Upload
others
View
8
Download
0
Embed Size (px)
Citation preview
JUCER Volume VI
May 2017
JUCER (Journal of Undergraduate Chemical Engineering Research)
Stony Brook University
Engineering 208
Stony Brook, NY 11794
Phone: (631) 632-6269
Cover Art: Clockwise from Top Right
Top Right: Surface of 60 nm SBS film on a glass disk after 48 hours of annealing (See Nicole
Passariello, Megan Lenore, Asim Rattu et al., The Effects of SBS Tri-Block Copolymer Surface
Roughness on Protein Adsorption)
Center Right: Graphical model of initial carbon dioxide and water system (oxygen atoms are
red, hydrogen are white, and carbon are cyan) (See Sandhiya Kannan, John Mikhail, Raphael
Prodromou et.al., Pressure-Time Dependence and Simulation of Carbon Dioxide Hydrate
Formation)
Lower Center: Schematic of micellization and gelation of Pluronic F127 in the presence of
Vancomycin (See Kim Nguyen, Jessica Kelly, Michael Callan, Eungyo Jang et. al.
Characterization of Pluronic F127 for the Controlled Drug Release Vancomycin in the Spinal
Column)
Center Left: Diagram of salting-out assisted liquid-liquid extraction of unreacted furfuryl
alcohol monomers from its oligomers and polymers (See Patrick Yang, Carmenn Ooi, Marolyn
Liang, Jeong Suk Buyn et. al., Effect of Anhydrous Sodium Sulfate Salt Loading on Aqueous
Furfuryl Alcohol Washes Extracted from Furfuryl Alcohol Oligomers using Deionized Water)
JUCER is an annual publication by the Chemical and Molecular Engineering program by State
University of New York at Stony Brook.
ISSN: 2373-4221
Author Inquiries:
For inquiries relating to the submission of articles please contact Miriam Rafailovich
1
2
Staff
Editor-In-Chiefs Patrick Yang
Angell Chee
Shweta Iyer
John Mikhail
Webmasters Sandhiya Kannan
Calvin Cheng
Steven Zhu
Faculty Advisor Dr. Miriam Rafailovich
Graduate Advisor Clement Marmorat
Associate Editors Adam Bennett
Veronica Burnett
Jeong Suk Byun
Mike Callan
Calvin Cheng
Vivian Cheng
Jillian Gannon
Henry Ho
Thomas Hurson
Anna Jang
Yunxiao (Shawn) Jiang
Sandhiya Kannan
Jessica Kelly
Steven Krim
Wooyoung Kwon
Christopher Lafergola
Kevin Lee
Megan Lenore
Jian Lin
Jinying Lin
Helen Liu
Jesse Matsuda
Mike McCutcheon
Kim Nguyen
Melanie Noye
Michael Okunewitch
Carmenn Ooi
Nicole Passariello
Raphael Prodromou
Dongni Qiu
Asim Rattu
Peter Ryzyk
Gurkirat Singh
Allen Tran
Danielle Wahl
Joshua Weinstein
Mailun Yang
Joo Yong Yi
Steven Zhu
3
A Letter from the Editors
We are pleased to present Volume VI of the Journal of Undergraduate Chemical Engineering
Research (JUCER). This volume is the culmination of a year of research from the Chemical &
Molecular Engineering class of 2017 at Stony Brook University. We hope that, in writing this
journal, the students have gained valuable insight in how to collaborate with their peers to
achieve scientific progress and how to communicate that progress to others in the academic
community.
JUCER is a peer-reviewed publication; every article included here has been approved by experts
in their respective fields. Each submission was written entirely by a team of undergraduate
students based on their own, unique work. This work involved studying the background of their
topics, performing experiments, and analyzing the data obtained from these experiments using
rigorous scientific principles learned over the course of their undergraduate career.
We would like to thank all of the students for their remarkable effort and the faculty for their
mentorship and guidance throughout this process. We are all incredibly proud of the results of
this endeavor and are delighted to present them here to be read and disseminated publicly.
Sincerely,
John Mikhail Patrick Yang Shweta Iyer Angell Chee
Co-Editors in Chief
Journal of Undergraduate Chemical Engineering Research
JUCER
4
Editor-In-Chiefs Patrick Yang is a graduating senior in the Chemical &
Molecular Engineering program at Stony Brook University
with Business Management specialization and Chemistry
minor. He has co-authored one conference publication through
Materials Today: Proceedings published by Elsevier and
another through ACS Sustainable Chemistry and Engineering
Journal it was also featured on the cover story of the journal.
Some of his research interests include renewable energy like
biomass conversion, catalysis, and polymers nanocomposites.
Following graduation, he will be participating in the Science
Undergraduate Laboratory Internship (SULI) at Argonne
National Lab and is planning to go into industry, possibly
returning for graduate studies.
Angell Chee is a 4th year student at Stony Brook University.
He is majoring in Chemical Engineering, minoring and
specializing in Computer Science. Angell is passionate about
challenging himself with new technologies and hopes to create
a business where he can contribute to the health, fitness and
computer science industries. He will work for Tata
Consultancy Services (TCS) as a software engineering
consultant and is looking forward to this new chapter of his
life.
John Mikhail is a student of Chemical & Molecular
Engineering and Applied Mathematics & Statistics with a
specialization in Computer Science. He does molecular
dynamics research with Professor Gersappe on interfacial slip
in polymers, Professors Koga and Endoh on carbon dioxide
hydrate formation, and Professor Michaelson on protein-
ligand docking. After his Bachelor's, John is pursuing
graduate studies in Chemical Engineering at MIT.
Shweta Iyer is the President of the Stony Brook University
chapter of AIChE. She has interned with Honeywell UOP, GE
Power, and GE Industrial Solutions, serving as a radionuclide
researcher, engineer, and pricing analyst, respectively. She
will be joining BAE Systems post-graduation and will pursue
an MBA.
5
Webmasters
Sandhiya Kannan is a graduating senior in Chemical and
Molecular Engineering with dual minors in Nanotechnology
and Astronomy. Her research experiences include her work at
Princeton University on a fractional distillation apparatus for
the removal of Po-210 from groundwater, an R&D internship
at Chemtura Corp. developing and testing petroleum
additives, her past research at Stony Brook on the synthesis of
gold and palladium nanoparticles, and her thesis project on the
formation of carbon dioxide hydrates. Upon
graduation, Sandhiya will be pursuing her Master's degree in
Chemical Engineering at Columbia University.
Calvin Cheng is a senior majoring in Chemical Engineering
with a specializing in computer science. He works on
simulations of a transdermal drug patch in order to build an
accurate model of a transdermal drug delivery system. He has
also researched under Dr. Taejin Kim to find an efficient
catalyst to help the reduction of NOx from fuel combustion.
After graduation, he plans to pursue a job in simulations and
modeling.
Steven Zhu is a Chemical and Molecular Engineering Major
of the Class of 2017. The research he worked on for his thesis
involved the simulation of a transdermal patch. His
specialization is in Energy and he plans to work in industry
post-graduation.
6
In Memory of Miguel Roncal
We dedicate Volume VI of the Journal of Undergraduate Chemical Engineering Research
to the memory of our beloved friend and classmate, Miguel Roncal. He embodied what it means
to be a true engineer: his energetic nature, ardent attention to detail, unwavering teamwork, and
ingenuity serve as continuous inspiration for our future academic and professional endeavors.
A Chemical and Molecular Engineering major with a Computer Science minor, Miguel
was a dedicated and hardworking student. He consistently scored well on exams and helped anyone
who struggled to grasp the material. Miguel was an excellent scholar and an even better colleague.
The years we were fortunate enough to share with Miguel were filled with happiness and
familial bonding. His many sayings, surprises, hugs, and jokes are forever missed but, greater still,
forever remembered.
7
Do it Up
By Shweta Iyer
A collection of memories from CME
When your car was parked too far, and you’d decided to drive a friend home
You bent the wheel of a bicycle, and laughed giddily away
Despite not knowing how to ride or brake
You never left anyone alone, if even for a minute you could stay.
Broken spokes could be mended; you built unbreakable ties
If a group splintered while crossing a street
You’d turn around and wait;
Even if the first half continued on, you’d never allow a break.
Perhaps the bike was one wheel too many
So naturally only a unicycle would do.
You determinedly fell and rode and fell again
Eliciting laughter from whomever were watching you.
A unique energy and a force of the purest love
Adventures with you are plenty.
You’ll heist toothpicks from the diner to make a proper gift
And ensure no hearts were empty.
Words cannot hope to express what even pictures won’t fully portray
The light, radiance, and gentle humor you embody inscribed in morning rays.
Your cheery nature and infectious laughter unfailingly constant and free
We’ll push up the tears and do it up forever in your memory.
8
Table of Contents Page(s)
9-19
20-29
30-40
41-49
50-57
58-66
78-84
85-91
92-101
102-110
111-121
Title
Characterization of Pluronic F127 for the Controlled
Drug Release Vancomycin in the Spinal Column
Effect of Surface Properties on Pluronic Deposition
Optimization of Synthesis Gas to Methanol
Conversion
The Effects of SBS Tri-Block Copolymer Surface
Roughness on Protein Adsorption
Transdermal Patch Simulation Using the Lattice
Boltzmann Method with Active Diffusion in the Cell
and Lipid Pathways
Effect of Anhydrous Sodium Sulfate Salt Loading
on Aqueous Furfuryl Alcohol Washes Extracted
from Furfuryl Alcohol Oligomers using Deionized
Water
Thermal Conductivity, Electrical Conductivity and
Mechanical Properties of Polypropylene/Graphene
and Polystyrene/ Graphene Nanocomposites
Effects of Graphene Oxide on Proton Exchange
Membrane Fuel Cells
Synthesis and Characterization of Gold-Palladium
Nanoparticles Catalyst For Improved Hydrogen Fuel
Cell Performance
Pressure-Time Dependence and Simulation of
Carbon Dioxide Hydrate Formation
Process Optimization for a Wood Stove with a
Combustion Catalyst
In-situ Water Management for the Optimization of
Methanol Dehydration to Dimethyl Ether
Author(s)
Michael Callan, Eungyo Jang,
Jessica Kelly, Kim Nguyen
Vivian Cheng, Christopher
Lafergola, Melanie Noye
Veronica Burnett, Shweta Iyer,
Steven Krim, Mike
McCutcheon, Gurkirat Singh
Megan Lenore, Nicole
Passariello, Asim Rattu
Angell Chee, Calvin Cheng,
Steven Zhu
Patrick Yang, Carmenn Ooi,
Marolyn Liang, Jeong Suk Byun
Peter Ryzyk, Thomas Hurson,
Joo Yong Yi, Joshua Weinstein
Henry Ho, Jesse Matsuda,
Mailun Yang
Adam Bennett, Helen Liu, Allen
Tran
Sandhiya Kannan, John Mikhail,
Raphael Prodromou
Yunxiao (Shawn) Jiang, Kevin
J. Lee, DongNi Qiu
Jillian Gannon, JinYing Lin,
Danielle Wahl 122-134
Characterization of Pluronic F127 for the Controlled Drug Release Vancomycin in the Spinal Column
Michael Callan1, Eungyo Jang1, Jessica Kelly1, Kim Nguyen1, Clement Marmorat2, Miriam Rafailovich1,2
1Department of Chemical and Molecular Engineering, Stony Brook University, Stony Brook, NY, 11794, USA
2Department of Materials Science and Engineering, Stony Brook University, Stony Brook, NY, 11794, USA
Abstract
Pluronic F127 is a unique hydrogel due to its sol-gel transition properties determined by temperature and concentration. By investigating its rheological properties and drug release capability, F127 can be used as a drug delivery system for the spinal column. In this study, rheological measurements including temperature sweeps and frequency sweeps were performed on various concentrations of F127 to characterize its physical structure at different conditions. For temperature sweeps, higher concentrations of F127 showed abrupt increase of viscosity at lower temperature. For frequency sweeps, the fracture point of Pluronic F127 with and without Vancomycin was observed at 18.42 Hz. The calculated flow rate at 18.42 Hz is (0.53 0.11) 10 ± × -6 m3/s, which is much higher than the flow rate of human spinal fluid. In addition, Pluronic F127’s drug release capability was tested via a flow chamber system where DI water is flowed at a constant rate over the hydrogel with antibiotic additive. UV-Vis spectroscopy was used to characterize the collected flow samples. From the flow test, a slow flow rate of 0.1 mL/min yielded higher release of F127 and Vancomycin as compared to higher flow rates of 0.5 mL/min and 1 mL/min.
Keywords: Pluronic, Vancomycin, hydrogel, drug release, rheology
1. Introduction
Hydrogels are three-dimensional networks formed by polymers that exhibit visco-elastic characteristics widely used in the field of cosmetic, pharmaceutical, and biomedical industries.[1] Pluronic is a triblock copolymer made up of hydrophilic end units, polyethylene oxide (PEO), with a hydrophobic center unit, polypropylene oxide (PPO).[2] Pluronic exhibits unique thermoreversible gelling capability.[3] At low
temperature in solution, Pluronic exists as single chain polymers (unimers) (Figure 1a).[4] As temperature is increased, the solubility of the PPO unit is lowered giving rise to formation of micelles (Figure 1b).[3] The hydrophobic blocks form the inner core while the hydrophilic blocks form the exterior corona of the micelle.[5] As more micelles are formed, the coronas of the micelles overlap leading to gel formation (Figure 1c).[5] Pluronic micelles self-assemble into a face-centered cubic (fcc) lattice structure.[6]
9
Figure 1. (a) Pluronic unimers in solution, (b) micelle formation, (c) micelles arranged in cubic structure.
Pluronic is biocompatible and FDA approved which makes it a desirable material to have as a drug delivery system. Pluronic is a non-cell adhesive material that can be used as a physical barrier to prevent scar tissue formation. Oh et al. studied the drug delivery of ibuprofen in Pluronic F127/F68 crosslinked with alginate as a physical barrier for adhesion prevention using in vitro and in vivo studies.[7] Their results showed a stable gel at 30 °C with sustained release of ibuprofen up to 45% of total loading amount. The Pluronic F127/F68/alginate/Ibuprofen gel was shown to be an effective material to prevent tissue adhesion formation. Therefore, it has been proposed to inject Pluronic after spinal surgery to prevent adhesion while delivering antibiotics.
In the spinal column of adults, there is continuous perfusion of cerebrospinal fluid at a rate of 400 to 600 ml per day,[8] which is equivalent to 0.28 to 0.42 ml per minute. Pluronic’s sol-gel phase transition is dependent upon the concentration and temperature of the material where Pluronic is a solution under a critical gelation concentration and temperature. Therefore, the gel stability of Pluronic in the spinal column is noteworthy to investigate to determine Pluronic’s effectiveness as a physical barrier to prevent cell adhesion.
Vancomycin is an antibiotic used to treat infections such as osteomyelitis.[9] The optimal dosage of Vancomycin as an
injectable solution is 5 mg/mL for adults.[10] This dosage can vary by age and weight of patients. Vancomycin does not impede bone fracture healing.[9] Hence, administering Vancomycin into the spinal column after spinal surgery can prevent surgical site infections.
In this study, we investigated Pluronic F127’s gel stability under various shear flows simulating cerebrospinal fluid flow to determine the dissolution time of Pluronic. Rheological measurements of Pluronic F127 were conducted to obtain critical gelation temperature of Pluronic at various concentrations and to observe the gel structure at increasing frequency sweeps. We conducted a drug release study of antibiotic Vancomycin to determine the drug elution of Vancomycin and the effects of flow on its release. The drug release study was performed using a flow chamber where DI water was flowed through the Pluronic gel containing Vancomycin at constant flow rates of 0.1, 0.5, and 1 mL/min. Samples collected from the flow chamber were then characterized via UV-Vis spectroscopy.
2. Methods
2.1. Chemicals and Materials
Pluronic F127 (MW = 12600 g/mol, Sigma-Aldrich, Saint Louis, MO, USA) and Vancomycin hydrochloride hydrate (CAS-No. 123409-00-7, Sigma Aldrich, Saint Louis,
10
MO, USA) were used as received without any further purification.
2.2 Preparation of Pluronic F127 Hydrogel
Pluronic F127 hydrogels were prepared by mixing together Pluronic F127 in DI water by weight/volume % (w/v %). For preparation of samples, an analytical balance with a precision of 0.001 g was used to weigh the desired amount of Pluronic F127. Samples with varying concentrations of Pluronic F127 in deionized (DI) water were prepared ranging from 1 - 30 (w/v %). Samples were also prepared with the addition of 1% Vancomycin added to the Pluronic F127/DI mixture. All solutions were stirred and kept in a refrigerator for at least one day to achieve homogeneous mixture.
2.3. Rheology of the Pluronic F127 Hydrogel
Rheological characterization of samples were performed on a Bohlin Gemini 150 HR Nano rheometer. Temperature sweeps were conducted using samples of varying Pluronic F127 concentrations (20%, 22.5%, 25%, 27.5%, 30% w/v%) in DI using the Peltier cylinder system with the temperature being controlled by the rheometer’s cooling system with a water pump. Samples were cooled in a refrigerator and then loaded onto the Peltier cylinder surrounded by the temperature control system. The temperature sweeps were carried out from 17 °C to 60 °C at a constant shear stress of 10 Pa and a strain of 10%. The condition of a constant shear stress of 10 Pa was selected for the temperature sweep after a preliminary test showed a plateau of G’ at 10 Pa both at low and high frequencies indicating that the hydrogel maintained its stable network under these conditions. The viscosity (Pa s) was • plotted against temperature (°C) from the data collected in the temperature sweeps. Samples were prepared for frequency sweeps by
placing 3 mL of the hydrogels into polystyrene (PS) petri dishes and placed into a Fisher Scientific IsotempTM IncubatorTM set at 37 °C for a duration of over 10 minutes to allow the samples to gel. Frequency sweeps were performed using the 2 cm Peltier plate spindle under the oscillatory mode of the rheometer. The rotating spindle allowed measurements of elastic modulus (G’), and viscous modulus (G”) to be collected following a logarithmic sweep of frequencies. Frequency sweeps of each modulus were evaluated between 0.01 to 100 Hz at a constant temperature of 37 °C, under a constant shear stress of 100 Pa with a set strain of 10%. A constant shear stress of 100 Pa was selected to identify the fracture point of the gel. The moduli were plotted against frequency.
2.4. Flow Tests for UV-Vis Spectroscopy
The flow tests were conducted in the Fisher Scientific IsotempTM IncubatorTM set at 37 °C. Prepared samples were loaded into a chamber for hydrogel testing that was created by the Garcia Center for Polymers at Engineered Interfaces at Stony Brook University. The schematic of the chamber can be found below in Figure 2.
Figure 2. Schematics of version 1.1 of the Tubing Apparatus for Hydrogel Testing. All dimensions are in inches.
A NE-1000 Single Syringe Pump (NE, Ringoes, NJ) was used to conduct the flow test to collect samples for UV-Vis
11
spectroscopy. A BD 20 mL syringe with a Lver-LoxTM tip (BD, Franklin Lakes, NJ) was loaded with DI and secured on the syringe pump. The chamber was loaded with 13 mL of hydrogel and secured by nuts and bolts. The hydrogel in the chamber sat in the incubator at 37 °C to gel. The syringe pump was then connected to the loaded closed chamber by amber latex tubing with an ID size of 3/16” and a wall size of 1/16” (VWR Scientific, West Chester, PA). The syringe pump was set to mL/min and flow tests of 0.1, 0.5, and 1.0 mL/min were conducted in order to simulate the flow rate range of spinal fluid (0.28-0.42 mL/min). For each single test, three separate samples were collected after 5 mL was collected so that for one test, a sample of 0-5, 5-10, and 10-15 mL were collected separately for UV-vis spectroscopy. UV-Vis of each 5 mL were taken and compared in order to characterize the time dependence of the release. A picture of the running setup can be found in Figure 3.
2.5 UV-Vis Spectroscopy
The UV-Vis spectroscopy of samples were conducted using a Evolution 220 (Thermo Fisher Scientific, Shanghai, China)
Figure 3. The running setup for the flow test. This setup is conducted in a Fisher Scientific IsotempTM IncubatorTM set at 37 °C.
UV-Vis spectrometer. A blank sample of the DI used in the solutions was used to calibrate the spectrometer. Samples were loaded on optical cuvettes and placed into the spectrometer. The absorbance of the UV-Vis spectrometer was performed between 200-400 nm at room temperature.
3. Results and Discussion
3.1. Gelation Temperature of F127
Rheology measurements were performed for Pluronic F127 aqueous solution for various concentrations: 20, 22.5, 25, 27.5, and 30 w/v %. Figure 4 shows the viscosity as a function of temperature from 18 °C to 60 °C. At low temperature, viscosity slowly increases as the copolymer starts to form micelles.[11, 12] Viscosity increases steeply when sol-gel transition takes place due to associations of the hydrophobic cores.[11, 13, 14] Temperature dependence of gel stability on Pluronic concentration was noticeable. With an increasing concentration of F127, the phase transition temperature decreased. This phenomenon is possibly due to a change in aggregation number with temperature and/or change in the micelle formation process at
Figure 4. Viscosity as a function of temperature at various concentrations of F127 aqueous solution (w/v %)
12
different concentration.[15] From the temperature sweeps, we chose the concentration of 30 w/v % to use for following experiments based on gel stability with the lowest opportunity to be liquefied at 37 °C.
3.2. Frequency Sweep on F127 with and without Vancomycin
Frequency sweeps measurements were performed on Pluronic F127 with and without the additive of 1% Vancomycin, and the effects of Vancomycin on the micellar structure of F127 was investigated. Frequency sweeps were conducted within the frequency range of 0.01 - 100 Hz at constant shear (100 Pa) and temperature (37 °C). Figure 5 shows the comparison between the frequency sweeps on F127 with and without Vancomycin. Both F127 systems, with and without Vancomycin, exhibited similar rheological behaviors in the examined frequency range. Regardless of the presence of Vancomycin, elastic modulus (G’) was greater than viscous modulus (G”) throughout the frequency range. This is associated with the highly entangled F127 gel networks, and reflects the gel behavior of 127.[16, 17] At low frequency ranges (0.01 - 10 Hz), G’ for both samples showed low-frequency plateau, and no significant increase was observed with frequency. For the viscous modulus (G”), decrease of G” is observed in both F127 systems. However, decrease of G” for F127 with Vancomycin was observed at lower frequency, which was at 0.01 Hz. Moreover, at this frequency, G’ of F127 with Vancomycin was twice higher than that of F127 without Vancomycin. These indicate that the presence of Vancomycin stabilized the F127 at lower frequency and F127 with Vancomycin exists as a stable gel. At frequency of 18.42 Hz, distinctive dips of G’ were observed in both systems. The possible explanation is that the frequency of 18.42 Hz was the limit for the material to
maintain its stable elastic domain at 100 Pa in shear. Due to this reason, at frequency higher than 18.42 Hz, the material degraded via breakage of micelle entanglements within shear at 100 Pa. As the micelles were no longer physically bonded at frequency higher than the gel fracture point, different slope of G’ were observed due to different forms of micellar solutions.
Figure 5. G’ (open symbols) and G” (closed symbols) moduli versus frequency for 30 w/v % Pluronic F127 with and without Vancomycin.
The frequency of 18.42 Hz was converted into flow rate, and the theoretical fracture point of F127 in fluid flow was obtained. From the frequency of the rotation in the rheometer, angular velocity of the material was determined as follows:
(1)⍵ = 2 · π · f In which is the angular velocity and ⍵
is the frequency of the rotation. By knowingf the angular velocity of the material, theoretical linear velocity was obtained from Equation 2 below:
(2)v = r · ⍵
where is the linear velocity and r is the v radius of the plate in the rheometer (Figure 6). Equations 1 and 2 were combined and rewritten as Equation 3, which gives the linear velocity in terms of the frequency of the rotation:
13
(3)v = 2 · r · π · f
Figure 6. Schematic representation for the theoretical calculation of the flow rate at the fracture point of F127.
Using the Equation 3, the theoretical linear velocity of the material was calculated for the frequency of 18.42 Hz, which is the fracture point of F127 at 100 Pa in shear. According to the Equation 3, the linear velocity of 0.53 0.11 m/s was obtained that ± corresponds to 18.42 Hz. This velocity is equivalent to the speed at which the outer point on the plate was rotating where the highest amount of mechanical deformation occurs. Assuming that 1 10-6 m2 unit area × from the outer point of the plate is moving at the velocity of 0.53 0.11 m/s (Figure 6), ± approximate flow rate of (0.53 0.11) 10 -6 ± × m3/s (= 31.8 6.6 mL/min) was obtained as a ± corresponding speed for frequency of 18.42 Hz. It is important to remark that the flow rate for the hydrogel fracture point is much greater than that of flow test (0.1, 0.5, 1 mL/min) or spinal fluid flow (0.28-0.42 mL/min).[8] This indicates that Pluronic F127 with and without Vancomycin maintains a stable gel. 3.5. Flow Test of Pluronic F127
UV-Vis for dilutions of 1 - 30 w/v % of Pluronic F127 were taken in order to obtain a baseline for comparison (Fig 8.a).
UV-Vis of Pluronic indicates that a higher and sharper peak corresponds to a higher concentration of Pluronic. Additionally, a broader peak range indicates a less pure sample. Thus, samples with lower concentrations of Pluronic are less pure, this might explain the shift observed in Fig 8.a at concentrations of 15% and less. However, since all concentration samples exhibited a peak around 220 nm, the known peak of Pluronic, we were able to create the calibration curve. Subsequently, UV-Vis of the flow tests at 1 mL/min, 0.5 mL/min and 0.1 mL/min were obtained (Fig 8.c). The time-dependence UV-Vis test showed that for each flow rate, all 5 mL samples exhibited similar curves. Thus we conclude that the release of Pluronic is not time dependent.
A qualitative comparison of the UV-Vis graphs showed that a very minimum amount of Pluronic was released during the flow tests. Therefore, a calibration curve using Beer Lambert Law was created using dilutions of 1 - 15 w/v % Pluronic in order to obtain a more accurate curve fit.
a(λ) A = · b · c (4)
where A is the measured absorbance, ) is (λa a wavelength-dependent absorptivity coefficient, b is the path length and c is the analyte concentration.
Figure 7. Calibration curve of 1, 2.5, 5, 7.5, 10, 15 w/v % dilutions of Pluronic F127 in DI.
14
From Figure 7 we can conclude that Beer’s Law was obeyed in the specified concentration range. The regression equation for the calibration curve was
having a correlation0.087x 0.0582y = + coefficient (R2) value of 0.99, and the standard deviation of the y-intercept was found to be 0.0263. A calibration curve with a
R2 value of 0.99 is considered to be linear, therefore using the curve we could reliably find the concentration of pluronic eluted during the flow test experiment. From the UV-Vis of Pluronic F127 dilutions we found that the max absorption peak occurred at 220 nm. Therefore, the absorbance at 220 nm was plotted against each flow rate using the regression equation.
Figure 8. (a) UV-Vis spectrum of 1-30 (w/v%) Pluronic in DI water at 37°C (b) UV-Vis spectrum of 0.001 – 5 (v/v %) Vancomycin in DI water at 37°C (c) UV-vis spectrum of 30% Pluronic F127 in DI water at 37 °C with aliquots of 5%, 2.5%, and 1% Pluronic F127 in DI at 37 °C for reference (d) UV-Vis spectrum of 30% F127 + 1% Vancomycin in DI with flow of 1, 0.5, 0.1 mL/min at 37°C, and UV-Vis spectrum of aliquots of 0.01 %, 0.05%, 0.025% Vancomycin for reference.
15
Figure 9. Concentration of 30 w/v % Pluronic F127 released as a function of flow rate at 0.1, 0.5 and 1 ml/min.
When we quantitatively analyzed the data, the calculated concentration of Pluronic released during the flow test was negative due to the fact that the concentrations of Pluronic were very low so there were many impurities that could have interfered with the UV-Vis spectra. Qualitatively, however, we observed that a minimal amount of Pluronic with concentrations far less than 1%, were released under flow. The readable data point at 0.1 mL/min showed that 0.62% Pluronic had eluted. This is because the micelle structure of Pluronic has ample time to disentangle at lower flow rates which allows the solvent of DI Water to degrade the Pluronic gel. Since there was no significant release of Pluronic, we conclude that our hydrogel is stable under flow and a good carrier for Vancomycin at the investigated flow rate. 3.6 Flow Test of Pluronic F127 and Vancomycin
UV-Vis spectroscopy of 0.001 - 5 v/v % Vancomycin were taken as a baseline (Fig 8.b). The same method used to create the calibration curve for pure pluronic was used to create the calibration curve for pure vancomycin. UV-Vis spectroscopy of the
flow test (taken at 0.1 mL/min, 0.5 mL/min and 1 mL/min) for 30 w/v % Pluronic F127 with 1% Vancomycin in DI water were also taken (Fig 8.d). The time-dependence UV-Vis also showed that for each flow rate, all 5 mL sample exhibited a similar curve. Therefore, we conclude that the release of vancomycin is not time dependent. A qualitative comparison of the UV-Vis of both the aliquots and Vancomycin flow tests also revealed minimal amounts of Vancomycin elution. Therefore a calibration curve was calculated using 0.01 - 0.1 v/v % Vancomycin for a better curve fit.
Figure 10. Calibration curve of 0.01, 0.025, 0.05, 0.1 v/v % dilutions of Vancomycin in DI.
Figure 10 obeys Beer - Lambert’s Law for the specified range. The regression equation for the calibration was found to be
having a correlation8.4253x .0402y = 1 − 0 coefficient (R2) of 0.94, and the standard deviation of the y-intercept is 0.102. An R2 value of 0.94 can be considered linear, therefore the calibration curve was used to plot concentration of Vancomycin against the flow rate. Concentration of Vancomycin released under flow was calculated by finding the absorbance at the known absorbance peak of Vancomycin, 280 nm[18], for each flow test and the concentrations were calculated using the regression equation.
For the flow rate of 1 mL/min we calculated a negative concentration, which is
16
theoretically impossible. We attributed this result to the extremely low concentration of Vancomycin that was released at this flow rate. Therefore the concentration was too minimal for UV-Vis analysis. However, since we obtained good results for the flow rates at 0.1 ml/min and 0.5 mL/min we decided to move forward with the quantitative analysis. The slope of concentration vs. flow rate for Vancomycin was calculated to be -0.0213 ± 0.0029 with a correlation coefficient of 0.98.
Figure 11. Concentration of 1 % Vancomycin in 30% Pluronic F127 released as a function of flow rate at 0.1, 0.5 and 1 mL/min.
Figure 11 shows small amounts, less than 0.02% concentration of Vancomycin eluting into DI water. Since Vancomycin is hydrophilic, it should have an affinity to release into DI water. However, because there is such a small release we can assume that the Pluronic micelle structure is highly entangled such that Vancomycin is trapped within the micelle branches. As a result, we can confirm that Vancomycin interacts with the hydrophilic corona of the Pluronic micelle structure. Furthermore, we can conclude that there is a decrease in the amount of Vancomycin released from Pluronic solution with increasing flow rate. This phenomenon is likely because at higher flow rates, the flow is such that there is not enough time for the micelle structure to disentangle and thus less
Vancomycin is released. In contrast, at slower flow rates the micelles have enough time to disentangle therefore more Vancomycin can be released from the solution. Qualitatively comparing the rate at which release rate is affected by flow rate for both Vancomycin and Pluronic, we observed similar behaviors. For both Pluronic and Vancomycin, there was a decrease in the amount of material released as flow rate was increased. Additionally, there was a very low release of both Pluronic and Vancomycin at 1 mL/min, indicating that both Pluronic and Vancomycin are resistant to fluid flow and remain stable at this flow rate. Thus, the Pluronic gel with Vancomycin will be retained in the spinal column as the flow rate of spinal fluid is 0.28 - 0.42 mL/min.
4. Conclusion
Pluronic F127 hydrogels possess the ability to serve as a carrier for the drug delivery of Vancomycin to the spinal column area. This is due to Pluronic F127’s biocompatibility, FDA approval, and the ability to manipulate the properties so that the release and stability of the hydrogel matrix will be favorable in the spinal column area. The rheological properties of Pluronic F127 were investigated, and flow tests were conducted to study the release of Pluronic and Vancomycin under fluid flow through UV-Vis spectroscopy. From temperature sweeps, the gelation temperature dependence on the hydrogel concentration was observed. As the concentration of Pluronic increased, the temperature of gelation decreased. The temperature of gelation for 30% w/v Pluronic F127 in DI was at 23 °C, while 20% w/v Pluronic F127 in DI was at 34 °C. These values are both below the in vivo temperature of the human body which means that a range of 20% to 30% w/v Pluronic can be injected into a person as a solution at room temperature, and then gels at body temperature. Frequency sweeps were
17
conducted which showed a thixotropic behavior of the hydrogel as G’ and G’’ diverged under a constant high shear stress of 100 Pa. Under the influence of a constant shear stress of 100 Pa, an interesting phenomena was observed after 10 Hz which can be attributed to a restructuring of the crystalline structure of the hydrogel matrix. The addition of 1% Vancomycin to the 30% concentration of Pluronic in DI did not alter the rheological properties significantly. This is important because hydrogels are sensitive materials that are affected by both their components, and the environments that the hydrogels experiences. If a significant change existed by the addition of the Vancomycin to the hydrogel matrix, then a further study would need to be conducted to understand the properties of the hydrogel created, and how those properties can be exploited for the controlled release. The flow tests showed that the slower the flow, the higher the release of the hydrogel components, which can be attributed to the duration for the test to occur, and the swelling of the components out of the hydrogel matrix.
These results show the efficacy of a Pluronic F127 hydrogel as carrier to deliver Vancomycin. The Pluronic F127 hydrogel maintains its stable network under the flow conditions of the spinal column area, and elutes the antibiotic Vancomycin. Further studies need to be conducted to establish further parameters for release that mimic in vivo conditions such as drug delivery rate of the body, pH, and the effects of enzymatic activity.
5. Acknowledgements
We gratefully acknowledge the financial support of the Department of Materials Science & Engineering and the Program in Chemical and Molecular Engineering at Stony Brook University. We also thank the Garcia Center for Polymers at
Engineered Interfaces at Stony Brook University for providing the hydrogel flow test chamber.
6. References
[1] Tatini D, Tempesti P, Ridi F, Fratini E, Bonini M, Baglioni P. Pluronic/gelatin composites for controlled release of actives. Colloids and Surfaces B: Biointerfaces 2015: 135: 400-407.
[2] Singla P, Chabba S, Mahajan R-K. A systematic physicochemical investigation on solubilization and in vitro release of poorly water soluble oxcarbazepine drug in pluronic micelles. In Colloids and Surfaces A: Physicochemical and Engineering Aspects 2016: 504: 479-488.
[3] Unosson J, Montufar EB, Engqvist H, Ginebra M-P, Persson C. Brushite foams-the effect of Tween® 80 and Pluronic® F-127 on foam porosity and mechanical properties. Journal of Biomedical Material Research Part B 2016: 104B: 67-77.
[4] Barba A, d’Amore M, Grassi M, Chirico S, Lamberti G, Titomanlio G. Investigation of Pluronic® F127-Water Solutions Phase Transitions by DSC and Dielectric Spectroscopy. Journal of Applied Polymer Science 2009: 114: 688-695.
[5] Giovanni F, Giovanna M, Marco C, Giulia B, Gianfabio G. Thermosensitive Self-Assembling Block Copolymers as Drug Delivery Systems. Polymers 2011: 3: 779-811.
[6] Kell M, Walther B, Søren H. Effects of PEO−PPO Diblock Impurities on the Cubic Structure of Aqueous PEO−PPO−PEO Pluronics Micelles: fcc and bcc Ordered Structures in F127. Macromolecules 2008: 41(5): 1720-1727.
[7] Oh SH, Kim JK, Song KS, Noh SM, Ghil SH, Yuk, SH, Lee JH. Prevention of postsurgical tissue adhesion by
18
anti-inflammatory drug-loaded Pluronic mixtures with sol-gel transition behavior. Journal of Biomedical Materials Research Part A 2005: 72(3): 306-316.
[8] Sakka L, Coll G, Chazal J. Anatomy and physiology of cerebrospinal fluid. European Annals of Otorhinolaryngology, Head and Neck Diseases 2011: 128(6): 309-316.
[9] Simoes S, Veiga F, Torres-Labandeira J, Ribeiro A, Sandez-Macho M, Concheiro A, Alvarez-Lorenzo C. Syringeable Pluronic-𝛼-cyclodextrin supramolecular gels for sustained delivery of vancomycin. European Journal of Pharmaceutics and Biopharmaceutics 2012: 80(1): 103-112.
[10] Wang Q, Shi Z, Wang J, Shi G, Wang S, Zhou J. Postoperatively administered vancomycin reaches therapeutic concentration in the cerebral spinal fluid of neurosurgical patients. Surgical Neurology 2008: 69(2): 126-129.
[11] Pham T, Liem C, Djabourov M, Ponton A. Mechanisms of micellization and rheology of PEO-PPO-PEO triblock copolymers with various architectures. Journal of Colloid And Interface Science 2008: 382(2): 278-287.
[12] Cohn D, Lando G, Sosnik A, Garty S, Levi A. PEO-PPO-PEO-based poly(ether ester urethane)s as degradable reverse thermo-responsive multiblock copolymers. In Biomaterials 2006: 27(9): 1718-1727.
[13] Jiang J, Li C, Lombardi J, Colby R, Rigas B, Rafailovich M, Sokolov J. In Polymer 2008: 49(16): 3561-3567.
[14] Escobar-Chavez J, López-Cervantes M, Nail A, Kalia Y, Quintanar-Guerrero D, Ganem-Quintanar A. Journal of Pharmacy & Pharmaceutical Sciences 2006: 9(3): 339-358.
[15] Bohorquez M, Koch C, Trygstad T, Pandit N. A Study of the
Temperature-Dependent Micellization of Pluronic F127. Journal Of Colloid And Interface Science 1999: 216(1): 34-40.
[16] Fernández V, Tepale N, Alvarez J, Perez-Lopez J, Macías E, Bautista F, Pignon F, Rharbi Y, Gamez-Corrales R, Manero O, Puig J, Soltero J. Rheology of the Pluronic P103/water system in a semidilute regime: Evidence of nonequilibrium critical behavior. Journal of Colloid And Interface Science 2009: 336(2): 842-849.
[17] Grassi G, Crevatin A, Farra R, Guarnieri G, Pascotto A, Rehimers B, Lapasin R, Grassi M. Journal of Colloid and Interface Science 2006: 301(1): 282-290.
[18] Tariq M, Naureen H, Abbas N, Akhlaq M. Development and Validation of a Simple, Accurate and Economical Method for the Analysis of Vancomycin in Human Serum Using Ultracentrifuge Protein Precipitation and UV Spectrophotometer. Journal of Analytical and Bioanalytical Techniques 2015: 6(2): 239.
19
20
Effect of Surface Properties on Pluronic Deposition Vivian Cheng1, Christopher Lafergola1, Melanie Noye1, Clement Marmorat1,
Miriam Rafailovich1 1Chemical and Molecular Engineering Department, Stony Brook, NY, 11794, USA
Abstract This paper investigates ways in which Pluronic F-127, a biocompatible triblock copolymer,
interacts with substrates based on their surface properties including surface tension, which can be
calculated from the resulting contact angle when F-127 is deposited on a given substrate. This
testing will aim to provide insight on the physics involved when F-127 dries on a given substrate,
since this is commonly seen in topical medications, drug delivery systems, and skin products. By
creating F-127 solutions of different weight percentages, and applying them to spin casted
polystyrene-silicon wafer substrates, the resulting contact angle can be measured in order to
determine surface tension. F-127 solutions were made in 1 weight percent, 5 weight percent, and
10 weight percent batches, and a 15 mg/mL solution of polystyrene was spin casted onto cut
silicon wafer squares. Each F-127 solution was dried onto this substrate, as well as DI water as a
control, and the contact angle was measured as each droplet dried. This is important because one
can determine how F-127 will deposit on a given surface depending on different surface
tensions, as well as with different concentrations of F-127 in solution. Optical microscopy and
atomic force microscopy (AFM) will also be utilized in order to obtain images of the F-127 that
is deposited on the substrate, so the pattern of the deposition can be observed. The coffee ring
effect is observed in the images because as a droplet dries, the height decreases as a function of
time, with the radius remaining the same, causing deposition to occur on the outer rim of the
drying droplet. We hypothesize that we can control the evaporation rate of water by using
varying concentrations of F-127, and that surface tension will affect how F-127 deposits on a
given surface. This research will ultimately serve the purpose of providing an in-depth analysis
of how F-127 deposits on given substrates, so that observed trends can then be applied to how
topical medications, cosmetics, and drug delivery systems will work when being deposited on
human skin.
Keywords
Pluronic, F-127, Droplet, Evaporation, Surface Pattern, Coffee Ring Effect.
1. Introduction Pluronic F-127 is a nonionic triblock
copolymer composed of a central
hydrophobic block of polypropylene oxide
surrounded by two hydrophilic blocks of
polyethylene oxide.
Figure 1. Chemical Structure of F-127 [1].
F-127 can be used as an antifoaming agent,
wetting agent, dispersant, thickener, or
emulsifier. By understanding how the
conditions of a given substrate affect F-127
deposition, it is easier for dermatologists as
well as the cosmetic industry to better
understand how people’s skin will be
affected. Currently, there are many different
products that use F-127, namely in drug
delivery, but in order to understand how
well it will be received by a given
individual, the physics involved with F-127
deposition needs to be researched. F-127
21
solutions were created for varying weight
percentages and spun cast onto PS-silicon
wafer substrates. Then contact angle
measurements were taken for the different
solutions. DI water was used as a control, so
that one could determine how F-127 will
deposit on the surface due to varying surface
tensions. Atomic force microscopy and
optical microscopy was conducted in order
to observe surface patterns of the F-127 that
was deposited on substrate. Temperature
was kept constant as well since this can also
affect the drying process, and adding this
variable into the experiment would only
serve to complicate results. This research
will help to explain how F-127 deposits on
given substrates, so that observed trends can
then be applied to a variety of application in
both the cosmetic and pharmaceutical
industry.
A common example of an
evaporating droplet is a coffee droplet on a
table. After evaporation, the droplet leaves a
ring-shaped stain often referred to as the
‘coffee ring effect’. Droplets placed on solid
surfaces can adopt different shapes ranging
from spherical to flat shaped based on
surface tension, which is excess surface
energy per unit area of liquid surfaces that
are in contact with another material. The
contact angle of the droplet on a solid can be
found by the equilibrium forces due to the
surface tensions at the liquid-gas, solid-gas
and solid-liquid interface.
Figure 2. Droplet Evaporating [2].
In figure 2, an evaporating droplet
which has a contact line pinned on the
surface is shown. The evaporation flux that
diverges at the contact line is represented by
the arrows above the droplet. It is important
to note that while the radius of the droplet
remains constant, the height, h(t), slowly
decrease as the water evaporate over time.
The arrows inside the droplet represent the
flow of the liquid in order to compensate for
the liquid that is evaporating from the edge.
The flow then brings dispersed particle to
the outer rim of the droplet until the droplet
is fully evaporated, leaving a ring shape
around the outer line of the droplet.
Interestingly, when the contact line is pinned
the evaporating is faster and when the
contact angle approaches zero it creates a
rush of particles that accumulates around the
outer rim which can help explain the coffee
ring effect. The top of the droplet is the
coldest due to longer conduction paths from
the substrate, so evaporation occurs faster
along the contact line [2]
Figure 3. Electron Micrograph of a sample of
10% F-127 solution [3].
Individual micelles can be visualized
on the micrograph of the F-127 sample in
figure 3, which shows that it is spherical in
shape and demonstrates that a core-shell
structure is present. This correlates to F-127
consisting of a hydrophobic poly (propylene
oxide) core and a shell comprising of
poly(ethylene oxide) and water. The core is
lighter due to the lower electron density
compared to the shell.
22
A previous study by Yeganehdoust
et al. looked at the kinetic energy and
average pressure of a droplet during the
evolution of the droplet. The study found
that the curvature can be calculated more
accurately when there are a larger number of
particles, which results in a higher surface
tension in that region. The higher surface
tension also correlates to a higher initial
kinetic energy and larger average pressures
[4].
Figure 4. Cross section and 3D projection of
single droplets [5].
The shape of the ink droplets at
different temperatures were observed in a
previous study by Soltman et al and can be
visualized in figure 4. The flux of a fluid to
the edge of the drop can lead to buildup of
solute as the drop evaporates. The geometric
nature of pinning and the curvature at the
drop’s edge, leads to increased evaporation
which can enhance the coffee ring effect.
The study explains how heat is easily
transferred from the substrate to the edge of
the drop which results in improved
evaporation near the edge of the droplet
compared to the center of the drop. In
addition, the study found that increasing the
temperature of the substrate also increases
the amount of solute transferred near the
contact line, while decreasing the substrate
temperature decreases rim evaporation. [5]
This removes the coffee ring formation since
the substrate slows edge evaporation more
than in the center.
Developing a better understanding of
how droplets dry and the coffee ring effect
can help us also better understand how F-
127 will deposit on different substrates as a
result of surface tension. We are specifically
interested in how F-127 will deposit since F-
127 can be used by many different
industries.
2. Materials and Methods
2.1 Creating Solutions for Pluronic Droplets
The solutions for our droplets were
created using Pluronic F-127 purchased
from Sigma-Aldrich and DI water. We
measured 1, 5, and 10 weight percent’s of F-
127 in relation to DI water and stored in test
tubes. A separate test tube was also prepared
for 100% DI water for the testing of a
control droplet.
2.2 Spin Caster
The Spin Caster was used to deposit
an even film over a silicon wafer purchased
from Sigma- Aldrich. The silicon wafers
were cut into 1x1 centimeter squares using
an X-Acto knife. Polystyrene (PS) was
purchased from Sigma-Aldrich. PS solutions
were dropped onto the surface of a wafer
immediately before spin casting. Tweezers
were used to handle wafers because it is
important to avoid creating defects on the
wafer. In total, thirty PS wafers were spin
casted.
2.3 Contact Angle Goniometer
The contact angle goniometer was
used to measure the contact angle and
volume of the droplets of F-127 on PS
wafers. A total of fifteen frames were taken
with 60 seconds in between frames, in order
to track and observe the evaporating droplet.
After the completion of all frames for the
droplet, each frame was individually
23
analyzed by using its contact angle and
volume in relation to the time the frame was
taken. The contact angle goniometer
acquired the core of our data by measuring
the contact angle of a droplet in relation to
decreasing time.
2.4 Atomic Force Microscopy
The surface patterns of each wafer
will be observed using AFM. AFM will
detect the patterns of the particles deposited
from F-127 solution onto the PS wafer.
Using data from AFM, we will be able to
compare the coffee ring effect amongst
different concentrations of F-127. The
coffee ring effect is an indication of particle
movement and deposition, thus advancing
our research on the varying effects of
surface tension on F-127.
2.5 Optical Microscopy
The surface patterns of each wafer
were also observed using optical
microscopy. Images will be taken using
different lenses. The microscopy images
allow for visualization of a dried droplet on
different surfaces. The surface patterns
detected using optical microscopy can help
to explain the coffee ring effect.
3. Results and Discussion
3.1 Contact Angle Goniometer
We started our data collection using
the PS wafers. We conducted trials to
calculate contact angles for 1 wt% F-127, 5
wt% F-127, 10 wt% F-127 and DI water.
For each concentration, we used droplet
sizes of 5 µL and 10 µL.
Figure 5 a. Volume vs. Time for 5 µL samples. The rate of evaporation can be calculated by finding the
slope of each sample.
Figure 5 b. Volume vs. Time for 10 µL samples. From this graph it is clear that the rate of evaporation is
greatest in the 1 wt% F-127, and slowest in 10 wt% F-127.
24
Table 1. The rate of evaporation is compared
between samples. For 5 µL samples, 5% wt
concentration evaporated the fastest and 10% wt
concentration evaporated the slowest. For 10 µL
samples, 1 wt% concentration evaporated fastest
and 10 wt% evaporated the slowest.
Table 1. indicates that for both
droplet sizes of 5 µL and 10 µL, the droplet
containing the highest concentration by
weight of F-127 took the longest to
evaporate. Compared to our control sample
of DI water, droplets with F-127 showed a
trend of evaporating slower. Therefore, a
higher concentration of F-127 hinders the
evaporation of water, thereby indicating that
F-127 delays the reduction of the height of
the droplet.
The 5 µL, 5% wt% F-127 data
deviated from the norm in regards to
evaporation rate due to some possible errors.
While gathering contact angle data, the
reflection from the light onto the droplet
could have skewed the contact angle
goniometer’s ability to calculate the volume
of the droplet. In addition, all 5 µL of the
solution may not have ejected from the pipet
when the droplet was released, thereby
altering from the desired results. The white
overcast in conjunction with an error with
the pipet could have caused the anomaly of
the 5 µL, 5 wt% data.
Time
(s)
Average of Right and Left
Contact Angles (°)
DI
Water
1%
Wt
5% Wt 10%
Wt
0 91.7
+/- 7.4
79.5
+/-
0.5
61.2
+/- 1.4
57.0
+/-
1.0
60 89.2
+/- 5.9
57.9
+/-
1.1
49.8
+/- 1.8
48.6
+/- 0.9
120 86.6
+/- 5.7
59.3
+/-
4.6
46.9
+/- 0.5
45.1
+/- 0.7
180 78.9
+/- 8.1
41.4
+/-
2.1
42.797
+/- 1.1
43.26
+/- 0.8
Table 2 a. Contact Angles for 5 µL samples for
the first 3 minutes. The contact angle was found
using the contact angle goniometer, and by
averaging the right and the left contact angles of
each individual droplet.
The errors found in tables 2 a. and b.
was the standard deviation of a single
sample’s left and right contact angle. The
right and left contact angles should similar,
if not exact since the liquid from the droplet
disperses outward from the center
symmetrically. The reflection from the light
blurred the edges of our droplet in certain
Concentration of F-
127
Rate of
Evaporation
( µL/s)
5 µL
DI Water 0.094
1% wt 0.062
5% wt 0.16
10%wt 0.048
10 µL
DI Water 0.12
1% wt 0.10
5% wt 0.048
10% wt 0.044
25
Time
(s)
Average of Right and Left Contact
Angles (°)
DI
Water
1%
Wt
5%
Wt
10% Wt
0 68.4
+/- 0.9
68.1
+/-
1.3
55.2
+/-
1.5
68.2 +/-
0.3
60 58.7
+/-
16.4
57.6
+/-
1.6
52.4
+/-
1.5
58.2 +/-
0.7
120 54.1
+/- 1.5
53.0
+/-
1.3
49.2
+/-
1.7
53.525
+/- 0.8
180 53.1
+/- 0.1
52.4
+/-
1.0
47.0
+/-
1.4
52.73
+/- 0.5
Table 2 b. Contact Angles for 10 µL samples
for the first 3 minutes.
frames, making it difficult for the
goniometer to exact the contact angles. A
discrepancy between the right and left
contact angle measurements could also have
resulted from existing scratches and
particles on the substrate. The standard
deviation was found by averaging the right
and left angles, subtracting the average from
each number, squaring the results,
calculating the mean of those squared
differences, and lastly square rooting the
mean.
A high contact angle indicates that
the liquid is hydrophobic and displays a low
wettability. A lower contact angle has higher
wetting which means the liquid spreads over
the surface compared to a higher contact
angle. The surface energy is the amount of
intermolecular force created at the surface of
a substrate. [6] For a low wetting substrate,
the surface tension of the liquid is stronger
than the surface energy. Thus, the higher the
contact angle, the lower the surface tension.
Figure 5 c. The volume of four samples of a 5
µL droplet of 5 wt% F-127 compared to its
height.
Figure 5 d. The volume of four samples of a 10
µL droplet of 5 wt% F-127 compared to its
height.
In Figure 5 c. for the 5 µL sample of
5 wt% F-127, it is apparent that the height
remains relatively constant as the volume
decreases. The drying of a droplet begins
when its height begins to decrease and its
liquid begins to flow out in opposite
directions. Since the height remains drops
slower than the control, the droplet can be
categorized as frustrated because the surface
tensions of the substrate hold on to the liquid
at the edges of the droplet. The other
samples of a 10 µL droplet and 5 wt% F-127
shown in figure 5 d. showcase normal
drying conditions.
26
Our contact angle results show that
the droplets with F-127 concentrations have
a smaller contact angle compared to DI
water. Therefore, according to our results,
droplets with F-127 have a higher surface
tension with the polystyrene substrate than
DI water alone does with the same substrate.
3.2 Atomic Force Microscopy (AFM) and
Optical Microscopy
Atomic Force Microscopy was used
to observe how the F-127 deposited on the
PS substrate for each concentration. This
reveals a lot about how the F-127 deposits
on a nanoscale. Optical microscopy is also
an important tool since surface patterns can
be observed, which provides insight as to
how different concentrations of F-127 will
deposit. From the images we obtained it was
clear that concentrations of F-127 that are 5
wt% or higher do not produce clear ridges,
rather, the surface is much more bulky and
rough. This is most likely due to the fact that
concentration was simply too high, hence
why the focus was on concentrations of 1%
and 0.1%, since these concentrations
produced images that were much more
telling about how F-127 will deposit on a
given surface.
Figure 6 a. 1 wt% microscopy image at 10X.
Figures 6 a,b are images found using
1WT% 5 µL F127 on PS substrate. The
Figure 6 b. 1 wt% AFM image, stripes region.
stripes in figure 6 a show that there is some
indication of a surface pattern from F-127
deposition, but there is too much F-127 to be
able to analyze the surface structure in more
detail. The AFM image in figure 6 b shows
one of the stripes in detail, but it is hard to
make out any details because of this bulk
amount of F-127.
Figure 7 a. 5 wt% microscopy image at 10X
Figure 7 b. 5 wt% AFM image, center region.
27
Figures 7 a,b correspond to images of the
5WT% 5 µL F127 on PS substrate. These
images were even more difficult to interpret
due to the higher concentration of F-127.
For this reason, we chose to create samples
for 0.1 wt%. Since this concentration
produced samples that were much easier to
analyze, since they had clearer surface
patterns.
Figure 8 a. 0.1 wt% microscopy image at 5X.
Figure 8 b. 0.1 wt% microscopy image at 50X.
Figures 8 a,b show optical
microscopy images for a 5 uL drop of 0.1
wt% F-127 on a PS substrate. From these
images, it is clear that the lower
concentration of F-127 results in a smoother
surface with more distinct surface patterns.
Interestingly, defects present on the surface
cause the F-127 to wrap around the defects
as it deposits, as a result of increased surface
tension in the area of the defect. This sheds
light on the how surface defects will affect
F-127 deposition which is important in
medical applications that utilize F-127 in
topical medications, or in the cosmetic
industry that use F-127 in topical cosmetic
products. A perfectly smooth silicon wafer
substrate is great in a lab setting, however,
in real-life applications the surface will
rarely be smooth, especially when
considering how unique each individual’s
skin is.
Figure 9 a. 0.1 wt% microscopy image at 5X.
Figure 9 b. 0.1 wt% microscopy image at 50X
(left) and an AFM image of the area (right).
Figures 9 a.b show microscopy and
AFM images of a 5 µL drop of 0.1 wt% F-
127 on a PS substrate. The dark spots seen
on the surface are most likely the substrate
showing through a very thin layer of F-127
deposited in the area. This is most likely due
to how F-127 wets a surface selectively,
since it will not wet a surface uniformly.
When water wets a surface, there is a
uniform coating that occurs, however, with
F-127 it will wet selectively because of the
nature of its amphiphilic structure. This
duality results in this unique surface pattern
when F-127 wets a surface.
4. Conclusions
28
In conclusion, the surface properties
of Pluronic F-127 were investigated in order
to better understand how it deposits on a
given substrate. The contact angles were
determined for different solutions of varying
F-127 concentration in order to observe the
effect of F-127 on the evaporation rate of the
droplet. A larger contact angle represents a
more hydrophobic substance and
demonstrates low wettability. The results
indicate that droplets of the F-127 have a
higher surface tension with respect to the
polystyrene substrate compared to the
control, which is DI water. The surface
properties of F-127 were observed using
AFM and optical microscopy at varying F-
127 concentrations, which showed many
stripes originating from the center of the
droplet for the 1WT% 5 µL F-127 on PS
substrate indicating a change in structure to
a hexagonal shape. The coffee ring effect
can be observed in the microscopy images
since the height of the droplet decreases
with respect to time while the radius remains
constant, which causes the solution to
deposit on the outer rim of droplet. From
figure 9 b, we also learn some interesting
things about how F-127 will wet a surface
since it is not uniformly distributed like
water is. Rather, because of F-127’s
amphiphilic nature small gaps form where
F-127 does not deposit as heavily, most
likely due to its selectivity when it comes to
the surface it is depositing on.
It is difficult to observe these trends
in concentrations of F-127 higher than 1
wt%, which is made clear by figure 7 b,
which shows bulk amounts of F-127,
without any clear surface patterns.
Interestingly, the 0.1 wt% AFM images
provide insight on how surface defects affect
the deposition of F-127. The F-127
deposited in a ring around the defects, in
figure 8 b, shows that increased surface
tension from the defect results in F-127
deposition that is not uniform. Thus, areas of
greater surface tension will have a greater
buildup of F-127 compared to the surface of
the rest of the sample which is shown to be
very smooth and uniform. We had
hypothesized that surface tension would
affect F-127 deposition, and this has been
proven from our research. We also have
proved that we can control the evaporation
rate of water by increasing the concentration
of F-127, since increased concentrations of
F-127 results in a slower evaporation rate of
water. When it comes to applications to the
cosmetic and medical industries, this plays a
vital role in how topical products should be
designed. Everyone’s skin is unique with a
variable number of defects, which is why
this research can lead to further study of
how products can be designed to take
advantage of how surface tension affects
Pluronic F-127 deposition.
Acknowledgements
We are grateful for the financial
support from the Department of Materials
Science & Engineering and the Program in
Chemical and Molecular Engineering at
Stony Brook University through research
funding.
References
[1] "Applications of Thermo-Reversible
Pluronic F-127 Gels in Pharmaceutical
Formulations." J Pharmaceut Science.
Canadian, 27 Nov. 2006. Web. 09 Apr.
2017.
[2] Mampallil, Dileep. "Some Physics Inside
Drying Droplets." Resonance: Journal Of
Science Education 19.2 (2014): 123-134.
Education Source. Web. 12 Oct. 2016.
[3] Lam, Yeng-Ming, Nikolaus Grigorie,
and Gerhard Goldbeck-Wooda. "Direct
Visualisation of Micelles of Pluronic Block
Copolymers in Aqueous Solution by Cryo-
TEM." The Royal Society of Chemistry, 28
May 1999. Web. 22 Feb. 2017
29
[4] F. Yeganehdoust, M. Yaghoubi. H.
Emdad, M. Ordoubadi. “Numerical Study of
Multiphase Droplet Dynamics and Contact
Angles By Smoothed Particle
Hydrodynamics.” Journal of Applied
Mathematical Modelling Vol. 40 (2016):
8493-8512. Academic Search Complete.
Web. 26 Oct. 2016.
[5] Soltman, Dan, Subramanian, Vivek.
"Inkjet-printed Line Morphologies and
Temperature Control of the Coffee Ring
Effect." Engineering Village. Elsevier, 4
Mar. 2008. Web. 16 Nov. 2016.
[6] Yuan, Yuehua, and Randall Lee.
"Contact Angle and Wetting Properties."
Springer-Verlag Berlin Heidelberg, 2013.
Web. 7 Dec. 2016.
30
Optimization of Synthesis Gas to Methanol Conversion
Veronica Burnett1, Shweta Iyer1, Steven Krim1, Mike McCutcheon1, Gurkirat
Singh1
1Chemical and Molecular Engineering Program, Materials Science and Engineering Department,
Stony Brook, NY, 11794, USA,
Abstract
As global markets shift from petroleum usage to other fuel sources, methanol presents itself
as a viable alternative. Methanol is a cleaner burning fuel than traditional hydrocarbons because it
produces no particulates and significantly less carbon dioxide. A common way to make methanol
is by the conversion of synthesis gas over a catalyst. Synthesis gas is a mixture of carbon monoxide
and hydrogen commonly derived from anaerobic digestion of biogas from biomass or from dry-
reforming of methane. This research utilizes the cleaner energy source of syngas to produce
methanol in an optimized conversion process by varying parameters of reactor temperature
conditions and catalyst particle size. Previous research has focused on surface area effects of the
catalysts but fails to thoroughly study reactor conditions in conjunction with the catalyst. In this
experimental design, a pressurized batch reactor was used to flow the syngas mixture over various
copper catalysts supported on zinc oxide and gamma alumina at different operating temperatures.
Two types of catalysts were synthesized by sonication: 1) a nanocatalyst and 2) a mixed oxide,
and compared these against a commercial catalyst. These catalysts were characterized by X-ray
diffraction and their particle size and surface morphology were imaged and measured by
transmission electron microscopy and scanning electron microscopy respectively. The
experimental runs in the batch reactor operated at three different temperatures of 220˚C, 230˚C,
and 240˚C for each catalyst totaling in nine experimental batch runs. Liquid samples were taken
after each run and analyzed by gas chromatography. Based on these analyses, we calculated for
each run the space time yield, the activation energy, and the percent conversion. Our results
showed that the mixed oxide catalyst achieved the highest overall percent conversion of 36.3% at
a reaction temperature of 240˚C and required the lowest activation energy of 138.2 J/mol. The
nanocatalyst and commercial catalyst achieved the next highest conversions of 22.6% and 18.9%
respectively when reacting at 220˚C but had higher respective activation energies of 197.1 J/mol
and 158.0 J/mol. Overall, the mixed oxide catalyst exhibited a trend of higher percent conversion
at higher reacting temperatures, while the nanocatalyst and commercial catalyst exhibited converse
trends of higher percent conversions at lower reacting temperatures.
Keywords: Catalysis, Methanol Production, Metal Oxides
31
1. Introduction
With the rapidly declining reserves of
remaining petroleum oil, there is a high
demand for alternative fuel sources to satisfy
increasing needs for the world’s expanding
population. China and India have recently
decided to shift to “Methanol Economy” and
this is helping renewed interest in methanol
research to help switch the fuel source to
methanol [1] [2].
One of the reasons these countries
and many researchers are studying methanol
as a replacement for traditional fuels is
because methanol is a more environmentally
friendly fuel.
Eq.1 2CH3OH + 3O2 → 2CO2 + 4H2O
In the combustion reaction 1, for
every methanol molecule burned one CO2
molecule is formed [3]. In comparison,
octane is a popular fuel and when burned, as
seen in Equation 2 below, 8 CO2 molecules
are produced when it is fully combusted.
Eq.2 2C8H18 + 25O2 → 16CO2 + 18H2O
Methanol creates less greenhouse
gases when burned, and since there are no
carbon to carbon bonds, there is no
possibility for particulates to be formed [3].
Additionally, the equipment or vehicles that
run on traditional fuels can be easily
converted to run on methanol [4]. For these
reasons, methanol is a great alternative to
traditional fuels.
Unlike fuels such as octane, methanol
is a light compound that can be easily
produced from a mixture of CO and H2, also
known as synthesis gas, instead of being
produced by the petroleum refining process.
Producing synthesis gas is a more
environmentally friendly process than
producing most of the common fuel types.
Traditionally, synthesis gas is synthesized by
dry reforming of methane. One group of
researchers studied the formation of synthesis
gas from the anaerobic digestion of biomass
as an alternative process [5].
The process of forming methanol
from synthesis gas has not been drastically
altered since the 1960’s. The commercial
catalyst in use now is copper on a zinc oxide
support (Cu/ZnO) that operates under
reaction temperatures ranging from 240˚C to
280˚C [6]. There is a search for a better
catalyst for a cheaper operation of the
conversion of synthesis gas to methanol.
Smaller catalyst particle sizes tend to
have the more surface area per unit weight,
which is directly related to an increase in
active sites. There is much research to be
done for the synthesis gas to methanol
conversion process using nanocatalysts at
different operating temperatures. For this
process many different catalysts have been
studied for the effect of their size on activity
or the effect of changing the temperature but
these parameters have not been studied
simultaneously for a single catalyst. We
propose utilizing the most common catalyst,
Cu/ZnO, for this reaction and altering both of
these parameters for true optimization of the
process.
As mentioned before, the process of
methanol synthesis from syngas is not new.
Developed in the 1920’s the process has not
changed much, but with a more active
catalyst that has the ability to reduce the time
and energy input to drive the reaction forward
to increase the per pass conversion to more
than 20%. The specific reaction mechanism
and how the carbon monoxide and hydrogen
interact on different catalytic surfaces are
only being able to be understood in recent
years due to advances in spectroscopic
techniques. Many recent papers in which
methanol was produced from carbon
monoxide include the secondary dimethyl
ether (DME) synthesis reaction [7]. The
focus on methanol as a product has recently
been done using CO2 as a reactant [7]. The
traditional catalysts have problems with
using CO2 instead of CO due to water
32
Fig 1. Inhibitory effects of H2O (O, O) and CO
(☐) on methanol synthesis (O, ☐) and CO
formation (☐) from the CO2 and H2 over a
Cu/ZnO/ZrO2 catalyst. Reaction conditions:
temperature = 523 K, SV = 1.8 x 105 (1/h),
pressure of CO2 + 3H2 = 5 MPa for H2O
addition and 3.2 MPa for CO2 addition. r0 and r
are the rates of methanol synthesis or CO
formation without and with the addition of H2O
or CO to the feed. O: methanol synthesis with
the addition of H2O to the feed. O: CO
formation with the addition of H2O to the feed.
☐: methanol synthesis with the addition of CO
to the feed. [7]
formation via the CO2 route, which has a
detrimental effect on the catalyst. Saito et al
quantified this detrimental effect [7]:
Liang et al [8] were able, through trial
and error, to discover the optimal ratio of
palladium onto zinc oxide of 1:10 onto the
carbon nanotube support, which gives a good
baseline for other metallic loading onto these
materials. Grobmann et al. [9] conducted a
similar experiment with Cu loaded onto ZnO
on carbon nanotubes, and this catalyst, while
showing excellent activity, was not very
stable.
Karelovic and Ruiz [10] conducted an
experiment that provides the largest basis for
our further testing which measured reaction
rates of methanol conversion based on Cu
particle size loaded onto ZnO. They
Fig. 2. Copper loadings, surface area and size of
crystallites obtained from XRD data. Data for
catalysts recovered after the catalytic test. [10]
Fig. 3. Methanol conversion rates of Cu/ZnO
nanocatalysts based on Cu %weight. [10]
synthesized several different catalysts with
the following Cu weight percentages and
characterizations by BET and XRD.
The same authors [10] found the following
methanol conversion rates (Figure 3).
To summarize, the results show the
highest methanol conversion rate for a
Cu/ZnO nanocatalyst with a 15% weight Cu
loading, the particle size was approximately
30 nm. However, their graph cuts off at
33
230˚C for the reaction temperature during
methanol conversion [10]. Our experiment is
novel in that it investigates the methanol
conversion and thus activity of slightly
different catalysts at these higher
temperatures where the conversion rates were
highest. From the graphs in Karelovic and
Ruiz’s experiment, it is unclear what the
following behavior of the conversion rate
graph is going to be; it could continue
exponentially, approach an asymptote, or
peak then decline. We expect a peak then
decline because temperatures that are too
high would cause catalyst sintering and thus
deactivation. Therefore, our experiment will
verify this optimum operating temperature
experimentally while also comparing the
difference between a synthesized
nanocatalyst, a mixed oxide catalyst, and a
commercially available catalyst. It is also
important to note that the aforementioned
experiment synthesized their catalysts using
the citrate method which is a form of wet
chemistry synthesis [10]. Our synthetic
methods will also differ from theirs since we
used sonolysis technique that provides a
much faster catalytic synthesis.
2. Experimental Section
2.1 Materials
Cu/ZnO supported on gamma-Al2O3
industrial catalyst was purchased from
Haldor Topsoe. n-Hexadecane (95%) was
purchased from Alfa Aesar. Copper (II) oxide
nanopowder (<50 nm particle size) was
purchased from Sigma Aldrich. Zinc oxide
powder (<5 micrometer 99.9%) was
purchased from Sigma Aldrich. Methanol
(99.8%) was purchased from Sigma Aldrich.
Polyethylene glycol 400 was purchased from
Sigma Aldrich. Pressurized CO, H2, He, and
N2 gas tanks were purchased from Airgas.
2.2 Experimental Methods
2.2.1 Catalyst Preparation
To prepare the industrial catalyst for
use in the batch reactor, pellets of the Haldor
Topsoe Cu/ZnO/gamma-Al2O3 catalyst were
placed into a mortar and pestle and ground
down to a powder. The powder was the
placed in an oven at 100˚C to dry before
being placed in the batch reactor.
The nanocatalyst was prepared by
taking 3.0183 g of the aforementioned
ground Cu/ZnO/gamma-Al2O3 catalyst with
80 mL of hexadecane and placing them into
a Misonix Liquid Processor Sonicator 3000.
The sonicator was purged with N2 gas for 15
minutes to ensure that there was no oxygen
present and then the sonicator was turned on.
The solution was kept cool using a Thermo
Scientific Neslab Merlin M33 chiller set to
20˚C. Measurements of temperature and
sonicator power were made every 15 min for
over four hours. After the sonication was
complete, the catalyst was separated from the
hexadecane by centrifuge, washed with
hexanes, and then centrifuged again to rinse
away the hexanes. The sonicated catalyst was
then placed in the oven at 100˚C for at least
24 hours to dry before being placed in the
batch reactor.
To prepare the mixed oxide catalyst,
3.97 g of ZnO and 0.047 g of Cu(II)O (1% by
weight) were added with 90 mL of
hexadecane to the sonicator. The sonicator
was purged with N2 gas for 15 minutes and
then the sonicator was turned on. The
solution was kept cool using the chiller set to
20˚C. Measurements of temperature and
power were taken every 15 min for four
hours. After the sonication was complete, the
hexadecane was removed and the catalyst
dried in the same manner as before.
2.2.2 Parr Batch Reactor
A Parr model 4575 batch reactor with
a 300 mL pressure vessel was used to conduct
all reactions. The reactor was connected to
and controlled by a Parr 4848 process
controller. 0.6756 g of each catalyst and 60
mL of polyethylene glycol were added into
34
the batch reactor, which was then sealed and
purged with nitrogen gas. 100 psi CO and 200
psi H2 were added to the reaction vessel and
the reactor was heated to the operating
temperature. Internal temperature, internal
pressure, and jacket temperature were all
recorded every 15 minutes. Each reaction
was run at the specified temperature for three
hours, then the heating jacket was turned off,
and the system was allowed to cool to room
temperature. The final pressure was noted
once the system had cooled, and liquid
samples were taken for analysis.
2.3 Analytical Methods
2.3.1. Gas Chromatography (GC)
For Gas Chromatography, a Gow-
Mac gas chromatograph series 580 was used.
Experiments were run using helium as the
carrier at a 24 ml/hr gas flow rate (measured
with ProFLOW 6000 from Restek). The GC
additionally used chromosorb 101 packing
material. And the GC was set as follows:
column temperatures: 150°C; detector
temperature: 190°C; current: 200mA. To
analyze the results, Clarity PC
Chromatography software was used to collect
the data for 10 min. Each set of data was
compared to a standard of methanol that was
injected under the same operating conditions
2.3.2. X-ray Diffraction (XRD)
X-ray Diffraction (PXRD) testing of
the catalyst powder was carried out using a
Rigaku Ultima IV diffractometer (Cu Kα; λ =
1.5405 Å). Additionally, a D/teX ultra high-
speed 1D position sensitive detector was
used. Our catalysts were dried in an oven at
100˚C before being packed and prepped for
XRD.
2.3.3. Scanning Electron Microscopy
Scanning Electron Microscopy (SEM)
was carried out using a LEO-1550 Gemini
microscope.
2.3.4. Transmission Electron Microscopy
Transmission Electron Microscopy
(TEM) was performed using the JEOL JEM
1400 microscope. It utilizes an acceleration
voltage of 40-120 kV with a magnification of
x5000-2,000,000/x120-4,000. Catalysts were
suspended in a solution of hexadecane and
mixed by sonic bath prior to TEM. Particle
size was found using Image-J software and
averaged across all images for each catalyst.
2.3.5. Energy Dispersive X-Ray Spectroscopy
Energy-Dispersive X-Ray Spectroscopy
(EDS) was jointly carried out under scanning
electron microscopy.
3. Results and Discussion
3.1 Catalyst Characterization
a. X-ray Diffraction (XRD):
Fig. 4. XRD spectra of catalysts tested for
methanol synthesis. (a) Nanocatalyst
Cu/ZnO/gamma-Al2O3.
(b) Mixed Oxide Catalyst: CuO/ZnO
35
The XRD data compares our two
synthesized catalysts, which we expected to
have a similar interference pattern and
crystallinity in the nanoparticle range. The
nanocatalyst showed similar peaks for both
ZnO and CuO standards with broader peaks,
indicating nano-sized crystallinity. However,
the mixed oxide catalyst showed a high
crystallinity with sharp peaks matching the
ZnO standard and no presence of CuO.
b. Energy Dispersive X-ray Spectroscopy
(EDS):
Fig. 5. EDS characterization of all three catalysts.
(a) Nanocatalyst Cu/ZnO/gamma-Al2O3. (b)
Mixed Oxide Catalyst: CuO/ZnO.
To further investigate the presence of
CuO in the mixed oxide catalyst, we
conducted EDS. The nanocatalyst in Fig. 5a
has well-defined Cu, Zn, Al, and O peaks
whereas the mixed oxide catalyst in Fig. 5b
only defines Zn and O peaks. We attribute the
Al in the nanocatalyst to the gamma-alumina
support and did not expect this in the mixed
oxide data. However, this confirms the lack
of Cu or CuO in the mixed oxide catalyst. It
is possible that this would not appear in either
XRD or EDS testing if the Cu did not attach
to ZnO during the sonication process. It is
also possible that CuO may have been present
in such a low amount from the 1% weight
addition that it did not manifest itself on
either spectrum. This occurred for Karelovic
and Ruiz’s study when they used low weight
percents of 0.5% and 1% of Cu on ZnO [10].
c. Scanning Electron Microscopy (SEM):
Fig. 6. SEM images of catalysts. (a) Nanocatalyst
Cu/ZnO/gamma-Al2O3. (b) Mixed Oxide
Catalyst: CuO/ZnO.
Since the EDS tests were coupled
with SEM images, we were also able to view
our catalysts’ surface morphology. The
images in Fig. 6 show that the mixed oxide
catalyst had more defined particles than the
nanocatalyst. There are clear, three-
dimensional agglomerations present in the
images in addition to singular particles
dispersed throughout.
800 nm
800 nm
(a)
(b)
(a)
(b)
36
d. Transmission Electron Microscopy
(TEM):
Fig. 7. TEM images of catalysts and
corresponding average particle size. From left to
right: (a) Nanocatalyst Cu/ZnO/gamma-Al2O3
(b) Mixed Oxide Catalyst: CuO/ZnO (c)
Quenched Nanocatalyst Cu/ZnO/gamma-Al2O3
(d) Quenched Mixed Oxide Catalyst (e)
Quenched Commercial Catalyst
Table 1. Average catalyst particle size
The mixed oxide catalyst was the
only catalyst to not agglomerate after being
subjected to the reaction conditions within
the batch reactor. The general trend for this
catalyst was that a smaller particle size was
correlated with a higher reaction temperature.
While the SEM images showed
agglomerations present prior to the catalysts
undergoing reaction conditions, the TEM
images still yielded some singular particles
that were averaged to calculate the sizes in
the chart. However, in the portions labeled
“Agglomeration” there were no singular
particles present, making measuring particle
size impossible. This meant we could not
determine whether sonication affected our
nanocatalyst enough to significantly change
particle size. In Fig. 7d, e, and f, the images
are from the post 240˚C batch runs for each
quenched catalyst.
Pre
Batch,
nm
Post
240˚C,
nm
Post
230˚C,
nm
Post
220˚C,
nm
Comm-
ercial
[data not
available]
Agglom-
eration
Agglom-
eration
14.7
Nano-
catalyst
12.0 Agglom-
eration
Agglom-
eration
Agglom-
eration
Mixed
Oxide
52.7 48.8 63.9 75.2
a)
c)
b)
e) d)
37
Table 2. Calculated kinetic results from batch data
3.2. Conversion Kinetics
The activation energies for the
nanocatalyst, commercial catalyst, and mixed
oxide catalyst were 197.1 J/mol, 158.0 J/mol,
and 138.2 J/mol respectively for each
reaction temperature. These were calculated
using the Arrhenius equation and finding the
slope of ln(k) vs. 1/T which gives EA/R. This
was then multiplied by R to give the
activation energy for each catalyst
throughout the three different reaction
temperatures. As indicated in the table above
the mixed oxide catalyst at a reaction
temperature of 240˚C showed the highest
percent conversion to methanol of 36.3%.
These represent the optimized reaction
conditions for highest conversion and lowest
initial energy input.
In terms of percent conversion, the
next best catalyst and reaction temperature
are the nanocatalyst at 220˚C. However, the
SEM and TEM data showed large
agglomerations for this, which most likely
contributed to inhibited catalytic activity,
and the main way this was overcome was by
the large activation energy of 197.1 J/mol.
The catalyst with the overall lowest
conversion was the nanocatalyst. According
to the Table 1, its particle size was 12 nm
which is significantly smaller than the mixed
oxide’s size of 52.7 nm. Typically smaller
particle sizes for catalysts yield higher
activity due to higher surface area exposed
during the reaction. However, our results
contradict this and show that depending on
the reaction conditions coupled with each
catalyst, sometimes a larger particle size is
favored for overall higher reaction efficiency.
It is possible that the large agglomeration of
the nanocatalyst could be due to its uniquely
small particle size, thus inhibiting its catalytic
activity.
Catalyst and
Reaction Temperature
MeOH Formed,
mol
Rate,
mol/hr
Space Time Yield,
kg MeOH /
kg catalyst / hr
Conversion, %
Commercial catalyst 240˚C 6.4 2.1 101.2 7.7
Commercial catalyst 230˚C 5.3 1.8 83.8 6.3
Commercial catalyst 220˚C 15.6 5.2 246.6 18.9
Nanocatalyst 240˚C 3.3 1.1 52.8 4.0
Nanocatalyst 230˚C 9.2 3.1 145.4 11.0
Nanocatalyst 220˚C 18.9 6.3 298.8 22.6
Mixed oxide catalyst 240˚C 30.3 10.1 479.0 36.3
Mixed oxide catalyst 230˚C 8.3 2.8 131.2 9.9
Mixed oxide catalyst 220˚C 8.9 3.0 140.7 10.7
38
Fig. 8 Gas Chromatography Data; a) Baseline methanol peak. b) Peaks for Nanocatalyst Post-220˚C.
3.3 Product Characterization
a. Gas Chromatography
Using the GC methods described
above, pure methanol was found to have a
retention time of 6-8 min as shown in Fig. 8.
For all of our liquid samples, we
observed peaks at similar retention times that
matched for a reference methanol sample.
These data confirmed that we successfully
produced methanol in all of our experimental
batch runs. The small peak between 2 and 4
minutes was observed on most samples,
which we attribute to unreacted syngas
remaining in the PEG liquid.
4. Conclusions
After thorough experimentation using
SEM, EDS, GC, XRD, and TEM for analysis
and characterization of multiple catalysts for
methanol synthesis from syngas, we
conclude that the mixed oxide catalyst is
most efficient of all tested when coupled with
a reaction temperature of 240˚C. This
combination of parameters yielded the fastest
conversion and the highest percent
conversion of methanol with the lowest
activation energy. The lack of particle
agglomeration as indicated by TEM is likely
a contributing factor to the success of this
material. This catalyst, like the others, was
tested at 220 ˚C, 230 ˚C, and 240 ˚C. While
the best-performing variation was mixed
oxide catalyst at 240 ˚C, the next best was the
nanocatalyst at 220˚C indicating that the
synthesis of methanol from syngas is a highly
temperature-dependent reaction and varies
(b)
(a)
40
depending on catalyst. At 240 ˚C, the mixed
oxide catalyst exhibited a 36.3% conversion
and formed 30.3 mol of methanol at 10.1
mol/hr. Since the general temperature
dependency trend showed that the
commercial catalyst and nanocatalyst had
poorer conversion at higher temperatures, we
believe their optimal operating temperature
was the lowest in our study of 220˚C. This
may be due to increased agglomeration at
higher temperatures, particularly with the
large gamma-alumina supports. The mixed
oxide catalyst showed the opposite trend of
increased conversion at higher temperatures,
which is why further research is necessary to
identify its optimal reaction temperature,
which may in fact be even higher than 240˚C.
Further experiments should study other
methods of catalyst synthesis exploring our
sonication technique as well as continue to
look into methods for preventing
agglomeration. It would also be beneficial to
continue to study these catalysts in broader
reaction temperature ranges and other reactor
types such as continuous or larger scale.
Acknowledgements
This project was supported by
funding from the National Science
Foundation. The authors thank the Low
Carbon Energy Laboratory at the Advanced
Energy Research and Technology Center for
providing the laboratory and equipment used
in conducting experiments and analysis. The
Material Science and Chemical Engineering
Department at Stony Brook University is
acknowledged for their financial and
academic support of this project. We would
like to thank Dr. Devinder Mahajan and
Nyima Choephell from the Material Science
and Engineering Department at Stony Brook
for their support and mentorship, Chung-
Chueh Chang in the Materials
Characterization Laboratory for the TEM
images, and Dr. Jim Quinn for the SEM,
XRD, and EDS spectra and information.
References
[1] Pti. "Nitin Gadkari Pushes for
Leapfrogging into Methanol Economy." The
Economic Times. N.p., 11 Sept. 2016.
[2] Yang, Chi-Jen, and Robert B. Jackson.
"China's Growing Methanol Economy and Its
Implications for Energy and the
Environment." Energy Policy 41 (2012):
878-84. Web.
[3] Smith, J. M., Van Ness H. C., and Michael
M. Abbott. Introduction to Chemical
Engineering Thermodynamics. Boston:
McGraw Hill, 2005.
[4] Bromberg, L., and W.K. Cheng.
"Methanol as an Alternative Transportation
Fuel in the US: Options for Sustainable
And/or Energy-secure Transportation." U.S.
Department of Energy: Alternative Fuels
Data Center. U.S. Department of Energy:
Office of Energy Efficiency & Renewable
Energy, n.d. Web.
[5] Mota, N., C. Alvarez-Galvan, RM
Navarro, and JLG Fierro. "Biofuels." Biogas
as a Source of Renewable Syngas
Production: Advances and Challenges:
Biofuels: Vol 2, No 3. Taylor and Francis
Online, n.d.
[6] J. B. Hansen and P. E. Hojlund Nielsen,
in Handbook of Heterogeneous Catalysis, ed.
G. Ertl, H. Knözinger, F. Schüth and J.
Weitkamp, Wiley-VCH, 2008, pp. 2920–
2949.
[7] Saito, M., Fujitani, T., Takeuchi, M., &
Watanabe, T. (1996). Development of
copper/zinc oxide-based multicomponent
catalysts for methanol synthesis from carbon
dioxide and hydrogen. Applied Catalysis A:
General, 138(2), 311-318.
39
41
[8] Liang, X., Dong, X., Lin, G., & Zhang, H.
(2009). Carbon nanotube-supported Pd–ZnO
catalyst for hydrogenation of CO2 to
methanol. Applied Catalysis B:
Environmental
[9] Großmann, Dennis, Axel Dreier,
Christian Lehmann, and Wolfgang Grünert.
"Methanol Synthesis over Cu–ZnO
Aggregates Supported on Carbon
Nanotubes." Applied Catalysis A: General,
n.d.
[10] Karelovic, Alejandro, and Patrico Ruiz.
"The Role of Copper Particle Size in Low
Pressure Methanol Synthesis via CO2
Hydrogenation over Cu/ZnO Catalysts." The
Role of Copper Particle Size in Low Pressure
Methanol Synthesis via CO2 Hydrogenation
over Cu/ZnO Catalysts - Catalysis Science &
Technology (RSC Publishing). The Royal
Society of Chemistry, n.d.
40
41
The Effects of SBS Tri-Block Copolymer Surface Roughness on
Protein Adsorption
Megan Lenore, Nicole Passariello, Asim Rattu, Tadanori Koga, Yizhi Meng, Mani
Sen, Derek Rammelkamp, Weiyi Li
Chemical and Molecular Engineering Program, Stony Brook University, Stony Brook, NY 11794
Abstract:
Polymers can be used for a variety of different tasks including biomedical applications.
However, most polymers are inherently incompatible with biomolecules such as proteins and
DNA. There is growing evidence to suggest that surface topography of a polymer can affect the
biocompatibility, due to the effects that the surface morphology has on the protein
adsorption. Greater protein adsorption to the surface is useful in applications such as orthopedic
implants. However, since immune reactions are common in contact medical devices, ways to
reduce protein adsorption are valuable to explore for blood or tissue contact devices. Here we aim
to develop a biocompatible polymer surface using a polystyrene (PS)-block-polybutadiene (PB)-
block-polystyrene (PS) triblock copolymer (SBS). SBS is not biocompatible on its own, but
allows us to change the surface topology by changing the annealing times of the triblock copolymer
or etching.
We prepared SBS thin films (60-80 nm thick) on both silicon substrates and glass
disks. For comparison purposes, PS and PB homopolymer thin films were also prepared on the
substrates. Atomic force microscopy images verified the concept that changing annealing time
effects the surface topography. Protein experiments using fluorescent BSA protein were
conducted to determine the effects of controlling surface topography on protein adsorption. The
results evidenced that the PS and PB homopolymers show very minimal protein adsorption,
while the SBS triblock copolymer show much more protein adsorbed on the surface. In addition,
we found as the annealing times increased, less protein adsorbed onto the polymer surface.
Keywords: Tri-block copolymer, protein adsorption, annealing, atomic force microscopy
1 Introduction
Polymers are widely used in everyday
life in products such as plastic bags, bottles,
rope, piping, packaging and coatings.
Additionally, they have an important role in
research for use in medical devices. Research
is being done to see if modifications to
polymers can either increase or decrease
protein adsorption on their surfaces. Many
have studied surface microtopography, and
its ability to influence cell behavior, and the
effects that changing these surface
micropatterns have on the amount of protein
adsorbed on the surface [2][8][9]. There are
many ways that surfaces can be modified in
order to be patterned, including spincasting,
molding, and sonication. Each of these
methods have their advantages in surface
patterning and can create different sized
surface structures. Previous literature
concludes that patterned surfaces adsorb
more protein on their surfaces than flat
surfaces [2]. Figure 1 shows this principle;
the pattern surface adsorbed 46% more
protein than the flat surface with only an 8%
increase in the surface area [2].
Many polymers are not
biocompatible and therefore cannot be used
for medical devices because they may be
harmful to living tissue [7]. Two polymers
42
Figure 1. The effects of creating a patterned
surface and protein adsorption [2].
used for the present study are polystyrene
(PS) and polybutylene (PB). In order to use
either of these two polymers with less
biocompatibility for uses in the body, they
must be modified for the specific use of the
medical device. For applications with
orthopedic implants, a surface that adsorbs
protein is necessary: however, less protein
adsorption is desirable for uses in blood or
tissue contact devices. Protein adsorption is
important for designs of biosensors, medical
diagnostics devices, and drug delivery
vehicles.
Although some other polymers are
naturally biocompatible, they are often very
expensive; PS and PB are both fairly
inexpensive polymers and would be ideal for
uses where high protein adsorption is needed.
A proposed solution to this problem would be
the use of a PS-block-PB-block-PS triblock
copolymer (SBS) as the coating for
biomedical devices. Ideally, the block
copolymer will adsorb proteins better than
the PS and PB homopolymers. Block
copolymers are appealing for this application
because of their ability to self-assemble into
ordered microdomains [1]. The way that
these copolymers pattern could be taken
advantage of in order to obtain the highest
possible protein adsorption [1].
Block copolymers have been studied
extensively due to their repulsion between
unlike blocks, which leads to the microphase
separation. Most research in this field
involves thermal annealing for long periods
of time in order to minimize surface-
thermodynamic effects, remove volatile
solvents, and maximize the chances of the
surface reaching thermodynamic equilibrium
morphology. Block copolymers can be used
as a template for surface patterning, and there
are many methods for changing their
surfaces. One study, “Fabrication of
nanopatterns using block copolymer and
controlling surface morphology,” tests
surface morphology changes through several
methods [3]. This study shows that
differences in polymer solution and spin
coating speeds affect the thickness of the
polymer layer, which in turn affects the
overall surface pattern. Additionally,
different sonicating solvents created different
surface pattern, and a combination of
solvents can create fingerprint regions along
the surface. The study also showed the ways
that polymer surfaces can be modified, and
the different methods for removing different
substances on the surface [3]. Research has
also shown that SBS can maintain periodicity
of spheres, cylinder, or lamellae [5].
While the chemical composition of
different polymers or copolymers affects the
ability of proteins to adsorb to a surface, the
surface itself can affect protein adsorption.
Previous research has shown that surface
topography on the nanoscale level can have
an effect on protein adsorption. There are a
few ways these surface patterns can be
created. Some of these techniques are photon
ad electron based lithography, nano-imprint
lithography, and etching and glancing angle
deposition [4]. The combination of polymers
and topography can enhance or diminish
protein adsorption on surfaces, resulting in
better surfaces for many medical
applications.
The goal of this experiment is to test
the effects of surface roughness via a thermal
43
annealing process on protein adsorption. We
hypothesize that greater the surface
roughness (i.e., annealing times) less proteins
will adsorb on the surface.
2. Materials and Methods
2.1 Materials
290k molecular weight polystyrene
was purchased from Sigma Aldrich. 38k
molecular weight polybutylene was
purchased from Polymer Source Inc. Toluene
was purchased from VWR. 85k molecular
weight SBS was purchased from Asahi-Kasei
Chemical Corporation. 30 mg/mL FITC-
BSA was purchased from Protein Mods. PBS
was purchased from Invitrogen. Glass disks
were purchased from Ted Pella, Inc. Silicon
wafers were purchased from University
Wafers. 9” Glass Pasteur Pipet was
purchased from VWR.
PS, PB, and SBS were all dissolved in
toluene and left overnight before spincasting
the polymer films. Samples were annealed in
a vacuum oven for varying durations of time
at 130 C. 30 mg/mL FITC-BSA was diluted
with PBS to the desired concentration of
1mg/mL before performing protein
incubation.
2.2 Experimental Methods
2.2.1 Spincasting
We first spincasted the polymers onto
substrates. The polymers used were PS, PB,
and SBS. The polymers were spincasted
(Headway Research, Garland, TX) onto both
glass disks and silicon wafers so they could
be imaged through optical microscopy as
well as atomic force microscopy. Before
spincasting, glass disks and silicon wafers
were cleaned. The glass disks were cleaned
by sonication (Ultrasonic Cleaner, Branson
200) in ethanol for one minute. Silicon
wafers were cleaned using a 1:1:1 solution of
water, hydrogen peroxide, and ammonium
hydroxide for 20 minutes followed by using
a 1:1:1 solution of water, hydrogen peroxide,
and sulfuric acid. After cleaning the glass
disks and silicon wafers, they were placed on
the spincasting apparatus and spun until
excess liquid was removed, generally 30
seconds. Once the excess liquid was
removed, the polymer was spincasted onto
the substrate for 30 seconds at 2500 rpm.
The silicon wafers were cleaned prior
to us working with them. The polymers were
spincasted onto the silicon wafers so that the
thickness of each polymer film could be
recorded using the ellipsometer. The number
of samples spincasted was based on how
many different annealing times were going to
be tested. For each annealing time that was
going to be tested, both a glass disk and
silicon wafer were spincasted.
2.2.2 Annealing
Once the substrates were spincasted,
they were annealed. They were annealed in a
vacuum oven (Sheldon Manufacturers,
Cornelius, OR) at 130 °C. Samples were
annealed for different time periods so that the
effects of annealing time on surface
topography could be observed. The time
periods they were annealed for were 0 hours,
8 hours, 24 hours, and 48 hours. High
temperature annealing of polymer thin films
improves their uniformity [6].
2.2.3 FITC BSA Protein Experiment
After samples are annealed, the
protein incubation was completed on the
glass disks. The protein incubation had to be
performed in a BSL-2 laboratory, which we
were all trained for. A 1 mg/mL solution of
fluorescein conjugate (FITC) Bovine Serum
Albumin (BSA) protein in PBS was diluted
from a 30 mg/mL stock solution. 100
microliters of this dilute solution were
pipetted onto a glass plate that had been
covered in parafilm. On top of each drop the
annealed spincasted sample was placed with
the surface side down, so the polymer
directly interacted with the BSA
44
protein. Each sample was incubated for 20
minutes, then rinsed in water and dried with
compressed air. The dry samples could then
be analyzed through optical microscopy.
2.3 Analytical Methods
2.3.1 Ellipsometer
The ellipsometer (Rudolph Research
AutoEL, Hackettstown, NJ) was used to
determine the thickness of the polymer
film. In order to find the thicknesses of the
films, the polymer must be spincasted onto
silicon wafers, as opposed to glass, because
the laser used to identify the thickness is
refracted off of the silicon surface. The index
of refraction used for SBS is 1.530.
2.3.2 Optical microscope
After the samples were rinsed and
dried with compressed air, images were taken
using a fluorescent optical
microscope. Since the type of BSA used was
fluorescein conjugated, areas glowed bright
green where protein adsorbed to the surface.
The microscope was set to FITC with a
magnification of 10x. The sample was
placed onto the center of the microscope with
the protein side up. The microscope was
adjusted so that the image on the computer is
clear. Once the image was clear, images at
three different locations on the sample were
taken. On the computer program, the color
was adjusted to be fluorescent green.
2.3.3. AFM
Atomic force microscopy was used to
show the surface topography of each
sample. AFM was taken at a variety of
different specifications including wide and
zoom magnification.
3. Results and Discussion:
First, PS and PB homopolymers were
spincasted and protein adsorption was tested
using the FITC BSA protein experiment. The
results of the protein experiment are shown
below in Figure 2. The PS and PB films were
spincasted to be 41 nm and 227 nm
respectively.
3.1 PS and PB Homopolymers
Figure 2. Optical microscope images of PS
(annealed 24 hrs at 130C: 40.8 nm thickness) and
PB (annealed 24 hrs at 80C: 227.7 nm thickness)
at different incubation times compared to glass
disks with and with BSA. untreated w/o BSA (I),
untreated w/ BSA 20 min (II), PS 10 min (III), PS
20 min (IV), PB 10 min (V), PB 20 min (VI)
The results from the protein
incubation of PS and PB both show that less
proteins adsorb to the substrates when there
is a polymer layer, as opposed to the
untreated glass disk, which does not have
polymer layer on it (see Fig. 2). These images
were used a basis for the next test of PS-
block-PB-block-PS triblock copolymer.
3.2 Thick Out-of-Period SBS thin films (78
nm thick)
3.2.1 AFM
45
AFM images were taken of the 78 nm
SBS thin films at three different annealing
times. We made samples with no annealing,
8 hour annealing, 24 hour annealing, and 48
hour annealing. The sample with no
annealing had dust particles on it, therefore
we could not get a good AFM image. The
other three AFM images are shown in Figure
3.
Figure 3. AFM images of SBS (annealed at
130℃: 78 nm thickness) at varying annealing
times. 8 hrs (I), 24 hrs (II), 48 hrs (III)
These AFM images show dewetting,
or the rupture of the thin film, which is caused
by the surface energy differences between the
surface and the film. In order to determine
whether the film dewetted down to the silicon
wafer surface, a scratch test was completed
and the AFM is shown in Figure 4.
3.2.2 Scratch Test
Figure 4. Scratch test AFM height image of SBS
(annealed at 130C: 78 nm thickness) with 40 μm
scale.
The height AFM image shown in
Figure 4 shows that the film did not dewet
down to the silicon wafer surface, based upon
the cross-sectional analysis. The cross-
sectional analysis shows that vertical distance
between the scratched and unscratched
surfaces is 37 nm, and the total film thickness
is 78 nm. Since the vertical distance is only
37 nm, which is not the total height of the
polymer film, we can conclude that the
polymer film did not dewet down to the
silicon surface. Rather, the SBS film show a
terrace structure, as previously reported [10].
We take advantage of this “rough” structure
for the study.
3.2.3 FITC BSA Protein Experiment on the
SBS 78 nm film
The FITC BSA protein experiment
was completed on 78 nm SBS thin films at
four annealing times. The results are shown
in Figure 5.
Figure 5. Optical microscope images of SBS
(annealed at 130C: 78 nm thickness) on glass
disks at varying annealing times. 0 hr (I), 8 hr (II),
24 hr (III), 48 hr (IV)
The images shown indicate that are
changes in protein adsorption when
introduced to varying annealing times. The
bright fluorescent areas show greater protein
adsorption. The problem with this method is
that the filter used on fluorescent microscopy
makes the entire image a green hue, making
it hard to determine which areas of the film
actually have adsorbed protein. From these
images, the no annealing and 8 hour
46
annealing (Fig. 5, I and II) seem to have
adsorbed the most protein, as seen by their
bright green areas. By that same logic, the
darkness of the 24 and 48 hour samples (Fig.
5, III and IV) show less protein adsorption.
3.3 Thinner In-Period SBS thin film (60
nm)
3.3.1 AFM
We first characterized the surface
structures with AFM. 60 nm SBS films were
spincasted, and the AFM images are shown
in Figures 6 and 7.
Figure 6. Wide AFM images of SBS (annealed at
130C: 60 nm thickness) at varying annealing
times. 0 hrs (I), 24 hrs (II), 48 hrs (III)
Figure 7. Zoom AFM images of SBS (annealed
at 130C: 60 nm thickness) at varying annealing
times. 0 hrs (I), 24 hrs (II), 48 hrs (III)
As seen with the 78 nm films,
different annealing times lead to a
topological change in the films. From here,
we tested the protein adsorption to the 60 nm
SBS films to see if there was a correlation
between the surface morphology and the
amount of protein adsorbed.
3.3.2 FITC BSA Protein Experiment on the
SBS 60 nm film
The FITC BSA protein experiment
was completed on the 60 nm thick films. The
samples were incubated for 20 minutes each.
The results can be seen in Figure 8 below.
Figure 8. Optical microscope images of SBS
(annealed at 130C: 60 nm thickness) on glass
disks at varying annealing times. 0 hr (I), 8 hr (II),
24 hr (III), 48 hr (IV)
Figure 8 shows the protein adsorption on the
60 nm films annealed at the four different
times. The exposure time for the samples was
kept consistent at 31 seconds. The bright
fluorescent green spots on the images
indicate protein adsorption to the surface.
The brighter spots on an image indicates that
more protein was adsorbed to the polymer
surface. Figure 8 images I and II have more
protein than images III and IV.
3.3.3 AFM for the SBS thin films on GLASS
In order to view the FITC BSA under
fluorescent microscopy, the thin films needed
to be spincasted onto glass disks. However,
to make the AFM imaging easier and to test
the thickness on the ellipsometer, the films
needed to be spincasted onto silicon wafers.
Since silicon wafers and glass have differing
chemical formulas (SiOx vs. SiO) we could
not logically assume that the interaction
between the polymer and the surface is the
same for the two. In order to test this, AFM
of the 60 nm SBS samples were tested on
glass disks, seen below in Figure 9.
The patterning of the SBS film is
similar on both the silicon wafers on glass
disks, as seen in Figures 7 and 9. Although
there are some differences in the 8 hour, the
47
Figure 9. AFM images of SBS (annealed at
130℃: 60 nm thickness) at varying annealing
times. 8 hrs (I), 24 hrs (II), 48 hrs (III)
24 and 48 hour samples are very similar in
morphology.
3.4 Discussion
From analysis of the AFM image in
Fig. 4, we concluded that the surface did not
dewet down to the silicon. This conclusion
allows us to confirm the fact that we have
created the rough surfaces for the proposed
protein adsorption experiments.
The AFM images of both 78 nm and
60 nm samples show that as annealing time
progresses, surface morphology changes.
Less annealing time creates more island-like
structures, while greater annealing times
shows these islands merging together into a
conglomerate. The darker areas of the AFM
images show the valleys in the film, while the
lighter areas show the peaks.
SBS has a domain spacing of 30 nm,
as clarified by small-angle x-ray scattering
performed by the Koga group. This spacing
refers to the distance between PB blocks in
the SBS triblock copolymer. Films that
follow the rule of being a multiple of the
domain spacing will show symmetrical
wetting. Since the 78 nm sample did not
follow this rule, the surface pattern could not
have symmetrical wetting, which caused
such a terrace structure to occur. Due to this,
we decided to try to make 60 nm SBS films
to note the difference in surface morphology.
There was a visible difference in
morphology between the 60 nm and 78 nm
films. The differences in morphology
contributed to protein adsorption. Figure 5
and Figure 8 show the protein adsorption on
the 78 nm and 60 nm films, respectively.
While it appears that about the same amount
of protein adheres to the samples based on the
annealing time, the patterns at which they
anneal are different. The protein adsorption
that occurs in Figure 8 images I and II
adheres in bands while in Figure 5 images I
and II still have the same amount of protein,
they do not seem to have any type of bands of
proteins. This occurs because of the actual
surface. The surfaces of the 60 nm films are
less random leading to the proteins to adhere
in the same way on the films.
Figures 6 and 7 show the AFM
images for the 60 nm SBS samples with white
dots across the frame. We attribute these
small dots to dust particles on the sample, or
small defects in the sample, but we are
assuming they had no effect on the protein
adsorption on the surface.
Our hypothesis stated that changing
the annealing time of polymer films would
change the amount of protein that adhered to
the surface. More specifically, the amount of
FITC BSA that adsorbed to SBS films would
change based on the annealing times the films
were subjected to. Our results show that the
longer a sample is annealed, the less protein
will adhere to the sample.
4. Conclusion Changing the surface topography of a
polymer film is one way to achieve the
biocompatible properties that a scientist or
researcher may want. Surface topography
can be changed by varying the length of
annealing times. The longer the SBS films
were annealed in the vacuum ovens, the less
protein adsorption to the polymer surface
occurred.
Biocompatible properties are
important in the study of medical devices and
applications. Coatings are needed to assure
that the devices being introduced into the
body are not rejected. Some devices need to
have the ability to adsorb proteins while other
48
devices do not want to adsorb proteins. The
different characteristics depend on where the
device will be going into the body and what
its purpose is. In order to create these
coatings, polymers are used on the devices.
However, many of the polymers that are
currently available are expensive and could
be further optimized. Our research showed
that controlling the surface topography has a
direct effect on the protein adsorption.
Future researchers can examine a
variety of different things. Researchers can
perform the experiment with different
proteins. Using different proteins will allow
for this polymer film to be used on medical
devices entering different parts of the body.
An ideal annealing time can also be observed
by performing the experiment at more
annealing times.
Acknowledgements This research was conducted at Stony
Brook University. Authors would like to
thank Stony Brook University’s Chemical
and Molecular Engineering Professors Dr.
Tad Koga and Dr. Yizhi Meng for their
guidance and the use of their
labs. Additionally, they would like to thank
graduate students Mani Sen, Derek
Rammelkamp, and Weiyi Li for their
assistance in the experiments.
References: [1] Gu, Xiaodan, "Self-Assembly of Block
Copolymers by Solvent Vapor Annealing,
Mechanism and Lithographic Applications"
(2014). Doctoral Dissertations May 2014 -
current. Paper 7.
[2] Chen, Hong, Wei Song, Feng Zhou,
Zhongkui Wu, He Huang, Junhu Zhang,
Quan Lin, and Bai Yang. "The Effect of
Surface Microtopography of
Poly(dimethylsiloxane) on Protein
Adsorption, Platelet and Cell Adhesion."
Colloids and Surfaces B: Biointerfaces 71.2
(2009): 275-81. Web. 10 Oct. 2016.
[3] Alam, Md. Mahbub, Yu-Rim Lee, Jin-
Yeol Kim, and Woo-Gwang Jung.
"Fabrication of Nanopatterns Using Block
Copolymer and Controlling Surface
Morphology." Journal of Colloid and
Interface Science 348.1 (2010): 206-10.
Web. 16 Nov. 2016.
[4] Lord, Megan S., Morten Foss, and
Flemming Besenbacher. "Influence of
nanoscale surface topography on protein
adsorption and cellular response." Nano
Today 5.1 (2010): 66-78.
[5] Kim, G. and M. Libera. "Morphological
Development in Solvent-Cast Polystyrene-
Polybutadiene-Polystyrene (SBS) Triblock
Copolymer Thin Films." Macromolecules,
vol. 31, no. 8, 21 Apr. 1998, p. 2569-2577.
[6] Pandey, Rajiv K., Arun Kumar Singh,
and Rajiv Prakash. "Directed Self-Assembly
of Poly(3,3’’’-dialkylquarterthiophene)
Polymer Thin Film: Effect of Annealing
Temperature." The Journal of Physical
Chemistry C 118.40 (2014): 22943-2951.
Web. 2 Apr. 2017.
[7] Thevenot, Paul, Wenjing Hu, and Liping
Tang. “SURFACE CHEMISTRY
INFLUENCE IMPLANT
BIOCOMPATIBILITY.” Current topics in
medicinal chemistry 8.4 (2008): 270–280.
Print.
[8] Sivaniah, E., Y. Hayashi, S. Matsubara, S.
Kiyono, T. Hashimoto, K. Fukunaga, E. J.
Kramer, and T. Mates. "Symmetric Diblock
Copolymer Thin
Films on Rough Substrates. Kinetics and
Structure Formation in Pure Block
Copolymer Thin
Films." Macromolecules 38.5 (2005): 1837-
849. Web.
[9] Bonduelle, Colin V., Solmaz
Karamdoust, and Elizabeth R. Gillies.
"Synthesis and Assembly of Butyl Rubber-
Poly(ethylene Oxide) Graft Copolymers:
From Surface Patterning to Resistance to
Protein Adsorption.” Macromolecules 44.16
(2011): 6405-415. Web. 5 Feb. 2017.
49
[10] Lambooy, P., T. P. Russell, G. J.
Kellogg, A. M. Mayes, P. D. Gallagher, and
S. K. Satija. "Observed Frustration in
Confined Block Copolymers." Physical
Review Letters. N.p., 01 May 1994. Web. 12
Apr. 2017.
50
Transdermal Patch Simulation Using the Lattice Boltzmann
Method with Active Diffusion in the Cell and Lipid
Pathways
Angell Cheea, Calvin Chenga, Steven Zhua, and Jiaolong Jianga a Materials and Science Engineering Department, Stony Brook University, Room 314 Old
Engineering, Stony Brook, New York 11794, United States
Abstract The transdermal patch is continuing to grow in popularity as a drug delivery system that
shows promising results. Since transdermal patch testing on humans is expensive, time consuming,
and limited due to ethical constraints, a simulation of this system would provide a relatively
cheaper option. An existing model simulates this system but only accounts for drug diffusion in
the skin lipid layer. Our model expanded on this simulation to also involve the drug diffusion
through the skin cell layer. Through this simulation, it was found that the cell layer introduces a
much more rapid drug diffusion through the stratum corneum (SC). A sensitivity test was done on
the lipid diffusion coefficient, cell diffusion coefficient, and partition coefficient to see their effect
on the overall properties of the transdermal diffusion. From these tests, the cell layer was
concluded to be the most important factor in transdermal diffusion. The cell diffusion coefficient
affected the drug diffusion the most while the lipid diffusion coefficient and the partition
coefficient showed negligible changes.
Keywords: Transdermal Patch, Lattice Boltzmann, Stratum Corneum, Drug Delivery, Brick and
Mortar
1. Introduction Computer modeling of transport
processes provided a relatively inexpensive
means to observe how interactions of
molecules on the atomic level affect a system
as a whole. It encompassed computation and
theoretical knowledge in order to create a
model that can simulate real life situations.
Numerical methods have been developed to
model various situations and help in
processing the simulation [1].
In the paper, we used a lattice
Boltzmann model to study the diffusion of
drugs from a drug patch through the human
skin. Getting data from experimentation can
be expensive and are prone to human error.
This, combined with human experimentation
can make it very difficult for scientists to test
for efficient transdermal patches [2]. By
creating a model, scientists will be able to
modify different parameters to find the key
parameters affecting the drug flux or peak
times in the patch. By running the parameters
through numerous simulations, the accuracy
of the human trials conducted can be greatly
improved.
The simulation will involve the
diffusion of drugs from a transdermal patch
to the outermost layer of the human skin, the
stratum corneum (SC). The stratum corneum
is the first barrier any foreign substance will
encounter when trying to enter the human
body. It has proved to be very efficient in
keeping these substances out [3]. In fact, this
barrier is often considered the rate-limiting
step of the drug diffusion.
51
Figure 1. Brick and mortar representation of the
stratum corneum layer
In order to create this model, we first
needed to create a structure that mimicked the
stratum corneum. The brick and mortar
model was is usually employed to model this
skin layer because of the resemblance [4].
When observed under the microscope, the
stratum corneum has many layers of skin
cells stack atop of each other, staggered in
various ways [5][6]. The brick and mortar
model created a similar structure, where the
bricks represented the skin cell while the
mortar represented the lipid layer in between
the skin cells.
To model the movement of the drug
within each layer, we used the lattice
Boltzmann method. The Lattice Boltzmann
Method (LBM) was developed relatively
recently as a mesoscopic method that could
bring together some properties from both
macroscopic and mesoscopic modeling
techniques [7,8,9,10]. This method was often
used to solve incompressible, time dependent
Navier Stokes equations. The advantage was
that it can simulate complex systems such as
multiphase flows and chemical reactions
between different fluids and their
surroundings. Instead of focusing on the
microscopic behavior of the molecules, the
LBM acted on the averaged microscopic
motion. Using the lattice Boltzmann method
and the brick and mortar model, a simulation
can be created to predict the drug flow.
Previous works have already created a
simulation using both lattice Boltzmann and
the brick and mortar model [11,12]. The
model created by Sun Ning only accounted
for the diffusion of the drug through the lipid
layer. Drug flow through the cell layer was
not computed by the simulation. Therefore,
this model was not completely accurate since
it does not represent the whole system.
In our simulation, the drug will be
allowed to flow through both lipid and cell
layer, while also including the diffusion in
between those layers. The data collected from
our simulation will be compared to the data
from the simulation that only includes the
lipid layer. The overall drug flow into the
stratum corneum should increase when the
drug is allowed to diffuse through the cell and
lipid layer as opposed to when the drug is
only allowed to diffuse through the lipid
layer.
2. Methods/Procedure Palabos is an open source
computation fluid dynamics (CFD) solver
that has a heavy reliance on the lattice
Boltzmann methods. Palabos deals with a
large range of topics, some of which are flow
through a porous media, and multi-phase
flow. By having references to alleviate
complex calculations Palabos is the perfect
software to help model a transdermal patch
diffusion.
We created two lattices to help
simulate the drug diffusion through the cell
and lipid layer. These layers are effectively
mirror images of each other. The first lattice
is the cell lattice, where the cells, represented
by brick-like shapes, are given a basic
dynamic for which the drug is allowed to
travel through. Most of the lipid pathways in
this lattice is set to have no dynamics,
meaning the simulation will ignore any
particles in this range and not compute them.
The lipid layer that immediately surrounding
the brick is given a bounce back dynamic.
From this setup, drugs in the cell will stay in
the cell. If they were to travel to the edge of
52
the cell, they will hit a bounce back layer that
will keep it in the cell and away from the lipid
layer.
Figure 2. Location of the types of dynamics
located in each lattice
The lipid lattice is similar except that
the lipid is given the basic dynamic and most
of the cell layer is given no dynamic. The
edge of the brick layer, however, is also
treated as a lipid layer and given normal
dynamics. The cell layer directly under this
edge layer is set to have a bounce back
dynamic. From this setup, the drug in the
lipid layer will stay in the lipid layer, but is
also allowed to transfer into the edge of the
brick layer. This edge is utilized as an
interface which will be used to model the
transport of drugs from lipid to cell layer.
From the edge of the brick layer, the drug will
not be able to transverse any deeper into the
cell since it will be immediately met with the
bounce back dynamic that is given to the
layer immediately under this edge.
The creation of two lattices is optimal
because it breaks a complex problem into
smaller problems, which are significantly
easier to solve. By separating the two lattices
with their respective conditions,
understanding the formation of the model and
managing each of its subparts becomes much
easier.
The resultant interface layer was
created from the two lattices. In both lattices,
the edge of the brick layer is given regular
dynamics. Because of this, a drug density will
exist in both lattices. All the points on the
model where the drug density exist for both
the cell and lipid lattice was regarded as an
interface layer. This interface layer allowed
the transport of drugs from the cell lattice to
the lipid lattice and vice versa. Using this
method, we modeled the transdermal drug
diffusion with two lattices that solely
computed the diffusion through their
respective layer, and an interface processor
that modeled the drug diffusion between
them.
Figure 3: Interface Layout. Interface layer is
shown in black.
A sensitivity test was performed on
the new model to determine which factors
affected the drug flux the most. Three
parameters were tested: diffusion coefficient
of drug through the lipid layer, diffusion
coefficient of drug through the cell layer, and
the partition coefficient of the drug between
the lipid and cell layer. In our simulation we
used coefficients that simulated the drug
fentanyl. A standard control case utilized a
lipid and cell coefficient of 1.2 * 10-7 cm2/s
[13]. The partition coefficient used for the
base case was 0.14. The partition coefficient
controls the amount of drug that travels
between the lipid and cell layers. The lipid
diffusion coefficient and partition coefficient
are the same values used in Sun’s Thesis [11].
We used the same diffusion coefficient for
the cell and lipid pathway for this sensitivity
test.
The total drug amount that was
measured in the patch and SC is recorded in
equivalent lattice units (ELU). This unit
53
represents the total amount of individual
lattices units in the simulation that contains
the drug. It is difficult to translate this drug
amount into physical units. This data was
calculated with the following equation:
𝐸𝐿𝑈𝑝𝑎𝑡𝑐ℎ = 𝜌𝑝𝑎𝑡𝑐ℎ ∗ 𝑊𝑝𝑎𝑡𝑐ℎ ∗ 𝐻𝑝𝑎𝑡𝑐ℎ (1)
𝐸𝐿𝑈𝑆𝐶 = 𝜌𝑆𝐶 ∗ 𝑊𝑆𝐶 ∗ 𝐻𝑆𝐶 (2)
In the above equations, ρ represents
the average drug density in the structure. W
and H represents the width and height
respectively of the structure. Together, they
represent the total area in which the ELU is
being evaluated over.
3. Results
Figure 4a, 4b and 4c showed the
effect fentanyl diffusing through the model at
different cell diffusion coefficients.
The cell diffusion coefficient
variation did not change the diffusion
through the patch but had a difference in the
amount of drug diffused through the SC
layer. At a higher cell diffusion coefficient,
the drug traveled through the system much
quicker. Because of this, the SC contained a
much lower overall concentration of the drug.
The cumulative flux of the patch did
not correlate very well with the cell diffusion
coefficient. A max peak was observed at a
two times base case, but the other cases
showed no changes. More cases needed to be
simulated regarding cell diffusion coefficient
on the overall flux before a conclusion can be
determined.
In Figure 5a, 5b, and 5c, the lipid
diffusion coefficient was another parameter
tested in the simulation.
Figure 4a. Drug distribution in the patch with
different cell coefficients
Figure 4b. Drug distribution in the SC with
different cell coefficients
Figure 4c. Flux for different cell coefficients
54
Figure 5a. Drug distribution in the patch with
different lipid coefficients
Figure 5b. Drug distribution in the SC with
different lipid coefficients
Figure 5c. Flux for different lipid coefficients
Similar to the cell coefficient test, the
patch concentration was largely unaffected
by lipid coefficients variations. An
increasing lipid coefficient decreased the
total drug amount in the SC, which was the
exact same trend observed in the cell
coefficient test. The variation of the lipid
coefficient however did not have as great of
an effect on the total drug in the SC as did the
variation in the cell coefficient. This may be
because the SC is mostly made up of cell
layer, so a change in the lesser lipid layer
would have an overall lesser effect.
The cumulative overall flux of the
drug decreased as the lipid diffusion
coefficient increased. It did follow the same
overall pattern of spiking and a resultant
gradual decrease, but the peak of this spike
seems to be controlled by the lipid
coefficient. The highest peak occurred when
the coefficient was the smallest and
decreased as the coefficient is increased.
Figure 6a, 6b, and 6c showed the
effects of varying the partition coefficient.
Figure 6a. Drug distribution in the patch with
different partition coefficients
Figure 6b. Drug distribution in the SC with
different partition coefficients
55
Figure 6c. Flux for different partition
coefficients
A higher partition coefficient should
allow the drugs to more easily move between
the lipid and cell layers. This will then free
more room in the SC for the drug to diffuse
into from the patch. Unlike the previous two
tests, varying the partition coefficient seemed
to have more of an effect on the drug amount
in the patch.
Similar to the lipid coefficient test,
the overall drug amount in the SC showed
little change when the partition coefficient
was varied. This was an interesting result
since the drug amount in the patch showed
some variation.
A lower partition coefficient caused
the drug to diffuse out the patch more slowly.
Also, a lower partition coefficient caused the
drug amount to peak at a lower value than the
other base conditions. This was apparent
since a lower partition coefficient did not
allow as much drug to diffuse to and from the
lipid and cell and vise-versa.
Cumulative overall flux and partition
coefficient seemed to have a positive
correlation. Similar to the analysis of
cumulative overall flux and lipid partition
coefficient, the overall flux followed the
same general pattern with a sharp peak and a
gradual decline. Unlike the negative
correlation observed in 5c however, an
increase in partition coefficient resulted in an
increase in the peak of the cumulative overall
flux.
4. Discussion
The human SC is primarily made up
of cells, which is also true in our model. The
“bricks” make up a large portion of our
structure. The lipid layer only existed in the
small surroundings between each cell.
Because of this, the cell layer dynamics will
probably cause the most drastic changes in
the results. Variation of the lipid coefficient
in figure 5a and 5b shows that the lipid
coefficient have minimal effects on the total
drug amount in SC and the patch. As seen in
figure 6a and 6b, the partition coefficient also
has minimal changes in the outcome. Since
the partition coefficient controlled the
movement of drugs between lipid and cell
layer, it should have more effect on the total
drug amount than the lipid coefficient, but
less effect than the cell coefficient. Figure 4b
showed that the cell coefficient did indeed
affect the total drug amount in SC by much
more than the lipid coefficient and the
partition coefficient. As a result, the amount
of drug present on the SC was largely
dependent on the diffusion of drugs through
the cell layers in the SC.
The flux of the drugs however, seems
to follow a different trend. According to the
figure 5c, the lipid diffusion coefficient had
the most effect on the flux. The highest lipid
diffusion coefficient resulted in the lowest
peak flux while the lowest coefficient
resulted in the highest peak. With a high lipid
diffusion coefficient, the drug is able to pass
through the lipid layer in much faster. This
gives the drug less time to diffuse into the cell
layer. The diffusion of the drug through the
cell layer is predicted to be the main pathway
for which the drug diffuses. Because of this,
a high lipid diffusion coefficient can actually
be detrimental to the flux because it decreases
the amount of drug that can enter this main
diffusion pathway, the cell layer. The
partition coefficient in figure 6c showed
negligible changes in flux for the range that
we tested. The cell diffusion coefficient data
in figure 4c remains inconclusive. There was
56
no general pattern that can be observed. The
base case, 0.5x base case, and 4x base case all
have the same peak height, but the the 2x base
case observed a sudden 50% jump in peak
height. This could be an optimal point where
the lipid diffusion coefficient and cell
diffusion coefficient may experience some
resonance effect resulting in more flux.
Further testing will need to be done with
these parameters.
Ning Sun’s model of transdermal
drug diffusion only accounted for the
diffusion of drug through the lipid layer. His
results showed that the cumulative flux
increases sharply, then steadily drops off.
Our results in figure 4c, 5c, and 6c all showed
a similar trend observed in Sun’s results [2].
Our results were similar in the sense that we
also have a large initial spike in flux. Adding
a cell layer however, seems to have given our
results a less steep increase when compared
to Sun’s results. The overall flux after this
peak also decreased more rapidly than Sun’s
results. Sun’s results showed the overall flux
to have a steady linear decrease after the
spike while our results in figure 4c, 5c, and
6c all showed a decrease that's more
comparative to an exponential decay.
The cumulative overall flux was
observed to have a positive correlation with
partition coefficient and a negative
correlation with lipid diffusion coefficient.
The lipid diffusion coefficient and
cumulative overall flux’s negative
correlation made sense when looking at the
data in 5b. The highest flux peak correlated
with the highest drug concentration in the SC.
Higher drug concentration in the SC will lead
to a higher flux since there was a greater drug
concentration that would want to diffuse out
of the SC. This pattern should ideally also be
observed in the sensitivity test for cell
coefficient, but that data is inconclusive as is.
Similarly, the partition coefficient
and cumulative overall flux’s positive
correlation agreed with the data in figure 6b.
The highest cumulative overall flux was the
four times base case, which is also where the
SC observed the highest peak in drug
concentration. All these cases followed the
same peak height order for concentration in
SC and cumulative overall flux. A higher
partition coefficient will allow more drugs to
pass from layer to layer. A higher transfer in
between these layers will allow the drug to
spend less time in the SC and therefore exit
the SC in greater numbers. This will speed up
the rate of drug transfer which can explain the
positive correlation observed between the
cumulative overall flux and the partition
coefficient.
5. Conclusion
By adding the cell layer into the
transdermal patch simulation, the resultant
drug in SC, patch, and accumulative flux
changed drastically. Diffusion of drugs
through the lipid layer only seemed to lead to
a steady and slow flushing of the drug from
the SC. The addition of the cell layer
increased this rate dramatically. Ultimately,
the addition of the cell layer in the
transdermal patch sped up the diffusion of
drugs through the system and lead to a steeper
decreases in overall cumulative flux. Our
various sensitivity tests have concluded that
the cell layer did indeed play the largest role
in the diffusion of the drugs. Variation in the
partition coefficient and the lipid coefficient
showed negligible changes when compared
to variations in the cell coefficient. The
diffusion of drugs from a transdermal patch
and through the stratum corneum was
ultimately most dependent on the cell
diffusion coefficient and least dependent on
the lipid coefficient and partition coefficient.
Acknowledgements
This research was supported Stony
Brook University. We would like to thank
Dr. Dilip Gersappe for providing us
mentorship throughout the project. We
57
would also like to thank our graduate
advisor, Jialong Jiang, for guiding us
through the simulation process. We would
also like to thank Miriam Rafailovich and
Clement Marmorat for guiding us through
the research.
References
[1] Rim, J. E.; Pinsky, P. M.; van Osdol, W.
W; Multiscale Modeling Framework of
Transdermal Drug Delivery. Annals of
Biomedical Engineering. 2009, 37, 1217 -
1229
[2] Zhang, C.; Luo H.; Lin, G.; Zhu, Z.;
Zhang, F.; Zhang, J.; Wu, Y.; Luo, J.; Wang,
H.; Transdermal patches for D-threo-
methylphenidate (free base): Formulation
aspects and in vivo pharmacokinetics.
Journal of Drug Delivery Science and
Technology, 2016, 35, 50-57.
[3] Wei, W.; Quan, P.; Liu, C.; Zhao, H.;
Fang, L.; Design of a Drug-in-Adhesive
Transdermal Patch for Risperidone: Effect
of Drug-Additive Interactions on the
Crystallization Inhibition and In  Vitro/In
 Vivo Correlation Study. Journal of
Pharmaceutical Sciences, 2016, 105, 3153-
3161.
[4] Heisig, M.; Lieckfeldt, R.; Wittum, G.;
Mazurkevich, G.; Lee, G.; Non Steady-state
Description of Drug Permeation Through
Stratum Corneum. I. The Biphasic Brick-
and-Mortar Model. Pharmaceutical
Research, 1996, 13, 421-426.
[5] Elias, P. M.; Structure and Function of
the Stratum Corneum Permeability Barrier.
Drug Development Research, 1988, 13, 97-
105.
[6] Harding, C.; The stratum corneum:
structure and function in health and disease.
Dermatologic Therapy, 2004, 17, 6-15.
[7] Sukop, Michael C.; Daniel T. Thorne,
Jr.; Lattice Boltzmann Modeling. An
Introduction for Geoscientists and
Engineers. N.p.: Springer, n.d. Print.
[8]Frapolli, N.; Chikatamarla, S. S.; Karlin,
S. S.; Entropic Lattice Boltzmann Model for
Compressible Flows. Physical Review E,
2015, 92, 1-5.
[9] Amir, B.; Mauzole, Y.; Hara, T.; Grilli,
S. T.; Janben C. F.; The Simulation of
Turbulent Particle-laden Channel Flow by
the Lattice Boltzmann Method. International
Journal for Numerical Methods in Fluids,
2015, 79, 491-513.
[10] He, X.,; Luo, L.S.; Theory of the lattice
Boltzmann method: From the Boltzmann
equation to the lattice Boltzmann equation.
Physical Review E, 1997, 56, 6811-6817
[11] Sun, Ning.; Applications of Lattice
Boltzmann Methods in Complex Mass
Transfer Systems. 2016 (Doctoral
Dissertation)
[12] Chen, L.; Lian, G.; Han, L.; Use of
Bricks and Mortar Model To Predict
Transdermal Permeation: Model
Development and Initial Validation.
Industrial & Engineering Chemistry
Research, 2008, 47, 6465-6472.
[13] Roy, S. D.; Gutierrez, M..; Flynn, G.L.;
Cleary, G. W.; Controlled Transdermal
Delivery of Fentanyl: Characterizations of
Pressure-Sensitive Adhesives for Matrix
Patch Design. Journal of Pharmaceutical
Sciences, 1996, 85, 491-495.
58
Effect of Anhydrous Sodium Sulfate Salt Loading on Aqueous
Furfuryl Alcohol Washes Extracted from Furfuryl Alcohol
Oligomers using Deionized Water
Patrick Yang, Carmenn Ooi, Marolyn Liang, Jeong Suk Byun, Xiaojun Chan,
Taejin Kim
Materials Science and Chemical Engineering Department, Stony Brook University, Stony Brook,
NY, 11794, U.S.A
Abstract Unreacted furfuryl alcohol (FA) monomers were successfully extracted from FA oligomers and
polymers through a liquid-liquid extraction coupled with the salting-out effect. Over three liquid-
liquid extractions, the initial concentration of FA at 65.1 wt.% and decreased to 21.0 wt.%,
showing net extraction of 44.1 wt.% FA out of the organic solution. The recovery and
enrichment of unreacted monomers is vital in reducing the total costs of chemicals as well as
generated wastes. Using deionized (D.I.) water as a green solvent, polar FA was extracted from
mostly nonpolar FA oligomers and polymers using polar D.I. water in a liquid-liquid extraction.
FA oligomers were enriched from 18.1 wt.% to ~ 46.6 wt.% while FA polymers decreased from
16.8 wt.% to 32.4 wt.%. The FA monomers were then extracted from aqueous solutions by the
salting-out effect. The addition of anhydrous sodium sulfate (Na2SO4) was investigated for
reaction time effect and salt loading on the aqueous wash by the salting-out process.
Graphical Abstract:
59
Keywords: Furfuryl Alcohol, Salting-out Assisted Liquid-Liquid Extraction (SALLE),
Anhydrous Sodium Sulfate, UV-Vis Spectroscopy
1. Introduction
Due to increasing energy
consumption and depletion of fossil fuels, it
is important to utilize renewable resources
for future energy. Biomass conversion is one
of promising pathways to develop renewable
energy. Among all possible sources of
biomass, lignocellulosic biomass, or second
generational biomass currently has the most
interest. First generation biomass includes
edible crops (e.g., corn and sugarcane) to
produce ethanol through a fermentation
process, while lignocellulosic biomass
conversion into fuels or chemicals does not
directly compete with food production. The
abundance of lignocellulosic biomass is
another drive for the development. It was
reported that the annual consumption of
fossil fuels can be fulfilled by 5 - 8% of
annual production of lignocellulosic biomass
[1].
Among all the platform chemicals
that are derived from lignocellulose, furfuryl
alcohol (FA) draws the most attention due to
its wide application in many other useful
chemical products, such as FA oligomers
(OFA) and FA polymers (PFA). Oligomers
or polymers can be synthesized through
dehydration and condensation reactions of
FA.2 Due to their strong heat stability, low
flammability, high acidity and basicity
resistance, PFA is widely applied in the
synthesis of polymer nanocomposites which
can be further applied as adsorbents,
membranes, catalysts, electrodes in fuel cells,
lithium batteries, and electric double-layer
capacitors [2-3].
Traditionally, liquid acid catalysts
such as sulfuric acid and phosphoric acid
have been used in FA oligomerization and
polymerization reactions [4-7]. Even though
homogeneous acid catalysts provide higher
conversion rates, heterogeneous catalysts
(e.g., metal oxides and zeolites) are
preferable due to its ease in catalyst
separation and then recycling [8]. Selectivity
of OFA and PFA can be controlled by
varying reaction time and temperature during
the acidic catalyzed FA condensation
reaction [9]. From the economic and
environment (chemical waste reduction)
points of view, unreacted furfuryl alcohol
should be separated from the products and
reused for OFA or PFA productions.
Compared to other separation
techniques such as membrane separation and
distillation, liquid-liquid extraction is proved
to have reduced extraction time, lower cost
on disposal and reagent purchases on
separation of desired solutes from solution
[10]. Due to the polarity characteristic, it can
be hypothesized that water will only react
with FA molecules and extract them from the
non-polar OFA/PFA solutions. Furthermore,
by using salting-out method, FA monomer is
separated from the aqueous solutions. The
interaction of salt ions with water molecules
causes water molecules unavailable to react
or bind to FA which is later separated into a
different phase [11]. The aim of this work
was to use the salting-out assisted liquid-
liquid extraction in the separation of FA from
OFA/PFA solution. By Hofmeister series, the
extractability of salts increased in the
sequence: SO42- ≈ CO3
2- > CH3COO- ≈ Cl-
[12]. Hofmeister series states that sulfate
anion has a relatively high extractability
compared to other anions. Therefore,
anhydrous sodium sulfate (Na2SO4) was
chosen in this study.
2. Experimental Section
2.1 Materials
Furfuryl alcohol (FA, 98%), sodium
sulfate (anhydrous, ≥99.5%) and methanol
(99%) were obtained from Sigma-Aldrich
60
and used without further purification for the
oligomerization reaction, salting out
experiment, and gas chromatography
respectively. Deionized (D.I.) water was
obtained from Thermo Scientific and was
used for the liquid-liquid extraction.
2.2 Methods
2.2.1 Furfuryl Alcohol Oligomerization
FA oligomers (OFA) were
synthesized over 1.2 g of molybdenum oxide
(MoO3) with 3.00 g of FA at 100oC for six h
and ambient pressure. The oligomerization
process was conducted in the 7 mL clear
glass vials placed within an aluminum block
designed to hold these glass vials which was
covered by heating tape (HTS/Amptek
1.30A, 156W). After six hours, the vials were
immediately transported into ice water and
held for five minutes to stop any further
reactions. To separate the oligomer solution,
which includes unreacted FA monomer, from
the solid catalyst, the vials were centrifuged
for five minutes at 4000 rpm and then the
solution was carefully pipetted off the top.
Furfuryl alcohol oligomers were then stored
in the refrigerator for further use.
2.2.2 Liquid-liquid Extraction
Liquid-liquid extraction of unreacted
FA monomers from its oligomers and
polymers was conducted using D.I. water as a
solvent. A constant 2:1 ratio by volume of
D.I. water to oligomerized solution was
applied during each liquid-liquid extraction.
Plastic centrifuge test tubes 15 mL)) with
built in plastic lids were used to conduct
liquid-liquid extraction. As an example, FA
oligomers were added to the plastic
centrifuge tube to the 4 mL mark and then
topped with 8 mL of D.I. water, until a total
volume of 12 mL was reached. The
oligomerized FA solution and D.I. water was
fully mixed by vigorous shaking for ten
minutes before being transported to a
centrifuge (UNIX Powerspin LX Centrifuge).
To counterbalance the centrifuge, another
centrifuge tube containing 12 mL of water
was used. The solution was then centrifuged
at 4000 rpm for five minutes to fully separate
the solution into two layers: an aqueous (top)
and an organic (bottom) solution. The
separated aqueous layers from each
extraction were then stored in separate large
glass bottle for use during the salting-out
separation. Remaining oligomer solution was
again washed with D.I. water and the
aqueous top layers were separated using the
same method described before. This liquid-
liquid extraction step was repeated three
times. Remaining oligomerized solution was
stored in the refrigerator for further
characterization.
2.2.3 Salting-out loading effect
The loading of anhydrous sodium
sulfate salt was varied from 0.45 g to 0.70 g
at room temperature with 4.00 g of the
aqueous wash solution that was separated
during the liquid-liquid extraction process.
To avoid salt agglomeration, the FA aqueous
solution was immediately mixed with sodium
sulfate at the beginning of the salting-out
process. After the salt had fully dissolved, the
solution was allowed to stand so that the
solution reached equilibrium. The organic
layer that is salted-out at equilibrium either
forms on the bottom of the vial or at the top
of the aqueous solution depending on salt
loading.
2.3 Analytical Methods
2.3.1 Gas Chromatography
Gas Chromatography/Mass Spectrum
(GC/MS) was used to identify the FA
oligomer composition. Tests were conducted
using the Perkin Elmer Clarus 680 GC/MS
equipped with a Perkin Elmer SQ8T mass
detector. The injector temperature of the
GC/MS was 250 °C and the split ratio 30:1.
Gas Chromatography (GC) was conducted
using a Perkin Elmer Clarus 680 equipped
61
with a flame ionization detector with
split/splitless injector. The GC has an Elite-
5MS capillary column (30 m x 0.25 mm x 1.0
um) with helium as the carrier gas. GC/FID
coupled with the GC/MS data can help
quantify the percent composition of the
oligomerized solution made up of unreacted
FA monomers and FA short chain oligomers
and long chain polymers.
2.3.2 UV-Visible Spectroscopy
Qualitative analysis of FA remaining
in the aqueous solution after the salting-out
process was performed using by Ultraviolet-
Visible (UV-Vis) Spectrometer (Evolution
220 UV-Vis Spectrophotometer, Thermo
Scientific). The UV-Vis spectrum was
obtained over the range of 200 - 800 nm
wavelength with scanning intervals of 1 nm.
Samples were prepared and tested within a
disposable semi-micro UV-cuvette
(BrandTech Scientific Inc) of dimensions
12.5 mm x 12.5 mm x 45 mm.
3. Results and Discussion
3.1 Extraction of FA from Oligomer Stock
Solution
Figure 1 shows total volume of the
organic layer decreasing as the number of
liquid-liquid extractions completed increased.
After three washes, the volume of organic
layer decreases from 3.8 ± 0.1 mL to 1.8 ±
0.1 mL. A decrease in the total volume of the
organic layer is due to interaction between
D.I. water and the hydroxyl group (-OH) of
FA. The polar end of FA allows movement of
FA molecules from the organic phase to the
also polar aqueous one, creating a partition of
FA between both layers. Apart from FA
oligomers that contain a terminal hydroxyl
group, the rest of the oligomers and long
chain polymers that are mostly nonpolar will
not have a high partition into the aqueous
layer and remain in the organic layer [13].
Figure 1. Total volume of the organic phase after
three liquid-liquid extractions using D.I. water.
Using gas chromatography coupled
with mass spectroscopy (GC/MS), a total of
five dimers and two trimers of furfuryl
alcohol were observed in the oligomerized
solution. The five FA dimers that were
observed were 2,2’-Difurylmethane, 2-(2-
furylmethyl)-5-methylfuran, Difurfuryl ether,
4-Furfuryl-2-pentenoic acid γ-lactone, 5-
Fufuryl-Furfuryl Alcohol, and two FA
trimers observed are 2,5-Difurfurylfuran and
2,2’-(Furylmethylene) bis(5-methylfuran).
The data obtained from GC/MS provides
helpful information in percent composition of
the solution through the identification of
corresponding retention time peaks of FA
and its dimers and trimers in the GC/FID
analysis.
Through the combination of data
collected from both GC/MS and GC/FID, the
composition of initial and final oligomer
solutions were determined as shown in
Figure 2. After a total of three liquid-liquid
D.I. water extraction washes, the
concentration of FA shows a decrease from
62
65.1 wt.% initially to 21.0 wt.%. This
indicates that D.I. water can be used
effectively as a polar solvent to extract water
soluble/polar unreacted FA monomers from
water-insoluble non-polar FA oligomers and
long chain polymers. We also observed that
the concentration of five dimers and two
trimers of FA increased from 18.1 wt.% to
46.6 wt.% while long chain FA polymers
increased from 16.8 wt.% to 32.4 wt.%
demonstrating that the FA was successfully
extracted from its oligomer and polymer
products.
Figure 2. A comparison of percent composition
of initial and final FA oligomerized mixtures,
including compositions of FA, FA dimers &
trimers, and FA long chain polymers.
3.2 Extraction of FA from Aqueous Solution
The reaction time effect for the
salting out effect was qualitatively
investigated and is shown in Figure 3 (a-f).
Wannachod et al.9 experimented with the
effect of salt on liquid-liquid extraction,
looking at the tie-line data for a mixture of
furfuryl alcohol, n-butanol, and water at
298.15 K. The authors observed that with
increasing sodium chloride (NaCl)
concentration in the system, the salt proved
to be beneficial in increasing the separation
of FA from water. In hydration theory, salt
ions bind with water molecules leading to a
decrease in the solubility of FA in water
leading to higher separation of FA in water.
Figure 3 (a-b) shows salt loadings of
0.45 and 0.50 g respectively, where the
organic layer is observed forming at the
bottom of the solution during the first two
hours. As the reaction time increases from
one minute to two hours, the amount and the
sizes of the sphere of the organic layer
separated from the aqueous washes increases.
The sizes of the spherical organic layer that
formed at bottom of the aqueous layer also
increased with increasing salt loadings.
Based on Figure 3 (c-f), salt loadings of 0.55,
0.60, 0.65, and 0.70 g respectively, the
organic layer is observed forming at the top
of the aqueous layer instead of the bottom.
With higher salt loadings, the aqueous
solutions could reach equilibrium at a much
faster rate.
In Figure 3 (b) with salt loading of
0.50 g, the solution reached equilibrium after
twenty hours. The organic layer turned from
a bunch of small bubbles at the bottom to a
much larger sphere with a few smaller ones
in the middle of the vial. The location where
the organic layer forms changes from the
bottom at a salt loading of 0.45 g to the top
with a 0.55 g loading. With a salt loading of
0.50 g, the organic layer formed
approximately at the middle of the aqueous
63
Figure 3. Addition of anhydrous sodium sulfate loadings from 0.45 g to 0.70 g with 0.05 g increment in
4.00 g aqueous wash over a period of 20 hours at room temperature (from left to right, a to f).
layer. This phenomenon suggests that the
movement of the organic layer, from the
bottom to the top of the aqueous solution, is
due to the increasing density of the aqueous
layer caused by increased salt loading.
The visual difference between the initial
aqueous wash solution and after anhydrous
sodium sulfate is added is presented
quantitatively in Figure 3 (a-f). Initially the
aqueous wash solution was an opaque light
yellow color, that yellow color being
attributed to FA and its oligomers/polymers
that were dissolved within the water during
the liquid-liquid extraction. However, the
addition of salt increases the transparency of
aqueous washes over time as the salting-out
effect occurs. The transition of an opaque
light yellow color of the batch aqueous wash
to a transparent aqueous layer as the reaction
time increases indicates achievement of
equilibrium in the system. The same
phenomenon could be observed when the salt
loading in the aqueous wash solution
increased. Higher salt loadings shortened the
total time needed to achieve equilibrium in
the solution. The salt solution reaches
saturation somewhere between a salt loading
of 0.65 to 0.70 g which is indicated by the
formation of salt at the bottom of the vial in
the 0.70 g sample.
64
Figure 4. UV-Vis spectra of aqueous washes with increasing salt (anhydrous sodium sulfate) loadings of
0.45 g to 0.70 g . For reference, a spectrum without any salt loading (0.00 g) is also shown.
3.3 Purity and Quantity of Extracted FA
with Increasing Salt Loading
Shown in Figure 4 are a series of
UV-Vis spectra collected between 300-800
nm of various aqueous wash solutions with
increasing sodium sulfate salt loadings. The
UV-Vis spectrum for the original aqueous
wash before any salt was added (0.00 g)
shows a spectrum that is much more shifted
to the right than any of the other samples
with salt loadings. On further inspection of
the data between wavelength ranges of 345 -
360 nm, the spectrum for the aqueous wash
shifts towards the left with increasing salt
loadings. This shift shows the trend that as
the concentration of FA in solution
decreases, the absorbance at a given
wavelength also decreases. As the
salt loading increases from 0.45 to 0.70 g,
the remaining quantity of organic FA
oligomer solution containing FA that was
dissolved in the aqueous washes decreases
which is indicated with the shift in spectra
towards the left. The spectrum shift is
relatively constant as the salt loading
increases with the exception of 0.50 - 0.55 g
when the FA layer moves from the bottom
to the top.
To make sure that the addition of
anhydrous sodium sulfate salt loadings did
not have any effect on the UV-Vis spectra
collected, a sample of pure D.I. water and
two other salt solutions samples were
prepared and tested to compare the
absorbance of the solutions using UV-Vis.
The two salt solutions were prepared by
dissolving 0.45 and 0.70 g of anhydrous
sodium sulfate in 4.00 g of D.I. water, the
same salt loading to the amount of liquid
used for the salting out of the aqueous wash
layers. There is no noticeable difference in
the spectrum between all three samples.
There is almost no absorbance detected for
the samples with salt dissolved inside. This
data suggests that the salt loading has no
effect on the final aqueous wash UV-Vis
spectrums collected. Based on Figure 4, a
higher salt loading decreases the amount of
the FA that is remaining within the aqueous
wash solution. This relationship suggests
that with increasing salt loading, more FA
was extracted and removed from the
aqueous wash. Therefore, increasing salt
loading is shown to help to enhance the
extraction of FA and other organic
compounds from the aqueous wash solution.
65
4. Conclusions
Unreacted FA monomer remaining
from the molybdenum oxide catalyzed
oligomerization reaction forming OFA and
PFA products was separated using a two-
step procedure: liquid-liquid extraction
followed by salting-out process. As the
number of extraction increased, the volume
of organic layer decreased which indicates
that FA monomer can be separated from the
organic solutions by D.I. water. The FA
concentration in the organic layer decreased
from 65.1 wt.% to 21.0 wt.% after liquid-
liquid extraction. Consequently, unreacted
FA monomers were separated from the
aqueous solutions using anhydrous sodium
sulfate for the salting-out effect. As salting-
out time and salt loading increased, the FA
collection efficiency increased. In addition
to the quantity of extracted FA, a higher salt
loading shortens the time needed to achieve
FA and D.I. water separation equilibrium.
The movement of the eluting unreacted FA
layer was observed to be forming in
different locations with different salt loading
which indicates a different density of salt
solutions. The UV-Vis spectroscopy
provided proof of the presence of unreacted
FA monomer extracted in the aqueous salt
solution. The qualitative data which
indicates the decreasing FA in the aqueous
layer with increasing salt loading was
provided and shown by the shifting of UV-
Vis spectrum towards the left. This
qualitative data shows that the amount of FA
being extracted is increasing with higher salt
loadings.
Acknowledgements
We gratefully acknowledge the financial
support from the Department of Materials
Science and Chemical Engineering at Stony
Brook University. We would also like to
thank the Advanced Energy Research and
Technology Center (AERTC) at the Stony
Brook University for providing us with the
facilities. We would especially like to thank
Dr. Miriam Rafailovich and Clement
Marmorat for all their help and guidance of
senior thesis course.
References
[1] Elgharbawy, A. A.; Alam, M. Z.;
Moniruzzaman, M.; Goto, M. Ionic liquid
pretreatment as emerging approaches for
enhanced enzymatic hydrolysis of
lignocellulosic biomass. Biochemical
Engineering Journal 2016, 109, DOI
10.1016/j.bej.2016.01.021.
[2] Wang, H.; Yao, J. Use of poly (furfuryl
alcohol) in the fabrication of nanostructured
carbons and nanocomposites. Industrial &
Engineering Chemistry Research 2006, 45
(19), DOI 10.1021/ie0602660.
[3] Marefat Seyedlar, R.; Imani, M.;
Mirabedini, S. M. Curing of poly (furfuryl
alcohol) resin catalyzed by a homologous
series of dicarboxylic acid catalysts:
Kinetics and pot life. Journal of Applied
Polymer Science 2016, 133 (43), DOI
10.1002/app.44009.
[4] Conley, R. T.; Metil, I. An investigation
of the structure of furfuryl alcohol
polycondensates with infrared spectroscopy.
Journal of Applied Polymer Science 1963, 7
(1), DOI 10.1002/app.1963.070070104.
[5] He, L.; Li, D.; Dong, D.; Yao, J.; Huang,
Y.; Wang, H. Effects of polymerization
conditions on the properties of poly (furfuryl
alcohol) composite membranes. Journal of
Applied Polymer Science 2012, 124 (4),
DOI 10.1002/app.35356.
[6] Chan, X.; Roy, A.; Ooi, C.; Yang, P.;
Morais, F.; Kim, T. Catalysts Loading Effect
of Tungsten Oxide Catalytic Furfuryl
Alcohol Oligomerization. Materials Today:
Proceedings 2016, 3, DOI
10.1016/j.matpr.2016.01.106.
[7] Kim, K.; Jeong, J.; Rahman, M.; Zhu, E.;
Mahajan, D. Characterizations of furfuryl
alcohol oligomer/ polymerization catalyzed
66
by homogeneous and heterogeneous acid
catalysts. Korean Journal of Chemical
Engineering 2014, 31 (12), DOI
10.1007/s11814-014-0322-x.
[8] Burket, C.L.; Rajagopalan, R.; Marencic,
A. P.; Dronvajjala, K.; Foley, H. C. Genesis
of porosity in polyfurfuryl alcohol derived
nanoporous carbon. Carbon 2006, 44 (14),
DOI: 10.1016/j.carbon.2006.05.029.
[9] Chan, X.; Nan, W.; Mahajan, D.; Kim,
T. Comprehensive investigation of the
biomass derived furfuryl alcohol oligomer
formation over tungsten oxide catalysts.
Catalysis Communications 2005, 72, DOI:
10.1016/j.catcom.2015.08.027.
[10] Ebrahimzadeh, H.; Yamini, Y.;
Kamarei, F.; Shariati, S. Homogeneous
liquid–liquid extraction of trace amounts of
mononitrotoluenes from waste water
samples. Analytica Chimica Acta 2007, 594
(1), DOI: 10.1016/j.aca.2007.05.013.
[11] Wannachod, T.; Hronec, M.; Soták, T.;
Fulajtárová, K.; Pancharoen, U.;
Arpornwichanop, A. Effects of salt on the
LLE and tie-line data for furfuryl alcohol—
n-butanol–water at T= 298.15 K. Journal of
Molecular Liquids 2016, 218, DOI:
10.1007/s10953-017-0619-6.
[12] Valente, I. M.; Gonclaves, L. M.;
Rodrigues, J. A. Another glimpse over the
salting-out assisted liquid-liquid extraction
in acetonitrile/water mixtures. Journal of
Chromatography A 2013, 1308, DOI:
10.1016/j.chroma.2013.08.014.
[13] Chan, X.; Yang, P.; Ooi, C.; Cen, J.
Orlov, A.; Kim, T. ACS Sustainable
Chemistry & Engineering 2016, 4 (8), DOI:
10.1021/acssuschemeng.6b01067.
Pages 67-77 currently unavailable
78
Thermal Conductivity, Electrical Conductivity and
Mechanical Properties of Polypropylene/Graphene and
Polystyrene/ Graphene Nanocomposites Peter Ryzyk1, Thomas Hurson1, Joo Yong Yi1, Joshua Weinstein1,
Miriam Rafailovich1, Xianghao Zuo1, Vladimir Samuilov 2 1Chemical and Molecular Engineering Department, Stony Brook University, Room 314 Old
Engineering, Stony Brook, New York 11794, United States 2Materials and Science Engineering Department, Stony Brook University, Room 314 Old
Engineering, Stony Brook, New York 11794, United States
Abstract
In recent years, many studies have been conducted regarding polymer
nanocomposites, namely polymers doped with various forms of graphene. The impact
toughness, tensile strength, and thermal conductivity of Polypropylene/ Graphene and
Polystyrene/ Graphene nanocomposites were compared at different Graphene concentrations
varying from 5% to 25%. The electrical conductivity of Polypropylene/ Graphene
nanocomposites of the same composition was also determined. The goal of these tests was to
find an optimal composition that enhances the electrical and thermal conductivity as well as
tensile strength of polypropylene without compromising the impact toughness for the
potential application of a battery anode. At 10% loading of GNP, an over 100% increase in
thermal conductivity and a 60% increase in tensile strength were achieved with only a 4%
loss to impact strength. The electrical conductivity increased from 260 S/m at 10% loading of
GNP to 2 * 107 S/m at 15% loading of GNP. The results conclude that the nanocomposites
with a load between 10% GNP and 15% GNP by weight displayed ideal material
characteristics.
1. Introduction
Polymer nanocomposite research
often has many different objectives, but a
common goal of creating materials with
improved properties in any of many
possible categories. Flame retardancy,
electrical and thermal conductivity smoke
suppression, tensile strength, impact
strength, glass transition temperature,
melting point, elasticity, etc; there are
applications where each of these properties
may be deemed more valuable. The
addition of additives to create a
nanocomposite is the most common way
to alter these properties. However, it is not
without drawbacks, and not very
predictable [2]. Thus, experimental
research is necessary.
Usually, the addition of additives
will disrupt at least partially the matrix of
the polymer [1]. This is a good example
because creating disruptions is a good
technique for fire retardancy and smoke
suppression, as the combustion may hit a
very small layer of non-combustible
graphene nanoparticles (GNPs), slowing
combustion [1]. However, this same layer
may be a point of weakness in terms of
impact strength, where it may break along
that layer.
Our research dealt with this issue
among others. The earlier stages of our
project attempted a composite using
Polystyrene (PS). PS is a polymer with a
density of 1.052 g/cm3 and a molecular
weight of the repeat unit of 104.15 g/mol
[15]. The PS did not handle the disruption
acceptably, and the result was a brittle
79
material with few possible useful
applications.
In this research, we originally
proposed to use graphene as a fire
retardant by synthesizing it with
polystyrene into a nanocomposite.
Graphene is a 2-dimensional layer of
carbon that has outstanding properties of
high thermal conductivity, superior
mechanical stiffness, and large specific
surface area [13]. There is previous
research that used functionalized graphene
oxide and obtained improved fire safety
data using a masterbatch-melt blending
approach, but we used pure graphene
nanoparticles in this research. We
expected that addition of high graphene
content into polystyrene can effectively
improve fire safety and mechanical
property of the nanocomposite. To
determine the most effective
graphene/polystyrene concentration ratio,
several samples with different
concentrations were prepared and tested.
Our project goal from the very
beginning was to use GNPs in unusually
high doses in a nanocomposite to improve
properties. Preliminary research suggested
up to 40 wt% GNPs could be added to a
polymer nanocomposite for some
potentially extraordinary properties.
Initial testing with polystyrene
produced a very brittle material, so much
so that it was difficult even to create molds
of for testing. In response, we decided to
try polypropylene (PP) as our principal
polymer. Polypropylene has a low glass
transition temperature which allows it to
act more like a fluid at room temperature,
making it better suited for these kinds of
composites [3]. PP has a density of 0.86
g/cm3 and a molecular weight of the repeat
unit of 42.08 g/mol [15]. PP has a
crystalline structure as opposed to the
amorphous structure of PS. The difference
in the degree of crystallinity attributes to
higher heat resistance and sharper melting
points. Moreover, a highly crystalline
structure is prone to being more brittle due
to the compactness and static organization
of the molecules [16].
The result was immediate
improvement and a promising material.
We then created samples of the same
graphene concentrations as the PS
nanocomposites but conducted with PP,
and were able to compare the two
materials afterwards. We successfully
improved properties in tensile strength and
thermal and electrical conductivities
without noticeable detriment to impact
toughness.
2 Experimental Section
2.1 Materials
XGNP-H-5 Graphene
Nanoplatelets were purchased from XG
Sciences. The nanoplatelets have an
average particle diameter of 5 microns.
Polypropylene pellets were purchased
from Amco Plastic Materials Inc.
Polystyrene pellets were purchased from
Amco Plastic Materials Inc.
Both the PP and PS
nanocomposites were prepared in the
following loadings of GNPs by weight:
0%, 5%, 10%, 15%, 20%, and 25% using a
Brabender mixer filled to its capacity of
200g run at 256 °C for twenty minutes.
The nanocomposites were then pelletized
by hand and molded to meet the various
testing standards using the appropriate
molds and a Carver Hot Press at 256 °C
and compressed to 6000 psi for five
minutes.
2.2 Experimental Methods
2.2.1 IZOD Impact Toughness Test
The nanocomposites were molded
into the appropriate ASTM D256-10e1
standard mold with dimensions 64 x 12.7 x
3.2 mm with a notch in the center and then
tested on a Testing Machines Inc. Monitor
Impact Tester that records the maximum
energy absorbed by the sample in J/m [4].
80
2.2.2 Tensile Strength Test
The nanocomposites were molded
into the appropriate ASTM D638-14
standard dogbone shaped mold and tested
on an Instron 5542 Single Testing Column
System. A plot of force versus stroke is
returned and from that the Young’s
modulus is calculated [5].
2.2.3 Thermal Conductivity Test
Thermal conductivity of the
samples was measured on a Unitherm
Model 2022 Thermal Conductivity
Instrument using the ASTM E1530
Guarded Heat Flow Meter Method. The
thermal conductivity tests were performed
at 25 °C and recorded in W/(m*K) [6].
2.2.4 Electrical Conductivity Test
The Electrical conductivity of the
samples was measured by cutting portions
of the nanocomposite into 1 cm2 blocks
and attaching two wire probes to each end
and using an EC meter to read the
conductivity.
Table 1. Various compositions of
nanocomposites after mixing the polymer and
graphene for both PP and PS.
3. Results and Discussion
The polymers, once combined with
graphene powder, turned from
clear/translucent white to black throughout
at every concentration due to the natural
black/grey color of graphene. This can be
assumed to mean the GNPs were present
on the surface covering the polymer and
within the matrix itself [7].
Five samples of each
concentration of GNPs for both
nanocomposites were prepared for the
IZOD impact test for consistency. After
the test was performed, the average impact
strengths were recorded in J/m and
averaged with their respective errors.
Using the results from Table 1,
impact strength of PP/GNP and PS/GNP
nanocomposites were plotted for
comparison with their respective
concentrations in Figure 1.
Both concentrations were found to
have an overall decreasing trend in impact
strength. This trend may be due to the
incorporation of graphene in the polymer
matrix, disrupting the crystal structure. It
is important to note that the shape of the
molds
for this test included a notch in the middle
along the edge to propagate fracturing the
material. Thus, the combination of the
disruption and the shape of the molds are
possible contributors to this decrease.
Beyond a load of 10 wt% GNP,
fluctuation in impact strength is present in
both nanocomposites – more so in PS.
This can be attributed to experimental
error such as surface deformities in some
of the samples. Most notably, the PS/GNP
composites undergo a 48% reduction in
impact strength when GNP concentration
is increased from 5 to 10%. On the
contrary, the impact strengths of the
%
GNPs
% Polymer
(PP and
PS)
Weight
GNPs
(g)
Weight
Polymer
(g)
0 100 0 200
5 95 10 190
10 90 20 180
15 85 30 170
20 80 40 160
25 75 50 150
81
PP/GNP nanocomposites were reduced by
only 8.6%.
Figure 1. Impact strengths of PP/GNP and
PS/GNP nanocomposites obtained from IZOD
impact test.
Similarly, five dogbones samples
of each concentration of GNPs for both
nanocomposites were molded for tensile
testing. Stress (extension %) and strain
(MPa) were recorded every 0.1 seconds.
45 to 50 data points from the elastic region
– the region before the material begins to
yield – were plotted on a stress versus
strain graph. The slope of the line these
points created is known as the elastic or
Young’s modulus [8]. The moduli for each
sample were calculated and then averaged
for each concentration with their
respective errors.
The elastic moduli for both
nanocomposites were then plotted against
each other for comparison in Figure 2.
The polystyrene nanocomposite
immediately starts a trend in decreasing
elastic modulus as GNP concentration is
increased. A study done by Qiu et al.
suggested that adding a small
concentration of functionalized graphene
oxide (FGO) (1 wt%) to polystyrene
increased the tensile strength by 23.3%
[9]. At such small concentrations, good
dispersion of FGO in the polymer matrix
occurs. However, when increased to 5
wt% FGO, the tensile strength decreased
significantly [9], which is also apparent in
the results of this research. As
concentration increases past a certain point
– in this case being 5 wt% – GNPs start to
reaggregate as part of their nature to self-
assemble; Thus, dispersion becomes poor
[10].
Figure 2. Young’s moduli for both PP/GNP
and PS/GNP nanocomposites obtained from
tensile testing.
On the other hand, the PP
nanocomposites had a gradual and steady
increase in elastic modulus as GNP
loading increased. Although the modulus
for PP is lower than PS, it steadily
increases as GNP concentration is
increased. The explanation of this
phenomenon is somewhat complex and
beyond the scope of this research. Simply
put, GNPs have a restrictive effect on the
movement of the polymers molecular
chains, thereby increasing its modulus
[11]. Compared to pure PP, GNP loading
of 10% increased the modulus by 1.5 times
its original value. GNP loading of 25%
increased the modulus by almost 3 times
its original value.
The thermal conductivity tests
were performed at 25 °C and recorded in
W/(m*K). The GNP loadings of the molds
for PP composites for this test were 0, 7.5,
13.5, 20, 26.5 and 32 GNP wt%
82
respectively. The PS nanocomposites were
consistent with the previous tests with the
addition of 30 and 35 GNP wt%. After the
results were obtained, trendlines were
given to the set of points in order to
establish a trend between GNP loading and
thermal conductivity, illustrated in Figure
3.
Figure 3. Thermal Conductivity for both
PP/GNP and PS/GNP nanocomposites.
This test resulted in a linearly
correlated relationship between GNP
loading and thermal conductivity in both
cases. Pure PP and PS started with a
thermal conductivity of 0.223 and 0.153
which are consistent with their literature
values of 0.22 and 0.13 at 25 °C
respectively [12]. According to the
trendline, PS with 30 wt% GNP loading,
compared to pure PS, increased thermal
conductivity four-fold while PP at the
same loading of GNP increased its original
thermal conductivity five-fold.
Electrical conductivity for the
PP/GNP nanocomposites were performed
for a variety of different frequencies.
However, for the sake of consistency, the
data for 100 mHz will be used for
comparison between the various graphene
concentrations. Each sample had an area of
1 cm2. Resistances along with their
corresponding frequencies were recorded
and compiled for comparison in Figure. 4.
The natural logarithm of electrical
conductivity was compared to the GNP
content in order to establish an overall
trend between them.
Figure 4. The natural log of Electrical
conductivities and Polypropylene with
different GNP concentrations at 100 mHz.
The electrical conductivity for the
PP/GNP nanocomposites have a sharp
spike from 10 to 15 wt%, increasing it
from 260 to 2✕107 S/m. This means that
around these concentrations, the
nanocomposites reached their percolation
threshold in that pseudo-circuits were
formed within the composites structure.
4. Conclusion
The influence by addition of
graphene nanoparticles(GNPs) into
polypropylene(PP) and polystyrene(PS)
matrix is deeply studied in this research.
Our primary goal of this research has
changed from studying PP/GNP
nanocomposite to comparing the
properties of PP/GNP and PS/GNP after
we figured out that PS/GNP
nanocomposite will have a limited use due
to its brittleness.
As shown in Figure 1, an overall
decreasing trend is found in both PP/GNP
and PS/GNP nanocomposites from IZOD
impact strength due to possible
incorporation of graphene in both
polymers. From tensile testing, it is found
that PP/GNP has a increasing trend while
83
PS/GNP has a decreasing trend.
Especially, increasing the graphene
content in PS/GNP to 5 wt% of graphene
significantly lowered. The increase in
graphene concentration has a linear
increase in thermal conductivity for both
PP and PS nanocomposites.
When choosing the optimal
concentration of GNP, IZOD Impact test
becomes a deciding factor. Based on the
analysis from all tests, it’s suggested that
the most applicable concentration for
PP/GNPs is 10 wt% because when the
wt% increases from 5 wt% to 10 wt%
thermal conductivity has a steady increase
by 28.6% from 0.35 W/m*k to 0.45
W/m*k and elastic modulus also increases
by 8.2% from 16.88 MPa to 18.38 MPa. In
contrast to other trends, decreasing trend is
shown in IZOD tests. Impact strength
significantly drops by 28% from 19.62 J/m
to 14.22 J/m when the concentration
changes from 10 wt% to 15 wt%.
Electrical conductivity test showed a huge
increase from 10 to 15 wt%, which
remains consistent with the optimal range
of GNP content chosen. A possible
application of the nanocomposite could be
a battery anode because of its high
electrical conductivity, corrosion
resistance [14], and high impact and
tensile strength.
Acknowledgements
We gratefully acknowledge the
support from the Program in Chemical and
Molecular Engineering at Stony Brook
University in the form of materials and the
use of their laboratories. We would also
like to thank our PI Xianghao Zuo for his
general assistance and Professor Vladimir
Samuilov for his assistance in electrical
conductivity testing.
References
[1] Kuryla, William C. “Flame
Retardancy of Polymeric Materials”. New
York, M. Dekker., 1973. EBSCOhost
[2] “Additives for Polymers” [Electronic
Resource]. New York, NY: Elsevier
Science Pub. Co., n.d. EBSCOhost
[3] Masaya Kawasumi, Naoki Hasegawa,
Makoto Kato, Arimitsu Usuki, and Akane
Okada. “Preparation and Mechanical
Properties of Polypropylene−Clay
Hybrids”. Macromolecules 1997 30 (20),
6333-6338
[4] ASTM D256-10e1, Standard Test
Methods for Determining the Izod
Pendulum Impact Resistance of Plastics,
ASTM International, West Conshohocken,
PA, 2010, www.astm.org
[5] ASTM D638-14, Standard Test
Method for Tensile Properties of Plastics,
ASTM International, West Conshohocken,
PA, 2014, www.astm.org
[6] ASTM E1530-11(2016), Standard Test
Method for Evaluating the Resistance to
Thermal Transmission of Materials by the
Guarded Heat Flow Meter Technique,
ASTM International, West Conshohocken,
PA, 2016, www.astm.org
[7] Zhang, W., et al., “Polydiacetylene-
Polymethylmethacrylate/Graphene
Composites as One-Shot, Visually
Observable, and Semiquantative Electrical
Current Sensing Materials.” ACS Appl.
Mater. Interfaces 5.11 (2013): 4603–4606
[8] Callister, W.D. “Materials Science and
Engineering: An Introduction.” Wiley
New York, 2007.
[9] Qi, X., et al., "Enhanced Electrical
Conductivity in Polystyrene
Nanocomposites at Ultra-Low Graphene
Content." ACS Appl. Mater. Interfaces
ACS Applied Materials & Interfaces 3.8
(2011): 3130-133.
[10] Yasmin, A., et al. “Processing of
expanded graphite reinforced polymer
84
nanocomposites.” Composites Science and
Technology. 66 (2006) 1182–1189.
[11] Huang, C., et al., “Processing of
expanded graphite reinforced polymer
nanocomposites” Appl. Sci. 5 (2015)
1196-1210.
[12] Harper, C.A. “Handbook of Building
Materials for Fire Protection”. New York:
McGraw-Hill, 2004.
[13] Tu, Zhaokang, Jiang Wang,
Changjiang Yu, Hanwen Xiao, Tao Jiang,
Yingkui Yang, Dean Shi, Yiu-Wing Mai,
and Robert K.y. Li. "A Facile Approach
for Preparation of Polystyrene/graphene
Nanocomposites with Ultra-low
Percolation Threshold through an
Electrostatic Assembly Process."
Composites Science and Technology 134
(2016): 49-56. Science Direct. Web. 17
Dec. 2016.
[14] Al-Tayyib, A-HJ, Al-Zahrani, M M.
“Corrosion of Steel Reinforcement in
Polypropylene Fiber Reinforced Concrete
Structures” American Concrete Institute.
American Concrete Institute Materials
Journal, 2, 87 (1990) 108-113.
[15] Paul J. Flory, Principles of Polymer
Chemistry, 1st Edition 1953 Cornell
University.
[16] Jordan, J., et al. “Experimental trends
in polymer nanocomposites—a review.”
Mat. Sci and Eng. A. 393.1 (2005): 1-11.
85
Effects of Graphene Oxide on Proton Exchange Membrane
Fuel Cells Henry Ho, Jesse Matsuda, Mailun Yang, Likun Wang, Miriam Rafailovich
Materials Science and Chemical Engineering Department, Stony Brook, NY, 11794, USA
Abstract
Proton exchange membrane fuel cells (PEMFCs) have a high power density that is
capable of producing clean and reliable energy. In order to commercialize PEMFCs, power
output must be improved while reducing costs therefore we focus on the use and viability of
graphene oxide on Nafion membranes and electrodes for use in PEMFCs. A GO solution is
prepared by measuring out methanol, water, and GO powder and sonicating the solution. After
sonication, the solution is applied as a monolayer onto Nafion using a Langmuir- Blodgett
Trough (LB-Trough) which produces our experimental proton exchange membrane (PEM). This
is conducted with the goal of increasing the efficiency of the fuel cell. The fuel cell was initially
tested on a small single cell test station. Afterwards the fuel cell was tested using our large cell
test station. Using the same technique to produce the GO PEM, a membrane electrode assembly
(MEA) was produced consisting of the PEM sandwiched between two platinum (Pt) electrodes
and subsequently hot pressed. Variables such as relative humidity, temperature and gas flow is
controlled. With the GO applied, the fuel cell is able to yield a higher power output compared to
the control, which is the standard PEMFC without GO modifications. The polarization curve that
was generated using current density and power density showed that the fuel cell with GO was
able to output more power per square centimeter at 0.299278 W/cm2, an increase of 22.1%,
compared to the power per square centimeter at 0.244678 W/cm2 of the control fuel cell.
1. Introduction
With the world’s increasing energy
demand, PEMFCs have the potential to
become a prominent source of renewable
energy production. An attractive aspect of
PEMFCs is the lack of toxic emissions from
the production of energy as the only
products are water and energy.
Figure 1 shows the basic diagram of
a fuel cell. The left side is the anode in
which the H2 gas enters. The cathode is the
right side in which it is either connected to
O2 or to atmosphere. The center consists of
the proton exchange membrane or a specific
electrolyte. Here a proton is stripped of its
electrons. The electrons are then used to
create power by forcing them through a
circuit as they are unable to pass through the
electrolyte due to the membrane. At the
cathode, the protons react with oxygen to
produce water.
Figure 1. Basic Diagram of a PEMFC. [1]
86
At the current state of research,
energy produced from PEMFCs is limited
by the financial costs associated with the
miniscule amount of energy produced in
relation to the initial cost. The major barriers
that hinder commercialization of PEMFCs
are associated with cost, performance, and
durability. The cost is due to the use of
expensive platinum catalyst to promote the
reaction; therefore, alternative catalyst
components are being investigated for lower
cost while maintaining performance.
Performance and durability of the PEMFC
can be improved by improving the MEA
through mitigating degradation. The
Department of Energy has imposed a
standard for commercial fuel cells to operate
for 40,000 hours when stationary and 5,000
hours when in transit [1].
Since 2006, the cost of automotive
fuel cells has decreased by more than 50%
while durability has doubled [2]. The cause
of such is done by utilizing a platinum alloy
to reduce the amount of platinum and by
improving the MEA. Yet it is still possible
to further improve on this technology.
Our goal is to enhance the power
output of PEMFCs. This can be done by
enhancing the power output by using
inexpensive material. The PEMFC output
can be increased by improving the
membrane or the electrode.
The membrane of PEMFCs is one of
the core components that has drawn
considerable attention to it in regards to
what material should be used. Currently,
Nafion membranes produced by Dupont
represent most PEMFCs. The unique
structure of the Nafion membranes provide
chemical stability as well as desirable proton
conductivity under 100% relative humidity
[3]. However, the proton conductivity of
Nafion membranes is extremely dependent
on the presence of water. At elevated
temperatures or low relative humidity, a
sharp decline in proton conductivity is
noticed for the Nafion membrane due to
dehydration [4]. As such, one approach in
aiding the membrane conductivity at low
relative humidity and elevated temperatures
is by incorporating hydrophilic inorganic
additives into Nafion membranes to improve
water retention capacity [5]. Graphene oxide
(GO) is considered as an amphiphilic
material with hydrophilic regions containing
oxygenic groups and may be utilized as an
inorganic additive in conjunction with
Nafion membranes for its unique two-
dimensional structure and high surface area
[6,7].
GO is a promising material for
PEMFCs due to its excellent mechanical and
chemical properties. The functional groups
of GO allow it to have exceptional
properties in reduction oxygen reactions as
well as making it a better support than
traditional platinum on carbon electrodes for
the catalysts involved in PEMFCs [8]. The
main functional group in GO is epoxide that
acts potential once water molecules bind to
the sites, this occurs even at low relative
humidity and at room temperature [9].
Figure 2 illustrates the interaction between
the epoxide group and water molecules as
well as how predicted movement of proton
transfer.
Figure 2. Illustration of proton conduction on
epoxide groups. [10]
Studies have indicated GO can be
implemented in MEAs to enhance the
performance of PEMFCs. We conducted
87
experiments to analyze the effects on GO on
Nafion membranes and determine if the
product meets and exceeds expectations.
Two methods of preparation are used;
applying GO solution with a Langmuir-
Blodgett Trough (LB-Trough) and by spray
coating the Nafion. The resulting coated
membranes were tested in fuel cell test
stations to determine current and power
densities.
2. Experimental Section
2.1 Experimental Methods
2.1.1 Creating GO solution
The Nano Graphene Oxide (NGO)
Powder (purchased from Graphene
Supermarket) is used to create a Graphene
Oxide Solution. The solution consists of 5
mL methanol, 1 mL DI water, 6 mg
graphene oxide. The solution is sonicated
using Branson 3510 for 40 minutes before
stored. This solution was homogenous.
2.1.2 Creating GO and Nafion Solution
A solution consisting of 5 mL
methanol, 1 mL DI water, 6 mg graphene
oxide, and 18 mL liquid Nafion is mixed
thoroughly using a sonication machine for
40 minutes. This solution was homogenous.
2.1.3 Coating Nafion 117 with GO solution
The graphene oxide solution was
applied to the Nafion 117 (Purchased from
DuPont) membrane by utilizing the
Langmuir-Blodgett Trough. One half mL
(0.50) of the GO solution was added to each
side of the trough. The barriers are set to
close at a rate of 8 cm2/minute until it hit a
target of 50 mN/m on the platinum sensor.
While the barriers were closing, 0.25 mL of
GO solution was added to each side every
100 cm2. The Nafion 117 membrane is then
lifted.
2.1.4 GO Nafion 117 on small test station
The Nafion 117 is hydrated in DI
water for 5 minutes. It is then placed in
between two Commercial Platinum Carbon
(0.1 mg/cm2) Gas Diffusion Electrodes
which were purchased from
FuelCellsEtc.com and tested using the small
test station. The cathode is open to
atmosphere while the anode connects to H2.
The H2 gas flow rate is set to 78 ccm.
2.1.5 Spraying Electrodes with GO solution
A similar solution of GO is sprayed
onto Nafion membrane using an air spray
gun.
2.1.6 GO electrodes on small test station
This MEA is tested under the same
conditions as the coated experiment on the
small test station. The test station is shown
in figure 3.
2.1.7 Coating Nafion 212 with GO solution
The same steps were conducted in
regards to coating the Nafion 117 with the
LB-trough. Once the coated membrane is
obtained, it is placed in between two
electrodes. This MEA is then hot pressed
using Carver hydraulic unit model #3912 at
130 °C for two minutes for the electrodes to
adhere to the membrane.
2.1.8 GO Nafion 212 in large test station
The MEA is placed into the large test
station, (University Test Stations from
FuelCellTechnologies.com) shown in figure
4. The large test station temperature is set to
60 °C, no back pressure, and 100% relative
humidity. Current density is increased
stepwise from 0.0 A/cm2 to 7.0 A/cm2 at set
intervals, minimum gas flow rate of 50 sccm
for H2 and minimum gas flow rate of 100
sccm for O2. The gas flows were concertedly
increased as the current increased.
88
2.2 Analytical Methods
2.2.1 Isothermal Curves
Isothermal curves are obtained from
the use of the LB-trough to analyze the
pressure as a function of surface area. This
is to analyze the ability of the solutions to
maintain a monolayer.
2.2.2 Polarization Curves
The data collected by the fuel cell
test stations allow us to create polarization
curves. Three assemblies are prepared:
control, sprayed, and coated. The control
refers to the MEA that has no GO applied.
The sprayed assembly has the electrodes
sprayed with GO solution. Lastly, the coated
membrane refers to the assembly with the
Nafion membrane coated with GO using the
LB-trough. The polarization curves allow us
to visualize which fuel cell will yield the
higher output. These polarization curves
were obtained using the following test
stations: Small Test Station and the Large
Test Station.
Figure 3. Small Test Station
The small test station provides
hydrogen to the fuel cell but is also open to
air. The fuel cell would be influenced by
ambient temperature and humidity. The
large test station allows us to control these
parameters.
Figure 4. Large Test Station
3. Results and Discussion
Figure 5 depicts voltage and power
density vs current density for the control
fuel cell, the GO coated fuel cell, and the
GO sprayed fuel cell. The fuel cell that has
the membrane coated with GO through the
LB-trough yields the highest voltage to
current density and power density. The next
is the fuel cell that has its electrode sprayed
with GO. Finally, the control has the lowest
values of the three as shown in Figure 7. We
determined that the use of GO spray is an
inefficient means of applying GO as the
particles tended to accumulate on the surface
of the nozzle preventing GO from being
applied as well as developing an uneven
coating on the membrane. A suggested
method in improving this process to make
this a viable option is to use an air spray gun
made of a material that GO does not adhere
to.
As it is determined that the GO coat
is more beneficial, we moved onto testing
the GO coat in the large test station. Figure 6
89
shows that the GO coat can output roughly
22.0% more than the control at the peak,
0.95 W/cm2 vs 0.78 W/cm2. Figure 6 shows
that when operating at 0.6V the standard
PEMFC can output 0.245 W/cm2 whereas
the PEMFC with GO coating can output
0.299 W/cm2. Tateishi et al had results of
their paper GO to reach power density
output at approximately 0.93 mW/cm2 at
0.6V, about one third of the output from the
LB-trough coated membrane [9].
From the isothermal curve that is
created during the application of GO via
LB-trough, we theorized that the GO
monolayer continuously deformed and
folded over, creating a multilayer. The
steady decrease shown in Figure 8. The
increase is a result of adding GO solution as
the surface area decreased every 100 cm2.
The monolayer formed from the GO
solution constantly fell apart meaning that a
stable monolayer is not present throughout
the application. However, by introducing
liquid Nafion into the GO solution, the
isothermal curve procured from the LB-
trough, Figure 9, shows that no collapse
occurred. Figure 9 instead shows that the
surface pressure increases showing that the
layers were being pushed on top of the water
rather than into the water. This means that
the monolayer formed with the addition of
liquid Nafion held up significantly better
than the GO solution without liquid Nafion.
The Nafion is able to support the
formation of a layer because it contains both
hydrophilic and hydrophobic areas. Liquid
Nafion contains sulfonic acid functional
groups that self-organize into arrays of
hydrophilic water channels. Interspersed
between the hydrophilic channels are
hydrophobic polymer backbones which
provide mechanical stability.
Figure 5. A graph of the current densities and
power densities of the control, GO coated, and
GO sprayed obtained from the small test station.
Figure 6. A graph of the current densities and
power densities of the control and GO coated
obtained from the large test station.
Figure 7. Histogram depicting the power density at 0.6V.
90
Figure 8. Isothermal curve that is obtained
from LB-trough for GO solution.
Figure 9. Isothermal curve that is obtained
from LB-trough for GO and Nafion Solution
4. Conclusion The gathered data indicates that GO
has a beneficial effect on the PEMFC.
Coating the Nafion membrane has a larger
effect on the PEMFC and efficiency than
spraying the electrodes. A suggested method
in improving the spraying process to make it
a viable option is to use an air spray gun
made of a material that GO does not adhere
to. Figure 6 shows that the power density
for the coated is 22.0% higher than the
control at peak output. Based on the
analysis, it is apparent that the most efficient
method in enhancing the output of PEMFCs
is the coating of the membrane with GO.
The GO and nafion solution is also shown to
be able to form layers whereas the GO
solution has a preference to enter the water
instead of creating a layer.
Acknowledgements
We gratefully acknowledge the
educational support from the Department of
Chemical and Molecular Engineering at
Stony Brook University. The guidance from
Miriam Rafailovich, Clement Marmorat, and
Likun Wang has been invaluable.
References
[1] Department of Energy. Fuel Cells.
https://energy.gov/eere/fuelcells/fuel-cells
(accessed Dec 21, 2016). [2] Department of
Energy. Fuel Cell Technologies Office:
Accomplishments and Progress.
https://energy.gov/eere/fuelcells/fuel-cell-tec
hnologies-office-accomplishments-and-prog
ress (acessed Dec 21, 2016). [3] Zhang, H.;
Shen, P. Recent development of polymer
electrolyte membranes for fuel cells. Chem.
Rev. 2012, 112, 2780–2832. [4] Hickner,
M.A.; Ghassemi, H.; Kim, Y.S.; Einsla,
B.R.; McGrath. J.E. Alternative polymer
systems for proton exchange membranes
(PEMs) Chem. Rev. 2004, 104, 4587–4611.
[5] Tripathi, B.P. ; Shahi, V.K. Organic-
inorganic nanocomposite polymer
electrolyte membranes for fuel cell
applications. Prog. Polym. Sci. 2011, 36,
945-979. [6] Chen, D.; Feng, H.; Li, J.
Graphene oxide: preparation,
functionalization, and electrochemical
applications. Chem. Rev. 2012, 112, 6027–
6053. [7] Chien, H.-C.; Tsai, L.-D.; Huang,
C.-P.; Kang, C.-y.; Lin, J.-N.; Chang, F.-C.
Sulfonated graphene oxide/Nafion
composite membranes for high-performance
direct methanol fuel cells. Int. J. Hydrogen
Energ. 2013, 38, 13792–13801. [8] Devrim,
Y.; Albostan, A. Graphene Supported
Platinum Catalyst-Based Membrane
Electrode Assembly for PEM Fuel Cell. J.
Electron. Mater. 2016, 45.8, 3900-907. [9]
91
Tateishi, H.; Hatakeyama K.; Ogata, C.; et
al. Graphene Oxide Fuel Cell. J. of the
Electrochem. Soc. 2013, 106, F1175-F1178.
[10] Koinuma, M.; Ogata, C.; Kamei, Y.; et
al. Photochemical Engineering of Graphene
Oxide Nanosheets. J. Phys. Chem. 2012,
116, 19822-19827
92
Synthesis and Characterization of Gold-Palladium
Nanoparticles Catalyst For Improved Hydrogen Fuel Cell
Performance
Adam Bennetta, Helen Liua, Allen Trana, Likun Wangb,
Miriam Rafailovicha,b* a,bChemical and Molecular Engineering Department, Stony Brook, NY, 11794, USA
Abstract
Alternative energy sources are becoming increasingly more important in meeting
current energy demands, due to the issues faced by conventional energy sources such as
rising costs, political volatility, depletion of resources and carbon emissions. One promising
renewable energy source is the proton exchange membrane (PEM) hydrogen fuel cell, which
creates only water as a major byproduct. Currently, platinum (Pt) is the predominant catalyst
being used for PEM hydrogen fuel cells. The total system cost is predominantly due to the
high cost of Pt. Pt catalyst activity in the hydrogen fuel cell suffers from exposure to carbon
monoxide (CO), which limits the viability of PEM fuel cells for vehicles and stationary
applications. Gold (Au) has been investigated as a promising catalyst for the oxidation of CO
to carbon dioxide (CO2), the addition of which to the PEM fuel cell can increase its longevity.
However, due to the high cost of Au, an alternative catalyst could greatly reduce the total cost
of a fuel cell system while maintaining high performance, making it a much more feasible
option. This research looks at the use of Gold-Palladium (Au-Pd) nanoparticles as a catalyst
for the improvement of hydrogen fuel cell performance. In this study the bimetallic Au-Pd
catalyst were synthesized via the Brust Method and characterized using extended X-ray
absorption fine structure (EXAFS) and transmission electron microscopy (TEM) techniques.
The subsequent testing on the catalyst was done at the fuel cell test station located at Stony
Brook University in order determine its power density. Through the characterization and test
station results it was determined that the synthesized Au-Pd catalyst had an alloy structure
and produced roughly 15% more power than the control catalyst.
1. Introduction
Fossil energy sources are currently
facing not only political issues such as
volatility and an unpredictable nature, but
also environmental concerns of resource
depletion and carbon emissions.
Alternative energy will be vital for our
increasing global energy demands and
renewable energy sources such as fuel
cells can play a key role in combating
these problems. A fuel cell is a device that
uses a fuel to convert chemical energy into
useable electricity. Fuel cells are an
attractive energy source due to the fact that
they don’t emit greenhouse gases and they
have a high energy density and efficiency
[1].
In a proton exchange membrane
(PEM) fuel cell, a polymer membrane is
held between the cathode and anode. This
membrane allows protons to pass while
preventing electrons from doing so.
Keywords: Energy, PEM fuel cell, Catalyst, Nanoparticles, Hydrogen fuel cell, Au-Pd
2
Nafion, a copolymer, is commonly used as
the membrane due to its high conductivity
and mechanical durability [2].
An external wire is placed on the
fuel cell connecting the electrodes and as
the electrons build up they travel from the
anode into the cathode producing an
electric current. Water and heat is also
produced during this reaction. As of now
PEM fuel cells are one of the most
promising alternative energy sources for
transportation as well as having
commercial and residential applications.
Figure 1. Schematic of PEM Fuel Cell [3]
Currently the most common type of
fuel cell is the hydrogen fuel cell.
Hydrogen is the most abundant element on
earth and can also be generated by splitting
a water molecule with a DC current
through the process of electrolysis [1]. In
the PEM hydrogen fuel cell (PEM HFC),
hydrogen is flowed through the membrane
on the anode side and reacts with oxygen
on the cathode side. The reaction occurs in
the presence of an electrolyte (the
membrane). The reaction occurring at the
anode is referred to as the hydrogen
oxidation reaction (HOR) and the reaction
occurring at the cathode is called the
oxidation-reduction reaction (ORR).
Table 1. Hydrogen PEM Fuel Cell Reactions
Anode
(Oxidation)
H2→ 2H++2e-
Cathode
(Reduction)
O2+2H++2e-
→H2O
Total Reaction H2+ O2 → H2O
In a fuel cell, the catalyst properties
are important in that they have a large
effect on performance. The rate-
determining step in the electricity
production of the fuel cell is usually the
ORR at the cathode [4]. In order for the
reaction to move forward, toward the
formation of electricity and water the
activation energy of the reaction must be
surpassed. Catalysts are needed for fuel
cells to lower the reactions activation
energy.
The catalyst is used at the cathode
and anode of the fuel cell and works by
promoting the HOR and the ORR.
Different catalysts and supports are widely
studied in order to find the most efficient
and cost effective fuel cell for operation.
Although current technology exists for
PEM HFC applications, its wide spread
implementation is not yet feasible due to
the high overall cost and questions of
durability. In order for the PEM HFC to
become a viable option for vehicle and
93
3
stationary applications the costs needs to
be greatly reduced [1].
Platinum (Pt) is the predominant
catalyst used in the PEM HFC and is
responsible for the bulk of the cost. The
durability of Pt catalysts is also
questionable due to carbon monoxide (CO)
poisoning [5]. The effects of CO poisoning
on the Pt catalyst are especially of interest,
since even just 25 ppm of CO can reduce
PEM HFC output by 50% [6]. Many
sources of hydrogen gas come from
natural gas reforming and thus contain a
considerable amount of CO2, which can
become CO while the PEM HFC is
operating and thus poison the Pt catalyst
[7]. Current research is geared towards the
cost reduction and increasing efficiency
and durability by addressing the effects of
CO poisoning. A major challenge that
remains is finding a suitable catalyst which
can oxidize CO. The application of a
suitable catalyst which can oxidize CO
allows the utilization of cheaper hydrogen
gas sources by reducing the purity
requirement of the feed hydrogen gas.
Nanoparticles are particles between
1 and 100 nm in size and are often seen to
have increased catalytic activity due to
their small size which prove a larger
surface area (source). Gold (Au) and
Palladium (Pd) nanoparticles have both
been shown to have catalytic properties in
CO oxidation. It has been shown that
supported Au nanoparticles are extremely
effective catalysts for oxidizing CO [6].
The catalytic activity of Au can even
further be enhanced by incorporating a
second metal as an alloy [5]. Pd has been a
metal of interest to combine with Au, since
it can add electrons to the system and thus
increase catalytic activity.
The most noticeable improvement in
catalyst performance occurs with a gold-
palladium core-shell nanoparticle
conformation. The palladium atoms on the
shell withdraw atoms from the gold core,
shifting the d-band center of palladium
such that the adsorption of O2 and O-O
bond breaking is promoted. The negative
charge on the palladium shell stabilizes
oxygen atoms as they dissociate, lowering
the energy barrier for O2 to dissociate
[9,10].
It has been shown that synthesis
methods greatly affect the size and
morphology of bimetallic nanoparticles.
Current methods to synthesize Pd
bimetallic catalysts lead to polydispersity
or clusters of nanoparticles. These
methods work, however, make
characterization difficult. The morphology
of nanoparticles has a great effect on the
catalyst activity [5]. Current research
involving Au-Pd nanoparticle catalysts
involve either core shell or alloy structure
and are synthesized in various ways. A
bimetallic core shell nanoparticle is
composed of two phases, one metal in the
core and the other surrounding it. A
bimetallic alloy is a random mixture of the
two metals [11].
Figure 2. Possible representative architectures
of bimetallic nanoparticles (a) alloy, (b) core-
shell, (c) cluster-on-cluster, (d) sub-shell, and
(e) intermetallic [11]
94
4
In this research the bimetallic
nanoparticle catalyst was synthesized via
the Brust Method, which is known to
produce small, high-surface area thiol-
stabilized nanoparticles through the
reduction of the metal. The Langmuir-
Blodgett (LB) trough was used to spread
the synthesized particles over water in a
trough that uses the surface tension to
compress them onto the Nafion membrane.
Bimetallic nanoparticles are
extremely difficult to characterize so the
purpose of this research is to determine
whether the Au-Pd nanoparticles
synthesized via the Brust Method have an
alloy or core shell structure. Extended X-
ray absorption fine structure (EXAFS) was
conducted at the Stanford Synchrotron
Radiation Lightsource (SSRL). EXAFS
experiments are used in order to model the
coordination environment around the
absorbing metal atoms. Transition electron
microscopy (TEM) was also done in order
to determine the size and dispersity of the
nanoparticles. The goal of this
characterization is to determine what the
exact structure of the synthesized Au-Pd
catalysts are.
After the catalysts were
characterized they were tested on the
hydrogen fuel cell test station at Stony
Brook University to determine the power
output of the fuel cell with the nanoparticle
catalysts. This was done by creating
polarization curves of the voltage against
current and power against current, which
allows us to compare the power output
between a control and the Au-Pd catalysts.
We hypothesize that PEM HFCs
assembled with the membranes with the
synthesized Au-Pd nanoparticles will have
increased power output as compared to
HFCS with just the platinum catalysts,
thus improving the performance of PEM
fuel cells. If successful this research can in
turn find ways to improve the durability of
the fuel cell system.
2. Experimental Section
2.1 Materials
HAuCl4 was purchased from Sigma
Aldrich (99%). K2PdCl4 was purchased
from Sigma Aldrich (99%). 0.1 mg/cm2 Pt
Loading Electrode Catalyst Paper was
purchase from Fuel Cell Store.
2.2 Experimental Methods
2.2.1 Nanoparticle Synthesis
In preparing the nanoparticle
solution, 393.83 mg (1 mmol) of
tetrachloroaurate (HAuCl4) and 326.43 mg
of palladous potassium chloride (K2PdCl4)
were dissolved in 36 mL of deionized
water. Following, 2654.59 mg of
tetraoctylammonium bromide (TOABr)
was dissolved in 96 mL of toluene. This
solution was added to the tetrachloroaurate
and palladous potassium chloride mixture
and then magnetically stirred for 20
minutes until the mixture separated into
two distinct layers. 200 µL of
dodecanethiol and 453.96 mg of sodium
borohydride (NaBH4) dissolved in 30 mL
of deionized water were added to the two-
layer solution, and magnetically stirred at
room temperature for 3 hours. The
aqueous layer was removed from the
solution via separatory funnel and the
remaining top layer was rotary evaporated
until 5 mL remained. 200 mL of ethanol
was added to the top layer solution and
refrigerated overnight at 4ºC. The top
solution was removed from top layer
solution and the remaining sample was
centrifuged at 5000 rpm for 10 minutes
then washed with ethanol multiple times.
The sample was dried in the vacuum
desiccator for 2 days.
95
5
2.2.2 LB Trough Coating
The Langmuir-Blodgett (LB)
Trough was calibrated to less than 0.25
mN/m surface pressure prior to
nanoparticle coating. The 212 Nafion
membrane is placed on the platinum plate,
attached to a hook in the center of the LB
Trough. 100 µL of nanoparticle solution
was added via micropipette to each side of
the LB trough. The solvent from the
nanoparticle solution was allowed 10
minutes to evaporate before starting the
coating process. The target pressure was
set to 5 mN/m with pushing rate of 6
mm/min.
2.2.3 Fuel Cell Test Station
The fuel cell test station was
operated at 60C. The 212 Nafion
membrane, was wetted with deionized
water and placed in a pre-fabricated
membrane electrode assembly (MEA). On
the fuel cell test apparatus (Fuel Cell
Technologies, Inc., SFC-TS), 78 ccm of
hydrogen gas was constantly flowed into
the MEA. The cell was held at 6V for 1
hour to stabilize it, and then cycled
between 0.5 A/cm2 and 1 A/cm2 9 times.
The cell was then operated at 0.2 A/cm2
for 6 hours to fully humidify the cell.
Following the break-in procedure, the VIR
software in LabVIEW was initiated to
conduct performance tests on the cell.
2.3 Analytical Methods
2.3.1 Transmission Electron Microscopy
(TEM):
The nanoparticle sample was
subjected to TEM observations using the
JEOL JEM 1400 Transmission Electron
Microscope. A few drops of nanoparticle
solution using a micropipette was diluted
in a petri dish of toluene and placed on a
TEM grid. Images were observed at 100kx
and 200kx magnifications. High
Resolution TEM was also performed on
the nanoparticle sample.
2.3.2 Extended X-Ray Absorption Fine
Structure (EXAFS):
Extended X-Ray Absorption Fine
Structure measurements were performed
on the nanoparticle solution at Stanford
Synchrotron Radiation Lightsource
(SSRL). X-Rays of narrow energy
resolutions were shone at the sample and
the transmitted x-ray intensity was
recorded. Dependent on sample thickness,
absorption coefficient, and atom type a
number of photons are absorbed by the
sample. When the incident x-ray energy
matches the binding energy of the
electrons of an atom in the sample, the
number of x-rays is increased and the
transmitted x-ray intensity drops, resulting
in the absorption curve.
3. Results and Discussion
The TEM images were observed to
be spherical nanoparticles in Figure 3. The
nanoparticle size distribution is relatively
similar, with TEM gridded AuPd
nanoparticle size averaged at 2.02 nm and
the LB trough AuPd nanoparticle size
averaged at 1.86 nm. This suggests a high
surface area for the nanoparticles.
(a)
96
6
(b)
(c)
Figure 3. TEM Images of (a) Au , (b) Pd, and
(c) AuPd Nanoparticles
Figure 4. AuPd NP TEM Gridded TEM Size
Distribution
Figure 5. AuPd NP LB Trough TEM Size
Distribution
(a)
(b)
Figure 6. (a) HR TEM of AuPd NPs (b)
magnified view of crystalline platelet structure
From the HR TEM, it can be seen
that the AuPd NPs have a crystalline
platelet structure, which is suggested in
literature to by the structure needed by
nanoparticle catalysts in order to
effectively oxidize CO. The platelet
structure has good contact with the
support, and has enough surfaces for the
reaction to occur, since the oxidation
reaction occurs on the edges and steps of
the platelet, rather than the smooth
surfaces [8].
97
7
The XANES (Fig. 7) analysis
shows that the Au nanoparticles samples
prepared through conventional methods
and LB trough were consistent with
metallic Au foil, with lower-intensity
features due to the presence of thiol-
stabilized NPs. The EXAFS spectra also
shows that the gold NP samples were close
to that of metallic gold, but showed lower
amplitudes again due to the presence of
Au-S bonds.
Figure 7. XANES for AuNP
Figure 8. EXAFS for AuNP
Figure 9. Pd K-Edge XANES for thiol-
stabilized PdNPs
Figure 10. Pd K-Edge EXAFS for thiol-
stabilized PdNPs
The Pd K edge XAS spectra for the
palladium nanoparticles are shown in
figures 8 and 9. The XANES spectra for
both samples are different from the
XANES for metallic Pd. They do however
resemble the XANES spectra for PdS very
closely. Similarly the EXAFS spectra and
the Fourier transformed EXAFS for the
nanoparticles show a very good indication
of a strong Pd—S contribution by the low
frequency oscillations and a maximum in
the low wavenumber range. The metallic
Pd is also present in the sample as shown
by the 2 and 3 Å peak indicating a Pd—Pd
bond. The XAS comparison of the
nanoparticle samples prepared with
conventional methods versus the LB
method show us there is no significant
difference between the two. No difference
between the samples was seen in
agreement with the XANES that both
samples had a large contribution of Pd—S
bonds.
From EXAFS analysis, it can be
seen that there are no Pd-Pd bonds in the
AuPd NPs sample. This suggests that the
AuPd NP is in an alloy configuration
rather than a core-shell configuration,
since a core-shell configuration would
show both Au-Au bonds and Pd-Pd bonds.
The final part of this research
involved determining the power output of
98
8
the synthesized Au-Pd nanoparticle
catalyst. This was done at the fuel cell test
station at Stony Brook University. 0.1
mg/cm2 Pt loading was used on the
electrode for the control and the Au-Pd
catalyst to compare the power output. The
membrane used was Nafion 212, which
has a thickness of 50.8 micrometers.
Hydrogen was flowed through the fuel cell
at a rate of 78 CCM (cubic centimeter per
minute). The operating temperature was
set at 60 ⁰C.
Figure 11. Voltage against current for Au-Pd
Catalyst with 0.1 mg/cm2 Pt catalyst loading
and control of only 0.1 mg/cm2 Pt catalyst
loading.
From the polarization curves we
determined that the hydrogen fuel cell had
a higher maximum power output with the
Au-Pd nanoparticle catalyst as opposed to
just the control. The maximum power
output of the Au-Pd catalyst was 0.314
watts whereas it was only 0.275 for the
control. Approximately 14.18% more
power was generated with the Au-Pd
nanoparticle catalyst.
Figure 12. Power against current for Au-Pd
Catalyst with 0.1 mg/cm2 Pt catalyst loading
and control of only 0.1 mg/cm2 Pt catalyst
loading on electrode.
9Table 3. Max power and current for Au-Pd
Catalyst with 0.1 mg/cm2 Pt catalyst loading
and control of only 0.1 mg/cm2 Pt catalyst
loading on electrode.
Catalyst Max
Power
(watts)
Max
Current
(amps)
Au-Pd 0.314 0.748
Control 0.275 0.699
Table 2: Structure parameters (coordination numbers N, interatomic distances R and disorder
factors σ2), obtained in fitting of experimental EXAFS data
99
9
In the fuel cell hydrogen is
oxidized at the anode and oxygen is
reduced at the cathode producing water.
Platinum catalyzes both of these reactions.
These results found that the use of an Au-
Pd nanoparticle catalyst catalyzes the fuel
cell reactions at a better rate yielding a
better power output.
4. Conclusions
The majority of the current work on
fuel cells is aimed towards their potential
use in vehicles, PEM fuel cells being the
most promising. In order for PEM fuel
cells to become a viable option for
vehicles and stationary applications the
cost needs to be reduced and the durability
needs to be increased. Fundamental
research is needed for the fuel cell
membrane and catalyst layer. The
characterization techniques of EXAFS in
conjunction with TEM showed that it is
highly likely the catalyst we synthesized
was an alloy. The XAS spectra showed us
that the obtained coordination numbers
and Pd—Pd and Pd—S distances are
consistent. Therefore it can be concluded
that the investigated samples contain both
metallic nanoparticles, as well as the low
molecular weight Pd-thiol complexes. The
absence of Pd-Pd bonds in the EXAFS
analysis of the AuPd NPs strongly
indicates that the synthesized catalyst is in
the alloy configuration.
The synthesized catalyst had an
average size of 1.98 nm indicating that the
Brust method was effective was effective
in synthesizing Au-Pd nanoparticles with
an effective surface area. From the HR-
TEM, it could be seen that the synthesized
AuPd NP had the desired crystalline
platelet structure, further indicating the
efficacy of the Brust method.
The research performed shows a
promising result of the Au-Pd alloy
catalyst in the PEM HFC. The Au-Pd alloy
nanoparticles increased the output of the
fuel cell showing that it is an effective
catalyst for fuel cell reactions. The power
output of the Au-Pd catalyst was
approximately 14.18% higher than the
control.
Acknowledgements
We gratefully acknowledge the
financial support from the Department of
Materials Science & Engineering and the
Program in Chemical and Molecular
Engineering at Stony Brook University
through research funding. We also thanked
the Advanced Energy Center (AERTC) to
provide laboratory and equipment and
SSRL to give us analysis of our
nanoparticles.
References
[1] Department of Energy Fuel Cell
Technologies Office Multi-year Research,
Development and Demonstration Plan
[2] Wang, Y., Chen, K. S., Mishler, J.,
Cho, S. C., & Adroher, X. C. (2011). A
review of polymer electrolyte membrane
fuel cells: Technology, applications, and
needs on fundamental research. Applied
Energy, 88981-1007.
[3] Jones, R. L. (2017, March 27). Energy
Storage & Delivery. Retrieved from
https://www.nist.gov/programs-
projects/energy-storage-delivery
[4] Reda, M. R. (1970, January 01). The
Rate Limiting Step (RLS) for the Oxygen
Reduction Reaction at the Cathode of
Polymer Electrolyte Membrane Fuel Cell.
[5] Hutchings, G., & Kiely, C. (n.d).
Strategies for the Synthesis of Supported
Gold Palladium Nanoparticles with
Controlled Morphology and Composition.
100
10
Accounts Of Chemical Research, 46(8),
1759-1772.
[6] Baschuk, J.J., and Xianguo Li.
International Journal of Energy Research
25, no. 8 (200): 695-713.
[7] Janssen, G.J.M., and N.P. Lebedeva. In
Presented at the Conference: Fuel Cells
Science and Technology vol. 2004, pp. 6-7.
2004.
[8] Haruta, M. Gold Bull (2004) 37: 27.
doi:10.1007/BF03215514
[9] Chen, D. et al. Core-shell Au@Pd
nanoparticles with enhanced catalytic
activity for oxygen reduction reaction via
core-shell Au@Ag/Pd constructions. Sci.
Rep. 5, 11949
[10] Staykov, A., Derekar, D., &
Yamamura, K. (2016). Oxygen
dissociation on palladium and gold
core/shell nanoparticles. International
Journal Of Quantum Chemistry, (20),
1486.
[11] Scott, R. J. (2015). Rational design
and characterization of bimetallic gold-
palladium nanoparticle catalysts. Canadian
Journal Of Chemical Engineering, 93(4),
623-630.
101
102
Pressure-Time Dependence and Simulation of Carbon Dioxide
Hydrate Formation
Sandhiya Kannan, John Mikhail, Raphael Prodromou,
Maya Endoh, Tadanori Koga
Chemical and Molecular Engineering Program, Stony Brook University, Stony Brook, NY 11794
Abstract:
Two concurrent objectives in modern industry – to acquire natural gas for energy and to minimize
the greenhouse effect by reducing carbon dioxide (CO2) concentration in the atmosphere – have
led to an interesting research opportunity. By replacing the methane naturally present in ocean-
floor hydrates with carbon dioxide, both concerns may be addressed simultaneously. The time
necessary for nucleation of these hydrates upon exposure to carbon dioxide gas is not yet known,
and would allow for more efficient process development for hydrate formation on the ocean floor.
This project focuses on an analysis of the formation time of CO2 hydrates as a function of pressure.
These hydrates are formed in the laboratory through replication of ocean-floor low temperature
and high pressure conditions using a cooled pressure chamber. Once water is loaded into the cell,
the system is isolated and pressurized with CO2 gas; a pressure transducer with a LabVIEW
interface is used to record the pressure of the isolated system over time. The observation of a
pressure drop indicates that some of the gaseous CO2 had been captured in a hydrate; the nucleation
time of the hydrate could be determined from these data as well. We found that the pressure drop
due to hydrate formation took place approximately twelve hours after the start of each trial. In
addition, computer simulations of the water and CO2 system at the nanoscale were performed to
establish a functional model which may be used for future work; Lennard- Jones interatomic
potentials and harmonic bond styles were used to calculate the force interactions. By analyzing
simulation results based on this model, it is possible to obtain an additional perspective of the
hydrate formation and visualize the clathrate structure at an atomic level. Continued study of the
hydrate formation process would allow its deployment as an efficient method for simultaneous
methane extraction and carbon dioxide sequestration.
Keywords: Carbon Dioxide Hydrates, Clathrates, Computational Modeling
1 Introduction
According to the United States
Energy Information Administration’s
International Energy Outlook 2016, the
world’s energy consumption has been
increasing in recent years and is predicted to
increase by 48% between 2012 and 2040,
with natural gas continuing to account for the
greatest increase in energy consumption
globally [1]. While natural gas tends to burn
cleaner than other fossil fuels, the
deployment of hydraulic fracturing, shale gas
production, and other technologies still
produce significant quantities of greenhouse
gas emissions, including carbon dioxide.
Methods of carbon dioxide sequestration are
being researched along with new and
unconventional sources of natural gas, but
recently there have been developments
towards a means of achieving both
simultaneously.
Methane seeping through faults on
the ocean floor forms hydrate/clathrate ice
structures, which consist of water molecules
arranged in crystalline cage structures
trapping methane gas molecules in the center,
103
Figure 1. Natural formation of methane hydrates
with clathrate structure on the ocean floor [2].
as seen in Figure 1.
These hydrates form through
exothermic reactions and are stable at low
temperatures and high pressures such as
those found at the ocean floor. By
destabilizing these hydrates, it would be
possible to collect the methane gas released
to the ocean surface. In our planned project,
we expect to observe a similar but more
exothermic hydrate formation of CO2
molecules within our system
and the formation of a Type I clathrate
structure. Based on previous studies, the heat
released from the hydrate formation should
be enough to dissociate methane hydrates
under the same pressure and temperature
conditions, allowing for a simultaneous
sequestration of carbon dioxide and
extraction of methane for natural gas.
1.1 Previous Work
Previous studies have focused on the
methane hydrate formation process, such as
one by Tsutomu, Ebinuma, and Ishizaki
which aimed to examine the equilibrium
conditions of methane hydrates to locate the
bottom of the gas hydrate stability zone
(BGHS) and establish the formation and
dissociation rates for the hydrates [3]. It was
hypothesized in this study that confining
methane hydrates in small pores of porous
glass would make it possible to observe the
effect of pore size and temperature on hydrate
dissociation. Pieces of porous silica glass
with pore diameters of 100, 300, and 500 Å
were dried and inserted in a high-pressure
cell along with distilled, deionized water
which filled the pores. The cell was
maintained at room temperature for one day,
and then the air was evacuated before
methane gas was introduced. A cryostat was
used to set the pressure cell to a
predetermined temperature, and hydrate
formation was observed over the course of
several days. Upon formation of the hydrate,
pressure in the cell was observed to gradually
drop to equilibrium while the temperature
rose.
By examining the equilibrium data, it
was found that smaller pore diameters result
in a larger temperature shift of the
equilibrium line. Pores that were smaller than
or on the order of 300 Å showed a heat of
methane diffusion different from the bulk
material. Dissociation temperature and
pressure trends in the experiment were,
however, accurately predicted using the
Clausius-Clapeyron equation; the shift in the
equilibrium lines of the methane hydrates
was similar to the results obtained from
Melnikov and Nesterov. The latent heat of
the confined hydrate was smaller than when
the hydrate was formed in bulk [3].
In a study by Jacobson et al., clathrate
nucleation was studied by varying the size
and hydrophilicity of guest molecules [4].
They examine clathrate formation using
molecular dynamics simulations with a
constant pressure and temperature
thermodynamic ensemble and a coarse-
grained model, meaning fine details of the
molecules were blurred into more
macroscopic properties. Clathrate formation
proceeds by a mechanism for nucleation,
which cannot be determined exactly from
empirical thermodynamic analysis of the
104
system. Molecular dynamics enables a much
more detailed picture of the experiment by
allowing the researchers to examine
individual atoms or molecules and their
trajectories over the course of the simulation.
The authors present the “blob
mechanism” of clathrate formation, in which
nucleation occurs in stages starting from a
small collection of molecules beyond a
certain critical density. The dissolved guest
collects into a blob, followed by an
amorphous clathrate, and then finally a
crystalline clathrate. For carbon dioxide,
relatively small clathrate cages are formed,
either 512 or 51262. Also, approximately 30
cages are needed to initiate formation of the
amorphous clathrate. Carbon dioxide and
methane, both being relatively small and
hydrophobic molecules, use the same cage
structures and are thus susceptible to the
memory effect, in which hydrates form
quickly because the appropriate cages are
already present in the medium.
A study done by Horvat and Mahajan
demonstrated the interactions associated with
CO2, CH4, and sand systems [5]. In the first
trial, CH4 hydrates were formed in a high-
pressure cell with no sand within 96 hours.
With the introduction of CO2 there was an
instant formation of CO2 hydrate as the
atmosphere in the cell became 99% CH4. In
the second trial, CO2 hydrate was allowed to
form before the introduction of CH4. CH4
hydrate formed initially, but over time the
hydrates dissociated and the CH4 entered the
gas phase. In the third trial, CH4 formed
hydrates in sand and then CO2 was
introduced. Gas analysis showed that most of
the hydrates were mostly CO2. In the fourth
trial, CO2 was allowed to form hydrates in the
sand and then shortly after CH4 was
introduced. The composition of the gas in this
run was low in CO2. Lastly, CH4 was allowed
to form hydrates in the sand before the
introduction of CO2. Upon depressurization,
CO2 concentration increased thus indicating
that most of it was in hydrate form. This study
showed that CO2 was the dominant hydrate
and gives insight into the relative speed of
hydrate formation.
In this study, we aim to analyze the
nucleation time of the carbon dioxide
hydrates as a function of pressure.
Simulations will also be used to predict and
model the hydrate structure and formation
process. By gaining a greater understanding
of the carbon dioxide hydrate formation and
structure, it would possible to optimize a
process for the simultaneous methane
extraction and carbon dioxide sequestration.
2 Materials and Methods 2.1 Hydrate Formation
In this experiment, ocean floor
conditions were emulated using a chiller and
high-pressure cell custom-designed per
specifications for a 2010 study by Koga et al.,
[7] as seen in Figure 2. The temperature was
set to 4 ℃ and a range of pressures between
the hydrate formation pressure and vapor-
liquid equilibrium of carbon dioxide. Carbon
dioxide flowed from a compressed gas
cylinder, which was monitored by pressure
gauges, into a high-pressure cell loaded with
water. The high-pressure cell was connected
to a chiller flowing cold water through its
walls to maintain the temperature of 4 ℃.
Figure 2. Pressure gauges and transducer (left)
and high-pressure cell with thermocouple (right)
The primary objective was to monitor
the pressure of the system as a function of
time to observe the exothermic hydrate
formation of CO2. The temperature in the cell
was also verified via the thermocouple
105
connected to the top of the cell. The
measurements were recorded using a
LabVIEW virtual interface, as seen in Figure
3.
Two trial runs were conducted under
specified experimental conditions, at
pressures of 420 psi and 550 psi, respectively.
Initially, 22 mL deionized water was added
on top of a Teflon platform in the chamber.
Figure 3. LabVIEW virtual interface displaying
pressure in cell as a function of time, recorded
with pressure transducer
The system was then sealed and purged of air
using carbon dioxide, prior to being
pressurized at the specified level for each
trial. The system was left pressurized for 1
day for the 420 psi trial and 2 days for the 550
psi trial, and data were collected using the
pressure transducer.
2.2 Model
To obtain an atomic-scale view of the
hydrate formation process, computational
models were used to represent carbon dioxide
and water interacting at specified temperature
and pressure. Interatomic interactions for
atoms i and j were modeled using both
Lennard-Jones and Coulombic potentials, per
Equation 1 [14].
𝐸𝑖,𝑗 =
{4𝜀𝑖,𝑗 [(
𝜎𝑖,𝑗
𝑟)12
− (𝜎𝑖,𝑗
𝑟)6
] +𝐶𝑞𝑖𝑞𝑗
𝜖𝑟, 𝑟 < 𝑟𝑐
0, 𝑟 ≥ 𝑟𝑐
(1)
Here, εi,j and σi,j are Lennard-Jones
parameters, C and ϵ are electrostatic
constants, q is partial charge, rc is cutoff
distance, and r is the distance between atoms
i and j. Bonding interactions for two atoms
are modeled using a harmonic bonding style,
according to Equation 2 [14].
𝐸𝑏𝑜𝑛𝑑 = 𝐾(𝑟 − 𝑟0)2 (2)
2.2.1 Derivation of K value
Equation 2 can also be written as
follows:
𝐸𝑏𝑜𝑛𝑑 =1
2𝑘(𝑟 − 𝑟0)
2 (3)
For simple harmonic motion, the frequency
may be written as:
𝑓 =1
2𝜋√
𝑘
𝑚 (4)
𝑘 = 𝑚(2𝜋𝑓)2 (5)
IR data was used to determine the equivalent
wavelength for light of a bond. Thus, the
following substitution was made for
frequency:
𝑓 = 𝑐𝜆−1 (6)
𝑘 = 𝑚(2𝜋𝑐𝜆−1)2 (7)
Finally:
𝐾 = 2𝑚(𝜋𝑐𝜆−1)2 (8)
2.2.2 Summary of modeling parameters
106
Table 1. Bonds [7][8][9][10][11][12]:
Molecule Bond r0 (nm) K (ag/ns2)
H2O O-H 0.10 3.9015e5
CO2 C=O 0.12 1.4105e6
Table 2. Interatomic potentials [13]:
Molecule Atom εi,i
(ag⋅nm2/ns2)
σi,i
(ag) qi (n⋅e)
H2O H 0 0 0.4238
O 78.18 0.3166 -0.8476
CO2 C 28.129 0.2757 0.6512
O 80.507 0.3033 -0.3256
For LJ parameters between different
types of atoms or atoms in different types of
molecules, ε and σ were combined using the
Lorentz-Berthelot rules [13].
2.3 Simulation
To set the desired pressure and
temperature, the system is first run under an
isothermal-isobaric ensemble. The timestep
is set to τ = 0.00045 nanoseconds and the
simulation is run for 10,000,000τ. After this,
a canonical ensemble is used and the
simulation is run for an additional
10,000,000τ. This second run simulates the
beginning of the hydrate formation process
once the water and carbon dioxide have been
pressurized.
3. Results
3.1 Hydrate Formation
Carbon dioxide hydrates were
observed to form in the range of pressures
specified previously. The hydrate formed
within the pressure cell with specific
crystalline structures, and upon
depressurization and warming of the chamber
the carbon dioxide began to dissipate from
the hydrate. An image of the hydrate within
the pressure chamber is seen in Figure 5.
We initially confirmed formation of
the hydrate from visual inspection of the
water during and after our experiments. After
depressurization, we observed the presence
of a white solid at 4 ℃ and less than 580 psi
– this solid could not be ice. Popping sounds
were heard coming from the ice after being
exposed to ambient conditions, which
indicated release of CO2 from the hydrate. In
addition, the pressure
Figure 5. CO2 hydrate under 550 psi pressure
and 4℃
data obtained in the experiments showed the
sharp change in slope associated with hydrate
formation. This sharp change in the pressure
vs. time graph is the result of carbon dioxide
hydrate forming rapidly after nucleation sites
are developed. The driving force for this
process is the excess pressure that is above
the hydrate transition.
107
Figure 6. Hydrate upon depressurization
Figure 7. Cell pressure as a function of time at 420 psi and 550 psi, respectively, at 4℃. The red box
shows the characteristic sharp change in slope in the graph of pressure vs time, which is indicative of
hydrate formation.
3.2 Pressure Data
Pressure data from the transducer for
each trial were plotted as a function of time.
As seen in Figure 7, a sharp drop in the
pressure was seen around 720-730 minutes
after the start of the 420-psi run, when the
pressure dropped due to hydrate formation. A
more significant pressure drop is seen in the
550-psi trial at a similar nucleation time.
3.3 Simulation Results
First, a computational model for the
simplified H2O/CO2 system was developed,
as seen in Figure 8. The positions and
orientations of water molecules were
generated using the solvate plugin of the
Visual Molecular Dynamics program
(VMD) [6]. Six molecules of carbon dioxide
were manually added to the model at
random positions and orientations. This
relatively small number of CO2 molecules
was chosen so each may be isolated from the
rest and analyzed independently after
hydrate formation. Additionally, data was
compiled on hydrate formation pressure as a
function of temperature, phase diagrams,
and the cage occupancy of carbon dioxide.
These findings are summarized in figures 9
108
Figure 8. Model of molecules of carbon dioxide
surrounded by water. Carbon (blue), oxygen
(red), hydrogen (white)
and 10. Figure 9
takes literature data and compares it
with the minimum pressure of hydrate
formation for CO2 in water. Data for pure
water was plotted using values obtained from
the Colorado School of Mines’ CSMHYD
software for prediction of hydrate
equilibrium [15].
Figure 10 shows the occupancy rates
of small (512) and large (51262) clathrate
cages, with water acting as the host
molecules and carbon dioxide the guests.
Large cages are nearly saturated with CO2 at
all tested temperatures, as indicated by the
occupancy rates in excess of 98% for all data.
Small cages do not show the same
consistency, but they begin with roughly 75%
occupancy at 0.5 ℃ and increase steadily
with both temperature and pressure. The
proposed rationale for this increase in
occupancy is that hydrate formation
pressures scale at nonlinear rates with respect
to temperature, and it seems this increase in
pressure is sufficient to negate the effects of
greater vapor pressure of CO2 at higher
temperatures.
Figures 9 and 10. CSMHYD data for pure water and CO2 hydrates.
4. Discussion
In the experimental portion of the
project, we demonstrated that a CO2 hydrate
can form at a temperature of 4 ℃ and a
pressure between 300 and 580 psi. From the
experimental data collected, the time
necessary for hydrate formation is between
720 and 730 min. The hydrate formation time
was not significantly different for the trials at
different pressures. A greater pressure drop
was observed for the trial with an initial
pressure of 550 psi. This is a result of the
greater driving force inducing the hydrate
formation of the carbon dioxide.
For our computational model and
simulation, we are able to represent the water
and carbon dioxide system using Lennard-
Jones and Coulombic interatomic potentials
109
and a harmonic bonding style. However, the
timescale of the simulation (about 10
microseconds) is much less than the time
required for hydrate formation, which is on
the order of one day. This was a necessary
restriction because the simulation time
representing this actual timescale would be
infeasibly long with our current model and
resources.
5. Conclusion
We were successful in detecting the
hydrate formation of carbon dioxide in a
pressure vessel at 420 and 550 psi. We
detected a pressure drop at the time of hydrate
formation, which was approximately twelve
hours after initiation for both trials. In
addition, we created a computational model
of the CO2 and water system. This model may
be used for future work, but in the present it
is unrealistic for us to simulate the full
hydrate formation with our modeling
parameters because of the large time relative
to the atomic scale. In future experiments,
with the addition of saline water and porous
silica we expect that the time of formation
would be less than that for our current
findings. In future experiments, we plan to
use saline water as well as porous silica glass
to more closely emulate ocean floor
conditions. In addition, we will perform
experiments at BNL and use XRD to confirm
our results. Overall, an understanding that the
nucleation time for these carbon dioxide
hydrates is approximately twelve hours will
allow for further development of the hydrate
formation process. This can then be
optimized for ocean-floor conditions with the
saline water and silica trials, eventually being
tested simultaneously with methane hydrate
formation.
Acknowledgements
We would like to express our sincere
gratitude to Mani Kuntal Sen for his guidance
and mentorship in the laboratory.
Additionally, we would like to acknowledge
Brookhaven National Laboratory for access
to their facilities and the 28-ID beamline at
NSLS II.
References
[1] United States. Energy Information
Administration. Office of Energy Analysis.
International Energy Outlook 2016. N.p.,
May 2016. Web.
[2] Kimantas, Janet. "More Methane
Surprises." AJ – Canada's Environmental
Voice. Alternatives Journal, Dec. 2014.
Web.
[3] Uchida, Tsutomu, Takao Ebinuma, and
Takeshi Ishizaki. "Dissociation Condition
Measurements of Methane Hydrate in
Confined Small Pores of Porous Glass." The
Journal of Physical Chemistry B 103.18
(1999): 3659-662. Web.
[4] Jacobson, Liam C., Waldemar Hujo, and
Valeria Molinero. "Nucleation pathways of
clathrate hydrates: effect of guest size and
solubility." The Journal of Physical
Chemistry B 114.43 (2010): 13796-13807.
Web.
[5] Horvat, Kristine, and Devinder Mahajan.
"Carbon dioxide-induced liberation of
methane from laboratory-formed methane
hydrates." Canadian Journal of Chemistry
93.9 (2015): 998-1006.
[6] Humphey, W., Dalke, A. and Schulten,
K., VMD - Visual Molecular Dynamics", J.
Molec. Graphics, 1996, vol. 14, pp. 33-38.
[7] Koga, Tadanori, et al. "Hydrate
Formation at the Methane/Water Interface
on the Molecular Scale." Langmuir 26.7
(2010): 4627-4630.
[8] “Vapor Pressure of Carbon Dioxide”.
Wolfram Alpha LLC. 2016. Wolfram|Alpha.
[9] “Bond Distance of Water”. Wolfram
Alpha LLC. 2017. Wolfram|Alpha.
[10] “Bond Distance of Carbon Dioxide”.
Wolfram Alpha LLC. 2017. Wolfram|Alpha.
[11] W.E. Acree, Jr., J.S. Chickos, "Phase
Transition Enthalpy Measurements of
110
Organic and Organometallic Compounds.”
NIST Chemistry WebBook, NIST Standard
Reference Database Number 69. Eds. P.J.
Linstrom and W.G. Mallard, National
Institute of Standards and Technology,
Gaithersburg MD, 20899. Web.
[12] Mohr, Peter J., and Taylor, Barry N., J.
Phys Chem. Ref. Data 28, 1713, 1999; Rev.
Mod. Phys. 72, 351, 2000. Web.
[13] Molecular dynamics simulation of CO2
hydrates: Prediction of three phase
coexistence line. Web.
[14] S. Plimpton, Fast Parallel Algorithms
for Short-Range Molecular Dynamics, J
Comp Phys, 117, 1-19 (1995). Web.
[15] Sloan, E. Dendy. CSMHYD. Computer
software. Center for Hydrate Research.
Colorado School of Mines, 26 Aug. 1998.
Web.
Process Optimization for a Wood Stove with a Combustion Catalyst
Yunxiao Shawn Jiang1, Kevin J. Lee1, and DongNi Qiu1
1. Chemical and Molecular Engineering Program, Materials Science and Chemical Engineering Department,Stony Brook University
Abstract Incomplete combustion of wood releases a substantial amount of carbon monoxide (CO), which is hazardous to human health. One retrofit option to reduce CO emission is the installation of a combustion catalyst. The operating condition of the catalyst affects its performance, hence particular attention is placed on the process optimization of wood stoves on improving the heating efficiency and decreasing the CO emission. In this study, a combustion catalyst was added into a wood stove that was modified with baseline features which included an airflow system (AFS) and dividers. The dividers increased the residence time of the combustion gases and slowed down the burn rate while the AFS provided better combustion gases mixing, and the combustion catalyst resulted in improved CO emission. Both the dividers and AFS features increased heating efficiency as well as CO emission. The configuration with AFS, dividers, and the combustion catalyst saw a decrease in time weight average CO (TWA-CO) concentration of 3682 ppm, a 25% decrease compared to the AFS and dividers configuration and simultaneously maintained the heating efficiency achieved from the AFS and divider configuration.
Keywords: wood stove, combustion catalyst, air flow systems, dividers, residence time, turbulence
1. Introduction During wood combustion,
specifically incomplete combustion, a substantial amount of CO is released into the air. Exposure to CO can be fatal since it deprives the body of oxygen and even when oxidized to CO2, it is still a greenhouse gas [1,2]. Limiting CO emissions can help protect human health as well as the environment. Additionally, standards for wood stoves emissions are being updated in various parts of Europe allowing for less CO emission [3].
New pellet stoves and wood stoves have improved performance compared to old wood stoves in order to meet these new standards [4]. Some ways that new stoves
improve their performance compared to old wood stoves include better mixing of fresh air and combustion gases, sufficient amounts of oxygen inside the stove, and optimal temperature inside the stove. Many different retrofit modifications are also available to improve the performance of an existing wood stove. One promising retrofit modification is to install a combustion catalyst into the stove. The purpose of installing a combustion catalyst into the wood stove is to use small quantities of noble metals, usually platinum, palladium or rhodium, on the surface of the catalyst as oxidation sites to effectively oxidize pollutants, such as CO [5]. Installing a catalyst into a domestic wood stove/boiler is
111
a viable approach in reducing emissions because of its low complexity and high compatibility. Many studies have been done to determine the viability of using a combustion catalyst. For example, the study by Hukkanen et al. showed that using a combustion catalyst can reduce emissions by oxidizing CO [6]. The catalyst used was a set of three metal wire mesh nets that consisted of platinum and palladium, encircled by a steel frame for inserting and removing from the flue stack [3]. CO emissions were reduced by 21% during the whole combustion process, and by 79% during the burn-out (the end of combustion) stage [3]. Another study with platinum cerium palladium catalyst on a wood stove showed a greater reduction of 70% for CO [3]. When considering the viability of using a combustion catalyst, how the catalyst is operated can affect performance. Studies have shown that the performance of a combustion catalyst is affected by how it is operated which affects its heating efficiency and CO emissions [7]. Side reactions may also occur inside the catalyst and must also be considered. Kaivosoja et al. showed that the side reactions that occur in the catalyst can actually increase polychlorinated dibenzo-p-dioxins and polychlorinated dibenzofurans (PCDD/F) emissions [8]. Therefore when considering the benefits of using a combustion catalyst to reduce CO, other emissions should also be taken into consideration. Additionally, the composition of the catalyst can also affect its usefulness. The study by Sudhanshu Sharma and M. Hegde looked at how to produce a three way catalyst and how it performed [9]. The catalyst used in their study was a “Ce0.98Pd0.02O2-𝛿 on γ-Al2O3,” the catalyst achieved 100% conversion of CO using a space velocity of 21,000 h-1 and below a temperature of 245 °C [9]. Lastly, process of operating a
catalyst inside the wood stove may further be optimized by installing additional features, such as a preheater for the catalyst or secondary air inlet, inside the wood stove to improve its performance.
In this study, four configurations of the wood stove were tested for the highest heating efficiency and lowest CO emission: the unmodified wood stove, the wood stove with dividers, the wood stove with AFS and dividers, and the wood stove with AFS, dividers, and catalyst. CO and O2 concentrations and flue stack temperature
were measured during the burn testsT stack and heating efficiency was calculated using the Stack Loss Method (SLM). 2. Material and Methods 2.1 Materials
Figure 1. Overview of the testing station
The wood stove used is an all cast iron, updraft, front loading stove, it has a height/width/depth of 32/28.7519.5 in. and a maximum heat output of approximately 50,000 Btu/hr, dependent on how operated [9]. A Testo 330 Flue gas analyzer was used
112
to measure CO, CO2, and O2 concentration, exhaust temperature, and draft in the flue stack through the gas analyzer port that was 3 ft above the top of the stove. The measuring range for CO was from 0 to 8000 ppm, with resolution of 1 ppm and accuracy of ± 10% of measured value; the measuring range for CO2 was from 0 to 10000 ppm, with accuracy of ± 3% of the measured value; the measuring range for O2 was from 0 to 21 vol.%, with resolution of 0.1 vol.% and accuracy of ± 0.2 vol.%. The measuring range for temperature was from -40℃ to 1200℃, with resolution of 0.1℃ and accuracy of ± 0.5℃ when measured temperature is between 0 to 100 ℃ and ± 0.5% of the measured value at the rest of range.
Approximately 20 ± 1 lb of red oak wood was used for the burn tests. Each wood log was cut into pieces of 9 inches in length, weight ranging between 1 - 4.5 lb. Moisture content was measured by Delmhorst BD-10 Pin Analog Wood Moisture meter one day in advance of the burn test. The moisture content of all wood logs fell into the range of 8 ± 1%. Wood logs were stacked uniformly and equally in weight in both chambers during the burn tests.
2.2 Dividers
Figure 2. Diagram showing the dividers and
divided firebox
The dividers separated the firebox into two equal-sized chambers for fuel load and exhaust, allowing separate ignition and forcing longer path lengths of combustion gases. The dividers were made of stainless steel. The air inlet allowed air into the firebox from the bottom of the first chamber, and combustion gases had to travel through two sharp turns to reach the outlet. The addition of dividers improved the turbulence and mixing of combustion gases and increased the residence time inside the firebox. 2.3 AFS
Figure 3. Diagram showing the components and
airflow of the AFS
The AFS was designed to increase the draft in firebox and thus increase the oxygen supply for complete combustion. The AFS was placed in the flue stack, above the gas analyzer port and injected a stream of fast moving air upwards by a blower through the venturi tube. The forced air inflow created a vacuum upstream in the flue stack and induced additional air intake in the firebox. Proof of concept experiments were done on a firebox model in lab to verify the induced draft. The additional air intake increased oxygen supply for complete combustion. It also resulted in a faster flow rate of combustion gases through the
113
firebox, which promoted better mixing in the stove. 2.4 Catalyst
A 2’’ 5.66’’ honeycomb catalyst × was obtained from FireCat Combustors, model Combustor ACI-43C. A small piece of the catalyst was taken for EDX analysis. 2.5 Burn Test
The wood stove was set up on a test stand (see Figure 1). The air draft inducer was used to regulate the air intake of the wood stove. Dividers were located inside the combustion chambers to be used to promote better mixing inside the wood stove as well as creating a secondary combustion. A combustion catalyst was installed along with a catalyst bypass arm control such that the catalyst could be inserted into and removed out of the direct path of the flue gas externally during the burn test in the back of the wood stove. Within the flue stack, Testo was inserted into the gas analyzer port to record CO and O2 concentrations. The gas analyzer port was located three feet above the top plate of the wood stove. In addition, the mass of wood loaded into the wood stove was recorded using the scale on the test stand just below the wood stove in order to standardize the test duration. For the AFS+Dividers+Catalyst configuration, the catalyst was inserted into the wood stove when the outside wall temperature of the first chamber reached 200 ℃.
Data was recorded for CO and O2 concentration and flue stack temperature every five minutes through the gas analyzer port using Testo gas analyzer and LabVIEW. The test was terminated when the amount of wood dropped below 10% of the initial weight.
2.6 Heating efficiency
The heating efficiency was determined using the Stack Loss Method, which estimates the heating efficiency of the wood stove by the flue stack temperature and the mass of dry O2, , inside the OD 2 flue stack [11].
is calculated by:OD 2
(EQ. 1) 4.7365 5.731 DO2 = 1 × O2%21%−O2%
+ 1
Which determines and , where LDG LW G and are the dry gas heat and wetLDG LW G
gas heat losses in the flue stack respectively:
(EQ. 2).001033 O T 0) LDG = 0 × D 2 × ( stack − 7
(EQ. 3).482 .004351 LW G = 9 + 0 × T stack The efficiency, HE, is then given by [11]:
E 00% L % H = 1 − DG − LW G − 1 (EQ. 4) Both O2% and were measured during T stack the burn test at a 5 minute interval. 2.7 Scanning Electron Microscope (SEM) for catalyst characterization
An SEM analysis was done to determine the composition of the catalyst used. The catalyst sample was taken from the catalyst by using metal pliers to chip off a piece of the catalyst. The sample was mounted on carbon tape and was spot coated with a 2 nm thick gold palladium coating to improve the conductivity of our sample. The SEM was done using a focused ion beam scanning electron microscope (ZEISS Crossbeam 340). During the SEM, eight different locations on the catalyst sample were observed. Electron images, EDS layered images, and a map sum spectrum was acquired for each of the eight locations.
114
3. Results and Discussion 3.1 Characterization of the Catalyst
Energy Dispersive X-ray Spectroscopy (EDS) layered images of the catalyst sample revealed how elements are dispersed along the surface of the catalyst.
Figure 4A-E. Electron Image from SEM test of catalyst, EDS layered images of Al, O, Na, and Ce
respectively
Figure 4A showed an electron image of sampled location on the catalyst surface while EDS layered images of the location revealed how different elements were concentrated on the catalyst surface. At this location, O and Al could be found in most places on the surface of the catalyst. This leads to the conclusion that O and Al are part of the inert support of the catalyst. Meanwhile Na was concentrated in small
clusters where Al and O was not present. This may be due to Na being introduced onto the catalyst surface as a contaminant and entering crevices along the surface of the catalyst. Very little Ce was found on the catalyst surface; however, figure 4E shows traces of Ce dispersed throughout the surface of the catalyst suggesting that it is part of the noble metal washcoat.
Figure 5. EDS analysis of the catalyst sample
115
Figure 5 showed that the catalyst sample had negligible amounts of many different elements. Some of these elements could be contaminants on the surface of the catalyst. For example, Fe might have originated from the metal pliers that were used to acquire the catalyst sample. Additionally, the sample was mounted using carbon tape, so carbon could have been introduced to the sample as well. While contamination may have been present on the
surface of the catalyst, the presence of Ce, Al, and O suggests that the catalyst used in this test may be similar to the one used by Sudhanshu Sharma and M. Hegde which had a Ce0.98Pd0.02O2-δ wash coat and a monolith consisting of γ-Al2O3 [9]. Since the composition of the catalyst used here is similar to the one used by Sharma and Hedge, the catalyst should also have similar characteristics, such as CO conversion rates.
Figure 6. CO concentrations of tests with different configurations
116
Figure 7. O2 concentrations of tests with different configuration
3.2 CO Emission and O2 Concentration
The graphs of CO emission and O2 concentration are shown above for each of the configurations and combustion phases are labeled with numbers and dash lines. Phase 1 and 2 are the start-up phases (ignition of the first chamber) of the AFS+Dividers configuration and Catalyst configuration respectively. Phase 3 and 4 represent the ignition of the second chamber for the AFS+Dividers configuration and Catalyst configuration respectively.
From the CO emission and O2 concentration data (Figures 6 and 7), there were multiple sharp decreases during the burn. Comparing Figures 6 and 7, nearly every sharp decrease in O2 corresponds to an increase in CO caused by the incomplete combustion of individual wood logs. The data from the AFS+Dividers, Dividers, and Catalyst configurations verified the
conclusion where oxygen decreases can be observed between 50 - 70 minutes when the second chamber was reported to be flaming from visual observation and temperature monitoring. Incomplete combustion resulted in higher CO due to insufficient combustion time, low temperature, and poor mixing between oxygen and the combustion gases. The dividers addressed the problems of insufficient time and poor mixing by increasing the residence time and turbulence inside the firebox. The AFS increased draft into the firebox, generated more turbulence, and therefore assisted the combustion.
All three modified configurations showed three major trends: the start-up phase, second chamber ignition phase, and the burn-out phase. The start-up phase from 0 - 20 minutes had high CO emission (see Figure 6) from incomplete combustion caused by low temperature (cold start) of the
117
wood stove. The CO emission decreased as temperature of the wood stove increased. A significant increase in CO emission occurred at 40 - 80 minutes due to the start-up phase of the second chamber. Temperature decreased as the burn test reached the burn-out phase. This resulted in another increase of CO emission that was lower than the two start-up increases in CO emission due to less wood mass.
The unmodified wood stove had the lowest overall CO emission compared to the other three modified configurations. Additionally, it had a relatively high overall O2 concentration due to the lack of dividers, which enabled more airflow despite the absence of AFS. Another reason for the high O2 concentration was due to the unmodified stove having a high burn rate. Figure 7 showed that after 50 minutes into the test, O2 concentration began increasing. This was because after 50 minutes, the burn test reached the burn out phase. The low CO emission from the unmodified test was because the unmodified stove reached higher temperatures than the other stove configurations which allowed for more CO to be converted into CO2.
The dividers configuration had less CO emission than the unmodified stove after the ignition of second chamber (60 - 80 minutes)because the combustion in the second chamber acted as an afterburner for the emissions from the first chamber, converting CO into CO2. Turbulence of combustion gases inside the firebox increased due to the introduction of dividers changing the pathway of air. The dividers also increased gases residence time inside the wood stove. The incomplete combustion caused by low temperature was most notable during the cold start phase of the tests. The addition of dividers created a second ignition which was a warm start due to the increased stove temperature. The O2
concentration after the second chamber ignition (60-80 minutes) was lower than the case when the first chamber was ignited (20-40 minutes). Similarly, the CO concentration after the ignition of second chamber (60 - 80 minutes) was lower than after the ignition of the first chamber (20-50 minutes). However, the drawback was that the overall oxygen concentration was low when compared to the unmodified stove. In order to increase oxygen concentration inside the wood stove, the AFS was installed along with the dividers.
The AFS+Dividers configuration had higher CO concentration compared with the dividers and the unmodified configurations. Two peaks could be observed during 20 - 40 minutes and 60 - 80 minutes, which correspond to the ignition of two chambers. The induced airflow through the stove helped better mixing of combustion gases in the stove, but shortening the residence time of gases. The disadvantage could be resolved by adjusting the power of AFS during the burn-out phase, but for the consistency of burn tests, AFS was kept at the maximum power throughout the tests. The higher O2 concentration, especially during 20 - 60 minutes, demonstrated the AFS was functioning as intended. Therefore, an addition of the catalyst could be used to oxidize CO.
The catalyst configuration had an overall lower CO emission than the AFS+Dividers configuration. The wall of the first chamber reached 200℃ approximately 15 minutes into the test, at this time the catalyst was moved into the pathway of the gas inside the wood stove at this time. Approximately 40 minutes into the test, there was a drastic decrease in O2 concentration in the wood stove. This is likely due to the second chamber igniting and the oxidation reaction occurring inside the catalyst. During this time, lower CO
118
emissions compared to the AFS+Dividers configuration were observed, likely due to oxidation of CO by the catalyst. However, the catalyst used during the test did not perform as well as the catalyst used by Sharma and Hedge. At 220℃, the catalyst used by Sharma and Hedge achieved 100% CO conversion, while the catalyst used in this test achieved a maximum drop in CO concentration of 2000 ppm or ~25% [9]. This could be due to the catalyst not being
sealed properly allowing gases to travel out of the wood stove without having to go through the catalyst. Additionally, CO emissions during the burnout phase were stable at 3000 ppm suggesting the catalyst was not active during burnout phase. The TWA-CO for AFS+Dividers and Catalyst configuration were 4612 and 3682 ppm respectively. The 25% decrease confirmed the effectiveness and aligned with the literature data (21%) [3].
Figure 8. Heating efficiency of tests with different configuration 3.3 Heating Efficiency
Figure 8 showed an overall decreasing trend of the heating efficiency as a function of time. Phases of all configurations are labeled with numbers and dash lines. Phase 1 and 2 represent the start-up phase of the AFS+Dividers configuration and the Catalyst configuration
respectively. Phase 3 and 4 represent the ignition of the second chamber for the AFS+Dividers configuration and Catalyst configuration respectively. Lastly, phase 5 represents the burn out phase of all configurations.
All three modified configurations showed an increase in heating efficiency at the onset of the second chamber, which
119
suggested dividers improve the heating efficiency of the wood stove. Meanwhile, the unmodified wood stove had lower heating efficiency when compared to each of the three modifications done. This is because there was no staged combustion in the unmodified stove and all of the wood was available for combustion at the beginning of the burn. It also further verifies the high temperature inside the flue stack of the unmodified configuration by equation 2 ~ 4 where high flue stack temperature corresponds to more CO oxidation and less heating efficiency. Therefore, combustion occurred at a much faster rate than staged combustion which decreased CO emissions but also decreased heating efficiency.
The increase in heating efficiency during the Dividers test compared to the unmodified configuration was due to staged combustion. Since not all of the wood was immediately available for combustion, the burn was prolonged, and maintained a higher heating efficiency. Figure 8 shows a jump in heating efficiency from 60-80 minutes into the test, this correlates to the second chamber of wood igniting during the test.
The AFS+Dividers and Catalyst configurations overlapped with each other throughout most time during the test, and their heating efficiency were higher than the unmodified and divider only configurations. The improved performance in efficiency was attributed to AFS as it increased the flowrate of gas into and out of the firebox, therefore lowered the temperature of the flue gas, which resulted in higher efficiency. The addition of a combustion catalyst caused additional exothermic reactions when oxidizing volatiles and CO. However, the Catalyst configuration did not produce significantly higher efficiency compared with AFS+Dividers, because the temperature of the flue gas had not reached
the optimum temperature required for the catalyst. From the data, the catalyst had noticeable better heating efficiency from 100 minutes onward when compared to the AFS+Dividers configuration from the blockage of the catalyst. 4. Conclusion
From the results of testing different configurations of the wood stove, the AFS+Dividers configuration provided the best heating efficiency, but also emitted the most CO. The unmodified wood stove gave the lowest CO emissions, but also had the lowest heating efficiency. Using the catalyst with the AFS and dividers provided high heating efficiency while emitting significantly less CO than the AFS and dividers modifications. Therefore, using a combustion catalyst with an AFS and dividers would be the optimal configuration for retrofitting an old wood stove.
Acknowledgements Funding for this project is provided
by the New York State Energy Research and Development Authority (NYSERDA award № 63043). Special thanks to NYSERDA and Robert Carver, Brookhaven National Laboratory and Dr. Tom Butcher, Stony Brook University, Jake Lindberg and Professors Devinder Mahajan, Sotirios Mamalis, Benjamin Lawler, Jon Longtin, and Simon Chung-Chueh Chang. References [1] Veronesi, Bellina, et al. "Particulate Matter Inflammation And Receptor Sensitivity Are Target Cell Specific." Inhalation Toxicology 14.2 (2002): 159-183. Academic Search Complete. Web. 16 [2] Environmental Protection Agency (EPA), Air Sources Emissions
120
[3] Ozil, F., Tschamber, V., Haas, F., Trouvé, G., 2009. Efficiency of catalytic processes for the reduction of CO and VOC emissions from wood combustion in domestic fireplaces. Fuel Processing Technology 90, 1053e1061. [4] Ozgen, Senem, et al. "Emission Factors from Small Scale Appliances Burning Wood and Pellets." Atmospheric Environment, vol. 94, Sept. 2014, pp. 144-153. EBSCOhost, doi:10.1016/j.atmosenv.2014.05.032. [5] Bensaid, S., et al. "After-Treatment of Household Wood-Fired Stove Emissions: From Catalyst Formulation to Full-Scale System." Catalysis Today, vol. 197, no. 1, 15 Dec. 2012, pp. 76-89. EBSCOhost, doi:10.1016/j.cattod.2012.06.026. [6] Hukkanen, A., et al. “Reduction of Gaseous and Particulate Emissions from Small-Scale Wood Combustion with a Catalytic Combustor.” Atmospheric Environment, vol. 50, Apr. 2012, pp. 16- 23. EBSCOhost, doi:10.1016/j.atmosenv.2012.01.016. [7] Thran, Daniela, et al. "Improvement Of Efficiency And Emissions From Wood Log Stoves By Retrofit Solutions." Chemical Engineering & Technology 2 (2017): 340. Academic OneFile. [8] Kaivosoja, T., et al. "Effects of a Catalytic Converter on PCDD/F, Chlorophenol and PAH Emissions in Residential Wood Combustion." Chemosphere, vol. 88, no. 3, July 2012, pp. 278-285. EBSCOhost, doi:10.1016/j.chemosphere.2012.02.027. [9] Sharma, Sudhanshu and M. Hegde. "Single Step Direct Coating of 3-Way Catalysts on Cordierite Monolith by Solution Combustion Method: High Catalytic Activity of Ce0.98Pd0.02O2-δ." Catalysis Letters, vol. 112, no. 1/2, Dec. 2006, pp. 69-75. EBSCOhost, doi:10.1007/s10562-006-0166-z.
[10] “Vermont Castings Operation Manual.” Hearth, https://www.hearth.com/i mages/uploads/vcresimanual9_17.pdf. 7 May 2017. [11] “Fired Steam Generators”, Performance Test Code 4: 2008, ASME, New York, NY; the US std. since 1998.
121
122
In-situ Water Management for the Optimization of
Methanol Dehydration to Dimethyl Ether
Jillian Gannon 1, JinYing Lin 1, Danielle Wahl 1
1. Chemical and Molecular Engineering Program, Stony Brook University, Stony Brook NY
11794
Abstract
The production of Dimethyl ether (DME) on a large scale is urgent as it is a promising
clean fuel. This study focuses on DME production through methanol dehydration, which is
catalyzed by a solid acid, typically γ-alumina (γ-Al2O3). Previous studies have found that the
byproduct water inhibits methanol dehydration through competitive adsorption with methanol on
the active sites of the catalyst. This poisons the catalyst and renders it inactive. In several
published papers, simulations of in-situ water removal in methanol dehydration, which is
designed to be accomplished by sweep-gas or membrane separation in fixed bed reactors, show
that methanol conversion is improved with in-situ water removal. Sweep-gas and membrane
separation methods cannot be performed in batch reactor so it was determined that using a water
absorbent/adsorbent might be a solution for solving this issue. Molecular sieves have been found
to be an excellent water adsorbent. This study aims to determine the ability of molecular sieves
that have a 5Å pore size to adsorb the water as it is generated in the batch reactor experiment in
order to improve the conversion of methanol to DME. In each of the nine experimental runs, the
reactor pressure was found to increase isothermally at a stable set temperature of 280 °C and
analyzing the product using gas chromatography (GC) gave results that proved that DME was
produced. The GC results also helped to determine the methanol conversion baselines: the γ-
Al2O3 baseline as 26.7% ± 8.8%, the molecular sieve baseline as 25.5% ± 6.2%, and the
experiment, ran with both γ-Al2O3 and molecular sieves, as 47.5% ± 8.6%. The results proved
that the molecular sieves improved methanol conversion. However, DME production data is
needed for further evidence that the improvement in the conversion of methanol to DME was
achieved by molecular sieves.
Keywords: Dimethyl ether, methanol dehydration, water removal, molecular sieves.
1. Introduction
Due to an increasing energy demand
and a reduction in the supply of fossil fuels,
focus has shifted to the use of bioenergy to
replace other energy sources. Due to its
abundance and renewable properties, it has
been shown that biomass could be a viable
source for energy [1]. Dimethyl ether
(DME) is a promising clean fuel alternative
with similar thermal efficiencies to diesel
fuel. At ambient conditions, DME is a
colorless, non-toxic, highly flammable gas.
It can be handled as a liquid when slightly
pressurized [1][2]. Compared to traditional
diesel fuel, DME has low NOx emissions,
almost no smoke and less engine noise. In
addition, DME is a useful chemical
123
intermediate for the preparation of many
important chemicals including dimethyl
sulfate and high value oxygenated
compounds [1][2]. Moreover, DME has
been used as an aerosol propellant to replace
chlorofluorocarbons which can destroy the
stratospheric ozone layer [3]. Consequently,
the utilization of DME is environmentally
favored.
DME can be produced through
synthesis directly from CO, CO2, and H2,
known as synthesis gas or syngas, or by
methanol (MeOH) dehydration catalyzed by
a solid acid, typically γ-alumina. For several
decades, the production of DME has been
focused on being synthesized directly from
syngas. Because of this, the role of the
active components and the reaction
mechanism is well understood. However,
the catalytic dehydration of methanol to
DME has not been as well studied [1].
Figure 1: Production of DME from syngas [10]
Many catalysts have been used and
modified to optimize the production of DME
from either syngas or methanol such as H-
ZSM-5 and Cu/ZnO/Al2O3 [1]. The problem
with the production of DME from methanol
or syngas is the byproduct of water shown in
Figure 1. Two methanol molecules combine
over a catalyst, which is typically γ-alumina
(γ-Al2O3), to form DME with water as a
byproduct. The water poisons and
deactivates the catalyst by blocking the
active sites on the catalyst and preventing
the consumption of methanol through
competitive adsorption. This results in lower
DME production and shortened catalyst
bedtime [1].
To address this issue, in-situ water
removal is proposed to improve the yield of
DME and avoid the poisoning of the
catalyst. Due to the impact water has on the
reaction, various methods have been
analyzed to limit the amount of water
present in the reaction. Many published
papers have shown that limiting the amount
of water present in the reactor can improve
the conversion of methanol to DME when
using methods such as the sweep-gas
method in fixed bed reactors, membrane
separation method in membrane reactors
with hydrophilic membranes, or the
combination of membrane and sweep gas
[4][11]. In one simulation, methanol
dehydration using a membrane reactor was
modified with a high water perm-selective
membrane and a conventional membrane
reactor. The result showed methanol
conversion in the conventional membrane
reactor was 68%, while with the modified
membrane reactor it was 82.5% [4]. The
methods proposed in these previous papers
are not applicable for a batch reactor,
however these results suggest that inorganic
water absorbents or adsorbents may be a
suitable component useful for water removal
in a batch reactor.
Molecular sieves have shown to
perform well in the adsorption of water.
124
Molecular sieves are a type of desiccant
primarily composed of alumina, silica, and
other metal oxides, with uniform pore-size
distributions. They are widely used in
heterogeneous catalysis [9]. It has been
shown that using molecular sieves alongside
a catalyst in the production of DME from
methanol can not only improve diffusion in
the pores, but also improve activity,
selectivity, and stability [6]. Additionally,
molecular sieves have been used to separate
water-based azeotropic mixtures such as
ethanol in water and isopropanol in water
[7]. Molecular sieves of 5Å diameter are
often used in the petroleum industry
specifically for separating compounds in
chemistry laboratories and drying starting
materials in a reaction. Due to the precise
and uniform sized pores, these molecular
sieves are commonly used as adsorbents for
both gases and liquids [9].
Our study focuses on the in-situ
water management for the optimization of
methanol dehydration to DME using γ-
Al2O3. We hypothesize that in situ water
removal can be assisted by using a
molecular sieve 5A desiccant which can
improve the conversion of methanol to DME
when using a batch reactor. To accomplish
in-situ water removal in a batch reactor,
molecular sieves with a 5 Å pore size were
used as a water adsorbent. Such molecular
sieves have been found to be particularly
suitable to absorb water [8]. In this study,
molecular sieves of a 5 Å diameter were
placed in a batch reactor in order to limit the
amount of water present to prevent the
poisoning of the catalyst. Methanol
dehydration reactions were run both with
and without the water adsorbent. The
methanol conversions of these products
were compared in order to confirm that
molecular sieve use improves the conversion
of methanol to DME.
2. Materials and Methods
2.1 Chemicals
Methanol (99.8%, anhydrous, 322415)
and polyethylene glycol (PEG-400, 81172),
were purchased from Sigma-Aldrich and
were used as the reactant and solvent,
respectively, for the synthesis of DME in a
standard batch reactor. Commercial gamma-
alumina, purchased from Acros Chemicals,
was used as a catalyst. Molecular sieves (5
Å, 8 to 12 mesh, 197295000), purchased
from Acros Chemicals, were used as water
adsorbents. Nitrogen gas, purchased from
Airgas, was used for purging and
pressurizing the system. DME (≥99%,
anhydrous, 295299), purchased from Sigma
Aldrich, was used as a reference sample for
GC-TCD analysis.
For GC-TCD analysis, Helium gas,
purchased from Airgas, was used as the
carrier. The molecular sieves 5A and γ-
Al2O3 were dried in a Thermo Scientific
HeraTherm oven at 120 °C before being
used for moisture removal.
2.2 Batch Reactor Setup
The batch reactor used to run all
experiments was a Parr model 4547 batch
unit fitted with a 300 mL pressure vessel.
This batch unit was connected to a model
4848 process controller and a Parr model
4831 temperature controller. The batch
reactor setup is shown in Figure 2 below. In
each run, the reaction took place in the
reactor vessel (A) that connected to the
125
reactor head (B) and placed on the reactor
stand (C). The reactor controller (D)
controlled the stirring speed and monitored
the reactor conditions. The heating mantle
(E) was connected to the temperature
controller (F) to control the temperature of
the system due to heating mantle
incompatibility with the reactor controller.
As shown in Figure 3, the reactor head had a
pressure transducer (1) and an automatic
stirrer (2), which was connected to the
reactor controller, a secondary pressure
gauge (3), and two thermocouples (4). One
of the thermocouples was connected to the
reactor controller for reporting the system
temperature and the other was connected to
the temperature controller for controlling the
temperature of the system. The reactor head
also included a gas inlet (5), gas outlet (6),
and dip tube (7).
Figure 2. Parr batch reactor setup Figure 3. Reactor head setup
Table 1. Type and amount of chemicals used in the three experiments
Experiment # 1 (Control Group 1) 2 (Control Group 2) 3 (Experimental Group)
Description γ-Al2O3 Baseline Molecular Sieves
Baseline
Experiment
Methanol (mL) 10 10 10
PEG-400 (mL) 60 60 60
γ-Al2O3 (g) 1.0 0 1.0
Molecular Sieve 5A (g) 0 11.0 11.0
1
2.3 Synthesis of Dimethyl Ether
Using the batch reactor, three
experiments were run using the same
quantities of methanol and PEG-400, but
with differing amounts of γ-Al2O3 and
molecular sieves. Specific amounts are listed
in Table 1. Each experiment was run three
times. In the experimental run, 60 mL of
PEG, 10 mL of methanol, 1.0 g of γ-Al2O3
and/or 11.0 g of molecular sieves were
placed into the 300 mL vessel. The reactor
was sealed and purged with N2. During
purging, the stirrer was turned on to ensure
that the dissolved air leaves the system.
After purging, the reactor was pressurized to
64 psig using N2 gas. The mixture was
stirred for fifteen minutes and an initial
liquid sample was obtained from the vessel
through the dip tube valve. Each reaction
started with initial conditions of 25 °C, 60
psig, and 300 rpm. Temperature and
pressure were monitored using the reactor
controller and were recorded every fifteen
minutes throughout each experiment. All
experiments were run isothermally at 280 °C
for three hours. After three hours, the batch
reactor was cooled to room temperature and
a final liquid sample was obtained. A gas
sample was obtained from the gas outlet
valve.
2.4 Methanol Adsorption Test
In order to determine the amount of
methanol that would be adsorbed by the
molecular sieves, 6 mL PEG-400, 1 mL
methanol, and 1.1 g molecular sieve were
added to a small glass vial. In another glass
vial, 6 mL PEG-400 and 1 mL methanol
were added. These two vials were stirred
and placed in the sonicator bath until the
mixture became uniform. Methanol
concentration in the vials was measured by
GC after a three-day period.
2.5 Calibration for Methanol
concentration and peak area
In order to calibrate the methanol
peak area from the GC results and the
methanol concentration in the sample, five
methanol/PEG-400 samples containing
different methanol concentrations were
prepared by mixing different volumes of
methanol with 6 mL PEG-400 in small vials.
The samples were placed in the sonicator
bath until the mixture became uniform.
2.6 Gas Chromatography (GC)
A Gow Mac series 580 Gas
Chromatograph with Thermal Conductivity
Detector (TCD) (column 8’, packing 0V101)
was used to measure the methanol and DME
concentrations in the samples. Helium gas
was used as the carrier gas with flow rate
measured by the ProFLOW 6000 from
Restek and was adjusted to 20-30 mL/hr.
The GC injector temperature, detector
temperature, and column temperature were
set to 150 °C, 190 °C, and 150 °C,
respectively. The detector current was set to
200 mA. For data acquisition and analysis,
the Clarify PC chromatography software
was set to collect the voltage output data for
10 minutes. For liquid samples, a 5 μL
sample was measured using a 10 μL
Hamilton model gas tight 1801 glass syringe
and was used to inject the sample into the
injector port. In the analysis of gas samples,
the sample size was 5 μL initially, but was
later increased to a sample size of 0.5 mL.
After a sample was injected into the GC, the
127
2
software was started for data acquisition.
Samples of pure methanol and pure DME
were used for calibration because GC is
based on comparative analysis. For each
sample, three measurements were made.
3. Results and Discussion
3.1 Batch Reactor Run Data
For the purpose of this study, the γ-
Al2O3 (catalyst) baseline is defined as
Experiment 1 and the molecular sieve
baseline is defined as Experiment 2. Three
runs were performed for each of the three
experiments, totaling nine experimental
runs. In each experimental run, reactor
temperature and pressure were recorded
every 15 minutes, as shown in Figure 4.
In Figure 4a, the reactor temperature
reached the desired set point of 280 °C in
roughly 30 minutes and generally
maintained that temperature for the
remainder of the reaction. In Figure 4b, the
reactor pressure increased sharply in the first
half hour due to the quick increase in
temperature. After t = 30 minutes, pressure
continued to increase gradually and then
leveled off at t = 90 minutes.
The increase in reactor pressure after
the temperature stabilized at the set point
indicated that methanol was converted to
DME. This is due to the fact that the vapor
pressure of DME is much higher than that of
methanol. When the pressure leveled off and
even slightly dropped, it indicated that the
equilibrium was reached and the reverse
reaction took place. Additionally, the results
show that the experiment time could be
shorten to 2 hours due to the leveling off of
the pressure.
128
2
Figure 4. Reactor Temperature (a) and Pressure (b) as a function of time for nine runs, Dotted lines with
square markers represent runs for Experiment #1 Catalysis Baseline, Dashed lines with diamond markers
represent runs for Experiment #2 Molecular Sieve Baseline, Solid lines with circular dot markers
represent runs for Experiment #3
3.2 GC Results
GC-TCD was used to determine the
methanol concentration in the liquid samples
and the DME concentration in the gas
samples. Pure methanol and DME samples
were used as references. Figure 5 shows the
voltage output was collected during GC data
acquisition. The software recorded the peaks
and reported the retention time, peak
intensity, and peak area. The recorded peaks
for different molecules have different
retention times which are listed in Table 2.
Table 2. Retention times of different molecules
Chemical Water MeOH DME
Retention
time (min) ~2 6-8 5-6
The peak intensity and area are
directly proportional to the chemical
concentration in the sample. As the
concentration of a chemical in the sample
decreases, the peak intensity decreases.
Figure 5 is a GC result example for the final
liquid samples of experiment #3. The DME
peak can be seen on the left and the
methanol peak on the right. The GC spectra
of the final liquid sample of all three
experiments have DME and methanol peaks
as shown in Figure 5. The presence of the
DME peak in the liquid sample may have
resulted from collecting the final liquid
samples during pressurization, therefore a
portion of DME product liquefied and
dissolved in the liquid mixture. The DME
peak helps to confirm that DME was
produced in all three experiments.
129
128
1
Figure 5. GC spectrum for Experiment #3 final liquid sample
The GC results for the gas sample (5
μL) of Experiment #1 and #2 only have a
water peak present. After the gas sample
size was changed to 0.5 mL, the GC result
of the gas sample (0.5 mL) showed peaks
for water, DME, and a very small amount of
methanol. The GC results for the gas
samples (0.5 mL) of Experiment #1 and #2
also showed a huge water peak and a small
DME peak.
The DME peak in gas samples also
helped to confirm that DME was produced
in all three experiments. The water peak is
higher than the DME peak in the 0.5mL GC
results and only a water peak was present in
the GC result of a 5 μL gas sample. This is
because water has a higher thermal
conductivity than DME. The gas samples
were collected using a syringe at different
pressures and flow rates because the flow
rate could not be controlled using the gas
outlet valve present in Figure 3. This means
that the gas samples were collected at
different conditions. Consequently, the
DME peaks in the GC results for gas
samples could not be compared to each
other, nor to pure DME. DME and water
concentrations and yields in the gas sample
could not be determined.
Experiment #2, involving only
molecular sieves, was not expected to have
any methanol conversion to DME, due to a
lack of a catalyst. However, the GC results
suggested that DME was produced.
Molecular sieves are primarily composed of
alumina and silica [9]. Methanol
dehydration to DME generally is catalyzed
by a solid acid catalyst. The catalyst activity
of a silica-alumina catalyst in DME
production was studied in previous work [1].
Therefore, DME production in Experiment
#2 should be a result of the methanol
conversion catalyzed by alumina and silica
in molecular sieves.
130
2
Figure 6. Methanol Peak Area vs. Methanol wt% in Methanol/ PEG-400 solution
The reported areas under the
methanol peak in the GC data of liquid
samples were used in the calculation of the
methanol conversion for the experiments.
By measuring the methanol peak areas for
five methanol/PEG-400 solution samples
containing different methanol
concentrations, a calibration of methanol
peak area and methanol concentration (wt%)
was performed (Figure 6). A fitting curve
was generated to determine the linear trend
line. The slope was 220.8 ± 2.3 and y-
intercept was -4.4 ± 18.2.
Table 3 shows the average methanol
peak areas of the initial and final liquid
samples for the three experiments and the
methanol concentration, which was
converted using the calibration above. The
percent methanol conversion was
determined by comparing the concentration
difference between the initial and final
liquid samples to the initial concentration.
A methanol adsorption test was
conducted by comparing the methanol
concentration of the sample with molecular
sieves to the one without molecular sieves.
The relative difference is 9.4 % ± 1.1 %.
This result showed that 9.4% of the
methanol in the sample was adsorbed by
molecular sieves. Methanol conversions of
Experiment #2 and #3 were adjusted based
on the adsorption test result. This adjustment
can be seen in Table 3.
Figure 7 shows that methanol
conversion in Experiment #3 is about twice
that of each of the baselines. The results
indicate that methanol conversion is
improved by the addition of molecular
sieves. However, in order to confirm that
methanol was only consumed in its
conversion to DME, additional DME
131
1
Table 3. Determination of methanol conversion
Sample
MeOH Peak
Area Avg.
(mV*s)
Area meas.
error (mV*s)
MeOH
wt%
MeOH
wt%
error
MeOH
Conversion
%
MeOH
Conversion
% error
Initial
liquid 2484.0 ±130.9 11.3 ± 0.8 - -
Exp. #1
Final
liquid
1820.4 ±122.3 8.3 ± 0.7 26.7 ± 8.8
Exp. #2
Final
liquid
1618.1 ±40.0 7.3 ± 0.3 25.5 ± 6.2
Exp. #3
Final
liquid
1071.8 ±128.2 4.9 ± 0.7 47.5 ± 8.6
production data is needed. It in unconfirmed
whether the amount of DME produced is
equivalent to amount of methanol consumed
in each experiment. It is also necessary to
characterize the fresh and used molecular
sieve samples in order to determine its
function in water removal and methanol
conversion.
Figure 7. Methanol conversion % for
Experiment #1 (with γ-Al2O3 only), Experiment
#2 (with molecular sieves only) and Experiment
#3 (with both γ-Al2O3 and molecular sieves).
4. Conclusions
In this study, three experiments were
run with the same amount of methanol and
PEG-400. Experiment #1 was run with γ-
Al2O3 only, Experiment #2 with molecular
sieves only, and Experiment #3 with both γ-
γ-Al2O3 and molecular sieves. From the
experiment data, the increase in pressure at
an isothermal temperature of 280 °C
indicated the conversion of methanol to
DME. The stabilization and then decrease in
pressure after 2 hours shows that the
experiment time could be reduced to 2
hours. This will reduce energy input.
The GC results indicated that DME
was produced in all three experiments. The
production of DME in Experiment #2 was
the result of methanol conversion catalyzed
by the alumina and silica components in the
molecular sieves. The methanol conversion
of three experiments was also determined
from the GC results: Experiment #1, 26.7 %
± 8.8%; Experiment #2, 25.5% ± 6.2%;
Experiment #3, 47.5% ± 8.6%.
The high methanol conversion in
Experiment #3 indicated that methanol
conversion is improved by the addition of
molecular sieves. However, additional DME
production data is needed in order to
132
123
confirm whether the amount of DME
produced is equivalent to amount of
methanol consumed in each experiment.
This information is needed confirm that
methanol is only consumed in the
conversion to DME.
Future work would include
additional runs for each of the three
experiments to obtain sufficient DME
production data and the characterization of
fresh and used molecular sieves to establish
its role as an in-situ water removal material
during conversion of methanol to DME. The
particle diameter of the molecular sieves in
relation to the particle size of methanol
should also be considered. The molecular
sieve used was 5 Å and the diameter of
water is only 3 Å. It could potentially be
beneficial to use a molecular sieve with a
smaller pore size (3 Å or 4 Å) in order to
reduce the amount of methanol absorbed by
the molecular sieves while still being able to
absorb water.
5. Acknowledgements
We greatly appreciate the support we
received from Dr. Devinder Mahajan,
Elizabeth Taveras and Nyima Choephell.
This work was performed at the Advanced
Energy Research and Technology Center
(AERTC) in Stony Brook, New York,
11794.
6. References
[1] Jun, K.W.; Lee H.S.; Roh, H.S.; Park,
S.E. Catalytic Dehydration of Methanol to
Dimethyl Ether (DME) over Solid-Acid
Catalysts. Bulletin of the Korean Chemical
Society 23.6 (2002): 803-07. Web. 29 Mar.
2017.
[2] Lee, S.G.; Sardesai, A. Liquid phase
methanol and dimethyl ether synthesis from
syngas. Topics in Catalysis, Volume 32,
Issue 3-4 (2005): 197-207. Paper.
[3] Jiang, S.; Hwang, J.S.; Jin, T.H.; Cai,
T.X.; Cho, W.; Baek, Y.S.; Park, S.E.
Dehydration of Methanol to Dimethyl Ether
over ZSM-5 Zeolite. Bulletin Korean
Chemistry Society, Volume 25, Issue 2
(2004): 185-189. Paper.
[4] Iliuta, I.; Larachi, F.; Fongarland, P.
Dimethyl Ether Synthesis with in-Situ H2O
Removal in Fixed-Bed Membrane Reactor:
Model and Simulations." Industrial &
Engineering Chemistry Research 49.15
(2010): 6870-877. Web. 29 Mar. 2017.
[5] Liu, D.H.; Yao, C.F.; Zhang,J.Q.; Fang,
D.Y.; Chen, D.S. Catalytic Dehydration of
Methanol to Dimethyl Ether over Modified
γ-Al2O3 Catalyst. Fuel 90.5 (2011): 1738-
742. Web. 29 Mar. 2017.
[6] Banat, F.; AI-Asheh, S.; AI-Lagtah, N.
Adsorptive Distillation Using Molecular
Sieves and Low-cost Biobased Adsorbents
for the Break-up of the Isopropanol–Water
Azeotrope. Adsorption Science and
Technology, Volume 21, Issue 9 (2003): 821
- 830. Paper.
[7] Tang, Q.; Xu, H.; Zheng, Y.Y.; Wang,
J.F.; Li, H.S.; Zhang, J. Catalytic
Dehydration of Methanol to Dimethyl Ether
over Micro/mesoporous ZSM-5/MCM-41
Composite Molecular Sieves. Applied
Catalysis A: General 413-414 (2012): 36-
42. Web. 29 Mar. 2017.
[8] Komarneni, S.; Pidugu, R.; Menon, V.C.
Water adsorption and desorption isotherms
of silica and alumina mesoporous molecular
sieves. Journal of Porous Materials, Volume
3, Issue 2, (1996): Pg. 99-106. Paper.
[9] n.a. Molecular Sieves - Technical
133
124
Information Bulletin. Sigma Aldrich. Sigma-
Aldrich Corp., n.d. Web. 31 Mar. 2017.
[10] n.a. DME(Dimethyl Ether). Toyo
Enngineering Corporation. Web. 21 Apr
2017. http://www.toyo-
eng.com/jp/en/products/energy/dme/
[11] Lee, K.H.; Youn, M.Y.; Sea, B.K.
Preparation of hydrophilic ceramic
membranes for a dehydration membrane
reactor. Desalination, Volume 191, Issues 1–
3 (2006): 296-302. Paper.
134
JUCER Volume VI
Spring 2017