2020 Ingenium - Journal of Undergraduate Research

Page 1

Undergraduate Research at the Swanson School of Engineering

University of Pittsburgh Swanson School of Engineering Undergraduate Research Benedum Hall, 3700 O’Hara Street, Pittsburgh, PA 15261 USA Spring 2020

The image on the cover shows collagen gels seeded with porcine aortic smooth muscle cells. Fibrillar collagen is shown in blue and fibroblasts are shown in red. (See page 92 by Hannah Schmidt, Department of Bioengineering.) Please note that neither Ingenium nor the Swanson School of Engineering retains any copyright of the original work produced in this issue. However, the Swanson School does retain the right to nonexclusive use in print and electronic formats of all papers as published in Ingenium.

Ingenium 2020

Table of Contents A Message from the Associate Dean for Research............................................................. 4 A Message from the Co-Editors-in-Chief............................................................................ 5 Graduate Student Review Board – Ingenium 2020............................................................. 6 Tumor derived exosomes regulate dendritic cell maturation and activation Rosh Bharthi, Carolina M. Gorgulho, Nils Ludwig, Theresa L. Whiteside, Michael T. Lotze DAMP Laboratory, Department of Pathology, Department of Immunology, Department of Bioengineering................................................................................................ 7 Evaluating carbon reduction strategies for the University of Pittsburgh Eli Brock, Sabrina Nguyen, Katrina Kelly, Robert Kerestes Department of Electrical Engineering.................................................................................... 11 Progress in bioplastics: PLA and PHA Samantha Bunke, Susan Fullerton, Erick Beckman Nanoionics and Electronics Laboratory, Department of Chemical and Petroleum Engineering........................................................................................................ 16 Three-dimensional nickel foam and graphene electrode in microbial fuel cell application: Study of biofilm compatibility Claire P. Chouinard, Felipe Sanhueza Gómez, Natalia Padilla Gálvez, Hermán Valle Zapata, R.V. Mangalaraja, Homero Urrutia Department of Chemical and Petroleum Engineering, Laboratorio de Cerámicos Avanzados y Nanotecnología, Departamento de Ingeniería de Materiales, Unidad de Desarrollo Tecnológico, Laboratorio Biopelículas y Microbiología, Departamento de Microbiología............................... 20

Changes to the maternal sacrum and coccyx during and after pregnancy and delivery Liam Martin, Megan R. Routzong, Ghazaleh Rostaminia, Pamela A. Moalli, Steven D. Abramowitch Translational Biomechanics Laboratory, Department of Bioengineering, Female Pelvic Medicine and Reconstructive Surgery, Division of Urogynecology, Department of Obstetrics, Gynecology, and Reproductive Surgery.......................................... 63 Laser-induced nanocarbon formation for tuning surface properties of commercial polymers Angela J. McComb, Moataz Abdulhafez, Mostafa Bedewy NanoProduct Lab, Department of Industrial Engineering........................................................ 68 The role of oxygen functional groups in graphene oxide modified glassy carbon electrodes for the electrochemical sensing of nicotinamide adenine dinucleotide hydride Ananya Mukherjee, Yan Wang, Leanne Gilbertson Environmental Engineering Laboratory, Department of Civil and Environmental Engineering..... 72 Characterization of hierarchical structures in remelted Ni-Mn-Ga substrates for directed energy deposition manufacturing of single crystals Tyler Paplham, Jakub Toman, Markus Chmielus Department of Mechanical Engineering and Materials Science............................................... 77 Wireless signal transmission through hermetic walls in nuclear reactors Jerry Potts, Yuan Gao, Heng Ban Multiscale Thermophysics Laboratory, Department of Mechanical Engineering and Materials Science............................................................................................................... 81

Extensions and analysis of a virtual balancing task for studying sensory-motor control Michael Clancy, Sudarshan Sekhar, Aaron Batista, Patrick Loughlin Department of Bioengineering.............................................................................................. 24

Genetically engineering ocular probiotics to manipulate ocular immunity and disease Yannis Rigas, Benjamin Treat, Anthony St. Leger Department of Bioengineering, Department of Ophthalmology, Department of Immunology...... 85

Feature validation and online visualization of forearm high-density EMG in an individual with spinal cord injury J. Sebastian Correa, Jordyn E. Ting, Douglas J. Weber Department of Bioengineering, Department of Physical Medicine and Rehabilitation................ 30

Effects of printing parameters on density and mechanical properties of binder jet printed WC-Co Pierangeli Rodríguez De Vecchis, Danielle Brunetta, Katerina Kimes, Drew Elhassid, Markus Chmielus Department of Mechanical Engineering and Materials Science............................................... 88

Tractography reveals patterns of hippocampal innervation in the human temporal lobe Lauren Grice, Chandler Fountain, Michel Modo McGowan Institute for Regenerative Medicine, Department of Radiology, Department of Bioengineering.............................................................................................. 34 Numerically resolved radiation view factors within thermoelectric generators via hybridized CPU-GPU computing Asher J. Hancock, Matthew M. Barry Department of Mechanical Engineering and Materials Science............................................... 38 Characterization of redox flow battery kinetics using a flow channel analytical platform Thomas J. Henry, Tejal V. Sawant, James R. McKone Department of Chemical and Petroleum Engineering............................................................. 42 Metformin administration impairs tendon wound healing Catherine Grace P. Hobayan, Arthur R. McDowell, Feng Li, Jianying Zhang, James H.C. Wang MechanoBiology Laboratory, Department of Orthopedic Surgery, Department of Bioengineering, Department of Physical Medicine and Rehabilitation...................................... 47 Mechanical characterization of silk derived vascular grafts for human arterial implantation Patrick Iyasele, Eoghan Cunnane, Katherine Lorentz, Timothy Chung, Justin Weinbaum, David Vorp Vascular Bioengineering Laboratory, Department of Bioengineering, Department of Cardiothoracic Surgery, Department of Chemical and Petroleum Engineering, McGowan Institute for Regenerative Medicine....................................................................... 52 Robust osteogenesis of mesenchymal stem cells in 3D bioactive hydrogel Eileen Li, Zhong Li, Colin Del Duke, Hang Lin Departments of Bioengineering, Department of Orthopedic Surgery, Center for Cellular and Molecular Engineering......................................................................................................... 57 Manufacturing a polyelectrolyte coating on contact lenses using automated vs. manual techniques for the treatment of dry eye disease Zixie Liang, Alexis Nolfi, Bryan Brown McGowan Institute for Regnerative Medicine......................................................................... 60

Monitoring the in-vitro extracellular matrix remodeling of tissue engineered vascular grafts Hannah Schmidt, Jonathan Vande Geest Department of Bioengineering, McGowan Institute for Regenerative Medicine, Vascular Medicine Institute.................................................................................................. 92 Analytical model validation for melting probe performance using applied computational fluid dynamics Michael Ullman, Michael Durka, Kevin Glunt, Matthew Barry Applied Computational Fluid Dynamics lab, Department of Mechanical Engineering and Materials Science............................................................................................................... 97 Crimped polymer microfibers produced via electrospinning: A review Nikolas J. Vostal Department of Mechanical Engineering and Materials Science............................................. 102 Adventitial extracellular matrix from aneurysmal aorta fails to promote pericyte contractility Kaitlyn Wintruba, Bryant Fisher, Jennifer C. Hill, Tara D. Richards, Marie Billaud, Amadeus Stern, Thomas G. Gleason, Julie A. Philippi Department of Chemical Engineering, Department of Cardiothoracic Surgery, Department of Bioengineering, McGowan Institute for Regenerative Medicine............................................. 106 Biotelemetry: a brief history and future developments in lowering cost Kevin Xu, Mark Gartner Department of Bioengineering............................................................................................ 110 Feasibility study of kinetic, thermoelectric, and RF enery harvesting powered sensor system Keting Zhao, Jiangyin Huang, Hongye Xu, Jingtong Hu Department of Electrical and Computer Engineering............................................................ 114 Index............................................................................................................................... 119

Category Definitions Computational research—using computational techniques to address a scientific question Device design—focusing on the development of a product or device Experimental research—using laboratory methods to achieve a novel overarching experimental aim Methods—developing new techniques and tools for research and design Review—summarizes the current state of knowledge on a particular topic


A Message from the Associate Dean for Research The second declension of the word Ingenium has several meanings – ‘innate or natural quality’, ‘intelligence, natural capacity’, ‘talent, art’, and ‘a man of genius’. In medieval times, Ingenium was held in contrast to the wisdom gained from practice and experience and considered to be a divine stamp on the soul delivered at the instant of one’s birth. Here at the University of Pittsburgh’s Swanson School of Engineering, we believe that our methods are tuned to reveal and cultivate the uniqueness of our students while impressing the habits of uprightness that guide their Ingenium to a common good. On behalf of the Swanson School of Engineering and Dean James R. Martin II, I proudly present the sixth edition of Ingenium: Undergraduate Research at the Swanson School of Engineering, a compilation of articles representing the achievements of our exceptional undergraduate students’ 2019 summer research. With each year and with each edition of Ingenium, we continue to see notable and impressive academic and professional growth and development in our undergraduate students when given opportunities to engage in scientific research. We witness students taking the knowledge, skills, and information that they learn in their coursework and apply it in a meaningful and intentional manner outside of the classroom. These thriving students are our future – of both our highly accredited institution and our world. Each will go on to be engineers, scientists, physicians, and whatever else they set their mind to, and they will, undoubtedly, make significant impacts in the fields of technology, medicine, travel, space, communication, and so much more.

David A. Vorp, PhD

The student authors of the articles contained within this issue of Ingenium studied mostly under the guidance of a faculty mentor in the Swanson School of Engineering. In some cases, the research took place at other institutions and even overseas. At the conclusion of the program, students were asked to submit an abstract summarizing the results of their research, and the abstracts were reviewed by the Ingenium Editorial Board, made up of Swanson School graduate student volunteers. The authors of the highest ranked abstracts were invited to submit full manuscripts for consideration to be included in Ingenium, and those that were submitted were “peer-reviewed” by the Editorial Board. Therefore, Ingenium serves as more than a record of our undergraduate student excellence in research, but also as a practical experience for our undergraduate students in scientific writing and the author’s perspective of the peer-review process. It also provides graduate students with an opportunity to experience the editorial review process and the reviewer’s perspective of the peer-review process. I would like to acknowledge the hard work and dedication of the Co-Editors-In-Chief of this issue of Ingenium, Monica Liu and Jianan Jian, as well as the production assistance of Marygrace Reder, Reiko Becker, and Jaime Turek. This issue also would not have been possible without the hard work by the graduate student volunteers who compromised the Ingenium Editorial Board and who are listed by name in this issue. It is also altogether fitting to thank the faculty mentors and other co-authors of each of the reports included in this issue. On behalf of the entire Swanson School of Engineering and Dean James R. Martin II, I hope that you enjoy reading this sixth edition of Ingenium and that the many talents of our students inspire the engineers of the future. Hail to Pitt!

David A. Vorp, PhD Associate Dean for Research Swanson School of Engineering University of Pittsburgh

4 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

A Message from the Co-Editors-in-Chief

Monica Liu

Jianan Jian

Greetings! We are excited to share with you the sixth edition of Ingenium: Undergraduate Research at the Swanson School of Engineering. The 26 articles featured in this year’s edition of Ingenium showcase the diversity and depth of research projects conducted by undergraduate students and their mentors in the Swanson School of Engineering (SSoE). This year, we aspired to include a broader swath of scientific contributions by inviting more undergraduate researchers to contribute review articles to Ingenium and by soliciting manuscripts from students involved in research outside of the summer research program at the SSoE. As a peer-reviewed publication, all manuscripts submitted to Ingenium are reviewed through a twostep, single-blind process. Graduate students across all departments of the SSoE volunteered their time to serve on the editorial board of Ingenium by reviewing abstracts and manuscripts. Student authors were asked to provide a point-by-point response to reviewer comments to mirror the peer-review process used by many journals. Based on this process, we categorized articles into one of five groups: Experimental Research, Computational Research, Device Design, Methods, and Review. Before each article, you can read bios of the authors and their faculty mentors to get to know the individuals behind the work. We have also included the students’ insights on how their work contributes to their field of study in the Significance Statement at the start of each article. We hope that with this information, you will see the dedication and hard work undergraduate students in the SSoE have put into addressing challenges in their field of study. Ingenium would not have been possible without the vision and guidance of the Associate Dean for Research, Dr. David Vorp. Further, we would especially like to express our gratitude for Jaime Turek, who offered invaluable support and assistance throughout the process. We would also like to thank our graduate student editorial board for their thoughtful and detailed comments to students. Finally, we are thankful for Reiko Becker, Marygrace Reder, and the entire team at the Office of University Communications who worked with us to produce and design this issue of Ingenium. As Co-Editors-in-Chief, the process of helping this year’s edition of Ingenium come to fruition has been rewarding and exciting. We are impressed by students’ scientific maturity, professionalism, and insight. We are further grateful to the faculty mentors for providing students with research opportunities and guidance along the way. We hope that as you read through this year’s edition of Ingenium, you will find it every bit as rewarding as we do.

Monica Liu Jianan Jian Co-Editor-in-Chief Co-Editor-in-Chief


Graduate Student Review Board – Ingenium 2020 Name Department Abhijeet, Gujrati................................................................ Mechanical Engineering and Materials Science Akinbade, Yusuf Akintayo................................................... Civil and Environmental Engineering Allen, Abigail..................................................................... Bioengineering Alrefaie, Abdulaziz............................................................. Mechanical Engineering and Materials Science Alshayeb, Suhaib.............................................................. Civil and Environmental Engineering Astbury, Matthew.............................................................. Industrial Engineering Baker, Andrew.................................................................. Mechanical Engineering and Materials Science Balmuri, Sricharani............................................................ Chemical and Petroleum Engineering Behrangzade, Ali............................................................... Bioengineering Bouzid, Zeineb.................................................................. Electrical and Computer Engineering Brantly, Nathan................................................................. Bioengineering Buettner, Nathanial............................................................ Civil and Environmental Engineering Cardoza, Alvaro................................................................. Electrical and Computer Engineering Coyle, Alex........................................................................ Civil and Environmental Engineering Dhamotharan, Vishaal....................................................... Bioengineering El-Hajj, Bashear................................................................ Civil and Environmental Engineering Geiger, Jason.................................................................... Civil and Environmental Engineering Gopalan Ramachandran, Rahul.......................................... Mechanical Engineering and Materials Science Greco, Ande...................................................................... Bioengineering Gregg, Robert................................................................... Chemical and Petroleum Engineering Grigsby, Erinn................................................................... Bioengineering Hashemi, Amirreza............................................................ Computational Modeling and Simulation Herrera, Angelica.............................................................. Bioengineering Hughes, Christopher.......................................................... Bioengineering Jalali Najafabadi, Hoda...................................................... Civil and Environmental Engineering Jantz, Maria...................................................................... Bioengineering Jian, Jianan*.................................................................... Electrical and Computer Engineering Johnson, Camille.............................................................. Bioengineering Karapetyan, Vahagn.......................................................... Bioengineering Kovalchuk, Matthew.......................................................... Mechanical Engineering and Materials Science Lee, Yoojin........................................................................ Bioengineering Li, Haoran......................................................................... Civil and Environmental Engineering Liu, Monica*..................................................................... Bioengineering Maldonado, Alex............................................................... Chemical and Petroleum Engineering McClain, Nicole................................................................. Bioengineering Mian, Sami....................................................................... Electrical and Computer Engineering Misra, Ritesh.................................................................... Electrical and Computer Engineering Miu, Evan......................................................................... Chemical and Petroleum Engineering Mutha, Pushkar................................................................ Electrical and Computer Engineering Nguyen, Preston............................................................... Mechanical Engineering and Materials Science Omecinski, Katelin............................................................ Bioengineering Patil, Rituja....................................................................... Chemical and Petroleum Engineering Pawar, Ritesh.................................................................... Chemical and Petroleum Engineering Pliner, Erika Mae............................................................... Bioengineering Pressly, Michelle............................................................... Chemical and Petroleum Engineering Rogers, Chase.................................................................. Civil and Environmental Engineering Routzong, Megan.............................................................. Bioengineering Shapiro, Monica................................................................ Chemical and Petroleum Engineering Sivakumar, Sruthi.............................................................. Bioengineering Ting, Jordyn...................................................................... Bioengineering Vishnubhotla, Sai.............................................................. Mechanical Engineering and Materials Science Yang, Timothy................................................................... Mechanical Engineering and Materials Science Zhou, Ziyi......................................................................... Mechanical Engineering and Materials Science *Co-Editors-in-Chief: Monica Liu and Jianan Jian

6 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

Tumor derived exosomes regulate dendritic cell maturation and activation Rosh Bharthia, Carolina M. Gorgulhoa, Nils Ludwigb,c, Theresa L. Whitesidee, Michael T. Lotzea, c-f DAMP Laboratory, UPMC Hillman Cancer Center, Pittsburgh, PA, bDepartment of Pathology, cUPMC Hillman Cancer Center, Pittsburgh, PA, dDepartment of Immunology, e Department of Pathology, fDepartment of Bioengineering, University of Pittsburgh, Pittsburgh, PA a

Rosh Bharthi

Rosh Bharthi is a bioengineering undergraduate at the Swanson School of Engineering. His current research interests include exosomes and how they play a role in tumor immunology. After graduating, he plans on attending medical school while still continuing his experience in bioengineering and research.

Michael T. Lotze, MD, is Professor of Surgery, Immunology, and Bioengineering; Vice Chair of Research within the Department of Surgery; Director of the DAMP Laboratories at the UPMC Hillman Cancer Center. His research work includes modern immunotherapy and gene therapy, Michael T. Lotze dendritic cell and cytokine therapies, and investigation of the role of mitochondria, metabolism, and unscheduled cell death in cancer. Dr. Lotze is the co-inventor of 10 patents in dendritic cell vaccines and antigen discovery and serves as associate editor of the Journal of Immunotherapy; he is also an award-winning NCI-trained scientist (1978-1990), the inaugural Director of Surgical Oncology at Pitt (1990-2000), former Vice President of Research at GlaxoSmithKline (2001), founding director of the UPCI Academy, former Chief Scientific Officer at Lion/Iovance Biotherapeutics and innovative educator as a prolific scientist/ tumor immunologist with over 500 publications and several books.

Significance Statement

It was initially unknown whether HMGB1-containing exosomes could trigger a dendritic cell response, potentially causing chronic inflammation. Understanding this would explain another mechanism for how cancer and chronic inflammation are related. From this study, exosomes do not contain HMGB1 nor do they stimulate dendritic cells. While this disproves the idea of tumor-derived exosomes causing chronic inflammation, the fact that tumor-derived exosomes did not activate dendritic cells could imply that exosomes are a potential vehicle for delivering antigens without stressing the immune system.

Category: Experimental research

Keywords: Tumor-derived exosomes, HMGB1, dendritic cells


Exosomes are nanovesicles ranging from around 50nm150nm that can carry various types of molecules, such as signaling molecules, lipids, and nucleic acids. In the context of cancer, exosomes are responsible for modulating the tumor microenvironment and signaling nearby cells to allow tumor development. HMGB1 is a damage-associated molecular pattern (DAMP) molecule known to be secreted by tumor cells and promote chronic inflammation when continually present in the body; some studies have also shown that HMGB1 is contained in exosomes. It has not been shown whether HMGB1 contained in exosomes can be responsible for causing dendritic cells to become activated, and thus starting the inflammation pathway. Exosomes were isolated from cell culture supernatants and characterized. Afterwards, Western blots were run to examine protein cargo, including but not limited to HMGB1, and flow analysis was used to determine whether exosomes from tumors might activate or suppress dendritic cells. HMGB1 was not present in exosomes, but RAGE and SQSTM-1/p62 was found to be present. Furthermore, exosomes inhibit dendritic cells, but not to the same magnitude that their corresponding tumor cells do.

1. Introduction

Exosomes are nanovesicles secreted by cells. They are formed by the cell membrane involuting and forming a vesicle inside the cytoplasm. This vesicle is then involuted once more, forming a multivesicular body (MVB), which contains many vesicles within it [1]. The MVB then fuses with the cell membrane, releasing the numerous exosome vesicles within. This pathway is mainly to package ubiquitinated proteins in exosomes and release them from the cell [2]. While exosomes were initially considered as a waste secretion system, research has shown that exosomes can carry a wide variety of cargo, such as lipids, proteins, mRNA, and miRNA which can then be taken up by nearby cells, thus acting as a signaling system [2]. The implications of this signaling system could be generalized to various physiological phenomena, such as angiogenesis. Exosomes play an important role in the context of the immune system and tumor immunology. Research has shown that exosomes secreted by antigen-presenting cells (APCs) are able to activate immune responses based on MHC-peptide complexes found on the surface of exosomes [3]. Additionally, Regulatory T cells (Tregs) release exosomes that suppress production of T cells [4]. Tumor-derived exosomes (TEX) can both stimulate and inhibit an immune response. These exosomes carry inhibitory proteins typically released by the tumor cells which can prevent activation of T cells [5]. On the other hand, TEX contain factors capable of stimulating APCs [5]. From these examples, it is clear that exosomal cargo plays a role in modulating the immune response, and understanding how this signaling works or could be utilized is crucial. It is known that HMGB1, a notable example of a damageassociated molecular patterns or DAMP, released from tumor cells can lead to chronic inflammation and that chronic inflammation 7

can promote tumors to further secrete HMGB1, creating a positive cycle [6]. Additionally, it is also known that TEX can play a role in tumor development and signaling; for instance, a study has shown that TEX from head and neck squamous cell carcinoma signal surrounding endothelial cells to promote angiogenesis [7]. This study sought to bridge these two ideas and determine whether TEX might be another form of delivering HMGB1 to surrounding cells. Doing so would show how exosome signaling plays a role in the development of chronic inflammation and cancer. In addition to finding what type of cargo is carried in TEX, the study also shows whether TEX activate dendritic cells (DCs). If TEX were to contain HMGB1 and activate DCs, it would further show evidence of a pathway for HMGB1 from exosomes to DCs. Lastly, in order to see if apoptotic deficiency leads to differences in exosome cargo or secretion, wild type HCT 116 human colorectal tumor cells were compared to a p53 KO, PUMA KO, and BAX KO (inhibiting apoptosis promotes autophagy, a degradation pathway that allowing molecules in the cell to be broken down into constituents and repurposed for other molecules; we believed that enhanced autophagy may thus reduce the cargo released in exosomes, hence why this was one of our goals).

2. Methods 2.1 Exosome Isolation HCT 116 human colorectal cancer cells were used for these experiments; in particular, wild type, p53 knockout, BAX knockout, and PUMA knockout cell types were used. 6 million cells were seeded in T125 flasks, with two flasks per HCT 116 cell variant. 25 mL RPMI cell media (Corning) with 10% exosome-depleted fetal bovine serum (FBS, Gibco) per flask was used as media; to make exosome-depleted FBS, 1x FBS was ultracentrifuged at 28,000 rpm to pellet exosomes, and the supernatant was used. After 72h, culture supernatant was collected and centrifuged for 2000xg for 10min at room temperature to pellet large debris. Supernatant was centrifuged again at 10,000xg for 30min at 4°C and filtered with .22μm filter to remove larger vesicles remaining. Supernatant was concentrated to 1mL using Vivacell 20 concentrators spun at 2000xg. Mini-SEC (size exclusion chromatography) columns were made by adding 10mL sepharose (GE Healthcare Bio-Sciences) to columns. An upper filter was added to the column, and the column was flushed with 10mL phosphate-buffered saline (PBS). Afterwards, the concentrated supernatant was added to the column. The fourth fraction was collected, using PBS to elute the fractions. Exosome characterization was performed by transmission electron microscopy (TEM) imaging, protein estimation via bicinchoninic acid assay (BCA), and Western blot testing for TSG101. 2.2 Western blot Using BCA to determine protein concentration, 10μg protein was loaded into each well. Bio-Rad TGX pre-made gels were used. TEX samples were concentrated using 0.5 mL 100 K Amicon Ultra centrifugal filters to reduce the sample volume needed to load 10μg protein. Cell lysates were used as a comparison for the TEX

8 Undergraduate Research at the Swanson School of Engineering

samples. TSG101 was used as a control molecule to test, since most exosomes contain this molecule. We tested for HMGB1, along with RAGE, LC3, PD-L1, and SQSTM-1/p62, which are other common molecules involved in tumor and immune system biology, to see if their presence could provide any insight on what role these exosomes may play in signaling. 2.3 TEX-Dendritic cell co-incubation Dendritic cells were isolated from buffy coats. Ficoll gradient centrifugation separated blood into plasma, RBCs, and PBMCs. PBMCs were extracted and washed 3 times with RPMI to remove platelets. RBC lysis buffer (Sigma Aldrich) was added during sequential washing to lyse any remaining RBCs. PBMCs were plated in 6 well plates in AIM V media (Gibco). After 2hr, nonadherent lymphocytes were removed, and RPMI with 10% FBS was added to the adherent monocytes along with 50ng/mL GM-CSF and IL-4 to differentiate monocytes into immature dendritic cells after 5 days. 12μg TEX protein was added to each well on either the fifth day (48hr incubation) or sixth day (24 hr incubation). On the seventh day of monocyte culture, the cells were stained with antibodies for flow analysis. Negative controls received no antigen, and positive controls received 2μg LPS. DCs co-incubated with tumor cells served as comparison for the corresponding DC-TEX co-incubation. 2.4 Flow cytometry Raw flow cytometry data was gated based on granularity, Zombie Live/Dead dye, and CD11c, a marker expressed by dendritic cells. Proportion of cells expressing marker of interest was obtained; markers of interested included HLA-DR, CD80, CD86, CD40, CD14, CCR7, CD25, CD274, and CD83. Data was plotted as mean±SD; ANOVA with Tukey post-test was performed on data.

3. Results 3.1 Exosome Characterization

Figure 1: TEM Images – Transmission electron microscopy of isolated exosome samples from each cell line. Exosomes are characterized by round vesicles ranging from around 50nm-150nm; they contain a dark, distinct outer border and may appear folded inwards, like a “crushed ping-pong ball” (the center cluster in the p53 KO image is a good example of this).

Exosomes were present in all samples; they are identified by circular structures ranging from 50-150nm with distinct darker borders. BCA protein estimation showed that protein was present in extracted exosome samples, around 100μg protein for each sample. Furthermore, Western blot confirmed the presence of TSG101 in all TEX samples.

Ingenium 2020

3.2 Western Blot

Figure 2: Western Blot – Membrane images of TSG101, HMGB1, RAGE, and SQSTM1/p62. Presence of SQSTM1/p62 was only tested for WT and p53 KO, but was not yet tested for PUMA KO and BAX KO. Blots were used to indicate the presence or absence of the molecules tested.

TSG101, the control marker, was shown consistently for all Western blots, confirming that the exosome isolation technique worked. When testing for HMGB1, PD-L1, and LC3, these markers were only present in cell lysates but not exosomal protein. However, the presence of RAGE, a receptor for HMGB1, was found in exosome samples, as well as SQSTM-1/p62 in WT and p53 KO exosomes. 3.3 Flow cytometry

Figure 3: Flow cytometry data – Data indicates the percent population containing DC markers. Data gated by granularity, Zombie L/D, and CD11c. “WT,” “p53,” “PUMA,” and “BAX” refer to DCs co-incubated with whole tumor cells from the respective cell line. Data is presented as mean±SD. ANOVA was run with Tukey post-test. Orange line shows mean of control group. Tumor cells appear to drastically inhibit expression of these markers compared to negative control. TEX from all samples also inhibit but not with the same magnitude as the tumor cells. Other markers tested did not yield meaningful, significant data.

The positive and negative controls showed a small difference in cell proportions for all markers tested. There was no significant difference between groups. However, the means for TEX-DC coincubation were higher than the means for tumor-DC co-incubation for all markers regardless of the cell type.

4. Discussion

The lack of HMGB1 in TEX shows that exosome release is not a pathway HMGB1 takes to signal other cells. This contradicts the results of a study showing that HMGB1 is carried in gastric cancer TEX [8]. However, this and other studies showing the presence of HMGB1 in TEX have used ultracentrifugation to pellet out TEX; this technique is not a reliable method of extracting exosomes due to the potential contamination of soluble protein [9]. This method and the use of exosome isolation kits have this risk, and the soluble form of HMGB1 released by the cells may have produced a false positive for these studies. Use of size-exclusion chromatography has been shown to be the optimal exclusion technique, which elutes a fraction enriched in exosomes with limited soluble protein [9]. According to the flow data, TEX co-incubation inhibited dendritic cell markers compared to the negative control but less so than their respective tumor cells. The data shows that TEX prevent differentiation of monocytes to DCs (indicated by lowered %CD14 expression, compared to negative control), increase DCs’ ability to bind IL-2 (indicated by increased %CD25 expression, compared to negative control), and prevent T cell communication (indicated by decreased %CD80 and %CD86, compared to negative control). However, there was a large variance in the data, and in order to reduce this variance, a larger sample size will be needed. The amount of treated protein for all groups might need to be increased to show a distinct difference between the positive and negative controls and among the four cell lines. If this pattern were to still hold true, this could imply that tumor-derived exosomes do not activate an immune response; if a therapeutic aimed to deliver a certain antigen to dendritic cells, exosomes may provide a safe way of doing so without initiating an adverse immune response. When using protein estimation to confirm the presence of exosomes and as preparation for the Western blot, the amount of exosomal protein extracted from each cell group (WT, p53 KO, PUMA KO, BAX KO) was similar across groups. This could indicate that there is no connection between apoptotic deficiency and exosome production. However, using a nanoparticle counter to directly count the number of exosomes extracted rather than use a protein estimation would provide direct results as to whether the exosome count differs across groups, making this a future direction to pursue. The presence of SQSTM-1/p62 is interesting in that it is involved in autophagy, a process in which large molecules are broken down in order for monomers to be recycled for cell processes. SQSTM-1/p62 is known to bind ubiquitinated protein [10]. Additionally, proteins that package exosomal cargo bind to ubiquitinated protein to release them from the cell. Because both pathways tend to target proteins marked with ubiquitin, the presence of SQSTM-1/p62 may show a possible relationship between autophagy and exosome release. The purpose of RAGE is to act as an embedded membrane receptor expressed on immune cells which causes the cells to release inflammatory cytokines when RAGE activated, though the purpose may change for other cell types [11]. However, soluble 9

RAGE (sRAGE) may play a role in inhibiting the inflammatory response; a study has shown that sRAGE acts as a decoy receptor in the context of AT1R binding to RAGE (AT1R activated by RAGE is responsible for downstream effects leading to inflammation, oxidative stress, and other effects) [12]. This same idea of sRAGE being a decoy receptor may also occur with TEX; TEX containing sRAGE may be taken up by DCs, and the sRAGE within exosomes may bind to DAMPs (damage associated molecular patterns) absorbed by DCs, which prevents RAGE from binding to DAMPs and thus prevents inflammation from occurring. However, it is unknown whether the RAGE in exosomes are soluble or membrane-bound, making this a possible future direction.

5. Conclusions

SQSTM-1/p62 and RAGE were present in TEX derived from HCT 116 human colorectal cancer cells but HMGB1, PD-L1, and LC3 were not observed. The presence or absence of these molecules provides some insight on pathways used for packaging exosome content. Functional experiments will need to be run to see whether SQSTM-1/p62 and RAGE serve a meaningful purpose in signaling. It will also be worthwhile to test other tumor cell lines or even patient samples to see whether the presence and absence of molecules from the HCT 116 cell line can be generalized to other tumor sources. TEX are able to suppress DCs but not to the same degree as tumor cells, and we know from the Western blot that this is not due to HMGB1. This is shown by the increase in percent expression of CD14, CD25, CD80, and CD86 from tumor-dendritic cell coincubation to TEX-dendritic cell co-incubation. Tumors suppress the presentation of these molecules when compared to the negative control; TEX show a weaker suppression since most of the average values still remain lower than the control group. Other markers tested did not yield meaningful, significant data. This shows that TEX may be involved in modulating cells of the immune system. Future experiments could involve isolating TEX from other tumor sources or from tumors grown in different culture conditions to see if the produced TEX show similar patterns in modulating dendritic cells. Co-cultures of TEX with other antigen-presenting cells may also be worth considering.

6. Acknowledgements

Funding was provided by University of Pittsburgh Swanson School of Engineering, the DAMP Laboratory at UPMC Hillman Cancer Center, and the Office of the Provost at the University of Pittsburgh. Microscopy resources were provided by University of Pittsburgh Center for Biologic Imaging, funded as Core within the P30 of the Hillman Cancer Center of UPMC.

10 Undergraduate Research at the Swanson School of Engineering

7. References

[1] R. Kalluri, The biology and function of exosomes in cancer, Journal of Clinical Investigation. 126 (2016). 1208-1215. [2] J.R. Edgar, Q&A: what are exosomes, exactly?, BMC Biology. 14 (2016). 1-7. [3] D.W. Greening et al, Exosomes and their roles in immune regulation and cancer, Seminars in Cell & Developmental Biology. 40 (2015). 72-81. [4] Q Li et al, Exosomes: versatile nano mediators of immune regulation, Cancers. 11 (2019). 1-21. [5] T.L. Whiteside, The effect of tumor-derived exosomes on immune regulation and cancer immunotherapy, Future Oncology. 13 (2017). 2583-2592. [6] C.M. Gorgulho et al, Johnny on the spot- chronic inflammation is driven by HMGB1, Frontiers in Immunology. 10 (2019). 1-18. [7] N. Ludwig et al, Exosomes from HNSCC promote angiogenesis through reprogramming of endothelial cells, Mol Cancer Res. 16 (2018). 1798-1808. [8] X. Zhang et al, Tumor-derived exosomes induce N2 polarization of neutrophils to promote gastric cancer cell migration, Molecular Cancer. 17 (2018). 1-16. [9] N. Ludwig et al, Optimization of cell culture conditions for exosome isolation using mini-size exclusion chromatography (miniSEC), Experimental Cell Research. 378 (2019). 149-157. [10] S. Pankiv et al, p62/SQSTM1 bids directly to Atg8/LC3 to facilitate degradation of ubiquitinated protein aggregates by autophagy, J. Biological Chemistry. 282 (2007). 24131-24145. [11] K. Kierdorf, G. Fritz, RAGE regulation and signaling in inflammation and beyond, J. Leukocyte Biology. 94 (2013). 55-68. [12] J. Jeong et al, Soluble RAGE attenuates AngII-induced endothelial hyperpermeability by disrupting HMGB1-mediated crosstalk between AT1R and RAGE, Experimental and Molecular Medicine. 51 (2019). 1-15.

Ingenium 2020

Evaluating carbon reduction strategies for the University of Pittsburgh Eli Brocka, Sabrina Nguyena, Dr. Katrina Kelly-Pitoub, Dr. Robert Kerestesa Electrical Engineering, University of Pittsburgh, Pittsburgh, PA Dr. Katrina Kelly-Pitou was part of the Center for Energy at the University of Pittsburgh when the project kicked off, but now works for SmithGroup in Pittsburgh, PA a


Eli Brock is a sophomore electrical engineering student. He is interested in power systems, mathematical modeling, and sustainable energy. He plans to pursue a PhD after graduating.

Eli Brock

Sabrina Nguyen is a senior electrical engineering student planned to graduate in the spring. Her interests lie in electric distribution systems, renewable energy, and sustainable energy practices. She plans to pursue a master’s degree in electric power systems and sustainability following her undergraduate career.

Sabrina Nguyen

Dr. Katrina Kelly-Pitou

Katrina Kelly-Pitou is an expert in energy systems and urban development. Currently, she acts as an Urban Systems Strategist at SmithGroup in Pittsburgh.Pulling together architects, landscape architects, engineers, and behavioral scientists, the Urban System’s group focuses on bridging the gaps between the technical and social development needed to future-proof cities across the globe. Katrina leads the interdisciplinary energy designs that are needed to transition society towards a zero-carbon future. In addition to her postdoctoral research in Electrical and Computer Engineering at the University of Pittsburgh, Katrina holds a PhD from the University of Nottingham in Energy Development Studies, a Masters in International Relations from Hult International Business School, and a BA/BSc from Duquesne University. She is a regular contributing to IEEE PES, IEEE Smart Cities, Energy Policy, and the European Economic Review.

Robert Kerestes received his Ph.D. from the Department of Electrical and Computer Engineering at the University of Pittsburgh in 2014. Prior to that he also received his M.S. (2012), and his B.S. (2010) all with a concentration in electric power systems at the University of Pittsburgh. Prior to his education, Robert served in the United States Navy as an interior communications electrician onboard the USS Constellation from 1998-2002 and then as a construction electrician in the Seabees Naval Construction Battalion from 2002-2006. Following his education, Robert worked at Emerson Automation Solutions (formerly known as Emerson Process Management) where he led the development and implementation of mathematical models for electric power applications related to embedded simulation. In 2016, Robert Robert Kerestes joined the University of Pittsburgh’s ECE Department where he focuses on the advancement of pedagogy and assessment in the field of electrical engineering, particularly in the areas of electric power distribution systems, distributed energy resources, and electromagnetism. Robert’s research interests are in the development of the next generation of electric power distribution systems and the integration of renewable energy resources into the distribution sector. Robert took on the role of Director of the Electrical Engineering Undergraduate Program in the Spring of 2017 and currently serves in that role.

Significance Statement

For an urban, cloudy campus like the University of Pittsburgh, unique factors come into play when deciding between carbon reduction strategies. For this university, this study finds on-campus, university-owned solar is a less effective initiative than green power purchase agreements.

Category: Computational research

Keywords: urban, campus, electric generation, efficiency, solar, decarbonization



The University of Pittsburgh has set goals to transition to renewable/low-carbon energy by 2030. As a large proportion of the University’s carbon emissions come from its purchased electricity, the purpose of this study is to explore methods Pitt could implement in order to reduce the emissions associated with its electricity consumption. This study compares three main strategies for decarbonization: on-campus solar installations, power purchase agreements, and bulk purchases of renewable energy credits. The project expands on data collected by other researchers, specifically data regarding the location, placement, and potential output of solar panels on campus. This data served as the foundation for a costand-emission dual-optimized model, created by using Python to derive output efficiencies for each building and prioritize more efficient rooftops. Visualization with ArcMap and MATLAB facilitated comparison between the model and its alternatives. After assessing relative costs and emissions reductions, the analysis concludes that entering PPAs would be more financially viable than investing in solar power.

1. Introduction

In the University of Pittsburgh’s current Clean Energy Plan, there are two overarching goals. The first mandates that 50% of Pitt’s electricity be renewably sourced by 2030. The second calls for a 50% reduction in greenhouse gas emissions by 2030 from the 2008 baseline [1]. This analysis will focus on the latter goal because it more comprehensively addresses the challenge of carbon reduction. Pitt is an urban school located in one of the cloudiest cities in the US. Lack of sunlight reduces the economic viability of solar panels; furthermore, urban areas have less room for solar panels and more shade from taller buildings. This university is not the only institution whose conditions are unfavorable for solar installation; however, with recent advances in photovoltaics, it seems possible that even installations in cloudy locations may offer cost-effective alternatives to traditional grid-paying fees. Therefore, a successful analysis of the production and cost impacts at Pitt can help to serve as a model for other universities in similarly cloudy locations. The University, like any climate-conscious economic actor, must weigh its benevolence against financial constraints and responsibilities. Therefore, the fundamental challenge of carbon reduction, and the purpose of this study, is to evaluate the cost efficiency of different strategies. Specifically, the decarbonization potential of the University’s purchasing power will be weighed against the prospects for on-site, university-owned solar generation. Producing energy on-site with university-owned assets offsets the cost of purchasing electricity but requires investment by the University. Conversely, purchasing green electricity simply requires restructuring electricity contracts, but Pitt must take care to ensure its actions are leading to emission reductions.

2. Methods

This study analyzes three alternatives for decarbonization of Pitt’s electricity consumption, which accounts for approximately half of its greenhouse gas emissions [2]. A HelioScope simulation conducted by university researchers in the summer of 2018 provides the foundation for analysis of the first alternative, onsite solar generation. Alternatively, the University could alter its purchasing practices to decarbonize its electricity consumption. While this approach could take several forms, this study will focus on two of the most common: entering into power purchase agreements (PPAs) with green power generation projects and purchasing voluntary renewable energy credits (RECs). A working model of the decarbonization potential and cost of solar panels will be juxtaposed with the price estimates for PPAs and RECs. 2.1 Working Solar Model On-site solar generation, as a decarbonization strategy, can be simply but comprehensively modeled with a single metric: the annual reduction in carbon emissions associated with each dollar invested, in kilograms per dollar per year. As this is a marginal metric, it is given by a derivative: (1) where ME is marginal emissions avoided in E is total emissions avoided in and P is installation cost in $. To go about finding E(P), the researchers began with the results of the HelioScope simulation. HelioScope is a solar panel simulation software which outputs data such as the panel capacity of a roof and its yearly energy output (in kWh), among other data. Due to location, shade, and other factors, some buildings are considerably more efficient than others (in other words, more energy production per installed watt). Assuming the University wants to optimize its investment, it should install panels on the most efficient buildings first. 2.2 Assumptions Important assumptions were made in evaluating the efficiency of the buildings. In the University’s simulation, each building was simulated with one of three panels, with wattage ratings of either 320W, 310W, and 280W. Note that wattage ratings do not reflect actual output. As commercial solar installations are priced per installed watt, it is assumed that the wattage capacity per building is consistent; in other words, the number of installed watts is independent of which panel is used. This assumption is important because the University would likely purchase only a single panel model for all buildings. It is also assumed that all installed watts on the same rooftop are equally efficient. 2.3 Annual Energy Production of Each Installed Watt With these assumptions, the data yielded a natural metric for the efficiency of a rooftop: YW, the marginal increase in energy output associated with each installed watt. YW is unique and constant for each building. YW for a given building is given by:

12 Undergraduate Research at the Swanson School of Engineering


Ingenium 2020

where YW is the energy output per installed watt in YR is the energy output of the roof in kWh, WP is the respective wattage rating of the simulated panel, and N is the panel capacity of the roof. 2.4 Annual Energy Production as a Function of Watts Installed To find total annual energy production as a function of watts installed Y(NW), the values of YW were added for each consecutive watt, starting with the most efficient, until NW watts were reached. The value of YW steps down each time NW exceeds the wattage capacity of the next most efficient building. 2.5 Total Emission Reductions as a Function of Investment E(P) was derived by scaling Y(NW). The installation price is assumed to be $2/watt, per conversations with representatives of installation firms. The University’s emission factor (the mass of CO2 emitted per unit of electricity consumed) is 0.487546 kg/kWh as of 2017 [2]. Therefore,


2.6 Marginal Emissions Reductions as a Function of Investment ME(P), found by differentiating E(P), served as the model of on-site solar to be compared with other alternatives. As ME(P) is an annual figure, it was assumed for simplicity that the emissions factor will be constant through the foreseeable future; in reality, it is likely to decrease. 2.7 Working Market Price of Renewable Power-Purchase Agreements An analysis of PPAs, specifically their prices, is difficult as their details tend to be confidential between the buyer and seller. A PPA is an agreement between an energy consumer, such as the University, and an energy generation project, usually one which has not been built yet. The consumer agrees to buy energy from the project for an agreed-upon price for an extended period. This price tends to be lower than the market price for electricity, thereby attracting the consumer, and the guaranteed demand for electricity draws in the project. These projects also tend to be renewable, usually wind and solar, as these sources can often offer electricity at a lower price and climate-conscious consumers create demand for them. Given the difficulty of finding PPA price data, the researchers turned to the relatively transparent LevelTen PPA marketplace, which operates in areas served by five major independent system operators. Their 2019 Q1 PPA Price Index reports the 50th percentile price for wind and solar PPAs in these areas to be $29.50/MWh [3] (about 3 cents per kWh, compared with the University’s market price of 8 cents). It was assumed that LevelTen was approximately representative of the PPA market and this price was used for comparison with the other alternatives. While a rough estimate, it is consistent with other approximate PPA price indices, including the National Renewable Energy Laboratory’s Q1 2019 Solar Industry Update, which puts US solar PPA prices between 2 and 3 cents [4].

2.8 Working Market Price of Voluntary Renewable Energy Credits The final alternative, RECs, are the United States’ mechanism for establishing a green electricity market. One REC is created for each MWh of renewable electricity generated. The REC can then be bought and resold by utilities and consumers, and whoever “retires” the REC can claim to have used 1 MWh of renewable power. Many RECs are bundled with the electricity itself. Bundled RECs are transferred in PPAs and are also purchased and retired by utilities to ensure compliance with state Renewable Portfolio Standards (RPS), with the RPS/REC structure behaving as a kind of renewable subsidy. However, some RECs, often produced in states without RPSs, are known as “unbundled” or “voluntary” RECs. The price for voluntary RECs fluctuates but tends to be between $0.50 and $1.00 [5]. As a conservative estimate, $1.00 will be the price of unbundled RECs used for this analysis.

3. Results

Figure 1 is a map of approximately 70 buildings on the University of Pittsburgh’s Oakland campus, symbolized by economic efficiency (annual kWh production/$). This metric was derived by scaling E(P) by the inverse of the emission factor. As seen, there is a 28% difference between the most and least efficient buildings, a testament to the value of the building efficiency metric for decision-making.

Figure 1: Pitt Buildings by potential solar efficiency


Figure 2 shows the annual carbon reduction per dollar spent as a function of installation cost, or ME(P) as derived in the “Methods” section. As would be expected, the marginal carbon reductions decrease as panels are installed on less efficient rooftops.

factor. The same $1, invested in solar panels, would create only 0.67 kg of carbon reductions each year, assuming it was at the optimized end of the solar model (the far left of Figure 2). Even with this generous assumption, after the 25-year panel lifetime, that is only 16.75 kg of emissions reductions—29 times less than that of the unbundled REC. However, most experts agree that unbundled RECs produce little to no additionality, meaning their purchase is unlikely to lead to new renewable generation capacity (this also explains their comparatively low price) [6]. As a result, while they may allow the University to claim environmental progress, their effectiveness as a decarbonization strategy is practically void.

4. Discussion

Figure 2: Marginal carbon reductions as a function of installation cost

University-owned solar, unlike other alternatives, represents an economic investment: for every unit of energy produced on-site, that unit does not need to be purchased from Duquesne Light Co., the utility servicing the University. As a result, over a certain time period, the panels will pay for themselves and save the University money. In Figure 2, the annual carbon reduction for any installation scale is given by the area under the curve up to the cost of installation. Dividing annual carbon reduction by the emissions factor and multiplying by the 8.1₵/kWh electricity price yields the annual savings by the university. Finally, dividing the installation cost by annual savings gives the payback period. This process revealed that payback periods vary from 18 years to 23 years, with longer paybacks associated with larger-scale installation. Furthermore, even for a full-campus installation project, the annual energy production from on-site solar panels would max out at 12.5 million kWh, less than 6% of the University’s annual energy consumption [3]. Like on-site solar, power purchase agreements save the University money on a per-kWh basis; however, they do so without requiring an initial investment. The LevelTen price index, as established, is around 3₵/kWh and the University’s electricity costs are around 8₵/kWh, creating 5₵ of savings per kWh. By contrast, university-owned solar panels offset the entire cost—8₵—of purchased electricity. Accordingly, the monetary return per kWh associated with an average green PPA is around 60% of that associated with on-campus solar panels. At the established price of $1, unbundled RECs would allow the University to claim 1 MWh of renewable energy, equivalent to 488 kg of emissions reductions per the 0.488 kg/kWh emissions

14 Undergraduate Research at the Swanson School of Engineering

The most striking findings of the study are the two-decade payback period for distributed solar and its relatively small impact, with a maximum of 6% of current demand being met. PPAs, on the other hand, involve no initial investment and facilitate decarbonization on a much larger scale. In fact, the University will be getting 25% of its electricity from a PPA with a local hydropower plant starting in 2023. Considering these advantages, the lower yearly return is a worthwhile trade-off for PPAs. Unbundled RECs, on the other hand, are an extremely cheap strategy to meet the University’s 50% renewable goal. However, in the context of the University’s second goal, which prioritizes carbon reduction instead of renewable penetration, purchasing voluntary RECs is not a viable solution. Directly purchasing renewable energy and the bundled RECs which accompany it leads to greater, but still imperfect, additionality. However, since they carry real market weight, bundled credits are more expensive. PPAs, since the project is unlikely to exist without the agreement, boast nearperfect additionality, as do on-site solar panels. For reference, the University is currently engaged in purchasing-based sustainability initiatives. These include the aforementioned hydropower dam and REC purchases (bundled and unbundled) which equated 14% of Pitt’s energy consumption in financial year 2019. The solutions discussed in this study would, obviously, expand on these preexisting initiatives.

5. Conclusion

Several of the proposed solutions in this project are typical for institutions trying to restructure their energy profile. In the case of solar panels, the return on investment decreases as panels are added to inefficient buildings, becoming prohibitively low after a certain point. Furthermore, this investment would impact only a small proportion of the University’s total electricity consumption. Purchasing large quantities of RECs would change the University’s energy profile, but there would be little to no measurable reduction in carbon emissions because of the weak legal structure surrounding RECs. PPAs are more viable than onsite solar generation and more effective than RECs, thereby making them the best decarbonization strategy examined in this study. As Pitt moves forward with sustainability initiatives, it will need to explore multiple strategies and implement the most viable ones

Ingenium 2020

concurrently. This is especially true when considering solar panels: as detailed in the results, a full-scale solar installation project would represent only a fraction of the University’s energy profile. The inefficiency of such a solar installation relative to PPAs make them a poor investment on any scale. This study recommends that PPAs serve as the primary solution, and that any potential investment in rooftop solar panels be spent on more potent sustainability initiatives. These other initiatives may include university-owned generation on a larger scale. For example, installing a solar park near campus would circumvent the building-by-building limitations and take advantage of the economies of scale which make PPAs so effective. Furthermore, advances in photovoltaics and associated drops in the price of solar panels will continue to make solar generation more affordable and accessible.

6. Acknowledgements

Funding was provided through the SSOE Summer Research Internship with advisors Dr. Robert Kerestes and Dr. Katrina Kelly. We would also like to thank Dr. Aurora Sharrard, who met with us to go over the specifics of the University’s decarbonization landscape.

7. References

[1] 2018 Pitt Sustainability Plan. University of Pittsburgh. 12.5.2017. Accessed 8.20.2019. https://issuu.com/upitt/ docs/2018_pittsustainabilityplan_final. [2] M.M. Bilec, H. Gardner, Greenhouse Gas Inventory of University of Pittsburgh for FY 2017. 5.29.2018. Accessed 10.25.2019. [3] LevelTen Energy Q1 2019 PPA Price Index. 5.8.2018. Accessed 10.25.2019. https://leveltenenergy.com/blog/ppa-priceindex/q1-2019/ . [4] P4 2018/P1 2019 Solar Energy Update. NREL. 5.2019. Accessed 10.25.2019. https://www.nrel.gov/docs/fy19osti/73992. pdf. [5] E. O’Shaughnessy, J. Heeter, J. Sauer. Status and Trends in the U.S. Voluntary Green Power Market (2017 Data). NREL. 10.2018. Accessed 10.25.2019. https://www.nrel.gov/docs/ fy19osti/72204.pdf. [6] D. Roberts. “It’s easy to buy ‘green power.’ Making a difference is a little harder.” Vox. 11.6.2015. Accessed 10.25.2019. https://www.vox.com/2015/11/16/9744620/supportrenewable-energy.


Progress in bioplastics: PLA and PHA Samantha P Bunkea, Dr. Susan Fullertona, Dr. Eric Beckmana DAMP Nanoionics and Electronics Laboratory, Department of Chemical Engineering, University of Pittsburgh, PA, USA


Samantha Bunke is a Chemical Engineering graduate from the University of Pittsburgh. During her time at Pitt, she conducted research in the Nanoionics and Eletronics Lab under Dr. Susan Fullerton for the past year, and she has also completed a co-op program with EQT Corporation. She will be working for Lubrizol as their Physical Properties Intern until she attends graduate school next fall at Stanford University.

Samantha Bunke

Susan Fullerton is an Assistant Professor and Bicentennial Board of Visitors Faculty Fellow in the Department of Chemical and Petroleum Engineering at the University of Pittsburgh. She earned her Ph.D. in Chemical Engineering at Penn State in 2009. Prior to joining Pitt, she was a Research Assistant Professor in the Department of Electrical Engineering at the University of Notre Dame from 2009 - 2015. Fullerton’s work has been recognized by an NSF CAREER award, a Marion Milligan Mason award for women in the chemical sciences from AAAS, a Ralph E. Powe Jr. Faculty Award from ORAU, and the 2018 James Pommersheim Award for Excellence in Teaching in Chemical Engineering at Pitt. Susan Fullerton

Eric Beckman

Eric Beckman received his Ph.D. in polymer science and engineering from the University of Massachusetts-Amherst in 1988. After postdoctoral research at Battelle’s Pacific Northwest Laboratory in 1987-88, Dr. Beckman assumed his faculty position at the University of Pittsburgh (1989). Dr. Beckman was promoted to associate professor in 1994, and full professor in 1997. Dr. Beckman received a Young Investigator Award from the National Science Foundation in 1992, and the Presidential Green Chemistry Award in 2002. Dr. Beckman’s research group examines the use of molecular design to solve problems in green engineering and in the design of materials for use in tissue engineering. In 2003, with support from the Heinz Endowments, the Bevier estate, and John Mascaro, Dr. Beckman created the Mascaro Center for Sustainable Innovation, a school of engineering institute that examines the design of more sustainable products and infrastructure. Dr. Beckman’s group has published over 200 papers and he has received over 40 patents.

Significance Statement

The findings in this article provide background on the plastic accumulation dilemma and insight on the current state of bioplastics. Bioplastics, such as PLA and PHA, have been proposed as a solution to this problem, and this paper assesses their pros and cons.

Category: Review paper

Keywords: Bioplastics, biodegradation, PLA, PHA

16 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020


Plastic has become a paradox and encumbrance in the modern age as a one-use product that is designed to last forever. The contradictory nature of plastic is why the material is now accumulating in landfills as well as the natural environment and is causing major issues in today’s society. Recycling is not a solution to this crisis since it is typically more expensive than manufacturing virgin plastics, and it only makes up 9% of the life cycle of all currently produced plastic [1]. However, a possible solution has presented itself in nature: bioplastics. Bioplastics are defined as plastics made from any biological material [2]. Their attractiveness lies in their potential to biodegrade in nature and to reduce greenhouse gas emissions in the production stage. However, research has determined that this is not always the case. This paper reviews the manufacturing process and degradation methods of two widely used bioplastics, polylactic acid (PLA) and polyhydroxyalkanoate (PHA). Considering the pros and cons of each biopolymer, the future of bioplastics may lie in expanding past these two materials to other biopolymers, possibly combining two or more into composites possessing the positive attributes of each, rather than looking at these bioplastics on an individual basis.

1. Introduction

Roughly 6,300 metric tons of plastic waste have been produced since 2015 alone [1]. 9% of this amount was recycled, 12% was incinerated, and 79% found its way into landfills or the environment [2]. Following this trend, the amount of plastic waste that will accumulate in landfills or the environment is projected to reach 12,000 metric tons by 2050 [3]. The most frequently used plastics are derived from petrochemicals and are not biodegradable [3]. The only methods of permanent disposal include combustion and pyrolysis, which also release highly toxic pollutants [3]. While most organic materials undergo biodegradation where bacteria break the material down into smaller, useful compounds, the only “decomposition” experienced by petrochemical-based plastics is the fracture into microplastic beads [3]. Despite the drawbacks, the demand for plastic continues to increase. Bioplastics have been introduced as replacement materials since they can possess similar properties to conventional plastics while offering other benefits, such as biodegradability and/or reduction in the carbon footprint [2]. The current global production rate for bioplastics is only at 912 thousand metric tons, but is predicted to rise to 1.3 million metric tons by 2023 [2]. This paper will review two major bioplastics currently being utilized, polylactic acid (PLA) and polyhydroxyalkanoate (PHA), and their environmental impact.

2. Production 2.1 PLA PLA is an aliphatic polyester produced from the fermentation of renewable biomass, such as corn and sugarcane [4]. This fermentation process produces lactic acid which can then be

converted into PLA using one of two processes: direct condensation of the lactic acid or ring-opening polymerization of the cyclic lactide dimer [5]. Due to the difficulties of removing the water by-product that would drive the equilibrium reaction forward, the condensation reaction approach limits the molecular weight of the final polymer [5]. The current method used in industry combines both production routes. Lactic acid is made from the fermentation of dextrose from corn and then undergoes a continuous condensation reaction to achieve a PLA prepolymer at low molecular weights [5]. Next, the prepolymer is introduced to a catalyst to cause a highly selective intramolecular cyclization reaction and purified in a vacuum distillation column [5]. This lactide stereoisomer mixture is then polymerized into high molecular weight PLA using a catalyzed ringopening reaction [5]. The monomer of PLA is shown in figure 1a. 2.2 PHA PHA is a natural polyester synthesized by bacteria to act as an energy reserve under nutrient stress [6]. This polymer can be extracted from the bacteria, processed, and then used as a plastic material in a range of applications. Currently, largescale production of PHA uses expensive sources of organic materials, making the polymer an unattractive substitute for cheap, conventional plastics [7]. Although these organic materials are often renewable resources, such as starch, cellulose, triacylglycerols, and sucrose, alternative substrates are being researched to lower the cost of production [8]. Waste streams, such as agriculture feedstock, industrial by-products, or food waste, are currently being tested as carbon sources for small-scale PHA production systems [7]. This method follows a circular-economy approach by reclaiming waste products rather than immediately disposing of them, which is considered a more environmentally-friendly practice. However, large-scale execution would require improvements in industrial equipment and technology for using waste streams, posing an additional obstacle [7].

3. Properties and Use 3.1 PLA PLA possesses similar physical properties to polypropylene (PP), polyethylene (PE), and polystyrene (PS), which is why it is currently being used to make a variety of plastic products such as cups, films, food containers, packaging, and more [4]. It has also been utilized in the medical industry as screws, pins, rods, sutures, and other medical devices that are required to biodegrade within the body [4]. PLA is a thermoplastic, meaning it becomes flexible and moldable at elevated temperatures and can be treated using the majority of polymer processing equipment already in existence [5]. It can be injection molded, thermally crystallized, or stress crystallized, making it simple to work with [5]. However, the cost of producing this bioplastic is roughly $4.40/kg, which remains greater than petroleum-based plastics and limits PLA’s success in the commodity sector [5].


3.2 PHA Over 100 different monomers have been discovered as building blocks for storage PHA, allowing this biopolymer to be produced with a variety of properties [8]. A generalized form of the PHA monomer is shown in figure 1b. Although the size and composition of the polymer depends on the bacteria used to produce it, PHA chains typically have a molecular mass high enough to portray similar mechanical properties to most synthetic plastics, such as polypropylene [9]. PHA is also highly crystalline, isotactic, piezoelectric, optically active, insoluble in water, and is a thermoplastic [8]. These valuable properties allow PHA to be used in a wide range of applications, including packing, commodity items, and medical devices [8]. The price of production is highly dependent on the carbon source used, but it currently ranges between $4/kg and $16/kg [8]. This is why waste streams are a desirable option for carbon substrates since they can considerably lower manufacturing costs [8].

Figure 1: The monomer structures of a) PLAand b) PHA [5,10].

4. Environmental Impact 4.1 PLA PLA dominates 43.1% of the global market share for bioplastics [11]. This large share is due to the attractive properties that PLA possesses and its potential as a sustainable alternative to petroleum-based polymers. A 2018 life cycle impact assessment conducted on the carbon footprint of certain bioplastics versus conventional plastics found that PLA has a 23% lower global warming potential (GWP) than low-density polyethylene (LDPE), based on a cradle-to-grave analysis [11]. This analysis was also based on the assumption that both ended their life-cycles in a landfill, rather than PLA’s intended “grave:” an industrial composting facility [11]. One of PLA’s main manufacturers, Cargill Dow, conducted its own assessment of the polymer’s environmental impact based on a cradle-to-pellet analysis. The comparison of PLA’s CO2 production versus petrochemical-based plastics can be found in figure 2.

What is supposedly the most attractive quality of PLA, and is the lead motivation behind pushing this material as a “green” alternative, is its biodegradability. Its degradation occurs in two steps. First, the PLA polymer chains must be broken up into lactic acid and smaller oligomers by hydrolysis under high heat (~58ºC) and moisture (50% relative humidity) [12, 13]. Once this occurs, microorganisms can feed on the compounds to produce water, carbon dioxide, and humus [15]. PLA cannot be broken down in residential composting piles or the natural environment since it requires extreme conditions, but the polymer should sufficiently degrade in industrial composting facilities within two months [14]. In spite of this fact, out of the 50,000 tons of PLA waste that was produced in 2013, only 10% was recovered through composting [16]. What seems to be the main obstacle preventing composting from being the primary disposal method of PLA is that many industrial composting facilities will not accept it [13]. PLA is often indistinguishable from conventional plastics making it difficult to separate the bioplastic from the rest, which can lead to contaminating compost piles with non-compostable materials [13]. 4.2 PHA A study completed in 2012 found that on average, PHA production emits 2 kg of CO2 less than petroleum-based plastics and reduces fossil fuel usage by 30 MJ per l kg of material manufactured [8]. Unlike PLA, PHA can biodegrade in almost any environment, not just industrial composting facilities [8]. This means that PHA can be composted in homes and will break down in the environment. What is more unique is PHA’s ability to degrade in aquatic environments, even with a maximum temperature of 6°C [8]. The time it takes to completely biodegrade depends on environmental factors, such as temperature, moisture content, microbial activity, and pH [8]. Biodegradation speed also depends on the properties of the polymer itself, such as crystallinity, composition, molecular weight, and surface area exposure [8]. In an industrial composting facility, biodegradation lasts about seven weeks [8]. In aerobic conditions, biodegradation of PHA produces carbon dioxide and water, while anaerobic conditions result in carbon dioxide and methane [8].

Figure 2: CO2 production of petrochemical polymers versus PLA. PC=polycarbonate; HIPS=high impact polystyrene; GPPS=general purpose polystyrene; LDPE=low density polyethylene; PET SSP=polyethylene terephtbalate, solid state polymerization (bottle grade); PP=polypropylene; PET AM=polyethylene terephthalate, amorphous (fibers and film grade); PLAl=polylactide (first generation); PLAB/WP (polylactide, biomass/wind power scenario) (15).

18 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

5. Conclusion and Outlook

Plastics acquired their popularity because of the durability, low costs, and convenience they offer to businesses and consumers. They laid the foundation for the linear-economy we live in today, in which production, consumption, and disposal follow a straight path right into landfills and the environment. This has now revealed itself as a major problem as these seemingly indestructible materials have been accumulating since they were first introduced and are now posing an environmental threat. Bioplastics offer an opportunity to maintain the convenience that consumers have grown accustomed to, without building upon the already existing issue. They are often manufactured using renewable sources, lowering greenhouse gas emissions during the production stage, and are biodegradable, preventing them from accumulating in the environment. PLA and PHA have shown a lot of promise as the future of the plastics industry, as one is already being mass-produced and the other is being researched extensively. However, each bioplastic comes with its own obstacles. PLA is highly durable and possesses the properties necessary to make it a prime replacement for conventional plastics, but these properties come at a cost; it cannot biodegrade on its own in the environment because it requires more extreme conditions. This obstacle can be overcome with the implementation of industrial composting facilities, but these facilities are still very limited and many of them will not even accept PLA because of the difficulty of separating it from non-compostable plastics. With the high likelihood of this bioplastic accumulating in landfills, critics have labelled PLA as corporate “green washing,” meaning it is a ploy to trick businesses and consumers into believing they are actually making a difference when in reality, they are still contributing to the plastic crisis. The major advantage of PHA is that it can biodegrade in a vast range of environmental conditions so it will not accumulate in landfills or in nature. It has many attractive properties which means it can replace conventional plastics in a variety of applications. However, due to the carbon sources and extraction methods being used to produce it, PHA is much more expensive then petroleumbased plastics and is currently not a viable option. Technological advancements need to be made in order to recover waste streams and use them as carbon sources for PHA production, not only lowering the costs of this bioplastic, but also developing a circulareconomy model. Despite these complications, biopolymers still have the potential to aid in the plastic struggle. PLA and PHA are on the correct track of becoming smarter, more environmentallysensible materials, but they have yet to reach the point of actually replacing conventional plastics. A possible solution to one of PLA’s obstacles is to require PLA products to be labelled to make them distinguishable from non-compostable plastics. To overcome PLA’s inability to biodegrade quickly and easily in industrial composting facilities, PLA could be blended with another biopolymer, such as PHA, polyalginate, lipids, starch, etc., to enhance its biodegradability, while maintaining PLA’s strength and durability. These additional biopolymers are currently being trialed, and

perhaps they will possess the correct properties needed to replace conventional plastics—even without blending—and succeed in reforming the “throwaway” society.

6. Acknowledgements

Funding was provided by Swanson School of Engineering, the Office of the Provost, and the Ellen MacArthur Foundation.

7. References

[1] “The Veolia Institute Review: Facts Reports.” Veolia Institute, 2019. [2] “Bioplastics—Facts and Figures.” European Bioplastics, 2017. [3] R. Geyer, J. R. Jambeck, K. L. Law, “Production, use, and fate of all plastics ever made.” Sci. Adv. 3, e1700782 (2017). [4] Rogers, Tony. “Everything You Need to Know about Polylactic Acid (PLA).” Creative Mechanisms, 7 Oct. 2015, https:// www.creativemechanisms.com/blog/learn-about-polylactic-acidpla-prototypes. [5] Henton, D. et al. “Polylactic Acid Technology.” Natural Fibers, Biopolymers, and Biocomposites, 11 Feb. 2005. [6] “Everything You Need to Know about PHA.” Creative Mechanisms, 22 Jan. 2017, https://www.creativemechanisms. com/blog/everything-you-need-to-know-about-phapolyhydroxyalkanoates [7] Rodriguez-Pereza, S. et al. “Challenges of Scaling-up PHA Production from Waste Streams. A Review.” Journal of Environmental Management, Academic Press, 6 Oct. 2017. [8] Reddy, C.S.K. et al. “Polyhydroxyalkanoates: an overview.” Biosource Technology, 2003. [9] Raza, Z. et al. “Polyhydroxyalkanoates: Characteristics, production, recent developments and applications.” International Biodeterioration & Biodegradation, 2018. [10] Li, Z. et al. “Polyhydroxyalkanoates: opening doors for a sustainable future.” NPG Asia Materials, 2016. [11] Choi, B. et al. “Carbon Footprint of Packaging Films Made from LDPE, PLA, and PLA/PBAT Blends in South Korea.” Sustainability, 2018. [12] Henton, D. et al. “Polylactic Acid Technology.” Natural Fibers, Biopolymers, and Biocomposites, 11 Feb. 2005. [13] Castro-Aguirre, E. et al. “Poly(lactic acid)—Mass production, processing, industrial applications, and end of life.” Advanced Drug Delivery Reviews, 2016. [14] “The Sustainability of NatureWorksTM Polylactide Polymers and IngeoTM Polylactide Fibersa: An Update of the Future.” Macromolecular Bioscience, 2003. [15] Vink, E. et al. “Applications of life cycle assessment to NatureWorksTM polylactide (PLA) production.” Polymer Degradation and Stability, 2002. [16] EPA, “Advancing Sustainable Materials Management: Facts and Figures.” 2013. Access date Oct 5th 2019.


Three-dimensional nickel foam and graphene electrode in microbial fuel cell application: Study of biofilm compatibility Claire P. Chouinarda, Felipe Sanhueza Gómezb, c, Natalia Padilla Gálvezd, Hernán Valle Zapatab, c, R.V. Mangalarajab, c, and Homero Urrutiad Department of Chemical and Petroleum Engineering, Swanson School of Engineering, University of Pittsburgh, Pittsburgh, PA, USA, bLaboratorio de Cerámicos Avanzados y Nanotecnología, Departamento de Ingeniería de Materiales, Universidad de Concepción, Concepción, Chile, cUnidad de Desarrollo Tecnológico, Universidad de Concepción, Coronel, Chile, dLaboratorio Biopelículas y Microbiología Ambiental, Departamento de Microbiología, Universidad de Concepción, Concepción, Chile a

Claire Chouinard

Claire Chouinard is a senior chemical engineering student from Ann Arbor, MI and is motivated by the pressing need to improve global water equality. After graduation, she plans to pursue a doctorate degree in environmental engineering with a focus on the biological aspects of water analysis and remediation. Dr. Felipe Sanhueza Gómez is a materials engineer in the field of materials science and engineering. He currently holds a research position at the Unidad de Desarrollo Tecnológico of the Universidad de Concepción, Chile.

Dr. Felipe Sanhueza Gómez

Significance Statement

Current methods of domestic and industrial wastewater treatment are energy intensive and unsustainable, and microbial fuel cells have the potential to create a paradigm shift in wastewater treatment by generating energy from wastewater itself. This study investigates the potential of nickel foam and graphene as a possible electrode material.

Category: Experimental research

Keywords: Microbial fuel cell, graphene synthesis, three-dimensional electrode, biofilm analysis

20 Undergraduate Research at the Swanson School of Engineering


Microbial fuel cells (MFCs) have the potential to transform both domestic and industrial wastewater treatment sectors, posing an alternative to existing methods of wastewater treatment. MFC technology generates energy from wastewater by exploiting the properties of electroactive microorganisms to generate a current while simultaneously decomposing organic matter. MFCs are not yet practical for wide-scale implementation due to high material costs and system inefficiencies. The objective of this project was to synthesize an improved anode material optimized for MFC systems from nickel foam (NF) and graphene. NF/graphene materials offer a low-cost alternative to traditional precious metal electrodes and electrical properties more robust than carbon papers and foams. The biocompatibility of the synthesized anode with bacterial communities was evaluated, and biofilm growth on NF and graphene electrodes was measured both for pure bacterial strains and for environmental samples. Although additional experiments are required to draw definitive conclusions, preliminary results demonstrate antibacterial activity in the NF and graphene materials, which could hinder MFC performance (i.e. current generation).

1. Introduction

Microbial fuels cells (MFCs), a type of bioelectrochemical system (BES), have the ability to revolutionize traditional methods of wastewater treatment by generating energy through the use of electroactive bacterial biofilms. In contrast, current methods of wastewater treatment are energy intensive, depleting a limited fuel supply and releasing environmental pollutants. Due to escalating water management concerns, water treatment was identified by the National Academy of Engineering as an Engineering Grand Challenge [1]. Despite prior work, MFCs still suffer from a wide variety of inefficiencies that prevent their immediate use in wastewater remediation at the industrial scale. Inefficiencies include both the slow rate of anaerobic respiration at the anode and formation of the biofilm itself [2]. Improvements to the anode material could mitigate these issues, increasing feasibility of wide-scale implementation of MFC technology. The objective of this project was to develop more effective anode materials for implementation in an MFC system and to analyze the compatibility of these anode materials with bacterial communities. Carbon materials, including three-dimensional carbon foams, have routinely been explored as a potential anode material for MFC systems, and Chen et al. found that reticulated carbon foam derived from a sponge-like natural product performed relatively well in an MFC application, with a current density of over 4.0 mA cm-2 [3]. Similarly, Zhao et al. identified a layer by layer synthesis method of three-dimensional nickel foam (NF) electrodes coated with graphene layers for use in oxygen evolution reactions, and electrical properties of the synthesized electrodes were comparable to that of state-of-the-art materials including Ir/C and Ru/C electrodes, demonstrating a notable current density of 10 mA cm-2 [4]. Although the Zhao procedure offers an economical three-dimensional anode material with superior electrical

Ingenium 2020

properties, the potential for use in an MFC setting has not yet been explored and could offer interesting results. The success of MFC technology fundamentally depends on the ability of electroactive bacteria, typically organized in a biofilm formation on the surface of the anode, to oxidize organic materials. Current density has been shown to have a positive correlation to biomass [2]. The present study completes a preliminary biocompatibility analysis of NF/graphene electrode material prepared using a modification of the Zhao procedure with analytical methods including planktonic cell counts and confocal microscopy [4]. This analysis is relevant due to the known toxicity of nickel and due to the compulsory biocompatibility of materials in MFC design and preparation [5]. It was hypothesized that antibacterial activity would be observed.

LSM780 NL0 Zeiss Spectral Confocal Microscope at magnifications of 25x and 40x. Confocal microscopy images were analyzed by the Centro de Microscopía Avanzada at the Universidad de Concepción. The percent volume of live and dead cells relative to the total electrode volume were determined using manually set controls and were held constant across samples, except for Pectobacterium sp., which was analyzed using a different magnification. Additionally, planktonic cell counts were collected at day 4 for both plastic and carbon cloth supports, used as controls for the NF/graphene materials. The plastic supports are known to encourage biofilm growth, and carbon is relevant given previous work in the field. All samples were stained using a LIVE/DEAD® BacLightTM Bacterial Viability Kit (N° L13152) purchased from ThermoFisher (Chile).

2. Materials and methods

Images collected using SEM technology are provided for the NF/graphene material prior to biofilm growth in Figure 1, revealing a uniform graphene coating on the surface of the NF and validating the coating and reduction methods used in the material synthesis.

Modifications were made to the Zhao procedure to optimize the synthesized material for use in an MFC system [4]. NF and graphene electrode samples of 1 cm by 1 cm were created by first submerging NF in acetone for 20 minutes, rinsing with deionized water, submerging in 0.10 M HCl for 20 minutes, and rinsing again with deionized water. After cleaning the NF base, graphite oxide (GO) layers were cyclically developed on the surface of the NF. Electrodes were submerged in a 6.25 mg/mL solution of poly(ethyleneimine) (PEI) at a pH of 10 for 20 minutes. The PEI solution was held at pH 10 rather than pH 7 due to the weak polyelectrolyte nature of PEI [4]. Following a rinse with deionized water, electrodes were then submerged in a graphite oxide (GO) suspension (GO-325) of 1.5 mg/mL at a pH>10 for 20 minutes. GO was prepared using a modified version of Hummer’s method [6]. Electrodes were removed from the solution and air dried. Steps were repeated until a Carl Zeiss 0.1-30 kV scanning electron microscope (SEM) revealed a uniform graphene layer, between 5 and 7 times depending on the trial. The GO coating was reduced to graphene with functionalized nickel nanoparticles using a mix of 20 mL of “Ni(NO3)2∙6H2O” 10 mM (58.2 mg/20 mL agua milli-Q) and 4 mL of L-ascorbic acid (LAA) 120 mg/mL. Electrodes were left in the solution for 4 hours at 80°C under gentle magnetic stirring. LAA was selected as the reducing agent rather than hydrazine as used by Zhao et al. because it exhibits a significantly lower inherent toxicity, enabling the implementation of the synthesized material in biological systems [4,7]. All chemicals used in this procedure were of high purity reagents and were purchased from Sigma-Aldrich (Chile). Following UV radiation for the purpose of sterilization, 2 NF and graphene electrodes were introduced in each of 6 cultures: Tryptic Soy Broth (TSB) only, Streptomyces sp., Pectobacterium sp., wastewater, sludge, and solid waste. A TSB control without NF/graphene electrodes was included as well. All environmental samples (wastewater, sludge, and solid waste) were collected from ESSBIO wastewater treatment plant facility in Concepción, Chile. Planktonic cell counts were collected at day 2 and at day 11 using an Olympus BX51 Epifluorescence Microscope at a magnification of 1000x, and biofilm images were collected at day 11 using a

3. Results

Figure 1: SEM imaging for GO reduced to graphene on NF. Uniform graphene coating after 5–7 cycles of PEI and graphite oxide, with LAA as the reducing agent.

Cell count data showed bacterial growth was absent from the TSB controls. Filaments present in the Streptomyces sp. culture prevented data collection for all types of biofilm supports and these data points are not included. Planktonic cell concentrations for day 2, day 4, and day 11 in are presented in Table 1 and Table 2, respectively. Reported planktonic cell concentrations were determined using the average of 4 manually collected counts, and 95% confidence intervals are included for each data point. Differences between days 2 and 11 are not statistically significant for Pectobacterium sp., while the cell counts between days 2 and 11 significantly increase for wastewater and sludge and significantly decrease for solid waste. However, both the carbon cloth and the plastic supports show consistently higher planktonic cell counts than the NF/graphene, reaching up to 7.94×109±2.04×109 cell/mL. 21

NF/Graphene (cell/mL) Culture

Day 2

Day 11

Pectobacterium sp.

3.51×109 ± 5.1×108

3.45×109 ± 5.3×108


1.90×109 ± 1.8×108

3.42×109 ± 4.6×108


2.27×109 ± 2.5×108

3.00×109 ± 6.8×108

Solid waste

3.46×10 ± 2.3×10

1.36×109 ± 3.2×108



Table 1: Planktonic cell concentrations for live growth on NF/graphene

Plastic (cell/mL)

Carbon Cloth (cell/mL)


Day 4


5.28×109 ± 1.25×109

5.20×109 ± 1.11×109


5.72×109 ± 1.79×109

5.17×109 ± 4.4×108

Solid waste

7.94×10 ± 2.04×10

6.50×109 ± 1.74×109



Table 2: Planktonic cell concentrations for live growth on support controls

Confocal microscopy images are included in Figure 2 for live cells at day 11 in the NF/graphene electrodes. Quantitative results for percent volume of live and dead cells are shown in Table 3. Overall, visual analysis revealed a higher concentration of live cells for the environmental samples than for Pectobacterium sp.

Figure 2: Confocal microscope images of adhered biofilm. Environmental samples provided more prominent biofilm than Pectobacterium sp. at day 11 of biofilm growth on NF/graphene. Culture

Pectobacterium sp.

Live (% volume)

Dead (% volume)









Solid waste



Table 3: Calculated biofilm growth on NF/graphene electrodes by percent total volume

4. Discussion

NF/graphene planktonic cell counts ranged from 1.36×109±3.2×108 cell/mL to 3.51×109±5.1×108 cell/mL, and both the plastic and the carbon cloth supports appeared to be more compatible with biofilm growth, reaching up to 7.94×109±2.04×109 cell/mL and 6.50×109 ± 1.74×109 cell/

22 Undergraduate Research at the Swanson School of Engineering

mL, respectively. Strictly within the NF/graphene materials, the planktonic cell counts across cultures were inconclusive. After 2 weeks of growth, confocal microscopy revealed biofilms colonizing from 10.58% to 31.62% of the total sample volume. These results, across all samples, were less prevalent than expected, supporting the likelihood of antibacterial activity in the NF/graphene electrodes. Similar work by Blanchet et al. reported biofilm growth of up to 39.3±1.1% on carbon cloth using activated sludge fed with food wastes, at nearly 10% greater than the 31.62% obtained by this work [8]. The environmental samples showed increased biofilm formation relative to the pure strain Pectobacterium sp., likely due to interspecies cooperation and greater adaptability within a multispecies biofilm [9]. This is observed in the confocal microscopy images in Figure 2, as the density of living cells (green) appears lower for Pectobacterium sp. than for the environmental samples (wastewater, sludge, and solid waste). The observed toxicity of the NF/graphene electrodes may be due to the NF or graphene, as nickel is known for its toxic effects and graphene has been shown to cause membrane and oxidative stress in bacterial cells [5,10]. It is recommended that the antibacterial effects of the synthesized materials be further explored with more extensive controls and larger sample sizes, with a focus on mixed culture inoculants from environmental samples. Future work will focus on more extensive data collection, and controls will be used to isolate the particular source of antibacterial activity. However, it is possible that the increased electrical performance of NF/graphene electrode will allow for a small amount of remaining antibacterial activity in the materials. Electrical activity could compensate for a slight decrease in biofilm growth, and a valuable MFC system will demonstrate improved current generation.

5. Conclusions

This work focused on a preliminary analysis of the antibacterial effects of NF and graphene-based anode materials synthesized using a modification of the procedure originally presented by Zhao et al. [4]. Pure strains and mixed cultures were included in the analysis, with biofilms grown using Pectobacterium sp., Streptomyces sp., and 3 types of environmental samples (wastewater, sludge, and solid waste) collected from the ESSBIO wastewater treatment plant facility in Concepción, Chile. Initial results support the presence of inherent antibacterial activity in the NF/graphene materials, although further work is required to verify these conclusions. Particularly, more thorough controls need to be tested to isolate and to subsequently reduce the source of the toxicity. However, the enhanced electrical properties of NF/ graphene electrodes could compensate for a minimal degree of antibacterial activity, and NF/graphene materials warrant future investigation. Developments in this technology could allow for the eventual implementation of MFCs in both municipal and industrial sectors, significantly lowering the energy cost of current wastewater treatment methods.

Ingenium 2020

6. Acknowledgements

This work was funded by the NSF REU Project #1757529 and the National Commission for Scientific and Technological Research FONDEF IDEA No. ID18I10337, Concepciรณn, Chile. Claire Chouinard would like to acknowledge the University of Maine FBRI REU program, as well its directors, Dr. Douglas Gardner and Dr. David Neivandt. She would also like to thank Dr. David Sanchez.

7. References

[1] NAE Grand Challenges for Engineering. Accessed October 14, 2019. https://www.nae.edu/187212/NAE-Grand-Challengesfor-Engineering [2] T. H. Pham, P. Aelterman, W. Verstraete, Bioanode performance in bioelectrochemical systems: Recent improvements and prospects, Trends in Biotech., 27 (2009) 168-178. [3] S. Chen, Q. Liu, G. He, et al., Reticulated carbon foam derived from a sponge-like natural product as a high-performance anode in microbial fuel cells, J. of Mater. Chem., 22 (2012) 1860918613. [4] M. Zhao, W. Yuan, C. M. Li, Controlled self-assembly of Ni foam supported poly(ethyleneimine)/reduced graphene oxide threedimensional composite electrodes with remarkable synergistic effects for efficient oxygen evolution. J. of Mater. Chem. A, 5 (2017) 1201-1210. [5] L. Macomber, R. P. Hausinger, Mechanisms of nickel toxicity in microorganisms, Metallomics, 3 (2011) 1153-1162. [6] W. S. Hummers and R. E. Offeman, Preparation of graphitic oxide, J. Am. Chem. Soc., 80 (1958) 1339. [7] K. D. Silva, H. Huang, R. Joshi, et al., Chemical reduction of graphene oxide using green reductants, Carbon, 119 (2017) 190-199. [8] E. Blanchet, B. Erable, M. Solan, et al., Two-dimensional carbon cloth and three-dimensional carbon felt perform similarly to form bioanode fed with food waste, Electrochem. Comm., 66 (2016) 38-41. [9] H. Flemming, J. Wingender, U. Szewzyk, et al., Biofilms: An emergent form of bacterial life. Nature Reviews Microbio., 14 (2016) 563-575 [10] J. Chen, F. Deng, Y. Hu, et al., Antibacterial activity of graphene-modified anode on Shewanella oneidensis MR-1 biofilm in microbial fuel cell, J. of Power Sources, 290 (2015) 80-86.


Extensions and analysis of a virtual balancing task for studying sensory-motor control Michael Clancya, Sudarshan Sekhar, PhDa, Aaron Batista, PhDa, Patrick Loughlin, PhDa Department of Bioengineering, University of Pittsburgh


Michael Clancy is an undergraduate student who has been studying Bioengineering at the University of Pittsburgh since 2016. His research interests include motor control and sensory feedback; and machine learning and neural networks.

Michael Clancy

Sudarshan Sekhar did his bachelor’s in electrical engineering from Anna University in Madras, India. He then did his masters & PhD at the University of Tuebingen, Germany, and is currently a postdoc at the University of Pittsburgh.

Sudarshan Sekhar, PhD

Aaron Batista is an Associate Professor in Bioengineering at the University of Pittsburgh. Prior to joining Pitt’s faculty he earned a PhD in Computation and Neural Systems at the California Institute of Technology, and pursued postdoctoral research at Stanford University. Dr. Batista’s research include the neurophysiology of skilled behavior.

Aaron Batista, PhD

Patrick Loughlin is a Professor and Associate Chair of Bioengineering at the University of Pittsburgh, which he joined in 1993 after earning a PhD in Electrical Engineering from the University of Washington. His research interests include sensory integration in motor control; control of human movement; sensory substitution, haptics, vibrotactile feedback; brain-computer interfaces; computational models; and signal processing. Dr. Loughlin is a Fellow of the American Institute for Medical and Biological Engineering (AIMBE), the Acoustical Society of America (ASA), and the Institute of Electrical and Electronics Engineers (IEEE). Patrick Loughlin, PhD

Significance Statement

Understanding how the brain encodes sensory information about its surroundings and in turn generates neural signals that drive movement could enable us to restore sensory-motor function to paralyzed individuals, through the development of brain-computer interfaces. Virtual object manipulation provides a means by which to experimentally investigate native sensory-motor function in controlled conditions that are a balance between natural yet highly variable movements versus highly repeatable yet simple movements such as point-to-point reaching. Here we analyze and extend one such task, the Critical Stability Task (CST).

Category: Methods

Keywords: motor control, sensory feedback, virtual object manipulation, neural networks, modeling

24 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020


Brain-computer interfaces (BCIs) aim to restore motor control to individuals with motor impairments by using recordings of brain activity to control the actions of a device, such as a prosthetic limb, robotic arm, or cursor on a computer monitor. To train a BCI to learn the mapping from neural activity to intended action, subjects are typically asked to perform highly stereotypical and repeatable tasks, such as repeatedly moving a cursor to a particular location on a screen while the activity of small groups of neurons is recorded. In contrast, most natural movements are more complex and varied, with motor and sensory actions working in concert in a feedback loop to control movement. However, acquiring and analyzing neural data while subjects perform natural movements in real environments is a very challenging undertaking. In the laboratory, virtual object manipulation provides a balance between natural yet highly variable movements versus simple yet highly repeatable movements. One such virtual task is the “Critical Stability Task” [1], which is a simulated balancing task akin to keeping a broom upright in the palm of one’s hand. Because the CST imposes a tight coupling between sensory feedback and motor action resulting in non-stereotypical movements, it provides a new experimental paradigm for the design and testing of BCIs. The two aims of this study were, first, to implement and test extensions to the CST, and second, to model and simulate the behavioral response of subjects performing CST. The extensions altered the underlying dynamics of the system, and as such they will enable future investigations into sensory integration and motor action to manipulate or control increasingly complex dynamical systems.

1. Introduction

Understanding how the brain decodes sensory information about its surroundings and in turn generates neural signals that drive movement could restore function to impaired individuals, through the development of brain-computer interfaces (BCIs) and advanced neuroprosthetics. To learn the mapping from neural activity to intended action, most BCIs are trained on highly reproducible and stereotypical tasks, such as repeatedly moving a cursor to a particular location on a screen while the activity of small groups of neurons is recorded. BCIs constructed in this way are very effective at allowing users to execute relatively simple but important actions such as moving a cursor on a screen to type or navigate a webpage. In contrast, most natural movements are more complex and varied, with motor and sensory actions working in concert in a feedback loop to control movement. Consider, for example, the intricate sensory-motor coupling required to generate the precise movements necessary to ride a bicycle without falling, or to balance a broom in the palm of one’s hand. However, acquiring and analyzing neural data from natural movements in real environments is a very challenging undertaking. Virtual object manipulation provides a reasonable compromise between natural yet highly variable movements versus simple yet highly repeatable movements. One such virtual task is the “Critical Stability Task”

(CST) [1], which is a simulated balancing task, akin to keeping a broom upright in the palm of one’s hand. Because the CST imposes a tight coupling between sensory feedback and motor action resulting in non-stereotypical movements, it provides a new experimental paradigm for the design and testing of BCIs, as well as the study of sensorimotor control in general. The purpose of the present study is to extend and refine the CST to enable the investigation of more complex behaviors. To achieve this, two aims were pursued: (1) to implement modifications to the CST, and test human performance on the modified task, and (2) to model and simulate the behavioral response of Rhesus monkeys performing the CST as a first step towards system identification of their underlying sensory-motor control. The modifications of the CST that were implemented altered the dynamics of the system. Specifically, the current implementation of the CST is a so-called ``first-order’’ system, which in theory is the simplest unstable system to control, requiring only sensory information about the position of the object. Higherorder dynamics as implemented here place greater demands on sensorimotor control that in theory require the subject to utilize additional sensory information about the system state, such as its position, velocity and acceleration. As such, these extensions of the CST should allow future investigations into sensory integration and motor action to control increasingly complex systems.

2. Methods The Critical Stability Task (CST) The CST is a virtual object manipulation task wherein subjects must make hand movements in order to keep a cursor on a screen from drifting left or right away from screen-center (Fig. 1).

Figure 1: An example from a standard CST trial, where the y axis indicates the time of the trial (stabilizing the cursor for 6 seconds is a success). The x-axis shows the horizontal onscreen cursor position over the duration of the trial. When the subject brings their hand to the center of the display monitor the trial starts, the cursor shown in gray then begins to drift horizontally off screen until corrective subject movements are made. The subject must aim to have their hand motions be equal and opposite to the cursor motions in order to stabilize the cursor. If the cursor drifts too far from the center, the trial ends and is counted as a failure.


As originally formulated, the CST is governed by a first-order differential equation [Jex 1966], (1) where the cursor velocity y'(t) depends on the cursor position, y(t), as well as the subject’s hand position, x(t), and the parameter λ>0 determines the task difficulty. In particular, the larger that λ is, the faster the cursor moves. In addition, because λ is nonnegative, the cursor will drift away from screen-center with increasing velocity in the absence of hand movements; specifically, when x(t) = 0, the cursor position (i.e., the solution to Eq. (1)) is given by y(t) = y0 eλt, where y0 is the initial cursor position. In order for the cursor not to move, i.e., such that its velocity is zero (y'(t) = 0 in Eq. (1)), the subject must make hand movements that are exactly equal and opposite to the cursor position: x(t) = –y(t). Hence, the CST is a feedback-driven task, in that sensory information about the cursor position is crucial for good performance (Fig. 2). Because of motor noise and sensorymotor delays, this ideal performance cannot be achieved in practice; hence, at some value of λ, the subject will eventually be unable to control the cursor, and it will rapidly drift off-screen, which we flag as a failure at the task. In this way, the parameter λ provides a quantitative measure of each subject’s sensory-motor control, with larger values being indicative of better performance.

As with the differential equation formulation of the CST in Eq. (1), the parameter a = eλTs > 1 determines task difficulty, with larger values causing the cursor to move faster. Additionally, in the absence of hand movements (x[n] = 0), the cursor will diverge from screen center, with position given by y[n] = any0. As with the original formulation in Eq. (1), in order for the cursor not to move in this discrete-time version of the CST (i.e., such that y[n+1] = y[n]), the subject must make hand movements that are exactly equal and opposite to the current cursor position: x[n] = –y[n]. This ideal performance is not possible because of sensory-motor delays and noise, and hence subjects will lose control of the cursor at some value of the parameter a, called the “critical instability value” (CIV). Aim 1: Extensions of the CST As shown above, for ideal performance (i.e., a stationary cursor), only information about the position of the cursor is necessary in the “first-order” instantiation of the CST (Eq. (2)). In theory, increasing the “order” of the CST will necessitate additional sensory information about the cursor (e.g., position, velocity, acceleration) in order to succeed at the task. For example, consider a second-order version of the CST given by y[n+1] = 2ay[n] – a2 y[n–1] + (a – 1)x[n] (3) Here, in order for the cursor not to move (y[n+1] = y[n]), hand movements must be generated according to

x[n] = K0y[n] + K1y[n–1]


or equivalently,

x[n] = Kpy[n] + Kv(y[n] – y[n–1])


where K0 = (1–2a) ⁄ (a–1), K1 = a2 ⁄ (a–1), Kp = (a–1), Kv = –a2/(a–1) Figure 2: A CST feedback control model. The blue pathway designates motor control based on sensory information (red) about the cursor. A future goal is to identify the unknown motor and sensory functions, Hm and Hs, by introducing motor and/or sensory perturbations (dm and ds).

The CST has been used to study sensory integration in balance [7], impairment of executive function and motor control [6], as well as mental workload [5] and neurological health [3, 4]. The CST was recently adapted for animal experiments in sensory-motor control via the discrete-time algorithm [1],

y[n+1] = ay[n] + (a – 1)x[n] (2)

where x[n] and y[n] are the hand and cursor position, respectively, at the current time nTs, y[n+1] is the cursor position to be rendered at the next update, and Ts is the update interval.

26 Undergraduate Research at the Swanson School of Engineering


Hence, in this second-order instantiation of the CST, hand movements depend on the current and prior cursor positions, which is analogous to requiring information about cursor position and velocity (Eq. (5), where the velocity is approximated by the discrete-time difference y[n] – y[n–1]). In the first aim of this study, we implemented higher-order versions of the CST, to allow future experiments to investigate the sensory requirements necessary to perform the task. Aim 2: Simulation and Modeling With reference to the feedback control model shown in Fig. 2, a desired future goal is to describe the monkey’s underlying sensory-motor control by computing the motor and sensory functions, Hm and Hs, respectively. As a step in that direction, the second aim of this study was to simulate the behavioral response of Rhesus monkeys performing the (first-order) CST.

Ingenium 2020

Simulations of CST performance were generated by a neural network designed using the Deep Learning and Machine Learning toolboxes provided by Mathworks. A Long Short-Term Memory (LSTM) neural network was chosen because it has “memory”, that is, the current output from the network can effect the next output. The LSTM was trained on Rhesus monkey behavioral data (hand and cursor positions) obtained during CST trials (140 training trials). At each time increment, the network was trained to predict the subject’s hand position at the next time increment, given information about the current cursor and hand. The cost function was the root mean squared error (RMSE) between the actual and predicted hand position. Different LSTM network architectures and types of training data (i.e., position only v. position + velocity) were investigated. The best network contained one LSTM layer with ten neurons, a dropout rate of 45%, and used the current hand position along with the current cursor position and velocity in order to predict the next hand position. After training, the optimized network was tested on held out data (37 trials) to assess its ability to predict the next hand position at each time increment. The network was also tested on its ability to perform the CST in place of a live subject; for these tests, the network was seeded with an initial hand position (i.e., hand position at the start of the trial) drawn from a distribution created from subject data. If the network was able to successfully perform the CST for first order, it could then be tasked with performing second order CST, as a first step towards quantifying the effect of increasing the system order.

generate a control action depends on task difficulty (i.e., the position and velocity gains in Eq. (5) are functions of the parameter “a”).

3. Results

Example trials of a subject (MC) performing the first- and second-order CST are shown in Fig. 3; a successful trial and a failed trial are shown in each case. We observed experimentally that a second-order implementation of the CST (Eq. (3)) was more difficult for the human subject to control than the first-order implementation (Eq. (2)). Fig. 4(a) shows the success rate as a function of increasing the value of the parameter “a” in equation 3 (i.e., increasing task difficulty) for a human subject (MC) performing the first-order CST, while Fig. 4(b) shows the success rate for the second-order CST. In both cases, performance degrades as the task becomes more difficult; the value of the system’s parameter “a” at which the subject fails 50% of the time on average (i.e., the CIV) is marked by ‘x’ in each plot. The CIV achieved for the secondorder case is substantially lower than for the first-order case. Our derivations in Eqs. (3)-(6) provide theoretical insights into why higher-order versions of the CST are more challenging to control. In particular, unlike the first-order CST for which ideal performance (i.e., stationary cursor) requires hand movements that depend only on the cursor position and are independent of task difficulty (x[n] = – y[n]), ideal performance at higher-order versions of the CST changes with task difficulty (parameter “a”), and moreover requires additional sensory information about the cursor beyond just its position. In particular, a second-order version of CST requires information about cursor position and velocity (Eq. (5)), and the degree to which this information is utilized to

Figure 3: Example trials showing the horizontal cursor and hand position versus time for a subject performing the first-order CST ((a) unsuccessful trial, (b) successful trial) and the second-order CST ((c) unsuccessful trial, (d) successful trial). The vertical scale is the same in all plots.


Figure 4 (above): Success rate of subject over a range of parameter values for the (a) first-order CST and (b) second order CST. The best fit line (black) to the data is a sigmoidal curve, displayed in the legend. The critical instability value (CIV) marked by a blue ‘x’ and displayed in the legend of each plot provides the value at which the subject can successfully complete the CST 50% of the time.

Figure 5 (right): (a) Cursor (red) and hand (green) position of a subject during one trial of the first-order CST, along with the predicted hand position (black dotted). (b) optimizing input parameters for the neural network, the red star indicates the network chosen to perform CST in place of a subject. The combination of cursor position and velocity and hand position had the lowest combination of RMSE and standard deviation among the 37 test trials. The abbreviations indicate the information the network was given at time t1 to predict the hand position at time t2 (C=cursor, H=hand, p=position, v=velocity, a=acceleration). (c) Three different trials of the neural network performing the CST.

28 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

Fig. 5(a) shows the cursor and hand position of a subject from one trial of the (first-order) CST, along with the hand position predicted by the neural network at each subsequent time increment, given the current cursor and hand information. The RMSE between the actual and predicted hand positions was 0.38 for this trial; the overall RMSE of the 37 test trials was 0.37 +/- .17. The bar graphs in Fig. 5(b) show the optimization of the combination of inputs to the network (cursor and hand position, CpHp, cursor position and velocity Cpv, etc.) Fig. 5(c) shows three different trials of the neural network performing the first-order CST.

4. Discussion and Conclusion

A benefit of the CST for studying sensory-motor control is that it challenges both the forward (motor) and the feedback (sensory) pathways of the control loop: successful task execution necessitates the use of sensory feedback about the cursor state in order to generate appropriate hand movements. By varying a single parameter, it can probe a subject’s skill and behavior over a range of task difficulty. Prior to this study, only a first-order version of the CST had been used (Eq. (1)), which in theory requires information about only the cursor position in order to succeed at the task. Manipulating a system with higher-order dynamics requires additional sensory information about the state of the system, and places a greater demand on sensory-motor performance. As shown by Eqs. (3)-(6), unlike the first-order CST for which ideal performance (i.e., stationary cursor) requires hand movements that depend only on the cursor position and are independent of task difficulty (x[n] = – y[n]), ideal performance at higher-order versions of the CST changes with task difficulty (parameter “a”), and moreover requires additional sensory information about the cursor beyond just its position. In particular, a 2nd order version of CST requires information about cursor position and velocity (Eq. (5)). Consistent with the theoretical predictions stemming from Eqs. (3)-(6), we found empirically that our second-order implementation of the CST was indeed more difficult for a subject to control than was the first-order CST, as evidenced by the much smaller parameter values for which the subject was successful (Fig. 4). Following derivations analogous to the 2nd order case, one can readily show that information about cursor position, velocity and acceleration would be necessary for ideal performance at a 3rd order CST, and 4th order would require that information plus the change in acceleration (i.e., “jerk”). Hence, higher-order versions of the CST become increasingly more demanding on the sensory feedback necessary to succeed at the task, and the degree to which this sensory information is combined to generate a motor action changes with task difficulty. These results provide some guidance on the possible forms of the transfer functions Hm and Hs in Fig. 2. In particular, if a subject is able to control an Nth-order CST, then the transfer functions must provide accurate estimates about the cursor states required for control (e.g., position, velocity, and acceleration for a 3rd-order CST). The neural network simulations provided a first step toward modeling sensory-motor control during CST performance. Utilizing

behavioral data from a rhesus monkey, we trained a neural net to perform the first-order CST. Interestingly, while in theory the firstorder CST requires sensory information only about the position of the cursor, best performance at the first-order CST by our neural network was achieved utilizing information about the cursor position and velocity, as well as the previous hand position; i.e., the neural network generated the next hand position as

x[n+1] = f{y[n],y[n-1],x[n]} (7)

Intriguingly, the trained network was unable to perform the second-order CST; our human subject also found the second-order CST much more difficult to perform, requiring additional practice even after having become adept at the first-order CST. Future research will explore what sensory information subjects use to perform the CST, and if/how this changes with task difficulty and the order of the system.

5. Acknowledgements

Funding provided by the University of Pittsburgh Swanson School of Engineering, and NIH R01 HD090125.

6. References

[1] Quick KM, Mischel JL, Loughlin PJ, Batista AP. ``The critical stability task: quantifying sensory motor control during ongoing movement in nonhuman primates.’’ J Neurophysiol 120 5 164 2181 2018 [2] Jex H, McDonnell J, Phatak A. ``A “critical” tracking task for manual control research.’’ IEEE Trans Hum Factors Electron HFE 7: 138 145, 1966. doi : 10.1109/THFE.1966.232660. [3] Potvin AR, Doerr JA, Estes JT, Tourtellotte WW. Portable clinical tracking-task instru- ment. Med & Bological Eng & Comput 15: 391–7, 1977. [4] Kondraske GV, Potvin AR, Tourtellotte WW, Syndulko K. A computer-based system for automated quantitation of neurologic function. Biomed Eng IEEE Transactions on 00: 401–414, 1984. [5] Burke MW, Gilson RD, Jagacinski RJ. ``Multi-modal information processing for visual work- load relief,’’ Ergonomics 23: 961–975, 1980. [6] Ramaekers JG, Kauert G, van Ruitenbeek P, Theunissen EL, Schneider E, Moeller MR. High-potency marijuana impairs executive function and inhibitory motor control. Neuropsychopharmacol 31: 2296–303, 2006. [7] Kadkade PP, Benda BJ, Schmidt PB, Wall C. ``Vibrotactile display coding for a balance prosthesis.’’ IEEE Transactions on Neural Syst Rehabil Eng 11: 392–9, 2003.


Feature validation and online visualization of forearm high-density EMG in an individual with spinal cord injury J. Sebastian Correaa, Jordyn E. Tinga, Devapratim Sarmab, Douglas J. Webera, b Department of Bioengineering, University of Pittsburgh Department of Physical Medicine and Rehabilitation, University of Pittsburgh a


Sebastian Correa is a bioengineering and Spanish student from Pittsburgh, PA. After graduating, he aspires to combine his passions for neural engineering and improving global health through his future career. Sebastian Correa

Douglas Weber, Ph.D. is an Associate Professor in the Department of Bioengineering, and he holds secondary appointments in Physical Medicine and Rehabilitation and Electrical Engineering. He is also a faculty member in the Center for the Neural Basis of Cognition, the Douglas Weber, PhD University of Pittsburgh Brain Institute, and the McGowan Institute for Regenerative Medicine. He established the Rehab Neural Engineering Lab in 2005 when he joined the University of Pittsburgh.

Significance Statement

Myoelectric signals can be recorded from the clinically paralyzed muscles of individuals who have been affected by spinal cord injury. With the optimization of signal processing, there is the potential to significantly improve the quality-of-life of patients by allowing them to control assistive devices through the use of these myoelectric signals.

Category: Experimental research

Keywords: High-density electromyography, signal feature extraction, spinal cord injury

30 Undergraduate Research at the Swanson School of Engineering


Spinal cord injury (SCI) results in damage to the corticospinal tract, weakening electrically active muscles that generate functional movements. The weak myoelectric signals produced by paretic (weakened) muscles due to SCI can be recorded through high-density electromyography (HDEMG) and used to understand pathological changes related to the injury. A custom HDEMG sleeve was used to measure electromyographic (EMG) signals from the forearms of participants with tetraplegia. Recorded EMG signals were filtered and processed to produce a set of signal features, including the root-mean-square, zero-crossings, and power. These features were used to quantify the strength of activation in forearm muscles which can allow us to discriminate activity patterns associated with different hand movements. The purpose of this study is to optimize the method of processing HDEMG signals with the future goal of enabling people with SCI to intuitively control assistive devices using EMG signals from their clinically paralyzed muscles.

1. Introduction 1.1 Motivation Every year about 18,000 people in the United States are directly affected by a spinal cord injury (SCI). This amounts to nearly 300,000 people living with spinal cord injuries in the U.S. [1]. This traumatic injury can cause life-long damage to several areas of the body, including motor and sensory impairments. One of the main components in the spinal cord that is responsible for the movement of the limbs is the corticospinal tract. Damage to the corticospinal tract can leave patients with paralysis. Paralysis in all four limbs is known as tetraplegia. As a result, affected individuals are often unable to independently perform activities of daily living and are therefore reliant on caretakers. In this study, we aim to help restore functional movements to individuals affected by SCI through the use of EMG-controlled assistive devices. 1.2 Using Electromyography to Classify Hand Movements Damage to the corticospinal tract weakens electrically active muscles responsible for performing functional movements. Impaired muscle fibers discharge spontaneously, or not at all, due to the pathological deficits found in the spinal cord. This hindered signal is what causes muscles to become paralyzed. The electrical potentials generated by muscle fibers, known as myoelectric signals, can be measured using electrodes placed at the surface of the skin. This method is known as surface electromyography (EMG) and is commonly used in clinical applications. High-density EMG (HDEMG) electrode arrays have been developed to measure these signals with a higher spatial resolution than traditional EMG devices. These systems use a large number of tightly spaced electrodes to provide a highdensity coverage across the surface of the limb. This allows for a more accurate assessment of deficits in people affected by neuromuscular disorders such as SCI. Through the processing of these myoelectric signals, there is the potential for the use of this technology to accurately control

Ingenium 2020

assistive devices. While the prospect of giving patients who have been paralyzed a more comfortable life is exciting, there are many obstacles to overcome. EMG signals found in the paretic (weakened) muscles of individuals with SCI are often abnormal and therefore hard to use for the discrimination of activation patterns across hand movements. To overcome this, using our custom HDEMG electrode sleeve array (Fig. 1A), we analyzed EMG signals found in the forearm of a patient with SCI for the accurate classification of hand movements. After processing the data through the use of digital filters, we extracted signal features used to evaluate the strength of the signals and the discriminability between movements. Features used in this analysis were the rootmean-square (RMS), zero-crossings, and power. We then analyzed the features through the use of the signal-to-noise ratio (SNR) and principal component analysis (PCA). This will provide insight into how well our processing methods work for the potential use in neuro-prosthetics. We hope to learn from the results of this study to help improve our method of signal processing, and to eventually help enhance the lives of people affected by SCI.

2. Methods 2.1 Experimental Setup Our HDEMG electrode sleeve array (Battelle Memorial Institute, Columbus, Ohio) is comprised of 150 monopolar electrodes made from 12 mm diameter steel discs as electrodes with approximately 15 mm inter-electrode spacing (Fig. 1A). Electrodes on the sleeve span from the wrist to the elbow joint (Fig. 1B). This allows for the measurement of signals across the flexor and extensor muscles of the forearm. The participant with SCI was a 28-year-old male with a C5 motor and C6/C5 (left/right) American Spinal Cord Injury Association Impairment Scale A spinal cord injury. The injury was sustained 9 years prior to the experiment. The able-bodied participant was a 20-year-old male. Only two participants were used for this analysis because only one individual with a SCI was available for testing. All data used in this analysis was recorded in one session for each participant. The experiment consisted of participants being cued to do several hand movements while wearing the HDEMG electrode sleeve array. Before sessions, the forearm of the participant was cleaned before a conductive gel was applied. During trials, participants were shown one of four different hand movements to do before being indicated to do the movement via a visual cue. The four hand movements used in this analysis include the cylindrical grasp, tripod grasp, lateral grasp, and hand open (Fig. 1C). Movement tasks were chosen because of their common use in daily activities. Movement order was random throughout trials.

Figure 1: HDEMG Sleeve and Movement Task. The HDEMG electrode sleeve uses 150 monopolar channels to measure EMG signals from muscles in the forearm. A. The electrodes in the array are 12 mm diameter steel discs with 15 mm inter-electrode spacing. B. The sleeve spans from the wrist to the elbow joint to measure signals from flexor and extensor muscles across the forearm. C. Hand grasps featured in this paper were chosen because of their common use in daily activities.

2.2 Electromyography Processing Raw data was collected using an Intan Technologies RHD2000 Recording System (Intan Technologies, Los Angeles, CA). Signals were sampled at a 10 kHz rate across 150 monopolar channels. Monopolar signals at adjacent electrodes were then differenced to remove common noise to create 135 bipolar channels after removing the most proximal row of electrodes. Bipolar signals were then filtered using a 4th-order digital Butterworth filter. Cutoff frequencies of the band-pass filter were varied to analyze the effects of different frequencies on the discriminability of hand movements. Frequencies tested were between 10-100 Hz and 410-500 Hz for the low and high cutoffs, respectively. This range was chosen in accordance with cutoff frequencies commonly used in analyzing surface EMG data [2]. Data was processed using MATLAB (Mathworks, Natick, MA).

Legend Noise/Rest Region Signal/Active Region

Figure 2: Data Evolution Data shown comes from one channel of a single trial recorded from both an able-bodied participant and a participant with SCI, respectively. Raw data recorded from the HDEMG sleeve was first differenced to remove noise across common electrodes to create bipolar signals. Bipolar data was then bandpass filtered using a 4th-order Butterworth filter. Data shown in this figure was filtered using 30 and 450 Hz cutoffs. Features were then extracted using filtered data. The signal-tonoise ratio was calculated using data from the rest and active periods.


2.3 Feature Extraction and Calculation of the Signal-to-Noise Ratio Signal features can be used to extract useful and dismiss unwanted information found in EMG signals. The selection of features is very important for this to be successful. Features used in this analysis were chosen because of their popularity in EMG signal analysis, their easy implementation, and their range among the time and frequency domains of signal features [3]. Root-Mean-Square (RMS): The root-mean-square (RMS) is a commonly used signal feature found in the time domain. The RMS can be expressed mathematically as [3]:

Zero-Crossings (ZC): The zero-crossings is defined as the number of times the signal passes the zero amplitude axis [4]. It measures frequency information defined in the time domain and it also a commonly used signal feature [3]. The zero-crossings can be expressed mathematically as: Power: The power was calculated through the use of the power spectral density and is a feature in the frequency domain [3]. The power can be expressed mathematically as:

Signal features can be visualized in Fig. 3 where they were calculated from one filtered channel of able-bodied and SCI data. Features were calculated using 200 ms overlapping bins with 50% overlap between bins. The SNR of each feature was then calculated across all 135 channels. The SNR is a measurement used to compare the average level of the desired signal to the unwanted signal, also known as the noise. This calculation can also be visualized in Fig. 3 and 4, with the data shown in the black shaded region used as noise/rest data while data in the red shaded regions was used as signal/active data. This value can be used to assess signal quality and strength.

2.4 Principal Component Analysis Principal component analysis (PCA) is a statistical procedure used to reduce the dimensionality of a large set of data. It uses eigenvector math to transform the original variables to new variables known as principal components. PCA is commonly used in pattern recognition and EMG signal applications [5]. We used PCA to analyze how well EMG signals could differentiate hand movements. The analysis was performed using a matrix of signal feature data points across all 135 channels. All three features were used in this analysis; however, only RMS data is shown.

3. Results 3.1 Differentiation of Hand Movements PCA can be used to visualize the differentiation of hand movements resulting from data calculated from signal features. As shown in Fig. 5, there is a clear distinction in the relative spacing amongst each hand movement between the two patients. When analyzing the able-bodied data, movement regions are more defined when compared to the SCI graph. In addition, the variance in the scale of the able-bodied plot is also much larger than in the SCI plot (Fig. 5). These results can also indicate the accuracy of the classification of hand movements for the two groups of data.

Figure 4: Principal Component Analysis Principal component analysis was used to analyze the discriminability between the activity recorded during different hand movements. A matrix of RMS data was input from all 135 channels. Results show RMS feature is much more differentiable in able-bodied data when compared to SCI data.

Legend Noise/Rest Region Signal/Active Region

Figure 3: Feature Extraction Data shown comes from one channel of a single trial of recording with an able-bodied participant and a participant with SCI, respectively. Signal features used in this analysis were the root-mean-square (RMS), zero-crossings, and power. Data shown in this figure was filtered using 30 and 450 Hz cutoffs.

32 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

3.2 Signal-to-Noise Ratio in Able-bodied and Spinal Cord Injury Participants There is a dramatic decrease in the SNR values of SCI to able-bodied participants. This trend is true across all four hand movements. The decrease in SNR was most notable in the tripod grasp, where the SCI SNR had a 118.9% difference to that of the able-bodied SNR. The trend is least notable in the hand open movement although the SCI SNR still only had an 83.1% difference to that of the able-bodied SNR. The two participants tested also had different movements yield the lowest SNR. For the participant with SCI the lowest SNR came from the tripod grasp, while the hand open movement provided the lowest SNR for the able-bodied participant (Table 1). Movement



Percent Difference (%)













Hand Open




Table 1: Calculated Signal-to-Noise Ratio Across Hand Movements The SNR can be calculated to assess the quality of data. Data shown in this table was filtered using 30 and 420 Hz cutoffs. A low SNR (≈1) means the data overall is noisy or the active data is not very differentiable from rest, while a high SNR (>>1) means the active data is much higher in amplitude than the corresponding rest data. When comparing the average percent difference in SNR of the two patients, signals were approximately 102.1% stronger in the able-bodied participant than the SCI participant.

4. Discussion

The poor results of the PCA and large difference in the SNR of the SCI data compared to the able-bodied data are due to the damage in the corticospinal tract after SCI. The clinically paralyzed muscle fibers are unable to produce myoelectric signals of significant strength, resulting in poor quality of data. The poor results of the PCA can also be due to the small level of control the participant with SCI has over his hand. The participant has a very limited ability to move his fingers, resulting in no displacement when attempting to do hand movements. As the patient is therefore effectively doing the same hand motion for each different movement, it further contributes to nearly no differentiation of hand movements in the SCI EMG data. Another factor adding to the difference in the quality of data from the individual with SCI is the different levels of fine finger control and strength needed to conduct each hand movement. For example, the tripod grasp, requires a higher degree of fine finger control, which is very difficult for the participant with SCI. In another movement, for example the hand open, the grasp requires more hand and finger strength rather than control. The individual with SCI did have relatively good strength of wrist extensor muscles. These facts most likely contributed to the varying differences in the SNR trends of the two participants.

5. Conclusion

Myoelectric signals can be recorded from participants with clinically paralyzed muscles due to SCI. After processing and analyzing signal features, the discrimination of hand movements proved to be very difficult in signals recorded from a participant with SCI when compared to able-bodied signals. Consequently, the features used in this analysis proved to be insufficient for the control of neuro-prosthetics. As our future goal continues to be to enable people with SCI to use EMG signals from affected muscles to control assistive devices, we hope to continue to refine our method of processing and filtering data. Future work will include testing other signals features so that we are able to discriminate hand movements with a higher accuracy. Accordingly, we hope to implement this method to allow for the online classification of hand movements, which will be a big step in helping improve the lives of people affected by SCI.

6. Acknowledgements

The author would like to thank Dr. Douglas Weber, the Swanson School of Engineering, and the Office of the Provost of the University of Pittsburgh for their contributions.

7. References

[1] Spinal Cord Injury Facts and Figures at a Glance. National Spinal Cord Injury Statistical Center. (2018). [2] Ting, Jordyn, et al. “A Wearable Neural Interface for Detecting and Decoding Attempted Hand Movements in a Person with Tetraplegia.” 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2019, doi:10.1109/embc.2019.8856483. [3] Phinyomark, Angkoon, et al. “Feature Reduction and Selection for EMG Signal Classification.” Expert Systems with Applications, vol. 39, no. 8, 2012, pp. 7420–7431., doi:10.1016/j. eswa.2012.01.102. [4] Zardoshti-Kermani, M., et al. “EMG Feature Evaluation for Movement Control of Upper Extremity Prostheses.” IEEE Transactions on Rehabilitation Engineering, vol. 3, no. 4, 1995, pp. 324–333., doi:10.1109/86.481972. [5] Güler, Nihal Fatma, and Sabri Koçer. “Classification of EMG Signals Using PCA and FFT.” Journal of Medical Systems, vol. 29, no. 3, 2005, pp. 241–250., doi:10.1007/s10916-005-5184-7.


Tractography reveals patterns of hippocampal innervation in the human temporal lobe Lauren Gricec, Chandler Fountainb, and Michel Modoa-c McGowan Institute for Regenerative Medicine, bDepartment of Radiology, cDepartment of Bioengineering, dUniversity of Pittsburgh Medical Center, University of Pittsburgh, Pittsburgh, PA, USA Regenerative Imaging Laboratory, McGowan Institute for Regenerative Medicine, University of Pittsburgh, PA, USA a

Lauren Grice

Lauren Grice is a senior Bioengineering major with minors in Electrical Engineering and Neuroscience. She is motivated by understanding the merger between healthcare and technology. After graduation, plans to attend graduate school to study the development of brain computer interfaces for the treatment of neurological disease.


Until recently, medical imaging technology has not been sophisticated enough to develop maps of neuronal connections within the human brain. Such information would be immensely useful as it could help to identify the causes and prognoses of neurological diseases. Diffusion tensor imaging (DTI) is a form of magnetic resonance imaging that can detect the water diffusion vectors in an image. Using tractography, said vectors can be computationally “traced” to create visual representations of neuronal connections. Although DTI has previously been used to identify white matter tracks in the human brain, little research has shown neuronal connectivity within and between grey matter regions. This study applies DTI and tractography to develop a map of hippocampal connections within a healthy human temporal lobe. The results show that the healthy, human hippocampus innervates with two main grey matter structures: the amygdala and the inferior temporal gyrus. This is a significant finding because in the case of disease, changes in the interconnectivity of the hippocampus with the amygdala and inferior temporal gyrus could be an indication of disease onset or prognosis.

1. Introduction Dr. Michel Modo holds a PhD in Neuroscience from King’s College London. He joined the Department of Radiology at the University of Pittsburgh in 2011 and his laboratory’s main focus is to develop novel imaging tools to visualize brain repair using stem cells and biomaterials. Dr. Michel Modo

Significance Statement

The temporal lobe is an important brain structure affected by Alzheimer’s disease and epilepsy. This study seeks to establish maps of neuronal connections between grey matter regions in the temporal lobe. Importantly, it is shown that diffusion tensor imaging can effectively characterize grey matter connections in the brain.

Category: Experimental research

Keywords: temporal lobe, diffusion tensor imaging, tractography

34 Undergraduate Research at the Swanson School of Engineering

Historically, research on neurological disease has focused on modeling whole brain structures to identify abnormalities. However, because of technological limitations, little research has been performed to establish a map of white matter connections between grey matter structures in the human brain. When considering connectivity within the brain, the temporal lobe (TL) region is of particular interest as it contains important grey matter structures like the hippocampus and amygdala which play significant roles in memory and processing of emotions. Additionally, the TL is targeted by debilitating conditions like epilepsy and Alzheimer’s disease. In an effort to better understand the function of the TL, this project addresses the need for visual maps of microstructural anatomy in a healthy, human TL to develop accurate representations of hippocampal connections and their patterns of cortical innervation. In this study, mesoscale, T2-weighted, diffusion tensor imaging (DTI) was used to quantify changes in neuronal connectivity. DTI is a specialized form of magnetic resonance imaging (MRI) that is sensitive to the magnitude and orientation of water movement [1]. Software can be used to detect the water diffusion vectors in an image to calculate both the mean diffusivity (MD) and fractional anisotropy (FA), or the directional dependence of diffusion in tissue [2]. Furthermore, using computational methods, the vectors in each voxel, or unit, of a DT image can be weighted by FA and “traced” to create streamlines, or three-dimensional representations of neuronal bundles. In healthy brain tissue, the flow of water molecules in white and grey matter demonstrates low diffusion and tends to be directionally-dependent, or anisotropic. Because of this diffusion behavior, constructed streamlines in DTIs of healthy brains should demonstrate high microstructural organization. However, in conditions like epilepsy and Alzheimer’s disease, brain cells die and

Ingenium 2020

axons degenerate, resulting in altered diffusion behavior and in turn altered streamline organization. Therefore, the knowledge of normal neuronal connections in the TL produced by this study will serve as a valuable point of comparison for future research on underlying structural abnormalities in cases of neurological disease.

2. Methods

The images analyzed in this study were from a whole, postmortem, left human TL. Diffusion weighted MR data were acquired on a 9.6T Bruker BioSpin MRI (DTI EPI; TR = 500 ms; TE = 0.384 ms; b-values = 1000, 2000, 4000, 8000 s/mm2; number of directions per shell = 40, slice thickness = 0.5 mm, FOV = 62.7 mm x 3.8 mm x 97.9 mm; Matrix = 128 x 128 x 196; total number of voxels = 1,218,380; Resolution = 0.01256 mm3/voxel). Using DSI Studio, a tractography software, three-dimensional maps of anatomical regions of interest (ROIs) were manually drawn on the MR images showing MD. ROIs delineated grey matter structures of the hippocampus, subiculum, amygdala, caudate, putamen, and TL gyri (i.e. the temporal pole, planum polare, superior temporal gyrus, middle temporal gyrus, inferior temporal gyrus, Heschl’s gyrus, fusiform gyrus, parahippocampal gyrus, planum temporale, transverse gyrus, occipito-temporal gyrus, and lingual gyrus). The ROIs were drawn by selecting appropriate voxels on each slice of the DT image. Then, the selected voxels for each ROI were compiled to create a 3-D representation from which a corresponding FA value and streamline count were extracted. Deterministic tractography was performed to trace streamlines within and extending from the hippocampus. Tractographic reconstruction parameters included a FA threshold of 0.02, angular threshold of 60°, step size of 0.25 mm (half the size of a voxel edge), minimum length of 1 mm (length of a connection between at least two voxels), and a maximum length of 60 mm (sample height). All seed orientations were considered with sub voxel seed positions and randomized sampling. All fibers passing through, originating, or ending within the seed ROI, or hippocampus, were considered. Streamline direction interpolation was determined using a Euler tracking algorithm terminated after 12,183,800 random seeds for the whole sample (10x the number voxels for total sample). Streamlines were overlaid on an MD map to provide an anatomical reference for their location.

3. Results 3.1 Anatomical verses diffusion images Figure 1 shows three anatomical views of the acquired image from the temporal lobe. The MD images are shown in the left-hand column and highlight a differentiation between white and grey matter, with grey matter regions delineated by a lighter grey color. The right column shows the corresponding diffusion encoded color (DEC) images. These images show a similar delineation between white and grey matter, possibly indicating a discrepancy in diffusion anisotropy between the two tissue subtypes. It is of note the central area of the lobe is shown as mostly blue or green, representing uniform diffusion direction. On the other hand, the cortex and hippocampus regions are multicolored, suggesting isotropic diffusion behavior occurred within the grey matter regions.

Figure 1: The left column shows a coronal, sagittal, and transverse view of the temporal lobe considered in this study. In these images the light grey regions represent areas of higher average diffusion and likely represent grey matter structures such as the hippocampus and cortex. The right column shows the diffusion encoded color images that correspond to the same slices shown in the MD images. Colors in these images group neurons with the same directional orientation.


3.2 Visualization of whole TL streamlines and ROIs Figure 2A shows a coronal slice of the MD image compared to the same image overlaid with all reconstructed streamlines within it. When comparing the streamlines shown in Figure 2A and the ROIs in Figure 2B, it can be determined that streamlines seem to be less dense in grey matter regions than in white matter regions. Additionally, it is shown that white matter regions appear to be mostly arranged in in the medial/lateral orientation.

In Figure 3B at the tail, or fimbria, of the posterior HC, axonal projections join the selenium of the corpus callosum (scc) and a telencephalic white matter tract, the cingulum bundle (cgb). This result verifies the previously known connection of the HC to the cgb and supports the fact that DTI can accurately contribute to the discovery of grey matter interconnectivity within the TL [3].

Figure 3: (A) Transverse view of one of the TL MD maps with the hippocampus delineated by a 3D ROI. (B) Result of streamline reconstruction by computational detection and tracing of the FA vector in each voxel in the HC ROI. Streamlines within, extending from, and ending in the HC are shown. Streamline colors represent its 3D direction in space, with red=lateral/medial, green=superior/ inferior, and blue=anterior/posterior. Intermediate colors represent a combination of two directions. Tractography reveals four bundles of neurons extending from the HC. At the anterior HC, streamlines connect to the ITG and AG while at the posterior HC, streamlines join the cingulum bundle (cgb) and selenium of the corpus callosum (scc). (C) Isolated streamlines seeded (beginning) in the ITC ROI and ending in the HC ROI. Figure 2: (A) Coronal view of one of the TL MD maps compared to the same image showing all generated streamlines from reconstruction by computational detection and tracing of the FA vector in each voxel in the image. Streamline colors represent its 3D direction in space, with red=lateral/ medial, green=superior/inferior, and blue=anterior/posterior. Intermediate colors represent a combination of two directions. (B) Coronal and transverse views showing the three-dimensional region of interest maps drawn on the DTI image. Here the superior, medial, and inferior temporal gyri are shown (labeled as STG, MTG, and ITG), as well as the parahippocampal gyrus (PHG), subiculum (SB), and hippocampus (HC).

3.3 Hippocampal tractography Figure 3B illustrates extra- and intrahippocampal streamlines were successfully reconstructed from the hippocampus ROI shown in Figure 3A. Furthermore, this approach demonstrates for the first time that visual representations of both white and grey matter connectivity in the TL can be probed using tractography. The visualization of neuronal connections in Figure 3B and 3C show a precise location for streamline termination as well as provide a qualitative understanding of their arrangement. In Figure 3B, at the anterior end of the HC, efferent neurons connect to two grey matter structures: the AG and ITG. Connections to the AG are few in number, but highly organized. Streamlines extending to the ITG terminate on both medial and lateral edges of an anterior segment of the gyrus. Conversely, as shown in Figure 3C, efferent connections from the ITG appear to fan out within the HC.

36 Undergraduate Research at the Swanson School of Engineering

4. Discussion

The MD images in Figure 1 support the fact that grey matter has less structural organization than white matter and therefore has higher average diffusion [2]. As a result of this increased diffusion, the grey matter regions show up brighter on the MD images in Figure 1. This result validates that diffusion tensor imaging can effectively delineate grey matter regions and can identify the diffusion directions within them. This is further supported by the DEC images in which the central white matter regions are shown in uniform colors suggesting neurons within the white matter branches have similar directional orientation; the cortex and hippocampus on the other hand are shown in multiple colors suggesting there exists a lesser degree of directional orientation of diffusion. The tractography image in Figure 2A may show a lack of streamline density within grey matter regions because of the limitations of diffusion tensor imaging to differentiate crossing or “kissing� fibers. In other words, although DTI can still identify neuronal bundles in grey matter, it may be limited in its ability to visualize the full complexity of fiber organization within these regions. However, the results in Figure 3 show that tractography was able to reveal that the main grey matter structures the healthy adult HC in this study innervates with are the AG and ITG. This is a significant finding as further identifying isolated

Ingenium 2020

connections between two or more grey matter regions could reveal neuronal circuits that have not been able to be detected before the use of DTI. Furthermore, previous studies have shown that functions such as perception, imagination, and episodic memory engage the anterior HC [4]. Since the HC plays a major role in memory formation and has been shown to atrophy in cases of Alzheimer’s disease and dementia, it is possible that a change in the interconnectivity of the HC with the AG and ITG could be an indication of disease prognosis [5]. Furthermore, previous studies have theorized that loss of connections precedes atrophy in Huntington’s disease. If this degeneration paradigm is consistent in Alzheimer’s disease, it may be possible to use DTI to identify the onset of dementia before the patient begins to show severe symptoms [6].

5. Conclusion

The methods in this study revealed that diffusion tensor imaging, although mainly used to identify white matter tracks in previous studies, can successfully reconstruct streamlines within grey matter regions as well as reveal patterns of grey matter region interconnectivity. Specifically, the results of this study showed that the main grey matter structures a healthy adult HC innervates with are the AG and ITG. Chiefly, though, the significance of reconstructing streamlines from DTIs in this study was to gain an accurate, comprehensive, and novel characterization of tissue microstructure within the TL and, specifically, the HC. Importantly, reconstructed streamlines from this study will later help to characterize the pathological basis of diseases that manifest in the TL, like Alzheimer’s disease.

7. References

[1] Mori & Zhang. “Principles of diffusion tensor imaging and its application to basic neuroscience research.” Neuron 51(5), 527539, 2006, doi: 10.1016/j.neuron.2006.08.012. [2] C.H. Sotak. “The role of diffusion tensor imaging in the evaluation of ischemic brain injury – a review.” NMR in Biomedicine 15(7-8), 561-569, 2002, doi: 10.1002/nbm.786. [3] Wu et al. “Segmentation of the Cingulum Bundle in the Human Brain: A New Perspective Based on DSI Tractography and Fiber Dissection Study.” Frontiers in Neuroanatomy 10, 10-84, 2016, doi: 10.3389/fnana.2016.00084. [4] Zeidman & Maguire. “Anterior hippocampus: the anatomy of perception, imagination and episodic memory.” Nature Reviews Neuroscience 17(3), 173–182, 2016, doi: 10.1038/nrn.2015.24. [5] Mu & Gage. “Adult hippocampal neurogenesis and its role in Alzheimer’s disease.” Molecular Neurodegeneration, 6, 6-85, 2011, doi: 10.1186/1750-1326-6-85. [6] McColan et al. ”Brain Regions Showing White Matter Loss in Huntington’s Disease Are Enriched for Synaptic and Metabolic Genes” Biological Psychiatry 83(5), 456-465, 2018, doi: 10.1016/j.biopsych.2017.10.019.

6. Acknowledgements

Lauren Grice was supported by the University of Pittsburgh Swanson School of Engineering and Office of the Provost throughout the entirety of this research.


Numerically resolved radiation view factors within thermoelectric generators via hybridized CPU-GPU computing Asher J. Hancocka, Matthew M. Barry, PhDa Department of Mechanical Engineering and Materials Science University of Pittsburgh, PA, USA a

Asher Hancock

Matthew Barry

Asher is a junior studying mechanical engineering with a minor in computer science. His research interests include multi-physics modeling, high-performance computing, and algorithm design for space applications. While not in the lab, Asher enjoys playing his violin and serving as Pitt Aero’s lead wing and aerodynamics engineer. Dr. Barry’s research focuses on multiphysics modeling of energy systems. This ranges from terrestrial thermal-fluidelectric coupled modeling of waste-heat recovery systems to thermal-electricmechanical coupled modeling of space power-generation systems, and includes phase-change modeling for extraterrestrial probe design and evaluation.

Significance Statement

Modeling the proportion of diffuse radiation exchanged between interacting surfaces, known as the radiation view factor, often proves too computationally intensive for numerical analysis. By creating a robust computational framework for resolving the view factors via GPU-accelerated programming, we improve the ability to achieve accurate radiation heat transfer models.

Category: Computational research

Keywords: thermoelectric generators, radiation view factors, GPU-accelerated programming


Thermoelectric generators (TEGs) are solid-state power generation devices that can convert thermal energy directly into electrical energy via an applied temperature gradient. Operated under large temperature differences to achieve high thermal efficiency, thermal radiation becomes increasingly dominant in heat transfer, requiring accurate characterization to correctly determine TEG performance. Numerical modeling is often employed to analyze a TEG’s radiative heat transfer. However, due to the computationally intensive process of characterizing the proportion of diffuse radiation exchanged between two surfaces, known as the radiation view factor, analysis is limited. In this work, a robust computational framework is developed and implemented to numerically resolve the radiation view factors within three-dimensional TEGs. The proposed numerical methodology utilizes a hybridized CPU-GPU ray-tracing algorithm to capitalize on the parallel nature of the view factor formulation, allowing for reduced computational runtimes in comparison to linear programming methods. The shadow effect, resulting from interference with the TEGs internal thermoelectric material and conductive interconnectors, is accounted for via the MöllerTrumbore ray-triangle intersection algorithm. To investigate the relationship between view factors and TEG geometry, the height to width ratios between the thermoelectric components and the thicknesses of the interconnectors were varied. Results indicate that view factor values decrease for increases in both height to width ratios and interconnector thicknesses. As grid-sizing decreases, residuals between corresponding runs also decrease, showcasing solution convergence.

1. Introduction 1.1 Thermoelectric Generators Thermoelectric generators (TEGs) are solid-state power generation devices constructed of thermoelectric (TE) modules that can convert thermal energy directly into electrical energy via an applied temperature gradient. These TE modules are connected electrically in series by highly conductive interconnectors, which operate thermally in parallel, between a heat source and a heat sink, to develop an output voltage. These devices show tremendous potential in power generation for space exploration and waste heat recovery [1]. The performance of a TEG is typically characterized by, ZT, the product of the figure of merit for thermoelectric materials and the average temperature gradient across the heat exchangers. Despite the nonlinear relationship of Z with temperature, TEGs operate under large temperature differences to achieve high thermal efficiency values [2]. Under these conditions, radiative heat transfer can dominate over conduction and convection, and thus must be properly analyzed in any TEG design. 1.2 Numerical Modeling of Radiative Heat Transfer Traditional methods of view factor calculation typically involve either deriving complex analytical solutions or utilizing computational tools, like the Monte Carlo method. Analytical

38 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

solutions become increasing difficult to formulate for complex geometries, such as TEGs, and are not commonly used. Computational methods, when restricted to the computer processing unit (CPU), require either small domains or supercomputer usage to achieve adequate runtimes: both, of which, are impractical in TEG heat transfer analysis. In prior works, Vujičić et al. employed a Monte Carlo method with the finite element technique to calculate the radiation view factor between parallel plates. However, they note that the computational intensity of the Monte Carlo method limited its application in complex models where fine meshes are required [3]. Walker et al. utilized a Monte Carlo ray tracing algorithm to resolve view factors, but linear programming, relying purely on CPU calculations, proved too demanding for non-symmetric geometries [4]. Narayananswamy derived an analytical expression for the view factors of arbitrarily oriented triangles, which is only applicable for primitive triangular surfaces [5]. As shown, there is a need for an efficient method to resolve the view factors for complex configurations.

2. Methods

To numerically resolve the view factors within a TEG, a computational framework, written in Java, was created. The program requires the user to input the following STL files, which are easily exported from CAD software: the surfaces undergoing radiative transfer and any geometry in-between the two surfaces. 2.1 Geometry Definition The view factor is defined as where θ is defined as

To numerically resolve the view factors within a TEG, each surface was tessellated into differential triangular areas based upon geometry and a corresponding desired refinement using a modified simple mesh generator or an in-house triangular sub-division script [6]. Differential triangular areas are utilized to maintain compatibility with the STL definition files. Each differential area of the emitter surface creates a corresponding radiation ray for each receiver differential area, which contributes individually for each view factor calculation (see Figure 1). 2.2 Shadow Effect Within a TEG, not all emitted radiation from one surface will reach the target surface due to the shadow effect (the possible interference of diffuse radiation with the internal thermoelectric legs or interconnectors). To account for the shadow effect, the Möller-Trumbore (MT) triangle-ray intersection algorithm was utilized to check for any interference with the geometry in-between the two participating surfaces [7]. If the MT algorithm detects an intersection, that emitted ray doesn’t contribute to the overall view factor calculation. The MT algorithm is calculated for every ray emitted. Since small refinement factors are necessary for accurate results, finely tessellated geometries are required. When refinements become small enough, determining the view factor between surfaces becomes too computationally intensive since emitted rays grow exponentially as refinement values decrease. Therefore, to effectively resolve many intersection checks for every emitted ray, the MT algorithm is computed on the graphic processing unit (GPU) to make use of the hundreds of cores available. The MT algorithm was desirable to implement due to its intrinsic low memory usage. These computations are parallelized across the device’s GPU via the Aparapi programming library package. 2.3 Numerical Validation Analytical solutions for known view factors were numerically computed for various cases: rectangular parallel plates, coaxial parallel plates, perpendicular plates connected at one edge, differential square area to a perpendicular plate, concentric cylinders, and arbitrarily sized perpendicular plates [8]. Parallel and perpendicular geometries are showcased to demonstrate code robustness. Figure 2 depicts the analytical curves for numerous cases of parallel plates and differential square areas to perpendicular plates. Plotted among the graphs are the calculated view factors for specific configurations. A grid independence study, as shown in Table 1, demonstrates the convergent nature of the code with an increasing number of tessellations for rectangular parallel plates. In comparison to Vujičić et al., the GPU accelerated code can calculate 6.87*1010 rays in 488 seconds whereas Vujičić et al.’s code only reported 3.20*106 rays in 1114 seconds when run on the CPU. Since calculations grow exponentially as refinement values decrease, the GPU-accelerated program becomes increasingly necessary when high accuracy is required.

Figure 1: Example of a Single-Junction TEG with tessellated a tessellated geometry


Figure 2: Geometrical configurations with analytical view factor values. A.) Coaxial Parallel Plates B.) Rectangular Parallel Plates C.) Differential Square Area to a Perpendicular Plate

Tessellations per Surface

Analytic Fij

Numerical Fij

Percent Error














Residuals Between Cases

Time [s] 45














Table 1: Numerical view factor values for parallel plates of X/L=1.00 and Y/L=1.00

2.4 Thermoelectric Generator Modeling Once the code proved valid for a variety of analytical configurations, the geometry of a single-junction TEG was modeled for view factor analysis (see Figure 1). TEG design often seeks to optimize TE material within the entire TEG geometry to improve efficiencies and decrease material costs. Therefore, to investigate the effect of geometry on the view factor calculation, the height to width ratios between the TE components and the thicknesses of the interconnectors were varied. The internal TEG geometry was exported as an STL file from SolidWorks, and the hot and cold sides were created using a modified simple mesh generator [6]. Highly tessellated emitting and receiving files are desirable for accuracy and convergence. The internal geometry, however, is coarsely meshed to lower the number of logical checks computed for triangle-ray intersection. Interconnector thicknesses and height to width (H/W) ratios were varied from 0.50in. to 1.50in. within a single-junction TEG. The packing density, a ratio between area of the semiconductors to the area of the entire TEG, was held constant to explore the

40 Undergraduate Research at the Swanson School of Engineering

relationship between the radiation view factor and the distance between opposing plates within a TEG unit cell. Each emitting and receiving plate was tessellated to a refinement of at least four million triangles for high accuracy.

3. Results

It is demonstrated from Figure 3 that as the H/W ratio increases, the radiation view factor decreases for a given device with a constant packing density. The same trend is demonstrated as the interconnector thickness increases within the device. To further validate the radiation view factor values calculated, the algorithm presented by Barry et al. was programmed and calculated for identical TEG configurations. As shown in Table 2, the percent difference between the values reported by Barry et al. and those calculated in this work approach a tenth of a percent, while the residuals between cases begin to converge on the correct numerical value [9]. The small difference between reported values gives confidence in the program’s accuracy.

Ingenium 2020

Figure 3: View factors vs. interconnector thickness for various H/W ratios within a single-junction TEG

5. Conclusion

4. Discussion

As shown in Figure 3, as the H/W ratio increases, radiation view factors decrease. This is due to the increased shadow effect from the internal TEG geometry disrupting the radiative transfer between the opposing ceramic plates. Similar logic is applied when increasing interconnector thickness. The algorithm originally presented by Barry et al. utilized an intersection algorithm useful for any convex (non-intersecting) polygon [9]. Polygon intersection was determined by checking the location of every edge of the polygon with respect to the ray’s directional vector: if the ray’s directional vector lies on the inside of every edge within the polygon, the algorithm executes. With STL definition files, this leads to various logical checks per tessellation, along with a precomputation of the plane equation containing each polygon. In comparison, the MT algorithm operates via mapping the original intersection coordinates into barycentric coordinates, which alleviates the need to precompute the plane equation for every individual tessellation. In consequence, the MT algorithm requires fewer logical checks, offering decreases in computational runtime. By calculating the view factors for identical configurations using the polygon intersection algorithm and the MT algorithm, the MT algorithm offered up to 13% runtime gains. Figure 4 demonstrates the MT algorithm’s superiority to the polygon intersection algorithm, as the disparity between runtimes grows as tessellations increase.

Figure 4: Computational runtime differences between the Möller-Trumbore intersection algorithm and the polygon intersection algorithm presented in [8]

Hybridized CPU-GPU computing, used in conjunction with the Möller-Trumbore triangle-intersection algorithm, proves successful in characterizing the radiation view factors between the opposing ceramic plates within a single-junction thermoelectric generator. In comparison to traditional computational methods for the radiation view factor, GPU-accelerated programming offers massive speedups in computational runtime. Pending further tests for other TEG configurations, numerically resolving the view factors within thermoelectric devices will prove tractable for radiation heat transfer analysis.

6. Acknowledgements

I’d like to thank Dr. Matthew Barry for his support throughout this project and for the opportunity to work in his lab. I’d also like to thank the Swanson School of Engineering for funding this research.

7. References

[1] D. Champier, “Thermoelectric generators: A review of applications,” Energy Conversion and Management, vol. 140, pp. 167-181, 2017. [2] H. J. Goldsmid, Theory of Thermoelectric Refrigeration and Generation, pp. 9-24 Berlin, Heidelberg: Springer Berlin Heidelberg, 2016. [3] M. Vujičić, N. Lavery, and S Brown, “View factor calculation using the Monte Carlo method and numerical sensitivity,” Communications in Numerical Methods and Engineering, vol. 22, no. 3, pp. 197-203, 2006. [4] T. Walker, S. C. Xue, and G. Barton, “Numerical Determination of Radiative View Factors Using Ray Tracing,” Journal of Heat Transfer, vol. 132, no. 7, p. 072702, 2010. [5] A. Narayananswamy, “An analytic expression for radiation view factor between two arbitrarily oriented planar polygons,” International Journal of Heat and Mass Transfer, vol. 91, pp. 841847, 2015. [6] P. O. Persson and G. Strang, “A Simple Mesh Generator in MATLAB,” SIAM review, vol. 46, no. 2, pp. 329-345, 2004. [7] T. Möller, B. Trumbore, “Fast, minimum storage ray-triangle intersection,” Journal of graphics tools, vol. 2, pp. 21-28, 1997. [8] 4 J. R. Howell, M. P. Menguc, and R. Siegel, Thermal Radiation Heat Transfer. CRC press, 2015. [9] M. Barry, J. Ying, M. J. Durka, C. E. Clifford, B. Reddy, and M. K. Chyu, “Numerical solution of radiation view factors within a thermoelectric device,” Energy, vol. 102, pp. 427-435, 2016.

Tessellations per Surface

Numerical Fij

Barry et al. Reported Fij Percent Difference Between Fij Values





Residuals in Numerical Fij





















Table 2: Comparison of numerical view factor values for a single-junction TEG device with H/W=1.50 and t=0.125[in]


Characterization of redox flow battery kinetics using a flow channel analytical platform Thomas J. Henrya, Tejal V. Sawanta, and James R. McKonea a Department of Chemical & Petroleum Engineering, Swanson School of Engineering, University of Pittsburgh, Pittsburgh, PA, USA

Thomas J. Henry

Thomas J. Henry is a sophomore in the Department of Chemical Engineering at the University of Pittsburgh, expecting to graduate in 2022. He joined the McKone research group in the summer of 2019 to work on the redox flow battery project. His professional interests involve the future of energy and sustainability, which his research is congruent with.

Dr. James R. McKone is an Assistant Professor of Chemical Engineering at the University of Pittsburgh, where he joined the faculty in the fall of 2016. He holds a bachelor’s degree in chemistry and music from Saint Olaf College and a Ph.D. in chemistry from the California Institute of Dr James R. McKone Technology, and he completed additional postdoctoral training at Cornell University. Dr McKone’s research group combines fundamental and applied research in electrochemistry to address long-standing challenges in energy and environmental sustainability.

Significance Statement

Renewable energy is becoming an increasingly integral component of the modern energy supply. As renewables become more prevalent, methods of efficiently storing large amounts of clean energy must be examined. This project seeks to better understand key factors influencing the efficiency of one such energy storage method, the redox flow battery.

Category: Experimental research

Keywords: Renewable energy, flow batteries, kinetics, electroanalysis

42 Undergraduate Research at the Swanson School of Engineering


The redox flow battery (RFB) is a promising industrial technology that offers the ability to store large quantities of renewable electricity at very low cost. However, conflicting reaction kinetics within these systems have been reported, even for battery chemistries that are nearing commercialization. The goal of this project is to construct and test an analytical platform that delivers highly precise measurements of the kinetic characteristics of RFB electrode-electrolyte combinations. To accomplish this, we are using a 3D-printed miniature flow cell that houses a microscopic Pt electrode to analyze the reaction kinetics of an aqueous Febased flow battery electrolyte. Our long-term goal is to integrate this apparatus directly into a working flow battery to characterize kinetics during long-term operation. Building on prior work involving rotating-disk electrode (RDE) techniques, this work focused on concentration-dependent kinetics analysis of the aqueous Fe(III/II) redox couple. Our measurements yielded exchange current densities of 5.6 ± 0.6 mA/cm2 at 10 mM Fe concentration, and 255 ± 3 mA/cm2 at 1000 mM Fe concentration. Thus, the apparent reaction rate was found to increase by a factor of ~45 while the Fe concentration was increased by a factor of 100, suggesting the emergence of kinetic or mass transfer limitation at higher concentrations. These values were also greater than those found using RDE voltammetry, which yielded a 33% slower reaction rate at 10 mM Fe concentration and nearly 80% lower at 1000 mM Fe. This is a valuable result that implicates the possibility that mass-transfer limitations confound measurements of reaction kinetics using both techniques.

1. Introduction

The redox flow battery (RFB) is a promising industrial technology that is capable of providing large-scale electricity storage to support the growth of renewable energy [1, 2, 3]. For example, solar and wind power are clean alternatives to fossil fuel combustion, but they produce inconsistent supplies of energy that vary based on external conditions [4]. To address this issue, RFB technologies can be implemented to store vast amounts of energy that can be released as needed, improving the operational efficiency of the power grid even as more renewables come online. RFBs operate through flow-based redox chemistry by manipulating the oxidation states of soluble electrolyte species [5]. While this is approach is fundamentally similar to a solid-state rechargeable battery, the RFB holds particular potential for its ability to be scaled up at very low cost with the use of large electrolyte storage tanks that enable operation on the same scale as a large powerproduction facility [6]. However, a concern with this technology involves inconsistent prior reports on the energy efficiencies (reaction kinetics) of the redox reactions within flow batteries, which makes further development difficult [7]. For example, previous work in the McKone Lab showed that the reaction kinetics for charge storage in a representative RFB electrolyte based on Fe(III/II) decreased as the concentration of the Fe was increased [8]. This trend is relevant for practical RFBs because very high concentrations are necessary for industrial systems [9].

Ingenium 2020

Our lab is working to develop an electroanalytical platform that consistently returns valid kinetics characteristics for a wide range of electrolyte species used in industrial RFBs. To accomplish this, we are using a 3D-printed, miniature flow cell that houses a microscopic Pt electrode to analyze the reaction kinetics. In this study, we used the same Fe-based RFB electrolyte over a range of concentrations to observe the dependence of concentration on kinetics. We further executed measurements under static and flowing conditions to assess the inter-relationships between transport and kinetics in this system. The results strongly implicate mixed kinetic-transport limitations in the redox chemistry of aqueous Fe salts at Pt electrodes. More broadly, we conclude that this flow cell platform can be used to characterize many common RFB electrolytes and electrode materials and accelerate future RFB device design.

2. Methods 2.1 Experimental Design All reagents were purchased from commercial suppliers and used as received. Deionized (DI) water was purified to >18 MΩ resistivity (Millipore, Milli-Q Advantage A10). Stock solutions of 0.5 M and 2 M HCl(aq) were prepared from 12 M starting material (Certified ACS plus) and used as the supporting electrolyte. The required quantities of FeCl2• 4H2O and FeCl3• 6H2O were obtained from Fisher Scientific and were weighed and added to the stock solutions of 0.5M HCl(aq) to produce 50 mL batches of electrolyte with total Fe concentrations of 10, 20, 40, and 100 mM. Additionally, 100 mM and 1000 mM Fe electrolytes were prepared with the 2 M HCl(aq) stock solution. A commercial platinum microelectrode (Bioanalytical Systems, 10 µm diameter) was polished successively using aqueous alumina slurries (Pace Technologies) of sizes 1, 0.3, and 0.05 µm. This was followed by cleaning in an ultrasonic bath (Branson M1800) for 30 seconds in DI water to remove any alumina residue between each polishing. The electrode was then submerged in DI water for 15 minutes prior to experimentation to encourage the removal of any final impurities on the electrode surface. A graphite rod (1/4 in. diameter) was used as the counter electrode along with a commercial Ag/ AgCl reference electrode (Fisher Scientific). To set up the flow channel, O-rings were attached to the tip of each electrode and threaded plugs were slid onto the electrodes. These plugs were then wrapped with polytetrafluoroethylene (PTFE) tape to provide a tighter seal before screwing them into a 3D-printed flow channel that was designed and constructed in house from Somos 11122XC SLA resin. Two lengths of MasterFlex pump tubing (1.6 mm ID) were connected to the in and out-flow of the channel using plastic fittings. A fully assembled flow apparatus is depicted in Figure 1.

Figure 1: Flow channel experimental setup with components fully installed.

A peristaltic pump (MasterFlex) was used to flow the electrolyte through the cell, and a 100 mL beaker was used as the electrolyte reservoir. The flow cell was cleaned by flowing DI water through the channel for 15 minutes immediately prior to experimentation. A Gamry Interface 1000E digital potentiostat was used to perform the experiments. After water cleaning, a separate 100 mL beaker was filled with Fe electrolyte. The water in the system was drained, and the two open-ends of tubing were transferred to the electrolyte reservoir, and the pump was set to produce a 10 mL/min flow rate. No attempt was made to remove atmospheric air from the electrolyte. Cyclic voltammetry experiments were performed to collect current versus applied potential data. A potential range from 0 V to 0.8 V versus Ag/AgCl was used at a scan rate of 30 mV/s. 2.2 Data Processing Raw data obtained from the potentiostat was translated to current density (mA/cm2) by dividing out the active area of the electrode (7.85x10-7 cm2). A mathematical model was used to fit current density as a function of applied potential via a leastsquared regression to the Butler-Volmer equation for a oneelectron transfer reaction: (1) where j is the observed current density, j0 is the exchange current density, αa and αc are the anodic and cathodic charge transfer coefficients, F is Faraday’s constant (96485 C/mol), η is the applied overpotential (referenced to the equilibrium potential of the reaction), R is the universal gas constant, and T is the temperature. From this fit, the behavior can also be expressed by the heterogeneous electron transfer rate constant, k0:


where C is the electrolyte concentration and the other terms have the same definitions as in Equation (1). Based on the assumptions inherent in the Butler-Volmer model, this metric is independent of electrolyte concentration, so data collection at multiple concentrations should yield identical values of k0.


3. Results 3.1 Kinetics in the Absence of Flow Initial experiments were performed on quiescent (non-flowing) electrolyte as a baseline to assess the effect of introducing flow into the system. Figure 2 depicts representative current density versus potential data for a full set of quiescent experiments in which the current densities have been scaled by electrolyte concentration.

Figure 2: Compiled current versus potential data for an Fe-based RFB electrolyte, normalized by concentration to capture kinetic trends.

When plotted this way, the data capture the reaction kinetics through the slope of the current-potential line near the zero-current point—a steeper slope is a qualitative indication of faster reaction kinetics. Table 1 provides quantitative data in terms of apparent k0 (cm/s) based on least-squares regression to the Butler-Volmer equation. As expected, the reaction rate remained near constant as the concentration of electrolyte was increased; however, a significant decrease was observed when transitioning from 100 mM to 1000 mM Fe concentration. Moreover, we observed a decrease in the maximum current density (relative to electrolyte concentration) for 1000 mM electrolyte, which suggests that increased electrolyte concentration gives rise to mass transfer limitations.

44 Undergraduate Research at the Swanson School of Engineering

3.2 Kinetics Under Convective Flow Figure 3 depicts analogous current density versus potential data for the same set of Fe electrolytes under 10 mL/min of electrolyte flow. This flow condition was chosen to mimic the operating conditions of several recently reported lab-scale RFB devices [10, 11, 12]. These data show a high level of noise, which is attributable to the peristaltic pump that was used to flow the electrolyte. The mode of fluid displacement in this pump produces large pulsations, which constantly changes the rate of electrolyte flow.

Figure 3: Raw current versus potential data for a set of flowing experiments, normalized by electrolyte concentration.

Ingenium 2020

To mitigate noise in the data, we employed a smoothing technique that treats the absolute value maximum current density over a narrow region of potential as the “true” current density. These values coincided with the maximum rate of electrolyte flow produced by the pump. This approach was chosen because it represents the best compromise between residual noise and our ability to resolve kinetics in the system (by maximizing the rate of mass transfer). The resulting smoothed data are depicted in Figure 4, which facilitate kinetics analysis analogous to that in Figure 2. Table 1 also presents quantitative results of the flowing trials against the static trials to view the differences in trends.

electrolyte flow can have on the efficiency of an RFB, especially for fast redox reactions that are generally mass-transfer limited, such as Fe(III/II). The trend of decreasing reaction rate with increasing electrolyte concentration was also more severe under electrolyte flow. In quiescent experiments, the reaction rates remained essentially constant up to 100 mM in Fe concentration. By contrast, under flow a progressive drop in kinetics was observed each time the concentration was increased. Figures 3 and 4 also demonstrate that this decrease in the apparent reaction rate constant coincides with a decrease in steady-state current density (after accounting for the change in concentration). This trend suggests a direct correlation between the concentration of electrolyte and mass transport of active species to the electrode. Thus, it is apparent from these data that the practical rate at which Fe-based RFB electrolytes can be oxidized and reduced at Pt electrodes depends on the interplay between kinetics and transport in the system.

5. Conclusion

Figure 4: Compiled current versus potential data for a series of flow trials, normalized by concentration to show kinetic trends. Data noise produced from the pump was filtered.

[Fe] (mM)

Static k0 (cm/s)

Flowing k0 (cm/s)



2.3 x 10

5.8 x 10-3 ± 6.1 x 10-4


2.7 x 10-3

5.5 x 10-3 ± 4.4 x 10-4


2.8 x 10

6.2 x 10-3 ± 2.2 x 10-4

100 (0.5 M HCl)

2.3 x 10-3

4.3 x 10-3± 1.8 x 10-4

100 (2 M HCl)

2.2 x 10

2.4 x 10-3 ± 1.4 x 10-4

1000 mM

1.5 x 10

2.6 x 10-3 ± 2.9 x 10-5


-3 -3

Table 1: Compiled rate constants for static and flowing trials, expressed in k0 (cm/s).

4. Discussion

The results show that there is an analytically significant increase in the apparent reaction rate when the electrolyte is flowed. The overall trend involved a consistent increase in rate by a factor of ~2. Because the electrolyte and electrode composition are nominally identical in each case, the only plausible explanation for this difference is that the reaction is under mixed kinetic-transport limitation, and the onset of solution convection increases the rate of transport. Hence, this result illustrates the significant effect that

Our flow cell analytical platform consistently returned precise kinetics data from this preliminary set of experiments using the Fe(III/II) redox couple at a micro-Pt electrode. Broadly, the results of the flow channel kinetics characterization agree with prior literature reports in which various methods were used to characterize Fe(III/ II) with platinum electrodes [13,14,15]. Our data further suggest that the flow cell increases mass transport as compared to other commonly used methods of kinetics analysis such as RDE and static microelectrode measurements. While flowing the electrolyte, kinetics values were measured at rates of around two times higher than in static trials or RDE experimentation, which is relevant for fast redox couples that engage in transport-limited reactions [8]. Given the successful results provided by this analytical approach, we expect that the method can be applied to many other electrolytes and electrodes to accelerate future RFB research. Our system was able to accommodate a flowing environment that a real RFB would employ, which will enable us to directly implement this apparatus into the flow loop of an operational RFB for real-time kinetics analysis. Eventually, similar methods could be applied to large industrial RFB systems, where significant challenge remains in developing online metrology tools. Our results further implicate potentially important emergent properties (e.g., solution viscosity, electrolyte speciation, or reaction mechanism) when the concentration of electrolyte is increased to levels that would be used in a real RFB. Platinum was initially used in our experiments for its chemical inertness and fast kinetics toward Fe(III/II) redox chemistry [16]. However, commercial RFBs cannot operate economically with noble metal electrodes, so a more economical material is used: carbon [17]. Thus, future studies will focus on transitioning to microscopic carbon electrodes. Initial work has already shown that the reaction rates of Fe(III/II) at carbon are significantly slower than at Pt; however, the method of characterization remains the same. Ongoing studies are now directed at developing and deploying surface treatments for carbon electrodes to enhance their catalytic 45

performance toward Fe and other RFB electrolytes under realistic operating conditions.

6. Acknowledgements

We gratefully acknowledge project funding and support provided by the Swanson School of Engineering, Oak Ridge Associated Universities, and the Office of the Provost at the University of Pittsburgh.

7. References

[1] Skyllas-Kazacos et. al. “Progress in Flow Battery Research and Development.” J Electrochem Soc. 158(8), 1945-7111, 2011. [2] Johansson et. al. “Renewable Fuels and Electricity for a Growing World Economy: Defining and Achieving the potential.” Energy Studies Review. 4, 201-222, 1992. [3] Dincer, I. “Renewable Energy and Sustainable Development: ACrucial Review.” Renewable Sustainable Energy Rev. 4, 157−175, 2000. [4] Hill et. al. “Battery Energy Storage for Enabling Integration of Distributed Solar Power Generation.” IEEE Transactions on Smart Grid. 3, 850-857, 2012. [5] Weber et al. “Redox flow batteries: a review.” J Appl Electrochem. 41, 1572-8838, 2011. [6] Dunn et. al. “Electrical Energy Storage for the Grid: A Battery of Choices.” Science. 334, 928-935, 2011. [7] Jahn and Vielstich. “Rates of Electrode Processes by the Rotating Disk Method.” J. Electrochem. Soc. 109, 849−852, 1962. [8] Sawant and McKone. “Flow Battery Electroanalysis: Hydrodynamic Voltammetry of Aqueous Fe(III/II) Redox Couples at Polycrystalline Pt and Au.” ACS Appl. Energy Mater. 123(1), 47434753, 2018. [9] Peng and Zawodzinski. “Describing ion exchange membrane-electrolyte interactions for high electrolyte concentrations used in electrochemical reactors.” J. Membrane Science. 592, 0376-7388, 2019. [10] Milshtein et. al. “High current density, long duration cycling of soluble organic active species for non-aqueous redox flow batteries.” Energy and Environmental Science. 11, 35313543, 2016. [11] Barton et. al. “Quantifying the impact of viscosity on mass-transfer coefficients in redox flow batteries.” J. Power Sources, 399, 0378-7753, 2018. [12] Milshtein et. al. “Quantifying Mass Transfer Rates in Redox Flow Batteries.” J. Electrochem Soc.,164(11), 3265-3275, 2017. [13] Angell, D. H.; Dickinson, T. “The Kinetics of the Ferrous/ Ferric and Ferro/Ferricyanide Reac-tions at Platinum and Gold Electrodes: Part I. Kinetics at Bare-Metal Surfaces.” J. Electroanal. Chem. Interfacial Electrochem. 35, 55–72, 1972. [14] Junsuke, S. “Hydrodynamic Voltammetry with the Convection Electrode. IV. The Measure-ments of the Kinetic Parameters of the Electrode Reaction. Part II.” Bull. Chem. Soc. Jpn. 43, 755–758, 1970.

46 Undergraduate Research at the Swanson School of Engineering

[15] Weber et. al. “The Effect of Anion Adsorption on the Kinetics of theFe3+/Fe2+ Reaction on Pt and Au Electrodes in HClO4.” J. Electroanal. Chem. 89, 271–288, 1978. [16] Chen et. al. “Solution Redox Couples for Electrochemical Energy Storage I . Iron (III)-Iron (II) Complexes with O-Phenanthroline and Related Ligands.” J Electrochem Soc., 128(7), 1460-1467, 1981. [17] Eifert et. al. “Characterization of Carbon Felt Electrodes for Vanadium Redox Flow Batteries: Impact of Treatment Methods.” J. Electrochem Soc., 165(11), 2577-2585, 2018.

Ingenium 2020

Metformin administration impairs tendon wound healing Catherine Grace P. Hobayan , Arthur R. McDowella, Feng Lia, Jianying Zhanga*, and James H-C. Wanga, b, c a, b

MechanoBiology Laboratory, Departments of Orthopaedic Surgery, Bioengineering, cPhysical Medicine and Rehabilitation, University of Pittsburgh, PA, USA



Grace Hobayan is a fourth-year bioengineering student at the University of Pittsburgh Swanson School of Engineering. She is interested in the intersection between cellular/tissue engineering and biomechanics, and she plans to attend medical school after graduation and apply Grace Hobayan her bioengineering background to the practice of medicine. Currently, she serves as a Teaching Assistant for Intramural Internship (BIOENG 1002), where she helps bioengineering students develop presentations and effectively communicate about their research projects to a wide variety of audiences. Dr. Jianying Zhang has interdisciplinary education background and overlapping research interests. Dr. Zhang’s research has resulted in over one hundred peerreviewed research publications and more than 35 patents, and some of the products generated from her patents have been Dr. Jianying Zhang approved by the FDA (USA) and put into clinical practice. Currently, Dr. Zhang is a research associate professor in the Department of Orthopaedic Surgery at the University of Pittsburgh School of Medicine.

Significance Statement

Obese patients with Type II diabetes often experience tendon injuries due to the exertion of higher loads on their tendons. Metformin is one of the most common medications for such prescribed for Type II diabetes. However, the effect of metformin on tendon disorders is largely unknown. This study has shown that HMGB1 is necessary for tendon healing, and the inhibition of HMGB1 by metformin impairs tendon wound healing. Thus, metformin may exacerbate healing of injured tendons in obese diabetic patients.

Category: Experimental research

Keywords: Metformin, HMGB1, cell tracking, tendinopathy Abbreviations: HMGB1 - High mobility group box 1, Scx- Scleraxis, α-SMA - α-smooth muscle actin, Met - metformin, PT - Patellar tendon, AT - Achilles tendon, TSC - tendon stem cell


Tendon injury is common, and injured tendons have a limited healing ability. High mobility group box-1 (HMGB1) has been found to enhance wound healing by recruiting cells to migrate to the wound area and increasing cell proliferation. However, the role of HMGB1 in tendon wound healing is currently unknown. Metformin, a hypoglycemic anti-inflammatory drug for Type II diabetes, inhibits HMGB1 activity by binding to its C-tail. Therefore, in this pilot study, we hypothesized that inhibiting HMGB1 by metformin slows the healing of wounded tendons due to its inhibition of HMGB1 activity. To test this hypothesis, a window defect was created in the patellar tendon and Achilles tendon of α-SMA-Ai9-Scx-GFP transgenic mice. The animals were injected either with saline every day (control group), or five days metformin before surgery (short term metformin), or metformin every day (long term metformin). Fluorescence microscopy images of tendon sections taken 4 weeks post-injury indicated that paratenon cells migrated into the wounded area of the tendon in the saline injection mice (control group), but not in the groups that received metformin injection. Cell densities in the wound area and HMGB1 serum levels were higher in the absence of metformin. This may indicate that metformin inhibited HMGB1 activity and reduced wound healing by blocking the recruitment and migration of paratenon cells to the wounded area. Thus, metformin administration limits the healing capacity for wounded Achilles tendons by limiting the migration of paratenon cells to wounded areas.

1. Introduction

Tendinopathy is a prevalent tendon disorder that affects a large proportion of people in both athletic and occupational settings [1]. However, the current treatment options for tendinopathy are largely palliative because the mechanisms causing the tendon disorder are not well understood [2]. Many intrinsic and extrinsic risk/causative factors can predispose to the development of tendinopathy. Among them, diabetes mellitus is an important risk/ causative factor [3]. Obese patients with Type II diabetes often experience tendinopathy due to the exertion of higher loads on their tendons. Metformin is a hypoglycemic anti-inflammatory drug commonly used for treatment of Type II diabetes. High mobility group box 1 (HMGB1) is an alarmin protein released from necrotic cells to induce inflammatory responses in the human body. Extracellular HMGB1 induces several responses, including the activation of proinflammatory cytokines, cell proliferation and stromal cell matrix responses [4]. Metformin has been shown to bind to the acidic tail of HMGB1 and inhibit its inflammatory activity in a concentration-dependent manner [5]. The inflammatory activity of HMGB1 has also been shown to contribute to tendon injury mechanisms by regulating inflammatory cytokines and matrix changes [6]. No prior studies have analyzed the effects of metformin in the context of tendinopathy.


Typically, tendinopathy is studied using the samples from the patients who choose surgical interventions to alleviate tendinopathy symptoms. The limitations of the sample size and source have made it difficult to devise an effective treatment modality for the tendon disease. Thus, animal models of tendinopathy are required to investigate the cellular and molecular mechanisms regulating the tendon disorder to devise effective treatment protocols. Moreover, to develop better treatment options for tendinopathy, it is essential to have a reproducible, cost-effective animal model of tendinopathy that will allow the evaluation of the tendon disease progression and enable the development of better treatment options. Tendon stem/progenitor cells (TSCs) are essential for the maintenance and repair of tendinous tissues when injured. Scleraxis (Scx) and α-smooth muscle actin (α-SMA) are the markers of tendon and mesenchymal progenitor cells, respectively, which are suggested to be involved in tendon wound healing [7]. However, little is known about the anatomical origin of these resident progenitors within the tendon that contribute to natural healing following injury. In this pilot study, we used α-SMA-Ai9-Scx-GFP transgenic mice that express both Scx and α-SMA to characterize the expansion of resident tendon progenitors that contribute to natural wound healing during adulthood and investigate the effect of HMGB1 on wounded tendon healing by inhibiting the activity of HMGB1 with metformin injection. We hypothesized that metformin would reduce the extent to which wounded tendons are healed because of its inhibitory function on HMGB1inflammatory activity. We also hypothesized that administration of metformin both before and after tendon injury would reduce healing capacity to a greater extent than only administering metformin before injury. This information will be beneficial for obese patients with Type II diabetes who regularly take metformin and are afflicted with tendinopathy.

2. Methods 2.1 Transgenic mice Tamoxifen-inducible α-SMA-CreERT2 mice were crossed with Scx-GFP mice, and then crossed with Ai9-Cre reporter mice to generate triple transgenic SMA-Ai9-ScxGFP mice. This allowed for tenocytes and paratenon cells to be visualized via fluorescence microscopy [7]. 2.2 Wound healing model Three intraperitoneal injections of tamoxifen (Sigma Aldrich, St. Louis, MO) were delivered on consecutive days to 2-month-old SMA-Ai9-ScxGFP mice at a dose of 100 μL of 20 mg/mL prior to patellar tendon (PT) injury by 1 mm diameter biopsy punch (Miltex, Inc., York, PA) and Achilles tendon (AT) injury using a 0.5 mm punch (Word Precision Instruments, Sarasota, FL). The animals were divided into three groups (6 mice per group): Group 1 (Saline) received intraperitoneal (IP) injections with saline every day; Group 2 (Met-Short term) received IP injections with Met (160 mg/kg/

48 Undergraduate Research at the Swanson School of Engineering

day) for 5 days before surgery; Group 3 (Met-Long term) received IP injections with the same dose of Met for 5 weeks (5 days before surgery and 4 weeks after surgery). For the Met-Short and MetLong groups, the same dosage of metformin was administered, but the frequency of administration is different. Met-Short involves only pre-surgery injection, and Met-Long involves both pre- and post-surgery injections to investigate whether higher frequency of metformin leads to further inhibition of tendon wound healing mechanisms. All mice were sacrificed at day 30 post-injury, their blood was collected, and both PT and AT of each mouse was harvested. The potential effect of HMGB1 on wounded tendon healing and cell migration was assessed by histological analysis. 2.3 HMGB1 determination in mouse serum The HMGB1 levels in serum of the mice at 30 days postsurgery were determined using an ELISA kit according to the manufacturer’s protocol (Shino-Test Corporation, Tokyo, Japan). Briefly, 0.5 mL of blood samples was collected from the heart of each mouse immediately after they were sacrificed, and the blood samples were stored at room temperature for 1 hour. The serum was separated by a centrifuge at 2000 g for 30 min. The supernatants were used for the measurement of HMGB1. All samples were analyzed from 3-6 mice in each group (n=3-6). 2.4 Histological analysis At 30 days after surgery, the tissue samples were harvested and placed in pre-labeled base molds filled with frozen section medium (Neg 50; Richard-Allan Scientific, Kalamazoo, MI). The base mold with tissue samples was immersed in liquid nitrogen cold 2-methylbutane and allowed to solidify completely. The tissue blocks were either stored in a deep freezer (-80ºC) or cut into 5 µm thick sections for histological analysis. The tissue sections were placed on glass slides and allowed to dry overnight at room temperature. The cell migration and the wounded tendon healing were analyzed under a fluorescent microscope (Nikon eclipse, TE2000-U) using SPOT imaging software (Diagnostic Instruments, Inc., Sterling Heights, MI). 2.5 Semi-quantification of the extent of cell migration The healing results were further analyzed by semiquantification on tendon tissue sections. Briefly, four views from each tendon were randomly chosen on a microscope with a magnification of 20 ×. Then, the Scx-GFP positive (green fluorescence) cells and α-SMA-Ai9 positive (red fluorescence) cells were identified manually and computed by SPOT IMAGING Software. Next, the proportion of Scx-GFP in total cells was calculated by dividing the green cell numbers by total cell numbers, or the sum of the green and red cell numbers. Similarly, the proportion of α-SMA-Ai9 in total cells were calculated by dividing the red cell numbers by total cell numbers. Twelve ratio values (four views/section × 3 sections/mouse) were averaged to obtain the percentage of Scx-GFP positive or α-SMA-Ai9 positive cells, which represents the extent of cell migration in the respective wound area.

Ingenium 2020

2.6 Data analysis The results presented in the figures are representative of these (mean ± SD, n = 3 to 6). Two-tailed student t-test (type=3) was used for statistical analysis to compare the cell numbers and serum levels of HMGB1 between Saline and Met-Short; Saline and Met-Long; and Met-Short and Met-Long groups. A p-value less than 0.05 was considered to be significantly different.

3. Results

Gross-inspection of the wound area showed that metformin injection inhibited wound healing in both PT and AT as evidenced by unhealed wound area size and color (Fig. 1A-F). The concentration of HMGB1 in sera of Met-injected mice was significantly lower than saline injection group (Fig. 1G).

The histological analysis on the frozen sections of mouse tendons showed that Met injection inhibited the cell migration in both wounded PT (Fig. 2) and AT (Fig. 3). Furthermore, saline treated mice had two times more α-SMA-Ai9 cells (red) migrated to the wound area than Met-Long term group (Figs 2, 3). Met injection decreased the migration of both Scx-GFP (green) and α-SMA-Ai9 cells (red) (Figs. 2, 3). Semi-quantification confirmed the decrease in cell migration in the wound area of PT due to long term Met treatment (Fig. 2J). Finally, Met injection decreased the serum levels of HMGB1 (Fig. 4).

Figure 1: Met injection inhibits wounded mouse tendon healing in a dosage-dependent manner. A-C: Gross view of wounded mouse patellar tendon (PT) at week-4 post-surgery; D-F: Gross view of wounded mouse Achilles tendon (AT) at week-4 post-surgery. Metformin injection decreases healing speed at a dosage-dependent manner as evidenced by wound area size (A-C) and color (D-F).

Figure 2: Metformin injection inhibits wounded patellar tendon healing by decreasing tendon cell migration. A-C: Scx-GFP cells; D-F: α-SMA-Ai9 cells; G-I: Merged images of A-C and D-F. Both green and red cell numbers are increased in the wound area of saline injection mice (A, D, G). However, reduced cell numbers and large empty areas are found in Met injections wounds (B, E, H, C, F, I). Semi-quantification shows that the cell numbers of α-SMA-Ai9 (red) in saline group are 2 times more than that in long-term Met injection wound area (J). *p<0.05 compared to saline group; #p<0.05 compared to Met-short group.


4. Discussion

Figure 3: Met injection inhibits wounded Achilles tendon healing by decreasing tendon cell migration. A-C: Scx-GFP cells; D-F: Îą-SMA-Ai9 cells; G-I: Merged images of A-C and D-F; J-L: Enlarged box areas in the images of G-I.

Figure 4: Met injection decreases HMGB1 levels in serum. *p<0.05 compared to saline group.

This study has directly demonstrated that Met was able to block HMGB1 release as evidenced by reduced levels of HMGB1 in the sera of wounded mice. In addition, Met injection to wounded mice impaired the healing as evidenced by low cell density and large, premature unhealed wound areas found in Met treated groups. These results showed that HMGB1 plays a critical role in wound healing by recruiting Îą-SMA positive cells in the paratenon to the wound area. Our results also indicate that metformin has inhibited the activity of HMGB1 and prevented proper wound healing by blocking the recruitment and migration of paratenon cells to the wounded area. This is consistent with the aforementioned literature in that the presence of metformin inhibited inflammatory activity that may have contributed to proper tendon wound healing [5]. Tissue regeneration is a well-orchestrated process that occurs after injury. Understanding the molecular events underlying the regeneration process and developing agents that aid regeneration is essential for patients with injured tissue in a variety of clinical settings [8]. It has been reported that HMGB1 is essential for life because HMGB1 knockout mice die perinatal [9]. Recent studies have shown that HMGB1 induces cell migration [4, 10] and promotes muscle regeneration after acute muscle injury [8]. These findings were further confirmed by our results. We have demonstrated that the metformin administration before tendon injury also inhibits the migration of paratenon cells to the wounded area of an Achilles tendon, and long-term administration of metformin (before and after surgery) slows down the healing. These results suggest that patients with Type II diabetes should stop taking metformin after sustaining tendon injuries. This study provided a useful animal model for tendon wound healing. Our results have demonstrated that both Scx+ resident progenitors and Îą-SMA+ progenitors are the main contributors to natural wound healing during adulthood. One limitation of this study is that proliferation of cells in the wound area was not determined since only a single early time point was analyzed (4 weeks postinjury) over the course of the tendon healing process; only the migration of the cells could be determined. Future studies should incorporate longer time points (e.g., 8-12 weeks after surgery). Hematoxylin & eosin (H&E) staining should be used to analyze alterations in tissue structure of wounded tendons in response to metformin. Immunohistochemistry should also be used to determine the activity of HMGB1 in future studies. Immunostaining and cell tracking should be used to determine both cell proliferation and migration at later time points in the tendon healing process. Furthermore, similar studies should be done to analyze the effects of metformin in female mice, as well as mice of different age groups.

5. Conclusion

We have demonstrated that the metformin administration inhibits the migration of paratenon cells to the wounded area of

50 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

patellar and Achilles tendons and impairs their healing via the inhibition of HMGB1 activity. Future studies are to investigate for similar effects on the tendons of female mice, measure cell proliferation, and verify tissue structure changes over multiple longer time periods to account for the slow healing process for tendons.

6. Acknowledgements

This work was supported in part by the NIH under award numbers AR061395, AR065949, and AR070340 (JHW). We thank Dr. Kelly Williamson, the Division of Laboratory Animal Resources (DLAR), and the Swanson School of Engineering (SSOE) Undergraduate Summer Research Internship for their support on this project.

7. References

[1] Scott A, Ashe MC: Common tendinopathies in the upper and lower extremities. Curr Sports Med Rep 2006, 5(5):233-241. [2] Abate M, Silbernagel KG, Siljeholm C, Di Iorio A, De Amicis D, Salini V, Werner S, Paganelli R: Pathogenesis of tendinopathies: inflammation or degeneration? Arthritis Res Ther 2009, 11(3):235. [3] Lui PPY: Tendinopathy in diabetes mellitus patientsEpidemiology, pathogenesis, and management. Scand J Med Sci Sports 2017, 27(8):776-787. [4] Schiraldi M, Raucci A, Munoz LM, Livoti E, Celona B, Venereau E, Apuzzo T, De Marchis F, Pedotti M, Bachi A et al: HMGB1 promotes recruitment of inflammatory cells to damaged tissues by forming a complex with CXCL12 and signaling via CXCR4. J Exp Med 2012, 209(3):551-563. [5] Horiuchi T, Sakata N, Narumi Y, Kimura T, Hayashi T, Nagano K, Liu K, Nishibori M, Tsukita S, Yamada T et al: Metformin directly binds the alarmin HMGB1 and inhibits its proinflammatory activity. J Biol Chem 2017, 292(20):8436-8446. [6] Akbar M, Gilchrist DS, Kitson SM, Nelis B, Crowe LAN, Garcia-Melchor E, Reilly JH, Kerr SC, Murrell GAC, McInnes IB et al: Targeting danger molecules in tendinopathy: the HMGB1/ TLR4 axis. RMD Open 2017, 3(2):e000456. [7] Dyment NA, Hagiwara Y, Matthews BG, Li Y, Kalajzic I, Rowe DW: Lineage tracing of resident tendon progenitor cells during growth and natural healing. PLoS One 2014, 9(4):e96113. [8] Tirone M, Tran NL, Ceriotti C, Gorzanelli A, Canepari M, Bottinelli R, Raucci A, Di Maggio S, Santiago C, Mellado M et al: High mobility group box 1 orchestrates tissue regeneration via CXCR4. J Exp Med 2018, 215(1):303-318. [9] Calogero S, Grassi F, Aguzzi A, Voigtlander T, Ferrier P, Ferrari S, Bianchi ME: The lack of chromosomal protein Hmg1 does not disrupt cell growth but causes lethal hypoglycaemia in newborn mice. Nat Genet 1999, 22(3):276-280. [10] Venereau E, Casalgrandi M, Schiraldi M, Antoine DJ, Cattaneo A, De Marchis F, Liu J, Antonelli A, Preti A, Raeli L et al: Mutually exclusive redox forms of HMGB1 promote cell recruitment or proinflammatory cytokine release. J Exp Med 2012, 209(9):1519-1528. 51

Mechanical characterization of silk derived vascular grafts for human arterial implantation Patrick Iyaselea, Eoghan M. Cunnanea, Katherine L. Lorentza, Justin S. Weinbauma, e, and David A. Vorpa, b, c, d, f, g Department of Bioengineering, bMcGowan Institute for Regenerative Medicine, cDepartment of Surgery, d Center for Vascular Remodeling and Regeneration, e Department of Pathology, fDepartment of Chemical and Petroleum Engineering, gDepartment of Cardiothoracic Surgery a

Patrick Iyasele is a Senior Bioengineer from Milwaukee, Wisconsin. He is pursuing a Ph.D. in Bioengineering with focus on cardiovascular tissue engineering. He is a Foundation board member for Pitt B.R.O.T.H.E.R.H.O.O.D and is highly active in the Pitt Excel Program. Patrick Iyasele

David A. Vorp, PhD

David A. Vorp, PhD, is the Associate Dean for Research at the Swanson School of Engineering. He is also John A. Swanson Professor of Bioengineering, with secondary appointments in the Department of Cardiothoracic Surgery, the Department of Surgery, the Department of Chemical and Petroleum Engineering, and the Clinical and Translational Science Institute at the University of Pittsburgh.

Significance Statement

Occluded blood vessels are commonly treated by revascularization surgery utilizing vascular grafts. Compliance mismatch between the native coronary artery and a vein autograft often leads to restenosis and failure of roughly half of all autografts implanted during revascularization surgery [1]. Due to the limited quantity of autografts and required invasive harvesting, tissue engineered vascular grafts (TEVGs) are being researched and developed as an alternative. This paper presents preliminary data that suggests silk TEVGs may be a suitable alternative as they have similar mechanical properties to native tissue at low strain ratios.

Category: Experimental research

Keywords: TEVG, Silk, Mechanical Testing, Tangential Modulus Abbreviations: TEVG- Tissue Engineered Vascular Graft

52 Undergraduate Research at the Swanson School of Engineering


Coronary artery disease occurs from the narrowing and blockage of the vessels supplying blood to the heart, leading to reduced blood flow and tissue damage. The preferred treatment for occluded small diameter arteries is revascularization surgery utilizing vascular autografts including the saphenous vein or internal thoracic artery. However, such autografts are limited in quantity, may be of poor quality and require invasive surgery to harvest and utilize. Additionally, compliance mismatch from the native coronary artery with a vein autograft often leads to restenosis and failure of roughly half of all autografts implanted during bypass surgery [1]. Tissue engineered vascular grafts (TEVGs) are currently being studied and developed as alternatives. A successful TEVG should mimic the mechanical properties of native vessels to reduce failure due to compliance mismatching, therefore, the objective of this study was to calculate the tangential modulus of a bombyx mori (BM) silk derived TEVG and compare it to the tangential modulus of the native vessels it is intended to replace. A BM scaffold was seeded with human cells and analyzed by uniaxial extension testing along with two explanted sheep carotid arteries. This entailed cutting the tube into ring specimens and tensile testing five replicates, recording the stress-strain curve and calculating the tangential modulus. The BM silk TEVG had a similar tangential modulus to a native sheep carotid artery at low strain ratios (1.3) but was significantly lower at high strain ratios (1.9).

1. Introduction

Cardiovascular disease is the leading cause of death worldwide, with most deaths associated with coronary heart disease, cerebrovascular disease, peripheral arterial disease, and deep vein thrombosis. These diseases often occur from the narrowing and blockage of blood vessels leading to reduced blood flow and tissue damage due to inadequate nutrient supply [1]. The preferred treatment for occluded small diameter arteries (such as the coronary arteries) is revascularization surgery utilizing vascular grafts. During this surgery, a graft is used to replace or bypass the damaged or occluded vessel. Around 400,000 coronary artery bypass grafting (CABG) procedures are performed each year in the United States alone [2]. The saphenous vein and internal thoracic artery are commonly used for autografts but those are limited in quantity, may be of poor quality and require invasive surgery to harvest and utilize [1]. Compliance mismatch from the native coronary artery and a vein autograft leads to restenosis and failure of roughly half of all autografts implanted during bypass surgery [1]. Tissue engineered vascular grafts (TEVGs) are currently being studied and developed as alternatives. An ideal TEVG should mimic the mechanical properties of native tissues. Compliance mismatch from the native coronary artery and a vein autograft lead to restenosis and failure of roughly half of all autografts implanted during bypass surgery [1]. Tangential modulus has been a useful property to ascertain in this regard as it describes the stiffness of a material at certain mechanical strains that are experimentally tested during uniaxial

Ingenium 2020

extension. Stiffness determines a vessels mechanical reaction to cardiac pressure waves. Replicating vessel stiffness of native tissue is critical to developing accurate blood vessel replacements [3]. The TEVG tested in this study is made from Bombyx mori (BM) silk. BM silk has considerable mechanical strength compared to its weight. Also, silk is resistant to digestion from most proteolytic enzymes, biocompatible, has controllable degradability, has the ability to be fabricated into different shapes, and has the ability for amino acid side chain modification [4]. The TEVG is seeded with human adipose-derived stem cells using a large-seeding device that perfuses cells in a rotating assembly that has been previously developed in our laboratory [5]. According to Issa et al., the number of cells seeded to a scaffold can change the mechanical properties of the scaffold [6]. The TEVG is seeded because seeding autologous endothelial cells onto the luminal surface of graft has been seen to improve patency rates. The absence of endothelial cells lining the lumen leads to the adherence of blood proteins and the activation of clotting mechanisms causing thrombosis [1]. The graft in this paper is seeded with adipose-derived stem cells which differentiate into smooth muscle cells and endothelial cells. Adipose-derived stem cells are used as they can be extracted in high quantities from adipose tissue aspirate and are easy to harvest [1]. The objective of this study is to quantify the tangential modulus of a BM silk derived TEVG and compare it to the tangential modulus of the native vessels it is intended to replace.

2. Methods 2.1 TEVG Fabrication 2.1.1 Graft Casting The grafts were fabricated by loading a 6% BM silk solution (by weight) into cylindrical molds. The silk solution was a generous gift from Dr. Biman Mandal, IIT-Guwahati. The mold was comprised of an outer stainless-steel tube that had a plastic straw inserted into it to provide a smooth surface to prevent damage to the graft during extraction. The mold had a 5.5 mm stainless steel rod that was inserted into the tube to provide the center alignment and stability for the tubular graft structure. This configuration was secured by two custom-designed 3D printed caps at the center of the tube that formed a watertight seal to keep the silk solution from leaking out of the mold. The filled mold was frozen at -20° C for 24 hours and then lyophilized for 12 hours. After lyophilization, the produced graft was submerged in an 80% ethanol solution. From the ethanol treatment, dense membranes of silk fibers were crystallized to the β-sheet conformation which are suitable for adherence and growth of endothelial cells [7]. The finished graft had an inner diameter of 5.5 mm, an outer diameter of 7.1 mm, and a length of 10 cm (Figure 1).

Figure 1: The finished silk TEVG. The final TEVG had an inner diameter of 5.5 mm, an outer diameter of 7.1 mm, and a length of 10 cm. An electrospun layer of silk/PCL was added around the internal structure to increase mechanical strength.

2.1.2 Addition of Electrospun Layer An electrospun layer made up of a 1:1 mixture of silk and polycaprolactone (PCL) was added to the outside of the graft to add additional mechanical strength to the graft using a custom electrospinning device [3]. The device consisted of a power drill on a linear actuator, a syringe pump, and a 12.5 kV voltage generator. The graft was put on a rod that was inserted into the drill head in order to rotate and move linearly. The voltage generator was connected to the rod on its negative terminal causing the graft to be negatively charged and the head of the syringe in the syringe pump was connected to the positive terminal giving the silk/PCL solution a positive charge. This polarization causes extruded silk/ PCL to be attracted to the graft due to electrostatic forces. The drill was set to 200 RPM and 50 linear cycles/min and the syringe pump was set at a .1 ml/min output rate. A total of 1ml of silk/ PCL was dispensed from the syringe. When the drill, pump, and generator were activated, a thin fiber of the silk/PCL mixture was stretched from the tip of the syringe pump to the graft causing the addition of a second silk layer onto the graft. 2.1.3 Vacuum Cell Seeding The scaffold was bulk seeded with 120 million human adipose-derived stem cells in order to produce a graft seeded at similar density to previous work using smaller-scale grafts [8]. This was done by using a custom vacuum rotational seeding device (Figure 2). The cultured cells were suspended and put into a syringe and pumped into the vacuum chamber. The vacuum chamber pressure was -127 mmHg. A cylindrical diffusing nozzle fabricated to match the internal diameter of the graft was used to spray the cell suspension onto the scaffold’s inner surface. The nozzle was attached to a rod that traveled the full length of the graft. The graft was rotated around the diffusion nozzle, inside an acrylic vacuum chamber. This hybrid method of rotational and vacuum seeding combined the best qualities of both methods. With the diffusion nozzle approximating very closely with the internal surface of the graft, the cells were better able to penetrate the graft. Because the graft was rotating under vacuum there was a pressure gradient to further enhance seeding while the rotation 53

helped to uniformly distribute the cells circumferentially. 30 ml of cell suspension was dispensed through the nozzle at 15 ml/min. The graft was rotated at 60 RPM and the linear displacement for the nozzle was 1 linear cycle /min.

Figure 3: Schematic for uniaxial ring testing. Ring tests are performed to obtain stress-strain curves and thereby determine the mechanical properties of the TEVG and carotid arteries. The diameter, width (W) and wall thickness (t) of each of the ring specimens were used to determine A0 and L0. These parameters were used to calculate corresponding stress, strain and tangential modulus values [3].

Figure 2: Custom cell vacuum rotational seeding device. A cylindrical diffusing nozzle sprays the cell suspension onto the scaffold’s inner surface inside an acrylic vacuum chamber. the cells were better able to penetrate the graft to uniformly distribute the cells circumferentially [9]. The scaffold was bulk seeded with 120 million human adipose-derived stem cells.

2.2 Uniaxial Tensile Testing Two sheep carotid arteries were explanted and the TEVG were cut into ring specimens which were tested under uniaxial tension to determine their stress-strain response and tangential modulus at both low (1.3) strain ratio and high (1.9) strain ratio (Figure 3). According to VĂŠronique Laterreur et al., data from a ring tensile test provides an accurate estimate of the failure strain and the stiffness of the graft when compared to measurements with the direct method which consists of pressurizing the graft with fluid until tissue failure [9].

The diameter, width and wall thickness of each of the ring specimens were measured in pixels using Image-J. A pixel to millimeter scaling factor from a metric ruler in captured images was used to convert measured pixels into millimeters. Force and extension values were measured from an Instron 5543A uniaxial extension tester and were converted into stress and strain ratio values (Figure 4). Stress was calculated by equation 1 where F is the load in Newtons (N) and A0 is the initial cross-sectional area: (1) Strain ratio was calculated using equation 2 where L0 is the initial length and Îť is the extension: (2) Tangential modulus was calculated by using linear regression on the low and high strain ratio regions. The low strain ratio region was designated as the 1.2 - 1.4 strain ratio region of the full stress-strain curve. The high strain ratio region was designated as the 1.8 - 2.0 strain ratio region. A linear fitting was applied to each region yielding tangential modulus calculated as the slope of the fitted line [10].

a) b)


Figure 4: Stress-strain curves for the three samples tested. The low strain ratio region is labeled orange, and the high strain ratio region is labeled red. a) Right sheep carotid artery stress-strain curve. b) Left sheep carotid artery stress-strain curve. c) Silk TEVG stress-strain curve. The silk TEVG is a linear sigmoidal shape (Figure 4c) but the arterial stress-strain curves are a non-linear logarithmic shape.

54 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

3. Results

Five replicates for both sheep carotid arteries as well as the TEVG were tested. The tangential modulus for the TEVG was 0.11 ± 0.02 MPa in the low strain ratio and 0.15 ± 0.03 MPa in the high strain ratio region (Figure 5). The high strain ratio region tangential modulus was significantly higher for the left and right carotids compared to the TEVG (4.3 ± 1.3 MPa and 3.6 ± 0.66 MPa vs. 0.15 ±. 03 MPa, respectively; both p= 0.0079 using two tailed Mann-Whitney Test, α= 0.05). The low strain ratio region tangential modulus was not statistically different between the left and right carotid artery and the TEVG (0.07 ± 0.03 MPa and 0.12 ± 0.04 MPa vs. 0.11 ± 0.02 MPa, respectively; p=0.095 and 0.8413). a)


Figure 5: Samples had similar modulus at low strain ratios, but the TEVG modulus was lower at high strain ratios. a) Tangential moduli between the left and right sheep carotids and the silk graft at 1.3 strain ratio region. b) Tangential moduli between the left and right sheep carotids and the silk graft at 1.9 strain ratio region. A two tailed Mann-Whitney Test indicated no significant statistical difference between any groups in the low strain region, but showed significant statistical difference in the high strain region. Asterisks specify statistically significance between groups. * p= 0.0079. **p= 0.0079

4. Discussion

The BM silk TEVG had a similar tangential modulus to a native sheep carotid artery at low strain ratios but was significantly lower at high strain ratios. The difference in tangential moduli between the high strain ratio regions of the TEVG and sheep carotids may come from the TEVG not being remodeled by macrophages and arterial cells as it would be post-implantation [11]. Vascular remodeling is an active process of structural change that involves changes in at least four cellular processes: cell growth, cell death, cell migration, and the synthesis and degradation of extracellular matrix [12]. Cells would not have added any structural proteins to the vessel to increase stiffness since they did not have an implantation or in vitro culture period. Even though the tangential modulus at low strain ratio is similar, the overall shapes of the stress-strain curves are different. The silk TEVG stress-strain curve is a linear sigmoidal shape (Figure 4c) but the arterial stress-strain curves are a non-linear logarithmic shape (Figure 4a, b). According to Peterson et al., the strain which the arteries undergo as a result of arterial pulse pressure variations is normally between 0.01 and 0.04, i.e., between 1% and 4% change in

circumference. The total strain associated with marked constriction and dilation does not usually exceed 10% [13]. Having similar tangential modulus at the 1.3 strain ratio region provides evidence that the graft would not rupture at physiological strains of 10%. It also provides evidence that at this strain region there is no compliance mismatch between the graft and native vessel which causes blood flow disturbances resulting in the development of atherosclerotic plaque. In Soffer, et al., they fabricated a silk scaffold made with an 8% silk solution and a 7.5% silk/ poly(ethylene oxide) solution for the electrospun layer [14]. They determined their graft to have a tangential modulus of 2.45 ± 0.47 MPa at the 50% strain region. Although the tangential modulus of their graft is significantly different from the graft fabricated in this paper, the tangential modulus for a human carotid artery and thoracic aorta at 50% strain is 0.04 ± 0.9 [15][16]. A limitation of this study is its use of uniaxial ring extension testing. The stationary cross heads used to pull the ring specimens may cut into the specimen instead of pulling it. This may happen due to the softness of the specimens and that the cross heads were made out of steel. Also, the scaffold was not cultured to allow for cellular remodeling. Cellular remodeling would change the mechanical properties of the graft. The change in mechanical properties of the graft due to remodeling may make it a closer match to native vessels. This would provide data on the optimum time point to implant a graft to minimize compliance mismatch. This study could be improved by increasing the sample size of all groups tested. The limited sample size hinders us from determining if this is a real effect and from determining if our fabrication technique is consistent. This paper presents preliminary data that suggests silk TEVGs may be a suitable alternative as they have similar mechanical properties to native tissue at low strain ratios. Even with the cells seeded in the graft, the graft seems to have similar enough mechanical properties to be immediately implanted in vivo. A future study would be to repeat the scaffold seeding and extension test to obtain more replicates for the data set. If the graft has mechanical stiffness that is no different than the tissue it is meant to replace, compliance mismatch which leads to restenosis could be mitigated. The graft would prevent the need for autografts which would prevent the required invasive surgery to harvest and utilize them.

5. Conclusions

This preliminary data suggests that our silk TEVG seeded with 120 million human adipose-derived stem cells has a similar tangential modulus to a native sheep carotid artery at low strain ratios but is significantly lower at high strain ratios. At low strains, the TEVG behaves similarly to native carotid artery which would reduce compliance mismatch after bypass surgery. Even though the low strain regions are above the physiological strain of 10%, the silk TEVG grafts behave linearly while native tissue behaves non-linearly. This difference may cause damage or rupture of the graft if strain on the vessels went above physiological strain.


6. References

[1] Pashneh-Tala, Samand et al. “The Tissue-Engineered Vascular Graft-Past, Present, and Future.” Tissue engineering. Part B, Reviews vol. 22,1(2016):68-100. [2] Go, Alan S et al. “Heart disease and stroke statistics--2013 update: a report from the American Heart Association.” Circulation vol. 127,1 (2013): e6-e245. [3] Stankus, John J et al. “Fabrication of cell microintegrated blood vessel constructs through electrohydrodynamic atomization.” Biomaterials vol.28,17(2007):2738-46. [4] Vepari, Charu and David L Kaplan. “Silk as a Biomaterial.” Progress in polymer science vol. 32,8-9 (2007): 991-1007. [5] Soletti, Lorenzo, et al. “A Seeding Device for Tissue Engineered Tubular Structures.” Biomaterials, vol. 27, no. 28, 2006, pp. 4863–4870., 2006.04.042. [6] Issa, Rita I., et al. “The Effect of Cell Seeding Density on the Cellular and Mechanical Properties of a Mechanostimulated Tissue-Engineered Tendon.” Tissue Engineering Part A, vol. 17, no. 11-12, 2011, pp. 1479–1487., 2010.0484. [7] Nogueira, Grínia M., et al. “Preparation and Characterization of Ethanol-Treated Silk Fibroin Dense Membranes for Biomaterials Application Using Waste Silk Fibers as Raw Material.” Bioresource Technology, vol. 101, no. 21, 2010, pp. 8446–8451. [8] Krawiec, Jeffrey T et al. “In Vivo Functional Evaluation of Tissue-Engineered Vascular Grafts Fabricated Using Human Adipose-Derived Stem Cells from High Cardiovascular Risk Populations.” Tissue engineering. Part A vol. 22,9-10 (2016): 76575. doi:10.1089/ten.TEA.2015.0379 [9] Véronique Laterreur et al, Comparison of the direct burst pressure and the ring tensile test methods for mechanical characterization of tissue-engineered vascular substitutes, Journal of the Mechanical Behavior of Biomedical Materials, Volume 34, 2014, Pages 253-263 [10] Khalil Khanafer et al, Determination of the elastic modulus of ascending thoracic aortic aneurysm at different ranges of pressure using uniaxial tensile testing, The Journal of Thoracic and Cardiovascular Surgery, Volume 142, Issue 3, 2011, 682-686 [11] Theodoridis, Karolina et al. “Six-Year-Old Sheep as a Clinically Relevant Large Animal Model for Aortic Valve Replacement Using Tissue-Engineered Grafts Based on Decellularized Allogenic Matrix.” Tissue Engineering Part C: Methods. vol: 23 Issue 12: December 1, 2017 [12] Renna, Nicolás F et al, “Pathophysiology of Vascular Remodeling in Hypertension,” International Journal of Hypertension, vol. 2013, ArticleID 808353, 2013. [13] Peterson, Lysle H., et al. “Mechanical Properties of Arteries in Vivo.” Circulation Research, vol. 8, no. 3, 1960, pp. 622–639., 8.3.622. [14] Soffer, Leah et al. “Silk-based electrospun tubular scaffolds for tissue-engineered vascular grafts.” Journal of biomaterials science. Polymer edition vol. 19,5 (2008): 653-64. doi:10.1163/156856208784089607

56 Undergraduate Research at the Swanson School of Engineering

[15] Adham M, Gournier J, Favre J, De La Roche E, Ducerf C, Baulieux J, Barral X, Pouyet M. Journal of Surgical Research. 1996;64:32. [16] Roeder R, Wolfe J, Lianakis N, Hinson T, Geddes L, Obermiller J. Journal of Biomedical Materials Research. 1999;47:65.

7. Acknowledgements

Thanks to the Swanson School of Engineering and Dr. Vorp for funding that made this work possible. Thanks to Dr. Jonathan Vande Geest for his generous gift of sheep arteries.

Ingenium 2020

Robust osteogenesis of mesenchymal stem cells in 3D bioactive hydrogel nanocomposites reinforced with graphene nanomaterials Eileen Lia, b, Zhong Lib, Colin Del Dukeb, Hang Lina, b, a Departments of Bioengineering, bDepartment of Orthopedic Surgery, Center for Cellular and Molecular Engineering, University of Pittsburgh School of Medicine, PA, USA

Eileen Li

Eileen Li is a junior bioengineering student who is currently pursuing the cellular and biomechanical engineering tracks. She has been working with Dr. Lin at the Center for Cellular and Molecular Engineering of the Department of Orthopaedic Surgery for 2 years. Her research interests focus on developing biomaterial scaffolds for bone regeneration and tissue engineering.

Dr. Lin is an assistant professor (tenure track) working in the Department of Orthopaedic Surgery. His research interests are to understand the relationship between aging and osteoarthritis (OA), develop disease modifying drugs to treat OA, and regenerate articular cartilage Dr. Hang Lin through tissue engineering strategy. Currently, Dr. Lin is supported by both internal and external grants, including several ones from the NIH.

Significance Statement

Treatment of bone defects is presently limited by the insufficient number of suitable bone grafts. A possible solution to this is the use of nanomaterial incorporated bioactive hydrogels to develop biosynthetic bone grafts. Our work demonstrates that the novel 2D nanomaterial, silica-coated graphene oxide (SiGO), can promote the osteogenic differentiation of human mesenchymal stem cells in 3D hydrogels and therefore holds promise in bone tissue engineering.

Category: Experimental research

Keywords: hydrogel, osteogenic differentiation, nanomaterials, silica-coated graphene oxide Abbreviations: silica-coated graphene oxide (SiGO), graphene oxide (GO), mesenchymal stem cells (MSC), methacrylated gelatin (gelMA)


A major challenge facing bone defect treatment is the limited availability of functional, natural bone grafts. To combat this issue, bone grafts engineered from stem cells and synthetic bioactive materials are attracting attention as an alternative approach to bone defect treatment. Silica-coated graphene oxide (SiGO) can be potentially used for the development of engineered bone grafts as silicon (Si) is essential for bone remodeling and growth. SiGO nanosheets were combined with methacrylated gelatin (SiGO/ GelMA) and used to resuspend MSCs. The solution was then photocrosslinked to create 3D cell containing scaffolds. . Usually, osteogenic growth factors are used to enhance osteogenic differentiation during cell culture, however there is the possibility that SiGO can enhance differentiation without the addition of these growth factors. The GelMA and SiGO/GelMA scaffolds were cultured in osteogenic medium for 4 weeks with no supplement of osteogenic growth factors The viability of cells encapsulated in the scaffolds were unaffected by SiGO addition. The expression levels of major osteogenic marker genes were generally higher in the SiGO/GelMA group than GO/GelMA. Calcein green staining, histology and immunohistochemistry results all indicated significantly more homogeneous and robust calcification in the SiGO/GelMA scaffolds. The results suggest that SiGO may hold immense potential in MSC-based bone tissue engineering and regeneration.

1. Introduction

Large bone defects, fracture-delayed unions, and non-unions have become more prevalent. A major issue for the treatment of bone injury is obtaining functional bone grafts for repair. Currently, autografts, allografts and xenografts are widely used clinically for bone defect management. Autografts remain the “gold standard� treatment, but have very limited availability. This method basically creates another area of injury for the patient and allows for the chance of many more complications to occur [1]. Allografts and xenografts are harvested from deceased donor and animals, respectively, thus making availability less of an issue. However, they pose high risks of disease transmission and immunological rejection [1]. To combat this, the applications of bone grafts engineered from stem cells and synthetic, bioactive materials are attracting more and more attention. Nanomaterials, with their unique physical and chemical characteristics, have been used in a broad array of biomedical applications. It has been reported that certain nanomaterials can help promote protein absorption and trigger signaling pathways, which may be exploited to direct cell behavior [3]. For example, if used appropriately, nanomaterials may assist in upregulating osteogenic differentiation of mesenchymal stem cells (MSCs) for creating tissue engineered bone, therefore eliminating the need and complications associated with the use of autografts, allografts or xenografts. Recent studies have utilized 2D graphene nanomaterials and their derivatives such as graphene oxide (GO) in the hope that these nanomaterials can provide satisfactory mechanical and biological environments for stem cell-based bone tissue engineering. Silica-coated graphene oxide 57

(SiGO) is of particular interest in bone tissue engineering as silicon (Si) is a key trace element that has been reported to be essential for maintaining healthy bone and promote bone remodeling [4]. In this research, we hypothesize that the incorporation of SiGO nanosheets in 3D hydrogel scaffolds can significantly enhance the osteogenic differentiation of human MSCs for potential bone repair and regeneration applications.

2. Methods

With IRB approval (University of Washington), MSCs were isolated from femoral heads and trabecular bone of human patients undergoing total knee arthroplasty. GO was synthesized using a modified Hummers method and then converted to SiGO nanosheets via a sol-gel method reported in our previous study [2]. The SiGO nanosheets were combined with 15% methacrylated gelatin (SiGO/GelMA) at 1mg/mL. This composite solution was used to resuspend MSCs at 20M cells/mL and photocrosslinked using 395 nm UV to create 3D cell-laden scaffolds. The scaffolds were cultured in osteogenic media (DMEM with 10% FBS, 1% AntibioticAntimycotic, 10 nM dexamethasone, 0.1 mM L-ascorbic acid 2-phosphate, and 10 mM beta-glycerophosphate, 10 nM vitamin D3) for 4 weeks with no supplement of osteogenic growth factors. The cytocompatibility of SiGO/GelMA was analyzed using the Live/ Dead cell viability assay. Real-time polymerase chain reaction (RT-PCR) was performed to analyze osteogenic gene expression. Histological staining and immunohistochemistry (IHC) were used to further assess osteogenesis quality. Pure GelMA scaffolds were prepared and cultured under identical conditions for comparative purposes.

Figure 1: Live/Dead staining: (A) GelMA, (B) SiGO/GelMA. After 4 day culture, SiGO showed no negative effect on the viability of MSC’s and is therefore biocompatible (scale bar = 200¾m)

Figure 2: Expression of osteogenic markers: A) OCN, B) BMP2. RT-PCR results indicate that SiGO/GelMA scaffolds led to higher expression of OCN and BMP2 in the cells than the GelMA scaffolds. (* P<0.05)

3. Results

The viability of cells encapsulated in the scaffolds were unaffected by SiGO addition, as proved by Live/Dead staining (Figure 1). The expression levels of major osteogenic marker genes, including osteocalcin (OCN) and bone morphogenetic protein 2 (BMP2) (Figure 2), were quantified with reverse transcriptase polymerase chain reaction (RT-PCR) and generally the highest expression of these proteins were in the SiGO/GelMA group. Alizarin red staining, Von Kossa staining and Calcein Green staining all indicated significantly more homogeneous and robust MSC calcification in SiGO/GelMA scaffolds than in the other groups (Figure 3 - 4). Through Immunohistochemistry staining, the highest amount of alkaline phosphatase (ALP) and OCN was identified in the SiGO scaffolds (Figure 5).

58 Undergraduate Research at the Swanson School of Engineering

Figure 3: Alizarin red staining: (A) GelMA, (B) SiGO/GelMA. The SiGO/GelMA scaffolds had significantly more calcium deposition after 4 weeks of osteogenic differentiation (scale bar = 1mm)

Ingenium 2020

4. Discussion

Figure 4: Calcein Green fluorescent imaging: A) GelMA, B) SiGO/GelMA. SiGO containing scaffolds showed more calcein green staining after 4 week culture, indicating more calcium minerals than the GelMA scaffolds

The high expression levels of major osteogenic marker genes for the scaffolds indicate that the MSC’s have undergone osteogenic differentiation. These marker genes were expressed at higher levels for SiGO-containing scaffolds, suggesting that SiGO is more osteoinductive, induces osteogenesis, than unmodified GO. Furthermore, the Alizarin red staining and Von Kossa staining, which indicate calcium and phosphate deposition, respectively, show more robust and homogenous mineralization in the SiGO group. Higher calcium and phosphate deposition is a good indicator of the presence of osteogenically differentiated MSC’s. This is further supported by the fluorescent calcein green staining, which demonstrates the homogeneous 3D distribution of calcium minerals throughout the SiGO/GelMA scaffolds in a larger quantity than in pure GelMA. This suggests that the scaffold biomaterial has an effect on transporting certain proteins that assist in upregulation of osteogenesis in the MSCs, which warrant further investigation. It is worth noting that the SiGO nanomaterial did not elicit any adverse effect on cell viability, as proved by the live/dead assay. Overall, SiGO provides a robust environment for the differentiation of MSC’s into bone tissue.

5. Conclusion

The results suggest that in comparison to other nanomaterials, SiGO may hold immense potential in MSCbased bone tissue engineering and regeneration. We believe the mechanically strong core and biologically active shell of SiGO nanoplatelets synergistically promote osteogenic differentiation. For future research, mechanical testing will be performed on the scaffolds to quantify their compressive modulus and western blot will be utilized to decipher the molecular mechanisms underlying the osteo-inductive properties of SiGO.

6. Acknowledgements Figure 5: Immunohistochemistry staining: A) ALP staining of GelMA, B) OCN staining of GelMA, C) ALP staining of SiGO/GelMA, D) OCN staining of SiGO/ GelMA. After 4 week cell culture, the SiGO/GelMA scaffolds showed significantly more ALP and OCN than the GelMA ones (scale bar = 100 µm)

Summer Research Internship (E.L.) funded by the Swanson School of Engineering and the Office of the Provost. This research is in part supported by the National Institutes of Health (NIH UG3TR002136).

7. References

[1] Aitken G, et al. Benefits and associated risks of using allografts, autograft and synthetic bone fusion material for patients and service providers. JBI Database of Systematic Reviews and Implementation Reports 8(8), 1-13, 2010. [2] Li Z, et al. Incorporating Silica-coated Graphene in Bioceramic Nanocomposites to Simultaneously Enhance Mechanical and Biological Performance. J Biomed Mater Res A, 2020, in press. [3] McMahon R, et al. Development of nanomaterials for bone repair and regeneration. J Biomed Mater Res 101B(2), 387-397, 2013. [4] Gotz W, et al. Effects of Silicon compounds of biomineralization, osteogenesis, and hard tissue formation. Pharmaceutics 11(3) 117-144, 2019. 59

Manufacturing a polyelectrolyte coating on contact lenses using automated vs. manual techniques for the treatment of dry eye disease Zixie Lianga, Alexis Nolfia, b, and Bryan Browna, b University of Pittsburgh, bMcGowan Institute for Regenerative Medicine, Pittsburgh, PA a

Zixie Liang is a Senior Bioengineering Student at the University of Pittsburgh with a cellular concentration and a chemistry minor. She is also an Undergraduate Research Assistant at McGowan Institute for Regenerative Medicine. Zixie Liang

Dr. Bryan Brown is an Associate Professor in the Department of Bioengineering with secondary appointments in the Department of Obstetrics, Gynecology, and Reproductive Sciences and the Clinical and Translational Science Institute at the University of Pittsburgh. He is also a core Dr. Bryan Brown faculty member of the McGowan Institute for Regenerative Medicine where he serves as the Director of Educational Outreach. Dr. Brown is also an Adjunct Assistant Professor of Clinical Sciences at the Cornell University College of Veterinary Medicine and Chief Technology Officer of Renerva, LLC, a Pittsburgh-based start-up company.

Significance Statement

Dry eye disease (DED) is a common ocular disease worldwide and is characterized by inflammation mediated by proinflammatory macrophages, however, treatment currently cannot provide an effective and sustained relief. This project aims to establish a new treatment for DED that creates a nanometer thick coating on contact lenses with a drug delivery system that releases immune modifying drugs efficiently to patients’ eyes. This immune modifying drug is capable of shifting macrophage phenotype, therefore reducing inflammation. The automated and manual manufacturing techniques to coat lenses are compared in this paper.

Category: Experimental research

Keywords: Layer-by-layer coating, macrophage, contact lenses, dry eye disease

60 Undergraduate Research at the Swanson School of Engineering


Dry eye disease (DED) is a multifactorial disease associated with a diminished quality of tear film and compromised ocular surface. It is recognized that inflammation mediated by macrophages is critical to the development of DED; therefore, we propose to coat an immune modifying drug that shifts macrophages from a pro-inflammatory (M1) to anti-inflammatory (M2) phenotype on to contact lenses with a polymer delivery system. This should provide for a sustained release of drug over time. Drug coated lenses were made either via automated coating method using a machine or in a manual way and then compared. Alcian blue staining and drug release kinetics show that immune modifying drug can be successfully coated on to lenses, and manually made lenses tend to be more uniformly coated but release less drug over time. The results of this study demonstrate that combination of current automated and manual methods should be used for the next-stage lens manufacturing for better drug coating and efficiency.

1. Introduction

Dry eye disease (DED) is a prevalent disease in the US and worldwide affecting millions of people, especially middle-aged and over [1]. Patients with dry eye disease often experience visual disturbances, eye dryness, irritation, and light sensitivity, which decreases the quality of life [2]. Treatments such as over the counter and prescription eye drops only provide transient and temporary relief and do not change the underlying disease [3]. Therefore, a new treatment for DED is desired. Past research revealed that DED is a disease with a core mechanism of inflammation and that this inflammation is mediated by macrophages [4]. A treatment for DED may be achieved by manipulating the polarization of macrophages to shift from a proinflammatory (M1) to an anti-inflammatory (M2) phenotype [4]. Previous studies focusing on polypropylene mesh have shown that immunomodulatory cytokines that promote an M2 phenotype can be released from a nanometer-thickness polymer coating loaded onto the surface of a biomaterial implant [5]. Therefore, we propose to use this immune-modifying drug with a polymer delivery system to create a coating on a contact lens. We hypothesize that this will give a sustained release of drug over time to remedy the defect of current DED treatment. To manufacture these coated lenses, we either produced them using a machine in an automated way or produced them manually by hand. The machine coated vs. hand coated method was compared during this experiment.

Ingenium 2020

Figure 1: Procedure of layer-by-layer coating performed using chitosan as polycation and dermatan sulfate as polyanion in cycles.

2. Methods 2.1 Layer-by layer coating A layer-by-layer procedure was conducted in order to deposit a uniform coating capable of releasing immune-modifying drugs. Chitosan (2mg/mL) was used as the polycation and dermatan sulfate (2mg/mL) was used as the polyanion in order to build up base layers (Fig.1). For manually produced lenses, lenses were placed in a cage and moved by hand between polymer solutions and washes (Fig.2). Residual polymer was tapped off in-between steps. For lenses made in an automated way, lenses were secured between tweezers and clamped into a Silar Controller (MTI Corporation, Richmond, CA) automated staining apparatus where the machine moved lenses between solutions and washes (Fig.3).

Figure 2: Set up of hand dipping method (i), and the lens cage (ii) used to secure lens.

After 10 cycles of non-drug-containing base polymer layers were added, a mixture of immune modifying drug (1.5µg/mL) and dermatan sulfate (2mg/mL), along with chitosan, was used to build up another 40 drug-containing layers on the lenses. Uncoated lenses were used as a control. 2.2 Coating characterization—Alcian blue staining Alcian blue, a blue stain that stains glycosaminoglycan components, staining was performed to confirm lenses had been successfully coated with polymers in a uniform and conformal way. Hand coated lens, machine coated lens, and uncoated lens were re-hydrated in distilled water and submerged into Alcian blue dye for 30 minutes at room temperature, and then the lenses were washed in running distilled water for 5 min. Stain was observed and captured by camera. 2.3 Drug cumulative release and kinetics—Controlled drug release assay with ELISA assay A controlled drug release assay and subsequent ELISA assay were performed to test if the drug was able to be released in a relatively more sustained way and if there were differences in release amount or kinetics due to differences in manufacturing. Immune modifying drug coated lenses (using manual and automated methods, n of 1 per condition) and uncoated lenses were immersed into a 500 μL solution of 1x PBS containing the enzymes chitosanase (0.05units/mL) and chondroitinase ABC (0.05 units/mL). Lenses were then incubated at 37°C with agitation. The solution was collected and changed every 24 hours in the first week, and then every 48 hours in the following 2 weeks. ELISA was then run using a commercially available kit (R&D Systems, Minneapolis, MN) in order to assess amount of drug release at each timepoint, followed by the generation of a graph showing cumulative release of drug over time.

3. Results

Images of Alcian blue stained lenses were captured and shown below in Figure 4. Both the hand dipped lens (i) and machine dipped lens (iii) were stained blue while the uncoated control lens (ii) remained clear. This experiment was repeated twice with a total of 2 lenses per condition. The hand dipped lens Figure 3: Set up of machine dipping method (i), and the tweezer (ii) used to secure lens.


appeared to be more uniformly and consistently stained, while the machine dipped lens had an unstained portion on the top where it was clamped to the machine. At the same time, the machine dipped lens showed a darker staining than the hand dipped lens.

Figure 4: Alcian blue staining of hand dipped (i), uncoated control (ii), and machine dipped (iii) lenses. Color blue indicate successful coating.

From ELISA, a graph of cumulative release of drug over time was made and shown in Figure 5. For both coating methods, the amount of drug released increased rapidly in the first 50 hours and reached its maximum at around 100 hours. However, the machine dipped method produced lenses that released approximately 2000 pg more immune modifying drug than the hand dipped method. The uncoated control lens shows no release of drug.

shows that manually dipped lenses, while having similar release kinetics, tend to release less drug overall than lenses dipped in an automated way. This could be attributed to the more thorough polymer removal and drying between steps in the manual method. Overall, the manual method showed a better ability of coating; however, the automated method is more efficient and convenient. Future study will try to incorporate the cage concept we used for manually coated lenses into the automated method to avoid direct contact from tweezer to lenses, which should help to make the coating more uniform. This combination could also make the manufacturing process more efficient. To achieve this goal, we would design a 3D printed cage that can be secured onto the machine and is able to hold several lenses at the same time. More experiments investigating the amount and release time of immunemodifying drugs from contact lenses that can be manufactured using numerous techniques would be also conducted in order to identify the best technique for project scale-up. Future studies also include in vitro and in vivo experiments to confirm maintenance of drug bioactivity and efficacy in dry eye animal models.

5. Conclusion

This investigation demonstrated that a uniform coating consisting of chitosan and dermatan sulfate can be coated onto contact lenses in order to deliver immune-modifying drugs and that this coating is more uniformly applied when using a nonautomated manufacturing technique. The length of release is the same for both automated and non-automated methods, but the automated produced lens released more immune modifying drug than the non-automated lens. Future studies will determine the best technique of coating to prepare the in vitro and in vivo study of the contact lenses.

6. References

Figure 5: Cumulative release of drug comparing lenses produced in an automated way (red circles) and lenses produced in a manual way (blue circles). Uncoated control lenses (green circles) show no release of drug.

4. Discussion

Blue stain on the hand dipped and machine dipped lenses indicate that both the manual and automated coating method can successfully load the coating on to the lens. The unstained portion on the machine dipped lens shows that the machine dipped method is less uniform due to the tweezer used for attachment to the machine. The darker staining on the machine dipped lens suggests a larger amount of polymer deposition. This result also corresponds with the cumulative release over time analysis that

62 Undergraduate Research at the Swanson School of Engineering

[1]. DA Schaumberg, DA Sullivan, et al. August 2003. Prevalence of dry eye syndrome among US women. Am J Ophthalmol 136(2):318–326. [2]. W Stevenson, SK Chauhan, et al. January 2010. Dry eye disease: an immune-mediated ocular surface disorder.Arch Ophthalmol 130(1):90-100. [3]. 2007. The definition and classification of dry eye disease: report of the Definition and Classification Subcommittee of the International Dry Eye WorkShop. Ocul Surf 5(2):75–92. [4]. I You, T Coursey, et al. August 2015. Macrophage Phenotype in the Ocular Surface of Experimental Murine Dry Eye Disease. Arch Immunol Ther Exp (Warsz) 63(4): 299-304. [5]. D Hachim, S LoPresti, et al. October 2016. Shifts in macrophage phenotype at the biomaterial interface via IL-4 eluting coatings are associated with improved implant integration. Elsevier Biomaterials 112(2017): 95-107.

7. Acknowledgments

Study conducted at the McGowan Institute for Regenerative Medicine in Dr. Bryan Brown’s Lab. Many thanks to Alexis Nolfi for guiding the experiments and for being an excellent mentor.

Ingenium 2020

Changes to the maternal sacrum and coccyx during and after pregnancy and delivery Liam Martina, Megan R. Routzong, BSa, Ghazaleh Rostaminia, MD, MScb, Pamela A. Moalli, MD, PhDc, Steven D. Abramowitch, PhDa Translational Biomechanics Laboratory, Department of Bioengineering, Swanson School of Engineering, University of Pittsburgh, Pittsburgh, PA, USA b Female Pelvic Medicine and Reconstructive Surgery (PFMRS), Division of Urogynecology, University of Chicago Pritzker School of Medicine, Northshore University HealthSystem, Skokie, IL, USA c Department of Obstetrics, Gynecology & Reproductive Surgery, University of Pittsburgh, Pittsburgh, PA, USA; Magee-Womens Research Institute, Pittsburgh, PA, USA a

Liam Martin

Liam Martin is a senior bioengineering student on the biomechanics track with a minor in mechanical engineering. He has worked for ten months in the Translational Biomechanics Laboratory where he has worked to help describe the effects of pregnancy and delivery on the maternal bony pelvis.

Dr. Abramowitch received his B.S. (1998) in Applied Mathematics and Ph.D. (2004) in Bioengineering from the University of Pittsburgh. Currently, he is an Associate Professor in the Department of Bioengineering and serves as the Director of the Translational Biomechanics Dr. Steven D. Laboratory. This past October he was the Abramowitch recipient of the Biomedical Engineering Society (BMES)diversity lecture award at the national conference in Philadelphia. Dr. Abramowitch’s research focuses on understanding the impact of pregnancy, delivery, and other life events (aging, menopause, etc.) on the structural integrity of the pelvic floor in women.

Significance Statement

Changes to the maternal pelvis during pregnancy and after delivery have yet to be robustly quantified, but could eventually allow for identification of women at risk of sustaining injury during vaginal delivery. By looking at the combined maternal sacrum-coccyx shape, we found significant posterior movement of the coccyx with respect to the sacrum during pregnancy and, in some women, after delivery.

Category: Computational research

Keywords: Delivery, Pregnancy, Maternal bony pelvis


Hormonal changes during pregnancy cause tissue remodeling, presumably to facilitate vaginal delivery. This study aimed to determine whether softening of maternal tissues results in sacrumcoccyx shape changes by comparing measurements between nulliparous (have never given birth), gravid (pregnant), and parous (have given birth) women. We hypothesized that these measures would differ significantly between groups and be consistent with remodeling that would facilitate vaginal delivery (i.e. posterior movement of the coccyx to accommodate the fetus). Assuming that some women do not fully recover from delivery, we expect to see differences between all three groups. Sacrum and coccyx features were measured by analyzing pelvic MRI scans. Of the 12 measures performed, 3 had significant univariate results: coccygeal curvature index (p<0.001), sacrococcygeal curvature index (p<0.001), and sacrococcygeal angle (p=0.010). Only the nulliparous and gravid groups differed significantly, while the parous values straddled both groups. The results of this study support the hypothesis that pregnancy results in significant changes to the combined maternal sacrum/coccyx shape that are consistent with those more favorable for vaginal delivery and implies that lasting changes occur during pregnancy and/or delivery. Additionally, when dividing these groups into subgroups defined by parity (number of deliveries), larger shape changes were quantified with increasing parity in the gravid group. Because our gravid patients had yet to give birth vaginally (vaginally nulliparous), these changes are likely due to pregnancy alone as a C-section is not expected to affect pelvic shape. This suggests that pregnancy, despite mode of delivery, can result in unrecoverable pelvic shape changes.

1. Introduction

Hormonal changes during pregnancy are known to cause tissue remodeling, resulting in connective tissue laxity at the pubic symphysis and sacroiliac joints presumably to facilitate vaginal delivery [1]. The tissue laxity and remodeling allow for the maternal pelvis to accommodate the growing fetus. Previous work by our lab demonstrates the need for this tissue laxity as the sacrum and coccyx significantly engaged with the fetal head during simulations of vaginal delivery: the mechanical load introduced by the fetal head pushed the coccyx posteriorly, forcing the muscles and connective tissues anchored and engaged with these bones to stretch [2]. This suggests three potential sources for persistent pelvic shape changes—increases in intraabdominal pressure due to the growing fetus, tissue remodeling during pregnancy, and/or injury during vaginal delivery. If the coccyx is moved during delivery, it is reasonable to assume that tissue remodeling during pregnancy may make this motion easier. Other studies have shown significant movement of maternal bony structures during pregnancy, though sacrum and coccyx shape specifically have yet to be investigated [3,4]. Tissue remodeling along with mechanical strain from the fetus have been found to cause lower spine and pubic symphysis pain that can persist after pregnancy and delivery [5]. Also, there are other studies that show that there are many changes that occur on the pelvic floor muscles [6]. It is 63

then believed that this pain may be less severe in women whose pelvises require less adaptation during pregnancy and delivery. If the pelvic shape does change significantly, can we quantity it, and does it return to “normal” afterwards? Thus, we aimed to determine whether tissue remodeling during pregnancy results in a measurable change in the combined sacrum-coccyx shape by comparing midsagittal angles and curvature indices between nulliparous (have never given birth), gravid (pregnant), and parous (have given birth) women. We hypothesized that these groups would differ significantly, consistent with remodeling and coccyx movement that would facilitate vaginal delivery.

2. Methods

This retrospective study was approved by the Institutional Review Board at the University of Pittsburgh and Northshore University HealthSystem. Images of 62 female patients between the ages of 20 and 49 that had a magnetic resonance imaging (MRI) pelvic scan with or without contrast for medical indications (such as abdominal/pelvic pain, appendicitis, abnormal placentation, or fetus anomalies) at Magee-Women’s Hospital or Northshore University HealthSystem between 2005 and 2018 were included in this study. Exclusion criteria were history of pelvic surgery (not including cesarean delivery), pelvic masses, and incomplete scans or birth history. Nongravid patients were imaged in the supine position while gravid patients were imaged in a lateral decubitus position. Patients were sorted into groups based on parity (number of births) and whether they were gravid (pregnant) resulting in 20 nulliparous (had never given birth), 16 gravid and vaginally nulliparous (pregnant and had never given birth vaginally), and 26 parous (had given birth at least once but were not currently pregnant) patients. Using HOROS v. 3.3.5 (Nimble Co LLC, Annapolis, MD USA) the midsagittal slice was identified, and 12 different sacrum and coccyx angle and curvature measurements were made using

definitions from previous literature [7]. All of these measures are either a count (the number of coccygeal vertebrae), length (coccygeal straight and curved length), or curvature index (ratio of straight to curved length). A curvature index is the ratio between the straight and curved length multiplied by 100, where the straight length is a straight line through two predetermine anatomical coordinates and the curved length is the average of the anterior and posterior edges of the shape. A structure with a curvature index value of 100 is perfectly straight while those with lower values are more curved. One of the measures that was defined in the previous literature was the sacral angle, but this measure was excluded here as many scans did not include necessary bony landmarks [7]. Three measures of interest are illustrated and defined by Figure 1 where the rest are defined as follows: Coccygeal curved and straight lengths measured from the middle of the upper border of the first coccyx vertebrae (Co1) to the coccygeal tip; Sacral curved and straight lengths measured from the middle of the upper border of S1 to the middle of the of the inferior border of S5; sacrococcygeal curved and straight lengths measured from S1 to the tip of the coccyx; Sacrococcygeal angle is the included angle between the middle of the superior portion of S1, the middle of the superior portion of the Co1, and the tip of the coccyx; Coccygeal angle is the included angle between the line drawn through the superior and inferior portions of Co1 and the line drawn through the superior and inferior of the most inferior coccygeal vertebrae [7].This study defined the sacrum as the 5 vertebrae below the lumbar spine. The sacrum-coccyx border is easily identifiable by the sharp change in orientation of the individual vertebrae, where the entire sacrum-coccyx shape is visible (Figure 1). Everything inferior to the 5 sacral vertebrae was defined as the coccyx, resulting in 3, 4, or 5 coccygeal vertebrae. The number of coccygeal vertebrae was included in the following statistical analysis and the difference between groups was found to be statistically insignificant.

A) B)


Figure 1: A) Shows the outline of the sacrococcygeal curvature index. The curvature index is the ratio between the straight length (yellow) to the average between the anterior and posterior curved lengths (pink and cyan respectively). B) Shows the sacrococcygeal curvature index the colors and the calculation are the same as in figure 1A. C) Shows the sacrococcygeal angle which is the included angle between the sacral and coccygeal straight lengths.

64 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

Statistical analyses were conducted in SPSS v.25 (IBM Corp. IBM SPSS Statistics, Armonk, NY) and consisted of a OneWay Independent MANCOVA followed by univariate ANOVAs and Benjamini-Hochberg (BH) corrections post hoc [8]. The covariates were age and parity. In a BH correction, a critical value is calculated, and if the p-value is smaller the measure is considered significant. Those variables are then considered in a univariate analysis. Measures with significant differences between groups (p<0.05) were followed-up with additional multiple comparisons. Homogeneity of variances were tested, and independent samples were assumed.

3. Results

Overall, the sacrococcygeal measures between groups (nulliparous, gravid, and parous) were significant at the multivariate level (p<0.001). After BH corrections, three of the measures

had significant univariate results: the coccygeal curvature index (p<0.001), sacrococcygeal curvature index (p<0.001), and sacrococcygeal angle (p=0.010). Table 1 shows the measures (average and standard deviation), the univariate p-values, and the BH critical values for all measures. The data shows that the sacrum-coccyx shape becomes straighter and the sacrum and coccyx more aligned during pregnancy. For all significant measures, only the nulliparous and gravid groups differed significantly (Figure 2a). When further separating those groups into subgroups based on parity, we can isolate the effect of pregnancy from that of vaginal delivery (Figure 2b). Because our gravid patients were vaginally nulliparous, this refers to non-vaginal delivery methods (i.e. cesarean delivery). However, the parous group was separated by vaginal parity. Qualitatively, we see that these subgroups differ more within the gravid group.


Nulliparous (Mean ± std)

Gravid (Mean ± std)

Parous (Mean ± std)

ANOVA p-value


Coccygeal Curvature Index

78.7 ± 6.6

89.2 ± 10.0

80.0 ± 5.5

< 0.001


Sacrococcygeal Curvature Index

73.3 ± 5.8

79.2 ± 3.7

77.6 ± 5.4

< 0.001


Sacrococcygeal Angle

92.8° ± 10.9°

109.3° ± 9.4°

101.9° ± 11.0°



Coccygeal Straight Length

3.2 cm ± 0.5 cm

3.7 cm ± 0.8 cm

3.2 cm + 0.5 cm



Sacrococcygeal Straight Length

11.4 cm ± 1.2 cm

12.3 cm ± 0.9 cm

11.9 cm +1.2 cm



Sacral Curvature Index

90.7 ± 6.6

89.2 ± 3.6

92.6 ± 4.6



Sacral Straight Length

10.8 cm ± 0.8 cm

10.5 cm ± 1.0 cm

10.7 cm ± 0.9 cm



Sacral Curved Length

12.0 cm ± 0.8 cm

11.8 cm ± 0.9 cm

11.7 cm + 0.8 cm



Coccygeal Angle

126.3° ± 17.2°

132.4° ± 21.9°

133.9° ± 14.5°



Coccygeal Curved Length

4.1 cm ± 0.6 cm

4.2 cm ± 0.8 cm

4.1 cm ± 0.6 cm



Coccygeal Vertebrae

3.5 ± 0.7

3.4 ± 0.5

3.5 ± 0.6



Sacrococcygeal Curved Length

15.6 cm ± 1.0 cm

15.5 cm ± 1.4 cm

15.4 cm ± 1.1 cm



Table 1: Table of all sacrococcygeal measures with significant differences bolded

Figure 2: A) boxplot of the coccygeal curvature index values with respect to their grouping (nulliparous, gravid, parous) with significant p-values shown B) shows the coccygeal curvature index normalized to the nulliparous average (shown by the dotted line at y = 1) separated by parity. For the gravid (which are vaginally nulliparous) the parity is non-vaginal delivery while the parous group is separated by number of vaginal deliveries.


Figure 3: A) boxplot of the sacrococcygeal curvature index values with respect to their grouping (nulliparous, gravid, parous) with significant p-values shown B) shows the coccygeal curvature index normalized to the nulliparous average (shown by the dotted line at y = 1) separated by parity. For the gravid (which are vaginally nulliparous) the parity is non-vaginal delivery while the parous group is separated by number of vaginal deliveries.

Figure 4: A) boxplot of the sacrococcygeal angle values with respect to their grouping (nulliparous, gravid, parous) with significant p-values shown B) shows the coccygeal curvature index normalized to the nulliparous average (shown by the dotted line at y = 1) separated by parity. For the gravid (which are vaginally nulliparous) the parity is non-vaginal delivery while the parous group is separated by number of vaginal deliveries.

The second significant measure was the sacrococcygeal curvature index (Figure 3). The figure shows that gravid group trends to have a straighter sacrum and coccyx (trending towards 100 on the y-axis). The values follow a similar trend to that of the coccygeal curvature index, as the nulliparous and gravid groups differed significantly while the parous group fell in between, differing significantly from neither. The third significant measure was the sacrococcygeal angle (Figure 4). Again, only the nulliparous and gravid groups differed significantly. A similar trend is followed as well when the groups are split by parity as subsequent non-vaginal deliveries in the gravid group resulted in larger shape changes.

4. Discussion

Overall these results support the hypothesis that pregnancy and childbirth result in significant change to the combined maternal sacrum-coccyx shape. Specifically, the coccyx moves posteriorly with respect to the sacrum during pregnancy as the coccygeal and

66 Undergraduate Research at the Swanson School of Engineering

sacrococcygeal curvature indices of gravid patients are closer to 100. This is reinforced by the fact that the sacral measurements did not differ significantly across any of the groups. This suggests that pregnancy and delivery would influence the coccyx more than the sacrum. These shape changes appear to meaningfully increase with each subsequent pregnancy. Figure 2b shows the coccyx becomes straighter (curvature index closer to 100) with subsequent pregnancies following Cesarean deliveries. However, the parous group does not differ from the nulliparous average, even after multiple deliveries. A curvature index closer to 100 shows that as the fetus grows the in the mother the sacrum/coccyx shape straightens out to become more vertical within the mother. Previous work has shown a similar phenomenon with cellular memory that results in many changes within the mother that allow for easier deliveries and postpartum recovery [9]. Another supports our findings that there is recovery in women after delivery, as at one year postpartum the mothers’ bodies had recovered significantly [10].

Ingenium 2020

In future research, longitudinal studies should focus on the effect of multiple pregnancies and mode of delivery to further explain the differences seen here. As the major limitation of this study is that the women are not the same across all groups. A future longitudinal study would be able to confirm the findings of this study by following the same patient throughout pregnancy and after delivery. This study only looked at midsagittal shape differences but provides motivation for a 3D analysis to investigate shape variations in the entire bony pelvis.

5. Conclusions

This study found that the coccyx is the main cause of variation in the combined maternal sacrum-coccyx shape as the coccyx moves posteriorly with respect to the sacrum during pregnancy. For some women, these changes persist after delivery, while others return to a more nulliparous shape. Continuation of this study may lead to an increase in understanding the ways that women are injured during delivery.

[8] Y. Benjamini and Y. Hochberg, “Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing,” J. R. Stat. Soc. Ser. B, 1995, doi: 10.1111/j.25176161.1995.tb02031.x. [9] D. Goldman-Wohl, M. Gamliel, O. Mandelboim, and S. Yagel, “Learning from experience: cellular and molecular bases for improved outcome in subsequent pregnancies,” American Journal of Obstetrics and Gynecology. 2019, doi: 10.1016/j. ajog.2019.02.037. [10] C. Reimers, J. Stær-Jensen, F. Siafarikas, J. SaltyteBenth, K. Bø, and M. Ellström Engh, “Change in pelvic organ support during pregnancy and the first year postpartum: A longitudinal study,” BJOG An Int. J. Obstet. Gynaecol., 2016, doi: 10.1111/1471-0528.13432.

6. Acknowledgements

I would like to thank Dr. Abramowitch and Megan Routzong for all of their help with research this past summer. I would also like to acknowledge the Swanson School of Engineering undergraduate research grant and NSF GRFP Grant #1747452 for supporting this research.

7. References

[1] P. Soma-Pillay, P. Soma-Pillay, C. Nelson-Piercy, H. Tolppanen, A. Mebazaa, and A. Mebazaa, “Physiological changes in pregnancy : review articles,” Cardiovasc. J. Afr., 2016, doi: 10.5830/cvja-2016-02. [2] M. R. Routzong, P. A. Moalli, S. Maiti, R. De Vita, and S. D. Abramowitch, “Novel simulations to determine the impact of superficial perineal structures on vaginal delivery,” Interface Focus, 2019, doi: 10.1098/rsfs.2019.0011. [3] S. D. Liddle and V. Pennick, “Interventions for preventing and treating low-back and pelvic pain during pregnancy,” Cochrane Database of Systematic Reviews. 2015, doi: 10.1002/14651858. CD001139.pub4. [4] T. Sipko, D. Grygier, K. Barczyk, and G. Eliasz, “The Occurrence of Strain Symptoms in the Lumbosacral Region and Pelvis During Pregnancy and After Childbirth,” J. Manipulative Physiol. Ther., 2010, doi: 10.1016/j.jmpt.2010.05.006. [5] J. Borg-Stein and S. A. Dugan, “Musculoskeletal Disorders of Pregnancy, Delivery and Postpartum,” Physical Medicine and Rehabilitation Clinics of North America. 2007, doi: 10.1016/j. pmr.2007.05.005. [6] M. Alperin, T. Kaddis, R. Pichika, M. C. Esparza, and R. L. Lieber, “Pregnancy-induced adaptations in intramuscular extracellular matrix of rat pelvic floor muscles,” Am. J. Obstet. Gynecol., 2016, doi: 10.1016/j.ajog.2016.02.018. [7] J. T. K. Woon, V. Perumal, J. Y. Maigne, and M. D. Stringer, “CT morphology and morphometry of the normal adult coccyx,” Eur. Spine J., 2013, doi: 10.1007/s00586-012-2595-2. 67

Laser-induced nanocarbon formation for tuning surface properties of commercial polymers Angela J. McComba, Moataz Abdulhafeza and Dr. Mostafa Bedewya NanoProduct Lab, Department of Industrial Engineering University of Pittsburgh, PA, USA a

Angela McComb is a sophomore undergraduate researcher (SSOE Fellow) in the Industrial Engineering department at the University of Pittsburgh. She is pursuing a degree in Industrial Engineering. Angela McComb

Moataz Abdulhafez is a PhD student at the Industrial Engineering department at Pitt. His research interests include laser material processing and in-situ characterization of nanomaterials. Moataz Abdulhafez

Dr. Mostafa Bedewy is an Assistant Professor of Industrial Engineering, Chemical & Petroleum Engineering, and Mechanical Engineering & Materials Science at the University of Pittsburgh, where he leads the NanoProduct Lab. Before that, he worked as a Postdoc at Dr. Mostafa Bedewy the Massachusetts Institute of Technology (MIT) in the area of bionanofabrication with Professor Karl Berggren in the Research Lab for Electronics. Also, he was previously a Postdoc at the MIT Laboratory for Manufacturing and Productivity, working with Professor John Hart on in situ characterization of carbon nanotube growth. In 2013, he completed his PhD at the University of Michigan in Ann Arbor, where he worked with Professor Hart on studying the population dynamics and the collective mechanochemical factors governing the growth and self-organization of filamentary nanostructures. He holds a Bachelor’s degree (honors) in Mechanical Design and Production Engineering (2006) and a Master’s degree in Mechanical Engineering (2009), both from Cairo University.

Significance Statement

Laser-induced nanocarbon (LINC) is a directly graphitized carbon nanomaterial on polyimide using laser irradiation. It has high potential for applications in antibacterial and antifouling surfaces. We demonstrate that controlling the laser parameters enables tuning the surface properties of LINC.

Category: Experimental research

Keywords: Laser-induced nanocarbon (LINC), Tunable properties, Laser writing, Nanocarbon

68 Undergraduate Research at the Swanson School of Engineering


The manufacturing of flexible materials with patterned functional nanostructures is required for wearable electronics and conformal devices. Laser-induced nanocarbon (LINC) is a typically porous material prepared by direct laser writing with an infrared laser on polymers like polyimide in ambient atmosphere. The resulting material is graphitic and hence conductive, making it suitable for applications in broad fields, such as microfluidics, sensors, and electrocatalysts. This paper aims to expand the range of surface properties for LINC by tuning laser processing parameters. While it was previously shown that different wetting behavior of LINC can be achieved by lasing in inert atmospheres, we uniquely fabricate hydrophobic and hydrophilic LINC by lasing in air, which underlie the simplicity, scalability, and cost-effectiveness of our process. We demonstrate that by varying laser speed, power, and line-to-line gap sizes, considerable changes in water contact angle can be obtained. Upon examining tensiometer, optical, and scanning electron microscopy (SEM) images, it is clear that different surface properties are due to the different morphologies of LINC. Additionally, it becomes evident that increasing gap sizes leads to an increase in contact angles and thus an increase in the hydrophobicity of the surface. The results demonstrate that simple changes to the lasing parameters, the sample hydrophilicity and hydrophobicity can be controlled without special environment control setups. The results also provide insight into the underlying mechanisms behind the various wetting behaviors of LINC.

1. Introduction

The applications of commodity polymers are ubiquitous in almost all aspects of modern life. Recently, it has been shown that nanoscale graphitic structures can be synthesized through the lasing of polyimide, a popular commodity polymer, with commercially available low-cost IR lasers [1]. These laser-induced nanocarbon (LINC) structures have been shown to exhibit excellent electrical and thermal conductivity as well as high surface area [1]. In addition, different morphologies have been observed at the microscale for LINC material, ranging from porous to fibrous [2]. Accordingly, wide-reaching potential applications of LINC such as anti-fouling [3], anti-icing [4] and antimicrobial [5] surfaces have been investigated. Recent research has shown that LINC surface properties depend on lasing parameters such as laser power and speed [6]–[8]. For example, it was shown that by changing the lasing environment, namely the process gas, the contact angle between water droplets and LINC surfaces can be switched between superhydrophobic (>150°) and superhydrophilic[8]. To fully exploit this feature, we generate nanostructured carbon directly on the surfaces of the commercial Kapton films by surface lasing with the objective of tuning surface properties and morphology in a large range of contact angles based on changing laser power, speed, and line-to-line gap in the LINC process. We explore the lasing parameter space to achieve control over the resulting contact angle of direct written LINC areas. Here, we test the hypothesis of whether it is possible to achieve contact angle control by only changing the lasing parameters in an air

Ingenium 2020

Figure 1: (a) Image of laser head and polyimide with LINC. (b) Schematic of experiment slide with LINC patches at different orientations.

environment. By showing this, we simplify the process since no special environmental setup is needed. Additionally, we propose a potential mechanism to the observed difference in contact angle measurements at different parameters by comparing it with SEMs of the LINC areas.

2. Methods 2.1 Sample preparation The experiment was conducted by laser scribing on polyimide (PI) tape, also referred to as Kapton, (Grainger, Cat. No. 15C616, thickness: 3.5 mil), which was applied to a glass slide with a rubber roller to eliminate air bubbles. A chemical wash consisting of acetone and isopropyl was applied and then quickly dried off with compressed air to clean the PI surface. 2.2 LINC Processing The laser scribing was conducted using a CO2 laser system (Full Spectrum Laser Pro-Series 20x12, 45 W, 3.81 cm focus lens) with 10.6 µm wavelength and 45W laser power. The system allows tuning the power by controlling the laser current. The beam diameter (1/e2) was measured to be 110 µm. The samples were placed on an XYZ stage with a maximum horizontal rastering speed of 500 mm/sec. 2.3 Experiment design As shown in Figure 1a, each slide has 4 square patches (1.27 cm x 1.27 cm) that are lased at different conditions. A single LINC patch is formed by lasing in the positive x-direction a distance (L),

then shifting by an increment (g) in the y-direction, followed by lasing in the negative x-direction by the same distance (L) (Fig. 1b). This sequence is repeated until the square is completed. The laser machine input design file is generated by programming a MATLAB script with lines and a certain number of pixels gap between each line, which represents a single patch. The study was conducted by varying power (28W and 12W), speed (v, 500 mm/sec, 350 mm/ sec), and the distance between lines lased (g, 0 pixels, 3 pixels, 5 pixels, 7 pixels), where one pixel represents a shift of 0.00508 cm (Fig. 1). Finally, the patches were rotated 90° for a total of two orientations in order to investigate contact angle anisotropy (Fig. 1b). Four drops are dispensed for each condition. A control sample without any lasing was also used to establish an experimental control measurement. 2.4 Characterization methods Quantitative results were obtained by measuring the Young-Laplace’s contact angle using a Biolin Scientific Optical Tensiometer. The process included ejecting four 5 µl water droplets on each sample and recording the results at 6.9 frames per second for 20 seconds. Afterwards, the program OneAttention was used to analyze each droplet by curve fitting the Young-Laplace equation to the droplet to yield a contact angle for each of the 140 frames. The average of the 140 contact angles was then recorded as the contact angle for the droplet. To explore the morphology of the LINC patches, SEM images were taken with a Zeiss SIGMA VP Scanning Electron Microscope. 69

Figure 2: Box plot representing the contact measurements resulting at different laser powers, speed and line gaps. Insets represent images of contact angle measurements at a wicking condition and a hydrophobic condition.

3. Results

The control sample had an average contact angle of about 88°. As shown in Figure 2, the parallel orientation almost always had a higher contact angle measurement than its perpendicular counterparts. The lowest contact angle measured was 0°, at power 28W, speed 350 mm/sec, g = 0 pixels, both orientations (Fig. 2c, inset). The highest contact angle measured was 111°, also at power 28W, speed 350 mm/sec, but this time at g = 7 and the parallel orientation (Figure 2d, inset). It is observed there is a general trend that as gap size increases, so does contact angle. Optical tensiometry of drops at two selected patches at identical power and lasing speed (28W power, 500mm/ sec speed) but different gap sizes (0, 3) exhibit significantly different contact angles as shown in Figure 3a-b (ANOVA test, CAg=0=56.70,CAg=3=94.65, n==4, p = 0.0271). Optical images of the patch show that the patches have different colors and different surface appearances. In the higher gap sample (Figure 3d) the patterns representing the dried water spots can be noticed, whereas in the lower gap case, (Figure 3c), they are not noticed. Further investigation using SEM showed that there are vast differences in the morphology of the structures, even between g = 0 and g = 3 at the same settings. The sample with g = 0 has a porous appearance, whereas sample with g = 3 has a fibrous appearance.

4. Discussion

For some samples, such as power 28W, speed of 350 mm/ sec, g = 0, wicking behavior was observed. When wicking occurs, the water is absorbed spontaneously into the LINC patch, driven by capillary forces. To explain the difference in contact angle with the different lasing conditions, we relate the observed behavior with the morphology of the patches observed in SEM images. It is evident from the differences between the samples with g = 0 and g = 3 that gap size plays an important role in morphology. The sample with g = 0 has high overlap between lines (Fig. 3c, e) whereas the

70 Undergraduate Research at the Swanson School of Engineering

Figure 3: (a-b) Contact angle measurements of power 28W, speed 500 mm/sec, gap (a) = 0, gap (b) = 3. (c-d) Top view optical images of (a) and (b). (e, g, i) SEM images of (a) at 250x, 5,000x, and 25,000x, respectively. (f, h, j) SEM images of (b) at the same magnification.

Ingenium 2020

sample with g = 3 has less overlap (Fig.3 d, f). High overlap occurs when distance between the lased lines is smaller than the laser spot size. In this case (g = 0) since the spot diameter is around 200 m, and the distance between the lines is around 50 m, each spot on the line is lased multiple lines due to the overlap between the lasing runs. The high overlap in the sample with g = 0 resulted in a porous morphology, looking much like a sponge (Fig. 3g, i). This potentially enhances water absorption through capillary effects, leading to lower contact angles (~53°) as opposed to the sample with g = 3 that has a smaller overlap, which exhibited a larger contact angle (~100°) and a more fibrous morphology (Fig. 3h, j). The results show that diverse morphologies are possible with changing the laser parameters, which in turn clearly affect the contact angle. Further work is needed to explore the different morphologies and the mechanism by which they are formed and affect the surface properties of LINC.

[6] L. Andrea et al., “New insights on laser-induced graphene electrodes for flexible supercapacitors: tunable morphology and physical properties,” Nanotechnology, vol. 28, no. 17, p. 174002, 2017. [7] E. Vasile, S. M. Iordache, C. Ceaus, I. Stamatin, and A. Tiliakos, “Morphic transitions of nanocarbons via laser pyrolysis of polyimide films,” J. Anal. Appl. Pyrolysis, vol. 121, pp. 275–286, 2016. [8] Y. Li et al., “Laser-Induced Graphene in Controlled Atmospheres: From Superhydrophilic to Superhydrophobic Surfaces,” Adv. Mater., vol. 29, no. 27, pp. 1–8, Jul. 2017.

5. Conclusion

We demonstrate that varying the gap between nanostructured carbon lines obtained by laser-induced nanocarbon formation (LINC) on polyimide surfaces has a significant impact on the contact angle of water. Our results show that a wide range of contact angles can be achieved, ranging from hydrophilic approaching 0° to hydrophobic surfaces approaching 120°. Characterizing the morphologies using a combination of optical and electron microscopy also related these properties to the structure of LINC. This highlights the versatility of our technique, which is promising for tuning the surface properties of commercial polymers for various applications, such as self-cleaning, anti-fouling, and water-repellant surfaces.

6. Acknowledgements

This research was funded by the Swanson School of Engineering and the Office of the Provost.

7. References

[1] J. Lin et al., “Laser-induced porous graphene films from commercial polymers,” Nat. Commun., vol. 5, pp. 1–8, 2014. [2] L. X. Duy, Z. Peng, Y. Li, J. Zhang, Y. Ji, and J. M. Tour, “Laser-induced graphene fibers,” Carbon N. Y., vol. 126, no. 7, pp. 472–479, Jan. 2018. [3] S. P. Singh, Y. Li, J. Zhang, J. M. Tour, and C. J. Arnusch, “Sulfur-Doped Laser-Induced Porous Graphene Derived from Polysulfone-Class Polymers and Membranes,” ACS Nano, vol. 12, no. 1, pp. 289–297, Jan. 2018. [4] D. X. Luong et al., “Laser-Induced Graphene Composites as Multifunctional Surfaces,” ACS Nano, vol. 13, p. acsnano.8b09626, Feb. 2019. [5] S. P. Singh, Y. Li, A. Be’er, Y. Oren, J. M. Tour, and C. J. Arnusch, “Laser-Induced Graphene Layers and Electrodes Prevents Microbial Fouling and Exerts Antimicrobial Action,” ACS Appl. Mater. Interfaces, vol. 9, no. 21, pp. 18238–18247, May 2017.


The role of oxygen functional groups in graphene oxide modified glassy carbon electrodes for the electrochemical sensing of nicotinamide adenine dinucleotide hydride Ananya Mukherjeea, Yan Wanga, and Leanne Gilbertsona, Environmental Engineering Laboratory, Department of Civil and Environmental Engineering, University of Pittsburgh, PA 15260, USA


Ananya Mukherjee is junior year environmental engineering student at the University of Pittsburgh. Inspired by the wetland ecosystems of her hometown in Princeton, NJ, she hopes to pursue a career related to sustainability and environmental design.

Ananya Mukherjee

Yan Wang is a PhD candidate in the Department of Civil and Environmental Engineering at the University of Pittsburgh. She obtained her bachelor’s degree at Sun Yat-sen University and master’s degree at Zhejiang University in China. Her current research interests are related to sustainable design of graphene-based nanomaterials for advanced energy, environmental, and biological applications.

Yan Wang

Leanne Gilbertson is an Assistant Professor of Environmental and Sustainable Engineering. She joined the faculty at Swanson School in the fall of 2015 after completing her postdoc and PhD, both at Yale University Department of Chemical and Environmental Engineering. Her research group focuses on the sustainable design of emerging materials and products for applications at the nexus of the environment and public health.

Leanne Gilbertson

Significance Statement

Epoxides on graphene oxide (GO) modified glassy carbon electrodes (GCEs) appear to be more electrochemically active than other functional groups in the oxidation of nicotinamide adenine dinucleotide hydride (NADH). This suggests that better NADH sensors can be designed by incorporating GO engineered with high epoxide content.

Category: Device design

Keywords: NADH, graphene oxide, glassy carbon electrodes, oxygen functional groups Abbreviations: Nicotinamide adenine dinucleotide hydride (NADH), nicotinamide adenine dinucleotide (NAD+), graphene oxide (GO), as received graphene oxide (ARGO), thermally annealed graphene oxide (TGO), carbon nanotubes (CNTs), cyclic voltammogram (CV)

72 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020


Nicotinamide adenine dinucleotide hydride (NADH) plays an important role in cellular metabolizing processes that generate energy. NADH levels can be indicative of cellular health so researchers are interested in developing better sensors for monitoring NADH. Electrochemical sensors are economically advantageous, but the irreversible oxidation process results in accumulation of nicotinamide adenine dinucleotide (NAD+) at the electrode surface. This leads to electrode fouling that inhibits further oxidation. Graphene oxide (GO) is a popular nanomaterial often used for modifying electrode surfaces to speed up electron transfer and prevent fouling. This research was focused on identifying what oxygen groups on GO are responsible for the enhanced sensing of NADH. Glassy carbon electrodes (GCEs) were modified with thermally annealed GO (TGO) and as received GO (ARGO). ARGO was found to cause faster oxidation in NADH and differed in composition from TGO mainly in its high percentage of epoxide moieties. This indicates that epoxide groups on GO may facilitate electron transfer in NADH. Should further research confirm these findings, more effective sensors could be designed by incorporating GO with high epoxide content.

1. Introduction 1.1 Medical Relevance The oxidation of NADH to NAD+ plays an important role in cellular metabolizing processes that generate energy in living cells. NADH is essential for mitochondrial respiration to produce ATP, as well as for over 200 other enzymatic reactions. [1] The levels of both NADH and NAD+ can be indicative of cellular health. In a study looking at NAD+ levels in patients with multiple sclerosis (MS) it was found that NAD+ levels declined as the diseases progressed. [2] Healthy controls had the highest NAD+ levels, followed by patients with relapsing remitting MS, secondary progressive MS, and primary progressive MS. The study concluded that reduced NAD+ levels indicate a loss of cellular function and metabolism which culminate in cell death. NAD+ naturally declines in animals and in humans during normal aging. It has been shown that enhancing NAD+ levels in mammals can improve mitochondrial function under cellular stress. This leads to improved protection against dietary and age-related metabolic complications. There is currently a lot of interest in researching more about NADH/NAD+ for possible therapeutic uses in treating a variety of diseases as well as improving overall health in elderly people. [3]

1.2 Graphene Modified Electrochemical Sensors Understandably, a lot of interest has been shown in developing sensors to better monitor NADH levels as they can be indicative of cellular health. Electrochemical sensing has been of particular interest to researchers because it is more cost effective, user friendly, and works with smaller sampling volumes than other sensing methods. [4] The problem with using electrochemical sensors is that NADH has sluggish electron transfer rates which leads to build up of the oxidation product NAD+. Fouling then occurs as NAD+ adsorbs onto the electrode surface and blocks electron transfer sites, hindering further oxidation. One method to speed up the reaction and reduce fouling is by modifying the electrode surface. GO is a popular nanomaterial often used for electrode modification because of its unique electrochemical properties. Graphene’s honeycomb lattice of carbon atoms has delocalized pi orbitals that allow electrons to move freely above and below the plane, making it extremely conductive. GO has many oxygen groups that provide abundant electron transfer sites which help to speed up oxidation. Zhang et al. has suggested that NADH oxidation is mediated by abundant oxygen groups at edge planes in GO. [5] However, there has yet to be research focused on identifying which oxygen groups are responsible for the enhanced oxidation at GO modified GCEs. The goal of this research is to inform better sensor design by identifying which functional groups in GO mediate electron transfer in NADH. Once active groups are identified, sensors can be optimized by modifying GO’s material properties to incorporate higher distributions of oxidation mediating moieties. It has been reported that NADH oxidation at ordered carbon nanotube (CNT) modified electrodes is mediated by quinone moieties. [6] Because CNTs are similar to GO in composition, we hypothesize that quinones mediate oxidation in GO as well. 1.3 Approach In this experiment ARGO and TGO dispersions will be drop casted in order to produce modified GCEs. The electrodes will be placed in solutions of known NADH concentration. CVs will be obtained by monitoring the current in response to an applied potential that steadily increases (anodic cycle) then decreases back to the initial value (cathodic cycle). When NADH is near the modified electrode surface, the current will increase indicating that it is being oxidized. The potential at which the current reaches a maximum value (peak potential) will be used to characterize the speed of the oxidation at either surface. A peak in the CV curve indicates that the majority of NADH near the electrode surface is oxidized. The magnitude of the peak will indicate how many NADH


molecules have oxidized, but what this research is interested in is the voltage at which the peak occurs. A fast reaction will be indicated by a peak that is generated early in the anodic cycle at a low potential i.e. a more negative peak potential indicates better electron transfer (Figure 1). The peak potentials of ARGO and TGO will be analyzed in conjunction with the surface group compositions of both materials (obtained by Wang et al. [7]) in order to determine what functional groups are electrochemically active.

Figure 1: Typical CVs of reactions occurring at different rates and corresponding peak potentials

Figure 2: ARGO/TGO dipersions (top) and ARGO/TGO modified electrodes showing molecular depictions of oxygen content (bottom)

2. Methods 2.1 Experimental Procedure Graphene ink dispersions were prepared using 2 mg of either TGO powder annealed at 600°C, or ARGO powder. 1.2 mL of DI water was added to the powder, followed by 792 μL isopropanol to aid with dispersion and 8 μL of nafion perfluorinated resin to act as a binding agent to the electrode surface, giving a concentration of 1 mg/mL. Both dispersions were probe sonicated at 10% power (400 W max power output) three times in 5 minute segments and cooled down in an icebath for 2-3 minutes in between rounds. Afterwards, both samples were placed in a bath sonicator (135 W) for 1 hour, resulting in well dispersed solutions. The ARGO and TGO dispersions are shown below (figure 2). Abundant oxygen groups were responsible for the brown color in the ARGO dispersion. On the other hand, TGO had most of its oxygen groups annealed off, resulting in a gray dispersion.

74 Undergraduate Research at the Swanson School of Engineering

GCEs were polished by rubbing 0.05 μM alumina slurry on microfiber cloth in a circular motion for 1-2 minutes. Electrodes were then cleaned off with isopropanol and then DI water. 10 μL of each dispersion was dropcasted onto a GCE and air dried. A 0.1 M phosphate buffer solution (PBS) at pH 7 was used for the electrolyte. The PBS was prepared by pouring 34.8 mL of 1 M K2HPO4 and 15.2 mL of KH2PO4 into a 500 mL flask and then filling the rest with DI water. 76 mL of the prepared PBS solution was used in each electrochemical cell and purged with Nitrogen gas for about 20 minutes. Afterwards the cell was cleaned by conducting a CV scan from -0.4 V – 0.9 V. The electrochemical experiments were conducted using a standard three electrode system with Ag/AgCl reference electrode, Pt wire external electrode, and ARGO or TGO modified GCE working

Ingenium 2020

electrode (Figure 3). The analyte was prepared by dissolving 64 mg of NADH in 4.5 mL PBS. 4 mL of the prepared NADH solution was added to the 76 mL PBS electrolyte to give a 1 mM concentration. CV scans of the 1 mM NADH cell were carried out with a 10 mV/s scan rate and a potential range of -0.4 V – 0.9 V.

3. Results

The peak potential for the ARGO/GCE was found to be 0.492 V, more negative than the TGO modified GCE which had a peak potential of 0.612 V (Figure 4).

Figure 4: Cyclic voltammograms of ARGO (red) and TGO (blue) modified GCE in the presence of 1 mM NADH in 0.1 M PBS from -0.4 V – 0.9 V with a scan rate of 10 mV/s.

Wang et al. used x-ray photoelectron spectroscopy (XPS) to obtain oxygen group distributions for materials similar to the ones used in this experiment. [7] The XPS results show that ARGO has similar C=O (quinone) and COOH (carboxyl) distributions, but substantially more C-O (epoxide) groups compared to TGO (Figure 5). Figure 3: NADH electrochemical cell setup with standard three electrode system

2.2 Quantifying Peak Potential Three separate overlapping CV scan measurements were used for the analysis after the system stabilized. In each case the CVs showed one nonreversible oxidation reaction resulting in a single current peak during the anodic cycle. The derivative graphs of the anodic scans were analyzed in order to quantify the peak potentials. The graphs were obtained by plotting the change in current (I) divided by the change in potential (V) on the y axis and the corresponding midpoint potential on the x axis. The peak potential was taken as the potential at which the derivative touched or crossed the x-axis after reaching a maximum.

Figure 5: Percent oxygen group distributions in ARGO and TGO obtained by XPS [7]


4. Discussion

The more negative peak potential at ARGO compared to TGO indicates faster electron transfer at ARGO. Since both materials had similar quinone distributions the results counter the initial hypothesis that quinones in GO mediate NADH oxidation in the same way they do in ordered CNTs. Because ARGO and TGO mainly differed only in their epoxide content, it seems logical to conclude that epoxides are more active in NADH mediation on GO. Further research will be needed to confirm that epoxides are truly the mediators in GO dispersions. If this is the case, then sensors should be engineered with high epoxide content GO and tested to see what the optimal material composition is for obtaining the fastest oxidation rate. Since our results indicated different functional group activity for GO than CNTs it seems as though composition might not be the only factor in determining which groups are most active in the oxidation process. One possible explanation is that the difference in shape between CNTs and GO may impact the activity of certain groups. The curved CNT structure could alter the orientation of moieties that otherwise stick out straight from the flat GO surface. As a consequence, some functional groups might have increased/ decreased exposure to NADH based on whether they are attached to a curved or flat surface. More research is necessary in order to determine if physical structure has any impact on surface group activity. Although ARGO was found to have a lower peak potential, the current peak was less defined and lower than that of TGO. The low magnitude of the ARGO peak indicates that fewer NADH molecules overall were able to oxidize at the surface, although the reaction proceeded at a faster rate. Additional experiments would be necessary to see if this low peak can be ascribed to adsorption of NAD+ at the electrode surface.

5. Conclusions

This research suggests that NADH oxidation in GO modified GCEs is mediated by epoxide moieties at the GO surface. GO/GCE NADH sensor designs might be improved by maximizing epoxide distributions to increase oxidation rates. Improved sensor designs could greatly aid research into the effects of NADH on human health as well as possible therapeutic uses. Further tests will be needed in order to confirm this study’s results. A diminished current peak was also observed in the ARGO/GCE. Additional research will be necessary in order to determine the cause of the lower current rise.

6. Acknowledgments

This research was conducted under the supervision of Dr. Gilbertson and mentorship of Yan Wang. Funding was provided by Dr. Gilbertson, the Swanson School of Engineering, and the Office of the Provost.

76 Undergraduate Research at the Swanson School of Engineering

7. References

[1] R.A. Fricker, E.L. Green, S.I. Jenkins, S.M. Griffin, The Influence of Nicotinamide on Health and Disease in the Central Nervous System, Int. J. Tryptophan. Res. 11 (2018), 1-11 [2] N. Braidy, C.K. Lim, R. Grant, B.J. Brew, G.J. Guillemin, Serum nicotinamide adenine dinucleotide levels through disease course in multiple sclerosis, Brain. Res. 1537 (2013), 267-272 [3] C. Canto, K.J. Menzies, J. Auwerx, NAD+ metabolism and the control of energy homeostasis - a balancing act between mitochondria and the nucleus, Cell. Met. 22.1 (2015) 31-53 [4] S.A. Kumar, S. Chen, Electroanalysis of NADH Using Conducting and Redox Active Polymer/Carbon Nanotubes Modified Electrodes-A Review Sensors, 8 (2008), 739-766 [5] L. Zhang, Y. Li, L. Zhang, D. Li, D. Karpuzov, Y. Long, Electrocatalytic Oxidation of NADH on Graphene Oxide and Reduced Graphene Oxide Modified Screen-Printed Electrode, Int. J. Electrochem. Sci. 6 (2011), 819-829 [6] J. Chen, J. Bao, C. Cai, T. Lu, Electrocatalytic oxidation of NADH at an ordered carbon nanotubes modified glassy carbon electrode., Anal. Chim. Acta. 516 (2004) 29-34 [7] Y. Wang, L. M. Gilbertson, Informing rational design of graphene oxide through surface chemistry manipulations: properties governing electrochemical and biological activities, Green. Chem. 19 (2017) 2826-2838

Ingenium 2020

Characterization of hierarchical structures in remelted Ni-Mn-Ga substrates for directed energy deposition manufacturing of single crystals Tyler Paplhama, Jakub Tomana, Markus Chmielusa a Department of Mechanical Engineering and Materials Science, University of Pittsburgh, Pittsburgh, PA 15261

Tyler Paplham

Tyler is a Materials Science and Engineering major from Buffalo, NY. He has worked in the Advanced Manufacturing and Magnetic Materials lab for nine months. After graduation, he plans to pursue his PhD in Materials Science and Engineering to research materials for sustainable energy generation and transmission.

Dr. Markus Chmielus has been an Associate Professor in the Mechanical Engineering and Materials Science Department since 2013. He previously worked at Cornell University (postdoc) and the Helmholtz Center for Materials and Energy in Germany (research Dr. Markus Chmielus scientist), and has degrees from the Technical University of Berlin, Boise State University, and the University of Stuttgart in Materials Science and Aerospace Engineering. He now leads the Advanced Manufacturing and Magnetic Materials Laboratory (AM³), where his group members perform experimental additive manufacturing and post-processing research on functional magnetic and structural metals.

Significance Statement

This work aims to provide a deeper understanding of how varying laser power and travel velocity affects the resolidification behavior in directed energy deposition, with a goal of successfully producing additively manufactured single crystals. Trends in properties of the remelted substrate as functions of each variable were suggested.

Category: Experimental research

Keywords: Additive manufacturing, directed energy deposition, magnetic shape-memory alloys


Magnetic shape-memory alloys (MSMAs) such as Ni-Mn-Ga Heusler alloys have been of great interest in the past decade due to the observed inverse magnetoplastic (IMP) effect, wherein an applied mechanical stress reorients the direction of easy magnetization of the crystal. However, practically large strains achieved with this mechanism may only be generated from single crystals, the manufacturing of which is traditionally time-intensive and delicate. A possible approach for more quickly producing single crystal MSMAs is to use directed energy deposition (DED), a type of laser additive manufacturing. This study focused on the laser remelting of an austenitic Ni51Mn24.4Ga24.6 single crystal substrate without deposition of an additional layer, so as to observe the effects of varying laser power and travel velocity on the resulting melt pool. Using digital microscopy and image analysis, various properties of the melt pool and contained microstructures were measured as functions of laser power and travel velocity. Some trends were suggested that provided insight into which parameter values are conducive to retention of a single crystal; however, a more comprehensive study is needed to confirm these trends.

1. Introduction

Magnetic shape-memory alloys are a unique class of metals which exhibit large reversible plastic deformations (up to 6-12% strain) upon a change of direction of easy magnetization in the crystal, which may be induced by a changing external magnetic field or an applied mechanical stress. This is made possible due to the presence of martensitic twin variants, which are alternating regions within the crystal structure with differing directions of easy magnetization. The lattices in each region are misaligned by almost ninety degrees, such that the twin boundary defines a dislocation line. Upon the application of an external magnetic field or mechanical stress, these boundaries “travel” throughout the crystal in a continuous series of dislocations, resulting in a macroscopic plastic deformation as the crystal structure transitions from one alignment to another [1,2]. These materials are of great interest due to their possible applications in powering sensors, actuators, and other small devices through the magnetic field induced strain and the IMP effect [2]. While the amount of power produced is small, MSMAs have notably high fatigue resistance and high energy density [3,4]. However, even this very small amount of power requires that the MSMA be a single crystal. Grain boundaries, inclusions and other dislocations introduce “pinning sites” which inhibit the motion of twin boundaries, limiting the magnetoplastic strain and thus the amount of power that can be produced. Single crystals are traditionally produced via Bridgman crystal growth, by which an ampule of molten metal is extremely slowly drawn across a very steep temperature gradient [5]. This is a time-intensive process that also leads to macrosegregation and thus inconsistency in composition across the crystal [6]. Additive manufacturing, specifically DED, may provide an avenue for more quickly producing single crystals with a lower level of macrosegregation. In this laser additive manufacturing method, 77

nozzles adjacent to the laser spray converging jets of metal powder into the laser beam, where it is melted along with the substrate surface to form a new layer of deposited material [7]. Before this method can be fully utilized in the printing of Ni-Mn-Ga single crystals, it must first be understood how the thermal history from the DED process affects the microstructures formed after remelting and deposition of a new layer. This experiment focused only on the remelting process, so as to determine individually the effect of varying the laser parameters on remelting of the original substrate without having to consider the complicating addition of a newly deposited layer of powder.

3. Results

Melt pools were observed in six of the tracks and are shown in Fig. 1. The remaining three tracks, all with a power of 100W, did not exhibit a melt pool and therefore are not shown.

2. Materials and Methods

The substrate used was assumed to be an austenitic single crystal and had a nominal composition of Ni51Mn24.4Ga24.6. An Optomec laser engineered net shaping (LENS) 450 system was used to remelt the substrate. For each track, the nominal laser power and travel velocity were chosen, then the laser was positioned and started a distance away from the substrate so that the full velocity had been reached by the time the laser made contact with the substrate. The laser was then shut off again a distance away from the substrate, then repositioned for the next track. Eight parallel tracks were made on the substrate surface. The substrate was then sectioned along a plane parallel to the top surface, revealing undisturbed material on which a ninth track (350-10) was made. The laser power and travel velocity were varied, and the chosen combinations are listed in Table 1. Each track was cut perpendicular to the travel velocity at approximately the halfway point in the track, mounted, polished, and etched to reveal the melt pool. The melt pools were then imaged on a digital microscope and analyzed in ImageJ. Laser Power [W]

Velocity [mm/s]


0.5 1 2.5




1 2.5 5 10


Figure 1: The six melt pools with parameters listed as power-velocity. The outline of each meltpool is shown in blue, while regions of different structure are outlined in yellow. All scale bars (lower right of each sub-image) read 200 m.

Of great importance to the determination of appropriate parameters for single crystal remelting and deposition are the nature of the hierarchical structures within the melt pool, namely the planar solidification region (PSR) and dendrites. The individual effects of laser power and laser travel velocity on the thickness of the PSR and the normalized depth of transition from [100] dendrites, oriented out of the page in Fig. 1, to [001] dendrites, oriented toward the top of the page in Fig. 1, are plotted in Fig. 2 and Fig. 3, respectively, where the normalized depth of transition is defined as the ratio of the depth of the [100] dendrites to the total depth of the melt pool.

Figure 2: Thickness of planar solidification region (PSR) as a function of laser power (left) and laser travel velocity (right). Note that tracks which did not exhibit melt pools are excluded.


Table 1: Laser power and travel velocity combinations for each track.

Figure 3: Relative depth of transition from [100] dendrites to [001] dendrites as a function of laser power (left) and laser travel velocity (right). As before, tracks which did not exhibit melt pools are excluded.

78 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

The thickness of the PSR appeared to increase with increasing laser power and decreasing travel velocity. However, there seems to be some minimum power or maximum travel velocity required for formation of the PSR, as the 200-2.5, 250-5, and 250-10 tracks did not have observable PSFs, suggesting that a threshold value for formation lies in the range of 200-250W and 2.5-5mm/s for varying laser power and travel velocity respectively. The relative depth of transition from [100] dendrites to [001] dendrites also increased with increasing power and decreasing velocity. A qualitative analysis was also performed on the observed dendrite angles. Dendrite regions, defined by the yellow subdivisions in Fig. 1, were characterized as vertical, diagonal, horizontal (all within the plane of the page), or [100] (out of the page). The melt pools were assumed to be sufficiently symmetric that no insight was lost by combining the right and left diagonal or horizontal (or vertical in the case of 250-1) groups into merely “diagonal” or “horizontal” (or “vertical”). The percentages of each type of dendrite orientation for the 250W melt pools are plotted in Fig. 4, which shows the effect of varying laser travel velocity on dendrite orientations with power held constant. A similar plot of the 2.5mm/s melt pools at 200W and 250W is also shown in Fig. 4, although this is less enlightening because only two tracks were created at the same velocity (at least out of those which resulted in melt pools).

Figure 4: Percentages of various dendrite orientations at constant power of 250W (top) and constant travel velocity of 2.5mm/s (bottom).

It is difficult to discern any conclusion from Fig. 4 other than the behavior of the planar solidification front, which was already shown in Fig. 2. It is interesting that 250-1 and 250-5 exhibit diagonal and horizontal dendrite groups but 250-2.5 does not.

A satisfactory albeit simplified explanation for the trend in relative depth of the [100] to [001] transition concerns how adjusting the laser parameters affects the shape of the melt pool. High power and low velocity result in a deep, abrupt melt pool whereas low power and high velocity result in shallower, more gradual melt pools. The solidification front advances with a velocity normal to itself, but the allowable directions of dendrite growth are limited to <100>. The dendrite growth must keep up with the solidification front, so the direction selected will be that which requires the lowest growth rate for the dendrites. As can be seen in Fig. 5, an abrupt, deep melt pool such as in the (250-1) and (350-10) tracks will have more dendrite growth in the [100] direction, whereas a shallow, gradual melt pool such as in (2502.5) and (250-5) will have more dendrite growth in the [001] direction. If a deep melt pool and shallow melt pool are sectioned at approximately the same distance along the track, this manifests as the deep melt pool showing a greater presence of [001] than shown by the shallow melt pool [8]. The behavior of the PSR is also complex. A high thermal gradient promotes planar growth, as thermal effects dominate over other influences and thus best preserves a quasi-steady state environment. According to the thermal model used in Gäumann et al., the Rosenthal solution predicts that the average of the ratio G3.4⁄V decreases with increasing power, where G is the thermal gradient and V is the velocity of the solidification front. This is in disagreement with what was observed in the present study, where planar growth increased with increasing power. However, it is crucial to note that the Rosenthal solution predicts that the average G3.4⁄V decreases with increasing power. The value of G3.4⁄V is very high at the beginning of solidification, but then decreases rapidly at a slowing rate with increasing z, where z is the depth from the surface of the melt pool (i.e. any point in the melt pool has a negative z-value) [9]. It is hypothesized that perhaps increasing the power results in a slower initial rate of decreasing G3.4⁄V. Because of the shape of the curve, this could result in maintaining a thermal gradient sufficient for planar growth across a greater range of z while still showing a lower overall average value.

4. Discussion

The trends shown in Fig. 2-3 are consistent with expected resolidification behavior. As a crystal solidifies, the solidification front is initially planar, as this results in equal surface energy across the entire front. However, protruding points caused by thermodynamic variation along the surface will be supported by constitutional undercooling present ahead of the interface. Thus, the planar front breaks down. Solidification tends to be more energetically favorable along the crystal lattice axes (usually defined along <100>), resulting in pointed structures called dendrites along the lattice axes. Varying the laser power and travel velocity alters melt pool shape, thermal gradient, velocity of the solidification front, etc. These parameters drive transition from planar to dendritic fronts and the selection of dendrite growth axis.

Figure 5: Schematic of preferred dendrite growth directions in deep, abrupt melt pools (left) and gradual shallow melt pools (right). Direction of easy growth, i.e. the observed growth direction, is shown in green, while the rejected direction is shown in red. The [010] direction has been omitted for simplicity.


5. Conclusion

Overall, the effects of varying laser power and travel velocity on many properties of the melt pools and their contained structures were able to be analyzed either quantitatively or qualitatively. Some trends were implied, many of which agreed with prediction. However, due to the small amount of available substrate and therefore the low number of samples, such trends would be highly vulnerable to any outliers or errors in measurement and are thus not reliable. A more comprehensive study in which either laser power or travel velocity in smaller increments should be performed to confirm the suggested trends. Additionally, since the overall goal of this project is to print single crystals with DED, a similar study should be conducted with samples that were remelted and had an additional layer of powder deposited.

6. Acknowledgements

This project was funded by NSF grant #1808082 including an REU supplement and the Mascaro Center for Sustainable Innovation. Additionally, the authors would like to thank all colleagues in the Advanced Manufacturing and Magnetic Materials lab for support during this project.

7. References

[1] Chmielus, M. (2010) “Composition, structure and magnetomechanical properties of NiMn-Ga magnetic shapememory alloys.” Logos Verlag Berlin. [2] Carpenter, D. (2008) “The application of ferromagnetic shape memory alloys in power generation devices”. Boise State Uni. Th. And Diss. 544 [3] Chernenko, V. A. et al. (2009) “Large magnetic-fieldinduced strains in Ni-Mn-Ga nonmodulated martensite”. Appl. Phys. Lett. 95 [4] Rizzello, G. et al. (2017) “An overview on innovative mechatronic actuators based on smart materials”. IEEE Africon 2017 (pp. 450-455) [5] Hage-Ali, M. & Siffert, P. (1995) Growth Methods of CdTe Nuclear Detector Materials. In T. D. Schlesinger, R. B. James (Eds.) Semiconductors and Semimetals (pp. 219-255) [6] Boonyongmaneerat, Y. et al. (2007) “Increasing magnetoplasticity in polycrystalline NiMn-Ga by reducing internal constraints through porosity”. Phys. Rev. Lett. 99 [7] Liu, D. et al. (2014) “Laser engineered net shape (LENS) technology for the repair of Nibase superalloy turbine components”. Metall. Mat. Trans. A 45 (pp. 4454-4469) [8] Rappaz et al. (1989) “Development of microstructures in Fe-15Ni-15Cr single crystal electron beam welds”. Metall. Mat. Trans. A 20A (pp. 1125-1138) [9] Gaümann, M. et al. (2001) “Single-crystal laser deposition of superalloys: processing microstructure map”. Acta Materialia 49 (pp. 1051-1062)

80 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

Wireless signal transmission through hermetic walls in nuclear reactors Jerry Pottsb, Yuan Gao, Heng Bana Multiscale Thermophysics Laboratory, bDepartment of Mechanical Engineering and Materials Science University of Pittsburgh, PA, USA


Jerry Potts is from Bensalem, Pennsylvania and is currently pursuing a Bachelor’s degree in Mechanical Engineering with a minor in physics. He has worked as a research assistant in the Multiscale Thermophysics Lab under Dr. Heng Ban since January of 2019. He plans Jerry Potts to continue his continue his education by pursuing a Ph.D. in Mechanical Engineering with a focus in clean energy systems.

Heng Ban

Dr. Heng Ban is the R.K. Mellon Professor in Energy at the University of Pittsburgh. He is the head of the Multiscale Thermophysics Lab, whose research includes developing novel techniques for measuring thermophysical properties at the micro-scale, as well as in-pile instrumentation to study nuclear fuels and materials.

Significance Statement

Using wired sensors in nuclear reactors leads to extremely high costs and downtime for maintenance. Wireless sensors are a promising solution, but can have difficulty operating in a radiative environment. This study seeks to verify wireless inductive transmission as an alternative and indicates it could be a viable method.

Category: Device design

Keywords: Wireless signal transmission, inductive coupling, LVDT, nuclear reactor


Nuclear reactors rely on traditional wired sensors for fuel and system monitoring. As wireless technology becomes more developed, it could potentially be a reliable workaround to the issues of maintaining wired sensors. However, many methods of wireless signal transmission are unable to operate properly in the harsh environment of the reactor. This paper assesses the viability of the use of near-field inductive coupling as an alternative wireless transmission method to measure fuel parameters in a nuclear reactor. A prototype was developed where wireless inductive coupling was used to supply power to a linear variable differential transformer (LVDT), which can be used to measure various fuel parameters, and to then record the output of the system. These tests were primarily concerned with the linearity, uncertainty, and repeatability of the measurements. Further testing was then conducted to observe if the wireless transmission would be able to penetrate the cladding of a nuclear fuel rod. The resulting data indicated that the measurements were highly repeatable and had a very strong linearity throughout the experiment. There was, however, a considerable increase in the uncertainty of the system.

1. Introduction

Equipment and fuel monitoring in nuclear reactors currently depend on hundreds of various sensors. The installation of these sensors can be very costly and lead to significant system downtime due to maintenance [1]. Wireless sensors can serve as a cheaper, more reliable option to the wired sensors that are currently used. These sensors are able to circumvent issues such as feedthroughs penetrating pressure barriers, corrosion, and other forms of cable degradation [2]. However, due to the harsh radiative environment of a nuclear reactor, the system needs to be encased in stainless steel cladding in order to protect the sensor components. So, any wireless sensors developed need a signal strong enough to penetrate its own cladding as well as the cladding of the fuel, neutron moderators, and potentially the wall of the reactor itself. This greatly limits the available methods of wireless transmission which can be used in this application. However, near-field inductive coupling can serve as a high efficiency wireless transmission mechanism for in-pile measurements of fuel parameters [3]. The tight electromagnetic coupling of the inductor pair maximizes the mutual inductance of the two coils and would allow the signal to penetrate the stainlesssteel cladding of the fuel rod. The ability of a signal to pass through a material can be quantified by that material’s skin depth, or the distance through a material at which the signal strength has decayed by a factor of e. For metals the skin depth is on the scale of micrometers, but the exact value is inversely proportional to the square root of the signal’s frequency [4]. So, by minimizing the operating frequency of the sensor the skin depth of the cladding will increase and limit the signal decay. This also ensures the sensor can be contained in its own cladding to avoid exposure to the reactor coolant and minimize the impact of external noise from the radiative environment without preventing signal transmission. 81

The purpose of this project was to design a system to assess the validity and accuracy of signal transmission using wireless inductive coupling. The system was designed around the use of a linear variable differential transformer (LVDT) as the sensing mechanism. An LVDT was selected for this system as it was already proven to be able to withstand the harsh environment of a nuclear reactor during the Halden Reactor Project [5]. Thus, using an LVDT makes this wireless system more applicable to short-term applications in reactors. If this prototype is successful, a new prototype will be developed using the same setup in order to further examine the signal transfer in more reactor-like conditions.

(3) For signal transfer, we need to model a second set of coils to transfer the output of the LVDT out of the circuit. The system can be modeled such that the LVDT circuit is the primary circuit providing the source voltage and the secondary circuit consists solely of the inductor coil L2. By doing so, the above equations can be rearranged to determine the output voltage from the wireless signal transfer, as shown in Equation 4. (4)

2. Methods

2.1 Theoretical Model of Wireless Signal Transfer Mutual inductance occurs when an alternating current in a coil induces a voltage across the ends of adjacent coils via its changing magnetic field. The strength of the mutual inductance depends on the spatial relationship between the two inductors, namely distance and orientation. This mechanism can be modeled using the equivalent circuit shown in Figure 1 [6].

2.2 Experimental Setup Using the theoretical models described in section 2.1, an experimental prototype was designed to verify the inductive coupling. The electrical model of this system is presented in Figure 2.

Figure 2: Electrical model of the prototype. Figure 1: An equivalent circuit showing the principle of mutual inductance.

By applying Kirchhoff’s voltage law to this system, we can derive equations to represent the coupling of these circuits as shown in Equations 1 and 2. (1)


Where L1, R1, and I1 refer to the inductance, resistance, and current of the primary circuit respectively and L2, R2, and I2 are the inductance, resistance, and current of the secondary circuit. Vs refers to the supply voltage, ω refers to the system frequency, M is the mutual inductance between coils L1 and L2, and Lm is the inductance of the LVDT. These equations can then be used to quantify wireless signal transfer as well as power transfer, both of which are used in this prototype. To supply power to the prototype, Equations 1 and 2 can be used to calculate the voltage across the LVDT—which is represented as the voltage drop across the inductor Lm in Figure 1—and the required source voltage, as shown in Equation 3.

The AC power source is used to provide power to the LVDT through the first set of inductor coils. Then, the output of the LVDT is transferred to a third circuit through another set of coils and read out via a multimeter. The LVDT used in this system is an HR500 from TE Connectivity. The core displacement ranges from 0 to 25.4 mm with a maximum non-linearity of 0.25%. This corresponds to a maximum uncertainty of 32 µm for the core displacement.

82 Undergraduate Research at the Swanson School of Engineering

Figure 3: (a) CAD model of the prototype sensor (b) 3D printed prototype partially assembled for initial testing.

Ingenium 2020

A physical model of system in shown in Figure 3a and is color-coded to better understand the various sections the prototype. The red section is the housing for the circuitry used to provide power and connect to the multimeter, while the dark blue section houses the LVDT assembly respectively. A hexagonal shape was chosen for these housings to represent the cladding of a nuclear fuel rod. The yellow components make up the mechanism through which the iron core of the LVDT is moved and are supported by the green components. A fine-pitch screw is used to externally control the core displacement and allows us to easily calculate the core displacement based on the number of turns of the screw. Lastly, the violet components represent the inductor pair responsible for power transmission, while the light blue indicates the pair of coils used for signal transmission. The inductor pairs have been designed such that a smaller coil is resting concentrically inside the larger coil to maximize the coupling coefficient between the two. The smaller coil is also wrapped around a ferritic core in order to maximize the inductance of the individual coils and the overall mutual inductance. The components shown in Figure 3a were produced using a 3D printer. For the purposes of these tests, the system was only partially assembled to support the LVDT and transmission coils, as shown in Figure 3b. Once assembled, the output voltage of the system was recorded over the full range of the core displacement for different input voltages from 1 to 4 V. The voltage was recorded after every half rotation of the screw for three trials per voltage tested. The input was provided at 5 kHz for every trial based on the recommended inputs of the LVDT used in the experiment. Linear regressions were generated for each set of data and then divided by their input voltage. The coefficients of these equations were then averaged together to obtain a general equation for the output voltage of the system. The standard deviation of each data point was also calculated, the highest of which was used to calculate the uncertainty of the overall system. Further testing was conducted to assess the signal’s ability to transmit through potential electromagnetic shielding. In this test, a 9/16” cylinder of 304 stainless steel was placed in the center of the inductor pair used for signal transfer to disrupt its coupling. 304 stainless steel was chosen to conduct this test as it is a common cladding material for nuclear fuel rods. The data collection process was identical to that of the previous test using an input voltage of 3 V. The regression of this data was then compared to that of the results from the tests without shielding and the uncertainty was also calculated based on the highest standard deviation calculated across the data collected.

Figure 4: The voltage output of the full system as a function of core displacement for a selection of different voltages supplied to the LVDT.

Input Voltage (V)

Max Standard Deviation of Output Voltage (mV)

Type A Uncertainty of Displacement (µm)













Table 1: Results of the uncertainty calculations for each input voltage tested.

The uncertainties of the voltage output for each input and their corresponding displacement uncertainty for listed in Table 1. From this, the maximum type A uncertainty was found to be 48 µm for the measurement of the core displacement. The type B uncertainty from the experimental design was found to be 26 µm, resulting in an overall uncertainty of 54 µm for this system over a range of approximately 20 mm. This uncertainty indicates the maximum deviation of a single displacement measurement from the true value. A general equation for the system output is shown in Equation 5, where the output voltage, V, is a function of the input voltage, Vin, from the function generator in volts and the core displacement, x, in millimeters.


Figure 5 shows the output voltage as a function of the core displacement for the shielded coils compared to the original unshielded output. The maximum standard deviation of the output voltages across all three trials was 1.2 mV, which leads to a Type A uncertainty of 57 µm and an overall uncertainty of 63 µm.

3. Results

The output voltage as a function of core displacement for the full system is shown in Figure 4, where each line represents a different input voltage. Note that input voltage here refers to the voltage input to the LVDT from the inductor coils, not the voltage supplied to the coils themselves. The data represented is the average values across all the trials. 83

5. Conclusion

Figure 5: The voltage output as a function of core displacement for both shielded and unshielded testing environments.

The experimental model was tested over a range of input voltages to observe its impact on the sensor’s output. The strong linearity and repeatability of the results indicate that the use of wireless inductive coupling is feasible for both power and signal transmission, so long as appropriate precautions are taken in the design to minimize the additional uncertainty. Tests were also conducted to assess the signal decay through material which simulates the cladding of a nuclear fuel rod. The signal’s ability to penetrate the material with no significant effects on the system’s linearity and is a strong indicator that this mechanism can be applied in nuclear reactors. Additional testing is currently being performed on the fully assembled system to assess how the system would operate in more reactor-like conditions and if the technology is still valid.

4. Discussion

6. Acknowledgements

Using Equation 5 to generate new linear regressions for each set of data in Figure 4 results in R2 values ranging from 0.9990 to 0.9996. This suggests that the results have a strong overall linearity and fit very well to the general equation for the system’s output. From the results in Table 1, there does not appear to be any correlation between the input voltage and the uncertainty of the displacement. The highest uncertainty value occurred at an input of 3 V, rather than 4 V, while all other values fluctuated with no apparent trend. Thus, the uncertainty can be attributed to random fluctuations and there is no concern of a larger uncertainty if the voltage input needs to be increased in future applications. In addition, the high linearity of the system indicates that these measurements are highly repeatable with very little noise interference. However, compared to the specified uncertainty of 32 µm for the commercial LVDT used in this test, this experimental setup significantly increases the uncertainty of the system. Despite the low noise levels, the systems output would vary significantly across different measurements if the wiring was disturbed, thereby increasing the uncertainty of the results. Additional precautions are therefore necessary to minimize the additional uncertainty from all aspects of the design, including the inductor coils. The introduction of the shielding led to a 53% decrease in sensitivity for the output voltage. However, there was no impact on linearity, as the linear regression generated for the shielding data had an R2 value of 0.9998. This is promising as the input voltage of the system can be increased to compensate for the loss of sensitivity without being concerned with the linearity of the measurements. However, this shielding does lead to a 16.6% increase in uncertainty from the unshielded results. This increase appears to be unavoidable as the system will encounter this type of shielding in any nuclear application. However, the effects of this increase can again be minimized by reducing the uncertainty of the original unshielded results.

84 Undergraduate Research at the Swanson School of Engineering

Funding for this project was provided by the U.S. Department of Energy. Special thanks to Dr. Heng Ban for his guidance throughout this project.

7. References

[1] H. M. Hashemian, Nuclear Power Plant Instrumentation and Control, Nuclear Power - Control, Reliability and Human Factors, p. 56, P. Tsvetkov, Ed., InTech, Rijeka, Croatia (2011). [2] F. Nekoogar, F. Dowla, Design Considerations for Secure Wireless Sensor Communication Systems in Harsh Electromagnetic Environments of Nuclear Reactor Facilities, Nuclear Technology, 202 (2018), 191-200. [3] L. Xie, Y. Hou, W Lou, Wireless Power Transfer and Applications to Sensor Networks, IEEE Wireless Communications, (2013) 140-145. [4] D. D. L. Chung, Materials for Electromagnetic Interference Shielding, Journal of Materials Engineering and Performance, 9 (2000) 350-354. [5] S. Solstad, R. V. Nieuwenhove, Instrument Capabilities and Developments at the Halden Reactor Project, Nuclear Technology, 173 (2011) 78-85. [6] I. Suh, J. Kim, Electric Vehicle On-Road Dynamic Charging System with Wireless Power Transfer Technology, 2013 International Electric Machines and Drives Conference (2013) 234240.

Ingenium 2020

Genetically engineering ocular probiotics to manipulate ocular immunity and disease Yannis Rigasa, b, Benjamin Treatb, Anthony St. Legerb, c Department of Bioengineering, University of Pittsburgh Swanson School of Engineering, Pittsburgh, PA, USA b Department of Ophthalmology, University of Pittsburgh Medical Center, Pittsburgh, PA, USA c Department of Immunology, University of Pittsburgh Medical Center, Pittsburgh, PA, USA a

Yannis Rigas is a current bioengineering major on the track of cellular engineering. His current research interests include microbiology, immunology, genetic and microbial engineering. Yannis plans to attend graduate school upon graduating from the University of Pittsburgh. Yannis Rigas

Dr. Anthony St. Leger

Dr. St. Leger is an ocular immunologist in the Department of Ophthalmology at the University of Pittsburgh. Since joining Pitt in late 2017, his lab has focused on understanding the relationships between the microbiome and ocular immune system during health and disease. Specifically, his lab currently investigates how the microbiome may be manipulated to alleviate ocular surface disease.

Significance Statement

Every year eye-related diseases account for a huge financial and medical burden. For example, in the U.S., approximately 1 million doctor visits and $175 million are spent annually on treating keratitis and contact lens disorders [1]. In an effort to characterize and manipulate the interactions between ocular immunity and microbiome, our lab is developing a model to characterize and manipulate the relationship between the microbiome and ocular immunity. To study this, we are developing a model that allows the study of commensal bacterium, Corynebacterium mastitidis, and its interactions with ocular immunity. The long-term aim of these studies is the development of a vehicle for long-term therapy delivery to the ocular surface in order to help treat ocular diseases.

Category: Methods

Keywords: Ocular immunity, genetic/microbial engineering, probiotic, and microbiology Abbreviations: C. mast (Corynebacterium mastitidis), IL-17 (Interleukin 17), γδTC (gamma delta T-cells), WT (Wild Type)


Recently, our lab discovered that the eye harbors a microbiome that includes Corynebacterium mastitidis, which can stimulate local immunity to protect the eye from more serious infections. From our previous work, we know that C. mast can remain on the ocular surface indefinitely. Therefore, the goal of this project is to genetically engineer C. mast, so that it can act as a natural vehicle to deliver therapeutics locally to alleviate or prevent ocular surface diseases. Here, we took initial steps towards this goal by genetically modifying the C. mast genome so that a fluorescent protein is selected for by utilizing an antibiotic resistance cassette. This fluorescence will allow for real-time in vivo detection of genetically modified C. mast. In this study we have discovered four possible mutants that are resistant to the antibiotic, kanamycin, and fluoresce with varying degrees intensity in the red channel. We hypothesized that these mutants would retain the ability to colonize the eye and induce immunity similar to wild type C. mast. Indeed, we observed that all four mutants were able to colonize the eyes of mice and elicit immune responses similar to wild type C. mast. We further demonstrated that genetically engineered strains of C. mast can effectively colonize the ocular mucosa and elicit an immune response similar to WT C. mast.

1. Introduction

Previously, research on the ocular surface has shown that Corynebacterium mastitidis is stably present on the conjunctiva, an ocular mucosal immune tissue, while displaying commensal properties by eliciting an immune response from γδT cells. The presence of this commensal bacterium, and its accompanying immune response were previously shown to protect the eye from pathogenic eye infections caused by Pseudomonas aeruginosa and Candida albicans. [2]. Through a series of experiments, we were able to link ocular colonization of C. mast with an induced antimicrobial immune response [2]. Due to the ability of this microbe to asymptomatically thrive at the ocular surface for indefinite periods of time, C. mast is an attractive candidate to engineer as a long-term drug delivery vehicle for inflammatory diseases like keratitis, Dry Eye Disease, and Sjogren’s Syndrome. A similar technique was used to deliver the immune regulating cytokine, interleukin (IL)-10, in mice with inflammatory bowel syndrome (IBS). More specifically, when Lactococcus lactis was engineered to express IL-10, disease associated with colitis was reduced compared to controls in two separate mouse models [3]. In this current study, we successfully genetically engineered C. mast by electroporation with a novel plasmid, and showed that C. mast can stably remain on the ocular surface and elicit an immune response similar to the WT strain. This discovery will pave the way for future modifications that will one day allow for a longterm delivery of therapeutics to the ocular surface.


2. Methods

We modified C. mast using a custom synthetically designed plasmid constructed from gBlocks (Integrated DNA Technologies, Coralville, Iowa) containing genes for the codon-optimized expression of mCherry, kanamycin resistance, and a mariner transposase.

We then tested the candidates for the ability to colonize C57BL/6 mice and induce an immune response that is similar to the WT strain of C. mast. Mice were inoculated three times over 6 days, and three weeks after the final inoculation, mice were sacrificed and flow cytometry was performed in order to determine if mice exhibited stimulation of IL-17 and neutrophils.

Figure 2: Timeline of mouse inoculation.

Figure 1: Visual representation of plasmid with transposase, mCherry, and Kanamycin resistance

Further iterations of this plasmid, containing a native origin of replication, were derived from the native cryptic plasmid already present in C. mast. We were able to integrate our plasmid into C. mast through both electroporation and conjugation. For electroporation, the induced pore formation in the bacterial cell wall allowed for the entry of the plasmid into the cytosol of the cell. From there, the transposase coordinates random transposon integration into the bacterial genome. For conjugation, a cell to cell interaction allowed for the transfer of our plasmid from the “donor” S17 E. coli to our target, C. mast. After successful uptake of plasmid DNA, the mariner transposase coordinated insertion of both the mCherry and kanamycin resistance gene. We screened for successful integration of our transposon by incorporating stringent selection with kanamycin (25 µg/mL) that would select for our transposon insertions. Candidates were further screened under a fluorescent microscope to detect red fluorescence which would indicate mCherry expression. Since mCherry expression will depend on genomic location of the transposon, we reasoned that the level of fluorescence would positively correspond with the activity of the promoter that would guide the expression of these genes. This system will allow us to identify novel promoters and genomic locations for future genetic modifications. Colony polymerase chain reaction for the mCherry gene eliminated any candidates that may have been autofluorescent due to stress or unwanted adaptive mutations and ensured only transposon mutants were chosen. In addition, flow cytometry was used in order to observe the amount of fluorescence in order to quantitatively analyze how bright our mutants are compared to WT C. mast.

86 Undergraduate Research at the Swanson School of Engineering

Flow cytometry was performed by isolating single cell suspensions of mouse conjunctiva and lymph node tissues, staining fixed and permeabilized cells with an antibody panel (that includes α-CD45, α-γδTCR, α-Ly6C, α-Ly6G, and α-IL-17 [Biolegend]) each labeled with unique fluorophores. Samples were run on a Cytoflex LX (check) and analyzed for cell populations. This response was compared to a control group of mice inoculated with WT C. mast to observe if the mutants had a similar impact. A one-way ANOVA was used to test whether any of our results were significant.

3. Results

Despite all the candidates maintaining kanamycin resistance, there were variable levels of fluorescence. Although all mutants had variable levels of fluorescence, each candidate expressed more fluorescence than the wild type C. mast. In fact, one candidate expressed 5-fold more fluorescence compared to other candidates as detected by flow cytometry for mCherry expression (Figure 3B-C).

Ingenium 2020

Upon inoculation of mice, all four mutants were able to successfully colonize the eye, suggesting that the transposon insertion did not affect any genes absolutely necessary for colonization. The decreased level of colonization correlates with the lower number of neutrophils and IL-17 producing cells as seen in Figure 4. In addition, mutants obtained from mice two weeks postcolonization were not able to colonize as efficiently.

4. Discussion

The varying amounts of fluorescence was to be expected as transposon insertions are notoriously variable, and fluorescence is more present when in an actively transcribed section of the genome. Although fluorescence was variable between the four mutants, it is promising that all candidates harbored more fluorescence than the WT C. mast. In addition, the fact that all four samples were able to colonize and induce an immune response means that we have not interrupted genes required for colonization and/or ocular immunogenicity. However, the reduced colony size and dampened immune response may correlate with changes to genes of intermediate importance. Genomic sequencing of transposon insertions will allow us to determine this.

5. Conclusion

In this study, we have genetically modified C. mast to express mCherry and Kanamycin resistance and showed that even with these modifications the ability to colonize the ocular mucosa and elicit an immune response in mice is still present. After obtaining these four mutants, we are attempting to optimize the integration of genetic material by altering the native plasmid present in C. mast in order to more efficiently introduce genes for mCherry and kanamycin resistance. This will allow us to screen a greater number of transposon mutants, and will further aid in rational design of future constructs. By genetically engineering C. mast we have also paved the way for further genetic manipulations that can lead to the development of future therapeutics for ocular diseases.

6. Acknowledgments Figure 3: Screening C. mast transposon insert candidates for mCherry expression. A. Fluorescence microscopy of individual WT C. mast vs transposon candidate bacteria. Flow cytometry measuring WT or transposon candidate mCherry expression levels (B) and mean fluorescence intensity (C) in WT and candidate strains (6,7,14,21) of C. mast. Experiment was repeated with similar results.

Figure 2: Flow cytometry measure amount of IL-17-producing γδTCR and neutrophils present due to inoculation of 4 candidates as well as WT C. mast. Dots represent individual values with value bars representing mean and standard deviation. Signifigance was determined by One-way ANOVA with Dunnett’s multiple comparisons test. * represents p<0.05, ** represents p<0.005 as compared to WT C. mast

Anthony St. Leger, PhD, Dana Previte, PhD, Kate Carroll, PhD, Hongmin Yun, PhD, Heather Buresch, the Swanson School of Engineering, and the Office of the Provost.

7. References

[1] Sarah A. Collier, MPH, Michael P. Gronostaj, MD, PharmD, Amanda K. MacGurn, MPH, Jennifer R. Cope, MD1 Kate L. Awsumb, MA, MPH, Jonathan S. Yoder, MPH, MSW, Michael J. Beach, PhD, Estimated Burden of Keratitis – United States, 2010, Morbidity and Mortality Weekly Report. 63(45) 1027-1031. [2] Anthony J. St. Leger, Jigar V. Desai, Rebecca A. Drummond, Abirami Kugadas, Fatimah, Almaghrabi, Phyllis Silver, Kumarkrishna Raychaudhuri, Mihaela Gadjeva, Yoichiro Iwakura, Michail S. Lionakis, Rachel R. Caspi, An Ocular Commensal Protects against Corneal Infection by Driving an Interleukin-17 Response from Mucosal γδ T Cells, Immunity, 47(1), 148-158. e5. [3] Lothar Steidler, Wolfgang Hans, Lieven Schotte, Sabine Neirynck, Florian Obermeier, Werner Falk, Walter Fiers, Erik Remaut, Treatment of Murine Colitis by Lactococcus lactis Secreting Interleukin-10, Science, 289(5483) 1352-1355.


Effects of printing parameters on density and mechanical properties of binder jet printed WC-Co Pierangeli Rodríguez De Vecchisa, Danielle Brunettaa, Katerina Kimesa, Drew Elhassida and Markus Chmielusa Department of Mechanical Engineering and Materials Science, University of Pittsburgh a

Pierangeli Rodríguez De Vecchis

Dr. Markus Chmielus

Pierangeli Rodríguez De Vecchis is a Materials Science and Engineering senior. She works in Dr. Chmielus lab on Binder Jet 3D Printing (BJ3DP). She has been awarded the sophomore and junior PCEASASM and WAAIME scholarships. She is the vice-president of materials advantage 2019-2020. Dr. Markus Chmielus is an Associate professor at the University of Pittsburgh (MEMS). His research is focused on binder jet printing of metal alloys and the study of magneto-caloric materials. He teaches Junior and Senior Materials Science classes and is the Advisor for the Material Advantage Chapter.

Significance Statement

Traditional manufacturing of WC-Co parts is slow and expensive, while Additive Manufacturing offers a commercially viable and fast solution. This project shows how choosing optimal printing parameters for binder jet printing (BJP) can produce parts with equally excellent mechanical properties, allowing to pair WC-Co (complex material) with BJP (innovative technology).

Category: Methods

Keywords: Binder jet printing, tungsten carbide, printing parameter, design of experiments.

88 Undergraduate Research at the Swanson School of Engineering


Tungsten carbide-cobalt (WC-Co) is a cermet material widely known for its excellent combination of mechanical properties including high hardness provided by small WC grains, and high toughness provided by the Co-matrix binder metal. Its applications range from mining and drilling tools to cutting gears. Traditionally, WC-Co parts are formed through powder metallurgy processes. The WC and Co mixed powders are typically pressed with added wax that results in a low-density part, which is later put through a de-waxing and hot isostatic pressing (HIP) process. However, this process demands mass production resulting in a slow and expensive process. Additive Manufacturing, particularly binder jet-printing (BJP) appeared as an option to supplement traditional WC-Co manufacturing, allowing the production of fast, specific, and highly detailed parts. A design of experiments was set-up to find the optimal printing parameters to form parts with high green densities, translating to high hardness and fracture toughness. The highest green densities were obtained with a 220% binder saturation, 45 s drying time, 100 µm layer thickness, 5 mm/sec roller speed, and a build-to-feed ratio of 2. Sintered-HIPed parts resulted in 99% relative density, 1310 HV hardness and 14.74 MPam0.5 fracture toughness.

1. Introduction

Tungsten carbide-cobalt (WC-Co), also known as cemented carbide is a cermet (ceramic-metal) material widely known for its excellent mechanical properties (including high density, hardness, toughness and flexural stress). It is used in wear resistant applications, including machining, cutting, and rolling, as well as mining and oil drilling tools [1]. Morphologically, the WC-Co microstructure is composed of hard/brittle, small, polygonal WC grains within a tough Co matrix. Co is known as the binder metal as it is chosen to melt at a lower temperature than WC to wet the grains and allow strong metallic bonds to form between WC particles during sintering, reducing brittleness without greatly decreasing hardness [2]. Traditionally, WC-Co parts are manufactured by powder metallurgy through which WC and Co particles are blended and ball-milled, after having previously carbonized W, typically with carbon monoxide (CO). Parts are formed through mechanical pressing or molding to obtain a green state. A small amount of paraffin is added to increase its density which is removed through a de-waxing process, followed by sintering and hot isostatic pressing (HIP), resulting in the final, fully dense cermet part [3]. This technology is slow and expensive, requiring the mass production of molds and limited in resolution. As a result, additive manufacturing (AM) appeared as an option to create WC-Co shapes fast and with specific design requirements. The focus of this project is the AM technology of binder-jet printing (BJP). BJP has the potential to produce WC-Co objects by selectively stacking layers of powder and binder alternatively, according to a computer design, as shown in Figure 1. It is a fast, cost-effective process that allows the formation of complex internal and external geometries [4]. In contrast with other AM

Ingenium 2020

methods, like Selective Laser Melting (SLM), this process does not introduce internal thermal stresses and does not need support pieces for complex hanging shapes. Nevertheless, BJP requires post-processing of curing and sintering to obtain a fully dense part without binder content. Only a few publications have BJPed WC-Co, like [5], who mainly addressed the wear resistance of the parts, but did not show a systematic approach to printing or tested other mechanical properties.

Figure 1: Binder jet printing process. (a) Roller spreads powder in x-direction, (b) Printhead deposits binder on design silhouette along y-direction, the binder is partially cured, (c) Feeder moves one layer down and builder one layer up to provide powder in z-direction. Repeat.

The objective of this project was to find the optimal set of printing parameters and characterize the microstructure of WC-Co parts resulting in similar properties to those obtained conventionally. This was achieved through two series of design of experiments (DoE). Firstly, the powder packing density was studied by comparing the effects of rolling speed, feed-to-build ratio and layer thickness, and secondly, the green density of the parts was studied when changing the binder saturation and drying time. Best parameters were chosen by judging the green density and quality of the part (integrity, edge and shape retention), and confirmed through post-sintering properties. Furthermore, the printed parts were prepared through metallography and their density (through Archimedes method), hardness and fracture toughness were compared. From previous experience, it was expected that low roller speed and intermediate feed-to-build ratio, paired with layer thickness of at least twice the size of the agglomerates would yield higher powder packing density, and that high levels of saturation and drying time will be needed to bind the porous powder.

The first DoE study involved comparing roller speed (5 or 15 mm/s), layer thickness (50 or 100 µm), and build-to-feed ratio (1.5 or 2) to obtain the highest powder-bed packing density without binder. Packing densities of the resulting eight “prints” were calculated as mass (OHAUS AX324) per volume. The combination yielding highest packing density was chosen: 100 µm layer thickness, 5 mm/s roller speed and a 2:1 build-to-feed ratio. The second DoE stage used the previous results as constants and consisted in varying binder saturation (100%, 160% or 220%) and drying time (30 s, 45 s or 60 s). These iterations resulted in nine prints of 12 cubes each that were cured at 200°C for 8 h. Green density was measured as before, as mass over volume. Green parts were sintered by General Carbide at 1440°C. One sample per print was sectioned, mounted, ground, polished (Struers LaboPol 5), etched with Murakami reagent, and imaged (Zeiss Smartzoom5 optical microscope). Relative density was calculated with ImageJ® software [7]. To measure hardness and fracture toughness, three indents were imprinted (Rockwell 724 Wilson) on each sintered sample, with a Vickers diamond tip and major load (P) of 60 kgf. Vickers hardness (H) was calculated by measuring the indent’s diagonals (Keyence VH-Z500 digital microscope). Toughness was calculated by plugging the measured edge cracks (L) in Shetty’s formula (Equation 1) for brittle materials exhibiting Palmqvist cracking [8]. (1)

3. Results 3.1 Powder characterization SEM images of the powder were taken to characterize its morphology. As shown in Figure 2, powders are collections of fine polygonal particles forming nearly spherical agglomerates, typical of WC-Co powders. Stereological analysis showed individual particles with an average diameter of 0.52 µm, forming agglomerates of about 21.3 µm.

2. Methods

Spray-dried, half-sintered WC-Co powder (12.5% Co) from General Carbide was characterized through scanning electron microscopy (SEM) and stereological analysis (line intercept method [6]). Cubes of 1x1x1 cm3 were BJP with an ExOne Lab varying settings in two DoE studies.

Figure 2: SEM of WC-Co powder. The submicron-sized particles form nearly spherical agglomerates.


Figure 3: Design of experiments main variables’ effect on green density. (a) Powder packing DoE, (b) Printing parameters DoE.

3.2 Design of Experiments The first DoE (Figure 3a) showed the effect of layer thickness, rolling speed and feed ratio on packing density. Green density increased with layer thickness and feed ratio but decreased with roller speed. Judging by the slopes of the lines, the most influential parameter was layer thickness. The parameters 100 µm layer thickness, 5 mm/s roller speed and feed ratio of 2 were chosen. For the second DoE (Figure 3b), it is observed that green density of the parts increased with binder saturation and with respect to drying time, it was a maximum at the midpoint. Again, judging by the slope of the lines, it seems that binder saturation has a greater effect than drying time on the density of the parts. From this experiment, nine resulting sets of prints, were obtained and evaluated to understand their properties beyond green density, although it is expected that the higher the green density the better the final density and properties.

3.3 Microstructure and Mechanical Properties Prints #4, #5 and #8 did not survive handling and transportation due to their fragility (very low binder saturation). From the remaining samples, all microstructures, after sintering at 1400° C and HIPing, were similar to that shown in Figure 4 (lighter WC grains in dark Co matrix) with an average WC grain size of ~1.72 µm. Prints #2 and #7 showed some porosity, especially near the corners, and samples #1, #3 and #6 showed some free C.

Figure 4: Sintered WC-Co microstructure showing WC grains in Co matrix (etched).

90 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

Both relative green and sintered densities are shown in Table 1. Sintered densities were approximately the same. Hardness and fracture toughness (Table 1) were calculated in order to compare their mechanical properties. Although some are very close, the samples with the best mechanical properties are those of print #1 (1310 HV and 14.74 MPam0.5).

5. Conclusion

Table 1: Settings, densities and mechanical properties of printed parts (color coded from best/green to worst/red).

It was found that BJPed WC-Co parts can be produced with nearly full density and mechanical properties similar to traditional parts. The two-step DoE allowed to formulate an optimal set of printing parameters (Print #1): 220% binder saturation, 45 s drying time, 100 um layer thickness, 5 mm/sec roller speed, and a build-to-feed ratio of 2. Through these settings a sintered density of 98.5%, hardness of 1310 HV and fracture toughness of 14.74 MPam0.5 were obtained. Microstructurally, a medium-grained structure was generated with some free C (not too harmful). These results prove BJP can be used to a wider variety of metallic powders and even expand into cermets that were differently treated. For both WC-Co manufacturing and BJP expansion, the combination of a traditional powder with a new technology shows the possibilities of combining them to reduce cost and time, while emphasizing detail and shape particularity. In the future, differently shaped WC-Co powders, as well as green part preservation will be studied.

4. Discussion

6. Acknowledgements

Drying time [s]


Binder saturation [%] 220





Print #

Sintered Density [%] 98.5

Hardness [HV]


Green Density [%] 21.9












Fracture Toughness [MPam0.5] 14.74











































Spray-dried WC-Co powder was originally chosen due to its uniform size distribution and flowability required in BJP. However, WC-Co is an intrinsically complex powder as agglomerates are porous, requiring binder saturation levels above 100%, which in turn is the driving force for the DoE. The first DoE step showed that the higher layer thickness was required as it cannot be smaller than the agglomerate size, and it is optimized when it is about three times bigger as layers can be composed of more than one agglomerate. The slower roller speed and higher feed ratio prevent the powder being pushed instead of spreading uniformly. The second DoE step showed that binder saturation is the variable with greater effect on the print’s density. The highest value (220%) was selected. This does not imply that the more binder the better density. ExOne solvent binder is C-based, and it is well known that excess C worsens WC-Co properties, by promoting WC grain growth and free carbon (graphite) precipitation. Drying time is also a relevant variable. Short times result in part bulging and long times in delamination as each layer is over-dried and does not bind to the next. Intermediate drying (45 s) was found to be optimal. High densities were achieved for almost all samples. The pores present in some samples, were small and aligned, suggesting delamination (layers of the green part may separate during handling and transportation due to low green density). Microstructurally, the samples are characterized as having a medium sized (1.4-3.4 µm) WC grain structure [9]. The presence of free C, might be due to retained binder or most likely because there is stoichiometrically more C than W in the prepared powder [9]. In terms of mechanical properties, both Vickers hardness and Palmqvist fracture toughness were comparable to those reported by General Carbide through traditional manufacturing (1245 1633 HV, 13-14 MPam0.5 [3]). It is important to note that calculated fracture toughness is a macroscopic approximation requiring deeper study. The load at which cracks start to appear (Pc) was approximated to 80 kgf, since there were at 100 kgf, but not at 60 kgf. Nano-indentation would give better values ([10]).

Thanks to the Swanson School of Engineering and the Office of the Provost, as well as PMFI, General Carbide and all Chmielus Lab group, especially Katerina Kimes.

7. References

[1] N. Alves Nery Balbino, E. Otoni Correa, L. Carvalho Valeriano, Development of the 90WC-8Ni-2Cr3C2 cemented carbide for engineering applications, Int. J. Adv. Manuf. Tech. (2018). [2] L. Fu, Two-step synthesis of nanostructured WC-Co powders, Scr. Mater. 44 (2001). [3] General Carbide, The Designer’s Guide to Tungsten Carbide, (2008). [4] A. Mostafaei, E.L. Stevens, J.J. Ference, D.E. Schmidt, M. Chmielus, Binder jetting of a complex-shaped metal partial denture framework, Addit. Manuf. 21 (2018). [5] R.K. Enneti, K.C. Prough, Wear properties of sintered WC-12%Co processed via Binder Jet 3D Printing (BJ3DP), Int. J. Refract. Met. Hard Mater. 78 (2019) 228–232. [6] C. Russ, John, T. Dehoff, Robert, Practical Stereology, 2nd ed., Plenum Press, 2000. doi:10.1088/1751-8113/44/8/085201. [7] C.T. Rueden, J. Schindelin, M.C. Hiner, B.E. DeZonia, A.E. Walter, E.T. Arena, K.W. Eliceiri, ImageJ2: ImageJ for the next generation of scientific image data, (2017) 1–26. [8] R. Spiegler, S. Schmauder, L.S. Sigl, Fracture Toughness Evaluation of WC-Co Alloys by Indentation Testing, J. Hard Mater. 1 (1990). [9] J. García, V.C. Ciprés, A. Blomqvist, B. Kaplan, International Journal of Refractory Metals Cemented carbide microstructures : a review, Int. J. Refract. Met. Hard Mater. 80 (2019) [10] T.A. Fabijanic, et al, Vickers Indentation Fracture Toughness of Near-Nano and Nanostructured WC-Co Cemented Carbides, Metals (Basel). 7 (2017). 91

Monitoring the in-vitro extracellular matrix remodeling of tissue engineered vascular grafts Hannah Schmidta, Jonathan Vande Geesta, b, c Department of Bioengineering, bMcGowan Institute for Regenerative Medicine, cVascular Medicine Institute a

Hannah Schmidt is a fourth-year undergraduate studying Bioengineering at the University of Pittsburgh. She conducted research on tissue engineered vascular grafts with Dr. Vande Geest’s lab for two years and, in the future, plans to attend medical school. Hannah Schmidt

Dr. Jonathan Vande Geest is a Professor in the Department of Bioengineering, Department of Mechanical Engineering and Material Science, the Department of Ophthalmology, the McGowan Institute for Regenerative Medicine, the Louis J. Fox Center for Vision Restoration, and Dr. Jonathan Vande the Vascular Medicine Institute at the Geest University of Pittsburgh. He received his BS in Biomedical Engineering from the University of Iowa in 2000 and his PhD in Bioengineering from the University of Pittsburgh in 2005. Dr. Vande Geest began his career at the University of Arizona in the Department of Aerospace and Mechanical Engineering and joined the U of A’s Department of Biomedical Engineering in 2009. Dr. Vande Geest returned to the University of Pittsburgh in January of 2016.

Significance Statement

It can be difficult to visualize longitudinal structural changes in-vivo in biological systems, oftentimes due to the necessity of destructive imaging. This research successfully overcomes this challenge for the study of vascular graft collagen remodeling by developing two-photon imaging parameters which can be generalized to broad in-vivo tissue engineering applications.

Category: Methods

Keywords: tissue engineered vascular graft, collagen remodeling, two-photon microscopy

92 Undergraduate Research at the Swanson School of Engineering


Heart disease is the leading cause of death in the United States for both men and women. Coronary artery bypass grafts (CABG) are frequently implanted to restore blood flow to the heart, but these small diameter vascular grafts frequently accumulate clots and become narrow, making them ineffective. The Soft Tissue Biomechanics Laboratory (STBL) is seeking to create a tissue engineered vascular graft (TEVG) to address these issues of compliance mismatch and thrombosis in small diameter grafts. It is particularly important to assess the extracellular matrix remodeling (ECM) capabilities of our TEVGs in order to monitor the in-vivo transition of TEVGs from synthetic graft to host remodeled tissue. This study therefore aims to develop an in-vitro imaging method for quantifying ECM remodeling of TEVGs. We were able to determine optimal imaging parameters and show that two-photon imaging can be used to characterize structural changes of collagen in the ECM, which will be used in the future to evaluate TEVG efficacy.

1. Introduction

Heart disease was responsible for 633,842 deaths in 2015 [1], making it the leading cause of death in the United States for both men and women. Coronary artery disease (CAD) in particular is responsible for nearly half of all heart disease cases [2]. Coronary artery bypass grafting (CABG) from autologous vessels is a common treatment used to restore blood flow to the heart in patients with CAD, but suffers from high rates of thrombosis and restenosis, with reintervention rates reported to be as high as 8.8% [3]. Providing a functional tissue engineered vascular graft (TEVG) for CABG surgeries would therefore result in substantial improvements in patient care. The Soft Tissue Biomechanics Laboratory is creating a TEVG for small diameter CABG applications to address these issues. Our team aims to create a primarily acellular, biocompatible, and compliance matched graft. This study will focus on evaluation of the overall function of the TEVGs. Because the grafts are primarily acellular when implanted, we must ensure that native cells can migrate and proliferate within the graft to transition TEVGs into living tissue. This transition is accomplished primarily through the production of a collagen extracellular matrix (ECM) by vascular smooth muscle cells as the TEVG degrades. It is therefore necessary to quantify the in-vitro ECM remodeling of TEVGs to understand the graft’s in-vivo transition from polymer-based scaffold to host remodeled tissue. Current approaches to imaging ECM remodeling include the work of Hjortnaes et al. 2009 [4], which showed that TEVG remodeling in-vivo can be monitored using laser scanning fluorescence imaging. The study relied on injection of nondestructive imaging agents which provided signal in response to proteolytic activity. Though they were successfully able to see changes in the enzymatic activity of TEVGs in a mouse model, graft degradation and new ECM formation cannot be directly measured using this method because it does not actually visualize the collagen or the TEVG itself. Rather, it uses enzymatic activity to make inferences about ECM remodeling. A different method is required to visualize degradation and collagen formation first-hand.

Ingenium 2020

Several other studies such as Raub et al. 2007 [5] and Quinn et al. 2016 [6] have developed methods for the assessment of collagen structure using multiphoton microscopy, but they have not sought to differentiate native collagen from synthetic graft materials or cells and compare their changes over time. The present study seeks to develop a method that provides this additional information which is lacking from the current approach, which is important to understanding how native tissue interacts with the graft. Understanding the coupled actions of graft degradation and new collagen formation is critical to evaluating the effectiveness of a TEVG. This study therefore will aim to develop a method for viewing and quantifying structural ECM changes through two-photon (2P) imaging. First, optimal imaging parameters will be determined, then the imaging method will be applied to study collagen gels. Collagen gels were chosen as a simpler model as they already possess a well-defined collagen matrix. Thus they provide a good model to critically evaluate the efficacy of the proposed ECM remodeling quantification method, toward eventual in-vitro and in-vivo applications.

2. Methods 2.1 TEVG Fabrication and Seeding TEVGs were fabricated by electrospinning a 10% (w/v) solution of gelatin (Sigma) and polycaprolactone (PCL, Sigma) in 1,1,1,3,3,3-Hexafluoro-2-propanol (HFP, Oakwood Chemical). A ratio of 80:20 (gelatin:PCL) was chosen based on the work of Ardila et al. 2015 [7], which showed that this ratio is favorable for smooth muscle cell growth and migration. The 80G:20P solution was electrospun at 15kV onto a rotating and translating target rod to create cylindrical constructs with an inner diameter of 1.4 mm and a radial thickness of 200 μm. Constructs were then crosslinked in 0.5% genipin (Fisher Scientific) in EtOH for 24 hours at 37°C. After crosslinking, constructs were washed three times in EtOH, then three times in 1X PBS. Porcine aortic smooth muscle cells (SMC), isolated as described by Ardila et al. 2015 [7], were then seeded onto 3 mm axial length constructs by dropwise seeding at a density of 4.5×106 cells/mL and incubated for 1 hour to allow cellular attachment. This seeding process was repeated twice. Seeded constructs were cultured in media containing 1 ng/ mL TGFβ2 (R&D Systems), which Ardila et al. 2015 [7] showed encourages SMC proliferation in electrospun constructs. Figure 1 shows a visual representation of this entire process.

Figure 1: Methods flow diagram for fabrication (A) and seeding (B) of TEVGs.


2.2 Creation of Excitation-Emission Spectra The 80G:20P constructs were excited at a range of wavelengths (750-1000 nm) with a tunable laser at 50 nm increments under a 2P microscope (LaVision BioTec). Excitation laser power at the sample was kept at a constant 40 mW throughout. At each excitation wavelength, the resulting emission signal was filtered from 375 nm to 680 nm. The intensity of each filtered signal was measured. In this way, the emission spectrum of a 80G:20P construct was determined over a range of excitation wavelengths. The same process was repeated for a 40 μm thick cross section of rat aorta, exciting from 750-1250 nm with 50 nm increments. The strength of the emission signal from the adventitial layer of the aorta, which is composed of collagen expressing SHG, was evaluated to determine the emission spectrum of fibrillar collagen. Seeded TEVGs were imaged using optimized 2P parameters based on the excitation-emission spectra of the construct and collagen.

Figure 2: Excitation-emission spectrum for 80G:20P electrospun constructs. Two distinct emission peaks are visible at 460 nm and 620 nm regardless of excitation wavelength.

2.3 Collagen Gel Fabrication After determining the efficacy of the imaging parameters on TEVGs, collagen gels were used to fully develop the method of ECM remodeling quantification. Collagen gels were used because they already have a well-defined fibrillar collagen structure, making them easier to study for the purpose of refining a method. To create the gel, type I collagen from rat tail (Fisher Scientific) was mixed with culture media to a final concentration of 3mg/mL. All materials were kept on ice to prevent premature polymerization. NIH 3T3 fibroblasts expressing red fluorescent protein (RFP) were then suspended within the collagen mixture, and the gels were allowed to polymerize at 37°C. 2.4 Image Analysis and Statistical Testing Collagen gels (n=9) were imaged using the same optimized 2P parameters, created as described in Section 2.2. Eight images per sample, each 10 μm depth apart, were then converted to binary using ImageJ [8]. The porosity of each binarized image was calculated and the average porosity of each collagen gel sample was determined. This process was repeated at time points of 1, 4, and 6 days to observe variations over time. The change in average porosity over time was evaluated by comparing time points using a student’s t-test (α = 0.05).

Figure 3: Excitation-emission spectrum for adventitial collagen in rat aorta. Emission peaks vary based on excitation wavelength due to second harmonic generation, corresponding to a maximum emission at roughly half the excitation wavelength.

3. Results 3.1 Imaging Parameter Optimization Optimal imaging parameters were chosen based on the two-photon excitation-emission spectra for each graft component. Figure 2 shows the excitation-emission spectra for an 80G:20P construct and Figure 3 shows the spectra for collagen from rat aorta. Parameters to simultaneously image constructs and collagen were established as (excitation/collection) 900/620nm and 900/425nm respectively. These optimized parameters were applied to image TEVGs as shown in Figure 4, which is a representative image of a seeded TEVG after one week of culture in media containing 1 ng/mL TGFβ2. The same parameters were applied to the rest of this study.

94 Undergraduate Research at the Swanson School of Engineering

Figure 4: Cross sectional (A) and en face (B) views of an SMC seeded TEVG after one week of culture in media containing 1 ng/mL TGFβ2. The 80G:20P construct is shown in green and DAPI stained cell nuclei are shown in blue.

Ingenium 2020

3.2 Porosity Images of seeded collagen gels over time, which were used as a model for ECM remodeling, are shown in Figure 5. From image analysis, gels showed slight variations in average porosity of the collagen matrix over time (Figure 6), but no statistically significant differences were found between time points, with p=0.48 from day 1 to 4 and p=0.21 from day 4 to 6.

Figure 5: Seeded collagen gels over time at time points of 1(A), 4(B), and 6(C) days. Fibrillar collagen is shown in blue and RFP fibroblasts are shown in red.

to the second harmonic generation (SHG) of collagen. Using these spectra, imaging parameters can be easily tuned to fit future needs by selecting excitation wavelengths and collection filters which preferentially collect TEVG or collagen signal. Thus, parameters for spectrally separate imaging of TEVG remodeling components were established. Porosity of collagen gel models, which was used as a metric for quantifying ECM changes, did not significantly change over time (Figure 6). This is primarily due to the high standard deviation of porosity between gels at the same time point and even between different areas of the same gel. Other measures of remodeling, such as cell count, had similarly high standard deviation and no statistically significant differences over time. This high variability makes it difficult to draw conclusions about ECM changes. Therefore this method could be improved by non-destructive imaging over several weeks with an emphasis on imaging the same location in both the x-y plane and in z. The next steps in this research will be to create a repeatable method of locating exact positions within 3D systems over long periods of time to more effectively monitor structural changes. This study was, as mentioned, limited by the high variability of collagen gels. It is also notable that collagen gels do not provide a perfect model for translating this work to the in-vivo study of TEVGs over time. They are, however, an effective and useful tool for continuing to develop this intravital imaging method. Overall, this imaging method improved upon current methods in the literature by seeking to visualize the interplay between synthetic materials and collagen formation in TEVGs. It can be used to improve the current methods of evaluating TEVG function as it relates to host integration.

5. Conclusion

Figure 6: Average porosity of NIH 3T3 seeded collagen gels over time. No statistically significant difference in porosity was observed from day 1 to 4 (p=0.48) or from day 4 to 6 (p=0.21). Error bars show standard deviation.

4. Discussion

The emission spectra of the 80G:20P construct and ECM collagen have very distinct signals which can be used to distinguish them during intravital imaging. This is evident from their excitationemission spectra. The 80G:20P construct displays two distinct emission peaks at 460 nm and 620 nm regardless of excitation wavelength (Figure 2) which are likely due to the autofluorescence of genipin within the graft. The collagen emission signal, on the other hand, varies depending on excitation (Figure 3). Each peak appears at roughly half the excitation wavelength, corresponding

The imaging parameters established here (900/620nm and 900/425nm in excitation/collection for constructs and collagen respectively) can be used to effectively image a TEVG as it is remodeled by creating spectrally separate signals for each component. This method of image binarization and porosity calculation is also a valid strategy for quantifying structural changes such as porosity, which did not significantly change over time in this study with p=0.48 and p=0.21. It will require further work to locate exact 3D positions repeatedly for more accurate observations of changes over time. This work will be used to evaluate the overall function of TEVGs, ensuring that the acellular grafts are populated and remodeled in-vivo, and has the potential to be applied to broader tissue engineering applications for in-vivo imaging.

6. Acknowledgements

This work was funded by NIH 1R56HL136517-01 to JPVG and made possible, in part, by the 2018 American Heart Association SURP award to Hannah Schmidt.


7. References

[1] C.J. Rothwell, T.E. Price, A. Schuchat, Health, United States Report 2016. Hyattsville, MD: National Center for Health Statistics, Centers for Disease Control and Prevention, 2017. [2] D. Mozaffarian et al., Heart disease and stroke statistics. Circulation. 2015;131:e29-e322 [3] P.W. Serruys, A.T. Ong, L.A. van Herwerden, et al., Fiveyear outcomes after coronary stenting versus bypass surgery for the treatment of multivessel disease: the final analysis of the Arterial Revascularization Therapies Study (ARTS) randomized trial. J Am Coll Cardiol, 2005. 46(4): p. 575-81. [4] J. Hjortnaes, D. Gottlied, J.L. Figueirdo, et al., Intravital Molecular Imaging of Small-Diameter Tissue-Engineered Vascular Grafts in Mice: A Feasibility Study. Tissue Engineering Part C: Methods. 16.4 (Aug. 2010) p597 [5] C.B. Raub, V. Suresh, T. Krasieva, et al., Noninvasive assessment of collagen gel microstructure and mechanics using multiphoton microscopy. Biophys J. 2007;92(6):2212–2222. doi:10.1529/biophysj.106.097998 [6] K.P. Quinn, K.E. Sullivan, Z. Liu, et al., Optical metrics of the extracellular matrix predict compositional and mechanical changes after myocardial infarction. Sci Rep. 2016;6:35823. Published 2016 Nov 7. doi:10.1038/srep35823 [7] D.C. Ardila, E. Tamimi, F.L. Danford, et al., TGFβ2 Differentially Modulates Smooth Muscle Cell Proliferation and Migration in Electrospun Gelatin-Fibrinogen constructs. Biomaterials. 2015 January; 37:164–173. doi:10.1016/j. biomaterials.2014.10.021. [8] C.T. Rueden, J. Schindelin, M.C. Hiner, et al. (2017), “ImageJ2: ImageJ for the next generation of scientific image data”, BMC Bioinformatics 18:529, doi:10.1186/s12859-017-1934-z

96 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

Analytical model validation for melting probe performance using applied computational fluid dynamics Michael Ullmana, b, Michael Durkaa, b, Kevin Glunta, b, and Matthew Barry, PhDa, b Applied Computational Fluid Dynamics Lab, bDepartment of Mechanical Engineering and Materials Science, University of Pittsburgh, PA, USA


Michael Ullman

Matthew Barry, PhD

Michael Ullman graduated from the University of Pittsburgh in December 2019 with a Bachelor’s degree in Mechanical Engineering and a minor in Physics. He plans to continue his education in the fall by pursuing a Ph.D. in Aerospace Engineering, focusing on computational fluid dynamics and fluid modeling. Dr. Barry’s research focuses on multi-physics modeling of energy systems. This ranges from terrestrial thermal-fluid-electric coupled modeling of waste-heat recovery systems to thermalelectric-mechanical coupled modeling of space power-generation systems, and includes phase-change modeling for extraterrestrial probe design and evaluation.

Significance Statement

NASA Jet Propulsion Laboratory is developing a melting probe to access Europa’s subterranean oceans in search of extraterrestrial life. This research advances the mathematical formulation of the melting process by identifying and quantifying discrepancies between models, ultimately providing insights into how the probe shall be designed to maximize its performance.

Category: Computational research

Keywords: Melting probe, Europa Clipper, Europa Lander, model validation


Observations of water vapor plumes ejected from the waterice surface of Jupiter’s moon Europa have prompted scientists to hypothesize that liquid water oceans lie beneath the surface, making Europa a primary focus in the search for extraterrestrial life. NASA Jet Propulsion Laboratory (JPL) is developing plans for its Europa Lander mission, during which a probe will melt through the surface ice sheets to access Europa’s subterranean oceans. An analytical model is being developed by JPL and the University of Pittsburgh to compute the probe’s melting performance envelope, but this model requires corroboration from numerical models. In this project, the boundary conditions of the analytical model were implemented into ANSYS-CFX finite-volume models to evaluate the analytical model’s validity. All of the numerical models exhibited heat flux distributions qualitatively similar to the analytical model on the front and side of the ice cavities, but the cavity shapes differed from what was desired. The greatest discrepancies occurred at the front corner of the probe, suggesting that radial dissipation of axial heat flux must be considered at this location. The results provide insights into the applicability of the analytical model and how the desired melt profile may be achieved.

1. Introduction

Because terrestrial life originated and thrives in Earth’s oceans, biologists hypothesize that the presence of water may be essential for life to emerge. To test this theory, astrobiologists look to examine extraterrestrial environments where water and organic compounds are plentiful. The discovery of life in these environments would help to elucidate how life developed on Earth and answer one of the most profound questions in science—are we alone in the universe? About 400 million miles from Earth, the smallest of the Galilean moons, Europa, orbits its home planet of Jupiter. This moon is notable for its thin, oxygen-rich atmosphere and fractured water-ice surface, which is splotched with red-brown hues believed to result from salt and sulfur compounds discolored by radiation [1]. Recent analyses of data from NASA’s Galileo orbiter suggest that the spacecraft flew through a plume of water vapor when passing close to Europa in 1997 [2]. Water vapor ejections have also been observed by the Hubble Space Telescope, with periodicity consistent with predicted variations in Europa’s tidal forces [3]. Scientists have hypothesized that these plumes originate within a water ocean beneath Europa’s icy surface, making it a primary focus in the search for extraterrestrial life. Because of this promise, NASA is developing plans for its Europa Lander mission, which will consist of a probe landing on and penetrating the moon’s surface to explore its subterranean oceans. The proposed method for penetrating the ice is a combination of drilling and melting—the latter of which will be caused by nuclear heat generation within the probe. Engineers at NASA Jet Propulsion Laboratory (JPL) and researchers at the University of Pittsburgh are developing a system of nondimensional equations—hereafter referred to as the JPL analytical model—to 97

determine the melting descent speed of the probe as a function of its internal heat generation and the characteristics of the ice environment. This model is comparatively simple to compute and allows for rapid trade-space studies by minimizing the number of system variables. Because of this utility, the model will eventually be implemented in a Monte Carlo simulation to ascertain the performance envelope of the probe while accounting for uncertainties associated with the ice environment. Many simplifications and assumptions are employed in the analytical formulation, so it requires validation from numerical modeling and experimental testing. While experimental data exists for terrestrial ice probes, steady-state melting probe performance in cryogenic ice has not been extensively studied. Thus, numerical modeling must play a critical role in the validation of the analytical model. The purpose of this project is to use finite-volume models within ANSYS-CFX to provide a framework for validating the analytical model and evaluating its shortcomings. These numerical models facilitate a more comprehensive understanding of the melting process, thereby providing insight into how the probe’s design can be optimized.

2. Methods 2.1 Assumptions and Conditions The foundation for the JPL analytical model was drawn from the work of Haldor W.C. Aamot [4]. This formulation models a cylindrical cavity in the ice with a constant radius and length, and calculates the heat required to produce a given descent speed in a given ice environment. The heat terms considered in this project are the axial heat beneath the front end of the probe, which maintains the prescribed descent speed, and the radial heat along the side of the probe, which maintains the desired water jacket thickness and prevents the probe from getting stuck on refrozen ice. The front end heat has two components—conduction and melting. The conduction term is the heat required to warm the ice beneath the probe from its far-field temperature to its melting temperature, while the melting term is the heat required to complete the phase change from solid to liquid. Both of these terms, as well as the side heat, depend upon the properties of the ice—namely, its density, specific heat, thermal conductivity, and latent heat of fusion. While future versions of the analytical model will include temperature-dependent ice properties, the model was only equipped to utilize temperature-independent properties at the time of this project. For the sake of comparison, the numerical models presented utilize temperature-independent water and ice properties, taken to be tabulated values at the melting temperature of ice—273.15 K. These values, listed in Table 1, were provided by project collaborators at JPL.

98 Undergraduate Research at the Swanson School of Engineering


Density (kg/m3)

Specific Heat (J/kg-K)

Thermal Conductivity (W/m-K)

Reference Specific Enthalpy (J/kg)











Table 1: Temperature-independent material properties, taken to be tabulated values at 273.15 K.

Apart from the melting heat at the front end, the analytical model only considers heat conducted into the ice surrounding the probe—convection in the water jacket which develops around the probe is omitted from the front and side heat terms. The model assumes an infinitesimal water jacket thickness along the front of the probe, such that it is in contact with the phase change region. To maintain negligibly small viscous frictional forces on the probe’s walls and allow debris to flow around the probe, a 6 mm gap is desired between the side of the probe and the ice cavity. Thus, because the probe design is a cylinder of radius 11.5 cm and length 2.1 m, the analytical model prescribes a 12.1 cm radius for the ice cavity. Another assumption of the model is that the melting process is at steady state. In this project, steady state was defined as the point at which the cavity profiles no longer changed with time in transient CFX simulations. All of the models created for this project used an advecting ice scheme. Like a car in a wind tunnel, the probe was held stationary, while ice was forced through the domain at the prescribed descent speed of 37.5 cm/hr and far-field ice temperature of 160 K. This allowed for greater computational speed, as a moving mesh model—i.e., using stationary ice and a moving probe—would require the domain to be remeshed after each time step. 2.2 Ice-Only Model The first numerical model developed for the analytical comparison was an ice-only model, which used a cavity geometry with a 12.1 cm radius and only solved the energy balance equation within the CFX solver. The front and side walls of the cavity were prescribed to be the melting temperature of the ice—273.15 K—while the rear wall was prescribed to be adiabatic—i.e., perfectly insulated. This model served as the most direct analog for the analytical model, as the ice cavity profile was explicitly imposed, and only the heat conducted into the surrounding ice was modeled. In this case, the analytical model would be corroborated if the conductive heat fluxes at the front and side of the ice cavity matched the values and profiles from the analytical calculations. 2.3 Water-Ice Model (Conduction-Only) The second model was a water-ice conduction model, which used the probe radius of 11.5 cm and only solved the energy equations within CFX. According to the analytical model, which calculates a heat flux profile conducted into the side of the ice cavity, the requisite heat flux on the side of the probe can be found by scaling the side cavity flux by the ratio of the cavity and probe radii. Thus, the analytical side cavity heat flux profile was scaled by this factor—(12.1 cm)/(11.5 cm) = 1.052—and applied to the

Ingenium 2020

side wall of the probe. On the front end of the probe, the sum of the analytically-calculated conduction and melting heat fluxes was applied. On the rear wall, an adiabatic condition was applied. In this scenario, the analytical model would be supported if there were a miniscule cavity thickness at the front end, a 6-mm-thick cavity along the side wall, and heat fluxes at the cavity interface which match the analytically-calculated values and profiles. 2.4 Water-Ice Model (With Convection) In order to capture the pertinent physics that the analytical model omits, the second numerical model was modified to include the effects of convection within the water annulus by enabling the fluid momentum solver within CFX. In previous simulations, it was found that placing fillets on the corners of the probe allowed for more substantial water flow from beneath the probe, facilitating complete ice melting in the front corner region and preventing

undesirable water pooling. Consequently, a 1 cm fillet was added to the front and rear corners of the probe in this model. The purpose of this model was to examine the effects of implementing convection on the shape of the ice cavity and the conductive heat fluxes at the cavity interface. This model illustrates whether the previously omitted physical effects play a negligibly small role in the analysis, thereby providing insight into whether the analytical model can be applied in a more realistic context.

3. Results

Once each of the models reached steady-state, the results were analyzed and compared to the values calculated by the analytical model. Figure 1 shows the conductive heat flux and cavity thickness profiles along the length of the probe, while Figure 2 shows the profiles along the front end. The integrated heat values and mean cavity thicknesses are given in Table 2. Figure 1: Left (a): Heat fluxes into side of ice cavity along probe height, Right (b): Thickness of ice cavity along side of probe.

Figure 2: Left (a): Heat fluxes into front of ice cavity along probe radius, Right (b): Thickness of ice cavity along front end of probe.


Side Wall: Model

Front End:

Ice Conduction (W)

Mean Ice Cavity Thickness (mm)

Ice Conduction (W)

Mean Ice Cavity Thickness (mm)

Aamot (Analytical)










Water-Ice: ConductionOnly





Water-Ice: With Convection





Table 2: Comparisons between analytical and numerical models: heat and cavity thicknesses.

4. Discussion

Figure 1a shows that all three models exhibit radial heat flux profiles qualitatively similar to the analytical model along the side of the ice cavity. This is especially true for the ice-only and water-ice conduction profiles, which are nearly indistinguishable from the analytical profile for nearly the entire probe length. However, near the front end, the radial heat fluxes at the ice interface are much smaller than the analytical values, which asymptotically approach infinity. Notably, the radial heat flux profile for the water-ice model with convection initially matches the profiles from the other models, but begins to increasingly deviate at a quarter of the probe’s length. Figure 1b shows that this increase in radial conduction facilitated by convection within the water annulus drastically increases the side cavity thickness. Indeed, the mean side cavity thickness in this model, 5.811 mm, is nearly the desired 6 mm value. The smaller side cavity thickness in the water-ice conduction model illustrates that the scaled analytical radial conduction profile does not yield an ice cavity with the desired 12.1 cm radius—an erroneous assumption of the analytical model. Also, neither of the water-ice models yield a vertical side cavity, which is necessary if the numerical models are to serve as analogous validation points for the analytical formulation. Figure 2a shows that the axial heat flux profiles along the front end of the cavity in the ice-only and water-ice conduction models closely agree with the analytical model for the first 10 cm of the cavity radius, but diverge at the corner of the cavity. This is likely due to radial dissipation of the front end heat flux, which is most pronounced at this location. In the ice-only model, the ice beneath the cavity must be warmed to 273.15 K to satisfy the prescribed front end boundary condition. When the heat used to warm the ice beneath the cavity radially dissipates into the far-field, more heat must be conducted through the cavity interface to maintain the prescribed temperature. In the case of the water-ice model conduction, the spike in heat flux at the corner of the cavity corresponds to incomplete melting at the probe’s front end, causing the conductive heat flux to approach the prescribed front wall heat flux. As shown in Figure 2b, this model yielded a cavity thickness of 0.178 mm at the front end of the probe, corresponding to the approximate thickness of the

100 Undergraduate Research at the Swanson School of Engineering

phase change region in the numerical models. This suggests that when convection is omitted, the analytical assumption of an infinitesimal water jacket thickness beneath the probe is correct, in a thermodynamic sense. However, as in the side cavity, the addition of convective heat transfer thickens the front cavity considerably. The flow within the water annulus carries heat away from the front cavity, corresponding to the smaller axial conductive heat fluxes illustrated in Figure 2a. These axial heat fluxes vanish at the front corner of the cavity because of the curvature of the cavity in this region, which itself is caused by the filleted corner of the probe in this model. As in previous simulations, the filleted corner allowed the ice in this region to full melt, thereby preventing water from pooling beneath the probe. As shown in Table 2, the ice-only model has the best agreement with the analytical model on both of the front and side of the ice cavity. On the front end, the numerical value is 7.4% higher than the analytical value, due to the spike in axial heat flux at the corner. On the side, the numerical value is 0.4% lower than the analytical value, due to the analytical model’s unrealistically large radial heat flux at the front of the cavity. In the water-ice conduction model, there is a similarly attributed discrepancy in heat at the side of the ice cavity. However, the front end heat falls below the analytical value because of the smaller ice cavity radius in this model, which corresponds to a smaller area over which the heat fluxes are integrated. As expected from Figures 1a and 2a, the largest heat discrepancies occur in the water-ice model with convection. Here, warm water carried out of the front cavity and into the side leads to a smaller axial conduction along the front end and a larger radial conduction along the side wall.

5. Conclusion

The numerical models implemented in this project indicate that the JPL analytical model is not an exhaustive description of the probe’s melt-driven descent. When convection in the water annulus is not considered, the analytical model successfully models conductive heat transfer along the majority of the front and side of the ice cavity. However, large discrepancies in both conductive heat terms occur at the front corner of the cavity. This suggests that the radial dissipation of the axial heat fluxes at this location must be considered in the analytical formulation. As demonstrated in the water-ice conduction model, this radial dissipation leads to an ice cavity radius which is smaller than desired near the front corner of the probe. When convection is considered, the heat fluxes at the cavity interface deviate from the analytical model, but the average width of the side cavity expands to nearly the desired value of 6 mm. Unlike in the analytical model, though, this side cavity is not vertical—a necessity for the model comparison. New probe boundary conditions are being investigated with the goal of creating a vertical 6-mm-thick side cavity within a regime which considers convection. Achieving this will create a robust numerical analog for the analytical model, which takes the pertinent physical phenomena into account and can then be used to identify correction factors for the analytical formulation. This revised analytical model will then be used in a Monte Carlo

Ingenium 2020

simulation, which will determine the probe’s melting performance envelope for a variety of possible Europan ice environments. To this end, this project provides key insights into understanding where the analytical model can be improved and how the discrepancies with the desired melt cavity profile may be mitigated.

6. Acknowledgements

I would like to thank Dr. Barry, the Swanson School of Engineering, and the Office of the Provost for sponsoring my SSOE Summer Research Internship. I would also like to thank Dr. Miles Smith and the team at NASA-JPL for their contributions to this project. Computational resources were provided by the University of Pittsburgh Center for Research Computing.

7. References

[1] Europa, NASA Science Solar System Exploration (2019). https://solarsystem.nasa.gov/moons/jupiter-moons/europa/indepth/. [2] J. Xianzhe et al, Evidence of a plume on Europa from Galileo magnetic and plasma wave signatures. Nature Astronomy 2 (2018) 59–464. [3] L. Roth et al, Transient water vapor at Europa’s south pole, Science 343 (2014) 171-174. [4] H.W.C. Aamot; Cold Regions Research & Engineering Laboratory, Heat transfer and performance analysis of a thermal probe for glaciers, Ft. Belvoir Defense Technical Information Center, 1967.


Crimped polymer microfibers produced via electrospinning: A review Nikolas J. Vostala Department of Mechanical Engineering and Material Science University of Pittsburgh, PA, USA a

Nikolas Vostal

Nikolas Vostal is a Junior at Pitt studying Materials Science and Engineering who grew up in Plymouth, Michigan. His interests lie in polymer engineering and composite material design, although outside the classroom he is the president of Pitt’s Material Advantage Chapter and a member of the Pitt Squash team.

Significance Statement

Materials with high strength often suffer from poor flexibility and vice versa. Composites reinforced with crimped microfibers can allow excellent flexibility at low strain but high strength at high strain. This work reviews methods to create such fibers by means of electrospinning and their potential applications.

Category: Review Paper

Keywords: electrospinning, crimped fibers, polymer microfibers


Until recently, the production of small-diameter fibers and other micro-scale materials have been expensive and difficult. However, in recent years electrospinning has become popular as a plausible, cost-effective means of creating microfibers for a number of different applications. It has also been found that by adjusting the setup of electrospinning, it is possible to create patterned microfibers with unusual properties. One possibility is the creation of crimped microfibers whose wavy nature allows them to be extremely flexible until sufficient strain is applied to straighten them. On a large-scale, mats comprising crimped fibers can be used to create materials which can reliably deform and return to their original position. Such properties have applications in many different fields, most notably the biomedical field, where crimped fibers can mimic the wavy collagen fibers found in organic tissue. This article reviews many of the successful methods of producing crimped nanofibers and their current applications.

1. Introduction

Electrospinning is a simple and versatile method of creating polymeric microfibers. The most basic electrospinning setup utilizes a syringe of polymer solution placed into a syringe pump [1]. The tip of the syringe is electrically charged via a high voltage generator while a nearby metallic collector is grounded. As the solution is slowly pumped out of the syringe, charge builds up around the droplet that forms at the tip of the needle. The built-up charge causes a portion of the droplet to jump across the gap, stretching and drying along its path and hence landing on the collector as a micron-sized fiber [2]. As the pump continues to expel solution, fibers land randomly across the collector, forming a nonwoven mat. While these random mats have found numerous applications, it is also possible to alter the shape and movement of the collector so that fibers land in an oriented fashion with transverse isotropy [3-5]. Aligned fiber mats can be woven to fabricate complex micropatterns that have greater tensile resistance [2].

Figure 1: Diagram of a typical electrospinning setup, Reproduced from Ref 6.

102 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

2.1. Heat Treatment When fibers are deposited during the electrospinning process, residual stresses may build up in the fibers. These stresses are caused by the whipping and stretching of the fibers when pulled from the needle and how the fibers land upon the collector [12]. In order to release these stresses, fibers can be subjected to temperatures above their glass transition temperature (Tg) [13]. This allows the stresses to relax and expand at different points along the fiber, creating a sinusoidal waviness. The degree of crimp depends upon how much higher the temperature gets above Tg [10]. It has also been shown that by spinning additional sacrificial fibers simultaneously, an interwoven fiber web can be produced. Once the sacrificial fibers are dissolved away, the stress in the remaining fibers leads to greater crimping [14]. A drawback of heat treating is that many polymers will begin to lose their crimp over time [13].

In recent years the concept of using electrospinning to produce crimped microfibers has emerged. The DuPont company first manufactured crimping fibers in the 1960’s using yarnpolypropylene through a process called melt spinning. While these fibers saw a large use in the textiles industry, they were limited in other fields by their large diameter [1]. With the increased use of electrospinning, old techniques are being used alongside new innovations to crimp micron-sized fibers for applications that were once impossible [7]. Crimped fibers are fibers which exhibit a waviness that allow them to flex and unbend before becoming taut and taking on stress [8-9]. On a stress-strain graph this is exhibited as a region of gradual increase in stress at small strains followed by a sharp increase. Most crimped fibers have a sinusoidal waviness but some exhibit a 3-D helical shape. The exact cause of crimping is still debated, but regardless of the shape, crimped fibers seem to form by either imposing external stress on the fiber during spinning or by releasing residual stresses in the fiber after spinning [10-11].

2.2. Plasticizer Treatment Much like the thermal treatment, the residual stresses of electrospinning can be released through a plasticizer treatment. This is achieved by taking spun fibers and clamping them at a length shorter than their initial length, inducing slack. The clamp and fibers are then submerged into a plasticizer of the polymer fibers, such as ethanol for polylactic acid (PLA) [15]. The plasticizer is then absorbed by the fibers and allows the polymer chains to release the residual stresses of spinning. This brings the chains to a lower energy conformation, resulting in the formation of crimped fibers [8]. No studies have been conducted concerning the longevity of such fibers: however, it would be interesting to examine the effects on crimping over time after the plasticizer is removed and begins to evaporate.

Figure 2: (a) An SEM image of sinusoidally crimped fibers reproduced from Ref 8, and (b) An SEM image of helically crimped fibers, reproduced from Ref 19.

This review seeks to document all reported methods of crimping electrospun fibers and attempts to explain the reasons behind crimping. It also discusses the few current applications of these fibers, although further research into the reliability, mass production, and exact physical properties is needed to truly understand their potential impact.

2.3. Bubble Electrospinning / Bubbfil Spinning Bubble electrospinning is a novel approach to producing crimped fibers, but few studies have explored it. Bubble electrospinning is different from the previous two examples because crimping occurs during the spinning process, before fibers land on a collector. In bubble electrospinning the syringe and syringe pump are replaced with a reservoir of solution. The entire reservoir is electrically charged and the collector plate is

2. Crimping Methods

Table 1 below offers a quantitative summation of the materials and techniques reviewed. Material

Material % in Solvent

Crimp Method Used

Crimp Type

Reported Spin Voltage (kV)

Average Fiber Diameter

Young’s Modulus, E



4.5% in hexafluoroisopropanol

Ethanol Treatment



600 nm

100 ± 22 MPa



8.5% in hexafluoropropylene




300 ± 100 MPa



14% in 3:2 TFA/DCM

Two- Spinneret Compounding


15 cm, 1 kV/cm


14% in 3:2 TFA/DCM

Two- Spinneret Compounding


+15 to +22 and -8 to -3



13% in DMF

Two- Spinneret Compounding


+15 to +22 and -8 to -3



5% in 3:1 DCM/ DMF

Heat Treatment in PBS


15 cm, 1 kV/cm

0.82 ± 0.02 µm

18 ± 2.5 MPa



5% in 3:1 DCM/ DMF

Heat Treatment in PBS


15 cm, 1 kV/cm

0.87 ± 0.03 µm

14 ± 2 MPa



5% in 3:1 DCM/ DMF

Heat Treatment in PBS


0.88 ± .002 µm

349 ± 69 kPa




Bubble Spinning



5 and 7% in dry DMF



115-590 nm


11% in hexafluoro-2-prpanol

Ethanol Treatment


400 nm

800 nm

18 & 19

17 & 20

Table 1: Information retrieved from various crimped fiber electrospinning articles. Omitted spaces denote untested areas.

10 575 ± 200 MPa



placed directly above. At the bottom of the reservoir a gas pump is connected which dispenses compressed gas slowly, allowing bubbles of air to form that float up through the solution. Once the air bubble pops the built-up charge in the solution launches small fibers toward the collector, landing with a sinusoidal crimp [16]. In order to create crimped fibers using this method large transverse vibrations are required in the fibers while they move to the collector. The only successful report of this was by Huang et al, who achieved large enough vibrations by increasing the temperature of the reservoir [17]. It was not stated whether the use of heating during spinning had the same long-term unraveling as other post-spinning treatments.

opposing charges cause the fibers to collide midair, forming a compound fiber. The compound fiber then continues horizontally towards the collector due to the remaining charge after the collision [19]. Since this method does not rely on chemical or mechanical post-processing the fibers maintain their crimp much longer than the previously mentioned methods [18].

Figure 4: Diagram of a two-spinneret electrospinning setup, reproduced from Ref 19.

3. Applications

Figure 3: Diagram of a bubble electrospinning setup.

2.4. Two Spinneret Compounding The final reported method of crimping electrospun microfibers was by using two separate spinnerets with different polymers and spinning them simultaneously. As the spinnerets eject solution the fibers meet in midair and land together on the collector. The fibers shrink in size before hitting the collector, but because the two fibers shrink at different rates, they curl around each other in a helical pattern [18]. There are a number of publications that have reported producing helical fibers, but the results vary drastically with spin conditions and may form large fiber loops and ribbons instead of crimped waves [19]. The two most successful techniques found differ by spinneret placement. In one method the spinnerets sit side-by-side, while the other places them perpendicular. Side-by-side spinning is the simpler of the two, as both needles can be placed next to each other and carry the same charge [18]. However, for perpendicular electrospinning the fibers must intersect in midair and continue to the collector. This is achieved by inducing the vertical spinneret with a smaller opposing charge to the horizontal spinneret. The

104 Undergraduate Research at the Swanson School of Engineering

In the biomedical field crimped microfibers are an important innovation for modeling organic tissue, specifically the collagen fiber. Collagen is a naturally wavy microfiber and is a crucial part of many different tissues such as the skin, tendons, ligaments, and veins [20]. The wavy (or crimped) nature of collagen allows tissue to flex and stretch freely within a certain range [21]. Beyond this range, the fibers start to resist strain and become taut [22-23]. Since electrospun fibers act similarly, they have been considered for a number of different supportive and regenerative applications. At this time most of the application research has been centered around using electrospun fibers as regenerative scaffolds. Regrowth of tissue like the anterior cruciate ligament (ACL) after reconstructive surgery can take up to a year [24]. By growing collagen cells on crimped biocompatible fibers replacement collagen fibers can be produced and transplanted in a fraction of the time it takes for the body to regrow naturally [12]. Few other applications have been explicitly explored at this time due to the relative newness and mass production restrictions of these methods [7]. However, crimped fibers have the potential to transform a number of different fields such as air and water filtration, radiation protection, fuel cell electrodes, and flexible electronics [1].

4. Conclusion

Electrospinning is a novel, but versatile process that can be used to create crimped fibers on the micron-scale. This can be accomplished in a number of different methods, including heating, plasticizer treatment, bubble electrospinning, and twospinneret compositing. Regardless of the method intermolecular stresses are used to bend the fibers at different points, forming

Ingenium 2020

crimped sinusoidal waves or 3-D helical shapes. While only briefly explored in the biomedical field as a regenerative scaffold, crimped microfibers have many potential applications. Compared to conventional methods the cost of electrospinning is much lower and its ability to create fibers at small diameters is unparalleled, although scaleup to larger volumes remains challenging. Further research into reliable fibers with larger yields is needed to explore the many diverse applications of crimped microfibers.

5. Acknowledgments

This paper would not be possible without the resources provided by the Swanson School of Engineering and the Office of the Provost. I would also like to thank Dr. Sachin Velankar for his guidance and insight.

6. References

[1] R. Kotek, “Recent Advances in Polymer Fibers”, Polymer Reviews 2008, vol 48, 221-229. [2] Li and Xia, “Electrospinning of Microfibers: Reinventing the Wheel?”, Adv Mater. 2004, vol 16, 1151- 1170. [3] Y Ishii et al. “A new electrospinning method to control the number and a diameter of uniaxially aligned polymer fibers”, Mater. Letters 2008, vol 62, 3370-3372. [4] P Katta et al, “Continuous Electrospinning of Aligned Polymer Microfibers onto a Wire Drum Collector”, Micro Letters 2004, vol 11, 2215- 2218. [5] K Zhang et al, “Bionic electrospun ultrafine fibrous poly(Llactic acid) scaffolds with a multi-scale structure”, Biomed. Mater. 2009, vol 4, 035004. [6] Bhattarai et al, “Biomedical Applications of Electrospun Nanofibers:Drug and Nanoparticle Delivery”, Pharamceutics 2019, 11, 5. [7] W. Teo et al “Technological advances in electrospinning of microfibers”, Sci. Technol. Adv. Mater. 2011, vol 12, 1-19. [8] W. Liu et al, “Generation of Electrospun Microfibers with Controllable Degrees of Crimping Through a Simple, PlasticizerBased Treatment”, Adv. Mater. 2015, vol 27, 2583-2588. [9] S.P. Rwei “Study of Self-Crimp Polyester Fibers”, Polymer Engr. And Sci. 2005, vol 45, 838- 845. [10] L. Peciulyte et al, “Thermal Imidization Peculiarities of Electrospun BPDA-PDA/ ODA Copolyamic Acid Microfibers”, Macromolecular Research 2013, vol 21, 419-426. [11] Demšar & Sluga, “Crimped Polypropylene Yarns”, Kovine Zlitne Tehnologije 1999, vol 33. [12] D. Surrao et al, “Biomimetic poly(lactide) based fibrous scaffolds for ligament tissue engineering”, Acta Biomaterialia 2012, http://dx.doi.org/10.1016/j.actbio.2012.07.012. [13] D. Surrao et al, “Self-Crimping, Biodegradable, Electrospun Polymer Microfibers”, Biomacromolecules 2010, vol 11, 3624–3629. [14] Szczesny et al, “Crimped Microfibrous Biomaterials Mimic Microstructure and Mechanics of Native Tissue and Alter Strain Transfer to Cells”, ACS Biomater. Sci. Eng 2017, vol 3.

[15] Pavlova et al, “Tuning the properties of electrospun polylactide mats by ethanol treatment”, Materials & Design 2019, vol 181. [16] R-X Chen et al, “Mini-review on Bubbfil spinning process for mass-production of microfibers”, Revista Mater 2014, vol 19, 325- 344. [17] JX. Huang et al, “Effect of Temperature on Nonlinear Dynamical Property ...”, Thermal Science 2014, vol 18, 10491053. [18] B. Zhang et al, “Curled Poly(ethylene glycol terephthalate)/ Poly(ethylene propanediol terephthalate) Microfibers Produced by Side-by-side Electrospinning”, Polymer Journal 2009, vol 41, 252–253. [19] C. Li et al, “Direct Formation of ‘‘Artificial Wool’’ Microfiber via Two-Spinneret Electrospinning”, Journal of Applied Polymer Science 2011, vol 123, 2992-2995 . [20] JX. Huang et al, “Transverse Vibration of an Axially Moving Slender Fiber of…”, Thermal Science 2015, vol 19, 14271441. [21] L.J. Gathercole and A. Keller, “Crimp Morphology in the Fibre-Forming Collagens” Matrix 1991, vol 11, 214−234. [22] Franchi et al, “Crimp morphology in relaxed and stretched rat Achilles tendon” Journal of Anatomy 2007, vol 210, 1-7. [23] M.B. Bennett et al, “Mechanical properties of various mammalian tendons” J. Zool 1986, vol 209, 537−548. [24] Mayo Clinic staff, ACL injury, https://www.mayoclinic.org/ diseases-conditions/acl-injury/diagnosis-treatment/drc-20350744, accessed 10/24/2019.


Adventitial extracellular matrix from aneurysmal aorta fails to promote pericyte contractility Kaitlyn Wintrubaa, Bryant Fisherb, Jennifer C. Hillb, Tara D. Richardsb, Marie Billaudb-d, Amadeus Sternd, Thomas G. Gleasonb-d, Julie A. Phillippib-d Department of Chemical Engineering, bDepartment of Cardiothoracic Surgery, cDepartment of Bioengineering, d McGowan Institute for Regenerative Medicine University of Pittsburgh, Pittsburgh, PA, USA a

Kaitlyn Wintruba is in her second year at the Swanson School of Engineering where she studies Chemical Engineering. After graduating, Kaitlyn plans to pursue her interest in regenerative medicine and biomedical research in graduate school. Kaitlyn Wintruba

Dr. Julie Phillippi is Associate Professor with Tenure in the Department of Cardiothoracic Surgery, School of Medicine with a secondary appointment in the Department of Bioengineering Swanson School of Engineering. Dr. Phillippi’s research is focused on the role of Dr. Julie Phillippi perivascular progenitor cells, adventitial biology, and vasa vasorum function in human aortic disease. Her work is currently funded by the National Heart, Lung and Blood Institute under award # R01HL131632.

Significance Statement

Ascending thoracic aortic aneurysm (TAA) is a lifethreatening condition, for which there are no strategies for early intervention. Our work supports the use of an extracellular matrix hydrogel as a potential regenerative biomaterial in the setting of TAA and translation towards microvascular regeneration in clinical applications.

Category: Experimental research

Keywords: Pericyte, Extracellular Matrix Hydrogel, Adventitia, Aortic Aneurysm Abbreviations: thoracic aortic aneurysm (TAA), extracellular matrix (ECM), aortic adventitia (Adv), porcine aortic adventitia (pAdv), human aortic adventitia (hAdv), fibroblast growth factor 2 (FGF2), platelet derived growth factor (PDGF), vascular endothelial growth factor A (VEGF-A), transforming growth factor beta 1 (TGFβ1), dimethyl sulfoxide (DMSO)

106 Undergraduate Research at the Swanson School of Engineering


Ascending thoracic aortic aneurysm (TAA) is a life-threatening condition lacking adequate diagnostics and risk adjudication for aortic dissection or rupture. Studying the adventitia, the outer most layer of the aorta, our group previously uncovered downregulation of several pro-angiogenic factors and reduced density of the microvascular network of vasa vasorum in aneurysmal aortic specimens. Knowledge of how adventitial extracellular matrix (ECM) influences the function of vasa vasorum-associated cells could potentially lead to developing a better diagnostics and therapies for TAA or aortic dissection. We hypothesized that adventitial ECM from normal aorta enhances pericyte function through a growth factor-mediated mechanism deficient in aneurysm-derived aortic adventitia. To test this hypothesis, we quantified pericyte contractility within a 3D hydrogel tissue culture scaffold as a measure of pericyte function. Human aortic adventitia (Adv)-derived pericytes were cultured within 2mg/mL bovine Type I collagen gels in the presence or absence of lyophilized human or porcine Adv ECMs. Addition of Adv ECMs to collagen accelerated gelation at 37˚C as evidenced by a higher optical density (p<0.05). Addition of normal human Adv ECM increased pericyte contractility when compared with aneurysm-derived Adv ECM (p<0.001). Inhibition of TGFβ-1R decreased porcine Adv ECM-induced contractility when compared with cells treated with vehicle and porcine Adv ECM (p<0.001). This work supports the use of porcine-derived ECM hydrogels to improve function of vasa vasorum-associated cells as a potential therapeutic biomaterial for microvascular regeneration in human aortic disease.

1. Introduction

Ascending thoracic aortic aneurysms (TAAs) are often asymptomatic, and the only treatment to prevent aortic dissection or rupture is elective replacement of the ascending aorta. Aortic dissection occurs in 5 to 30 cases per million of the population annually [1] with the mortality rates of aortic dissection worldwide being 1% per hour without surgical intervention [2]. Interactions between the cellular and extracellular matrix (ECM) components of the intimal, medial, and adventitial layers of the aorta mediate blood flow throughout the body. Currently, the complex cellular molecular mechanisms inciting TAA and driving disease progression remain incompletely understood. Improved understanding of these mechanisms could help develop less invasive treatment options. Prior work in the Thoracic Aortic Disease Research Laboratory identified involvement of the adventitial microenvironment in the pathogenesis of TAA. Growth factors within adventitial ECM such as fibroblast growth factor 2 (FGF2), platelet derived growth factor (PDGF), vascular endothelial growth factor A (VEGF-A) can influence vasa vasorum-associated cells. These pro-angiogenic growth factors involved with vasculogenic function were found to be downregulated in human aneurysmal aortic specimens and porcine-derived adventitial (pAdv) ECM exhibited FGF2-mediated angiogenic activity in vitro and in vivo [3]. Transforming growth factor beta (TGFβ) is involved in various vasculogenic functions

Ingenium 2020

and induced a contractile phenotype in aortic smooth muscle cells in vitro [4]. Within the adventitia, TGFβ serves as a pro-angiogenic growth factor influencing perivascular progenitor cell function. The vasa vasorum exhibited microvascular remodeling and lower density suggestive of reduced neovascularization in aortic specimens from TAA patients [5]. Furthermore, several populations of vasa vasorum-associated progenitor cells reside within the adventitia using analytical flow cytometry and immunohistochemistry to identify pericytes with the surface marker profile of CD146+/CD34±/CD31- [6]. Pericytes serve as perivascular progenitor cells within capillaries and micro vessels and in ex vivo culture exhibited unique spheroid formation and sprouting on Matrigel substrates [6], two behaviors that mimic vasculogenic function. Pericytes cultured in the presence of TGFβ1 and PDGF developed a spindle-shape morphology and increased expression of markers for smooth muscle lineages [6]. In the present study, we hypothesized that adventitial ECM from normal aorta promotes pericyte function by increasing pericyte contractility through a growth factor-mediated mechanism deficient in aneurysm-derived aortic adventitia. To test this hypothesis, we quantified pericyte contractility within a 3D hydrogel tissue culture scaffold as a measure of pericyte function.

2. Methods

Human ascending aortic adventitial tissue specimens were collected from patients undergoing ascending aortic and/or aortic valve replacement, or heart transplantation with Institutional Review Board approval and using an informed consent process. Pericytes (Figure 1) were isolated from the adventitia of non-aneurysmal human aorta and then immortalized using a lentiviral vector to deliver HPV-E6/E7 to avoid senescence and phenotypic changes at high passage. A surface marker profile of CD146+/CD31- indicative of pericytes was routinely enriched with magnetic bead separation, during which biomolecules are attached to magnetic beads of specific affinity and filtered through a magnetic separation column such that only positive marker expression is captured. Adventitial ECMs were prepared from decellularized and lyophilized adventitial specimens from human and porcine aorta. Human aortic tissue specimens were collected from patients with ages ranging from 17 to 82 years. The Adv ECMs were then finely ground to produce an ECM particulate. These ECMs were previously shown to contain minimal DNA content while retaining numerous bioactive growth factors [3].

Pericytes were cultured within 2mg/mL bovine Type I collagen gels in the presence or absence of normal or aneurysmal human adventitial (hAdv) ECMs (250µg/mL) within the collagen gel. Parallel experiments cultured pericyte-embedded collagen gels in the presence or absence of pAdv ECM (250µg/mL) or with TGFβ-1 receptor inhibitor SB431542 (100nM) added to the culture medium or DMSO as the vehicle control. Optical absorbance readings (405nm) over 3h of dry heat incubation (37˚C) were obtained to calculate gelation kinetics of collagen blended with hAdv and pAdv ECMs. Normalized absorbance was calculated at two-minute intervals using the following equation: (1) where A represents the absorbance reading at a particular time point, Amax represents maximum absorbance, and A0 represents the initial absorbance. Pericyte-embedded collagen gels were imaged with bright-field microscopy (Nikon SMZ25, 1X SHR Plan Apo Objective), and cell contraction was quantified by measuring the initial gel area and after 48 hours of culture (Nikon Elements AR 4.60). All experiments were performed with four assay replicates and cells obtained from two different patients (n=4). Statistical analysis was carried out utilizing one-way ANOVA with Tukey’s post-hoc test (SigmaPlot 12.5) and a p<0.05 was considered statistically significant.

3. Results 3.1 Adventitial ECM Promotes Pericyte Contractility All ECM treatments increased pericyte contractility as evidenced by decreased gel area when compared with pericyteembedded collagen gels lacking ECM treatment (p<0.001). Addition of normal hAdv ECM doubled the degree of pericyte contractility when compared with gels lacking any ECM treatment and aneurysm-derived hAdv ECM (Figure 2). Aneurysm-derived hAdv ECM failed to induce contractility above baseline (Control). Inhibition of TGFβ-1 receptor decreased pAdv ECM-induced contractility when compared with cells treated with vehicle (±DMSO) and pAdv ECM (Figure 3).

Figure 1: CD146+/ CD31- Pericytes. Representative phase contrast image of pericytes with uniform elongated cell morphology.


Figure 2: Pericyte Contractility of Human Adv ECMs. Contractility of collagen gels without pericytes (No Cell), with pericytes (Control), and with pericytes treated with 250µg/mL hAdv ECMs, respectively. Bars represent the mean contractility of four assay replicates ± standard error of the mean. * indicates p<0.001.

Figure 3: pAdv ECM-induced Pericyte Contractility ± TGFβ-1 Inhibitor. Contractility of 250µg/mL pAdv ECM treated pericyte-embedded collagen gels with TGFβ-1 inhibitor (+SB431542) or vehicle (+DMSO) and respective acellular and cell controls. Bars represent the mean contractility of four assay replicates ± standard error of the mean. * indicates p<0.001.

3.2 hAdv ECM Accelerates Collagen Gelation Calculation of normalized absorbance during gelation revealed peak gelation was reached within 100 minutes for acellular hydrogels, and in the presence of cells and Adv ECMs (Figure 4). Addition of hAdv ECM to Type I collagen accelerated gelation as evidenced by a higher optical density (Figure 4) and decreased time to 50% gelation (Table 1) (p<0.05). Conversely, addition of pAdv ECM decreased the gelation rate compared to collagen alone. Furthermore, the gelation rate was greater with the addition of aneurysmal hAdv ECM compared to normal hAdv ECM.

Figure 4: Collagen Gelation ± Adv ECMs. Lines depict normalized absorbance readings of pericyte-embedded collagen (2mg/mL) blended with normal and aneurysmal hAdv and pAdv ECMs (250µg/mL) over 100 minutes.

108 Undergraduate Research at the Swanson School of Engineering

Ingenium 2020

Rate (OD/min)

T1/2 (min)

No Cell

0.0659 ± 0.0041

54.76 ± 1.11


0.0616 ± 0.0024

50.84 ± 1.37

Normal hAdv ECM

0.0656 ± 0.0035

47.11 ± 2.16

Aneurysmal hAdv ECM

0.0675 ± 0.0039

*41.10 ± 0.78

pAdv ECM

0.0563 ± 0.0029

63.14 ± 1.20

Table 1: Gelation Kinetics of Collagen Blended with Adv ECMs. Gelation rate (optical density divided by time) and time required for 50% gelation (T1/2) of pericyte-embedded collagen blended with Adv ECMs. Data represents mean of four assay replicates ± standard deviation. * indicates p<0.05 compared to acellular collagen gel.

4. Discussion

Less pericyte contractility with aneurysm-derived hAdv ECM compared with normal hAdv ECM treatment demonstrates adverse effects in pericyte function may be caused by growth factor deficiencies in adventitia from TAA [3]. Porcine Adv ECMinduced pericyte contractility and SB431542-mediated inhibition of contractility indicates that Adv ECM influences pericyte function via a TGFβ1-mediated mechanism. TGFβ1-mediated pericyte contractility supports the hypothesis that normal hAdv ECMinduced increase in pericyte function is growth factor dependent. Notably, the validity of these results could be improved with the testing of additional cell lines. Wrinkler et al. showed pericytes wrap capillaries and express contractile proteins to potentially regulate capillary diameter by constricting or relaxing [7]. However, it remains unclear whether this is true in vivo. Another study has shown functional cell behavior in response to ECM hydrogels to be comparable to collagen substrate for cardiac applications [8]. Our work expands previous studies by examining functional effects of aortic adventitial pericyte-ECM interaction in a 3D microenvironment. ECM hydrogels represent a candidate biomaterial for less invasive treatment in the setting of TAA due to their potential to replenish growth factors and restore a pro-angiogenic adventitial microenvironment. Pericytes comprise a subpopulation of cells within the adventitia that localize to the vasa vasorum and serve as progenitor cells with vasculogenic behavior. The therapeutic potential of an ECM hydrogel likely lies, at least in part, in its ability to promote pericyte function. Future experiments should determine baseline and pAdv ECM-induced contractility of pericytes isolated from adventitia of aneurysmal human aorta. Furthermore, additional studies will seek to explore how adventitial ECM affects other pericyte functions including sprouting, migration, and chemotaxis.

5. Conclusions

Knowledge that adventitial ECM promotes pericyte function by significantly increasing pericyte contractility and accelerating hydrogel gelation might assist in developing a novel therapeutic intervention for patients at risk for TAA or aortic dissection. The pAdv-induced and TGFβ1-mediated increase in pericyte contractility supports the use of porcine-derived ECM hydrogels to improve function of vasa vasorum-associated cells as a potential therapeutic biomaterial as part of a microvascular regeneration strategy to treat human aortic disease.

6. Acknowledgements

This study was supported by the National Institutes of Health under award #HL131632 (JAP), #HL127214 (JAP) and #HL109132 (TGG), UPMC Health System Competitive Medicine Research Fund (JAP) and the McGowan Institute of Regenerative Medicine Summer School Program (KLW).

7. References

[1] Davies RR, Goldstein LJ, Coady MA, et al. Yearly rupture or dissection rates for thoracic aortic aneurysms: simple prediction based on size. Annals of Thoracic Surg. 73 (2002) 17-28. [2] Masuda Y, Yamada Z, Morooka N, et al. Prognosis of patients with medically treated aortic dissections. Circulation. 84 (1991) 117-113. [3] Fercana GR, Yerneni S, Billaud M, et al. Perivascular extracellular matrix hydrogels mimic native matrix microarchitecture and promote angiogenesis via basic fibroblast growth factor. Biomaterials. 123 (2017) 142-154. [4] Crosas-Molist E, Meirelles T, López‐Luque J, et al. Vascular smooth muscle cell phenotypic changes in patients with Marfan syndrome. Arteriosclerosis, Thrombosis, and Vascular Biology. 35 (2015) 960-972. [5] Billaud M, Hill JC, Richards TD, et al. Medial hypoxia and adventitial vasa vasorum remodeling in human ascending aortic aneurysm. Front. Cardiovasc. Med. 5 (2018) 124. [6] Billaud M, Donnenberg VS, Ellis BW, et al. Classification and functional characterization of vasa vasorum-associated perivascular progenitor cells in human aorta. Stem Cell Reports. 9 (2017) 292-303. [7] Winkler EA, Rutledge WC, Kalani MYS, et al. Pericytes regulate cerebral blood flow and neuronal health at a capillary level. Neurosurgery. 81:5 (2017) N37-N38. [8] DeQuach JA, Mezzano V, Miglani A, et al. Simple and high yielding method for preparing tissue specific extracellular matrix coatings for cell culture. PLoS One. 5 (2010) 130-139.


Biotelemetry: A brief history and future developments in lowering cost Kevin Xua, Mark Gartnera Department of Bioengineering, University of Pittsburgh, Pittsburgh, PA a

Kevin is a junior in the Department of Bioengineering on the cellular engineering track from Mars, PA. Kevin has always had an interest in connecting biomedical concepts to technology and takes pride in helping to find novel approaches to existing problems. Kevin Xu

Mark Gartner

Dr. Gartner is a professor in the Department of Bioengineering, primarily teaching the Senior Design course. After receiving degrees from both Pitt and CMU, Dr. Gartner began his work in medical product design and development at UPMC. Dr. Gartner also co-founded Enison, Inc., a vertically-integrated incubator that developed medical products based on a “surface first� philosophy.

Significance Statement

The study of wildlife behavior is extremely important but is hard to accomplish in its current state due to the lack of accessibility for biotelemetric devices. Through the development and adaptation of current tracking systems, researchers are creating lower-cost devices that will allow for more widespread study of animal species.

Category: Review/perspective paper

Keywords: Biotelemetry, wildlife, tracking, history

110 Undergraduate Research at the Swanson School of Engineering


Biotelemetry is crucial to a variety of wildlife and conservation-related assessments. The three main systems used in biotelemetry today (very high frequency transmitters, global positioning system tracking, and satellite tracking) all have advantages and disadvantages, but the expense and difficulty of implementing biotelemetric hardware remains a barrier to entering the field. The recent research and development into newer systems has made it easier to get involved in biotelemetry, yet the cost of hardware still makes it difficult to study many animals in diverse and widespread areas. Current research into creating lower-cost tracking devices using off-the-shelf, open-source hardware have helped pushed for more access to biotelemetric devices. The continuation of this research and the push for more accessible biotelemetric devices will allow researchers not to only learn more about wildlife behavior, but also factors such as wildlife biology and ecology.

1. Introduction

Wildlife research is a long-standing and extensive field, but the study of animals and wildlife has certainly not been easy. Humans have lived among animals for thousands of years, but for most of this time, information about animals was gathered by simple observation and chance, rather than finding a systematic or quantitative approach to observation. However, with the development of biotelemetry in the 1960s, researchers have been able to improve their study of the general movement and behavior of animals. Biotelemetry involves the capture and tagging of a species of interest with a transmitter device. Once tagged, the device transmits radio signals to reveal the location of the transmitter and also to relay any other information or data that may be collected. Location data is extremely important and can be used to study an animal’s preferred habitat, home range, and to understand population dynamics. Details into animal movement can reveal fundamental behaviors such as how the animal acquires food, shelter, or mates or how they survive in general. Typically, biotelemetry involves three different techniques: very high frequency (VHF) transmitters, global positioning system (GPS) tracking, and satellite tracking. VHF tracking is also known as direct tracking and is used in close proximity to the tracker in order to find the exact location of a tagged animal. GPS and satellite tracking allow an animal to be tracked globally and is useful for remote tracking or for tracking migrating animals, since locations can be accurately determined regardless of distance. While exciting developments are being made in the field of biotelemetry in the improvement of technologies, biotelemetric hardware is extremely expensive and frequently challenging to implement [1]. As a result, only a small number of units are typically purchased and only the most at-risk animals are studied. In addition, the access to these technologies can be extremely difficult in developing countries, and as a result, limitations on sample size must be made in studies [2]. In this paper we will explore (1) the progression and development of new technologies

Ingenium 2020

in the field of biotelemetry and (2) the pursuit of designing and developing low-cost wildlife trackers in order to improve access to studying a larger number of species.

2. Methods

This paper will first aim to explore the development of technologies and techniques in the field of biotelemetry, first starting with the advent of wildlife tracking devices in the use of VHF tracking. The development of very commonly used systems such as GPS tracking and satellite tracking will also be explored, as well as future goals and ambitions for tagging and tracking. This paper will also delve into the development of lowcost wildlife tracking devices, which typically use off-the-shelf hardware that has been modified to serve as a substitute for the more expensive systems currently in use. Many research articles regarding low-cost wildlife tracking devices are written in the study of specific species (e.g. pampas deer, brushtail possum, etc.), but the techniques and goals are applicable to the study of low-cost devices in general. This paper will serve as a general discussion of low-cost tracking devices in the interest of increasing access to studying all animals, not just specific species.

3. Discussion 3.1. History of Wildlife Tracking Devices 3.1.1. VHF Tracking

Figure 1: A park ranger employing VHF tracking to triangulate and track mountain lions.

Very High Frequency (VHF) technology was the first main technique used in biotelemetry to track and identify individual animals. The first successful system was tested in 1963 on rabbits, striped skunks, and raccoons [3]. A VHF tracking system consists of two main components that include the transmitter device used to tag the animal as well as the remote receiver which is usually a hand-held antenna. The location of the transmitter and animal can be determined by triangulation, which requires the collection of transmission data from three or more different locations around a single point.

VHF tracking is a relatively cheap and low-cost method of biotelemetry that can also be long lasting due to the low-power requirements of transmitters. It can be used on a variety of animal species, from insects to large mammals, which makes it very effective and widely applicable. However, VHF tracking is incredibly laborious and requires on-site tracking, since the transmission signal can only be received nearby with an antenna. Furthermore, triangulation of data can be extremely tedious and while the initial cost of a device can be inexpensive for VHF tracking, the individual cost of each data point can become extremely costly [4]. 3.1.2. Satellite Tracking

Figure 2: Various species fitted with Argos transmitters, demonstrating its widespread use.

Satellite tracking offers an immediate advantage to VHF tracking in that it allows for global tracking. This is mainly due to the existence of the ARGOS satellite system, which is almost universally used for wildlife tracking. Satellite biotelemetry has been around since the mid-1980s, after an agreement between the French Space Agency, National Oceanic and Atmospheric Administration (NOAA), and National Aeronautics and Space Administration (NASA) allowed ARGOS to be used exclusively for the collection of environmental data [5]. ARGOS satellites identify signals sent by transmitters that are called Platform Transmitter Terminals (PTT), which have been miniaturized over time to allow attachment to animals. They do so using the Doppler effect, which allows the satellites to iteratively integrate signals, using the Doppler shift to measure the receive frequencies of incoming messages. Finally, the location of the PTT is estimated based on the satellite location combined with these integrated Doppler shift signals. Unfortunately, while the ARGOS satellite system allows for global tracking, PTT units are extremely expensive and can be quite inaccurate. The use of Doppler shift signals means that it is hard to obtain a successful location fix, as Argos processing centers require multiple messages from a transmitter to make the proper estimations. Furthermore, due to the scarcity of satellites in the ARGOS system, only a limited number of fixes can be acquired in a day and researchers must hope that a satellite passes overhead and obtains a successful fix for accurate data.


3.1.3. GPS Tracking

Figure 3: A dog fitted with a GPS collar. GPS biotelemetry has become so successful that it has even been commercialized for everyday use.

Since the launch of the Global Positioning System (GPS) in 1993, it has been widely used and adopted in wildlife tracking due to its widespread availability and ease of access. Compared to ARGOS, GPS technology utilizes over 24 satellites to provide location data to users. If a GPS receiver picks up signals from three satellites (using triangulation), the receiver can be located in two dimensions (latitude and longitude). If it instead picks up signals from four satellites, the receiver can be located in three dimensions. GPS data can be stored and/or received in multiple ways. First, the data can simply be stored on the receiver over time until the animal carrying the tag is recaptured and the data is retrieved. Second, the data can be wirelessly downloaded off of the tag to another receiver. Finally, the GPS data can be relayed over the ARGOS satellite system, which allows the data to be globally accessed. The use of GPS has revolutionized biotelemetry and made it much easier, allowing researchers to obtain data anywhere in the world with excellent accuracy. Furthermore, researchers can obtain truly continuous data without having to wait for a satellite to be overhead. However, GPS systems can still be quite expensive, especially when combined with the usage of satellite transmission. GPS tracking devices also consume lots of power, due to the high power demands of locating and determining a GPS satellite fix. 3.1.4. Future Developments in Tracking Devices Much of the development in tracking devices is currently focused on optimization of current devices, which include battery optimizations and size reductions. As power consumption becomes more optimized, the inclusion of sensors that will allow researchers to not only track location but also study environmental conditions will increase which will allow the simultaneous study of not only wildlife behavior, but also wildlife biology and ecology. However, there are specific developments in animal tracking devices that could be useful. Nanotechnology Researchers have explored using nanometer-sized fluorescent probes called quantum dots in order to study microorganisms including zooplankton and phytoplankton [6]. Controlled experiments showed that the implantation of these devices did

112 Undergraduate Research at the Swanson School of Engineering

not affect the overall behavior of the affected animal species and may allow researchers to study microorganism response to light, food, and predation, all which could not be studied before in such microorganism animal species. Transceivers Currently, receiving biotelemetric data requires the offloading of data onto land-based centers, which requires some sort of physical interaction with tags or receivers. However, with the use of transceivers, or devices that act as both transmitters and receivers, researchers plan to have tags communicate with each other or to satellites in order to increase the ability to both transmit and receive data. 3.2. Development of Low-Cost Wildlife Tracking Devices While wildlife tracking has become more accessible due to GPS and satellite systems, these systems are still extremely expensive and make it hard to track many animals at once. Furthermore, in developing countries, it can be near impossible to acquire devices or justify the cost of a device when it can easily be lost or broken by animals in the wild. For this reason, it has been the focus of researchers to develop low-cost tracking devices using existing technology. Now that open-source hardware and software have become so widespread, it has become increasingly easy for researchers to create their own devices. In multiple occasions, researchers have taken existing devices that contain a GPS unit and a transmission system and modified them to create a low-cost tracking device [2, 7, 8]. A focus on many of these devices is the use of cellular communications, which is now widely accessible and much cheaper than transmitting data over the ARGOS satellite system. Devices can also be easily modified to the need of the researcher and can include sensors that allow for tracking of further data than just location, including speed, temperature, altitude, and battery status. Finally, specific software is not required to access data, and data is often published on a simple Internet-connected web browser or application. Compared to current tracking devices, which can cost upwards of thousands of dollars, most of these low-cost devices can be built or modified for under US$500. Furthermore, the use of existing accessible GPS devices meant that accuracy or resolution did not have to be sacrificed in the creation of these low-cost devices. The main disadvantage of these devices are the time required to build each device and the lack of commercial optimization in devices that may require additional maintenance or adjustments compared to the typical tracking device.

4. Conclusions

Biotelemetry is a developing field that is important in the study of wildlife behavior and movement. Since the formal start of biotelemetry in 1963 with the use of VHF devices, the development of the ARGOS system and today, GPS devices have allowed biotelemetry to become more accessible and widespread in this field. While these devices are extremely expensive, there is widespread movement to make devices cheaper. With continued research into the field and into future technologies, it will be

Ingenium 2020

incredibly exciting to see costs plummet in the construction of these devices to allow for the widespread study of animal species. Now, more than ever, the future of animal tracking is that of the simple hobbyist who can buy a GPS unit and a microcontroller to build a device that is just as reliable and accurate as a commercial system. The use of these modular, customizable systems will not only allow researchers to track more than just location, but also learn more about the behaviors of animals which have yet to be explored to this day. With the further development of biotelemetric devices, the still unknown world of wildlife biology will become more familiar, even to the microorganism scale.

5. Acknowledgements

Thank you to Dr. Gartner, the Swanson School of Engineering, and the Office of the Provost that provided the original funding for the project that led to this article. Thanks also to the Pittsburgh Zoo & PPG Aquarium for the support throughout this project as well and for the opportunity to dive into this exciting world of wildlife research.

6. References

[1] B. Thomas, J.D. Holland, E.O. Minot, Wildlife tracking technology options and cost considerations, Wildlife Research. 38 (2011) 653-663. [2] C.A. Zucco, G. Mourao, Low-Cost Global Positioning System Harness for Pampas Deer, J. of Wildlife Management. 73 (2009) 452-457. [3] W.W. Cochran, R.D. Lord, A Radio-Tracking System for Wild Animals, J. of Wildlife Management. 27 (1963) 9-24. [4] A. Markham, On a Wildlife Tracking and Telemetry System: A Wireless Network Approach. 2008. [5] R. Farve, Demonstration of Satellite/GPS Telemetry for Monitoring Fine-Scale Movements of Lesser Prairie-Chickens, United States Department of Agriculture. 2002. Accessed 24 Oct 2019. https://www.fs.fed.us/t-d/programs/im/satellite_gps_ telemetry/wildlifetrackingtelementry.htm [6] M. Lard, J. Backman, M. Yakovleva, B. Danielsson, L. Hansson, Tracking the Small with the Smallest – Using Nanotechnology in Tracking Zooplankton, PLoS ONE. (2010). [7] M. Fischer, K. Parkins, K. Maizels, D.R. Sutherland, B.M. Allan, G. Coulson, J.D. Stefano, Biotelemetry marches on: A costeffective GPS device for monitoring terrestrial wildlife, PLoS ONE. (2018). [8] B.M. Allan, J.P.Y. Arnould, J.K. Martin, E.G. Ritchie, A costeffective and informative method of GPS tracking wildlife, Wildlife Research. 40 (2013) 345-348.


Feasibility study of kinetic, thermoelectric and RF energy harvesting powered sensor system Keting Zhaoa, Jiangyin Huanga, Hongye Xua and mentor Dr. Jingtong Hua Department of Electrical and Computer Engineering, University of Pittsburgh, Pittsburgh, PA a

Team: Keting Zhao, Hongye Xu, Jiangyin Huang (left to right)

Keting Zhao is an electrical engineering student concentrating in communication and signal processing with a minor in computer science. She is interested in pursuing further training in signal processing and robotics controls. Hongye Xu is a computer engineering student concentrating in digital system and computational modeling. He is interested in studying further about neural network and machine learning. Jiangyin Huang is an electrical engineering student concentrating in communication and signal processing. He wants to learn deeper into computer vision and signal analyzing during his graduate level education.

Dr. Jingtong Hu

Dr. Jingtong Hu is currently an Assistant Professor in the Department of Electrical and Computer Engineering. His main research interests include embedded systems, FPGA, and cyber-physical systems. He is a recipient of Employer Diversity Recognition Award from Pitt Career Center.

Significance Statement

Using the combination of kinetic, thermoelectric and RF energy harvesting to power a communication system has not yet been tested. This feasibility study of this system will further indicate the feasibility of implementing the energy harvesting communication system into devices or equipment which involve in human daily activities for long-term usage.

Category: Device Design

Keywords: Energy Harvesting, Software Duty Cycle, Power ORing Achitecture Abbreviation: microcontroller unit (MCU), radio frequency energy harvester (RFEH), kinetic energy harvester (KEH), thermoelectric energy harvester (THE), piezoelectric energy harvester (PEEH), electromagnetic energy harvester(EMEH), low power mode (LPM)

114 Undergraduate Research at the Swanson School of Engineering


Batteryless and wireless energy harvesting systems are critical to internet of things (IoT) vision as well as the sustainability of long-lived, untethered systems. The systems can be implemented into wearable devices, which are used for monitoring vital body signals. They can also help the process of making insectwearable devices that monitor changes in the environment. In general, this kind of system can be divided into three levels: energy harvesting subsystems (EH), an embedded microcontroller unit (MCU), and peripherals (sensors, radios, etc.) [1]. This research will provide a feasibility study of integrating all the levels mentioned above. This paper will focus on the radio frequency energy harvester (RFEH), kinetic energy harvester (KEH) and thermoelectric energy harvester (THE) as well as an instruction on designing the rest of the system to be compatible with all three EHs and able to handle frequent power shortages during the process. The reason for choosing KEH, TEH and RFEH as power sources is because these energy are closely related to the human activities. Thus, this feasibility study could contribute towards further studies on energy harvesting communication system within wearable devices. The main result of this study indicates a positive and promising future of integrating multiple energy harvesters into a wearable communication system.

1. Introduction

The IoT serves to relate mechanical and digital machines through a large interconnected system, which, in some cases, must collect data on even the minutest details. This is costly in both the number of devices required and the power to run them. It stands to reason that in order to progress the field something must be introduced to combat this limiting factor. To that end, the feasibility of designing these devices to harvest ambient energy instead of being supplied from a separate device comes to mind. There are a number of methods to perform this such as TEH, KEH and RFEH, but the uncertainty comes from whether it is feasible to rely on such methods. It would only be unfeasible if these combined systems cannot generate enough power. Radio frequency energy harvesting is a technique that harvests energy from the electromagnetic field in the air and converts it into the electrical domain, voltage and current. TEH generates voltage when differing temperatures are placed side by side. For the energy harvester, this study will follow the insight that is provided by Leonov’s research on TEH of human body heat [2] and Gorlatova’s research on KEH of human activities [3] as well as Lu’s survey research paper on RFEH [4]. For integrating the sensor, we will take reference from the research Mementos: System Support for LongRunning Computation on RFID-Scale Devices [5]. This study will show whether energy harvesting methods can sustain the needs of IoT devices.

Ingenium 2020

2. Methods 2.1 Equipment Setup 2.1.1 PowerCast RFEH A RFEH is made of three parts, an antenna, a rectifying circuit and a matching circuit to give a DC output [4]. The RF energy harvester module used in this research is the model TX91501 PowerCast transmitter, which can broadcast unlicensed 915MHz ISM band radio wave and can produce 3 watts of EIRP and data up to 40-50 feet by using a matched receiver. The module P1110B receiver can rectify a 915MHz center frequency at a maximum of 23 dBm input power giving 4.3V/100mA as its maximum output. The beam pattern of TX91501 transmitter is 60 by 60 feet. Therefore, a patch directional antenna with 122 horizontal and 68 vertical energy pattern and 6.1 dBi gain is used to capture power from the transmitter. In order to boost the power, the DC-DC booster module BQ25570 is connected to the RFEH. 2.1.2 EMEH and PEEH For PEEH, device PPA-1011 is used in this research. However, there is no developed EMEH for commercial use. Therefore, the EMEH used in this project is built based on Kwon’s research [6]. This EMEH is mainly structured as repulsively stacked multilayer magnets with washer in between each one and placed in the middle of an independent coils which stands on compression springs for the purpose of oscillation as shown in Figure 1. With this setup, the magnetic flux density alternates multiple times depends on the number of stacked magnets during oscillation. This design can give a relative higher oscillating frequency of magnetic flux which can help produce AC power more efficiently [6]. To use these in conjunction, AC/DC converters are needed in the system since they output different forms of electricity. After the initial testing, a zener diode regulator is therefore implemented. Two zener diodes are connected back to back and with voltage output limited by Vz.

2.1.3 TEH with Different Scales There are multiple available TEHs in the market with similar designs, only different in sizes. The efficiency of all products are highly correlated to their sizes. Based on the data sheet, the power generated by TEH is stable but relatively small compared to the other two kind of EHs. Thus, power management module LTC-3108, which is designed for boosting ultra-low voltage input to 3.3V is used with TEHs. The output of the TEH is connected to LTC-3108, and four 1 F capacitors are connected in parallel to the output of LTC-3108 as storage. 2.2 Multi-inputs EHs Design Figure 2 shows a Power ORing architecture, which is a modular design that allows multiple EH sources to be connected in parallel through diodes [7]. A supercapacitor is used to smooth the raw output voltage from EHs. The modified final system schematics are based on this architecture with the testing result as seen in Figure 3. Each EH module is connected to the switch so that the user can choose which sources or sources to use. The diode connected to the RFEH has a cut off voltage at 0.7V. The MCU used for this research is a model MSP430G2553. A magnetometer sensor, a radio communication module, and a UART to USB converter, are connected to the MCU. The idea is to put MCU in stasis when the EHs are charging the supercapacitor. Once the supercapacitor is charged up to the operation voltage, 3.3V, the MCU will switch to its active mode, reading and transmitting data repeatedly until the supply power is below threshold. It will then revert to sleep mode until the next time the supercapacitor is fully charged. A software duty cycle is implemented to realize this idea.

Figure 2: Power ORing architecture

Figure 1: EMEH cut section indication and self-made EMEH final product


Figure 3: Schematic of the System Design

2.3 Software Duty Cycle Control An interruptible computation is integrated to complete the duty cycle control. The source code of the interrupt is from course CSE466 at University of Washington [8]. In the MSP430 architecture, there are several types of interrupts: timer, port and ADC interrupts. In this case, the timer and port interrupts are configured and used. MSP430G2553 has four different low power modes (LPM), and LPM0 is used. During LPM0, the CPU and the MCLK are disabled, while the ACLK and SMCLK remain active. Additionally, MSP430 uses a watchdog timer to reset itself after a certain amount of time in order to avoid a counting overflow. In order to avoid an unexpected reset during an interrupt, once the CPU receives a digital 0 from pin P1.4, the watchdog timer will be put on hold from counting and an interrupt will set on pending. Then pin P1.4 selects the high-to-low edge and clears the interrupt pending flag as the interrupt begins. ISR saves the state and redirects the stack pointer to interrupt functionality. After an interrupt happens, the interrupt service routine continues sensing the input from pin P1.4 and compares it against digital 1. When the input voltage is under its minimum operation voltage requirement, the input from P1.4 will indicate as digital 0, and vice versa. Now if input indicates the digital 0, the system remains in stasis; however, if input indicates a digital 1, the MCU exits LPM0 and resumes the main program’s next state. The clock will begin running from where it had been halted, and the MCU will execute the instructions from main program code again, reading data from sensors repeatedly, until the next interrupt is triggered. Additionally, a voltage divider is implanted so that when the input source voltage reaches 3.3V, the voltage across pin P1.4 will be 2.4V. In order to minimize the loss on input current, resistors with large resistance value are used to build the voltage divider. The resistors with value of 0.9kΩ and 2.4 kΩ are connected into the circuit as shown in Figure 4.

116 Undergraduate Research at the Swanson School of Engineering

3. Results 3.1 RFEH Modify with DC-DC Booster The testing result of output voltage and current with different distances between the RF transmitter and RF receiver is shown in Figure 4. The result shows that the maximum output power of RF energy harvester is 1.037mW and rapidly decreases as the distance from the source increases. When the distance between the transmitter and the receiver is greater than 150 cm, the output power is stable at average of 0.04mW. The minimum requirement for MCU to execute the program is 3.3V and 190µA, which the testing result clearly does not meet.

Figure 4: Raw outputs of RFEH vs Distance

After interageting the DC-DC booster with the RFEH, the retesting result shows that the output voltage from the receiving end is stable at 4.1V upon 487cm from the source. The relationship between output current and distance is shown in Figure 6. As distance increases to 7 feet from the source, the current drops down to 0.159 mA, which is lower than the threshold of powering up the rest of the system. Therefore, the RFEH now powers up the system around 200cm away. As shown in Figure 5, the modified

Ingenium 2020

RFEH is connected to the diode of cutoff voltage 0.7V. Thus, the maximum voltage across the supercapacitor when using RFEH alone will be 3.5V, which is in its safe range.

Figure 5: RFEH Output Current vs Distance with Booster

3.2 PEEH Modify with Zener Diode Regulator and EMEH Modify with Diodes The average raw output power from PPA-1011 is RMS voltage at 14V and AC current at 100A. However, PPA-1011 does overshoot unexpectedly and has peak voltage as high as 40V. Based on datasheet of LTC3588, the maximum input of this device is 18V. Thus, the raw voltage of PPA-1011 can damage the power management device. Zener diodes with Vz at 17V are then used to build the regulator and connected to PEEH, so that the input voltage of LTC3588 will not overshoot. As a result, the output average RMS voltage of PEEH is 8.25V which is in a safe range. AC current is negligible and maintains at 100µA. When the vibration frequency is in the range of 1.5Hz to 3.1Hz, the EMEH can produce output RMS voltage from 3 to 7 V and AC current from 5 to 7mA. After the raw EMEH power goes through the bridge rectifier with the capacitor, the average DC output voltage is 5.5V and the average DC current is 2.05mA. The operating voltage range of the MCU is from 1.8V to 3.6V. Thus, the output voltage of EMEH needs to be regulate before connects to the MCU. In order to reduce the output voltage of EMEH, three diodes which have cut off voltages at 0.7V are connected in series between EMEH and the super capacitor. EMEH produce the DC voltage at average of 3.4V and no significant changes on output current after connecting diodes into the design. 3.3 Three TEHs Testing Result Comparison Three different TEHs are all placed on arms, metals under sun, in room and in wind in order to test and compare the efficiency of harvesting. According to the result in Table 1, it can conclude that the efficiency of a TEH is higher when either the fluid speed around it is faster or the temperature difference between two sides is larger. Module EHA-PA1A is integrated into the system since it generates considerable power with only the coin size.

Module EHA-PA1A can stably generate 25mV with 3.5mA when placed on arms according to the experiment. To simulate the charging process, 25mV with 3mA is connected to the input of the LTC3108. After 2 hours charging, the voltage at the input of the MCU stabilizes at 2.4V and the input of the GPIOs stabilizes at 3.3V. DS18B20, 4.5mW thermal sensor, is connected to MCU, used as a test peripheral. With stable 25mV input, the sensor works for half an hour. 3.4 Voltage Divider Implement The datasheet indicates that the operation voltage range of MSP430 is from 1.8V to 3.6V. However, after a few tests, it shows that if the input voltage is lower than 2.3V, the input from pin P1.4 will indicate as digital 0, vice versa. This is higher than the expected minimum voltage, 1.8V. After the voltage divider has been introduced, the system can functional properly when RFEH keeps generating power. Other two kind of EHs could support the system after RFEH initializes the system properly. Once EHs stops supplying power, the supercapacitor starts discharging and the MCU could still be active as long as 45 seconds before switching to LPM. The MCU could stay in LPM for another 90 seconds before completely shutting down. If RFEH or other EHs start generating power again in this time period, MCU will be able switch back to active mode without being reset. The final design demonstration is shown in Figure 6.

Figure 6: Multi-inputs EHs powered sensor system demonstration




Metal under sun (stable)




Human body room temp (first touch)




Human body room temp (stable)




85ºC metal with room temp




Voltage under wind comparing to no wind

130% ↑


Not significant

Size (mm)




Table 1: Three TEGs’ efficiency under different circumstance


4. Discussion

The result shows a success to power the MCU with sensor module by RFEH indecently. TEH and KEH, however, needs the support from RFEH to power the rest of system initially. There are still few issues still remain with this design. RFEH is easily affected by kinetic motions along the energy collection path. When people pass by, the voltage will rise for a short amount of time. It is also easily affected by cellphones and other devices that generate radio waves. Adding the dc-dc booster, these issues can be eliminated when the distance from receiver to transmitter is at most 200cm, but exist otherwise. TEH takes approximately 2 hours to charge up four 1 F capacitors to 3.3V when the system is placed on human body. Even with smaller and fewer capacitors, due to the low input voltage, it still takes a long time to charge. The solution taken in this experiment is to charge the capacitors to 3.3V first and use the TEH to stabilize the voltage. Since the voltage generated would change the temperature difference, the hot side of the TEH (the side on skin) gets colder as time goes on. Thus the TEH is very inefficient, considering the amount energy it absorbs. The current design of EMEH is relatively big for wearable device. Due to the limitation of available material and handcraft technique, only two third of the wire around the coil is used while the rest of the wire is disconnected, but adds extra weight to the EMEH. The frequency of vibrating and number of turns of wire do have effect on the amount of power KEHs can generate. A shaker and coils with different number of turns are used to perform further analyzing in later testing which will help optimizing the volume of EMEH. The software duty cycle control only functions when the MCU is not completely shut down, which means the power shortage cannot be longer than 8.3 minutes. In this research, four EHs are in use, so power shortage is controlled under this threshold. In some case, this can be challenge and the possible way to solve this issue is to implement a duty cycle with logic ICs in the hardware design. In addition, the digital read for MCU I/O pin shows an inconsistent when super-capacitor is charging and when it is discharging. This makes the determination of when MCU is supposed to enter Low Power Mode vague and unstable. The power loss of the voltage divider is also not necessary. In order to eliminate this problem, a software ADC interrupt or a hardware duty cycle are going to be added into the system.

118 Undergraduate Research at the Swanson School of Engineering

5. Conclusion

The result shows it is feasible to integrate all three levels of the system together with commercial products on a relatively large scale. TEHs and KEHs needs further research and adjusting on method of support sufficient power independently. RFEH will need customized the antenna and matching circuit in order to capture the energy from Wi-Fi rather than the special radio frequency generator. A hardware duty cycle should be tested as well as making PCB based on this design to scale down the system. The radio module is still in design and has not yet been calibrated. Once it is connected into the system, the power consumption will increase. A super-capacitor with a larger capacitance might be needed in the modified the design.

6. Acknowledgement

Funding for this project was provided by the Swanson School of Engineering and the Office of the Provost. Special thanks to Dr. Hu and his PhD student, Yawen Wu, for providing the source of devices and the help we needed on this project.

7. Reference

[1] J. Hester and J. Sorber, “Flicker,” Proceedings of the 15th ACM Conference on Embedded Network Sensor Systems - SenSys 17, 2017. [2] 1. V. Leonov, “Thermoelectric Energy Harvesting of Human Body Heat for Wearable Sensors,” in IEEE Sensors Journal, vol. 13, no. 6, pp. 2284-2291, June 2013. doi: 10.1109/ JSEN.2013.2252526 [3] 2. M. Gorlatova, J. Sarik, G. Grebla, M. Cong, I. Kymissis and G. Zussman, “Movers and Shakers: Kinetic Energy Harvesting for the Internet of Things,” in IEEE Journal on Selected Areas in Communications, vol. 33, no. 8, pp. 1624-1639, Aug. 2015. doi: 10.1109/JSAC.2015.2391690 [4] Lu X., Wireless Networks with RF Energy Harvesting: A Contemporary Survey, vol. 17,2015 [5] B. Ransford, J. Sorber, and K. Fu, “Mementos,” Proceedings of the sixteenth international conference on Architectural support for programming languages and operating systems - ASPLOS 11, 2011. [6] Soon-Duck K., “Electromagnetic energy harvester with repulsively stacked multilayer magnets for low frequency vibrations”, March 2013. [7] Estrada-López JJ, Multiple Input Energy Harvesting Systems for Autonomous IoT End-Nodes. 2018 [8] Hemingway B. (2015, Fall), Lecture 4: MSP430 Interrupt.

Ingenium 2020

Index Category: Computational research

Category: Methods

Eli Brock and Sabrina Nguyen Evaluating carbon reduction strategies for the University of Pittsburgh.......................... 11

Michael Clancy Extensions and analysis of a virtual balancing task for studying sensory-motor control....................................................................................... 24

Asher J. Hancock Numerically resolved radiation view factors within thermoelectric generators via hybridized CPU-GPU computing....................................................................................... 38 Liam Martin Changes to the maternal sacrum and coccyx during and after pregnancy and delivery...................................................................................................................... 63 Michael Ullman Analytical model validation for melting probe performance using applied computational fluid dynamics.......................................................................................... 97

Category: Device design Ananya Mukherjee The role of oxygen functional groups in graphene oxide modified glassy carbon electrodes for the electrochemical sensing of nicotinamide adenine dinucleotide hydride...................................................................... 72 Jerry Potts Wireless signal transmission through hermetic walls in nuclear reactors....................... 81 Keting Zhao, Hongye Xu, Jiangyin Huang Feasibility study of kinetic, thermoelectric, and RF enery harvesting powered sensor system................................................................................................................ 114

Yannis Rigas Genetically engineering ocular probiotics to manipulate ocular immunity and disease........................................................................................... 85 Pierangeli RodrĂ­guez De Vecchis Effects of printing parameters on density and mechanical properties of binder jet printed WC-Co.................................................................................................. 88 Hannah Schmidt Monitoring the in-vitro extracellular matrix remodeling of tissue engineered vascular grafts.................................................................................................................. 92

Category: Review Samantha Bunke Progress in bioplastics: PLA and PHA............................................................................... 16 Nikolas J. Vostal Crimped polymer microfibers produced via electrospinning: A review.......................... 102 Kevin Xu Biotelemetry: a brief history and future developments in lowering cost........................ 110

Category: Experimental research Rosh Bharthi Tumor derived exosomes regulate dendritic cell maturation and activation...................... 7 Claire P. Chouinard Three-dimensional nickel foam and graphene electrode in microbial fuel cell application: Study of biofilm compatibility.......................................................... 20 J. Sebastian Correa Feature validation and online visualization of forearm high-density EMG in an individual with spinal cord injury................................................................................. 30 Lauren Grice Tractography reveals patterns of hippocampal innervation in the human temporal lobe................................................................................................. 34 Thomas J. Henry Characterization of redox flow battery kinetics using a flow channel analytical platform............................................................................................................ 42 Catherine Grace P. Hobayan Metformin administration impairs tendon wound healing............................................... 47 Patrick Iyasele Mechanical characterization of silk derived vascular grafts for human arterial implantation............................................................................................. 52 Eileen Li Robust osteogenesis of mesenchymal stem cells in 3D bioactive hydrogel.................... 57 Zixie Liang Manufacturing a polyelectrolyte coating on contact lenses using automated vs. manual techniques for the treatment of dry eye disease......................... 60 Angela J. McComb Laser-induced nanocarbon formation for tuning surface properties of commercial polymers....................................................................................................... 68 Tyler Paplham Characterization of hierarchical structures in remelted Ni-Mn-Ga substrates for directed energy deposition manufacturing of single crystals.......................................... 77 Kaitlyn Wintruba Adventitial extracellular matrix from aneurysmal aorta fails to promote pericyte contractility....................................................................................................... 106


Swanson School of Engineering 151 Benedum Hall 3700 O’Hara Street Pittsburgh, PA 15261 412-624-9800 engineering.pitt.edu

The University of Pittsburgh is an affirmative action, equal opportunity institution. Published in cooperation with the Office of University Communications. 112627-0320

Ingenium: Undergraduate Research at the Swanson School of Engineering, 2020

Spine art

Articles inside

A Message from the Associate Dean for Research article cover image
A Message from the Associate Dean for Research
page 6
A Message from the Co-Editors-in-Chief article cover image
A Message from the Co-Editors-in-Chief
page 7
Graduate Student Review Board – Ingenium 2020 article cover image
Graduate Student Review Board – Ingenium 2020
page 8
Tumor derived exosomes regulate dendritic cell maturation and activation article cover image
Tumor derived exosomes regulate dendritic cell maturation and activation
pages 9-12
Evaluating carbon reduction strategies for the University of Pittsburgh article cover image
Evaluating carbon reduction strategies for the University of Pittsburgh
pages 13-17
Progress in bioplastics: PLA and PHA article cover image
Progress in bioplastics: PLA and PHA
pages 18-21
Michael Clancy, Sudarshan Sekhar, Aaron Batista, Patrick Loughlin article cover image
Michael Clancy, Sudarshan Sekhar, Aaron Batista, Patrick Loughlin
pages 26-31
with spinal cord injury article cover image
with spinal cord injury
pages 32-35
Lauren Grice, Chandler Fountain, Michel Modo article cover image
Lauren Grice, Chandler Fountain, Michel Modo
pages 36-39
Metformin administration impairs tendon wound healing article cover image
Metformin administration impairs tendon wound healing
pages 49-53
Mechanical characterization of silk derived vascular grafts for human arterial implantation article cover image
Mechanical characterization of silk derived vascular grafts for human arterial implantation
pages 54-58
Robust osteogenesis of mesenchymal stem cells in 3D bioactive hydrogel article cover image
Robust osteogenesis of mesenchymal stem cells in 3D bioactive hydrogel
pages 59-61
techniques for the treatment of dry eye disease article cover image
techniques for the treatment of dry eye disease
pages 62-64
Liam Martin, Megan R. Routzong, Ghazaleh Rostaminia, Pamela A. Moalli, Steven D. Abramowitch article cover image
Liam Martin, Megan R. Routzong, Ghazaleh Rostaminia, Pamela A. Moalli, Steven D. Abramowitch
pages 65-69
Laser-induced nanocarbon formation for tuning surface properties of commercial polymers article cover image
Laser-induced nanocarbon formation for tuning surface properties of commercial polymers
pages 70-73
The role of oxygen functional groups in graphene oxide modified glassy carbon article cover image
The role of oxygen functional groups in graphene oxide modified glassy carbon
pages 74-78
Characterization of hierarchical structures in remelted Ni-Mn-Ga substrates for directed energy deposition manufacturing of single crystals article cover image
Characterization of hierarchical structures in remelted Ni-Mn-Ga substrates for directed energy deposition manufacturing of single crystals
pages 79-82
Wireless signal transmission through hermetic walls in nuclear reactors article cover image
Wireless signal transmission through hermetic walls in nuclear reactors
pages 83-86
Genetically engineering ocular probiotics to manipulate ocular immunity and disease article cover image
Genetically engineering ocular probiotics to manipulate ocular immunity and disease
pages 87-89
WC-Co article cover image
pages 90-93
Monitoring the in-vitro extracellular matrix remodeling of tissue engineered vascular grafts article cover image
Monitoring the in-vitro extracellular matrix remodeling of tissue engineered vascular grafts
pages 94-98
fluid dynamics article cover image
fluid dynamics
pages 99-103
Crimped polymer microfibers produced via electrospinning: A review article cover image
Crimped polymer microfibers produced via electrospinning: A review
pages 104-107
Adventitial extracellular matrix from aneurysmal aorta fails to promote pericyte contractility article cover image
Adventitial extracellular matrix from aneurysmal aorta fails to promote pericyte contractility
pages 108-111
Biotelemetry: a brief history and future developments in lowering cost article cover image
Biotelemetry: a brief history and future developments in lowering cost
pages 112-115
Feasibility study of kinetic, thermoelectric, and RF enery harvesting powered sensor system article cover image
Feasibility study of kinetic, thermoelectric, and RF enery harvesting powered sensor system
pages 116-120
Index article cover image
pages 121-125