectrometer
that are computer controlled. At the center of the system is the MSSF reactor, that
dumps solution to waste after assay completion.
Activate System
Inject Fluid A1
Inject Fluid A2
Pause to
Incubate
Detector
Filled?
No
Fluid B1'
Yes
Fluid B1
Inject Fluid A1
Inject Fluid A2
Inject Fluid B2
Fluid C’ in
Detector?
Pause to
Incubate
Yes
Fluid C’
No
Detect
Conversion
of C’ to C
Fluid C
System
Deactivates
Detect
Conversion
of B1’ to B1
Figure 38: A
ow of control diagram representing the MSSF system automation pro-
tocol.
78
6.2.2 Design of a Spectrophotometer MFIC
At the core of the MFFS system is a novel spectrophotometer MFIC that enables real-time
quantication of reaction products in Solution C (Figure 39). As with other MFIC's, industrial-
grade stereolithography was used to fabricate the component. This enables fabrication of three-
dimensionally complex micro
uidic channels and incorporation of o-the-shelf mechanical and
optical components. In this manner, a wide variety of detectors, emitters, and lters can be used
to outt this MFIC to automate many biochemical protocols. Here, self-aligned, broad-spectrum,
ber-based light management is enabled by incorporating precision manufactured sapphire win-
dows and SMA ber ports (Figure 39A).
z
x
y
y
x
z
y
Light Path Light Blocks
A B C
Figure 39: A computer aided design (CAD) model of the spectrophotometer MFIC. (A)
Exploded view with the sapphire window and SMA adapter shown. Inset is an iso view
of the fully assembled component. (B) A cross-section of the assembly in the XY plane
detailing the aperture channel section through which light transits between emitter and
detector bers. (C) A cross-section of the assembly in the XZ plane showing the snaked
inlet and outlet channels that serve to block ambient light from entering the aperture
section of the channel.
This component is fabricated using an opaque photoresin to eliminate stray light from entering
the detection ber that may contribute to background signal, degrading detection nonlinearity.
More specically, stereolithography was leveraged to create channel sections in orthogonal planes
that act as an optical aperture or contain features that block to ambient light (Figure 39B, C).
The opaque walls of the aperture section act to absorb much of the non-parallel rays of light
79
exiting the emitter ber and transiting the sample resident to the channel. The S-shaped features
near the inlet and outlet of the MFIC body eliminate ambient light that may enter the aperture
section due to imperfect light-absorbing capabilities of the photoresin.
The channels of the spectrophotometer MFIC are designed with a 642.5 m channel side length
like many other components in the standardized MFIC library (Figure 40). Channel corners and
bends are rounded where the channel is sealed to the windows to reduce the immobilization of
small gas bubbles commonly present in liquid reagents. Facile interconnection with previously
developed MFIC's is accomplished by maintaining the interference-t based interconnect system
previously described in [78{81,98]. The total
uidic path length for the spectrophotometer MFIC
is 36.7 mm and the aperture channel length is 8.0 mm. These parameters yield a hydraulic
resistance of 6.1192 GPa-s-m
3
and residence volume of 15.15 L.
80
MFIC Component
Hydraulic Resistance
(GPa-s-m
3
)
Connector 1
Mixer (5G) 5
Mixer (10G) 10
Spectrophotometer 6.1192
Port 1
T-Junction 0.5
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10: 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10: 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10: 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10: 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10: 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10: 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10: 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10: 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10: 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
6:1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
6:1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
6:1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10 : 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10 : 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10 : 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10 : 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10 : 1
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
A A
B B
C C
D D
SHEET 1 OF 1
DRAWN
CHECKED
QA
MFG
APPROVED
KrisnaBhargava 4/16/2018
DWG NO
TITLE
SIZE
D
SCALE
REV
10 : 1
Figure 40: Library of the MFICs used in the MSSF system. Each MFICs function, wire
frame, and associated hydraulic resistance is indicated.
6.2.3 MSSF Benchmarking on HAase Assay
HAase was assayed by quantifying the spectral shift in absorbance caused by the charge inter-
action between cationic plasmonic AuNPs and repeating polyanionic HA units (Figure 41) [97].
Unadulterated HA causes AuNPs to bind to one another, increasing their proximity and causing
81
the peak plasmon absorbance of solution to shift towards longer wavelengths. This does not oc-
cur when HA is previously incubated with a sample containing HAase, enabling the detection of
HAase in that sample. The HAase cleaves the HA into fragments that reduces, if not eliminates,
the ability of HA to aggregate the AuNPs. The performance of the MSSF system was bench-
marked by directly comparing the limit of detection, turnaround time, and operator involvement
against execution of this assay by conventional, by-hand, microtiter plate-based protocols.
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
- -
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
HA
N
Br
-
+
HAase
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
+ +
+
+
+
+
+
+
+
+
+
+
+ + +
+
Figure 41: The principle of operation used to determine HAase activity though the
electrostatic interactions between HA and AuNPs. In the absence of HAase, repeating
units of polyanionic HA will bind cationic CTAB stabilized AuNPs through charge in-
teractions, leading to a shift in their plasmon absorbance. In the presence of HAase,
HA is cleaved into smaller fragments that do not cause AuNPs to aggregate.
The incubation periods, dilution ratios, and reagent concentrations used during experiments were
optimized in the MSSF system and then translated to by-hand protocols. Standard working con-
centrations of HA25 and AuNP solutions were adopted in order to determine controls parameters
and system performance. The optimal incubation period necessary to induce a shift in absorbance
due to HA and AuNP interaction was determined experimentally. Solutions of HA were loaded
at Inlets A1 and A2 and a solution of AuNPs was loaded at Inlet B2 of the MSSF reactor (Figure
36B). The protocol described in Figure 38 was run with no incubation periods, and the spectral
absorbance of Solution C was monitored over time (Figure 42A). The absorbance peak was ob-
82
served to shift from approximately 530 nm to just under 580 nm in the rst ten minutes, indicating
a change in the plasmon resonance of the AuNPs as they aggregate (Figure 42B). After twenty
minutes, no further shift was observed. However, the intensity of the absorbance peak decreased
rapidly after 10 minutes, indicating precipitation of AuNPs out of solution (Figure 42A). Hence,
a standard incubation time of ten minutes was adopted for Solution B and AuNP solution in all
subsequent experiments.
83
450 500 550 600 650 700
0.0
0.2
0.4
0.6
0.8
Wavelength (nm)
Absorbance
AuNP
1 min
3 min
5 min
7 min
10 min
20 min
30 min
45 min
0 10 20 30 40 50
540
560
580
600
Time (min)
Peak Absorbance Wavelength (nm)
B
A
Figure 42: (A) A time series study demonstrating the spectral absorbance for unadul-
terated HA and AuNP solutions mixed and incubated in the MSSF system. (B) The
peak absorbance wavelength as a function of incubation time.
84
HAase dose-response curves were generated to compare the limit of detection of the MSSF system
and plate-based, by-hand legacy techniques (Figure 43). The protocol described in Figure 38
was run with varying dilutions of HAase loaded at Inlet A1 while maintaining standard HA and
AuNP solutions at Inlets A2 and B2 respectively (Figure 36B). The ratio of absorbance at 580 nm
and 530 nm was used to construct the curve, indicating the extent to which a plasmonic shift has
occurred. The expectation is that samples containing low concentrations of HAase would present
substantial and maximum absorbance near this wavelength (A580), while samples containing
a high concentration of HAase would demonstrate peak absorbance near the peak absorbance
of unadulterated AuNPs at 530 nm. A four-parameter logistic t was used to determine the
performance of the system, indicating that the MSSF system showed nearly equivalent detection
limits as by-hand techniques (Figure 44). The system demonstrates higher resolution with greater
signal changes produced by varying enzyme concentration, as compared to by-hand techniques.
Plate based
MSSF system
4PL
4PL
0 10
0
10
1
10
2
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
1.3
log[HAase] (U/mL)
A580/A530
Figure 43: Dose-response curves and ts for HAase activity comparing by-hand tech-
nique (black) and the MSSF system (blue). Error bars represent one standard deviation
for the ratio between the absorbance averages of 570-590nm to 520-540nm.
85
Performance Metric By-hand Technique MSSF System
Limit of detection (LOD) 5 U/mL 6 U/mL
IC
50
13.52 U/mL 9.64 U/mL
Hill Slope -3.05 -7.56
Detectable range 10 – 15 U/mL 8.5 – 12 U/mL
Total reagent
consumption
200 L 37.5 L
Turnaround time ~ 2.5 h ~ 1.3 h
User steps 10 2
Figure 44: Comparison of performance metrics for HAase activity determined in a by-
hand, plate-based assay and by the MSSF system.
6.3 Materials and Methods
6.3.1 Fabrication of MFICs
The MFIC library utilized here was fabricated using stereolithography (SLA) services from Proto
Labs Inc.. MFIC's used here include 5G and 10G mixers, that have been previously described
and characterized [98], interconnects, 1/4-28 interfaces and a spectrophotometer component
(Figure 40). The spectrophotometer MFIC was fabricated using Huntsman RenShapeSL 7820
SLA resin, a low-viscosity resin that generates black components. All components were designed
using Autodesk Inventor 2017. The open face channel on the spectrophotometer MFIC is sealed
with circular sapphire windows that are set onto designed slots (Esco Optics, Oak Ridge, NJ)
and bonded to the component surface with a thin layer of Loctite 5 min epoxy (Westlake, OH).
Sapphire windows were selected due to their chemical inertness and favorable optical character-
istics, allowing a broad transmission range (UV to IR). Male ended SMA adapters (Thor Labs,
Newton, NJ) were stacked on top of each of the sapphire windows and bonded to the designed
86
component lip with a thin layer of Loctite 5 min epoxy (Figure 36).
6.3.2 MSSF System Architecture and Maintenance
A system diagram of the MSSF system is presented in Figure S1. Key components include the
MSSF reactor, a spectrometer unit (Flame-S-UV-VIS-ES, Ocean Optics, Largo, FL), a tungsten
halogen light source (DH Mini, Ocean Optics, Largo, FL), several computer-controlled syringe
pumps (KDS Legato 950 OEM), and optical bers (QP Series, Ocean Optics, Largo, FL). The
pumps and spectrometer are controlled using custom software written in Python 3 using SciPy
and Pyserial packages.
Prior to all tests, the spectrometer is calibrated. Water is loaded into the network until the
channel within the spectrophotometer MFIC is lled. A bisection search routine is implemented
to locate the spectrometer integration time at which 90% of the maximum output signal of the
spectrometer is achieved. This integration time is then stored and used for all detection steps
during subsequent operation of the MSSF system.
Channel surfaces of the MSSF reactor are blocked prior to assays using Pierce Protein-Free (PBS)
Blocking Buer (ThermoFisher Scientic, Waltham, MA). This is accomplished by setting pumps
to withdrawal mode, loading blocking buer in the outlet of the network, and withdrawing from
all inlets at 40 L min
1
for a total volume of 600 L. Channel surfaces are cleaned after assays by
running 50mM phosphoric acid, 5% Contrad solution, and Milli-Q water sequentially through the
them in a similar manner.
6.3.3 HAase Activity Assay Parameters
All aqueous solutions in this report are prepared with puried water (18.2Mohm resistivity; Milli-
Q Advantge A10, Millipore, Burlington, MA). Hyaluronic acid sodium salt from bovine vitreous
humor (HA) and hyaluronidase from bovine testes (HAase) are obtained from Sigma Co. (St.
Louis, MO). For all experiments, HA is diluted in 1x PBS (Mediatech, Inc., Manassas, VA) and
HAase is diluted in a 0.1M MES buer solution pH 6.0. Accurate Spherical Gold Nanoparticles,
35nm diameter, stabilized with cetyl trimethyl ammonium bromide (CTAB) were purchased from
Nanopartz (Loveland, CO). HAase activity measurements for the by-hand assays are produced
using a BioTek Synergy H1 Hybrid Multi-Mode plate reader (Winooski, VT). To determine the
87
optimal incubation period necessary to induce a plasmon shift in absorbance, solutions of HA were
loaded at Inlets A1 and A2 and a solution of AuNPs was loaded at Inlet B2 of the MSSF reactor
(Figure 1B). HA was infused through each inlet for a total
ow rate of 50 L min-1 and AuNPs
were infused at 333 L min
1
. Mixed solution was incubated in the detector and measurements
were taken between 1 and 45 minutes (Figure 42).
The by-hand assay protocol is summarized by Figure 45. First, 75uL of a 3.75 g mL
1
HA
solution is mixed with 25uL of an HA sample in a 96 well plate. Solution is gently mixed and
incubated for 1h, which we refer to as Solution A. After incubation, 15 L of Solution A is mixed
with 100 L of AuNPs. Solution is gently mixed and left to incubate for 10 minutes. Once
incubated, we refer to this solution as Solution B. Spectral measurement is then performed in a
plate reader. To determine HAase activity with the MSSF system, similar reagent concentrations
are loaded onto the systems syringe pumps. The automation software then rapidly infuses HA
and HAase through Inlets A1 and A2, as per Figure 1, maintaining a 3:1 ratio at a total
ow rate
of 383 L min
1
. After mixing and incubation, AuNPs are infused at 333 L min
1
and Solution
A is infused at 50 L min-1, maintain a 6.66:1 AuNP to Solution A ratio. No more than 12 L
of Solution A are dispensed into the 2nd T-junction with AuNPs. After mixing and incubation,
spectral measurements are taken in the MFIC spectrophotometer. The peak absorbance ratio
(A580/A530) as a function of HAase concentration for both assay formats (Figure 43) is tted
to a 4-parameter logistic (4PL) model to estimate IC50 values using Prism Version 7 (Graphpad
Software).
88
Pipette
enzyme and
substrate
into a well
Gently mix
solution
Incubate 1 h
(Solution A)
Transfer
solution A to
a new well
and add
AuNPs
Gently mix
solution
Incubate 10
min
(Solution B)
Transfer
plate to
Plate Reader
Wait for
results
Export data ~ 2.5 hours
Load
reagents
onto
syringes
Run
software
protocol
Wait for
results
1.3 hours
Plate-based assay
MSSF System
Figure 45: Comparison of operator work
ow for micro-titer plate-based, by hand tech-
nique (top) and use of the MSSF System (bottom). Turnaround time is indicated in
the circular arrow.
6.4 Discussion
The MSSF system exhibits the potential to overcome the accessibility and resource limitations
inherent to by-hand, microtiter plate-based techniques for executing plasmonic nanoparticle ag-
gregation assays. A summary comparison of performance metrics is presented in Figure 44.
The MSSF system is portable, eliminates the need for specialized disposables and skills, and can
manufactured at scale in a cost eect manner. Furthermore, it performs equivalently to by-hand
techniques while drastically reducing operator involvement and lowering turnaround time. Op-
erator errors are also reduced through automation of all
uid handling steps, including mixing,
transport, incubation, and optical detection. As a result, assay turnaround time is cut to a frac-
tion of the total run time while achieving equivalent detection limits, with a wide linear dynamic
range, and higher detection accuracy and resolution.
In-line detection is made possible by the novel spectrophotometer MFIC, which eliminates the
need for specialized dark room facilities to operate accurately and with optimal sensitivity. Fur-
thermore, the MFIC is designed to rapidly couple to peripheral optical hardware without requiring
specialized ber alignment equipment. The
uidic network design in this MFIC leverages SLA
techniques to fabricate out-of-plane channel routing to achieve light blocks and control of light
path length equivalent to classic cuvette dependent spectrophotometer instruments, while mini-
mizing reagent consumption. This component ts into an arbitrary number of nodes in an MFIC
assembly, making feedback control in complex
uidic networks possible. System feedback control
89
would serve to rapidly optimize system
ow rates and incubation periods for assays with limited
experimental parameter detail. The broad programmability of dilution ratios and inlet reagents
suggests that the MSSF system can act as a platform for monitoring a large menu of enzyme
biomarker targets. To expand on this, we will have to address
uorescence and chemilumines-
cent detection for less abundant biomarkers that require higher detection sensitivity. This will
be accomplished by making modications at the hardware level by exchanging sensor devices or
supplementing optical lter sets to optical bers coupled to the MFIC. This would increase the
systems versatility, serving as a potential diagnostic platform with multiple modes of detection
to monitor a wide range of biomarkers with simple assay parameter changes made in the systems
software.
90
7 Conclusions
In this report, a novel micro
uidic platform of Modular Fluidic and Instrumentation Components
has been introduced. With advancements in stereolithography print resolution, additive manu-
facturing practices have been leveraged to dramatically reduce component cost to the sub-dollar
mark. The library of MFICs described here allows for rapid, recongurable assembly of 3D mi-
cro
uidic assemblies that have been previously dicult, if not impossible, to fabricate. Using
Monte-Carlo analysis, the output performance of high-precision mixing devices has been simu-
lated and validated experimentally to demonstrate the ability to accurately predict performance
of a system assembled from mass-manufactured MFICs. Likewise, surface chemistry modica-
tion of MFIC surfaces has been accomplished via iCVD to enable highly customized, scalable
modication of components. Lastly, the MSSF system exhibits the potential to overcome the
accessibility and resource limitations inherent to by-hand, microtiter plate-based techniques for
executing plasmonic nanoparticle aggregation assays. This new MSSF system is demonstrated in
an analytical application: quantifying an enzyme biomarker of bladder cancer. We demonstrate
equivalent performance metrics to an equivalent experimental setup performed by-hand, on a
microtiter plate.
On a nal note, recent advancements in 3D printing techniques have demonstrated the ability to
achieve micro
uidic channels with cross sections in the tens of microns [99]. Feature sizes on this
scale enables the opportunity to generate MFICs with sub 100 m channel cross section lengths.
The gap between micro
uidics fabricated via SLA and being able to achieve micron level channel
features is beginning to quickly close.
91
8 Acknowledgments
I would like to acknowledge:
Dr. Noah Malmstadt
Dr. Krisna Bhargava
Dr. Celine Billerit
Dr. Carson Riche
Dr. Gertrude Gutierrez
Lu Wang
Nareh Movsesian
Andrew Friedman
Anoop Tembhekar
Danish Iqbal
Roya Ermagan
Bin Pan
and all of the folks in the BME department.
92
References
[1] S.C. Terry J.H. Jerman and J.B. Angell. A gas chromatographic air analyzer fabricated on
a silicon wafer. IEEE Transactions on Electron Devices, 26(12):1880{1886, 1979.
[2] R. Karlsson A. Michaelsson and L. Mattsson. Kinetic analysis of monoclonal antibody-
antigen interactions with a new biosensor based analytical system. Journal of Immunological
Methods, 145:229{240, 1991.
[3] G.M. Whitesides. The origins and the future of micro
uidics. Nature, 44:368{373, 2006.
[4] H.A. Stone A.D. Stroock and A. Ajdari. Engineering
ows in small devices: Micro
uidics
toward a lab-on-a-chip. Annu. Rev. Fluid Mech., 36:381{411, 2004.
[5] A. Manz N. Graber and H.M. Widmer. Miniaturized total chemical analysis systems: A
novel concept for chemical sensing. Sensors and Actuators A, B1:244{248, 1990.
[6] T. Squires and S. Quake. Micro
uidics: Fluid physics at the nanoliter scale. Reviews of
Modern Physics, 77:977{1026, 2005.
[7] E.K. Sackmann A.L. Fulton and D.J. Beebe. The present and future role of micro
uidics in
biomedical research. Nature, 507:181{189, 2014.
[8] X.-M. Zhao Y. Xia and G.M. Whitesides. Soft lithographic methods for nano-fabrication.
Journal of Materials Chemistry, 7(7):1069{1074, 1997.
[9] D.C Duy C. McDonald J.A. Schueller and G.M. Whitesides. Rapid prototyping of micro
u-
idic systems in poly(dimethylsiloxane). Analytical Chemistry, 70:4974{4984, 1998.
[10] Haubert K Drier T and Beebe D. Pdms bonding by means of a portable, low-cost corona
system. Lab on a Chip, 6:1548{1549, 2006.
[11] J.C. McDonald and G.M. Whitesides. Poly(dimethylsiloxane) as a material for fabricating
micro
uidic devices. Accounts of Chemical Research, 2002.
93
[12] J.R Anderson D.T Chiu R.J Jackman O. Cherniavskaya J.C McDonald H. Wu S.H Whitesides
and G.M. Whitesides. Fabrication of topologically complex three-dimensional micro
uidic
systems in pdms by rapid prototyping. Analytical Chemistry, 72:3158{3164, 2000.
[13] B.J. Pine. Mass Customization: The New Frontier in Business Competition. Harvard
Buisness School Press, Boston MA, 1993.
[14] K. Ishii and B. Lee. Design for product retirement and modularity based on technology
life-cycle. In ASME Winter Annual Meeting, volume 2-2, pages 921{933, San Francisco CA,
1995.
[15] Y. Lu C.A. Mirkin K.A. Shaikh K.S. Ryum E.D. Goluch J-.M. Nam J. Liu C.S. Thaxton T.N.
Chiesl A.E. Barron and C. Liu. A modular micro
uidic architecture for integrated biochemical
analysis. Proceedings of the National Academy of Sciences, 102(28):9745{9750, 2005.
[16] P.K. Yuen. Smartbuild - a truly plug-n-play modular micro
uidic system. Lab on a Chip,
8:1374{1378, 2008.
[17] M. Rhee and M.A. Burns. Micro
uidic assembly blocks. Lab on a Chip, 8:1365{1373, 2008.
[18] K. Sun Z. Wang and X. Jiang. Modular micro
uidics for gradient generation. Lab on a
Chip, 8:1536{1543, 2008.
[19] G.M. Whitesides and A.D. Stroock. Flexible methods for micro
uidics. Physics Today,
54(42-38), 2001.
[20] S. Miserendino and Y-.C. Tai. Modular micro
uidic interconnects using photodenable
silicone microgaskets and mems o-rings. Sensors and Actuators A, 143:7{13, 2007.
[21] P. Grodzinski J. Yang R.H. Liu and M.D. Ward. A modular micro
uidic system for cell
pre-concentration and genetic sample preparation. Biomedical Microdevices, 5(4):303{310,
2003.
[22] S.M. Langelier E.L-. Dahl A.J. Manzo B.N. Johnson N.G. Walter and M.A. Burns. Flexible
casting of modular self-aligning micro
uidic assembly blocks. Lab on a Chip, 11:1679{1687,
2011.
94
[23] K.W. Oh K. Lee B. Ahn and E.P. Furlani. Design of pressure-driven micro
uidic networks
using electric circuit analogy. Lab on a Chip, 12(515-545), 2012.
[24] T. Stiles R. Fallon T. Vestad J. Oakey D.W.M Marr J. Squier and R. Jimenez. Hydrodynamic
focusing for vacuum-pumped micro
uidics. Micro
uidics and Nano
uidics, 1:280 { 283,
2005.
[25] K. Lee C. Kim G. Jung T.S. Kim J.Y. Kang and K.W. Oh. Micro
uidic network-based
combinatorial dilution for high throughput screening and optimization. Micro
uidics and
Nano
uidics, 8(577 - 685), 2010.
[26] P.J. Kitson M.H Rosnes V. Sans V. Dragone and L. Cronin. Congerable 3d-printed mil-
li
uidic and micro
uidic 'lab on a chip' reactionware devices. Lab on a Chip, 12, 2012.
[27] A.K. Au W. Lee and A. Folch. Mail-order micro
uidics: evaluation of stereolithography for
the production of micro
uidic devices. Lab on a Chip, 14:1294{1301, 2014.
[28] A.K. Au N. Bhattacharjee L.F. Horowitz T.C. Chang and A. Folch. 3d-printed micro
uidic
automation. Lab on a Chip, 15(1934-1941), 2015.
[29] K. Lee C. Kim B. Ahn R. Panchapakesan A.R. Full L. Nordee J.Y. Kang and K.W. Oh. Gen-
eralized serial dilution module for monotonic and arbitrary micro
uidic gradient generators.
Lab on a Chip, 9:709{717, 2009.
[30] H. Bruus. Theoritical Micro
uidics. Oxford University Press, 2007.
[31] S. Choi M.G. Lee and J-.K. Park. Micro
uidic parallel circuit for measurement of hydraulic
resistance. Biomicro
uidics, 4(034110):1{9, 2010.
[32] S.A. Eccles. Targeting key steps in metastatic tumour progression. Current Opinion in
Genetics and Development, 15:77{86, 2005.
[33] P.C. Wilkinson. Assays of leukocyte locomotion and chemotaxi. Journal of Immunological
Methods, 216:139{153, 1998.
95
[34] N.L. Jeon H. Baskaran S.K.W. Dertinger G.M. Whitesides L.V.D. Water and M. Toner.
Neutrophil chemotaxis in linear and complex gradients of interleukin-8 formed in a micro-
fabricated device. Nature Biotechnology, 20:826{830, July 2002.
[35] T.M. Keenan and A. Folch. Biomolecular gradients in cell culture systems. Lab on a Chip,
8(34-57), 2008.
[36] X. Zhang W. Wang W. Yu Y. Xie X. Zhang Y. Zhang and X. Ma. Development of an
in vitro multicellular tumor spheroid model using microencapsulation and its application in
anticancer drug screening and testing. Biotechnology Progress, 21:1289{1296, 2005.
[37] B. Heit S. Tavener E. Raharjo and P. Kubes. An intracellular signaling hierarchy determines
direction of migration in opposing chemotactic gradients. The Journal of Cell Biology,
159(1):91{102, October 2002.
[38] H. Chen Z. He A. Bagri and M. Tessier-Lavigne. Semaphorin-neuropilin interactions under-
lying sympathetic axon responses to class iii semaphorins. Neuron, 21:1283{1290, 1998.
[39] E.F. Foxman J.J. Campbell and E.C. Butcher. Multistep navigation and the combinatorial
control of leukocyte chemotaxis. The Journal of Cell Biologya, 139(5):1349{1360, 1997.
[40] R.W. Gundersen and J.N. Barrett. Neuronal chemotaxis: chick dorsal-root axons turn toward
high concentrations of nerve growth factor. Science, 206(4422):1079{1080, 1979.
[41] C. Kim K. Lee J.H. Kim K.S. Shin K.-J. Lee T.S. Kim and J.Y. Kang. A serial dilution
micro
uidic device using a ladder network generating logarithmic or linear concetrations.
Lab on a Chip, 8:473{479, 2008.
[42] S.C. Jacobson T.E. McKnight and J.M. Ramsey. Micro
uidic devices for electrokinetically
driven parallel and serial mixing. Analytical Chemistry, 71(20):4455{4459, 1999.
[43] G.M. Walker N.Monteiro-Riviere J. Rouse and A.T. O'Neill. A linear dilution micro
uidic
device for cytotoxicity assays. Lab on a Chip, 7(226-232), 2007.
[44] C. Neils Z. Tyree B. Finlayson and A. Folch. Combinatorial mixing of micro
uidic streams.
Lab on a Chip, 4:342{350, 2004.
96
[45] KC. Bhargava B. Thompson and N. Malmstadt. Discrete elements for 3d micro
uidics.
Proceedings of the National Academy of Sciences, 111(42):15013{15018, 2014.
[46] L.M. Fidalgo C. Abell and W.T.S. Huck. Surface-induced droplet fusion in micro
uidic
devices. Lab on a Chip, 2007.
[47] L.L. Lazarus C.T. Riche B.C. Marin M. Gupta N. Malmstadt and R.L Brutchey. Two-phase
micro
uidic droplet
ows of ionic liquids for the synthesis of gold and silver nanoparticles.
Applied Materials and Interfaces, 2012.
[48] S. Okushima T. Nisisako T. Torii T. Higuchi. Controlled production of monodisperse double
emulsions by two-step droplet breakup in micro
uidic devices. Langmuir, 20:9905{9908,
2004.
[49] A.A. Darhuber and S.M. Troian. Principles of micro
uidic actuation by modulation of surface
stresses. Annu. Rev. Fluid Mech.
[50] H.C. Shum D. Lee I. Yoon T. Kodger and D.A. Weitz. Double emulsion templated monodis-
perse phospholipid vesicles. Langmuir, 24:7651{7653, 2008.
[51] S. Hu X. Ren M. Bachman C.E. Sims G.P. Li and N. Allbritton. Surface modication
of poly(dimethylsiloxane) micro
uidic devices by ultraviolet polymer grafting. Analytical
Chemistry, 2002.
[52] R.S. Kane S. Takayama E. Ostuni D.E. Ingber and G.M. Whitesides. Patterning proteins
and cells using soft lithography. Biomaterials, 1999.
[53] B. Regenberg U. Kruhne L.H. Pedersen M. Simon O.R.T. Thomas J. Nielsen and T. Ahi.
Use of laminar
ow patterning for miniaturized biochemical assays. Lab on a Chip, 2004.
[54] S. Takayama E. Ostuni P. LeDuc K. Naruse D.E. Ingber and G.M. Whitesides. Subcellular
positioning of small molecules. Nature, 2001.
[55] S. Takayama J.C. McDonald E. Ostuni M.N. Liang P.J.A. Kenis R.F. Ismagilov and G.M.
Whitesides. Patterning cells and their environments using multiple laminar
uid
ows in
capillary networks. Proceedings of the National Academy of Sciences, 96:5545{5548, 1999.
97
[56] V. Barbier M. Tatoulian H. Li F. Are-Khonsari A. Ajdari and P. Tabeling. Stable modica-
tion of pdms surface properties by plasma polymerization: Application to the formation of
double emulsions in micro
uidic systems. Langmuir, 22(12):5230{5232, 2006.
[57] A.R. Abate and D.A. Weitz. High-order multiple emulsions formed in poly(dimethylsiloxane)
micro
uidics. Small, 5(18):2030{2032, 2009.
[58] S. Matsumoto. W/o/w-type multiple emulsions with a view to possible food applications.
Journal of Texture Studies, 17:141{159, 1986.
[59] A. Edris and B. Bergnstahl. Encapsulation of orange oil in a spray dried double emulsion.
Food/Nahrung, 45:133{137, 2001.
[60] A. Benichou A. Aserin and N. Garti. Double emulsions stabilized by new molecular recognition
hybrids of natural polymers. Polymers for Advanced Technologies, 13:1019{1031, 2002.
[61] V. Muguet M. Seiller G. Barratt O. Ozer J.P. Marty J.L. Grossiord. Formulation of shear
rate sensitive multiple emulsions. Journal of Controlled Release, 70:37{49, 2001.
[62] S.S. Davis and I.M. Walker. Multiple emulsions as targetable delivery systems. Methods in
Enzymology, 149:51{63, 1987.
[63] K. Bernath M. Hai E. Mastrobattista A.D. Griths S. Magdassi and D.S. Tawk. In vitro
compartmentilizaton by double emulsions: sorting and gene enrichment by
uoroscence
activated cell sorting. Analytical Biochemistry, 2004.
[64] Y. Zhang Y.P. Ho Y.L. Chiu H.F. Chan B. Chlebina T. Schuhmann L. You and K.W. Leong.
A programmable microenviornment for cellular studies via micro
uidics-generated double
emulsions. Biomaterials, 2013.
[65] H.F. Chan Y. Zhang Y.-P. Ho Y.-L. Chiu Y. Jung and K.W. Leong. Rapid formation
of multicellular spheroids in double-emulsion droplets with controllable microenvironment.
Scientic Reports, 2013.
98
[66] A. Zenchenko S.R.A. Devenish B. Kintses P.-V. Colin M. Fischlechner and F. Hollfelder. One
in a million: Flow cytometric sorting of single cell-lysate assays in monodisperse picolitre
double emulsion droplets for directed evolution. Analytical Chemistry, 2014.
[67] A.S. Utada E. Lorenceau D.R. Link P.D. Kaplan H.A. Stone and D.A. Weitz. Monodisperse
double emulsions generated from a microcapillary device. Science, 308:537{541, 2005.
[68] H.C. Shum Y.-J. Zhao S.-H. Kim and D.A. Weitz. Multicompartment polymersomes from
double emulsions. Angewandte Chemie, 50:1648{1651, 2011.
[69] E. Lorenceau A.S. Utada D.R. Link G. Cristobal M. Joanicot and D.A. Weitz. Generation
of polymerosomes from double emulsions. Langmuir, 21:9183{9186, 2005.
[70] Y. Wang H.H. Lai M. Bachman C.E. Sims G.P. Li and N.L Allbritton. Covalent micropat-
terning of poly(dimethylsiloxane) by photografting through a mask. Analytical Chemistry,
2005.
[71] C.T. Riche B.C. Marin N. Malmstadt M. Gupta. Vapor deposition of cross-linked
uoropoly-
mer barrier coatings onto pre-assembled micro
uidic devices. Lab on a Chip, 11:3049{3052,
2011.
[72] C.T. Riche C. Zhang M. Gupta and N. Malmstadt. Fluropolymer surface coatings to control
droplets in micro
uidic devices. Lab on a Chip, 14:1834{1841, 2014.
[73] S. Seidel C.T. Riche and M. Gupta. Encylopedia of polymer science and technology.
[74] Y. Mao and K.K. Gleason. Hot lament chemical vapor deposition of poly(glycidyl methacry-
late) thin lms using tert-butyl peroxide as an initiator. Langmuir, 2004.
[75] K. Chan and K.K. Gleason. Initiated chemical vapor depositon of linear and cross-linked
poly(2-hydroxyethyl methacrylate) for use as thin-lm hydrogels. Langmuir, 2005.
[76] G.K. Singh and M. Siahpush. Widening rural-urban disparities in life expectancy, u.s., 1969-
2009. American Journal of Preventitive Medicine, 2004.
99
[77] C.J. Easely J.M Karlinsey J.M Bienvenue L.A. Legendre M.G Roper S.H Feldman M.A.
Hughes E.L Hewlett T.J. Merkel J.P. Ferrance J.P. Landers. A fully integrated micro
uidic
genetic analysis system with sample-in-answer out capability. Proceedings of The National
Academy of Sciences, 2004.
[78] K. C. Bhargava, B. Thompson, and N. Malmstadt. Discrete elements for 3d micro
uidics.
Proc. Natl. Acad. Sci. U.S.A, 111(42):15013{15018, 2014.
[79] K. C. Bhargava, B. Thompson, D. Iqbal, and N. Malmstadt. Predicting the behavior of
micro
uidic circuits made from discrete elements. Sci. Rep., 5:15609, 2015.
[80] K. C. Bhargava, B. Thompson, A. Tembhekar, and N. Malmstadt. Temperature sensing in
modular micro
uidic architectures. Micromachines, 7(1):11, 2016.
[81] B. Thompson, C. T. Riche, N. Movsesian, K. C. Bhargava, M. Gupta, and N. Malmstadt.
Modular, discrete micromixer elements fabricated by 3d printing. Micro
uid. Nano
uidics,
20(5):78, 2016.
[82] A. Gomez-Hens and D. Perez-Bendito. The stopped-
ow technique in analytical chemistry.
Analytica Chimica Acta, 242:147{177, 1991.
[83] K. S. Chakrabarti, R. V. Agafnov, F. Pontiggia, R. Otten, M. K. Higgins, F. X. Schertler,
D. D. Oprian, and D. Kern. Conformational selection in a protein-protein interaction revealed
by dynamic pathway analysis. Cell. Rep., 14(1):32{42, 2016.
[84] H. Roder, K. Maki, H. Cheng, and M. C. Ramachandra Shastry. Rapid mixing methods for
exploring the kinetics of protein folding. Methods, 34:15{27, 2004.
[85] R. Bleul, M. Ritzi-Lehnert, J. Hoth, N. Scharpfenecker, I. Frese, D. Duchs, S. Brunk-
laus, T. E. Hansen-Hagge, F. J. Meyer-Almes, and K. S Drese. Compact, cost-ecient
micro
uidics-based stopped-
ow device. Anal. Bioanal. Chem., 399:1117{1125, 2011.
[86] H. Song and Ismagilov R.F. Millisecond kinetics on a micro
uidic chip using nanoliters of
reagents. JACS, 125:14613{14619, 2003.
100
[87] V. Srinivasa, V. K. Pamula, and R. B. Fair. Droplet-based micro
uidic lab-on-a-chip for
glucose detection. Anal. Chima. Acta, 507:145{150, 2004.
[88] L. Jiang and S. Pau. Integrated waveguide with a micro
uidic channel in spiral geometry
for spectroscopic applications. Appl. Phys. Lett., 90:111108, 2007.
[89] P. S. Dittrich and A. Manz. Lab-on-a-chip: micro
uidics in drug disovery. Nat. Rev. Drug
Discov., 5:210{218, 2006.
[90] J. R. Gregg, C. Lai, R. Dmochowki, T.R. Talbot, and D. A Barocas. Recent antibiotic
treatment increases the risk of urinary tract infection after outpatient cystoscopy. Urol.
Prac., 3(2):90{96, 2016.
[91] K. Koo, L. Zubko, B. E. Sirovich, P. P. Goodney, D. J. Robertson, J. D. Seigne, and F. R.
Schroeck. The burden of cytoscopic bladder cancer surveillance: Anxiety, discomfort, and
patient preferences for decision making. Urology, 108:122{128, 2017.
[92] V. B. Lokeshwar, T. Habuchi, H. B. Grossman, W. M. Murphy, S. H. Hautmann, G. P. Hem-
street III, A. V. Bono, R. H. Getzenberg, P. Goebell, B. J. Schmitz-Drager, J. A. Schalken,
Y. Fradet, M. Marberger, E. Messing, and M. J. Droller. Bladder tumor markers beyond
cytology: International consensus panel on bladder tumor markers. Urology, 66(6):35{63,
2005.
[93] V. B. Lokeshwar, D. Rubinowicz, G. L. Schroeder, E. Forgacs, J. D. Minna, N. L. Block,
M. Nadji, and B. L. Lokeshwar. Hyaluronan in peritumoral stroma and malignant cells
associates with breast cancer spreading and predicts survival. Am. J. Pathol., 156(2):529{
536, 2001.
[94] P. Auivinen, R. Tammi, J. Parkkinen, M. Tammi, U. Agren, R. Johansson, P. Hirvikoski,
M. Eskelinen, and V.-M. Kosma. Hyaluronan in peritumoral stroma and malignant cells
associates with breast cancer spreading and predicts survival. Am. J. Pathol., 156(2):529{
536, 2000.
101
[95] V. B. Lokeshwar, C. Obek, M. S. Soloway, and N. L. Block. Tumor-associated hyaluronic
acid: a new sensitive and specic urine marker for bladder cancer. Cancer Res., 58(4):773{
777, 1998.
[96] H. T. Pham, N. L. Block, and V. B. Lokeshwar. Tumor-derived hyaluronidase: a diagnostic
urine marker for high-grade bladder cancer. Cancer Res., 57(4):778{783, 1997.
[97] J. W. Kim, J. H. Kim, S. J. Chung, and B. H. Chung. An operationally simple colorimetric
assay of hyaluronidase activity using cationic gold nanoparticles. Analyst, 134(134):1291{
1293, 2009.
[98] K. C. Bhargava, R. Ermagan, B. Thompson, A. Friedman, and N. Malmstadt. Modular,
discrete micromixer elements fabricated by 3d printing. Micromachines, 7(5):137, 2017.
[99] H. Gong, B.P. Bickman, A.T. Woolley, and G.P Nordin. Custom 3d printer and resin for 18
um x 20um micro
uidic
ow channels. Lab on a Chip, 2017.
102

Abstract (if available)

Abstract In this report, a novel microfluidic platform is introduced. By leveraging the benefits of advanced stereolithography over traditional methods of fabricating microfluidic devices, a library of standardized, Modular Fluidic and Instrumentation Components (MFIC’s) is designed, manufactured, and tested. This manufacturing approach allows for the large-scale production of 3D microfluidic components that were previously difficult, if not impossible, to generate. In this report we apply MFIC assembled systems towards high-precision mixing, the control of multi-phase flows, and real-time spectrophotometry. A statistical approach to analyze the performance and predictability of microfluidic mixing in the context of mass manufacturing is demonstrated. This study serves as a basis for considering massively parallelized MFIC systems that are able to execute high-precision mixing for general bench top applications. Furthermore, the surface chemistry of MFIC’s are modified via initiated chemical vapor deposition (iCVD). Surface channel modification via iCVD further enables this platform to fit the demands of mass customization and manufacturing in the context of scalable microfluidic design. Lastly, an MFIC enabling real-time optical sensing is introduced to expand the functionality of the MFIC library. This component is scalable and incorporates off-the-shelf optical detection for absorbance-based sensing. This novel MFIC is integrated into a “Multi-Stage Stopped Flow” system. This system has been developed to collapse the bench top work flow necessary to execute Gold Nanoparticle (AuNP) based aggregation assays commonly used to monitor enzyme activity. Here, the activity of a urinary biomarker for bladder cancer, Hyaluronidase, is monitored.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Predictable microfluidic mixing using discrete element microfluidics

PDF

Intricate microfluidic devices for biopharmaceutical processes: forging ahead with additive manufacturing

PDF

A modular microscale laboratory

PDF

High-resolution optical instrumentation for biosensing and biomechanical characterization

PDF

Vapor phase deposition of dense and porous polymer coatings and membranes for increased sustainability and practical applications

PDF

Scalable polymerization additive manufacturing: principle and optimization

PDF

Continuous flow synthesis of catalysts with custom made reactor with flow and temperature studies

$Metasurfaces in 3D applications: multiscale stereolithography and inverse design of diffractive optical elements for structured light$

PDF

Metasurfaces in 3D applications: multiscale stereolithography and inverse design of diffractive optical elements for structured light

PDF

Biological and chemical detection using optical resonant cavities

PDF

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

PDF

Hybrid vat photopolymerization: methods and systems

PDF

Development of optical instrumentation and signal analysis for biomedical applications

PDF

Modeling and simulation of complex recovery processes

Asset Metadata

Creator Thompson, Bryant (author)

Core Title Expanded functionality and scalability of modular fluidic and instrumentation components

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 10/18/2018

Defense Date 07/23/2018

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

Tag 3D printing,additive manufacturing,microfluidics,OAI-PMH Harvest,rapid manufacturing,sensors

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Malmstadt, Noah (committee chair), Chung, Eun Ji (committee member), Roberts, Richard W. (committee member)

Creator Email bryantth@usc.edu,bryantthompson05@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-86449

Unique identifier UC11675643

Identifier etd-ThompsonBr-6889.pdf (filename),usctheses-c89-86449 (legacy record id)

Legacy Identifier etd-ThompsonBr-6889.pdf

Dmrecord 86449

Document Type Dissertation

Format application/pdf (imt)

Rights Thompson, Bryant

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 a...

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