You are on page 1of 22

University of Waterloo

Faculty of Engineering
Nanotechnology Engineering

Thin-Walled PDMS Membranes for Use in Reducing T-junction Droplet Size


Fluctuation

Professor Derek Rayside


Department of Electrical and Computing Engineering
University of Waterloo
200 University Ave W
Waterloo, ON, N2L 3G1

Prepared by
Alex Willerth
ID: 20560828
User ID: awillert
2A Nanotechnology Engineering
May 1, 2017
330 Phillip Street
Waterloo, Ontario, Canada
N2L3W9

May 1, 2017

Dr. Shirley Tang, Program Director


Nanotechnology Engineering
University of Waterloo
Waterloo, Ontario
N2L3W9

Dear Madam,

This report, entitled Thin-Walled PDMS Membranes for Use in Reducing T-junction Droplet Size
Fluctuation was prepared as my 2A work term report for research under Professor Derek Rayside
of the Electrical and Computer Engineering Department at the University of Waterloo. This report is
in fulfillment of the course NE250.

Professor Rayside is currently designing a microfluidic design automation toolchain, named


Manifold, to ease the process of microfluidic circuit design. During Winter 2017, I was working
under his supervision as an NSERC USRA recipient. His research team included a fourth-year design
project team from ECE, a masters thesis student, a PhD student, and myself. One of the core elements
of Prof Raysides research is utilizing a universally applicable frontend to describe a users circuit,
and then simulating and optimizing for relevant parameters using a domain-specific backend. The
microfluidic-specific backend is critical for the initial application of the toolchain, and includes
simulation of a T-junction including thin-walled membranes.

The purpose of this report is to outline the motivation and process of estimating the effects of these
membranes on droplet size fluctuation. Estimation is the first step to simulation, the preponderant
functionality of the Manifold backend. With an effective empirical model of how membranes affect
droplet size fluctuation, the users requirement of a max fluctuation can be satisfied without trial and
error. The target audience of this report is Professor Derek Rayside, my supervisor, and future
employees thereof, to be used as a reference while refining their microfluidic design automation tool.

I would like to thank Derek Rayside for his guidance and the opportunity to work with a product with
such potential. I would also like to thank Atulan Zaman for his guidance, which enabled me to learn
the relevant concepts for this report. I would also like to thank Cody Chen for his generous help in
fabricating several T-junctions, testing and imaging their droplet generation, and allowing me the
chance to learn the relevant procedures. Lastly, I would like to thank Stacy Gaikovaia and Lauren
Meadowcroft for proofreading this report and improving its quality.

I hereby confirm that I have received no further help other than that mentioned above in writing this
report. I also confirm that this report has not previously been submitted for academic credit at this or
at any other academic institution.

Sincerely,

Alex Willerth
ID 20560828
Contributions

Professor Rayside employs several students at various levels of education and on a rotating basis in
the case of co-ops. The original vision for Manifold was created by Murphy Berzish and Derek
Rayside in 2016. Subsequently, the bulk of the programming work for the end-to-end toolchain was
completed by a team of five fourth-year ECE students, who presented the framework of the project
at the ECE capstone symposium in April 2017. Professor Derek Wright is also currently affiliated
with the project and working on creating a finite element method of simulating a T-junction.

The objective of the team is to design a microfluidic design automation program to ease the design
and specification of microfluidic circuits. Mr Berzish was responsible in 2016 for outlining the need
for the program, an overview of its functional flow, the language of the front-end design script utilized
by the user, and several applications for the microfluidic backend specifically. Mr Zaman, with whom
I worked closely, focused on developing the SMT solver used as a core piece of the Manifold
architecture. Mostly using dReal, Mr Zaman was tasked with solving underconstrained problems
necessary for most of the simulation steps and helping integrate the functionality into Manifold.

As a nanotechnology student joining a team of graduates and soon-to-be-graduates with extensive


experience in software engineering, my tasks focused on the microfluidic backend of Manifold. I was
tasked with simulating a T-junction in COMSOL to begin building the finite element analysis section
of the toolchain. Subsequently, I worked with Mr. Chen to assist in and learn about the fabrication,
testing, and imaging of T-junctions using various amounts and sizes of thin-walled PDMS
membranes. With the data collected during these experiments, I performed extensive data analysis to
find the effect of the membranes on the droplet size fluctuation, which forms the bulk of this report.

My report presents an important part of the vision for Manifold, which was the focus of my work
term. My main objective by demonstrating the effects of membranes on T-junctions is to demonstrate
that they can be just as precise as other, more expensive, alternatives. In writing this report, I have
consolidated my findings in such a way as to accurately portray these effects. This will allow future
development of Manifold to offer membrane-equipped T-junctions as a node in the microfluidic
backend. Furthermore, as Prof Rayside plans to publish a paper including my findings, creating this
work term report has given me the opportunity to practice academic writing in preparation. It has also
developed my organizational, communication, and data analysis skills. Microfluidic circuit creation
is not closely related to any academics in Nano so far, but in later years I intend to pursue similarly-

iv
focused elective courses. Therefore, conducting this research and preparing this report will likely
expose me to important experience for the future of my degree.

This report has potential to have a significant impact on Prof Raysides goals. T-junctions are
currently regarded as the cheapest but least precise droplet generation method. They are entirely
passive and easily fabricated, but the droplets created fluctuate in size such that they are unusable in
applications where the testing criteria require low margins of error. More active designs,
incorporating manual pinching of the droplets by automated methods, exist and generally yield more
precise volumes while being much more expensive to manufacture. T-junctions with low size
variation are therefore an attractive alternative if they can be created, and coupled with an automated
design process via Manifold would present a massively cheaper design alternative for droplet
generation in any microfluidic circuit. Therefore, if I can demonstrate that thin-walled membranes
create T-junctions with low size variation, and the precise relation can be included in the automated
simulation within Manifold, it will significantly improve the attractiveness of Prof Raysides final
product.

v
Summary

The context of this report is microfluidic droplet generation. For many applications, there exists a
large demand for droplet generation, for which one of the easiest and most inexpensive methods is
the T-junction. However, the problem addressed in this report is the main limitation of T-junctions:
the low precision of generated droplets. Therefore, improving the precision of this generation without
drastically increasing cost would provide major benefits to the relevant microfluidic circuits.

The motivation of this report is to find a cheap method of improving the precision of droplets
generated by a T-junction, to make them viable for more applications. The difficulty of this area lies
in finding an effective solution without increasing manufacturing cost or complexity. T-junctions are
only utilized when cost must be minimized, and so any more precise design that also costs more will
likely be ignored. The impact of this solution would be the ability to fabricate multi-way T-junctions
at low cost and with high droplet precision. This would enable a drastic increase in the cost-efficiency
of pharmaceutical drug testing, discrete microfluidics, and other droplet-based research.

The approach of this report involves three main steps: designing a circuit with membranes and a
circuit without, running the same droplet generation experiment with each, and analyzing the images
collected to find droplet size fluctuation. The circuits were designed using PDMS soft lithography,
the experiments were run using oil and water at various pressure levels, and the image analysis was
conducted using ImageJ and MATLAB.

The major result of this report is that the membranes tested reduced droplet size fluctuation from
19.09% to 2.11%. The fundamental principle causing this effect is the way the membranes act as
capacitors to absorb pressure change, which solves the problem of size change by maintaining
constant pressure at the droplet generation site. The evaluation of this report is that the problem of
low droplet size precision can be solved by including enough membranes to any circuit.

The conclusions of this report are that the membranes reduced droplet size fluctuation from 19.09%
to 2.11%, that there was an unexpected macroscale pattern observed in the fluctuation, and that the
baking procedure of the PDMS negatively impacted the membranes performance.

This report recommends an investigation to quantify the effects of membrane variables on precision,
an investigation into the macroscale pattern observed, and a decrease in baking time of PDMS.

vi
Table of Contents
Contributions...................................................................................................................................... iv

Summary ............................................................................................................................................ vi

List of Figures .................................................................................................................................. viii

List of Tables ..................................................................................................................................... ix

1.0 Introduction ................................................................................................................................... 1

1.1 Current Difficulties in Microfluidic Circuit Design .................................................................. 1

1.2 Current Difficulties in Microfluidic Droplet Generation .......................................................... 1

2.0 Background ................................................................................................................................... 3

2.1 Manifold .................................................................................................................................... 3

2.2 Application of Membranes within Manifold............................................................................. 3

2.3 Problem Statement .................................................................................................................... 4

2.4 Constraints and Requirements................................................................................................... 4

3.0 Circuit Design ............................................................................................................................... 5

3.1 Hypothesis................................................................................................................................. 5

3.2 Design ....................................................................................................................................... 5

3.3 Fabrication ................................................................................................................................ 6

4.0 Experimental Procedure ................................................................................................................ 8

4.1 Economic Considerations ......................................................................................................... 8

5.0 Image Analysis .............................................................................................................................. 9

5.1 Data ........................................................................................................................................... 9

5.2 Sources of Error ...................................................................................................................... 10

6.0 Impacts on the Manifold Software .............................................................................................. 11

6.2 Region of Impact ..................................................................................................................... 11

7.0 Conclusions ................................................................................................................................. 12

8.0 Recommendations ....................................................................................................................... 13

References ......................................................................................................................................... 14

vii
List of Figures

Figure 1: Formation of a bubble at a planar T-junction. ..................................................................... 2


Figure 2: Greyscale image of a T-junction ......................................................................................... 6
Figure 3: Membrane deformation ....................................................................................................... 7
Figure 4: Images of droplet breakup ................................................................................................... 8
Figure 5: Average droplet size over time .......................................... Error! Bookmark not defined.

viii
List of Tables

Table 1: Summary of maximum size fluctuation.9

ix
1.0 Introduction

Manifold is a microfluidic design automation program which addresses the difficulty of designing a
microfluidic circuit. It is detailed in the vision paper titled Manifold: An SMT-Based Declarative
Language for Electronic and Microuidic Design Synthesis by M Berzish, A Khan, A Zaman, V
Ganesh, and D Rayside [1]. This report presents a unique method of improving T-junctions for use
in Manifolds simulations.

1.1 Current Difficulties in Microfluidic Circuit Design

Currently, microfluidic circuit design faces two large problems. Firstly, designing a circuit is
challenging due to the lack of effective solutions for the simulation of entire systems to achieve
specific desired parameters. This can result in circuit designers solving for a handful of variables and
leaving the rest to chance, or alternatively, solving for all relevant parameters of their module of
interest but without the ability to anticipate how each component of the system will affect the rest of
it.

Secondly, it is time consuming to achieve the desired design. Current fabrication workflows include
trial and error synthesis due to the challenges outlined above. To test each design iteration can take
weeks, as the fabrication procedure often involves soft lithography, a process which requires lengthy
periods of time to create masks used. A circuit must be designed to the best of ones ability, sent off-
site to develop a mask, and finally several copies must be made using the mask. Despite best efforts
in this procedure, there is a possibility the circuit will not perform as expected due to the
interdependent and complex nature of microfluidic systems. This could result in the necessity of
beginning the entire process again, with only a few tweaks to your design and no further certainty of
success.

1.2 Current Difficulties in Microfluidic Droplet Generation

Within microfluidic circuits, there exists a large interest in droplet generation techniques. These are
divided broadly into passive techniques, without moving parts, and active techniques, which require
automated mechanical motion. Passive techniques are cheaper to manufacture, less precise, and
require less maintenance. The most widespread passive method is the T-junction, wherein a

1
continuous and discrete phase of liquids combine in a T shape to produce droplets at a consistent rate.
The mechanism of droplet formation is detailed below, in Fig. 1.

Figure 1Formation of a bubble at a planar T-junction [2]

In Figure 1, oil and water, respectively, enter the main channel (width: w) and side channel (width:
win) at rates of ow qc and qd. A formation cycle comprises two periods: a lling period in which the
bubble lls-up the junction and grows to a size Vll and a squeezing period in which the forming
bubble is squeezed by the liquid until it pinches-off. (Figure, caption, and description copied without
alteration [2]).

T-junctions, as stated above, sacrifice precision in their droplet generation due to the size fluctuation
inherent to their operation. This size fluctuation is not well understood. It is often theorized to be
associated with the pinching period of droplet production, when a sudden change in the channel could
cause a difference in subsequent droplets. However, as shown in section [3], the size fluctuation
occurs on a macro scale spanning an average of 8.6 droplets, effectively dismissing the possibility of
the trigger lying within the production of a single droplet. Regardless of the source, the imprecise
nature of T-junctions makes them unusable in many applications which require consistency.

If both problems outlined above can be solved, it will facilitate a massive increase in the accessibility,
cost efficiency, effectiveness, and speed of microfluidic research and development, specifically
droplet generation.

2
2.0 Background

The necessary technical background for this paper encompasses two areas: the need for and
functionality of the Manifold software toolchain, and the functionality offered by membranes via
Manifold. Additionally, the problem statement is presented.

2.1 Manifold

The problems facing microfluidic circuit design are not new. Electronic circuits once faced similar
hurdles, with most design done entirely by hand and with frequent flaws. The solution, and the reason
electronic circuits are so advanced, widespread, and inexpensive today, is automated design
programs. These programs vary widely, but their main purpose is to receive information from the
user on their target circuit, solve for all missing variables, simulate the result sufficiently to ensure it
will yield the intended features, and finally return the refined design to the user in a format easily
fabricated.

Manifold presents a similar toolchain, without restricting the user to electronic circuits. The front end
of the program is a design language engineered to be applicable to any type of circuit that has an
available backend. After the circuit is described by the user via this design language, it is passed to
the microfluidic-specific backend, which simulates all required parameters according to the
specification and returns a refined circuit. This eliminates both the difficulty and uncertainty inherent
in microfluidic circuit design, drastically reducing the manufacturing time and the cost.

2.2 Application of Membranes within Manifold

The entirety of Manifold constructs circuits as a combination of nodes, ports, and connections, which
function exactly as their names suggest. Therefore, the microfluidic backend contains nodes which
represent every modern circuit element. This includes simple pipes, valves, pressure sources, inlets
and outlets, and loads such as T-junctions. It is important that each node be governed by all relevant
equations and laws known to interact with the fluid flow within.

3
2.3 Problem Statement

The problem being solved in this report is the definition of control over droplet size fluctuation within
T-junctions, for use within the Manifold framework. As membranes are a novel method of control,
no current literature exists relating their type, size, quality, and quantity on the droplet size
fluctuation. The end goal of finding this relation is to enable users to define their required precision.
Currently, T-junctions offer a questionable degree of precision at best. If membranes effect is
quantized, it will allow users to specify a maximum acceptable size fluctuation and Manifold will be
equipped to return with a capable specification.

To know if the problem has been solved, a clear effect of the membranes upon droplet size fluctuation
must be observed. A reduction averaging at least 50% of the fluctuation across all tested pressure
steps is desired to ensure the potential for effective control.

2.4 Constraints and Requirements

As mentioned in section 1.2, there already exist several more precise but more expensive methods of
droplet generation. Therefore, if an improvement in T-junction precision is to be attractive to circuit
designers, it must not increase cost more than 10% or fabrication time more than 25%. Additionally,
it must not slow the rate of droplet generation more than 5%, as this parameter is usually extensively
maximized. Lastly, it must still allow the circuit to function at full capability: up to a pressure of
5atm, and useable with a variety of liquids.

4
3.0 Circuit Design

The circuits designed for this experiment were unique and specifically designed. The hypothesis
being tested as well as the circuits design and fabrication processes are detailed here.

3.1 Hypothesis

To sufficiently test the objective of this paper, two circuits were designed. Only one contained twelve
membranes (detailed below) but all other variables were kept as equal as possible. The intention of
the membranes was to deform in response to the pressure fluctuations responsible for the droplet size
fluctuation regularly observed T-junctions. By deforming in this manner, it was hypothesized that the
membranes would absorb the pressure waves before they affected the droplet breakup site,
maintaining size precision.

3.2 Design

To reduce cost and increase turnaround time, the same internal structure was used for both circuit: a
continuous phase connected directly to an inlet, and a discrete phase connected through a series of
switchback pipes to another. After intersecting, the circuit proceeded directly to an outlet. The
switchbacks preceding the discrete phase, through which water would travel, were necessary to
equalize the hydraulic resistance with the oil inlet, leading to the continuous phase. Figure 2, below,
shows an overview of the area of interest, encompassing the T-junction where droplet breakup was
imaged. This figure used a lens with focus 1.5x - 10x, for a scale of 1.348 m/pixel. The continuous
phase, running vertically centered, is flanked by four membranes above the junction, while the
discrete phase, entering frame from left, is flanked by eight visible membranes.

5
Figure 2: Greyscale image of a T-junction

The channels were fabricated with a width of 100m and a height of 50m, while the membranes
each had a length of 75m and a height of 50m. The placement of the membranes was an important
design decision. Due to the exploratory nature of the experiment, the exact location of the pressure
changes couldnt be anticipated. Therefore, membranes were placed along both incoming channels
so that their deformations could be observed and the placement analyzed.

Mentioned above, the two circuits had the same internal design. The membranes within one of the
circuits were disabled by filling and sealing the holes outside of them.

3.3 Fabrication

The exact fabrication procedure of the circuits was heavily modified from accepted literature
methods, and so warrants explanation. Soft lithography was used to construct the circuits entirely of
PDMS, with the membranes necessarily fabricated as part of the surrounding walls to preserve their
integrity. The baking procedure used a curing agent with a ratio of 1:10 to solvent. Additionally, after
curing the circuits were each baked at 1100C for 72 hours each. This is drastically different from
the conventional baking procedure, which usually operates at a similar temperature but for times as
short as 15 minutes.

The reason for this modified procedure was consistency. Due to the exploratory nature of the
experiment, all efforts were made to precisely preserve every variable between experiments to

6
accurately portray the effects of the membranes. Therefore, Dr Chen aimed to avoid the slight
differences that could all too easily occur with such a short baking time. Even a few seconds
difference in baking time at these early periods could result in a significantly different Youngs
Modulus, which was a critical feature of the circuits. Therefore, the circuits were baked until it could
be assumed they had nearly reached steady-state, such that any difference in the baking procedures
would not affect the material.

This modified procedure resulted in a significantly stiffer preparation of PDMS than is normally
encountered. Figure 3, below, demonstrates the change in deformation of a membrane. Both images
are taken at scope 1.5x 40x, 3.03 pixels/m. The left image shows no deformation, at rest with no
applied pressure. The right image shows deformation due to an applied pressure of 1800 mbar.
According to image analysis, an average deformation of 4.41 m was observed.

Figure 3: Membrane deformation

Using a plate theory equation from [3], the PDMS was found to have a Youngs Modulus of 1856
kPa. This is drastically different from the literature value of 360-870 KPa [4].

7
4.0 Experimental Procedure

Due to the intent of the experiment, all variables were kept constant between two distinct runs. The
T-junctions ran for several minutes at each of five pressure steps to allow them to reach steady-state,
with approximately 1-2 seconds of images captured.

The two circuits each were supplied with oil through their continuous phase, and water through their
discrete phase. Pressure for the oil and water in mbar, respectively, were 300x199, 600x410,
1000x720, and 1500x1070.

The images captured were focused on junction where droplet breakup occurred, as shown below in
Figure 4. Droplets were analyzed throughout their subsequent trajectory to gather enough data for an
accurate size measurement. Water is outlined with dark lines, while oil fills all remaining space in
the channels.

Figure 4: images of droplet breakup

4.1 Economic Considerations

As discussed in section 2.4, design and fabrication of these membranes must not increase cost more
than 10% or fabrication time more than 25%. After creating the circuit, no significant cost difference
was observed, because the material requirements and precision requirements are almost identical to
a standard PDMS circuit. Fabrication time for previous circuits in the lab was approximately one
week, including more than two days spent baking. While fabricating circuits with membranes, baking
time was slightly lowered, and should be lowered further in subsequent experiments (see section 8.0).
Therefore, this method of increasing precision is acceptable economically.

8
5.0 Image Analysis

Image analysis of particles is a well-documented area of research, and the author of the report relied
on a specific publication to appropriately extract meaningful data from the experiments [5].

The primary consideration while collecting results was to keep the process the same between the two
experiments. As such, steps were taken including maintaining consistent camera focus, exposure, and
ensuring the surface of the chip was clean each time to minimize artifacts.

5.1 Data

The data collected is summarized here. A succinct summary is shown in Table 1, below. The first
two rows show % size fluctuation.

Table 1: Summary of maximum size fluctuation


# membranes Pressure 1 Pressure 2 Pressure 3 Pressure 4
0 10.45 24.73 22.58 18.60
12 3.95 1.62 1.90 0.99
Membrane effect
(% reduced) 62.20 93.45 91.59 94.68

As clearly demonstrated, the membranes had the theorized effect. In all pressure steps, they decreased
the droplet size fluctuation far beyond what would have been possible with no effective capacitors.

Additionally, a macroscale pattern was observed in the size fluctuation in both circuits. Upon
preparing the experiment, the expectation was that size change was caused by pressure change
resulting from each pinching phase, causing fluctuation in a pattern encompassing perhaps two to
three droplets. However, a much broader period was observed, as shown below, in Figure 4. This
data was collected during generation without membranes, but adding membranes did not significantly
alter the period. Each scatter point signifies one droplet. A macro repetition can be observed over an
average period of 18.2 droplets.

9
18000

16000

14000
Droplet Size, m2

12000

10000

8000

6000

4000

2000

0
0 20 40 60 80 100 120 140 160 180
Droplet #

Average Droplet Size, m2


Oil 600mbar, water 410mbar, captured at 500 fps
5.2 Sources of Error

The most notable source of error was observed in the image analysis: there was no observed
membrane deformation in response to the droplet size fluctuation. In fact, after extensive checking
and utilizing several comparison methods, the membrane doesnt appear to adjust at all during the
entire capture period. Deformation was clearly observed between rest state and applying 1800 mbar
of pressure, but the pressure differences causing the size fluctuation were clearly not enough to affect
the membranes.

10
6.0 Impacts on the Manifold Software

Manifolds current workflow is important to understand to realize the potential impact of finding a
positive relationship between membranes and drop control. Firstly, every piece of the circuit
specified by the user takes the form of a node. Pipes, loads, T-junctions, inlets, outlets, and filters are
all examples of nodes currently included in the microfluidic backend. Each of these nodes is packaged
with its own set of equations that govern the internal mechanics and effect on adjacent nodes, via the
ports and connections available. Thus, is necessary to build a complete picture of each node before
it can be effectively added to any Manifold circuit design.

6.2 Region of Impact

The T-junction node, still currently in development within Manifold, is underrepresented. There does
not currently exist any complete simulation method for solving for desired parameters of a T-junction,
nor is there recorded information on methods of reducing droplet size fluctuation. The conventional
method employed with marginal success is lengthening the outlet pipe to an arbitrary length (several
mm) such that any pressure changes caused by droplets leaving the channel have less effect on the
breakup occurring at the T-junction.

Improving the T-junction node with an accurate method of finding droplet size fluctuation would
drastically improve its viability for use in circuit design. While Manifold presents a uniquely simple
and powerful approach that removes much onus from the user, it is still currently necessary for the
user to decide upon an appropriate droplet generation method and coordinate its precision to their
satisfaction. With the ability to designate a target precision within their T-junction node during front-
end circuit design, the entire process becomes extremely accessible, simple, and cheap to
manufacture.

11
7.0 Conclusions

From the analysis in the report body, it was concluded that, upon adding 12 membranes to the circuit,
the droplet size variation decreased from 19.09% to 2.11%, for a reduction of 88.9%. This presents a
sweeping improvement to droplet size precision, such as cannot be achieved through any other known
methods. It also meets the criteria defined in the section 2.3 of this report, as 88.9% is far greater than
50%.

Additionally, as detailed in section 5.2 of this report, a macroscale pattern was observed during
droplet generation. Contrary to initial hypotheses, which assumed that each individual droplet would
cause a fluctuation in internal pressure, the only fluctuation observed occurred over the course of an
average of 18 droplets. This is a significant conclusion, as it requires abandoning the prior widely
held belief.

Lastly, according to section 3.3, it was concluded that the altered fabrication procedure used for
PDMS negatively impacted the membranes performance. Due to the radically increased baking
timing, the Youngs Modulus similarly increased, resulting in much stiffer membranes with less
response to pressure than expected. This stiffness stifled any response to small pressure fluctuations.

12
8.0 Recommendations

The recommendations of this report revolve around the work currently being conducted by the author
under supervision of Prof Rayside.

Based on the analysis and conclusions in this report, further study is recommended to quantify the
effect of membrane dimensions, number, position, and flexibility. Several more experiments should
be designed, each keeping all variables the same except for one of these, such that its effect can be
observed. After sufficient experimentation, a second report should be prepared summarizing these
effects, isolated into a single set of equations relating all relevant membrane parameters with
maximum droplet size fluctuation.

Secondly, based on the findings in section 5.2, more work is warranted to understand the macroscale
pattern observed. An experiment should be conducting including several microscale pressure sensors
placed strategically throughout an enlarged T-junction design. Both with membranes and without,
data should be collected during droplet generation to understand the way pressure changes propagate
throughout the fluid medium.

Lastly, based on the observations in section 3.3, it is recommended to decrease the baking time and
ratio of the curing agent in subsequent fabrications of thin-walled PDMS membranes. With the
objective of minimizing the Youngs Modulus, baking time should be decreased, temperature
increased, and ratio decreased.

With sufficient further study into these areas, a precise, cheap, and simple droplet generation
technique can be made easy to incorporate to the Manifold framework.

13
References
[1] Manifold: An SMT-Based Declarative Language for Electronic and Microuidic Design
Synthesis, M Berzish, A Khan, A Zaman, V Ganesh, and D Rayside, 2016
[2] Volkert van Steijn, et al. Predictive model for the size of bubbles and droplets created in
microfluidic T-junctions Lab on a Chip, issue 19, Oct, 2010
[3] Exploring Microuidic Design Automation: Thin-wall Membrane Regulator, Atulan Zaman,
2016 pp. 40-41
[4] Re-configurable Fluid Circuits by PDMS Elastomer Micromachining, D. Armani ; C. Liu ; N.
Aluru, 1999
[5] Qualitative and Quantitative Evaluation of Two New Histogram Limiting Binarization
Algorithms, J. Brocher, Int. J. Image Process. 8(2), 2014 pp. 30-48

14

You might also like