MULTISCALE MODELS IN BIOELECTROCHEMICAL ENGINEERING

By

Christina M. Wark

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

Chemical Engineering - Doctor of Philosophy

2025

ABSTRACT

Future technological progress will increasingly rely on biomaterials and bioinspired mech-

anisms. We can draw inspiration from nature to meet industrial needs because biological

materials and mechanisms are often highly efficient and superior to other synthesized mate-

rials. Overall, the work herein strives to advance understanding of bio-inspired materials for

a versatile set of engineering challenges and solutions.

First, we study the highly efficient reaction cascade of the tricarboxylic acid (TCA), or

Krebs, cycle, which converts nutrients in food to usable energy in fractions of seconds. Of

particular interest in the TCA cycle is the conversion of oxaloacetate (OAA) between malate

dehydrogenase (MDH) and citrate synthase (CS). Based on previous experimental and com-

putational studies of recombinant and mutant MDH-CS complexes, a time-dependent finite

difference model was developed to predict each complex’s transfer efficiency and determine

the surface’s reaction pathways. Utilizing the kinetic parameters of recombinant and mu-

tant complexes determined experimentally and surface transition probabilities of OAA from

a Markov state model, the lag time of MDH-CS was determined computationally for recom-

binant and mutant complexes. Additional implications of the reaction path and reversible

reaction at MDH are also considered. This model study furthers the understanding of dy-

namic enzymatic cascades and points toward approaches to cascade design.

Second, we address the upgrading of bio-oil from pyrolyzed lignocellulosic biomass. Bio-

oils yield a complex mixture of organic constituents that can be used as feedstock for valuable

chemicals. However, these molecules are often oxygenated with low energy density. Recently,

electrocatalytic hydrotreatment (ECH) of bio-oil was used to reduce oxygen content and in-

crease energy density. Mechanistic understanding of interaction effects in mixtures is required

for effective process design. Here, the mechanisms by which ECH of bio-oil constituent 4-

propylphenol (4-PP) is inhibited by furfural (FF) are studied computationally. Inhibition is

elucidated through potential dependent studies of adsorption and reaction mechanisms on

a platinum/ruthenium electrocatalyst. Thermodynamic studies suggest that the FF path-

ways are competitive in adsorption and more favorable in reactions due to fewer barriers

and smaller limiting potentials than the 4-PP pathways. Prediction of activation barriers by

reaction energy scaling techniques found the FF pathways also to be favored kinetically over

the 4-PP pathways. Reaction thermodynamics and kinetics models suggest that FF inhibits

4-PP hydrotreatment because the FF pathways are more favorable than 4-PP pathways on

a platinum/ruthenium catalyst.

Finally, a new biomaterial is evaluated as a high performance substrate in tissue regen-

eration applications. A biomaterial was previously developed where gelatin was modified by

methacrylic anhydride (GelMA) for stability at physiological conditions and silver-bioactive

glass (Ag-BG) was added for antibacterial properties. Chemically linked GelMA and Ag-BG

(GAB) was hypothesized to exhibit superior structural and cell viability behavior to GelMA

in extrusion printed applications, which we seek to validate. Scaffold printing parameters

were optimized for acellular GelMA, GAB, and a GelMA + Ag-BG composite bio-ink. Then,

cell-laden media was introduced to the bio-ink to generate scaffolds. The viability of the cells

within the scaffold was observed over time after printing. Ultimately, bio-ink performance

and cell viability were poor for the selection of materials during cellular printing. These

challenges suggest further optimization steps are needed within the material synthesis and

printing processes, which is a point of more recent work that shows promising results.

Copyright by
CHRISTINA M. WARK
2025

To my parents, Joe, Ayla, and Lily...

v

ACKNOWLEDGMENTS

First and foremost, I would like to express my deepest gratitude to my advisor, Dr. Scott

Calabrese Barton, for his unwavering support and encouragement to persist throughout the

years. The person and researcher that I am today is undoubtably due to his guidance and

mentorship. I was always challenged to ask ”why?” which has made a lasting impression

that will help me to push the edge of science throughout the rest of my career and life. I am

a better person and scientist from my time here. I would also like to extend my appreciation

and gratitude to my committee members: Dr. James (Ned) Jackson, Dr. Jose Mendoza

Cortes, Dr. Michael Murillo, and Dr. Chris Saffron. Though my research trajectory made

some substantial shifts throughout my time at MSU, each of my committee members have

been figures of guidance and expertise on a versatile set of theoretical and physical concepts.

They have all made a lasting impact on how I think about and execute my pursuit of science.

Additionally, I want to thank my current and past groupmates, and collaborators. To

Brandon, Yunlu, Yan, Kanchan, Manali, Alex, Mindy, Pranavi, and Meiers: I am so thankful

for all of our chats, research related and otherwise. I was lucky to learn alongside so many

wonderful people. I was also fortunate to collaborate with many incredible people throughout

my research work. Thank you to the MTRAC and beyond team members: Dr. Xanthippi

Chatzistavrou, Dr. Adam Marsh, Dr. Nuredin Ashammakhi, Dr. Chris Contag, Dr. Victoria

Toomajian, and Anthony Tundo, all of whom have taught me so many things about materials

engineering and how to probe the interface between design and application. Further thanks

to Dr. Per Askeland, Emily Pawlowski, Amy Albin, and Dr. Dan Holmes who helped

me in access to equipment necessary to the work and analysis of the results. I also want

to acknowledge Johannes Schavikin and Romila Manchanda who I’ve been able to continue

learning through as they pursue further development of these biomaterials. I am also grateful

vi

to the other members of the NSF-DFG team: Dr. Meheryar Kasad, Dr. Uwe Schr¨oder, Dr.

Keyvan Malaie, and Jannelle Casanova. It has been an absolute pleasure to work alongside

all of you. Finally, I want to thank the entire Schr¨oder group at the University of Greifswald

for their kindness, friendship, and all of the things I learned during my visit.

I am so thankful for the friends that have been around to support me for a long time

and the new friends that I made along the way. To Sarah, who has been my buddy in all

things academic and otherwise since our time as undergrads through working on our PhDs

together. This was easier knowing I was never alone. To Kyra and Allie, two of the friends

I made along the way, thank you for being such wonderful pillars of emotional support and

reminders to take care of my wellbeing when the going got rough. To the rest of my core

friend groups: Tim, Bre, Brad, Lacie, Chris, Sean, Ryan, Andrew, Greg, Dan, Robert, Ben,

Jeremy, and the Mega Group, thank you all for standing with me for all of these years. Even

at the lowest points, time with you all made things brighter. Thank you for always reminding

me of who I am and what I’m working toward, but also being so fun when I needed a break.

I am deeply grateful to my family: Joe, mom, dad, Michelle, and my in-laws-to-be: Tim,

Marsha, Ben, and Maggie. They all have been nothing but supportive and encouraging

throughout my life. There are not enough words to express how thankful I am for my soon-to-

be husband, Joe, for all that he has done to support me through this process, and constantly

encouraging me to be my best and keep pushing toward my dreams. This accomplishment

was made possible because he stood with me through all of the highs and lows by making sure

everything else was taken care of so that I could focus during the busy times, tagging along

when I had to check on my syntheses at odd hours, and celebrating every accomplishment

along the way. My parents have worked so hard to give me the opportunities to continue

learning and pursue science. I wouldn’t have made it this far without their support. To my

vii

sister, Michelle, thank you for indulging me as my assistant in all the times I wanted to play

”Science Lab” as kids. She helped me find my passion in science. Thank you to the family

I gained when I met Joe. I am so lucky to have been welcomed in to their family from the

start. They all have been such a strong source of support, encouragement, and love through

this process and in life. Lastly, to my cats, Ayla and Lily, who sat with me through all of

the early mornings, late nights, and Zoom lectures and meetings. Their emotional support

helped me through a lot of the challenges of grad school and their presence made it easier

to talk out my code logic when debugging.

With many thanks and a lot of caffeine,

Christina

viii

TABLE OF CONTENTS

Chapter 1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Enzyme-inspired Efficient Reaction Cascades . . . . . . . . . . . . . . . . . .
1.2 Multi-scale Modeling of Electrochemical Conversion of Biomass
. . . . . . .
1.3 Manufactured Cell-Laden Tissue Scaffolds
. . . . . . . . . . . . . . . . . . .
1.4 Overview of Included Works . . . . . . . . . . . . . . . . . . . . . . . . . . .

Chapter 2

Finite Difference Model of Electrostatic Channeling in TCA
Cycle Enzymes . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Chapter 3

Quantum Chemical Insight into Electrochemical Upgrading
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
of Bio-oil
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1
1
16
35
39

43
43
47
54
64

65
65
67
78
90

Kinetic Evaluation of Electrochemical Upgrading of Bio-oil 93
Chapter 4
93
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1
95
4.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 Results and Discussions
99
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

Chapter 5

Toward Validation of a New Bioink for 3D Printed Cell-
Laden Tissue Scaffolds . . . . . . . . . . . . . . . . . . . . . . . 107
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.1
5.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

Chapter 6

Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.1 Enzyme-inspired Efficient Reaction Cascades . . . . . . . . . . . . . . . . . . 123
6.2 Multiscale Modeling of Electrochemical Conversion of Biomass . . . . . . . . 125
6.3 Manufactured Cell-Laden Tissue Scaffolds
. . . . . . . . . . . . . . . . . . . 127
6.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

ix

APPENDIX A SUPPLEMENTARY INFORMATION FOR CHAPTER 2:
FINITE DIFFERENCE MODEL OF ELECTROSTATIC
CHANNELING IN TCA CYCLE ENZYMES . . . . . . . . 148

APPENDIX B SUPPLEMENTARY INFORMATION FOR CHAPTER 3:

QUANTUM CHEMICAL INSIGHT INTO ELECTROCHEM-
ICAL UPGRADING OF BIO-OIL . . . . . . . . . . . . . . . 154

x

Chapter 1

Introduction

Technical industries constantly pursue more efficient solutions to engineering problems. Sev-

eral sectors of industry seek sustainable materials due to climate change and diminishing
resources.1 Recently, a trend emerged in engineering toward sustainability aspects that of-

ten rely on efficient processes and clean materials. One aspect focuses on understanding
the efficiency of biological processes. 2,3 Another aspect investigates use of renewable and
alternative resources to generate supplies that meet current demands. 4 A third application
lies in the design of bio-compatible materials in medical technologies.5 Versatile solutions

can be explored through exploiting mechanisms and functions of bio-derived and bio-inspired

materials. The ability to apply bio-inspired materials to meet diverse needs relies on ad-

vancements in computational modeling, material performance optimization, and an improved

fundamental understanding of mechanisms. The research summarized herein aims to ad-

vance understanding of bio-inspired materials for a versatile set of engineering challenges

and efficient, sustainable solutions.

1.1 Enzyme-inspired Efficient Reaction Cascades

Complex living organisms utilize highly efficient metabolic pathways to execute several vi-
tal functions required for survival. 6,7 Biological cycles and naturally occurring mechanisms

efficiently convert chemicals into energy sources and other materials needed to complete
the metabolon, or metabolic cycle.6 The desire to generate novel chemicals (i.e. biofuels)
at high yields has placed focus on understanding biochemical mechanisms.3,7 A cascade of
reactions is required to mimic the various steps in these metabolon.6,7 However, the inclu-

sion of biocatalytic surfaces in the form of enzymes adds complexity to reaction systems.

1

Figure 1.1: Enzyme cascades in (a) ethanol production from biomass(adapted from Ref 89),
(b) methanol generation from CO2 from Ref 10), (c) DNA processing for cancer detection
from Ref 11), and (d) biofuel cells (adapted from Ref 12).

This complexity can be overcome with computational methods that can reveal phenomena
influencing the yield of the cascade. 3,8

Enzymatic cascades are broadly implemented in various applications for efficient chem-

ical processes. Some examples of these applications are summarized in Figure 1.1. In the

commodity chemical sector, enzymatic cascades are under development to convert biomass

and carbon dioxide into platform chemicals. Specific examples include the production of
ethanol from biomass-derived pyruvate (Fig. 1.1a) 9 and methanol from carbon dioxide (Fig.
1.1b)10. Pyruvate is formed via a glycolysis pathway on α-amylase and glucoamylase. 9 The

pyruvate is first converted to acetaldehyde on pyruvate decarboxylase. Then, ethanol is

formed by the reduction of acetaldehyde by alcohol dehydrogenase. This pathway is ob-

served in yeasts utilized for fermentation. The conversion of CO2 to methanol is similarly

inspired by nature through a photo-electrochemical process consisting of a three-enzyme cas-
cade designed to mimic photosynthesis.10 Here, carbon dioxide is first reduced to formate

on formate dehydrogenase (FDH), formaldehyde is produced from formate on formaldehyde

dehydrogenase(FaldDH), and finally, methanol is formed on alcohol dehydrogenase (ADH).

Enzymatic cascades are also utilized in technological applications, including medical di-

2

+‐+→+++‐Δ‐‐‐+3oxoglutaratearginine3O2ethenesuccinateguanidinepyrroline5carboxylate7COLL212Thisethene-formingenzymehasbeenexpressedinvariousotherorganismssuchasE.coli,43Saccharomycescerevisiae,44andPseudomonasputida.45However,thethreeaforementionedpathwayswillleadtoyieldsofetheneonglucosebelow0.12g/g,44whichisunattractiveforlargescaleethenesynthesis.Enzymaticdecarboxylationofacrylicacidtoethenehasbeensuggestedtooccurinsomeorganismsduetoasideactivityofpyruvatedecarboxylase,46buttheproofforthisisconsideredtobeweak.47Currently,thekeyoptiontoapproachatheoreticalyieldofetheneonglucoseof2mol/mol(0.31g/g)isbyusingacid-catalyzeddehydrationoffermentation-derivedethanoltoethene.Thisiscommerciallyappliedasbiobasedroute.38Itwouldbeattractiveifenzymaticdehydrationwouldbepossible,butsuchactivityisonlydescribedinpatentapplications.48,493.2.PropenePropene(propylene)isoneofthemostimportantchemicals.In2002,about53milliontwasobtainedfrompetrochemicalprocessesforconversionintopolymersandnumerousderivatives.50Biologicalformationoftracesofpropenehasbeenobservedinaerobicculturesofmanydifferenttypesoforganisms,inparticularRhizopusstrains,butthesourcewasnoteluci-dated.51,52Inotherstudies,rabbitcytochromeP-450-catalyzedformationofpropenefromisobutyraldehydehasbeendemonstrated,withanactivityof98mmol/minofproductpermolofP-450.53Itisassumedthatbymono-oxygenaseaction,O2isactivatedusingNADPHasanelectronsource.Oneoxygenatomisthenreducedtowater,andtheotheratomistransferredtotheisobutyraldehyde,leadingtopropeneandformicacid.54Usingsuchareactionforbiobasedpropeneproductiondoesnotseemtobepursued,althoughtherequiredisobutyraldehydecanbeproducedfromglucose(seesection7.4).Itwillbeeasiertousearouteinvolvingacid-catalyzeddehydrationofisopropanol(seesection5.1.4)intopropene.55Besides,patentapplicationsclaimthatthisdehydrationcanbeenzyme-catalyzed.48,49CommercialbiobasedpropeneproductionispursuedbyBraskem,usingfermentativeethanolproduction,followedbychemicalconversionintoethene,dimerization,andmeta-thesis.563.3.IsobuteneCurrently,morethan10milliont/aofisobutene(2-methylpropene)isproducedviapetrochemicalroutes.There-fore,productionfrombiomassviabiocatalyticpathwaysisofconsiderableinterest.57Biologicalisobuteneformationhasbeenknownsincethe1970s.Thehighestproductionrate,merely0.45mg/(Lh),wasfoundforastrainoftheyeastRhodotorulaminuta.58Isobutenewasformedbyreductivedecarboxylationofisovalerate,whichisproducedinthecatabolicpathwayofL-leucine.59,60ThereactionseemstobeasidereactionofacytochromeP450monooxygenasethatisinvolvedinhydroxylatingbenzoate.61Besidesthelowrateofisobuteneformation,theusedpathwayhasnoobviouspotentialtoapproachatheoreticalstoichiom-etryof1isobutene+2CO2+2H2Opermolofglucose.Tworecentdevelopmentsmayenablemuchhigheryields,althoughtheydonotdescribereasonableisobuteneproductionlevelsyet.Ithasbeenshownthat3-hydroxyisovaleratecanbeconvertedtoisobuteneasaside-activityofmevalonatediphosphatedecarboxylase(Figure4).62,63Also,patentapplicationsdescribeisobutanoldehydrationasasideactivityofengineeredoleatehydrataseandotherhydratases.48,49Fermentativeproductionofisobutanolisdescribedinsection5.1.7.Ifaone-potbiochemicalconversionofcarbohydratetoisobutenewouldbecomepossible,isobutenegaswouldbeemittedfromafermentor,andisobutanolrecoveryanditschemocatalyticdehydrationcouldbebypassed.573.4.Butadiene1,3-Butadieneisaveryimportantmonomerforsyntheticrubbersbutalsoforsomeplastics.Besides,itisanintermediateintheproductionofsomeimportantchemicals.Productionandofbutadienein2009wasabout9.2milliont.64Severalmetabolicpathwaysto1,3-butadienehavebeenclaimedinapatentapplicationofGenomatica,65butwithoutFigure4.Pathwaysfrompyruvatetoarangeofcompounds,indicatedinboldface.ChemicalReviewsReviewdx.doi.org/10.1021/cr400309c|Chem.Rev.2014,114,1871−19081876+‐+→+++‐Δ‐‐‐+3oxoglutaratearginine3O2ethenesuccinateguanidinepyrroline5carboxylate7COLL212Thisethene-formingenzymehasbeenexpressedinvariousotherorganismssuchasE.coli,43Saccharomycescerevisiae,44andPseudomonasputida.45However,thethreeaforementionedpathwayswillleadtoyieldsofetheneonglucosebelow0.12g/g,44whichisunattractiveforlargescaleethenesynthesis.Enzymaticdecarboxylationofacrylicacidtoethenehasbeensuggestedtooccurinsomeorganismsduetoasideactivityofpyruvatedecarboxylase,46buttheproofforthisisconsideredtobeweak.47Currently,thekeyoptiontoapproachatheoreticalyieldofetheneonglucoseof2mol/mol(0.31g/g)isbyusingacid-catalyzeddehydrationoffermentation-derivedethanoltoethene.Thisiscommerciallyappliedasbiobasedroute.38Itwouldbeattractiveifenzymaticdehydrationwouldbepossible,butsuchactivityisonlydescribedinpatentapplications.48,493.2.PropenePropene(propylene)isoneofthemostimportantchemicals.In2002,about53milliontwasobtainedfrompetrochemicalprocessesforconversionintopolymersandnumerousderivatives.50Biologicalformationoftracesofpropenehasbeenobservedinaerobicculturesofmanydifferenttypesoforganisms,inparticularRhizopusstrains,butthesourcewasnoteluci-dated.51,52Inotherstudies,rabbitcytochromeP-450-catalyzedformationofpropenefromisobutyraldehydehasbeendemonstrated,withanactivityof98mmol/minofproductpermolofP-450.53Itisassumedthatbymono-oxygenaseaction,O2isactivatedusingNADPHasanelectronsource.Oneoxygenatomisthenreducedtowater,andtheotheratomistransferredtotheisobutyraldehyde,leadingtopropeneandformicacid.54Usingsuchareactionforbiobasedpropeneproductiondoesnotseemtobepursued,althoughtherequiredisobutyraldehydecanbeproducedfromglucose(seesection7.4).Itwillbeeasiertousearouteinvolvingacid-catalyzeddehydrationofisopropanol(seesection5.1.4)intopropene.55Besides,patentapplicationsclaimthatthisdehydrationcanbeenzyme-catalyzed.48,49CommercialbiobasedpropeneproductionispursuedbyBraskem,usingfermentativeethanolproduction,followedbychemicalconversionintoethene,dimerization,andmeta-thesis.563.3.IsobuteneCurrently,morethan10milliont/aofisobutene(2-methylpropene)isproducedviapetrochemicalroutes.There-fore,productionfrombiomassviabiocatalyticpathwaysisofconsiderableinterest.57Biologicalisobuteneformationhasbeenknownsincethe1970s.Thehighestproductionrate,merely0.45mg/(Lh),wasfoundforastrainoftheyeastRhodotorulaminuta.58Isobutenewasformedbyreductivedecarboxylationofisovalerate,whichisproducedinthecatabolicpathwayofL-leucine.59,60ThereactionseemstobeasidereactionofacytochromeP450monooxygenasethatisinvolvedinhydroxylatingbenzoate.61Besidesthelowrateofisobuteneformation,theusedpathwayhasnoobviouspotentialtoapproachatheoreticalstoichiom-etryof1isobutene+2CO2+2H2Opermolofglucose.Tworecentdevelopmentsmayenablemuchhigheryields,althoughtheydonotdescribereasonableisobuteneproductionlevelsyet.Ithasbeenshownthat3-hydroxyisovaleratecanbeconvertedtoisobuteneasaside-activityofmevalonatediphosphatedecarboxylase(Figure4).62,63Also,patentapplicationsdescribeisobutanoldehydrationasasideactivityofengineeredoleatehydrataseandotherhydratases.48,49Fermentativeproductionofisobutanolisdescribedinsection5.1.7.Ifaone-potbiochemicalconversionofcarbohydratetoisobutenewouldbecomepossible,isobutenegaswouldbeemittedfromafermentor,andisobutanolrecoveryanditschemocatalyticdehydrationcouldbebypassed.573.4.Butadiene1,3-Butadieneisaveryimportantmonomerforsyntheticrubbersbutalsoforsomeplastics.Besides,itisanintermediateintheproductionofsomeimportantchemicals.Productionandofbutadienein2009wasabout9.2milliont.64Severalmetabolicpathwaysto1,3-butadienehavebeenclaimedinapatentapplicationofGenomatica,65butwithoutFigure4.Pathwaysfrompyruvatetoarangeofcompounds,indicatedinboldface.ChemicalReviewsReviewdx.doi.org/10.1021/cr400309c|Chem.Rev.2014,114,1871−19081876+‐+→+++‐Δ‐‐‐+3oxoglutaratearginine3O2ethenesuccinateguanidinepyrroline5carboxylate7COLL212Thisethene-formingenzymehasbeenexpressedinvariousotherorganismssuchasE.coli,43Saccharomycescerevisiae,44andPseudomonasputida.45However,thethreeaforementionedpathwayswillleadtoyieldsofetheneonglucosebelow0.12g/g,44whichisunattractiveforlargescaleethenesynthesis.Enzymaticdecarboxylationofacrylicacidtoethenehasbeensuggestedtooccurinsomeorganismsduetoasideactivityofpyruvatedecarboxylase,46buttheproofforthisisconsideredtobeweak.47Currently,thekeyoptiontoapproachatheoreticalyieldofetheneonglucoseof2mol/mol(0.31g/g)isbyusingacid-catalyzeddehydrationoffermentation-derivedethanoltoethene.Thisiscommerciallyappliedasbiobasedroute.38Itwouldbeattractiveifenzymaticdehydrationwouldbepossible,butsuchactivityisonlydescribedinpatentapplications.48,493.2.PropenePropene(propylene)isoneofthemostimportantchemicals.In2002,about53milliontwasobtainedfrompetrochemicalprocessesforconversionintopolymersandnumerousderivatives.50Biologicalformationoftracesofpropenehasbeenobservedinaerobicculturesofmanydifferenttypesoforganisms,inparticularRhizopusstrains,butthesourcewasnoteluci-dated.51,52Inotherstudies,rabbitcytochromeP-450-catalyzedformationofpropenefromisobutyraldehydehasbeendemonstrated,withanactivityof98mmol/minofproductpermolofP-450.53Itisassumedthatbymono-oxygenaseaction,O2isactivatedusingNADPHasanelectronsource.Oneoxygenatomisthenreducedtowater,andtheotheratomistransferredtotheisobutyraldehyde,leadingtopropeneandformicacid.54Usingsuchareactionforbiobasedpropeneproductiondoesnotseemtobepursued,althoughtherequiredisobutyraldehydecanbeproducedfromglucose(seesection7.4).Itwillbeeasiertousearouteinvolvingacid-catalyzeddehydrationofisopropanol(seesection5.1.4)intopropene.55Besides,patentapplicationsclaimthatthisdehydrationcanbeenzyme-catalyzed.48,49CommercialbiobasedpropeneproductionispursuedbyBraskem,usingfermentativeethanolproduction,followedbychemicalconversionintoethene,dimerization,andmeta-thesis.563.3.IsobuteneCurrently,morethan10milliont/aofisobutene(2-methylpropene)isproducedviapetrochemicalroutes.There-fore,productionfrombiomassviabiocatalyticpathwaysisofconsiderableinterest.57Biologicalisobuteneformationhasbeenknownsincethe1970s.Thehighestproductionrate,merely0.45mg/(Lh),wasfoundforastrainoftheyeastRhodotorulaminuta.58Isobutenewasformedbyreductivedecarboxylationofisovalerate,whichisproducedinthecatabolicpathwayofL-leucine.59,60ThereactionseemstobeasidereactionofacytochromeP450monooxygenasethatisinvolvedinhydroxylatingbenzoate.61Besidesthelowrateofisobuteneformation,theusedpathwayhasnoobviouspotentialtoapproachatheoreticalstoichiom-etryof1isobutene+2CO2+2H2Opermolofglucose.Tworecentdevelopmentsmayenablemuchhigheryields,althoughtheydonotdescribereasonableisobuteneproductionlevelsyet.Ithasbeenshownthat3-hydroxyisovaleratecanbeconvertedtoisobuteneasaside-activityofmevalonatediphosphatedecarboxylase(Figure4).62,63Also,patentapplicationsdescribeisobutanoldehydrationasasideactivityofengineeredoleatehydrataseandotherhydratases.48,49Fermentativeproductionofisobutanolisdescribedinsection5.1.7.Ifaone-potbiochemicalconversionofcarbohydratetoisobutenewouldbecomepossible,isobutenegaswouldbeemittedfromafermentor,andisobutanolrecoveryanditschemocatalyticdehydrationcouldbebypassed.573.4.Butadiene1,3-Butadieneisaveryimportantmonomerforsyntheticrubbersbutalsoforsomeplastics.Besides,itisanintermediateintheproductionofsomeimportantchemicals.Productionandofbutadienein2009wasabout9.2milliont.64Severalmetabolicpathwaysto1,3-butadienehavebeenclaimedinapatentapplicationofGenomatica,65butwithoutFigure4.Pathwaysfrompyruvatetoarangeofcompounds,indicatedinboldface.ChemicalReviewsReviewdx.doi.org/10.1021/cr400309c|Chem.Rev.2014,114,1871−19081876Theextensiveliteratureonthistopicwillalsonotbetreatedhere.Oneofthechallengesassociatedwithutilizationofbiomass-derivedsugars,particularlythoseproducedbychemicalhydrolysis,istheformationofdegradationproductsthatmaybeinhibitoryduringfermentation.23Generally,microbialcellsarelessproductivewhenconsumingcrudemixturesofhydrolyzedmonosaccharidesratherthanglucose.Withfewexceptions,thisreviewwillonlygivedataforglucose,becauseglucosecanbeusedasbenchmarksubstrateformanymetabolicroutes.Themainpathwayforglucoseconversionistheglycolysis,andFigure1showshowthismergeswithpathwaysforconversionofsomeotherbiomassmonomers.Table2includesallbiomasscomponentsthatarehighlyrelevantforrenewablechemicalsynthesis,exceptlignin,becauseligninisnotconsideredtobeenzymaticallyhydro-lyzable.Ligninbiosynthesisoccursbyradicalpolymerizationofguaiacyl,syringyl,andp-hydroxyphenylunitsfromtheprecursorsconiferyl,sinapyl,andp-coumarylalcohol,respec-tively,allcontainingphenylpropenoidbuildingblocks.19Theobtainednetworkofaromaticgroupscanbedegradedbysomefungiandbacteriausingenzymessuchasperoxidasesandlaccase.InthepresenceofO2,theseenzymesgenerateoxidativeradicals,leadingtosubstitutedaromaticcompoundsthatcanbefurthermetabolized.19InthepresenceofH2O2,lipaseshavebeenusedtoformperoxycarboxylicacidsinsituthatcauselignindegradation.24Thelignincatabolismoftheproteobacte-riumSphingobiumsp.SYK6wasrecentlyproposedtoincludeaglutathionedependentcleavageofaβ-O-4aryletherlinkagebondbyanetherase.25Withtheaidofenzymes,somecomponentsusefulasfinechemicalshavebeenobtainedfromlignin,andaquesthasbeguntoidentifymetabolicroutesfromlignintocommoditychemicals.26Chemicaldegradationoftheligninfractionofbiomassismorestraightforward,andthedegradationproductsmaybefurtherconvertedusingbiochemicalmethods.Inthiscontext,gasificationoflignin(andotherbiomass)tosyngas,suchasmentionedinTable1,isparticularlyrelevant.Thisbiosyngas,amixtureofCOandH2,likesyngasoffossilorigin,canbeconvertedintocommoditychemicalsnotonlybyusingchemicalmethodsbutalsobysomemicroorganisms.27Amongdifferentsyngasconvertingpathways,theWood−Ljungdahlpathwaycanbeusedtoproduceacetyl-CoAfromsyngas.28,29Inshort,oneCOisreducedtoamethylgroupthatislinkedtoanotherCO.Subsequently,allkindsofmetabolicroutescanleadfromacetyl-CoAtocommoditychemicals.ThecompaniesLanzaTech,Coskata,andINEOShaveproducedethanolbyfermentationofsyngasonpilotscaleandaremovingtowardcommercialization.29Syngasfermentationperformancedoesnotyetseemtobeatthelevelofcarbohydratefermentationperformance,butithastheadvantagethatitalsousesthenoncarbohydrate(lignin)portionoflignocellu-losicbiomass.2.ALKANES2.1.MethaneMethanefromfossilsourcesisanimportantrawmaterialfortheindustrialproductionofacetylene,synthesisgas,H2,methanol,andmanyothercompounds.Biologicalmethaneproductionisalsowell-known,andhasbeenappliedsince1881.30Methanecanbeproducedfromaceticacidbymethanogenicbacteria(Figure2).Thisoccursspontaneouslyduringbiomethanation,atypeofanaerobicdigestionofbiomass.Mixedculturesofmicroorganismsfirsthydrolyzecomplexorganicmoleculesintosimplesugars,aminoacids,andfattyacids.Inthesubsequentacidogenesisstage,theseareconvertedintocarboxylicacidssuchasacetic,propionic,butyric,andvalericacid,alongwithammonia,carbondioxide,andhydrogensulfide.Inthethirdstage,acetogenesis,productsfromacidogenesisarefurtherdigestedtoproducecarbondioxide,hydrogen,andorganicacids,mainlyaceticacid.Finally,aceticacidisconvertedintomethaneandcarbondioxideduringmethanogenesis.Thisfinalconversioninvolvesseveralintermediateswithenzyme-catalyzedmethyltransferstocoenzymesMandB.31Themaximumyieldofmethaneperglucoseequivalentis0.27g/g,iftherestbecomesCO2.Figure1.Metabolicpathwaysofsomemajorbiomassmonomersuptopyruvate.Glycolysisisthedirectroutefromglucosetopyruvate.FormationandconsumptionofADP,ATP,phosphate,andwaterhavenotbeenindicated.Electrons(e−)aretakenupbyNAD+orNADP+.Componentsinboldfacearetreatedexplicitlyinthetext.Figure2.MetabolicpathwaystoH2andsomeC1compounds.ChemicalReviewsReviewdx.doi.org/10.1021/cr400309c|Chem.Rev.2014,114,1871−19081874GermanEdition:DOI:10.1002/ange.201611379BiocatalyticPhotosynthesisHotPaperInternationalEdition:DOI:10.1002/anie.201611379PhotoelectrochemicalReductionofCarbonDioxidetoMethanolthroughaHighlyEfficientEnzymeCascadeSuKeunKuk+,RaushanKSingh+,DongHeonNam,RanjithaSingh,Jung-KulLee,*andChanBeumPark*Abstract:NaturalphotosynthesisisaneffectiverouteforthecleanandsustainableconversionofCO2intohigh-energychemicals.Inspiredbythenaturalprocess,atandemphoto-electrochemical(PEC)cellwithanintegratedenzyme-cascade(TPIEC)systemwasdesigned,whichtransfersphotogeneratedelectronstoamultienzymecascadeforthebiocatalyzedreductionofCO2tomethanol.AhematitephotoanodeandabismuthferritephotocathodewereappliedtofabricatetheironoxidebasedtandemPECcellforvisible-light-assistedregenerationofthenicotinamidecofactor(NADH).ThecellutilizedwaterasanelectrondonorandspontaneouslyregeneratedNADH.TocompletetheTPIECsystem,asuperiorthree-dehydrogenasecascadesystemwasemployedinthecathodicpartofthePECcell.Underappliedbias,theTPIECsystemachievedahighmethanolconversionoutputof220mmhˇ1,1280mmolgˇ1hˇ1usingreadilyavailablesolarenergyandwater.Photosynthesisoccursingreenplantsthroughahighlyefficient,photoinducedenergy-transfercascadeconsistingoftheZ-schemeandtheCalvincycle.Thebeautyofthisnaturalprocesshasinspiredmanyresearcherstodevelopeco-friendly,solarreductionsofCO2tohydrocarbonfuels.[1]Methanol,ausefulreducedformofCO2,[2]canbesynthesizedfromCO2bytheuseofdifferenttypesofphoto-orelectro-chemicalcatalysts,suchasmetalcomplexesandoxide-basedsemiconductors.[3]Thesecatalysts,however,exhibitcriticalproblems:forexample,metalcomplexesoftensufferfromseverephotocorrosioninanaqueousmediumwhenillumi-natedbylight,[4]whilesemiconductorsgeneratelowyieldswithpoorselectivityandrequireharshconditionssuchashightemperaturesandpressuresforbetteroutput.Recently,redoxbiocatalystshavebeenutilizedforthephotochemicalreduc-tionofCO2.Thesebiocatalystshaveseveraladvantages,includingtheabilitytofunctionundermildconditionswithextremelyhighspecificitywithoutsidereactions,althoughtheysufferfromissuesofhighproductioncost,poorstabilityunderharshconditions,andO2sensitivity.[5]Inprinciple,acascadeinvolvingthreedifferentenzymes,namelyformatedehydrogenase(FDH),formaldehydedehydrogenase(FaldDH),andalcoholdehydrogenase(ADH),canreduceCO2toformate,formaldehyde,andmethanolinastepwisemanner.However,eachoftheenergeticallyuphillreductionsrequiresthesupplyofastoichiometricamountofexpensivenicotinamidecofactor(NADH)toachievecatalyticturnover,andthebiocatalyticphotochemicalsystemsreportedtodatefortheconversionofCO2intomethanolarelimitedbytheirlowyieldsofNADHregeneration.Here,wereportaninorganic/bioorganichybridphoto-syntheticplatformconsistingofanNADH-regeneratingphotoelectrochemical(PEC)cellandamultienzymecascade(Scheme1).ThetandemPECcellwithanintegratedenzyme-cascade(TPIEC)systemmimicsthenaturalphotosyntheticZ-schemebyacquiringphotogeneratedelectronsfromtheoxidationofwateronilluminationwithvisiblelightandtransferringthemtotheenzymecascadeforthebiocatalyticreductionofCO2tomethanol.Inthephotoelectrode-typeZ-schemesystem,theanodicandcathodicphotoelectrodesareseparatedintotwocompartmentsbyanion-exchangesaltbridgeandconnectedbyaconductingwiretopreventproductScheme1.Schematicillustrationofthesolar-assistedproductionofmethanolfromCO2andwaterthroughathree-enzyme(CcFDH,PcFaldDH,YADH)cascade.Underilluminationbyvisiblelight,waterisoxidizedatthephotoanode(Co-Pi/a-Fe2O3)andservesasanelectrondonor.Photoexcitedelectronsaretransferredtothephoto-cathode(BiFeO3),followedbyreductionofNAD+toNADHviaarhodium-basedmediator(M=[Cp*Rh(bpy)H2O]+,Cp*=C5Me5,bpy=2,2’-bipyridine).Ultimately,theexcitedelectronsfromreducedNAD+aredeliveredtothethree-enzymecascadeforthebiocatalyticsynthesisofmethanol.[*]S.K.Kuk,[+]Dr.D.H.Nam,Prof.Dr.C.B.ParkDepartmentofMaterialsScienceandEngineeringKoreaAdvancedInstituteofScienceandTechnology335ScienceRoad,Daejeon305-701(RepublicofKorea)E-mail:parkcb@kaist.ac.krDr.R.K.Singh,[+]R.Singh,Prof.Dr.J.K.LeeDepartmentofChemicalEngineering,KonkukUniversity120Neungdong-ro,Seoul143-701(RepublicofKorea)E-mail:jkrhee@konkuk.ac.kr[++]Theseauthorscontributedequallytothiswork.SupportinginformationandtheORCIDidentificationnumber(s)fortheauthor(s)ofthisarticlecanbefoundunder:http://dx.doi.org/10.1002/anie.201611379.AngewandteChemieCommunications3827Angew.Chem.Int.Ed.2017,56,3827–3832⌫2017Wiley-VCHVerlagGmbH&Co.KGaA,WeinheimEnzyme cascades in biofuel cells Macazo and Minteer 115 Figure 1 (Left) Schematic representation of reactions occurring at the methanol bioanode and air-breathing cathode. (Right) Representative power curves obtained for methanol/air biofuel cells employing three bioanodes that contained different cascades of dehydrogenase enzymes: a) NAD–ADH, b) NAD–ADH and NAD–AldDH, c) NAD–ADH, NAD–AldDH, and NAD–FDH, that were tested in pH 7.40 phosphate buffer with 100 mM methanol fuel solution. Reproduced with permission from Wiley from reference [32] . phosphate pathway. These natural enzyme cascades have been employed in biofuel cells. The first enzymatic cascade-based biofuel cell was reported by Palmore and Whitesides in 1998 [15 ●●] . This cascade oxidized methanol to carbon dioxide utilizing nicotinamide adenine dinucleotide (NAD)-dependent alcohol dehy- drogenase (ADH), aldehyde dehydrogenase (AldDH), and formate dehydrogenase (FDH). This is an exam- ple of a linear series cascade , where the fuel is oxidized through the first enzyme to form an intermediate that is oxidized by the second enzyme, followed by the third enzyme, as shown in Figure 1 . This cascade not only im- proved the current ( ∼2600 µA cm −2 ) and power density ( ∼680 µW cm −2 ) performance of the biofuel cell, but also the energy density, since the biofuel cell is harnessing all six electrons from the oxidation of methanol. This early research work, however, did not immobilize the enzymes on an electrode, thus, recent work has concentrated on im- mobilization of the cascade to increase the active lifetimes of the enzymes and further improve the performance [4,5,16] . Other linear series cascades include the oxidation of ethanol to acetate (current density: ∼3.0–5.4 µA cm −2 ; power density: ∼1160–2040 µW cm −2 ) [17] , sucrose to gluconolactone (current density: ∼140 µA cm −2 ; power density: ∼22.2 µW cm −2 ) [18] , the electron transport chain (current density: ∼10 µA cm −2 for ETC metabolon with oxidized cytochrome c and ∼6.1 µA cm −2 for com- plex IV with reduced cytochrome c) [19] , trehalose to glu- conolactone (current density: ∼365 µA cm −2 ; power den- sity: ∼285 µW cm −2 [20] ; current density: ∼120 µA cm −2 ; power density: ∼12 µW cm −2 [21] ) and glycolysis [22] . The most impressive coulombic efficiencies have come from recent work by the Zhang group on the pentose phosphate pathway (current density: 6000 µA cm −2 ; power density: ∼800 µW cm −2 ) [23 ●●,24] , which focused on producing biofuel cells for portable power. The work on trehalose to gluconolactone (current density: 460 µA cm −2 ; power density: ∼55 µW cm −2 ), however, is probably the most popular natural enzyme cascade in recent literature, which centered on implantable applications in insects for sensing and cyborgs [25 ●] . Not all metabolic pathways are linear. The citric acid cycle or Krebs cycle is a cyclic metabolic pathway with a multitude of different substrates along the cycle that can be oxidized, including: pyruvate (current density: 3920 µA cm −2 ; power density: ∼930 µW cm −2 ) [26] , lac- tate (current density: 2700–3320 µA cm −2 ; power den- sity: ∼800–827 µW cm −2 ) [27] , ethanol (current density: 3600 µA cm −2 ; power density: ∼1010 µW cm −2 ) [28 ●] , and malate. It is important to note, however, that a linear cas- cade may be employed first in order to generate the in- termediates that can enter the cyclic metabolic cascade. As shown in Figure 2 , these pathways allow the usage of a single pathway to oxidize a wide variety of intermedi- ates. However, most of the cyclic pathways have built- in metabolic control that requires the entire cycle to be present and functioning for the system to have high flux (current density) through the pathway. Therefore, these systems are more complex for preparation and function. There are also parallel cascades in natural metabolic path- ways. For instance, the three-enzyme cascade oxidation of sucrose uses invertase first to separate the disaccha- ride sucrose into two monosaccharides (glucose and fruc- tose) and then uses glucose oxidase and fructose dehy- drogenase to simultaneously oxidize the monosaccharides in parallel (current density: 344 µA cm −2 ; power density: www.sciencedirect.com Current Opinion in Electrochemistry 2017, 5 :114–120 Three-EnzymeCascadeBioreactorforRapidDigestionofGenomicDNAintoSingleNucleosidesJunfaYin,†TianXu,†,‡NingZhang,†andHailinWang*,†,‡†StateKeyLaboratoryofEnvironmentalChemistryandEcotoxicology,ResearchCenterforEco-EnvironmentalSciences,ChineseAcademyofSciences,Beijing100085,China‡UniversityofChineseAcademyofSciences,Beijing100049,China*SSupportingInformationABSTRACT:Structure-basedDNAmodificationanalysispro-videsaccurateandimportantinformationongenomicDNAchangesfromepigeneticmodificationstovariousDNAlesions.However,genomicDNAstrandsareoftenrequiredtobeefficientlydigestedintosinglenucleosides.Itisanarduoustaskbecauseoftheinvolvementofmultipleenzymeswithdifferentcatalyticacitivities.Hereweconstructedathree-enzymecascadecapillarymonolithicbioreactorthatconsistsofimmobilizeddeoxyribonucleaseI(DNaseI),snakevenomphosphodiesterase(SVP),andalkalinephosphatase(ALPase).Bytheuseofthiscascadecapillarybioreactor,genomicDNAcanbeefficientlydigestedintosinglenucleosideswithanincreasingrateof∼20folds.Theimprovementismainlyattributedtodramaticallyincreaseenzymaticcapacityandactivity.Withadesignedmacro-porousstructure,genomicDNAof5−30Kb(∼1.6−10millionDaltons)canbedirectlypassedthroughthebioreactorsimplybyhandpushingoralow-pressuremicroinjectionpump.Bycouplingwithliquidchromatography-tandemmassspectrometry(LC-MS/MS),wefurtherdevelopedasensitiveassayfordetectionofanoxidativestressbiomarker8-oxo-7,8-dihydro-2′-deoxyguanosine(8-oxodG)inDNA.Theproposedthree-enzymecascadebioreactorisalsopotentiallyapplicableforfastidentificationandquantitativedetectionofotherlesionsandmodificationsingenomicDNA.ExposuretochemicalcarcinogensmayresultinDNAdamages,akeysteptowardtheonsetofcancer.1−4DNAdamagescanoccurviastructuralmodificationofthenucleo-sidesandphosphatemoietiesinDNAchains.5ItisdifficulttodirectlysearchforvariousmodificationsingenomicDNAsinceDNAstrandsarelongbiopolymersconsistingofalargenumberofnormalnucleosides(deoxyadenosine,dA;deoxygunosine,dG;deoxycytidine,dC;andthymidinemonophosphates,dT)interspersingwithasmallquantityofmodifiednucleosides(e.g.,5-methylcytosine,5mC)ortracemodifications(e.g.,5-hydroxymethylcytosine,5-formalcytosine,and5-carboxylcyto-sine).TodeterminetheseDNAmodifications,genomicDNAsampleswereoftenrequiredtobeefficientlydigestedintomononucleotidesorsinglenucleosides.6−10Liquidchromatog-raphy-massspectrometry(LC-MS)canexplorewhethernucleotideismodified,andissensitiveenoughtoquantifythesemodificationsonDNAchains.9,11−15Forexample,duringLC-MS/MSanalysisofglobalDNAmethylationandhydroxylmethylation,genomicDNAwasdigestedintothemixtureofclassicalDNAnucleosides(dA,dT,dG,dC)andmodifiednucleosides.16−18Recently,wediscoveredanewDNAmodificationindrosophila,N6-methyladenine.19Toaccuratelyidentifyitsstructureandvalidateitsidentity,theDrosophilagenomicDNAwasalsorequiredtobedigestedintosinglenucleosidesforhighsensitivityanalysis.Anotherexample,identificationandanalysisofoxidativedamagedDNA(e.g.,8-oxo-7,8-dihydro-2′-deoxyguanosine,8-oxodG)completelydi-gestionofgenomicDNAtomaximalrelease8-oxodGwasnecessaryforaccurate8-oxodGquantification.20−25Analysisofnucleosidescanprovideincreasedsensitivity,accuracyandprecisionratherthandetectionofDNApolynucleotidesormononucleotides.Ofnote,DNAdigestionisverydifferentfromproteomicdigestion,inwhichtheproteinscanbequicklydigestedintosmallpeptidesandaminoacidsbyoneenzyme(e.g.,thepotentenzymetrypsin),whichcanthenbedirectlyanalyzedbyLC-MS.26−32Comparedwiththemostcommonlystudiedproteins,genomicDNAstrandsaremuchlargersupramacromolecules,whichpossessmolecularweightgenerallymorethan10millionDaltions.TocharacterizechemicalstructuralchangeofnucleotidesinDNAortoquantifythestucturallychangednucleoties,longDNAchainsareoftenrequiredtobedigestedintosinglenucleosidesfacilitatingfollowedMSanalysis.Nevertheless,thedigestionofgenomicDNAintosinglenucleosideswasrequiredatleastthreeenzymes,involvingdeoxyribonuclease(e.g.,DNaseP1orDNaseI),phosphodies-terase(e.g.,snakevenomphosphodiesterase,SVP),andalkalinephosphatase(ALPase).10,16−19,33−37DNaseIareappliedtoReceived:April28,2016Accepted:June24,2016Published:July14,2016Articlepubs.acs.org/ac©2016AmericanChemicalSociety7730DOI:10.1021/acs.analchem.6b01682Anal.Chem.2016,88,7730−7737Downloaded via MICHIGAN STATE UNIV on March 24, 2025 at 14:12:33 (UTC).See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.a.b.c.d.agnostics11 and biofuel cells12. In medical diagnostics, a three-enzyme cascade for DNA con-
version to single nucleosides exposes DNA damage due to the onset of cancer (Fig. 1.1c). 11

The DNA is first broken into smaller mono-nucleotide and oligonucleotide fragments by de-

oxyribonuclease I (DNase I). The resulting oligonucleotides are broken into mono-nucleotides

via snake venom phosphodiesterase (SVP). Finally, the mono-nucleotides are broken down

into single nucleosides by alkaline phosphatase(ALPase). Biofuel cells utilize enzyme cas-
cades to electrochemically oxidize fuel sources such as methanol (Fig. 1.1d). 12 A process

similar to the reverse of the CO2 to methanol conversion is used here. Methanol is first con-

verted to formaldehyde via ADH. Then formic acid is generated on aldehyde dehydrogenase

(AldDH). Finally, CO2 is generated from formic acid on FDH. These examples illustrate the

versatility of enzyme-catalyzed reaction cascades for chemical processes and in technology.

The processes can be inspired by natural phenomena or utilize the selectivity of enzymes to

develop new, efficient processes.

1.1.1 Enzymatic Reaction Mechanisms

In enzyme-catalyzed reactions, the mechanisms must account for the interaction of reactants,

or substrates, and intermediates with the enzyme’s active site and the reaction at the active

site. For single-substrate reactions, this is accomplished via a Michaelis-Menten rate law,

which captures the formation of an enzyme-substrate complex, E · S, as depicted below.

E + S −−→←−− E · S → E + P

(1.1)

Where the enzyme is denoted by “E”, a simple diagram of the Michaelis-Menten mechanism

is shown in Figure 1.2a. Also in this figure (Fig. 1.2a-b) are two forms of Bi-Bi mechanisms

for dual-substrate reactions. These mechanisms are discussed in more detail in Section

1.1.1.1.

For a simple Michaelis-Menten mechanism, the overall reaction rate or reaction velocity,

ν, can be defined as the rate of increase in the product concentration, CP. Reaction velocity

3

Figure 1.2: Schematic of enzyme reaction mechanisms: (a) standard Michaelis - Menten, (b)
Sequential Bi - Bi, and (c) Ping - Pong Bi - Bi mechanisms.

depends on the substrate concentration (CS) plus other terms for active site interactions.13–15

ν =

VmaxCS
KM + CS

(1.2)

Where Vmax represents the maximum rate of product generation when the enzyme is satu-

rated with substrate as a function of turnover frequency (kcat) and enzyme concentration(Cenz).

Vmax = kcatCenz

(1.3)

The Michaelis-Menten coefficient, KM, describes the adsorption equilibrium between the

substrate and the enzyme as a function of kcat. Further, at a product generation rate of

one-half of Vmax, the substrate concentration will be equivalent to KM.

KM =

kdes + kcat
kads

4

(1.4)

Figure 1.3: Example Michaelis-Menten Plot where kcat = 10 s-1, KM = 10−1 M, and
Cenz = 10−1 M.

The rate constant of substrate adsorption is kads and desorption is represented by kdes. An

example of a Michaelis-Menten plot is provided in Figure 1.3.

The rate law can be further simplified in the case of very large Michaelis-Menten constant

relative to the substrate concentration. Here, the generation of the product would be pseudo-

first order and take the form of:

ν =

Vmax
KM

CS

(1.5)

1.1.1.1 Bi-Bi Mechanisms

The rate law becomes more complex when multiple substrates are consumed in an enzymatic

reaction. This is often the case in naturally occurring enzymatic reactions like those within
the Krebs cycle. 16 The bi-bi mechanism describes a reaction in which two substrates (S1
and S2) are consumed to produce two products (P1 and P2).

S1 + S2

Enz
−−−→←−−− P1 + P2

(1.6)

Within a bi-bi reaction system, the reaction may proceed as either a sequential mechanism
or a ping-pong mechanism, as shown in Figs 1.2b-c. 17,18 In the sequential mechanism, the

binding of substrates and release of products are classified as ordered sequential, where

5

substrates bind and react in a specific order, or random sequential, where the substrates

bind at different times in either order before the dissociation of products. One cycle of

substrate consumption and product dissociation must occur before a new cycle occurs. The

same forward rate law describes both categories of sequential mechanisms, as the kinetics
are indistinguishable for the two categories of sequential mechanisms 17:

νf =

KM,S2

CS1

+ KM,S1

VmaxCS1
CS2

CS2
+ CS1

+ Ki,S1

KM,S2

(1.7)

CS2

This expression considers the Michaelis-Menten kinetics for each substrate and an inhibition

factor of S1, Ki,S1
, to describe the likelihood of inhibition of the second substrate reaction
by the first substrate. 13 Only one inhibition term is necessary because there is no inhibition

of whichever substrate that binds first. For a reversible reaction, the reverse rate law would

be:

νr =

KM,P2

CP1

VmaxCP1
CP2

CP2
+ CP1

+ KM,P1

+ Ki,P1

KM,P2

(1.8)

CP2

The ping-pong mechanism describes a case in which the first substrate reaction occurs

by chemically modifying the enzyme, and then the second reaction occurs, renewing the

enzyme. Since the second reaction is dependent on the first reaction, there is no inhibition,

and the rate law does not consider an inhibition term.

ν =

KM,S2

CS1

VmaxCS1
+ KM,S1

CS2
CS2

+ CS1

CS2

(1.9)

The system may be saturated with one of the two substrates to probe just one of the two
reactions within a bi-bi mechanism.16 In this case, the bi-bi rate law may be simplified to the

single-substrate Michaelis-Menten equation because the saturated substrate’s concentration

change is negligible within the context of the reaction environment. For example, if the

system is saturated with S2, the rate law in Eq. 1.8 can be written as:

ν =

Vm,appCS1
KM,app + CS1

(1.10)

For both simultaneous and ping-pong bi-bi mechanisms, the applicable maximum rate,

6

Vm,app, is the same for a fixed concentration of the second substrate, C0
S2

:

Vm,app =

VmaxC0
S2
+ C0
S2

KM,S2

(1.11)

The applicable Michaelis-Menten constant, KM,app, varies for the two subsets of bi-bi

mechanisms. In a simultaneous mechanism, the applicable constant is:

KM,app =

Ki,S1

KM,S2
KM,S2

+ KM,S1
+ C0
S2

C0
S2

In the ping-pong mechanism, applicable Michaelis-Menten constant is:

KM,app =

KM,S1
KM,S2

C0
S2
+ C0
S2

(1.12)

(1.13)

Similar to the standard expression of the Michaelis-Menten equation, and in the case

of a KM,app that is much larger than the concentration of S1, the rate expression may be

simplified to a pseudo-first order rate law:

ν =

Vm,app
KM,app

CS1

(1.14)

1.1.2 Reaction Cascades and Lag Time

A reaction cascade is a general classification for reaction systems in which a reactant is
converted to a stable intermediate, then a product without substantial side reactions. 3 These

“one-pot synthesis” processes can be highly efficient with minimal byproducts or competing
side reactions.3,19 The efficiency of such systems is evaluated by the system lag time, or the

time that the systems requires to achieve steady state. Steady state is defined by constant
bulk intermediate concentration and linearly increasing product generation.7 For an example

of lag time calculation, consider an arbitrary system in which intermediate generation follows

a zeroth order rate law and its consumption to form product follows a first order rate law.

7

Figure 1.4: Example product and intermediate concentration curves with lag time fit to
product curve (green dashed line) for k1 = 0.9 s-1 and k2 = 1 s-1.

The rates of intermediate (CI) and product (CP) generation are:

dCI
dt

= k1 − k2CI

dCP
dt

= k2CI

(1.15)

(1.16)

where k1 and k2 are the rate constants of the intermediate and product generation,

respectively, and k2 is larger than k1. The time dependent solutions to the two differential

equations are:

CI(t) =

k1
k2

(cid:16)

1 − e−k2t(cid:17)

CP(t) = k1t

(cid:16)

1 − e−k2t(cid:17)

(1.17)

(1.18)

The resulting concentration curves are plotted in Figure 1.4 where the lag time is approx-

imately 1 second. At steady state, the concentration of intermediate is constant (CI = CI,ss).

8

The lag time is proportional to the intermediate concentration.20

tℓ =

CI,ss
k1

(1.19)

From this relationship, one can deduce that lower steady-state bulk concentrations of in-

termediate will reduce the lag time. Since this is also indicative of a faster approach to

steady state, a lower lag time and lower bulk intermediate concentration result from more

efficient systems. Therefore, a well-designed catalyst with favorable intermediate transport

properties, thus reducing the intermediate lost to the bulk, is imperative for highly efficient

reaction cascade processes.

1.1.3 Transport Mechanisms in Enzymatic Reaction Cascades

The design of such catalysts is inspired by enzymatic cascades occurring in natural biochem-

ical processes. In particular, the highly efficient intermediate transport mechanisms from

the reactant consumption site to the product generation site are of interest. Two mecha-

nisms found in enzymatic reaction cascades are electrostatic channeling and tunneling as
exemplified in Figure 1.5a-b. 2,21–23 Electrostatic channeling utilizes an electrostatic inter-

action between the oppositely charged enzyme surface and a charged stable intermediate
to keep the intermediate very close to the surface, preventing losses to the bulk.20,21,24–27

Molecular tunneling occurs when the proteins of the enzyme complex form a tunnel between

intermediate generation and consumption sites to force the intermediate to the next step of
the reaction.22,23,28–31 The relationship between surface transport and surface charge on a

catalyst makes the electrostatic channeling mode especially intriguing for reaction cascade

engineering and will be the focus of the remaining discussion on this topic.

In the study of enzymatic systems that exhibit electrostatic channeling, three design cases

can be explored, as visualized in Figure 1.6. The first represents one of the extreme cases

in which intermediate does not channel on the surface, which we refer to as “free enzymes”

(Fig. 1.6a). In the case of free enzymes, the intermediate and product generation sites are

not physically connected. Thus, the intermediate can only reach the product generation site

by desorbing from the first site, traversing the bulk, then adsorbing on to the second site.

9

Figure 1.5: Transport mechanisms in enzyme cascades: (a) electrostatic channeling on malate
dehydrogenase - citrate synthase where the intermediate traverses the surface along the
highlighted line and (b) molecular tunneling in tryptophan synthase where intermediate is
tunneled along the blue line to the second active site. Figures are adapted from Ref2.

Figure 1.6: Electrostatic channeling design cases (a-c) and (d) product generation rates for
each case.

This case represents a lag time maximum for a given system, with lag times on the order
of 1 to 10 seconds.16,32 In Fig 1.6d, the free enzymes exhibit the lowest product generation

rates at early times.

The other cases represent a system where the intermediate generation and consumption

sites are connected by some surface on which the intermediate can be electrostatically chan-

neled. Such a surface may be formed by an enzyme complex or a peptide bridge connecting

the two enzymes. The second extreme case is where intermediate is perfectly channeled on

the surface, i.e.when no intermediate is released to the bulk but is completely consumed to

create product (Fig. 1.6b). Surface transport occurs very quickly at rates several orders of

10

magnitude faster than chemical reactions. 26,33,34 Because of the timescale difference between

transport and reactions, and lack of intermediate bulk concentration, the perfect channeling

case results in nearly zero lag time and constant product generation, as depicted in Fig.

1.6d.

The final case in Fig. 1.6c represents the more realistic middle ground of channeling

systems. In this mixed desorption case, a mixture of intermediate channeling and desorption

occurs. Lag time and product generation rate (Fig. 1.6d) in this case reside somewhere

between that of the free enzymes and perfect channeling. The resulting transfer efficiency of

the system describes how close the mixed desorption lag time is to the free enzyme lag time.

A more efficient system with a higher tendency to channel would have a lower lag time and

greater transfer efficiency.

1.1.4 Experimental Studies of Enzymatic Cascades

Enzymatic reaction cascades of recent interest that utilize electrostatic channeling include the
engineered cascade of Hexokinase - Glucose-6-Phosphate Dehydrogenase (HK-G6PDH)7,24,27,32
and naturally-occurring Malate Dehydrogenase - Citrate Synthase (MDH-CS).16,26,35 Some

recent works also suggest that mechanisms of electrostatic channeling exist within the metabolism
of E. Coli, but mechanistic studies for this system are not yet available.36,37 HK-G6PDH

is an example of a bridging system, as shown in Figure 1.7a. Details of the peptide bridge,

colored green in the Figure, such as length and protein have been engineered to maximize
transfer efficiency.7,24,27,32 MDH-CS, responsible for the conversion of malate to citrate via
an oxaloacetate intermediate in the Tricarboxylic acid (TCA) or Krebs cycle 38, is a com-

plexed enzyme system, as shown in Fig. 1.7b. Several variations of the MDH-CS complex
structure have been studied to determine the implications on transfer efficiency.16

In 2017, Liu et al. studied electrostatic channeling within the transformation of glucose to
6-phospho-gluconolactone via a glucose-6-phosphate (G6P) intermediate on HK-G6PDH.32

The bridge, where the channeling of negatively-charged G6P occurs, was anchored to cysteine

residues on each enzyme. In addition to the free enzyme systems, two peptide bridges were

probed experimentally: the first with 5 positively-charged lysine residues and the second

with 5 neutral glycine residues. The glycine bridged system exhibited a comparable lag time

11

Figure 1.7: Electrostatic channeling enzyme systems of interest include (a) HK-G6PDH
(adapted from Ref 32) and (b) MDH-CS (adapted from Ref16).

to the free enzymes at around 100 s. The lysine bridge afforded a shorter lag time of 70 s,

exemplifying the advantage of engineered electrostatic channeling. This affect is observed

across a range of enzyme concentrations with lower, and more similar, lag times at higher

enzyme concentrations. Additionally, the lag time of the charged lysine bridge decreased with

lower ionic strength, more so than the neutral glycine bridge. This suggests that electrostatic

channeling of a negatively charged intermediate on a positively charged bridge is particularly

efficient under low enzyme concentration and low ionic strength conditions.

The conversion of malate to citrate via an oxaloacetate (OAA) intermediate on MDH-CS
was experimentally studied by Bulutoglu et al. in 2016.16 In this work, complexed MDH-

CS was found to exhibit electrostatic channeling. The efficiency of a complex, and the

impact of channeling, varied for a variety of different MDH-CS complexes. The variants

included native bovine, commercial wild-type porcine, recombinant porcine, and a mutation

of the recombinant. Implications of electrostatic channeling were determined via transient

analysis of product concentration, resulting in lag times for each variant. Native bovine

and recombinant porcine complexes were the most efficient, with lag times for both almost

ten fold lower than the commercial. The mutant exhibited the longest lag time, nearly 30

fold that of the recombinant. In addition to lag time analysis, Bulutoglu et al.’s work also

12

parameterized the kinetics at MDH and CS for the recombinant and mutant complexes.

This study indicates that OAA transport and overall system efficiency is reliant on effective

surface transport modulated by consistent surface charge.

1.1.5 Computational Studies of Enzymatic Cascades

Various computational methods have been developed to address and model complex reac-

tion systems. Molecular and atomistic level simulations have been employed to advance the

understanding of mechanisms that occur in complex chemical reaction environments, par-
ticularly catalytic systems.8,27,39 Transport of an intermediate on a complex or bridge has

been studied via Brownian dynamics, Molecular Dynamics (MD), and Markov State Mod-

els (MSM). The implications of chemical reactions and diffusion on a system’s efficiency has

been probed via some Kinetic Monte Carlo (KMC) models. However, use of these techniques
for biological systems is limited to short time scales. 40 Time scale limitations compromise a
simulation’s ability to capture transport events and some nuances of chemical reactions. 40,41

In focusing computational efforts on the bridge of HK-G6PDH for simplicity, multiple

studies have combined the previously mentioned computational methods to exploit design

parameters within our group. The 2017 work of Liu et al. complemented experiments with

MD simulations to determine the thermodynamics of adsorption and transport of G6P, and
to screen peptide bridges. 32 Upon determining that a lysine bridge would sufficiently channel

the intermediate, the adsorption energy, desorption probability, and diffusivity of G6P on

the lysine bridge were calculated. Subsequently, MD and KMC methods were utilized in

continuation of the previous work to determine rates of adsorption, desorption, and diffusion

of G6P on the lysine bridge while considering chemical reaction rate constants calculated
from their previous experimental data.7 Here, the lag time of HK-G6PDH was simulated

as a function of ionic strength, bridge length, and desorption on G6PDH after transport

across the bridge. Although the “hopping rate” of transport from one bridge residue to

the next was higher for a longer bridge, the desorption rate was also higher, resulting in

longer lag times for the longer bridge. Lag times were further extended when desorption

on G6PDH was also considered. The previously reported increase of lag time with ionic

strength was also captured in these simulations. Both effects can be observed in Figure 1.8.

13

Figure 1.8: Lag time as a function of bridge length and ionic strength for a lysine bridge in
the HK-G6PDH system (from Ref 7).

The implications of desorption on both enzymes and the bridge were further probed by MSM
and KMC models in 2019.24 These simulations most closely matched experimental lag times

when desorption was considered on the bridge and both enzymes instead of only in other

smaller segments of the system. MD and KMC methods were later utilized to assess the

arginine bridge, which was predicted to yield lower lag times across a range of ionic strengths
than the lysine bridge.27

While the HK-G6PDH system can be simplified to enable macroscopic simulations in

alignment with experiment (Figure 1.9a), the same methods are not as easily applied to

MDH-CS, which has a much larger network of intermediate positions (Fig 1.9b). As a re-

sult, models of MDH-CS to date primarily focus on capturing the details of surface transport

within this network. Surface transport of OAA was first studied computationally with Brow-
nian dynamics. 35 Similar to HK-G6PDH, increasing ionic strength negatively impacted the

transfer of the intermediate. Negatively charged OAA was found to interact with the sur-

face, but charge-neutral OAA did not. More recently, our group used metadynamics and a

MSM to analyze the transfer efficiency of OAA on a recombinant and mutant variation of
MDH-CS.26 Metadynamics enabled long-time MD simulations by introducing a bias energy

that forces OAA to visit states near the active site in shorter MD simulations. These states

14

Figure 1.9: Network models of (a) HK-G6PDH (from Ref 24)and (b) MDH-CS. The HK-
G6PDH system may be modified to this simple model representing each enzyme as a single
node with additional nodes for residuals on the bridge. The MDH-CS system consists of a
much more complex network with 500 sites depicting the four active sites, additional surface
site, and bulk positions of intermediate.

were then used to initiate longer simulations. The collection of these trajectories was ana-

lyzed by a MSM to produce state transition probabilities for states categorized as one of the

four active sites, other surface states, or desorbed intermediate states (Fig. 1.9b). Here, the

mutation, which disrupts the positive charge on the surface, had a transfer efficiency less

than half that of the recombinant.

Previous models of MDH-CS consider surface transport, but are incapable of including

chemical reactions due to the size of the network combined with large differences in time

scale. To overcome the limitations of other methods to handle transport and reaction events,

finite difference models can observe longer time scales while handling the complexities of the
reaction environments.42–45 The work in Chapter 2 utilizes a finite difference model to study
the transient behavior of MDH-CS, as summarized in Figure 1.19a. 46

15

1.2 Multi-scale Modeling of Electrochemical Conver-

sion of Biomass

Bio-derived materials are also capable of providing alternative solutions as value added ma-

terials. With threats of global warming and diminishing supplies of fossil fuels, there exists
a strong need for sustainable and clean means for energy production. 1 Although advance-

ments in alternative energy technology have been made, society still relies on liquid fuels
to supply energy, particularly for transportation applications.47 As such, we turn to na-

ture to supply alternative resources for fuels and related value-added chemicals. The use

of bio-oils from plant-derived lignin has become a popular solution in addressing the world
energy crisis.4,47 Extraction of usable chemicals from lignocellulosic biomass first requires

pyrolysis to produce the bio-oil. Pyrolysis is an energy and heat-intensive process that yields
highly oxygenated molecules with low energy density. 48 Recent work suggests electrochemi-

cally upgrading bio-oils to more value added chemicals may be a viable secondary processing
method. 49–51 However, deconvoluting and understanding the plethora of reaction mecha-

nisms that can take place in these systems is a complex task that remains to be explored in

depth.

1.2.1 Bio-oils from Pyrolyzed Lignocellulosic Biomass

Bio-oils are extracted from biomass via pyrolysis.48 During fast pyrolysis processes, temper-

atures around 500C are applied to small biomass particles for approximately 1 second to

separate volatile chemicals from the undesired mass, which forms biochar. Fast pyrolysis is

advantageous because it yields higher liquid content with shorter residence times and less
biochar content than slow or intermediate pyrolysis.4,48,52 After the reaction, the volatile

output is cooled to form bio-oil. These oils are rich with a variety of molecules. The pyrolysis

process and composition of lignocellulosic bio-oil is depicted in Figure 1.10. The constituents,

however, are often oxygenated and low in energy density. Pyrolyzed oils must be further

processed in order to achieve a greater variety of molecules with characteristics such as high

energy density.

16

Figure 1.10: Schematic of lignin pyrolysis and composition of bio-oils.4,48

While there is a vast array of applications for all of the constituents types, furfurals and
phenols are of particular interest.53–57 Phenol-type molecules can be used in fuels owed to the

high carbon density. Specifically, 4-propylphenol has been shown to yield favorable products
that can be directly implemented in sustainable jet fuels.55–57 Furans and furfural-derived

compounds are of interest because of their versatile applications, including solvents, polymer
substrates, fuels, fuel additives, and textiles.53,54 To achieve the derivates of the platform

molecules in bio-oil, further treatment is required, which comes with a set of engineering

challenges. In fuel applications, the high oxygen content of the molecules creates the first

challenge in efficient post-treatment, or “upgrading”, of pyrolysis oils. The second challenge

arises in the complexity of the bio-oil mixtures with many different molecules that can

potentially inhibit processing of other molecules. An understanding of molecular interactions

is vital to engineering efficient upgrading processes, regardless of application.

17

1.2.2 Electrocatalytic Hydrotreatment of Bio-oil

Electrocatalytic hydrotreatment (ECH) has been utilized to remove the oxygen from the bio-
oil molecules using applied electric potential to drive the hydrogenation reactions.50,51,58–62

This approach is preferred over the traditional method of thermocatalytic hydrogenation

(TCH) because, in addition to the high temperatures utilized to drive reactions, pressurized
hydrogen gas provides the hydrogen for reactions.63 ECH is a more favorable approach than

TCH because it can be operated in ambient conditions, and the aqueous electrolyte supplies

hydrogen.

1.2.2.1 Electrochemical Reactions

During ECH processes, many mechanisms are involved in processing even a single molecule.

For an illustration of the steps involved, consider some reversible reduction reaction where

the equilibrium of an oxidized species (Ox) and a reduced species (Red) exists at some

equilibrium potential, E0, and n moles of electrons are involved.

Ox + ne− + nH+ −−→←−− Red

(1.20)

First, the oxidized species must diffuse through the bulk electrolyte to arrive near the

surface and react. As the molecules approach the electrode surface, the electrolyte and

oxidized molecules must be oriented to favor interaction with the surface. This regime is

called the double layer or diffuse layer, with the oxidized species concentration gradient

dependent on whether the reaction rate is electron transfer or mass transfer controlled, as
shown in Figure 1.11.64

For the reversible electrochemical reaction, the net reaction rate, νnet, is proportional

to the current in response to an applied potential, divided by the number of electrons, n,
Faraday’s Constant, F , and the surface area of the electrode, A.64

νnet =

Inet
nF A

(1.21)

In the electron transfer controlled regime, the net reaction rate is a function of the forward

18

Figure 1.11: (a) Current versus Potential curve and (b) Oxidized species concentration
profile in the diffusion layer with the distance from the electrode interface, x, normalized
to the width of the diffusion layer for a reversible reduction reaction. The electron transfer
controlled regime is colored orange, the mass transfer controlled regime is colored green, and
the mixed control region is colored purple.

and reverse electron transfer rate. Assuming that the electron transfer reactions are first-

order with respect to the reactant concentration (COx & CRed) at the electrode interface,

the net reaction rate is the difference between the forward, or reduction, reaction rate, νf ,

and the reverse, or oxidative, reaction rate, νr.

νnet = νf − νr = kf COx − krCRed

(1.22)

The Butler-Volmer equation relates the electron transfer rates to the observed current64:

Inet = nF A [kf COx − krCRed]

(1.23)

where, the rate constants are related to the over potential, η, or the difference between

applied potential and the equilibrium potential, by an electron transfer constant, α, which

describes the change in free energy barrier with polarization.

kf = k0exp (−αnF η/RT )

(1.24)

19

kr = k0exp ((1 − α)nF η/RT )

(1.25)

where k0 is the standard rate constant which describes the rate when the interface and

solution are at equilibrium and kf = kr.

Although the net current at equilibrium is zero, there is still a need to balance the

electrochemical, or faradaic, activity at the electrode. From this arises the exchange current

where electrons are exchanged at equal rates between the electrode and oxidized or reduced

species. Often the current is normalized to the area of the electrode to yield the exchange

current density. This exchange current density is:

j0 =

I0
A

= F k0C1−α

Ox Cα

Red

(1.26)

In the mass transfer controlled regime, the reaction rate is equivalent to the diffusion

rate of the oxidized species.

νnet = mOxCOx =

DOx
δdl

COx

(1.27)

The mass transfer coefficient, mOx, is the diffusion coefficient, DOx, divided by the thickness

of the diffusion layer, δdl.

1.2.2.2 Adsorption Mechanisms in Electrochemical Reactions

Although adsorption of an organic (Ox species) occurs much faster than electron transfer or

mass transfer and does not impact the reaction rate, the orientation of the oxidized species
on the electrode can affect the details of the reaction mechanism. 65–67 Determination of

organic adsorption configuration is complicated by molecule size and coverage of the surface.

The possibilities of adsorption configuration are expansive, and at moderate coverages, the

configurations change from one in which the ring is parallel to the surface to a vertical one
in which the ring is perpendicular to the surface.66,67 This shift can affect the reactivity of

the organic due to less interaction with the surface.

Organic ECH further requires a proton presence. Most commonly, a Langmuir-Hinshelwood

20

Figure 1.12: Depictions of (a) a Langmuir-Hinsehlwood mechanism of organic and hydrogen
co-adsorption and reaction and (b) an Eley-Rideal mechanism of organic adsorption and
reaction with protons in solution. 13,68,69

mechanism is assumed in which the adsorbed organic reacts with co-adsorbed hydrogen. 68

However, in practice, it is possible that the adsorbed organic could also be hydrogenated by
protons in solution, per an Eley-Rideal mechanism.69 The two mechanisms are illustrated

in Figure 1.12 for some organic molecule A.

1.2.2.3 Hydrogen Evolution Reaction

In electrochemical processes, the reducing potentials required for the hydrogenation of most
organics occur near that of hydrogen deposition and the hydrogen evolution reaction (HER). 70

While the presence of hydrogen aids in more efficient hydrogenation, it also enhances the

competition for the active sites necessary for organic hydrogenation. The net reaction of

HER is:

2H+ + 2e− −−→←−− H2

(1.28)

21

The equilibrium potential of HER is a function of the standard hydrogen electrode potential,
E0, which is zero when utilizing a reversible hydrogen reference electrode. 70 The activities
of protons in solution (αH+) and hydrogen gas (αH2

) are also accounted for:

Eeq = E0 +

1
2f

ln

(cid:33)

(cid:32)α2

H+
αH2

(1.29)

where f = F

RT , the ratio of Faraday’s constant to the gas constant - temperature product.
During HER, protons can interact with the electrode through a Volmer, Heyrovsky, or
Tafel mechanism. 70,71 The Volmer and Heyrovsky mechanisms (Eqs. 1.30 & 1.31) entail

electrochemical adsorption where the former exhibits only proton adsorption and the latter

exhibits the dissociation of H2 and adsorption of a proton. An Eley-Rideal mechanism is

comparable to the Heyrovsky step. A Tafel mechanism (Eq. 1.32) represents the dissociative

chemisorption of H2.

Volmer : H+ + ∗ + e− −−→←−− H∗

Heyrovsky : H ∗ +H+ + e− −−→←−− ∗ +H2

Tafel : 2H ∗ −−→←−− 2 ∗ +H2

(1.30)

(1.31)

(1.32)

In acidic media, and assuming a Langmuir adsorption isotherm, the reaction rates of
Volmer, Heyrovsky, and Tafel steps in units of mol · cm−2 · s−1, are a function of the fraction
of free surface sites, θ, and the fraction of active sites covered with hydrogen, θH.70

νVolmer = kvf CH+θ(1 − θH)e−βvf (E−E0

v) − kvrθθHe(1−βv)f (E−E0
v)

(1.33)

νHeyrovsky = khf CH+θθHe−βhf (E−E0

h) − khrCH2

θ(1 − θH)e(1−βh)f (E−E0
h)

(1.34)

22

Figure 1.13: A generic Tafel plot of current density vs. overpotential from Ref73.

νTafel = ktf θ2θ2

H − ktrθ2(1 − θH)2CH2

(1.35)

In these expressions, the forward and reverse rate constants are represented by kif and kir,

respectively. The standard potentials of the Volmer and Heyrovsky mechanisms are denoted
by E0

h. The symmetry coefficients of the Volmer and Heyrovsky mechanisms are βv

v and E0

and βh.

The Volmer and Heyrovsky mechanisms are differentiated by analysis of a Tafel plot, as

shown in Figure 1.13. The value of the slope of the linear portion can suggest the utilized
mechanism. A slope of 30 mV dec-1 suggests a Volmer-Tafel mechanism and a slope of 40
mV dec-1 suggests a Volmer-Heyrovsky mechanism. 72 If the slope is 120 mV dec-1, then the

rate-limiting step is dependent on surface coverage of hydrogen where, at low coverages, the

Volmer step is rate-limiting and at saturation, either the Volmer or Heyrovsky mechanism

can be rate-limiting.

During electrochemical hydrotreatment of organic molecules, the hydrogen and organics

are in competition for adsorbed sites. The number of remaining free sites, θ∗, and those

occupied by either hydrogen (θH) or an organic (θorg), all relative to the total number of

sites, sum to unity:

1 = θ∗ + θorg + θH

(1.36)

23

Therefore, there will be competition between the organic reactions and hydrogen evolution

because the available number of active sites for organic adsorption is dependent on hydrogen

coverage. Additionally, the standard potentials of the Volmer and Heyrovsky steps are 0

V versus a standard hydrogen electrode.

If the applied potential is positive against the

same reference, favoring oxidation instead of HER, protons in solution will be favored for

the Volmer step, and adsorbed hydrogen will be favored in a Heyrovsky mechanism due

to the negative exponent in the exponential factor for the forward reactions. When the

applied potential is reduced to negative values against the standard hydrogen electrode,

the favorability of the Volmer and Heyrovsky steps is reversed, favoring adsorbed protons

and hydrogen gas, respectively. In either case, the Tafel mechanism presents a competitive

hydrogen pathway due to hydrogen coverage at positive potentials and hydrogen gas presence

at negative potentials.

1.2.3 Thermodynamic Evaluation

Reaction mechanisms in the conversion of bio-oil molecules to value added chemicals may

be explored by quantum modeling. Energetics and thermodynamics of reactions in compar-

ison to other mechanisms drive the outcome of the reaction system. Using quantum level

simulations, we can identify rate limiting steps and reaction pathways based on the reac-

tion barrier, adsorption energy, and reaction energy. Further work may extend modeling to

multiple scales to enable further comparison to experiment.

1.2.3.1 Density Functional Theory

Quantum-level simulations may be executed with Density Functional Theory (DFT) to un-

derstand the electronic structures of atoms within a molecular system. Specifically, since

the electron clouds of an atom are much more sensitive to environmental changes than the

nucleus, the electron density and changes therein in response to external factors are the
focus. 74 The Born-Oppenheimer approximation describes this separate treatment of elec-

trons and nuclei when searching for the electrons’ lowest energy state, or the ground state.

The ground state is the electronic structure that yields the lowest energy while solving the

24

Schr¨odinger equation. Using the variation of Kohn and Sham, a set of Schr¨odinger’s equa-
tions are written for single electrons in the form of74,75 :

(cid:20) h2
2m

(cid:21)
∇2 + V (r) + VH(r) + VXC(r)

ψi(r) = Eiψi(r)

(1.37)

where h is Planck’s constant, m is the electron mass, r describes the coordinates of an

electron, ψi is the wave function of an electron i, which describes the probability of an electron

being at some position, r, and Ei is the electronic energy. The four terms within the square
brackets represent the potentials to be considered in the solution. 74,75 The first term, h2
2m ∇2,
represents the kinetic energy of the electron. The interaction between the electron and the

nuclei around it is described by V (r). The Hartree potential, VH(r), accounts for repulsion

between the electron and the remaining electron density. The final term, VXC(r)corrects

for exchange and correlation contributions within the single-electron equations, such as the

non-physical self-interaction energy that results from the Hartree potential.
Kohn-Sham equations are solved in four steps 74:

In practice,

1. Select an initial electron density.

2. Solve the single electron Schr¨odinger equation for the electrons from Step 1.

3. Calculate an electron density from the solution determined in Step 2.

4. Compare the initial electron density to the one calculated in Step 3.

If the two electron densities from Steps 1 and 3 are the same, then the ground state has

been found. If not, the process should be repeated with a different guess at electron density

in Step 1 until the ground state is found. System attributes such as total energy can be

calculated once the electron densities have been described as a function of nuclear positions.

Changes in the electronic structure and factors like total energy, due to the movement of

nuclei, can also be described through this process.

1.2.3.2 Adsorption Thermodynamics

To probe adsorption thermodynamics by DFT, three states must be assessed including a

bare surface slab with some vacuum layer perpendicular to the top layer, the free organic

25

Figure 1.14: Three states considered in calculation of adsorption energy:
molecule, bare slab, and the adsorbed state.

free organic

in vacuum, and an adsorbed state in which the organic is bound to the surface slab. A

schematic depicting the states is in Figure 1.14. Energy prediction for the first two states,

the bare slab and organic in vacuum, is relatively straightforward. The lattice parameters

of various facets of many surface crystals are available in databases and are experimentally
determined.76,77 The slab size used varies somewhat in literature, but a 4x4 atom surface
with 3 to 5 layers is commonly used to model a surface. 77–80 Similarly, the atomic positions
of organic molecule isomers in vacuum are available through PubChem.81 Both the bare slab

and the free organic undergo complete geometric optimization, or relaxation, in which the

energy and force are both optimized to a given threshold.

Calculation of the adsorbed state energy is a more tedious task. Using the relaxed

geometry of the surface slab, layers beneath the top 2 are fixed to represent the bulk surface
that would not react to an adsorbate.82–86 The adsorbate is then placed near the top of the

surface. The exact configuration of the adsorbate relative to the surface can vary, resulting
in substantial differences in adsorption energy. 85 Consequently, when working with a new

system, one must test an array of possible adsorption configurations. The horizontal, or

“ring-down”, configuration exhibits stronger binding than a vertical configuration and is
the focus of most studies.85 Even still, the placement of the ring and substituent group(s)

can vary. The center of the ring and substituent group(s) can take atop, hollow, or bridge
configurations on most surfaces. 87 Combination of ring and substituent group(s) positions

relative to the surface can lead to an expansive list of possibilities that must all be assessed.

An example of possible positions for a five member ring with one substituent on an FCC(111)

26

Figure 1.15: Surface-parallel adsorption configurations of a five member ring with one sub-
stituant group. Ring positions are color coded and each have two substituent positions. The
substituant positions are (a) hollow and (b) atop for an atop ring, (c) bridge and (d) hollow
for a ring over a bridge, (e) bridge and (f) hollow for a ring over a hollow, and (g) atop and
(h) FCC hollow for a ring over a FCC hollow.

slab is shown in Figure 1.15. Whichever adsorption configuration, upon relaxation, that

yields the lowest energy is considered the preferred state. The adsorption energy of some

organic, i, is:

i(g) + ∗ → i∗

Eads,i = Ei∗ − Ei(g) − E∗

(1.38)

(1.39)

where the adsorbed state energy is Ei∗, the energy of the free organic is Ei(g), and the

slab is represented by E∗.

1.2.3.3 Reaction Energy

The thermodynamics of reactions are similarly predicted by the comparison of state energies.

However, this will focus more on successive hydrogenation steps that occur on the surface.

As such, the relaxed states of the various hydrogenation products are required. The prod-

uct state can initially assume the positions of the reactant state to enhance computational

27

efficiency. The energy of reaction, for some reactant, i∗, and some product, (i + H)∗, is:

i ∗ +H∗ → (i + H) ∗ +∗

Erxn = E(i+H)∗ + E∗ − Ei∗ − EH∗

(1.40)

(1.41)

1.2.3.4 Gibbs Free Energy Corrections in Adsorption and Reaction

The internal energies calculated in the geometric optimization process represent that at ab-

solute zero temperature and zero pressure. To assess the thermodynamics of these systems in

line with experimental conditions, steps must be taken to align the energies with what would

occur at some non-zero temperature and pressure. Computationally, this is accomplished

by prediction of Gibbs free energy through vibrational corrections to the electronic energies
previously discussed. For a given state, j, the Gibbs free energy is88,89:

Gj = Ej + EZPE − T S + H0→T

(1.42)

where Ej is the electronic energy calculated in the relaxation, EZPE is the zero point energy,

or the accounting of vibrational energy of a system at absolute zero, T S is the entropic

contribution, and H0→T are additional enthalpic contributions beyond Ej + EZPE due to

heating the system from 0K to some arbitrary temperature, T .

The Gibbs free energy of some elementary reaction step would then be:

∆Grxn = GFS − GIS

(1.43)

where GFS and GIS are the free energies of the final state and the initial state, respectively.

This can be applied to reactions and adsorption or desorption.

1.2.3.4.1 Potential Dependence To apply the thermodynamic evaluation to electro-

catalytic processes, one must further account for the impact of applied potential. The pres-

ence of protons and electrons in solution are accounted for by the Computational Hydrogen

28

Electrode (CHE) method.90 Here, the chemical potential of a proton-electron pair (H+ +
e−) in solution is assumed to be equivalent to half the chemical potential of a hydrogen gas

molecule (H2(g)) at 0 V vs. RHE (0 URHE) and any value of pH. At any value of URHE, the

resulting relationship is described by Eq. 1.44.

µ(H+ + e−) = 0.5µ(H2(g)) − |e|URHE

(1.44)

Here, |e| is the elementary charge in units of coulombs and the term |e|URHE is converted

from units of joules to eV.

Elementary electrochemical steps, assumed to include proton-coupled electron transfer
(PCET)91,92, are also corrected for applied potential. The product of elementary charge and

applied potential are similarly accounted for in the change in free energy expression, such
that Eq. 1.42 is expanded to be93 :

∆G = ∆E + ∆EZPE − T ∆S + ∆H0→T + n|e|URHE

(1.45)

where the |e|URHE is again converted to have units of eV and n is the number of electrons

used in the reaction.

1.2.3.5 Experimental Adsorption Thermodynamics

The free energy and heat of adsorption of an organic species on a Pt electrode can be probed

experimentally through organic-blocked hydrogen electrochemical under potential deposition
(Hupd).94,95 In the reductive window during a cyclic voltammogram or linear voltage sweep,
the amount of adsorbed hydrogen is a function of the sum of peaks, as highlighted in Figure

1.16. When organic adsorption blocks hydrogen adsorption, the sum of peaks will decrease

and enables the calculation of fraction of Hupd blocked, which is equivalent to the amount

of organic adsorbed.

The relationship between fraction Hupd blocked (θ) and the heats of adsorption of the

organic is described by a Frumkin isotherm. Here, the fraction of Hupd inhibited as a function

of concentration is fit to the isotherm to determine the saturation coverage, θsat, the free

29

Figure 1.16: Example of hydrogen adsorption peaks in a CV. The inset particle presents the
various facets of a Pt nanoparticle, which correspond to the 3 sub peaks highlighted in the
CV. Adapted from Ref94.

energy of adsorption at zero coverage, ∆Gaq
of adsorption, α.95

ads,θ=0, and the linear scaling slope for free energy

C
C0

=

θ
θsat
1 − θ
θsat





exp

−∆Gad,aq,θ=0
RT

−





α θ
θsat
RT

(1.46)

For the aqueous phase experiment, the Gibbs free energy (∆Gad,aq) and heat of adsorption

(∆Had,aq) are related to Keq,aq with an assumed entropy of adsorption value (∆Sad,aq).

Keq,aq = e

−∆Gad,aq/RT

= e

−∆Had,aq/RT

−∆Sad,aq/R
e

(1.47)

To compare with the gas phase free energies calculated by DFT, the aqueous phase

relationships are extrapolated to the gas phase through Henry’s Law. Here, Keq,gas is deter-

mined from the aqueous equilibrium coefficient and Henry’s constant, KH, which becomes

unitless when multiplied by RT, and used in Eq. 1.47 to determine the values of ∆Gad,gas
and ∆Had,gas.95

(1.48)

Keq,aq
KH

Keq,gas =

30

Henry’s Law enables estimation of the solvation of the organic from gas to aqueous

phases. The resulting isotherms are a function of θ, and therefore, can be evaluated at a

given coverage to correspond to the coverages used in DFT models. Chapter 3 employs

Hupd analysis in comparison to a potential dependent thermodynamic assessment by DFT

to understand mechanisms involved in the electrochemical hydrotreatment of a furfural and

4-propylphenol binary mixture (Fig. 1.19b).

1.2.4 Kinetic Evaluation

Thermodynamic evaluation of hydrogenation reactions provide a substantial amount of infor-

mation toward understanding competing mechanisms. However, the varied rates of reaction

can also affect the competition between reactions and, therefore, must also be assessed.

1.2.4.1 Activation Barrier

Prediction of reaction kinetics requires information about the activation barriers. Computa-

tionally, several methods are utilized to search the reaction path and find the transition state

(TS), as summarized in Figure 1.17. The TS with the highest energy represents the highest

barrier in the reaction. Some common methods include nudged elastic band (NEB, Fig.
1.17a) or the climbing image (CI-NEB) variation, and the dimer method (Fig. 1.17b).96,97

However, the TS search methods are computationally demanding, especially for large sys-

tems representing heterogenous catalysis. More recently, linear scaling relationships have

been developed to correlate the energy of an activation barrier (Fig. 1.17c) or TS (Fig
1.17d) with that of the initial and/or final states. 98 This approach offers the advantage of

quick estimation of transition state energetic information where the relationship has been

defined for a specific reaction type using a method like NEB, CI-NEB, or the dimer method.

Elastic band methods search for a TS along a pathway between an initial state (IS) and

final state (FS). The subsequent images are connected via “springs” with spring constants to

set the forces that determine how stiff the pathway is, or how far apart images are from each
other.96 The primary difference between NEB and CI-NEB is that, in the climbing image

variation, the optimized TS is found by moving the highest energy image to the appropriate

31

Figure 1.17: Summary of methods to determine transition states and activation barriers.
(a) NEB, (b) Dimer, (c) Brønsted - Evans -Polyani (BEP) relationship, and (d) Transistion
State Scaling (TSS) relationship. Figs a-b are adapted from Ref96 and Figs c-d are from
Ref98.

32

place in the reaction path. The standard NEB method may get close to the actual TS, but

does not explicitly optimize the path to find the highest energy state. Commonly, studies

using NEB to find a transition state will request an odd number (often 5) of images along

the path to increase the likelihood of finding the saddle point. For greater computational

efficiency, one may also combine techniques to conduct an initial pathway search using NEB,

then refine the pathway with CI-NEB from the optimized pathway found in the initial
search.99,100

In comparison to the elastic band methods, the dimer method predicts the transition
state and reaction pathway from only the initial state image and a closely related image.97

The pair of images, which are almost identical, are referred to as the “dimer”. The dimer is

then moved along the potential energy surface assuming that the IS is an energy minima. As

the dimer moves, the curvature of the potential energy surface is determined. The motion

continues until a minimum curvature is determined and the transition state is found. While

this approach may be beneficial in regard to the lack of final state requirement, this is only

a local search whereas the NEB approaches are global methods. NEB and CI-NEB would,

therefore, have more flexibility to search along a specified path. For any approach, the

activation barrier, Ea is the difference in energy between the TS, or the highest energy state

found along the pathway, and IS.

Ea = ETS − EIS

(1.49)

Linear scaling relationships have been developed to generate quick information about
the energetics of the activation barriers.98 Brønsted - Evans -Polyani (BEP) relationships

correlate the activation energy to the reaction energy of a given step. Transition state scaling

(TSS) relationships correlate the TS energy to the energy of the final state. The relationships

of BEP and TSS are defined as, respectively:

Ea = αErxn + β

ETS = α′EFS + β′

33

(1.50)

(1.51)

where α and α′ represent linear slopes and the βs are the y-intercept. The two relation-

ships can be combined to establish a unified expression:

ETS = αEFS + (1 − α)EIS + β

(1.52)

The slopes are equivalent and the y-intercepts are related.

β′ = (1 − α)EIS + β

(1.53)

Application of linear scaling relationships appears straightforward, but it is imperative

that constants used in further calculations are consistent with the type of reaction and

surface. Initial use of BEP relationships focused primarily on small molecule bonds such

as C-O bonds in CO2, COOH, and CO, and C-H bonds in ethane and methane. Similar
systems were also probed with TSS.98 More recently, relationships have been determined for
bonds within more complex molecules such as phenol and furan derivatives.101–103 These

insights are promising for estimation of transition energies, but should be used cautiously in

application due to the limited information on less popular surface facets.

Similarly to reaction energies, the implications of variable applied potential on activa-

tion barriers can be assessed. This is accomplished through application of Marcus theory

where the energy of a transition state references the potential dependence of the proton in
electrolyte.93 The arbitrary elementary steps for the hydrogenation of organic species A are

as follows:

A ∗ +H+ + e− → A ∗ +H∗

A ∗ +H∗ → AH∗

(1.54)

(1.55)

While assuming the CHE method, the equilibrium potential, U0, is a combination of
state free energies. 90,93 The free energy change of the activation barrier, ∆Ga, can then be
predicted at any potential, U .93 Details of this analysis are reported in Chapter 4.

34

1.2.4.2 Computational Predictions of Rate Constants

Once activation barriers are predicted by DFT, the kinetic rate constants can also be pre-

dicted. Transition State Theory postulates that a rate constant, k, is related to the activation
barrier free energy, ∆Ga, at a given temperature, T .104,105 Similarly to the activation bar-
riers, potential dependence can also be further considered in kinetics. 104 Chapter 4 extends

the work within Chapter 3 by employing DFT activation barriers to predict reaction kinet-

ics in the electrochemical hydrogenation of a furfural and 4-propylphenol binary mixture.

Chapter 4 is summarized in Fig 1.19c.

1.3 Manufactured Cell-Laden Tissue Scaffolds

In a further desire to mimic biology for effective engineering solutions, bio-inspired materials

are a current focus in medical applications. Development of high-performance materials as a

solution to treat bone degradation resulting from osteoporosis has been a prevalent research
direction in the biomedical field.106 A successful material must meet several criteria to be

considered for these applications. Tissue-regenerative materials must exhibit biocompat-

ibility, maintain stability in physiological conditions, mimic the extracellular matrix with
structure tunablity, and promote cell viability. 107–110 Collagen, and the cheaper alternative,
gelatin, meet most of the requirements as a biomaterial.108

1.3.1 Bio-ink Development for Tissue Regenerative Materials

While initially attractive, gelatin degrades at physiological temperatures (˜37C). 111 To over-

come stability constraints, gelatin methacryloyl (GelMA) was developed to enhance material

stability by bonding methacrylate to the gelatin backbone. Chemical modification of gelatin
increases the melting point above physiological temperatures. 111 Hydrogels of GelMA also
take longer to experience the same degree of mass loss as gelatin. 107,110 To address cell

viability, GelMA was further modified by chemical linkage to silver-doped bioactive glass
(Ag-BG).112 Bioactive glass is a traditional component in tissue-regenerative materials with
appreciable cell viability properties. 106 Addition of silver has also been shown to increase

35

antibacterial properties of biomaterials. 109,112 Chemically linked GelMA-AgBG (GAB) is

therefore anticipated to have superior cell viability properties to GelMA alone.

1.3.2 Development and Initial Validation of a New Bio-ink for 3D

Printed Cell-Laden Tissue Scaffolds

The dissertation work of Marsh sought to develop and validate the use of GAB within 3D
cell laden, or 3D Multi-scale (3DBP), tissue regenerative scaffolds.5 Early work utilized a
polymer foam method to support a coat of Ag-BG sol-gel.113 A follow-up study sintered Ag-

BG nanoparticles derived from the sol-gel to a polymer foam base scaffold and compared the
performance implications of sintering temperature. 112 A fused filament method was further
used to enhance the compressive strength of the Ag-BG-polymer scaffold. 109 Finally, Ag-

BG was incorporated with gelatin based inks for 3D bioprinting for enhanced tunablity of
scaffold design while protecting the cells from the forces present during printing.110 Optical

and SEM images of the scaffolds produced through Marsh’s methods are provided in Figure

1.18. This work is the foreground for further validation reported herein.

The polymer foam supported Ag-BG scaffold incorporated the bioglass to promote an-

tibacterial properties of the scaffold while studying the implications of processing on the
structural characteristics (Fig. 1.18a).113 A polyurethane foam provides can provide a tun-

able 3D network with a high surface area. An increased surface area is beneficial for methods

using a sol-gel application because it enables an increased amount of antibacterial or ther-

apeutic components that increase the bioactivity of the scaffold. The Ag-BG components

were found to be homogeneously distributed on the scaffold at the micron-scale. The bioac-

tivity of the scaffold against methicillin-resistant Staphylococcus aureus (MRSA) was also

assessed by direct and indirect tests. The direct test where MRSA was exposed to the scaf-

fold showed a decrease of nearly 6 orders of magnitude in bacteria viability within 24 hours.

Indirect tests where MRSA was exposed to extracts of the scaffolds also showed decreases in

bacteria viability. The bioactivity of the Ag-BG decreased over the 21 day time span of the

indirect tests, which is attributed to a decrease in Ag concentration over time. While these

scaffolds expressed bioactivity for an extended amount of time, the compressive strength of

36

Figure 1.18: Optical and SEM immages of Ag-BG scaffolds produced via (a) polymer foam
method, (b) Ag-BG sintered polymer foam, (c) fused filament method, and (d) 3D printing.
Image adapted from Marsh.5

37

the Ag-BG scaffold is reported to be low for similar structures, indicating that it may only

be viable for low load bearing applications. Nonetheless, this confirms that the inclusion of

Ag-BG promotes antibacterial properties when incorporated in to a polymer scaffold.

In the follow up study, particles of Ag-BG were derived from the sol-gel and sintered to the
polymer foam(Fig. 1.18b).112 The sintering temperature was varied from 900C to 1000C and

the implications on material and anti-bacterial characteristics were observed. Compared to

the unaltered foam, sintering at 900C decreased the pore size and 1000C sintering increased

the pore size. The 900C sintered structure could not be assessed for compression strength

because it was too brittle and prematurely collapsed. While the 1000C sintered scaffold

could be tested for compression, the compression strength was still low. Further, the 1000C

sintered scaffolds exhibited slightly better anti-bacterial properties against MRSA than the

900C sintered scaffold. The 1000C scaffolds also expressed the ability to revive antibiotics

during the first 24 hours, but did not see significant difference with or without antibiotics

present beyond 48 hours. The material and bacterial characteristics of the polymer foam

based Ag-BG scaffolds have been shown to be dependent on process variables. These scaffolds

show promising results for low load bearing applications, but could be enhanced by improved

compression strength.

The fused filament method is a printing method in which a polymer filament is printed

to form the desired scaffold (Fig. 1.18c). This method also offers more control over matrix

design than the polymer foam. In 2020, Marsh et al. reported using this approach with Ag-
BG for the first time.109 Polyolefin and Ag-BG powder were combined in to the base filament
and printed to form a scaffold of approximately 1 cm3. The scaffolds were determined to

have homogenous elemental distribution to the micro-scale, and porosity and compression

strength comparable to that of human bone. Similarly to the polymer foam work, MRSA

was exposed to the scaffolds for up to 48 hours. At the 24 hour mark, viability of the

bacteria had decreased, but only by 2 orders of magnitude compared to the almost 6 order

of magnitude decrease of the polymer foam. Some further decrease in bacteria viability was

observed at the 48 hour mark, but it was not as substantial as the first 24 hours. Antibacterial

properties were present and the scaffolds were found to be comparable to bone in porosity and

compression strength, making this approach a viable option for tissue regenerative scaffold

38

manufacturing.

Synthetic polymer approaches have shown promising results as a support for Ag-BG in

antibacterial properties for tissue generation, but synthetic materials are often not compati-

ble with cells. To improve the compatibility with cells for tissue growth, bio-based polymers

were explored. In 2021, Marsh et al. reported initial studies of Ag-BG incorporated in to
gelatin based bio-inks in 3D bioprinting manufactured scaffolds (Fig. 1.18d). 110 A mixture

of GelMA and Ag-BG has limitations in homogeneity of ink composition and may impose

further risk to the cells due to extra force from the Ag-BG particle size. Marsh et al.’s work

instead chemically linked the Ag-BG to GelMA, forming GAB, to overcome this. The GAB

exhibited superior printability and rheological characteristics compared to GelMA alone or

a nano-composite blend of GelMA and Ag-BG acting as a control. Specifically, the GAB

ink had viscosity, storage modulus, and loss modulus orders of magnitude larger than the

GelMA or control. This suggests that GAB is able to absorb and recover from the forces

of printing better than the other materials. Combining the ability to protect cells from the

forces of printing and the antibacterial properties from the Ag-BG, GAB is predicted to be

a superior bio-ink for 3DBP of tissue generative scaffolds. Chapter 5 presents further work

toward validation of GAB as a superior bio-ink in 3DBP of tissue scaffolds including 3DBP

parameter optimization and some cell viability studies in comparison to GelMA and the

nano-composite control. This chapter is summarized by Fig. 1.19d. Necessary future work

for complete validation of this approach are also discussed.

1.4 Overview of Included Works

The remainder of this dissertation will discuss three research topics in the area of bioelectro-

chemical engineering. Chapter 2 presents the first project yielding a manuscript on the topic
of electrostatic transport within TCA cycle enzymes.46 Chapter 3 reports the details of a

second project that resulted in a manuscript on the topic of thermodynamic assessment and

mechanistic insight in the electrocatalytic upgrading of bio-oil constituents. Chapter 4 adds

a kinetic assessment to the system of Chapter 3. The work in Chapter 5 leans more toward

biochemical engineering and reports validation work of a new bio-ink material for manu-

39

Figure 1.19: Summary of Included Research Projects: (a) Substrate channeling pathways in
TCA Cycle enzymes46, (b) interactions in ECH of Furfural (FF) and 4-Propylphenol (4-PP)
binary mixture on a Pt/Ru electrocatalyst, (c) kinetic evaluation of FF and 4-PP ECH on
a Pt/Ru electrocatalyst, and (d) extrusion printing of gelatin-hybrid hydrogels.

40

factured tissue regenerative scaffolds. Chapter 6 summarizes the conclusions of Chapters 2

through 5 and discusses future work directions from this collection of studies.

Chapter 2 (Fig. 1.19a) explores the impact of electrostatic channeling in the conversion

of malate conversion to citrate via oxaloacetate (OAA) intermediate on malate dehydroge-
nase - citrate synthase (MDH-CS). 25 Kinetic parameters determined experimentally16 and
Markov state transition matrices produced by previous molecular dynamics simulations26

are combined in a finite difference model to calculate time trajectories of surface and bulk
occupancies by OAA in the system. 46 Solution of a system of differential mass balance equa-

tions, based on surface transport rates, provide occupancy probability at each node in the

network. Lag time of MDH-CS was determined computationally to be comparable to exper-
iment for recombinant and mutant complexes. 16 Using the model, dynamics of each of four

possible reactiond pathways between two source (MDH) active sites and two sink (CS) sites

could be studied independently. This analysis provides a dynamic model for intermediate

transport in an electrostatically channeled system, and can be used as a predictive tool to

provide mechanistic insight in to path dominance.

Chapters 3 and 4 evaluate the mechanisms in the electrocatalytic hydrogenation (ECH)

of a 4-propylphenol (4-PP) and furfural (FF) binary mixture on a Pt/Ru electrocatalyst.

In Chapter 3, the thermodynamics of each reactant’s reaction pathways are assessed com-

putationally (Fig. 1.19b) . Competitive adsorption between FF and 4-PP along with the

respective stable intermediates is assessed with DFT for the FCC (100), (110), and (111)

facets of Pt and Ru(0001). The experimental adsorption free energies of select stable in-

termediates are further explored by organic-blocked hydrogen under potential deposition on

Pt. The reaction free energies are also calculated for the experimentally suggested path-

ways of FF to tetrahydrofurfuryl alcohol (THFA) and 4-PP to propylcyclohexane (PC). The

affect of applied potential on the reaction free energies was a further point of exploration.

This mechanistic study provides insight into the competition between two bio-oil constituent

molecules.

Chapter 4 adds a kinetic analysis from the activation barriers of the energy pathways

determined in Chapter 3 (Fig 1.19c). Here, linear scaling relationships for furanics and

phenol are used to predict transition state energy from the energy of intermediate states on

41

the Pt and Ru facets studied in Chapter 3. Through predicting the transition state energy,

activation barriers are determined which enable prediction of kinetic rate constants. The

implications of variable applied potential are assessed for the transition states and barriers

similar to the methods used in the previous work.

Chapter 5 (Fig. 1.19d) compares cell viability in extrusion printed scaffolds of GelMA,

GAB, and blended GelMA + AgBG bio-inks toward validation of GAB as a novel bio-ink

with enhanced cell viability properties. First, GelMA and GAB materials were synthesized

and characterized as reported in literature. Then the extrusion printing parameters were

optimized for the three bio-inks. Upon printing parameter optimization, cell viability studies

were conducted for GelMA, GAB, and blend bio-inks with fibroblast(3T3) cells. GelMA

bionic was the only one to enable cell viability, which lasted less than 24 hours. The poor

performance of these bio-inks was attributed to the acidic pH of the materials in inks,

resuspension issues resulting from non-optimal lyophilization, and the potential toxicity from

residual toluene after synthesis. Correction of these factors are a point of future work.

42

Chapter 2

Finite Difference Model of

Electrostatic Channeling in TCA

Cycle Enzymes

2.1

Introduction1

Complex living organisms utilize the highly efficient metabolic pathways of biocatalyzed
reaction cascades to perform vital functions necessary for survival3,6,7,114. These reaction

cascades can convert chemicals into energy sources and other materials needed to satisfy
metabolic requirements3,6. The natural potential of catalytic cascades to produce novel

chemicals (e.g., biofuels) at high yields has ignited research efforts aimed at understanding
their fundamental mechanisms3,7.

Within such catalytic cascades, interactions of reaction intermediates with biocatalytic

surfaces can control intermediate transmission between active sites, a phenomenon known as
channeling2. The traversal of charged intermediate species on an oppositely charged surface,
or electrostatic channeling, is one channeling mechanism featured in natural cascades7,27.

Electrostatic interactions maintain short distances between the intermediate and the sur-
face, allowing for the shortest path from one enzyme to another24. Transport via surface
interactions often occurs quickly, contributing to efficiency26. Without these interactions, it

can take excessive time for the intermediate to desorb from the first enzyme, diffuse through

1This chapter

is adapted from Wark, C. M.; Xie, Y.; Calabrese Barton, S. Finite Dif-
Electrochimica Acta.

in TCA Cycle Enzymes.

ference Model
of Electrostatic Channeling
https://doi.org/10.1016/j.electacta.2025.146131.

43

Figure 2.1: Reactions on MDH-CS Complex.

the bulk solution, and adsorb onto the second enzyme before being converted to a product.

The tricarboxylic acid (TCA) cycle, also known as the Krebs cycle, is a biocatalytic
cascade vital to efficient energy production in metabolism 115,116. One step of the TCA cycle

involving the malate dehydrogenase - citrate synthase (MDH-CS) complex, is a research
focus 117,118. As shown in Figure 2.1, MDH-CS converts malate to citrate by way of an
oxaloacetate (OAA) intermediate and a CoenzymeA (CoA) cofactor 25.

The MDH-catalyzed reaction tends to favor the reverse conversion of OAA back to malate
within the reaction equilibrium, contrary to the desired direction119. The reverse reaction

becomes significant as the concentration of NADH (the reduced form of Nicotinamide adenine
dinucleotide ,NAD+)) increases with OAA generation. This challenge is addressed through

electrostatic channeling, utilizing the attractive forces between negatively charged OAA and
the positively charged area on the surface of the complex26,118. This attraction effectively
pulls OAA away from the MDH active site and directs it toward the CS active site 118.

Previously, experimental and computational approaches have increased understanding of

the mechanisms of transport and key factors that drive the efficiency of MDH-CS. Bulutoglu

and coworkers have shown that the complex does utilize electrostatic channeling, and have
reported kinetic parameters 16. The dynamics of reaction and transport between the two

enzymes were characterized by measurement of lag time, which represents the time required

for the overall citrate production rate to achieve steady state. For native bovine, commercial

wild-type porcine, and recombinant porcine complexes, lag time was determined via transient

analysis of CoA concentration. Additionally, a mutated version of the recombinant porcine

complex was considered in which a positively-charged arginine residue (Arg65, Fig. 2.2) is

replaced with charge-neutral alanine. The native tissue and recombinant complexes had the

shortest lag times of 40 and 30 ms, respectively, both lower than that of the commercial

complex (290 ms). By mutating the recombinant complex, a substantial increase in lag

time to 880 ms was observed. This comparison of lag times suggests that the complex does

44

Figure 2.2: Electrostatic potential surface of MDH-CS Complex adapted from Xie et al 26,120.
Positively charged areas are colored blue, negatively charged areas are red, and neutral
charges are colored white.

exhibit electrostatic channeling and that modifications to the electrostatic surface is capable

of dramatically affecting intermediate transport.

Computationally, molecular and atomistic level simulations have been employed to ad-
vance the understanding of mechanisms that occur in catalytic cascades 8,27,39. The use of

these techniques for biological systems is, however, primarily limited to time scales of less
than 1 µs 40. Time scale limitations result in a compromised ability of the simulation to

capture transport events and most aspects of chemical reactions having rate constants on
the order of 1 second 40,41.

Molecular Dynamics (MD) and Kinetic Monte Carlo (KMC) simulation methods have

been used by our group to study transport phenomena at varying time scales in biocat-
alytic cascades 7,24,26,27,32. Generally, MD simulations are limited to time frames less than 1

µs, whereas KMC simulations can encompass longer time scales but become slow when the
number of rate phenomena or the range of time scales considered becomes large 41,121–124.

We have used KMC to study a bridged enzyme system, namely hexokinase and glucose-6-
phosphate dehydrogenase bridged by a polyarginine peptide 7,24,27,32. This model success-

45

fully predicted experimental lag time measurements as a function of ionic strength and bridge

length. Similarly, ionic strength was found to affect reaction probability in the computa-

tional Brownian Dynamics study of OAA transfer mechanisms on MDH-CS by Huang et
al.35. They also showed that OAA would interact with the surface only if it carried negative

charge and that transfer efficiency and reaction probability decreased with increase of ionic

strength.

Recently, we have studied the electrostatic channeling of OAA on the MDH-CS com-

plex with a Markov state model (MSM) to determine the dominant pathways and transfer
efficiency of OAA26. For the MSM, 400,000 frames from MD trajectories of OAA on the

MDH-CS surface were clustered into 500 micro-states. The micro-states were grouped into

basins to analyze transport efficiency. The basins categorize active site states, desorbed

intermediate states, and surface intermediate states. With two active sites on each of the

enzymes, four surface paths are possible, each of which was analyzed. The path from the

first MDH active site to the first CS active site was found to be the dominant pathway with

the lowest probability of desorption for both the recombinant and mutant complexes. Upon

mutation, the transfer efficiency from the first MDH active site to either CS active site was

found to decrease from 96% to 47% for the recombinant and mutant complexes, respectively.

This previous Markov-state study of MDH-CS only considered equilibrium conditions and

did not extend to transient simulations or incorporate reaction rate laws that would enable

the prediction of lag time. The KMC approach used previously to model transient conditions
is not compatible with an MSM due to the large number of rate phenomena (5002 to describe
transport from the MSM transition matrix alone) 41,121–124. Alternatively, finite difference

methods can be utilized to numerically solve a system of ordinary differential equations

derived from the MSM transition matrix, combined with the rate laws describing the kinetics
at each active site 42–45. Solution of this system of equations enables the prediction of

intermediate transport dynamics and can provide information on the system’s transient

behavior, such as lag time.

Here, we study the transient behavior of MDH-CS computationally by calculating the

cascade lag time from a finite difference model. The model produces a time record of OAA

and CoA concentrations, which are used to extrapolate the lag time after steady state is

46

achieved. Recombinant and mutated MDH-CS are considered using the MSM transition
matrix and basin definitions of Xie et al.26. We also utilize the model to study the transport

along individual transport paths and estimate their unique lag time. Finally, we explore the

implications of reversible reaction at MDH on the recombinant complex under channeling and

free enzyme conditions. This approach provides further insight into the cascade’s transport
efficiency and allows for model results to be compared against experiment16. An enhanced

understanding of efficient natural systems can then enable the design of reaction cascades for
novel chemical processes, such as upconversion of biomass or CO9,10, biosensors and biofuel
cells12,125, and medical diagnostics11.

2.2 Methods

Possible positions in the transport of OAA from the MDH source sites, SM 1 and SM 2 (Fig.

2.2), to the CS consumption sites, SC1 and SC2, are represented by 500 micro-states in a

Markov state model (MSM). The probability of the intermediate state of OAA transitioning
42,126. The

from state i to state j is described by the transition probability matrix, TP

transition probability matrix is combined with an assumed time scale to generate a set of

differential equations describing the mole balance for each state. The kinetic rate law is

used at each active site to incorporate the mole balance associated with the reaction into the

set of differential equations. The differential equations describing state occupancy by OAA,

combined with an assumption of zero initial occupancy, represents an initial value problem

that is solved using the initial value problem solver solve ivp from the Python package
SciPy127. The output of solve ivp is a prediction of state occupancy as a function of time.

2.2.1 State Transition Probability and the Network

The transition probability matrix, TP, was generated as described by Xie et al26. Briefly,
400,000 MD frames were summarized by the distance between the intermediate and the four

active sites, and the intermediate distance to the surface. These frames were clustered into

500 micro-states (or states, used interchangeably). The Python package PyEMMA was then

used to calculate the transition probability matrix, TP

128.

47

Figure 2.3: Connectivity of the 500 states in the Markov state network for (a) Recombinant
and (b) Mutant complex, as defined by the transition probability matrix, TP. Image gener-
ated using pyvis 130.

The 500 states are further divided into six basins. Network representations of the tran-

sition matrices for the recombinant and mutant complexes are shown in Figure 2.3. Only

connections between states where the transition probability is non-zero are shown. The nodes

are color-coded according to the basin to which they belong. The probability of remaining

in the same state is not represented. This graph was produced using the Python packages
NetworkX and pyvis129,130.

Of the six basins shown in Fig. 2.3, four represent active sites. The active site basins on

MDH are named SM 1 and SM 2, and the CS basins are named SC1 and SC2. A fifth basin

represents OAA desorbed from the surface. A final basin collects states that are neither in
the active site basins nor desorbed, denoted as “Surface Intermediate” 26.

2.2.2 System of Equations

The transition probability matrix, TP, is converted to a transition rate matrix, T , according
to126:

T =

TP − I
τ

(2.1)

where I is an identity matrix of equal dimension to TP and τ is a sufficiently small transport
time constant such that TP may be considered a stochastic matrix 126. Xie et al. determined

48

a sufficiently small transport time constant, τ , for MDH-CS to be 10 ps 26. Analysis of the

effect of time scale on lag time demonstrates that lag time calculations are unaffected by

a τ of up to 100 ns (Appendix A, Figure A.2). Here we assume τ = 100 ns to minimize

the time scale difference between surface transport and other phenomena without affecting

the lag time. This assumption increases computational efficiency and minimizes problem

stiffness, or the need for very small time steps to obtain a smooth solution, by decreasing

the difference between transition and reaction time scales. Using the transition rate matrix,

the mole balances for the set of states, θ, generally take the form:

dθ
dt

= T θ + rrxn + rads − rdes

(2.2)

The time-dependent occupancy array, θ(t), describes the probability that each state is

either occupied by OAA (0 < θi(t) ≤ 1) or vacant (θi(t) = 0) as a function of time. Initially,

we assume that there is no OAA in the system, θi|t=0 = 0.

The rrxn term in Eq. 2 represents the rate of OAA generation or consumption via

enzymatic reaction, described below. For nodes in the surface or desorbed basins, there is

no generation or consumption, so rrxn = 0. At MDH basin nodes, the production of OAA

is represented by a reversible reaction rate law. At CS basin nodes, OAA consumption is

controlled by an irreversible rate law. The terms rads and rdes represent the intermediate

adsorption rate to and desorption rate from bulk concentration, also described below.

The rates of production and consumption of OAA are calculated by rate laws defined at

each active site basin. At MDH, OAA can be both produced and consumed. The forward

rate of OAA generation, rMDH,f , follows bi-bi kinetics:

rMDH,f =

kcat,MDHf CNAD+CL−malate
KiNAD+KL−malate + KL−malateCNAD+ + KNAD+CL−malate + CNAD+CL−malate
(2.3)

Because rMDH,f is independent of OAA concentration, this rate represents a zeroth-order

rate law in COAA. The maximum turnover frequency of malate to OAA is kcat,MDHf . The
Michaelis Menten constants of the two substrates are KL−malate and KNAD+. The enzyme
interaction parameter of NAD+ is KiNAD+. These parameters are specific to the forward

49

reaction at MDH and are defined in Table A.1 (Appendix A). Concentrations of substrate
L-malate and NAD+ are provided in Table A.2 (Appendix A).

The reverse reaction rate by which OAA is consumed at MDH, rMDH,r, is also defined

by bi-bi kinetics:

rMDH,r =

kcat,MDHrCNADHCOAA
KiNADHKOAA + KOAACNADH + KNADHCOAA + CNADHCOAA

(2.4)

The maximum turnover frequency of OAA back to malate is kcat,MDHr. The Michaelis

Menten constants of the two substrates are KOAA and KNADH. The enzyme interaction

parameter of NADH is KiNADH. These parameters are specific to the reverse reaction at

MDH and are defined in Table A.1 (Appendix A). OAA concentration is initially zero,

and accumulation is tracked over time.
In the experiments, the initial concentration of
NADH is zero, and NADH accumulation is not reported 16. For this reason, when modeling

experimental results, we assume a fixed CNADH = 0 µM .

The combined rate of all MDH sites is given by the difference between rMDH,f and rMDH,r,

and the contribution of each MDH state, accounting for the total enzyme concentration,

CMDH−CS, is given by:

rrxn,MDH =

rMDH,f − rMDH,r
CMDH−CSNMDH

(2.5)

where the total number of states in the two MDH active site basins is represented by NMDH.

The consumption of OAA at CS is described by an applicable rate constant determined

from bi-bi kinetics in Bulutoglu’s work and reduced to Michaelis Menten kinetics for constant

CaCoA

16. The resulting expression is:

rCS =

kcat,CS ∗ CaCoA
KaCoA + CaCoA

(cid:88)

θ

CS

(2.6)

where the maximum turnover frequency of OAA at CS, kcat,CS, and Michaelis Menten

constant, KaCoA, are provided in Table A.1 (Appendix A). The concentration of acetyl-

Coenzyme A (CaCoA) is provided in Table A.2 (Appendix A). Like the MDH states, the

50

Figure 2.4: Diagram of modified surface network and homogenous bulk interactions.

overall rate law is evenly divided amongst the CS states, numbering to NCS, as:

rrxn,CS = −

rCS
NCS

(2.7)

This work uses two approaches to describe the relationship between channeling and des-

orption. The first utilizes the network directly to account for the transition between the

surface and desorbed basins.

In this case, the rate terms for adsorption, rads, and des-

orption, rdes, in Eq. 2.2 are zero, as adsorption and desorption are already considered in

the transition rate matrix, T . The second approach considers a single, homogenous bulk

concentration instead of a desorbed basin. Interaction between the remaining surface net-

work and the bulk is determined based on surface-bulk terms in the transition matrix. This

modification is described in the diagram of Figure 2.4.

Variable adsorption scaling is considered, wherein rate terms for adsorption, rads, and

desorption, rdes, describe the state-specific interaction between the surface network and the

bulk. However, the adsorption rate constants are scaled by a factor, Kads. These terms are

calculated from the transition matrix such that:

rdes,isurface

= kdes,isurface

· θisurface

(2.8)

51

rads,isurface

= kads,isurface

(1 − θisurface

)

COAA,bulk
CMDH−CS

kdes,isurface

=

(cid:88)

jdesorbedi̸=j

Tisurfacejdesorbed

kads,isurface

=

1
Kads

(cid:88)

isurfacei̸=j

Tisurfacejdesorbed

(2.9)

(2.10)

(2.11)

In the MSM network model, the bulk concentration of OAA intermediate, COAA, desorbed,

is calculated from the occupancies of nodes in the desorbed basin, θidesorbed

:

COAA,desorbed = CMDH−CS

(cid:88)

idesorbed

θidesorbed

(2.12)

In the homogenous bulk model (Fig. 2.4), an additional material balance is required for the

bulk concentrations of OAA intermediate, COAA, bulk, which is dependent on the net rate

of desorption of OAA from the surface as:

dCOAA,bulk
dt

= CMDH−CS

Nsurface
(cid:88)

(cid:16)

isurface

rdes,isurface

− rads,isurface

(cid:17)

(2.13)

The production rate of CoA is taken by summing the OAA occupancy in all states of both

CS active sites. This leads to an additional material balance for product CoA equivalent to

rCS in Eq. 2.6:

dCCoA
dt

= rCS

(2.14)

The system of differential equations to be solved numerically consists of mole balances for

each of the 500 micro-states, a mole balance for CoA, and, in the case of the homogenous

bulk model (Fig. 2.4), a mole balance for OAA.

52

2.2.3 Numerical Solution

To solve the system with solve ivp, a function containing the system of differential equa-

tions, the initial θ matrix, and the time span are required inputs. The implicit Runge-Kutta
method, ‘Radau’, was chosen to overcome the problem stiffness efficiently 127. The system

of equations is solved by assuming zero presence of OAA or CoA as the initial condition.

A repository is available detailing the calculation of concentration profiles and lag times for
the homogenous bulk model with fixed adsorption scaling131.

2.2.4 Lag Time Calculation

Lag time is determined by fitting the final 30% of the CoA concentration profile to a linear

function. CoA concentration profiles were predicted for a time span of 40 seconds to ensure

that the system achieves steady state within the lag time window. When the CoA concen-

tration profile is curve fit with a slope, m, and y-intercept, b, the concentration of product

will be zero at the lag time. The lag time, tℓ, is thus calculated as:

The lag time, tℓ, can also be estimated from the transport efficiency, α, as 20:

tℓ = −

b
m

tℓ = tℓ0(1 − α)

(2.15)

(2.16)

where tℓ0 is the lag time of the uncomplexed, free enzymes. Details on the calculation of free

enzyme lag times are provided in the Supplementary Information. Eq. 2.16 was used in our
previous work to predict lag times from estimated transport efficiency.26 It is worth noting
that Eq. 2.16 is derived assuming that the enzyme kinetics are pseudo-first order 20. The

pseudo-first-order assumption is valid if the bulk OAA concentration is substantially smaller

than the applicable Michaelis Menten constant. For recombinant and mutated free enzymes,

the steady-state concentration of bulk OAA is on the order of 1 to 10 µM (Fig. A.1). The

applicable Michaelis Menten constant for both complexes is in a similar range (Table A.1).

Therefore, pseudo-first-order kinetics cannot be assumed for this system, and the approach

53

described in this work represents a potential improvement over Eq. 2.16.

2.3 Results and Discussion

The Markov state transition matrix of Xie et al. 26 was combined with the kinetic data
obtained by Bulutoglu et al. 16 to produce a transient finite difference model that predicts

the prevalence of OAA intermediate molecules on the surface of the MDH-CS complex and

in the bulk. The model also calculated transient product CoA concentrations that, once the

reaction reached a steady state, were used to calculate the lag time for the overall reaction.

The model is applied to conditions wherein both electrostatic channeling and desorption

occur. First, the Markov state network (Fig. 2.3), describing the transition between the

five surface basins and the desorbed basin, is applied with catalytic reactions occurring at

the active site basins. Then, the system is modified to treat the desorbed basin as a single

homogenous state with variable concentration. We study the effect of scaled adsorption rate

constants to mimic a variable adsorption equilibrium coefficient. We then study the flux

through individual transport paths for fixed adsorption rate scaling. Finally, we assess the

implications of reversible kinetics at MDH of the recombinant complex under channeled and

free enzyme conditions as a function of fixed NADH concentration.

2.3.1 Transport analysis using the Markov state model

To analyze intermediate transport in the complex system, we first employ the Markov state

model using the connectivity matrix, T , between surface and desorbed states, as illustrated

in Fig. 2.3 for the recombinant and mutant complexes. The resulting CoA and OAA

concentration profiles compared to experiment are shown in Fig. 2.5.

When the Markov transition matrix, T , is utilized directly to model transport, the cascade

displays near-perfect channeling (tℓ ≈ 0 s) for both the mutant and recombinant complexes.

Due to fast intermediate transport within the network via both the surface and desorbed

basins, the concentration of desorbed OAA (Eq. 2.10) remains near zero. Consequently, no

lag time is observed before achieving a steady state.

Previously, Xie et al. predicted an intermediate transport efficiency of 96% for the recom-

54

Figure 2.5: Citrate and OAA concentration profiles for the mixed channeling cascade using
the network and combined with reaction rate laws.

binant complex but only 47% for the mutant. Based on Eq. 2.16, the lag times associated

with Xie’s efficiency predictions are 0.24 s and 1.71 s for the recombinant and mutant, re-

spectively (Table 2.1). The unmodified MSM model matches the short lag times of the

experiment and Xie’s prediction for the recombinant complex, but disagrees with the long

lag times for the mutant. This is because transport via the bulk basin of the network does

not introduce any time lag. In contrast, Xie’s analysis assumed that any intermediate lost

to the bulk was lost entirely, introducing significant lag time. To bridge this disagreement,

we must, at minimum, add capacitance to the bulk solution that will slow transport through

the bulk. This might be expected to increase the lag time of the mutant without affecting

that of the recombinant.

2.3.2 Homogenous Bulk Model

To enhance differentiation of transport characteristics between the recombinant and mutant

complexes, we modify the network to replace the nodes of the desorbed basin with a single

homogenous bulk concentration, COAA,bulk. The desorption rate constants (Eq. 2.10) de-

scribe the interaction between the surface and bulk. Adsorption is considered via a variable

55

Figure 2.6: Recombinant and Mutant lag times for a range of Kads values. Experimental
lag times are marked by horizontal dashed lines and the optimized Kads value is marked by
a vertical line.

adsorption scaling (Eq. 2.11). Adsorption scaling factors from the network (Kads = 1) may

not capture the energy barrier for OAA to exit the surface to the bulk. Instead, we find a

fixed adsorption scaling factor to scale adsorption and capture a more realistic barrier in the

desorption of OAA.

2.3.2.1 Effect of Adsorption Scaling Factor, Kads

Lag times for adsorption scaling factor, Kads, ranging from 1 to 108 were calculated for both
recombinant and mutant complexes. Adsorption rate constants that are defined directly by

the transition rate matrix, T , are represented by Kads = 1. However, this results in low lag

times that do not exploit the channeling differences between the recombinant and mutant.

By increasing the value of Kads, as shown in Fig. 2.6, larger observed lag times and more

significant differentiation in lag time between the recombinant and mutant complexes are

observed.

The optimal value of Kads is determined by minimizing the total absolute error between
experimental and modeled lag times. The point of minimal error exists at a Kads = 3.4×104,

56

Figure 2.7: Citrate and OAA concentration profiles for the mixed channeling cascade using
the bulk concentration model for Kads = 3.4×104. The inset figure displays the concentration
curves at experiment time less than 1 second.

shown in Fig. 2.6 as a vertical line. This value is assumed in the definition of kads,isurface
(Eq. 2.11) for the remainder of the analysis.

In our previous work, surface adsorption between enzyme active sites was calculated from
an assumed desorption rate constant and an adsorption equilibrium coefficient7,27. There,
the adsorption equilibrium coefficient ranged from approximately 103 to 105. Practically, an

adsorption equilibrium coefficient scales the adsorption rate constant relative to desorption

rate constants in the same fashion as we consider here with a scaling factor.

Assuming a constant adsorption scaling factor, Kads = 3.4×104, recombinant and mutant
complex behaviors are predicted by OAA and CoA concentration profiles. The simulated

concentration profiles are compared to experimental CoA concentration profiles in Fig. 2.7.

Here, the predicted CoA concentration profiles, and resulting lag times, of the modeled

complexes are comparable to the experimental outcomes. The recombinant complex model

displays faster accumulation of CoA and a lower lag time than the mutant model. The

modeled recombinant CoA profile exhibits a steady-state accumulation rate that is similar

to the experimental result. However, the mutant’s modeled CoA profile only matches the

experiment initially (< 1s, Fig. 2.7 inset) but continues accumulating CoA faster than the

57

experimental counterpart. The steady-state rate of CoA accumulation for the mutant model

matches that of the recombinant, as expected because the rate of CoA accumulation depends
solely on the MDH reaction rate and is the same (rMDH = 3.1µM s−1) for both complexes.
The second key observation is that, in this case, the modeled mutant complex exhibits

a higher bulk OAA concentration than the recombinant complex (Fig. 2.7, dashed lines).

When operating with CNADH = 0 µM , MDH is operating with a constant velocity, rrxn,MDH

(Eq. 2.5). At these conditions, the steady-state bulk concentration of OAA is proportional
to the lag time20:

tℓ =

COAA, bulk|steady state
rrxn,MDH

(2.17)

Thus, higher bulk OAA concentrations, indicating a greater probability of desorption, lead

to longer lag times.

2.3.3 Lag Time Analysis of Recombinant and Mutant MDH-CS

The two extreme cases of (1) enzyme complex with perfect channeling, and (2) free enzymes

with no channeling, define the range of lag times that can be expected for each complex,

as detailed in Table 2.1. Based on efficiencies predicted by Xie et al., the recombinant lag
time should be close to zero, matching the experimental data 16,26. For the mutant complex,

lag time is expected to fall about halfway between the free enzyme and perfect channeling

values.

Experimental 16

Free Enzymes 16†

MSM model 26‡

MSM Network Model

Homogenous Bulk Model

Table 2.1: Recombinant and Mutant Lag Times.

Recombinant Lag Time, tℓ / s

Mutant Lag Time, tℓ / s

0.03 ± 0.01

2.5

0.24
0.025 ± 5e-13
0.11 ± 1e-10

0.88 ± 0.06

0.9

1.71
0.044 ± 4e-12
0.81 ± 5e-9

†Predicted from kinetic data ‡ Calculated using Eq. 16

When considering adsorption and desorption only as defined by the MSM, a pseudo-

perfect channeling case is found in which both complex models exhibit very low lag times

(Fig. 2.5). By scaling the adsorption rate constants to minimize absolute lag time errors,

58

a mixed channeling case is found. In this case, the model predicts a substantially lower lag

time for the recombinant than the mutant. This observation aligns with the findings of our
previous work and experiment 16,26.

2.3.4 Path Lag Time Analysis

The existence of four active sites, two for each enzyme, suggests four possible paths by which

OAA can traverse the surface. Considering OAA source sites on MDH, SM 1 and SM 2, and

the OAA consumption sites on CS, SC1 and SC2, the paths can be labeled as described

previously by Xie et al.: Path 1: SM 1 → SC1, Path 2: SM 1 → SC2, Path 3: SM 2 → SC1,

Path 4: SM 2 → SC2

26.

The present model includes reaction kinetics at each active site, allowing us to use the

model to estimate lag time and, therefore, reaction efficiency for each path. These can

be calculated by zeroing the reaction rate at inactive sites. Xie et al. did not calculate

individual path efficiencies directly, but the efficiencies of each OAA-producing active site

were explored. Here, the parallel operation of paths for sites SM 1 and SM 2 are assessed in

addition to the study of individual paths.

Transport of OAA from site SM 1 was predicted by Xie et al. to result in higher transfer
26. Therefore, Paths 1 and 2 are expected to be more efficient and

efficiency than from SM 2

have lower lag times than Paths 3 and 4. The lag times for each path are compared against

the overall models and the origination sites, SM 1 (Paths 1+2) and SM 2 (Paths 3+4) as

provided in Fig. 2.8. The results in Fig. 2.8 suggest minimal variation exists among the

individual paths of the recombinant, and the mutant is much more sensitive to the reaction

path. Although there is not much difference between the individual paths of the recombinant,

the parallel operation of Paths 1 and 2 results in slightly lower lag times than the parallel

operation of Paths 3 and 4. Opposite trends are observed in the mutant. Parallel operation

of two paths overcomes the decreased efficiency observed for individual Paths 2 and 4, as

shown by much higher lag times for Paths 2 and 4 compared to Paths 1 and 3.

59

Figure 2.8: Overall complex and path lag times. Error clouds represent the experimental
uncertainty in lag times as reported by Bulutoglu et al 16. Model lag time uncertainties are
determined by slope and y-intercept fit error propagation through Eq 2.15. Model results
show uncertainty in lag time (vertical axis) and are very small.

2.3.5 State Occupancy Observations

In addition to determining the time to steady state, the numerical model can predict locations

of OAA buildup and transport at steady state conditions. Channeling efficiency can be

further explored by comparing the amount of bulk, or desorbed, OAA to the amount in the

five surface basins. It is important to note that in these analyses, basins and paths can have

occupancy greater than unity because each basin has tens of nodes. Therefore, the total

occupancy of each basin can have a value up to the number of nodes.

2.3.5.1 Basin Occupancy

To understand the tendency of OAA to transport on the surface versus desorbing off the

surface near MDH active sites and adsorbing on to or near CS active sites, the proportionality

of basin occupancy can be compared. The steady-state occupancy of each basin for both

complexes is provided in Figure 2.9. Less OAA remains on the surface for the recombinant

complex than for the mutant complex. The mutant complex has a downstream turnover

60

Figure 2.9: Recombinant and Mutant Basin Occupancy by OAA at steady state conditions.

frequency, kcat,CS, half that of the recombinant (Table A.1), so there is expected to be more

OAA buildup on that complex. The higher amounts of OAA in CS sites than MDH sites

indicate efficient transport.

2.3.5.2 Path Occupancy

The relative channeling efficiency of paths can be elucidated by comparing steady-state total

surface occupancy, or path occupancy, and bulk OAA concentrations with only desired active

sites “turned on” (Eqs. 2.13 and 2.17).

The total surface occupancy of each path is shown in Figure 2.10a. For the recombinant

complex, the amount of OAA on Paths 1 and 2 is lower than on Paths 3 and 4 and is

consistent with the expectation that the M1 site is more efficient. The mutant complex has

the lowest occupancy on Path 1, followed closely by Path 3. Paths 2 and 4 have nearly

double the occupancy. The effect of the reaction path is more distinguishable for the mutant

and supports the hypothesis that Path 1 is the most efficient of the four paths. Paths 3 and

4 have higher occupancy than Paths 1 and 2, and indicate higher efficiency of SM 1 than

SM 2. The occupancy of dominant trajectories on each path is also assessed in Appendix A

(Fig. A.3).

61

Figure 2.10: Recombinant and Mutant Path Surface Occupancy at steady state by (a) OAA
and (b) Path Bulk OAA Concentration.

The bulk OAA concentration resulting from individual activation of each path (Fig.

2.10b) follows a similar trend to that of surface occupancy (Fig. 2.10a). For the recom-

binant, Paths 1 and 2 generate lower bulk OAA concentration than Paths 3 and 4. The

mutant generates comparable concentrations for Paths 1 and 3, which have lower bulk OAA

concentrations than either Paths 2 or 4. For all paths, substantially increased bulk OAA

concentration was observed for the mutant complex compared to the recombinant, which
aligns with previous desorption probability observations 26.

Combining the bulk OAA concentration observations with those of the surface occupancy,

the recombinant complex transforms OAA to CoA more efficiently than the mutant, indi-

cated by lower operating concentrations of OAA for the recombinant complex. Transport in

the recombinant complex is more sensitive to the origination site than the particular path.

Transport in the mutant is most sensitive to the OAA consumption site and is most efficient

with Path 1. The effect of the reaction path is most visible on the mutant, confirming Path

1 as the most efficient.

2.3.6 Reversible Kinetics at MDH

The impact of reversible kinetics at MDH is explored for the recombinant complex in chan-

neling and free enzyme conditions. A non-zero, fixed concentration of bulk NADH, CNADH,

is specified, representing the ambient condition. In Fig. 2.11, results for the channeled com-

plex are shown in comparison to the free enzymes as a function of bulk NADH concentration,

62

Figure 2.11: Steady state rate of CoA production as a function of fixed NADH concentration
for reversible kinetics at MDH. (a) Keq = 3.3 × 10−2, derived from data of Bulutoglu et
al16. (b) Keq = 3 × 10−5, derived from data of Guynn et al 119. (c) Amplification factor of
CoA generation rates in channeled versus free enzyme conditions.

CNADH, for two values of the NADH equilibrium coefficient, Keq, defined as:

Keq =

kcat,MDHf
kcat,MDHr

(2.18)

The case of Keq = 3.3 × 10−2 (Fig. 2.11a), arises from the kinetic values reported by
Bulutoglu et al., as listed in Appendix A Table A.116. The case of Keq = 3 × 10−5 is taken
from data reported by Guynn et al119.

In the absence of NADH, represented by the extreme left of Fig. 2.11a-b, the channeled

and free enzyme systems produce CoA at the same rate. In the absence of NADH, equivalent

rates for channeled and free systems are expected since the reverse reaction at MDH is

inactive. At high concentrations of NADH, a constant amplification factor due to channeling

is observed (Fig 2.11c) that is independent of CNADH. This amplification increases with Keq,
being 1.8 for Keq = 3.3 × 10−2 and 4.3 for Keq = 3 × 10−5. At moderate values of CNADH,
reaction rates for both the complex and free enzymes decrease with increasing CNADH, and

the amplification factor increases from 1 to its high-concentration value. In the case of the

low value of Keq, however, the free enzyme rate decreases more rapidly, and a maximum

amplification near 10 arises at an intermediate NADH concentration (CNADH ≈ 0.2 µM).

This intriguing result is likely due to the interplay between electrostatic channeling and the

63

detailed kinetics at each active site, and will be the subject of future study. In all cases,

however, the channeled system’s production rate exceeds that of the free enzyme system,

demonstrating the impact of channeling in enzyme systems with reversible kinetics.

2.4 Conclusion

In this work, the dynamics of intermediate channeling in the malate dehydrogenase cit-

rate synthase (MDH-CS) complex is modeled using a finite difference approach applied to

a Markov state model. The model predicts lag times for the recombinant and mutant com-

plexes of approximately 0.11 seconds and 0.81 seconds, respectively, by assuming a fixed

adsorption scaling. These lag times are comparable to their experimental counterpart. Ad-

ditionally, we explore details of the transport mechanism on the surface through analysis

of specific active sites and paths. The recombinant complex was found to have comparable

efficiency across the individual paths as marked by similar lag times of the origination site to

the overall and amongst individual paths. Transport through the mutant was insensitive to

the OAA origination site but exhibited more significant variation in efficiency in the separate

paths.

Analysis of surface occupancy identifies Path 1 as the preferred and most efficient path.

The recombinant complex’s greater efficiency than the mutant is due to its higher preference

for Path 1. In contrast, the mutant has a more distributed preference across the paths and an

observed tendency to leak OAA. Finally, considering reversible kinetics at the MDH active

site, the complex employing electrostatic channeling is demonstrated to be significantly less

sensitive to the presence of reaction product NADH than the free enzyme.

The transient model reported here quantitatively relates complex structures to interme-

diate transport within cascade reactions. While the connectivity network established by the

Markov state model enables the model, building the MSM is computationally intensive. Ex-

panding these methods to other biocatalytic cascades may elucidate the generalized behavior

of similar systems, allowing for dynamic and rapid prediction of reaction cascade dynamics.

64

Chapter 3

Quantum Chemical Insight into

Electrochemical Upgrading of Bio-oil

3.1

Introduction

With threats of global warming and diminishing supplies of fossil fuels, there exists a strong
need for sustainable energy and chemical production.1 Although alternative energy tech-
nology has advanced, society continues to rely on fossil energy supplies.47 Alternatively,

renewable biomass represents a sustainable resource for organic fuels and value-added chem-

icals. The use of pyrolytic bio-oils derived from plant-based lignocellulosic biomass is a
developing option for replacing fossil fuels. 4,47 However, bio-oil, made from lignocellulosic

biomass via fast pyrolysis, is highly oxygenated, which leaves it chemically unstable and low
in energy density.48 Recent experimental work and system analysis suggests that electro-

chemically upgrading bio-oils by hydro-deoxygenation can increase the stability and energy
density of pyrolyzed bio-oil. 49–51,57,61,63,132–141

A sufficient understanding of electrochemical deoxygenation for proper control and opti-

mization requires overcoming two major hurdles. The first lies in the chemical complexity

of bio-oil, comprising roughly 25% water and 25% lignin fragments, with the remainder a

mix of aldehydes, carboxylic acids, carbohydrates, phenols, furans, alcohols, and ketones.

Several deoxygenation/hydrogenation reaction paths can proceed simultaneously and inter-
actively in these systems, leading to a broad selection of possible products. 61,85,132,142,143

Secondly, the electrocatalytic interface plays a large role in product reaction rate and se-

lectivity. The interplay of these factors requires deep mechanistic examination, which, in

65

conjunction with validating experiments, can be exposed through atomistic and quantum
simulations.67,83,93,142,144–148

Our previous experimental work investigated the electrochemical conversion of various

bio-oil constituent molecules on both single- and mixed-metal electrocatalysts. Oxygenated

aromatics, furans, and mixtures were electrocatalytically upgraded on Pt and/or Ru cat-
alysts supported on activated carbon. 55,59,149 Mixed-metal Pt-Ru catalysts were found to

improve performance as compared to single-metal catalysts in the conversion of a variety of

oxygenated aromatics to cycloalkanes. Specifically, a 75/25 wt% Ru/Pt catalyst supported

on activated carbon cloth (Ru/Pt/ACC) produced yields of cyclohexanes greater than 50%
and nearly 100% conversion of the associated oxygenated aromatic.59 Similarly, furfural

(FF) conversion to tetrahydrofurfuryl alcohol (THFA) was favored under mildly acidic con-
ditions with moderate yields.55 However, low carbon mole balances were observed under

all conditions, an indication of significant side reactions involving FF, intermediate furfuryl

alcohol (FA), and THFA. Nevertheless, by also studying the reduction of 2-methylfuran

(MF), FA, and 2-furoic acid, the activity of a Ru catalyst toward furan ring saturation was

demonstrated.

Studies of single molecules informed a recent study of a binary mixture of 4-propylphenol
(4-PP) and FF on a Ru/Pt/ACC electrocatalyst.149 Here, 4-PP ECH was performed in the

presence of possible inhibitors under a constant current of 60 mA and 50°C. When stud-

ied individually, 4-PP produces two major products: 4-propylcyclohexanol (4-PCOL) and
propylcyclohexane (PC), in yields of 70% and 20%, respectively.59 However, when processed

in a mixture with FF, product yields are both reduced to less than 10%. Conversely, the

yields of 4-PP products are relatively unaffected by the presence of other categories of bio-

oil constituent molecules. To understand the factors resulting in inhibition of 4-PP electro

reduction by FF and products, mechanistic understanding is required.

The previous work on experimental electrocatalytic hydrotreatment (ECH) of a binary

mixture of bio-oil constituents suggests that 4-PP proceeds to 4-PCOL and PC. At the same
time, FF proceeds to THFA via a FA intermediate on a Pt/Ru catalyst.149 These pathways
are consistent with other mechanisms studied in the literature.56,68,145,150 However, FF and
product molecules are known to dimerize and produce other products.53,54 The major side

66

products that have been studied include the production of methylfuran (MF) from FA along

with dimerization of either FF or FA to form furoin (FUI) or hydrofuroin (HF). In the

work of Kasad et al., it was suspected that dimers fouled the catalyst surface so mechanistic

studies should also consider these byproducts.

Density functional theory (DFT) calculations of various aspects of electrochemical hy-

drotreatment have been conducted for various single-molecule bio-oil constituents and cata-
lysts. Phenolics 56,57,66,79,83–85,135,140,150–153 and furans49,59,67,102,137,138,142,145,148,154–159

are popular models given their prevalence in bio-oil and the vast array of value-added chem-
icals that can be produced52,53. Owing to the high activity of platinum electrocatalysts,
the Pt(111) surface is a common choice for these models 56,84,85,146,160–163. More recently,

ruthenium and non-precious metal catalysts such as copper and nickel have been consid-
ered in DFT models.50,51,58,62,83,84,135,142,144,155,156 Platinum and ruthenium have both

been shown to effectively catalyze the electrochemical hydrotreatment of bio-oil in DFT
studies.57,83,85,156

This work is a follow-up computational investigation to the previous experimental study

aimed at understanding the reaction mechanisms in the hydrotreatment of a furfural and

4-propylphenol mixture on a Ru/Pt electrode. This contrasts with previous works that

have focused on single compounds and single metal catalysts. We utilize quantum-scale

studies to explore the mechanisms by which model molecules compete in the ECH process,

analyzing reaction energetics and rate-limiting steps. Adsorption energetics predicted by the

model are validated against experimental results obtained by blocking of hydrogen under

potential deposition (Hupd). These mechanistic findings suggest possible explanations for

the experimentally observed outcomes.

3.2 Methods

The electrocatalytic hydrotreatment reactions of furfural (FF) and 4-propylphenol (4-PP)

were assessed computationally by calculation of adsorption energies and reaction energies.

The potential dependence of pathway reaction energies was additionally explored. Further,

the adsorption energies of some of the molecules are verified experimentally by blocked

67

hydrogen adsorption on a platinum electrocatalyst under hydrogen underpotential deposition

(Hupd) conditions.

3.2.1 Reaction Mechanism Energy Calculations

Adsorption and reaction energies are derived from optimized state energy calculations of

gas phase organic molecules in vacuum, bare catalyst surfaces, and molecules adsorbed on

catalyst surfaces. Each state underwent a geometric relaxation via the plane-wave pack-
age Quantum ESPRESSO (QE), version 7.2. 99,100 The Python package Atomic Simulation

Environment (ASE) was used to visualize the state structures and write input files for QE
calculations.88 Input parameters were established from reported models exploring the hy-

drotreatment of organic molecules on platinum catalysts such as benzene on Pt(111) by
Shayeghi et al.66 and furfural on Pt(111) bimetallic surfaces by Khouini and Vahedpour. 156
The Python package PyEnergyDiagrams is used to build the relative energy diagrams. 164

The relaxations minimized energy and force to a threshold of 0.0001 Ry and 0.001 Ry

per Bohr, respectively. Core electrons were described by ultra-soft pseudo-potentials from
the SSSP PBE Efficiency v1.3.0 library. 165 Valence electrons are described by Kohn-Sham

wave functions expanded to a kinetic energy cutoff of 35 Ry and a charge density cutoff

of 279.3 Ry. Cutoffs were selected based on the highest recommended value in the pseudo-

potentials for a given atom considered. The Brillouin zone was sampled for the gamma point

only for each calculation. The mixing factor for self-consistency was reduced to 0.5 for each

calculation. Forces due to van der Waals interactions were corrected through Semi-empirical
Grimme’s DFT-D3 method166. Gas phase species were relaxed with one molecule centered

in a 10 x 10 x 10 ˚A vacuum.

Regarding the bare surfaces, ruthenium is highly stable in the HCP(0001) facet, which
is the only surface considered for this metal.83 Lattice parameters a and c are set to 2.733
˚A and 4.314 ˚A, respectively. 167–169 Platinum, however, is studied in FCC(111), (110), and

(100) facets to account for its polycrystalline nature. The experimental lattice parameter of
3.92 ˚A is used for all platinum slabs.85,170 Bare slab and adsorbed state structures utilize a

4x4 metal surface with 3 atomic layers. A 10 ˚A vacuum is placed along the z-axis above the

metal surface for all calculations. After a bulk relaxation of the bare metal slab, the bottom

68

layer is fixed in adsorbed structure calculations to allow the top 2 layers to respond to the

adsorbate presence. Fermi-surface effects are treated with Methfessel-Paxton smearing and
a degauss factor of 0.1. 171

Adsorption energies are calculated by comparing the energy of the adsorbed state to that

of the bare slab and free organic molecule:

Eads = Ei+slab − Ei − Eslab

(3.1)

where Ei+slab is the energy of the relaxed adsorbed structure, Ei is the energy of the organic

in vacuum, and Eslab is the energy of the relaxed slab.

The energy of a surface reaction step is:

Erxn = EF S − EIS

(3.2)

where EIS is the energy of the initial state where the reactant molecule and any necessary

hydrogen atoms are adsorbed to the metal surface. EF S represents the final state of the

adsorbed product(s).

State energies are further corrected for non-zero temperature and pressure to yield Gibbs
free energies through vibrational frequency analysis.93 In doing so, the Gibbs free energy of

any state j relative to a chosen reference free energy, Gref , is:

∆Gj = Ej + EZPE − T ∆S + ∆H0→T − Gref

(3.3)

where Ej is the relaxed energy of the state such as those used in Eq. 3.1. EZP E is the

zero point energy of the state and represents the vibrational modes of the state at 0K. The

terms T ∆S and ∆H0→T represent the change in entropy and enthalpy, respectively, when

increasing the temperature to any point above 0K. The reference Gibbs free energies, Gref ,

for the FF and 4-PP pathways are:

Gref,FF = GFF(g) + 3GH2(g) + Gbare slab

(3.4)

69

Gref,4PP = G4PP(g) + nGH2(g) + Gbare slab

(3.5)

where n equivalents of H2 are considered to represent the maximum number of protons

required for the pathways. When 4PP is reduced to 4PCOL, three equivalents of H2 are

consumed, and 3.5 equivalents are consumed when 4PP is reduced to PC.

Vibrational modes and frequencies were calculated for the preferred adsorption configu-
rations via the QE package PHonon. 99,100 The thermochemistry module of ASE was utilized

to process the set of vibrational frequencies of each image and calculate EZPE, T ∆S, and
∆H0→T .88 Gas phase corrections are calculated with the IdealGasThermo function and ad-
sorbed state corrections are calculated by the HarmonicThermo function. The calculations

are executed with a pressure of 101,325 Pa and a temperature of 293.15 K, respectively.

3.2.2 Determination of Preferred Adsorption Configurations

Molecules can adsorb on surfaces in multiple configurations. For aromatic molecules, these

configurations can be defined by the placement of the ring center in hollow, atop, and bridge

configurations. With each ring configuration, the molecules are rotated about the axis per-

pendicular to the surface to consider the substituent group(s) placement. The configuration

of each molecule on each surface with the lowest adsorption energy is considered the pre-

ferred geometry. For example, FF is placed onto Pt(111) with 2 aldehyde configurations for

each ring configuration, resulting in 8 unique possible adsorption configurations. Of the 8

configurations, the one with the lowest adsorption energy is the preferred geometry. This

method was applied to FF and 4-PP on each facet. Adsorbed intermediates and products

were relaxed from the preferred configuration of the respective reactant.

3.2.3 Potential Dependent Reaction Thermodynamics

In electrocatalytic hydrotreatment, an applied potential impacts the rate of reaction steps

involving electron transfer. The impact of applied potential on the reaction thermodynamics

is assessed utilizing the computational hydrogen electrode (CHE) reported by Nørskov et
al. as an effective reference electrode. 90 Briefly, the chemical potential of a proton-electron

70

pair in solvent is assumed to be half the chemical potential of H2(g) at 0 V vs. RHE. At any
applied potential, URHE, the chemical potential of a proton-electron pair, H+ + e−, is:

µ(H+ + e−) = 0.5µ(H2(g)) − |e|URHE

(3.6)

where |e| is the elementary charge.

The reactions at the working electrode require an additional term to represent the elec-

trons consumed in electrochemical steps. Eq. 3.3 is modified to Eq. 3.7 account for this
where n is the number of electrons consumed up to and including state j 93,142.

∆Gj = Ej + EZPE − T ∆S + ∆H0→T − Gref − n|e|URHE

(3.7)

3.2.4 Reaction Mechanisms

A Langmuir-Hinshelwood mechanism is commonly assumed in organic electrochemical hy-

drotreatment (ECH), where an adsorbed organic compound reacts with a co-adsorbed hy-
drogen atom. 68 Here, we consider the energy of a hydrogen atom co-adsorbed onto each

facet with the organic to be equivalent to the energy of the proton-electron pair in solution
per the CHE. 93 The number of electrons consumed in these reactions is also considered as

the total consumed in a given state relative to the reference state. Generally, the relative

energy of a given organic adsorbed state, j, is represented by:

∆Gj = Gj + nH2

GH2(g) + G∗ + n|e|URHE − Gref

(3.8)

and GH2(g) represent the number and free energy of H2 molecules that have
where nH2
not yet been consumed from the reference state. G∗ represents the free energy of bare slab

images, which is only present before adsorption or after desorption of the organic. The

mechanisms for FF to THFA, 4-PP to 4-PCOL, and 4-PP to PC, accounting for bare slabs

and proton-electron pairs per the CHE, are summarized below.

71

The net reaction for the reduction of FF to THFA is:

FF(g) + 6 (H+ + e−) + ∗ −−→ THFA(g) + ∗

(3.9)

The elementary reaction steps for FF reduction to THFA are:

FF(g) + ∗ −−→ FF∗

FF∗ + H+ + e− −−→ (FF + H)∗

(FF + H)∗ + H+ + e− −−→ FA∗

FA∗ + H+ + e− −−→ (FA + H)∗

(FA + H)∗ + H+ + e− −−→ (FA + 2 H)∗

(FA + 2 H)∗ + H+ + e− −−→ (FA + 3 H)∗

(FA + 3 H)∗ + H+ + e− −−→ THFA∗

THFA∗ −−→ THFA(g) + ∗

(3.10)

(3.11)

(3.12)

(3.13)

(3.14)

(3.15)

(3.16)

(3.17)

4-PP can either be converted to 4-PCOL or PC. To produce 4-PCOL, the net reaction

is:

4-PP(g) + 6 (H+ + e−) + ∗ −−→ 4-PCOL(g) + ∗

(3.18)

72

The elementary reaction steps are:

4-PP(g) + ∗ −−→ 4-PP∗

4-PP ∗ + H+ + e− −−→ (4-PP + H)∗

(4-PP + H)∗ + H+ + e− −−→ (4-PP + 2 H)∗

(4-PP + 2 H)∗ + H+ + e− −−→ (4-PP + 3 H)∗

(4-PP + 3 H)∗ + H+ + e− −−→ (4-PP + 4 H)∗

(4-PP + 4 H)∗ + H+ + e− −−→ (4-PP + 5 H)∗

(4-PP + 5 H)∗ + H+ + e− −−→ 4-PCOL∗

4-PCOL∗ −−→ 4-PCOL(g) + ∗

(3.19)

(3.20)

(3.21)

(3.22)

(3.23)

(3.24)

(3.25)

(3.26)

The mechanism to produce PC proceeds through a propylbenzene (PB) intermediate for

a net reaction of:

4-PP(g) + 7 (H+ + e−) + ∗ −−→ PC(g) + (OH)∗

(3.27)

The elementary reaction steps are:

4-PP(g) + ∗ −−→ 4-PP∗

4-PP∗ + H+ + e− −−→ (PB + H, OH)∗

73

(3.28)

(3.29)

(PB + H, OH)∗ + H+ + e− −−→ (PB + 2 H, OH)∗

(3.30)

(PB + 2 H, OH)∗ + H+ + e− −−→ (PB + 3 H, OH)∗

(3.31)

(PB + 3 H, OH)∗ + H+ + e− −−→ (PB + 4 H, OH)∗

(3.32)

(PB + 4 H, OH)∗ + H+ + e− −−→ (PB + 5 H, OH)∗

(3.33)

(PB + 5 H, OH)∗ + 2 (H+ + e−) −−→ (PC, OH)∗

(PC, OH)∗ −−→ PC(g) + (OH)∗

(3.34)

(3.35)

A decision tree method is utilized to predict the radical intermediates in the elementary

reaction steps (i.e.(FF +H) ∗). Here, adsorbed states for all possible hydrogenation locations

on a given ring are relaxed for the first hydrogen addition. For a given hydrogenation step, the

state with the lowest energy compared to the others is considered the most surface-stabilized

and, therefore, is the most thermodynamically likely. The most stable intermediate of a given

step is the basis for defining possible intermediates for the next step. The preferred paths,

comprising the most likely intermediates for each hydrogenation step, on each facet, are

visualized in Appendix B: Figures B9-B12 for the FF pathway and Figures B13-B16 for the

4-PP pathway. These pathways correspond to the data presented in the reaction energy

diagrams. Examples of the decision trees for the FF and 4-PP pathways on Pt(110) are

described in Appendix B, Figures B17 and B18. The energies of all possible intermediate

states are provided in Appendix B Tables B1-B3.

74

3.2.5 Hydrogen Underpotential Deposition (Hupd) Inhibition by

Organic Molecules

The adsorption of organic molecules on a Pt electrode can be experimentally measured by

interfering with hydrogen adsorption on the same electrode. The adsorption of organic

and hydrogen on a Pt electrode is studied using cyclic voltammetry (CV) and linear sweep

voltammetry (LSV) near the hydrogen reversible potential. The experiments yield estimates

of the free energy of adsorption for the organic compound, which can then be compared with

computed free energies that account for solvation effects expected to be present during the

experiments.

3.2.5.1 Materials

The platinum black electrode is prepared with a platinum sheet (Mateck). The platinum

black solution contains chloroplatinic acid hydrate (Sigma Aldrich), with lead acetate from

ammonium acetate (Sigma Aldrich) and lead sulfate (Sigma Aldrich) as additives for sta-

bility. An aqua regia solution, composed of 3:1 hydrochloric acid (Honeywell Fluka) and

nitric acid (Fischer Chemical), is used for electrode cleaning. Electrochemical experiments

utilized a bare platinum sheet counter electrode and an Ag/AgCl reference electrode in

saturated potassium chloride (Sensortechnik Meinsburg). Perchloric acid (0.1 M, Fischer

Chemical) is the electrolyte. Furfural (Sigma Aldrich), furfuryl alcohol (Sigma Aldrich),

and 2-methylfuran (Sigma Aldrich) are introduced to the electrochemical cell to assess the

adsorption free energy.

3.2.5.2 Electrode preparation

Platinum black was galvanostatically deposited on a 3 cm × 2 cm × 0.1 mm platinum

sheet. The platinum black precursor solution was prepared as 3.5 wt/wt% Chloroplatinic

acid hydrate in DI water. A lead acetate solution was added to the precursor solution in

a quantity of 0.005 wt/wt% for stabilization. The lead acetate solution was prepared from

2 moles of ammonium acetate per mole of lead sulfate in DI water. Before the electrode

was coated with platinum black, it was cleaned in two steps. First, the bare electrode

75

was washed in a boiling aqua regia mixture to ensure any previously deposited matter was

removed. Then, the electrode was cleaned electrochemically with 0.5 M sulfuric acid while

collecting a background cyclic voltammogram (CV).

3.2.5.3 Cell setup and CV parameters

Electrochemical experiments were conducted using a Bio-logic SP-50e Potentiostat with an

equally sized bare platinum sheet as a counter electrode and a Ag/AgCl electrode as the

reference. A background CV was recorded between 1.2 V and -0.25 V vs. Ag/AgCl at a

scan rate of 100 mV/s in 0.1 M HClO4. CVs were collected accordingly unless otherwise

noted. To determine the electrochemical surface area (ECSA), cyclic voltammetry (CV)

measurements were performed at scan rates of 2, 5, 10, 20, 50, 100, and 200 mV/s.

3.2.5.4 Pt Black Coated Pt Electrode

The platinum sheet working electrode was coated with platinum black through various cycles

totaling to 3 minutes of coating time. Chronoamperometric techniques were used for coating
at a constant current of -30 mA/cm2. The coating time variable is detailed in Table 3.1. The

3 electrodes were rinsed and patted dry between coating and collecting CV measurements.

Table 3.1: Pt Electrode Coating Schedule.

Coat

Time per Side (s)

Coat Time (s)

Total Time (s)

1

2

3

4

15

15

30

30

30

30

60

60

30

60

120

180

3.2.5.5 Hupd experiments

The heat and free energy of adsorption of an organic molecule on the platinum surface can

be determined electrochemically by observing hydrogen adsorption blocked by an organic

molecule in the window of hydrogen under potential deposition. Using the electrochemi-

cal cell setup, linear potential sweeps are collected after the introduction of various con-

centrations of organic compounds. Here, the potential was scanned from 0.2 V to -0.3 V

76

vs. Ag/AgCl at a scan rate of 2 mV/s. Before the introduction of organic, the electrolyte

is purged with nitrogen, and an initial cyclic voltammetry (CV) scan was run to verify the
potential window associated with hydrogen adsorption.94

3.2.5.6 Calculation of Free Energy and Heat of Adsorption

Experimental free energy and heats of adsorption of organic species were calculated by
methods previously reported 94,95. The fraction of Hupd inhibited, θ, is first calculated by
integrating the hydrogen adsorption curves obtained in the presence and absence of organic

species to achieve the charges associated with each case, QHupd
tively. The resulting value of θ is

and QHupd+Organic, respec-

θ = 1 −

QHupd+Organic
QHupd

(3.36)

The aqueous free energy of adsorption is related to coverage in a Frumkin adsorption
isotherm.94,95 This modification provides the advantage of linearly scaling the free energy

of adsorption with the organic coverage by the slope α, as described below. The fraction

of Hupd inhibited as a function of concentration is fit to the isotherm to determine the
saturation coverage, θsat, the free energy of adsorption at zero coverage, ∆Gaq
ads,θ=0, and the
linear scaling slope for free energy of adsorption, α.95

C
C0

=

θ
θsat
1 − θ
θsat





−∆Gaq
ads,θ=0
RT





α θ
θsat
RT

−

exp

(3.37)

The standard state concentration, C0, is assumed to be 1 M. The free energy of adsorption

at any evaluation coverage (∆Gads,θeval

) is:

∆Gads,θeval

= ∆Gads,θ=0 + αθeval

(3.38)

The gas-phase free energy of adsorption is calculated from the aqueous-phase free en-
ergy of adsorption through coverage-dependent adsorption equilibrium coefficients (Kgas
eq,θ,
Kaq
eq,θ) and the Henry’s constant (KH) of the organic in water.172 Henry’s constants used

77

in this analysis are adjusted to account for electrolyte through Setschenow’s law.173–175 By

assuming ideal gas behavior, Henry’s constant of volatility, KH, in units of bar/M, becomes
dimensionless when divided by RT. 95,172 Dimensionless and electrolyte-corrected Henry’s

constants for each organic are reported in Table 3.3.

Kgas

eq,θ =

Kaq
eq,θ
KH /(RT )

The free energy of adsorption is related to the equilibrium coefficient.

Keq,θ = exp

(cid:19)

(cid:18)−∆Gθ
RT

(3.39)

(3.40)

With this equivalence, the value of ∆Ggas

ads,θ is calculated below and used with α to predict

the gas-phase free energy of adsorption at an evaluation coverage (Eq. 3.38).

∆Ggas

eq,θ = ∆Gaq

eq,θ + RT ln(KH /RT )

(3.41)

The gas-phase free energy of adsorption, calculated from experiments, can then be di-

rectly compared to the same value obtained from DFT calculations.

3.3 Results and Discussion

Previous work on experimental electrochemical hydrotreatment (ECH) of a binary mixture of

bio-oil constituents suggests that 4-propylphenol (4-PP) proceeds to propylcyclohexane (PC)

and 4-propylcyclohexanol (4-PCOL). Furfural (FF) proceeds to tetrahydrofurfuryl alcohol
(THFA) via furfuryl alcohol (FA) intermediate on a Pt/Ru catalyst. 149 However, FF and
product molecules are known to dimerize and produce other products.53,54 The main side

products that have been studied include the production of methylfuran (MF) from FA along

with dimerization of either FF or FA to form furoin (FUI) or hydrofuroin (HF). Previous

work suggests that dimers were produced, so the adsorption behavior of these byproducts

is considered. However, the production of THFA is the main pathway observed by Kasad

et al. and is the only reaction pathway fully explored in this work. A reaction network

78

Figure 3.1: Proposed reaction pathways from (a) 4-propylphenol (4-PP) and (b) furfural
(FF).

summarizing the species studied in this work is shown in Figure 3.1.

3.3.1 Adsorption Energies by DFT

Upon computing the energies of the adsorbed reactants, FF and 4-PP, the preferred con-

figurations on each surface were determined based on the minimum energy obtained for a

range of possible configurations on four metal surface facets. These minimum-energy con-

figurations are shown in Figure 3.2. The remaining non-preferred adsorption configuration

images are provided in the Supplementary Information. Configurations of product species

are relaxed from the configuration of their respective reactant. The atomic positions of the

optimized reactant, stable intermediates, and side product adsorbed states, along with the
optimized bare slabs, are available in a repository.176

FF molecules prefer a hollow or bridge-type configuration on each of the four surfaces
shown in Fig. 3.2, which is consistent with previous models of FF on Pt(111)102. Similarly, 4-

PP prefers a hollow-type configuration on Ru(0001), Pt(110), and Pt(100), though it prefers

an atop-type configuration on Pt(111).

The adsorption energies of the reactants, stable intermediates, and products are reported

79

Figure 3.2: Preferred adsorption configurations of furfural and 4-propylphenol on platinum
and ruthenium surfaces.

in Figure 3.3 for the four facets studied. Both the electronic adsorption energy and free en-

ergy of adsorption are presented. The electronic adsorption energy accounts for the change in

chemical potential of the three states, capturing the implications of the surface - organic in-

termolecular interaction. The free energy of adsorption adds the implications of temperature

on the intramolecular interactions to the electronic adsorption energy.

Comparing the electronic adsorption energies of preferred configurations of species in

the proposed pathways provided in Fig. 3.3a, we find that 4-PP binds more strongly to all

surfaces than FF. However, on Ru and Pt(111), the reactants have comparable electronic

binding strengths. This suggests competitive intermolecular surface-organic interactions that

can cause competitive adsorption between FF and 4-PP and potential inhibition of active

sites by either reactant.

On Ru(0001) and Pt(110) surfaces, intermediate FA exhibits weaker binding than FF or

4-PP at the electronic level. This may explain reclaimed 4-PP product yields in the presence

of FA instead of FF. The binding of FF and 4-PP is competitive and strong on Ru and

Pt(110), whereas FA binds less strongly on these facets. This computational observation is

consistent with the experimental observation that FF inhibited 4-PP ECH more than FA,

which suggests that Ru(0001) and Pt(110) are key facets of this process. Finally, intermedi-

ates and products with a greater degree of hydrogenation bind more weakly to the surface

80

Figure 3.3: Adsorption electronic and free energies of (a,b) molecules involved in the proposed
FF and 4-PP pathways and (c,d) other FF products on platinum and ruthenium surfaces.

than the reactants. The carbon atoms of THFA or 4-PCOL and PC have bonds with more

hydrogens than their respective reactants, which reduces the availability of the carbon to

bind with the surface. Owed to this reduced binding availability, it is expected that the more

hydrogenated species would have a weaker adsorption energy.

From the free energy standpoint at 20°C (Fig. 3.3b), the free energies of adsorption of

the 4-PP molecules suggest stronger binding than the FF molecules. The contribution of

vibrational corrections (EZPE − T ∆S + ∆H0→T , Eq. 3.7) increase with molecular weight

(MW), thus further reducing the free energy of the larger 4-PP molecule as compared to

the smaller FF. However, it is worth noting that both pathways have more negative free

energies on Pt(100) and Pt(110) than the other facets. This is more apparent from the free

energies (Fig. 3.3b) than the electronic energies (Fig. 3.3a). With vibrational corrections,

the adsorption energies of FF and 4-PP are also closest on Ru compared to the Pt facets.

Thermodynamically, the adsorption of these species on Pt and Ru is favorable, with the same

facet comparison trends observed in both electronic and free adsorption energy calculations.

While the free energies suggest that the adsorption of 4-PP is more favorable, this effect could

be negated by more competitive reaction thermodynamics in the FF pathway, as discussed

81

in the Reaction Thermodynamics section.

The adsorption of additional intermediates and products of FF hydrogenation were con-

sidered, as shown in Fig. 3.3c-d. These include methylfuran (MF) and dimerization products

furoin (FUI) and hydrofuroin (HF). MF, a possible hydrogenation product of FA, exhibits

stronger binding than THFA at the electronic level (Fig. 3.3c) and is comparable when vi-

brationally corrected (Fig. 3.3d). Due to this, FA may be converted to MF instead of THFA.

Since Kasad did not observe MF experimentally, we do not further consider MF production

here. Both dimers (FUI and HF, Fig. 3.3c-d), bind most strongly on all surfaces, likely

owing to their high MW. This suggests that dimerization may also be a possible inhibition

mechanism in this reaction system. The presence of such dimers was not measurable by

Kasad. Therefore, the side reactions producing FUI and HF are not considered below.

3.3.2 Reaction Thermodynamics

The thermodynamics of the FF to THFA and 4-PP to 4-PCOL and PC reactions were

assessed and compared. First, an energetic decision tree method developed in this work

determined the intermediate states for each hydrogenation step between the stable interme-

diates. Such intermediate states differed primarily in the position of hydrogen addition. The

intermediate state with the lowest energy was considered the preferred intermediate. The

pathways were compared on each Pt and Ru facet. Then, the potential was varied to assess

the implications of potential dependence on the reaction thermodynamics.

3.3.3 Proposed Pathway Reactions

In addition to binding energies of key species along both FF and 4-PP pathways, the ener-

getics of each hydrogenation step were probed. The energy landscapes in Figure 3.4 comprise

the most favorable step determined by the decision tree. An example decision tree for the

FF to THFA and 4-PP to 4-PCOL pathways on Pt(110) is provided in Appendix B, Figures

B17 and B18. The energies and SMILES notation of all intermediate steps are also provided

in Appendix B, Tables B1-B3. Hydrogen presence is also accounted for through the CHE

method described in Eq. 3.8.

82

Figure 3.4: Relative free energy landscapes for the (a) FF to THFA pathway with Eref =
EFF(g) + E3H2(g) + E∗, (b) 4-PP to 4-PCOL pathway with Eref = E4−PP(g) + E3H2(g)
+ E∗ and (c) 4-PP to PC pathway with Eref = E4−PP(g) + E3.5H2(g) + E∗ on Pt(100),
Pt(110), Pt(111), and Ru(0001).

83

The FF to THFA pathway is most favorable on Pt(110) (Fig. 3.4a), as denoted by the

lowest energy at every hydrogenation step besides bound THFA. This coincides with FF

and FA binding more strongly to Pt(110) than to other surfaces. However, this energy

pathway is not strictly downhill, indicating barriers to this transformation. The relative

energy of the pathway continually decreases on Pt(111) and Pt(100) until the desorption of

THFA. Although both of these surfaces exhibit weaker binding with the three key species

than Pt(110), there are no positive reaction energies on Pt(111) or Pt(100). On Pt(110), the

reactions to generate FA and THFA both exhibit positive reaction energies. Positive reaction

energies indicate that larger activation barriers may exist for that step. To assess activation

barriers, a kinetic analysis with transition state barriers is a point of future studies.

On Ru(0001), the FF to THFA pathway exhibits higher state energies than on platinum

facets. The high state energies can be attributed to the larger zero point energies compared

to those of the platinum facets (SI, Tables S7-S10). A slight barrier exists in the formation

of the first radical. However, after the first hydrogenation step, the energy decreases with

each proceeding step. Although the state energies are lower on Pt facets, the pathway on

Ru(0001) would be competitive due to continually decreasing energy and a more negative

net reaction energy than on other facets.

Similar trends are observed for the 4-PP to 4-PCOL pathway on Pt surfaces. Like the

FF pathway, state energies of the 4-PP pathways are also lowest on Pt(110), and barriers

still exist for the pathway on this facet (Fig. 3.4b). The most significant difference for the

4-PP pathways is also in comparing the pathways on Pt(110) and Ru. On platinum surfaces,

4-PP is converted to 4-PCOL through the hydrogenation of the ring. In contrast, on Ru,

the hydroxy group is immediately cleaved from 4-PP, and then the ring is saturated with

hydrogen. For the 4-PP to PC pathway on Ru (Fig. 3.4c), reaction energies are initially

similar for the adsorption of 4-PP and the first two reaction steps. The third hydrogenation

step exhibits a barrier before the state energies rapidly decline for the remaining reaction

steps. Considering such reaction energetics, it is plausible that 4-PCOL is preferentially

produced on Pt and PC is produced on Ru(0001).

Like the competitive binding observed between FF and 4-PP, the reaction landscapes

have the lowest energy on Pt(110). Contrary to the trends in adsorption in which Pt(110)

84

exhibited the strongest binding, the reaction energies of both pathways are least preferable

on Pt(110). This is because, for both the FF and 4-PP pathways on Pt(110), there are the

most indications of barriers throughout the transformations, as denoted by upticks in energy

throughout the reaction. The FF to THFA pathway generally decreases on all surfaces,

whereas the 4-PP pathways on all facets exhibit barriers. While the state energies are

lower for the 4-PP pathways, there are more barriers to reaction than for the FF pathway.

Therefore, it is conceivable that the FF to FA reaction blocks the transformation of 4-PP to

4-PCOL on Pt and 4-PP to PC on Ru.

3.3.3.1 Potential Dependence

Applied potential also impacts the relative energies of the various states via the potential

term in Eq. 3.7. Potential-dependent reaction thermodynamics are predicted for potentials

between 0.5 VRHE and -1 VRHE, as shown in Figure 3.5.

The presence of a limiting potential, UL, or the potential in which the largest barrier is

equal to zero, can also help to identify where barriers may have a larger impact on the overall
thermodynamic favorability.177 This implies that if the applied potential is more negative

than the limiting potential (U < UL), the reaction pathway will no longer exhibit barriers.

Limiting potentials for the pathways on each facet are provided in Table 3.2.

Table 3.2: Limiting Potentials (in V vs. RHE) of FF and 4-PP Pathways on Pt and Ru
Facets.

Facet

FF Pathway UL

4-PP Pathway UL

Pt(100)

Pt(110)

Pt(111)

0.19

-0.09

0.14

Ru(0001)

-0.11

-0.14

-0.25

-0.10

-0.25

Regarding the FF to THFA pathways on each surface, the limiting potentials are pos-

itive on Pt(100) and Pt(111). The Pt(110) and Ru(0001) facets exhibit negative limiting

potentials of approximately -0.1 VRHE. The limiting potential for the 4-PP pathway is more

negative on all surfaces than for the FF pathway. The limiting potential of the 4-PP path-

way is most negative on Pt(110) and Ru(0001) at -0.25 VRHE. On all facets, the limiting

85

Figure 3.5: Potential dependant relative free energy landscapes for the two proposed path-
ways on (a and b) Pt(100), (c and d) Pt(110), (e and f) Pt(111), and (g and h) Ru(0001).

86

potentials of the FF and 4-PP are dissimilar, where the 4-PP potentials are 0.15 to 0.3 VRHE

lower than the FF potentials. The more negative limiting potential of 4-PP pathways would

require a more negative potential, or greater overpotential, to ensure thermodynamic favor-

ability. Consequently, a more negative applied potential will further drive the FF pathways,

resulting in more competition during electrode polarization.

Although Pt(110) and Ru exhibit the strongest binding with the reactants, the pathways

on these facets will require the most overpotential. This is due to the larger barriers on

these facets compared to Pt(111) or Pt(100). Regarding FF reduction, the only barrier

or positive reaction energy on Pt(111) or Pt(100) is the desorption of THFA. On Pt(110),

positive reaction energies are observed in the formation of FA and THFA. The FF pathway

on Ru exhibits a barrier to the formation of FF radical. The 4-PP pathways exhibit barriers

on each of the facets, which require a larger overpotential to overcome. The more negative

limiting potentials on Pt(110) and Ru can be attributed to the relatively large barriers in

the second and third hydrogenation steps.

3.3.4 Experimental Validation with Hupd Experiments

Estimates of adsorption energies for FF, FA, and MF were obtained experimentally by ob-
serving hydrogen adsorption inhibition by the organic presence on a Pt electrode. 94,95 First,

the clean electrode was used to produce a CV, which represents the normal hydrogen adsorp-

tion behavior. Then, varied concentrations of organic molecules were added to the solution

and cyclic voltammograms (CVs) were collected for each concentration. CVs are provided

in Figure B19 of Appendix B.

Using Eq. 3.36, the fraction of Hupd inhibited by an individual organic species, θ, is

calculated for varying bulk organic concentration, COrganic, as shown in Figure 3.6. The

lines in the figure represent the fit of Eq. 3.37. FF and FA exhibit similar fit curves in

which Hupd is completely inhibited by organic concentrations larger than about 1 mM. MF

completely inhibits Hupd at concentrations greater than 0.1 mM. Adsorption of hydrogen

is most sensitive to the presence of MF, followed by FA, and then FF. From the fit curves

in Fig. 3.6, the Gibbs free energy of adsorption is predicted for the aqueous phase and

correlated to the gas phase to probe the energy of adsorption in comparison to the DFT

87

Figure 3.6: Normalized organic coverage as amount of hydrogen blocked versus concentration.
The coverage is normalized to the saturation coverage determined for each molecule. Using
the coverage profiles from experiments, the parameters in Table 3.3 were fit for concentrations
of 1 µM to 1 mM.

calculations presented here.

3.3.4.1 Gibbs free energy of adsorption

For each molecule, the aqueous and gas phase free energies of adsorption for zero coverage,

the coverage dependence slope, α, the saturation coverage, θsat, and the Henry’s constant

are provided in Table 3.3. Figure 3.6 suggests that FF and FA exhibit similar behavior,

with MF binding slightly more strongly due to the complete inhibition of Hupd at lower

MF concentrations. This is supported by comparable free energies of adsorption on a clean

surface in the aqueous phase at approximately -20 kJ/mol. However, these trends do not

hold in the extrapolation to the gas phase where MF is predicted to adsorb to a clean surface

less strongly than that of FF or FA. The Henry’s constant is largest for MF, followed by FF 3

orders of magnitude smaller, then FA. This would cause the gas phase adsorption equilibrium

constant to increase the least for MF from the aqueous prediction and subsequently weaken

the free energy of adsorption.

88

Table 3.3: Constants for FF, FA, and MF Hupd Inhibition Curves.

Molecule

KH

FF

FA

MF

1.5E-4 ± 0.2
3.3E-6 ± 2E-3

0.27

∆Gad,aq,0

kJ/mol
-19.4 ± 0.2
-18.9 ± 0.7
-22.9 ± 1.5

∆Gad,gas,θ0

kJ/mol
-40.8 ± 0.2
-49.7 ± 0.7
-26.1 ± 1.5

α

θsat

-5.4 ± 0.6
-5.3 ± 1.7
-5.3 ± 1.9

0.93 ± 0.05
0.60 ± 0.05
0.36 ± 0.004

The DFT calculations are considered to be a coverage (θ) of 0.0625 to account for the

one organic molecule per 16 surface atoms. To appropriately compare the experimental

adsorption energies to the computational energies, the free energy and heat of adsorption at

θeval = 0.0625 must be evaluated. The free energy of adsorption for each molecule at the

evaluation coverage are calculated for the aqueous phase and extrapolated to the gas phase

to be compared with the computational free energies of adsorption on each Pt facet in Table

3.4.

Table 3.4: Experimental and DFT calculated Gibbs free energy of adsorption of FF, FA,
and MF at θeval = 0.0625.

Experimental

Experimental

DFT - Pt(100)

DFT - Pt(110)

DFT - Pt(111)

Molecule

∆Gad,aq,eval

∆Gad,gas,eval

∆Gad,gas,eval

∆Gad,gas,eval

∆Gad,gas,eval

kJ/mol
-19.7 ± 0.2
-19.3 ± 0.7
-23.3 ± 1.5

FF

FA

MF

kJ/mol
-41.1 ± 0.2
-50.0 ± 0.7
-26.4 ± 1.5

kJ/mol

-225

-272

-251

kJ/mol

-275

-303

-265

kJ/mol

-138

-226

-207

In the gas phase, the experimentally predicted free energy follow similar trends at the

evaluation coverage as in the case of a clean electrode, provided in Table 3.4. For the three

surfaces, the computational adsorption free energies do not align with the experimental

free energy prediction. Although, the binding strength trend is similar on Pt(110). Here,

and on Pt(110) where the molecules are predicted to adsorb most strongly compared to

other facts, FA exhibits the strongest binding, followed by FF, then MF. This opposes the

experimentally predicted aqueous phase free energies of adsorption, suggesting that MF

binds most strongly, followed by FF and FA. In the DFT calculations and the gas phase

experimental extrapolation, MF binds more strongly than FF with the exception of Pt(110)

DFT calculations.

89

Discrepancies between the computational trends and experimental outcomes for MF,

specifically, may be due to the higher volatility and hydrophobicity of MF than FF or

FA. Experimentally, higher volatility could result in evaporated MF that does not interact

with the electrode. Additionally, as the compounds are hydrogenated, they become more

hydrophobic. The increased hydrophobicity of an organic compound in an aqueous environ-

ment leads to a decrease in solubility, which has been shown to enhance the binding between
the organic compound and the surface. 178 This would explain the stronger binding of MF

than FF or FA observed experimentally in the aqueous phase. Furthermore, computational

free energies are also several hundred kJ/mol more negative than the experimental values.

The reasoning behind the discrepancy remains unclear. Still, theories suggest that additional

energies are associated with the displacement of water by the organic phase and the solvation
between the aqueous and adsorbed phases, which are not accounted for by Henry’s Law.95

3.4 Conclusions

Computational and experimental methods were utilized to assess and determine a mechanis-

tic explanation for the inhibition of the electrocatalytic hydrotreatment of 4-propylphenol

to propylcyclohexane in the presence of furfural on a Pt/Ru electrode. First, DFT meth-

ods were used to predict the adsorption energy and reaction free energies of the 4-PP to

PC and FF to THFA pathways on Pt(100), Pt(110), Pt(111), and Ru(0001). An electronic

energy decision tree determines the hydrogenation intermediate states of each path. Stable

intermediates and other intermediate states are considered in reaction energy assessments.

Finally, free energies and heats of adsorption are experimentally determined for FF, FA,

and MF through competitive binding with hydrogen underpotential deposition on Pt. The

aqueous phase results are correlated to the gas phase and 1/16th coverage to compare with

the computational free energies of adsorption.

The electronic adsorption energy of reactants 4-PP and FF are most competitive on

Pt(110) and Ru(0001) where 4-PP always binds more strongly to a given facet. Vibrational

frequency analysis enabled the correction of the internal energies of the adsorption energy

calculations to Gibbs free energy, which exhibits stronger binding of the larger molecules.

90

Namely, where 4-PP and FF were competitive, the free energy of adsorption favors 4-PP.

The free energies of reactions along both pathways are further assessed under varied applied

potential. The reactions for both pathways are most favorable on Pt(110) and competitive

on Pt(100). However, the FF pathway requires less overpotential to have favorable reaction

energies than the 4-PP pathway on each of the facets studied. Additionally, there exist more

possible barriers in the 4-PP pathways, as indicated by positive reaction energies, than in

the FF pathways. These could be elucidated by analysis of activation barriers and kinetics,

and is a point of future study.

Furfural is determined to proceed to THFA through an FA intermediate on all four facets

studied, as suggested by the previous experimental work. The 4-PP pathway proceeds to

4-PCOL on Pt, and is consistent with previous works. However, on Ru, the hydroxy group

is cleaved off of 4-PP to form propylbenzene, which then proceeds to PC. Due to this, 4-

PCOL is favored on Pt, but the PC observed in the experiment is likely to form on Ru. The

pathway reaction free energies and the adsorption energies of FF and 4-PP are competitive

on Ru. Due to this, it is thermodynamically likely that FF would block the adsorption and

reaction of 4-PP on Ru, resulting in a decrease in PC product.

Experimental free energies of adsorption on Pt support some of the conclusions deter-

mined through the computational assessment. MF, FF, and FA exhibit comparable binding

strengths in the aqueous phase, but discrepancies arise in extrapolating to the gas phase.

These can be attributed to the variations in Henry’s constant for each species, resulting in

different shifts in adsorption equilibrium. Experimentally, FA exhibits weaker binding than

FF in the aqueous phase, but stronger binding in the gas phase, which is consistent with

the computational results that suggest FA would bind more strongly than FF. Further, the

isotherms of FF and FA are similar and suggest that either molecule would completely block

Hupd at higher organic concentrations than MF. Quantitatively, the experimental and com-

putational results are dissimilar, which may be due to effects not accounted for in Henry’s

Law. Understanding the effects of solvation near the catalyst’s surface may correct for some

of this, and is a point for future studies.

Ultimately, this work sought to understand the complexities of interacting bio-oil molecules

during electrocatalytic hydrogenation processes. Some details have been elucidated which

91

indicate competitive activity between furans and phenols. However, processing a real bio-oil

would have many more molecules and is likely to involve several more interaction effects. In

addition to considering activation barriers and solvation effects, there exists a need to scale

up the model both in time or space and in the number of interacting reaction pathways. An

immediate expansion of this work involves the prediction of reaction rate constants from the

potential dependent activation barriers, which experiments utilizing a rotating disk electrode

could validate. By similar studies of other bio-oil constituents, it would become possible to

predict the behavior of a real bio-oil, but would be time intensive. Prediction of adsorption

energies and reaction barriers by high-throughput means would expedite the prediction of

reaction rate constants that enable larger-scale models of interacting constituent molecules.

92

Chapter 4

Kinetic Evaluation of Electrochemical

Upgrading of Bio-oil

4.1

Introduction

In Chapter 3, an evaluation of adsorption and reaction energies in the competitive elec-

trocatalytic hydrogenation (ECH) of furfural (FF) and 4-propylphenol (4-PP) on a Pt/Ru

electrocatalyst was performed. The mechanisms of FF to tetrahydrofurfuryl alcohol (THFA)

via furfuryl alcohol (FA) intermediate and 4-PP to propyl cyclohexane (PC) via 4-propyl

cyclohexanol (4-PCOL) intermediate were assessed on the fcc facets of platinum and an hcp

facet of ruthenium. The electronic adsorption energies of the species indicate competitive ad-

sorption mechanisms, and that dimerization of FF may further block the surface, inhibiting

the 4-PP reaction. Further, the reaction energies of the FF pathway are more thermody-

namically favored than the 4-PP pathway. Comparing each path on the facets studied, the

reactions are most favored on the Pt(110) surface, as marked by the low relative free energies

of the states. However, the FF pathway may be more favored due to fewer barriers along

the reaction coordinates than the 4-PP pathway.

Within the FF to THFA and 4-PP to PC pathways on each Pt and Ru facet (Chapter 3,

Figure 3.4), most paths experience at least one unfavorable reaction energy of positive value,

indicating locations of potentially high barriers. To computationally determine activation

barriers of reactions, the transition state (TS) must be determined for each reaction step.

Through DFT methods, the TS can be determined by a search of the reaction path to find

the highest energy state between an initial state (IS) and final state (FS). Some of these

93

methods include nudged elastic band (NEB), climbing image NEB (CI-NEB), and the dimer
method. 96,97 The elastic band methods search for a TS along a pathway between an initial

state (IS) and final state (FS). The subsequent images are connected via “springs” with

spring constants to set the forces that determine how stiff the pathway is, or how far apart

images are from each other. CI-NEB offers the advantage of optimizing the details of the
highest energy state, which NEB does not do.96 The dimer method moves a pair of images

consisting of the IS and an image closely related to the IS, called the “dimer”, along the
potential energy surface.97 The curvature of the potential energy surface is calculated with

the motion of the dimer. The transition state exists at the point of minimum curvature.

Although transition state search methods are effective, they can become prohibitively

computationally demanding for complex reaction systems. To overcome this, linear scaling

relationships have been developed to correlate the energy of a TS or activation barrier to the
IS or FS.98 This approach offers the advantage of quick estimation of transition state energy

where the relationship has been defined for a specific reaction type using a method like NEB,

CI-NEB, or the dimer method. Two methods of linear scaling relationships include Brønsted

- Evans -Polyani (BEP) and Transition state scaling (TSS). BEP relationships correlate the

activation energy to the reaction energy of a given step. TSS relationships correlate the TS

energy to the energy of the final state.

Linear scaling relationships are only viable if trends have been determined for the class of

molecules, bond break or formation, and catalyst to ensure that implications of these factors

are accurately captured. Initially, BEP and TSS relationships were primarily determined for

small molecule bonds, such as C-O bonds in CO2, COOH, and CO, and C-H bonds in ethane
and methane.98 More recently, relationships have been determined for bonds within more
complex molecules such as phenol and furan derivatives.101–103 Kanchan and Banerjee report

a TSS relationship for furan ring hydrogenation on Pt(111) and Ru(0001). Other surfaces and
ring-opening reactions were also assessed.101 Liu et al. also determined a TSS relationship
for hydrogenating FF’s aldehyde group to form FA on noble metal catalysts. 102 Finally,

Mao et al. determined BEP relationships for cleavage of the C-O bond and hydrogenation
of the ring for phenol. 103 There are limited studies of scaling relationships for 4-PP, mainly

because the propyl group is unreactive, as demonstrated in Chapter 3. Therefore, we will

94

use the scaling data for phenol here.

Thermodynamic evaluation of hydrogenation reactions provides substantial information

for understanding competing mechanisms. However, reaction kinetics may be a further cause

of competitive mechanisms and inhibition of certain pathways. Computationally, reaction

rate constants can be calculated from the free energy reaction barriers, per Transition State
Theory. 104,105 Similarly to the state energies, the potential dependence of activation bar-

riers and, by extension, rate constants may be calculated by applying Butler-Volmer type
kinetics.93

The work herein extends that reported in Chapter 3 by calculating facet-dependent rate

constants for the transformation of FF to THFA and 4-PP to PC. The activation free en-

ergies are first calculated from the initial and final state free energies of Chapter 3 using

previously reported linear scaling relationships. Then, zeroth-order reaction rate constants

are calculated for each reaction step on the three facets of Pt and the one Ru facet. Finally,

the potential dependence of the rate constants is predicted. These results lay the groundwork

for predicting transient product concentration profiles for the two pathways independently

and in competition.

4.2 Methods

The free energies of initial, intermediate, and final states are predicted for the FF to THFA

and 4-PP to 4-PCOL or PC reactions on Pt(100), Pt(110), Pt(111), and Ru(0001) in Chapter

3. The activation-free energies are predicted by combining these energies with linear scaling

relationships for transition state or activation barrier energies reported in the literature.

Transition state theory is applied to calculate the zeroth-order reaction rate constants for

each reaction step from the activation free energies. The activation energies of the pathways

on the four facets and the resulting rate constants are compared for a given pathway and

across the two pathways. A Butler-Volmer type expression is utilized to further predict the

potential dependence of these parameters by first determining an equilibrium potential for

each reaction step, then predicting potential-dependent rate constants.

95

4.2.1 Activation Barrier

As discussed in Chapter 1, the energy of a reaction is defined as the difference between the

energy of the initial state (IS) and final state (FS).

Erxn = EFS − EIS

(4.1)

The activation barrier of the reaction, Ea is the difference in energy between the TS, or the

highest energy state found along the pathway, and IS.

Ea = ETS − EIS

(4.2)

Activation barriers or TS energies are predicted by a combination of BEP and TSS linear

relationships. BEP predicts the activation energy, Ea, from reaction energy:

Ea = αErxn + β

A TSS relationship correlates the TS energy to the eneergy of the FS:

ETS = α′EFS + β′

(4.3)

(4.4)

Where α and α′ represent linear slopes and the βs are the y-intercept. The constants for

each reaction type are summarized in Table 4.1. Parameter errors were not reported, but

goodness of fit measures were provided, which are included in the Table.

Table 4.1: Linear scaling constants for FF and phenol hydrogenation, and C-OH bond
cleavage of phenol.

Reaction Type

Relationship Type

α

β

β′

Goodness of Fit

furan ring hydrogenation 101

FF aldehyde hydrogenation 102

phenol hydrogenation 103

phenol C-OH bond cleavage 103

TSS

TSS

BEP

BEP

-3.94

0.31

-1.00

0.64

1.226

1.01

R2 = 0.73

SE = 0.13 eV

R2 = 0.93

R2 = 0.96

α′

1.192

1.01

The phenol hydrogenation is a slight deviation from the equations reported above. Here,

96

the activation energy of hydrogenation is instead correlated to the adsorption energy of
hydrogen.103

Ea,Ph+H = αEads,H − β

(4.5)

Free energies along the reaction pathways are also considered from Chapter 3. For the

transition states, the correction of the respective FS is applied to the TS for simplicity.

4.2.2 Computational Predictions of Rate Constants

Once activation barriers are predicted using computational approaches or scaling relation-

ships, the kinetic rate constants can also be predicted. Transistion State Theory postulates

that a zeroth-order rate constant, k0, is related to the activation barrier free energy, ∆Ga,
at a given temperature, T , and the reaction specific equilibrium potential, U0.90,93,104,105

k0 =

kBT
h

exp

(cid:19)

(cid:18)−∆Ga
kBT

(4.6)

Where kB is the Boltzmann constant and h is Planck’s constant.

The zeroth order rate constants predicted here are independent of reactant concentration

and reactant concentration would increase linearly. As calculated here, a generation rate for

some product concentration, CPwould look like:

dCP
dt

= k0

(4.7)

However, non-zeroth order reactions are possible in experiment. To fit these conversions

to first or second order rate laws that are dependent on concentration, experimental insight

is required. Specifically, concentration profiles over time for generated products must be fit

to determine turnover of reactants and realistic reaction orders.

4.2.3 Potential Dependence

Similarly to reaction energies, the implications of variable applied potential on activation

barriers can be assessed. This is accomplished through application of Marcus theory and

97

Butler-Volmer kinetics where the energy of a transition state references the potential depen-
dence of the proton in electrolyte.93 The reversible hydrogenation reaction of some organic

species A is:

A ∗ + H+ + e−

kf
−−→←−−
kr

AH∗

(4.8)

Using the computational hydrogen electrode (CHE) method, the equilibrium potential,

U0, is a combination of free energies and the elementary charge, e.90,93:

U0 =

GAH∗ − GA∗ − 0.5GH2(g)
|e|

(4.9)

The free energy change of the activation barrier, ∆Ga, at any potential, U , is 93:

∆Ga(U ) = ∆Ga(U0) + βF (U − U0)

(4.10)

where F is Faraday’s constant and β is the symmetry factor, generally assumed to be 0.5

for highly reversible (i.e. fast) reactions.

To impose potential dependence for some potential, U , the rate constant is defined as a
function of the equilibrium potential and rate constant, U0 and k0.104 When U < U0, the
reductive, or forward, reaction is favored. The oxidative, or reverse, reaction is favored when

U > U0. The forward and reverse rate constants, kf and kr, are defined below:

kf = k0exp

(cid:18) −β|e|(U − U0)
kBT

(cid:19)

kr = k0exp

(cid:18) (1 − β)|e|(U − U0)
kBT

(cid:19)

(4.11)

(4.12)

where k0 is defined by Eq. 4.6 and is the same for both Eqs. 4.11 and 4.12. By assuming

that the reaction is symmetric (β = 0.5), the reverse rate constant is equivalent to the

forward rate constant.

98

Figure 4.1: Reaction Free Energy diagram for (a) FF to THFA on the four facets (b) 4-PP
to PC on platinum facets and (c) 4-PP to PC on ruthenium.

4.3 Results and Discussions

4.3.1 Reaction Free Energy and Activation Barriers

To predict kinetics computationally, the reaction free energies must first be extended to in-

clude the activation barriers. Activation free energy barriers (Ga) are predicted as described

in Eqs. 4.3-4.5 with constants for each reaction type in Table 4.1. Figure 4.1 presents the

relative free energy diagrams of the pathways at 20°C from Chapter 3, which are updated

with the transistion state energies, denoted by “TSX”.

Fig 4.1 suggests that the barriers are generally lower for the FF pathways than those of

4-PP on each surface. The two steps in FF hydrogenation to FA exhibit notably higher bar-

99

Figure 4.2: Activation Barrier Free Energies at 20°C and 50°C for (a,c) FF to THFA and
(b,d) 4-PP to PC on the four surface facets.

riers than the remaining hydrogenation steps. In the 4-PP pathways, barriers are relatively

consistent along the reaction on platinum and also large for most steps on ruthenium. A

comparison of activation free energies is provided in Fig 4.2 for temperatures of 20°C and

50°C, the two temperatures used in the experiments of Chapter 3 and in the previous work,

respectively.

The activation free energies of both pathways are on the order of 0 to 3 eV and are

comparable with others reported in literature calculated from traditional methods (i.e. NEB,
dimer method). 104,142,145 Comparison of ∆Ga values also confirms that the FF pathways

exhibit lower barries than those of the 4-PP pathways by approximately 1 eV. There is also

a sizeable decrease in the FF pathway barriers to nearly 0 eV for the hydrogenation steps of

FA. Conversely, the 4-PP barriers are relatively consistent for the platinum facets. However,

while the initial deoxygenation step of 4-PP on Ru has a higher barrier, there is a consistent

decrease during the hydrogenation steps on ruthenium. Although the barriers of the 4-PP

reaction on Ru are still higher than those on FF, there may be competition amongst the

various pathways due to the shift in favorability of 4-PP to PC on Ru. Finally, the barriers

are not substantially impacted by increase of temperature from 20°C to 50°C, suggesting that

100

Figure 4.3: Zeorth order rate constants at 20°C for (a) FF to THFA and (b) 4-PP to PC on
the four surface facets.

the thermodynamics of these reactions would be minimally impacted in that temperature

range. The remainder of the analysis will only be conducted at 20°C.

4.3.2 Rate Constants

Rate constants were further calculated from the activation barriers via Eq. 4.6. The con-

stants provided in Figure 4.3 are tabulated for each TS, as labeled in Fig 4.1, and facet

for the FF and 4-PP pathways, respectively. The FF pathway rate constants (Fig. 4.3a)
vary from a minimum of 5.4E-06 s-1 (Ru(0001) TS1) to a maximum of 1.8E24 s-1 (Ru(0001)
TS6). Most constants are between approximately 1E4 and 1E15 s-1. Due to the larger acti-

vation barriers, the rate constants of the 4-PP pathway are much smaller than those of the
FF pathway. In Fig. 4.3b, the minimum is as low as 7.7E-30 s-1 (Ru(0001) TS1) and the
maximum for platinum surfaces is 0.79 s-1 (Pt(111) TS6). Ruthenium sees a maximum rate
of 1.8E04 s-1 (TS4), but is an outlier due to the relatively small activation barrier for that
step. Most rate constants are smaller than 1E-10 s-1.

As in discussed in Chapter 3 and observing the intermediate state free energies of Fig.

4.1, the FF and 4-PP reaction free energies are comparable. The lower state free energies

of the 4-PP pathway than those of the FF pathway would initially suggest thermodynamic

favorability. However, combining the tendancy of the 4-PP pathways to increase in energy

101

along the reaction and the larger barriers, the FF pathway is predicted to be more thermo-

dynamically favored on the Pt/Ru catalyst.

The rate constants of the FF pathway are generally also predicted to be higher than

those of the 4-PP pathway. The minimum constant in 4-PP to PC is 24 orders of magnitude

smaller than the minimum in FF to THFA. The maximum constant in FF to THFA is also

more than 20 orders of magnitude larger than that of the 4-PP pathway. The median rate

constant across the facets is similarly larger for the FF pathway than the 4-PP pathway.

These comparisons suggest that in addition to thermodynamic favorability of the FF to

THFA path, it is also kinetically favored.

4.3.3 Potential Dependence

In the electrochemical hydrogenation reactions, there can be further implications from ap-

plied potential on the reaction pathways. First, the equilibrium potential of the reactions

were determined. A reaction’s equilibrium potential, U0, is the potential at which the re-

versible electrochemical reaction is in equilibrium. An equilibrium potential is calculated for
each reaction step. 93 At a potential of U0, the forward and reverse reactions proceed at a
rate of k0 s−1. These values are further used to predict potential dependent rate constants.
In Figure 4.4, the rate constants are provided as a function of potential for each reaction

step on the facets, as calculated by Eq. 4.11. The equilibrium potential and rate constant

of each step is presented as a point.

On Pt facets, U0 for FF reactions is generally around 0 VRHE with a handful of reactions

finding equilibrium around -0.5 VRHE. 4-PP reactions exhibit equilibrium around similar

potentials, but are slightly more negative than that of FF reactions. On Ru, most equilibrium

potentials occur near the same potential as the same reaction on Pt, but some steps exhibit

substantially more negative equilibrium potentials as low as -2 VRHE. This indicates that

around an applied potential of 0 VRHE, the reactions on platinum will be pushed forward

in their equilibrium while a more negative applied potential is required to have the same

effect on Ru. While the FF pathway reactions have a slightly more positive equilibrium

potential than those of the 4-PP reactions, equilibrium conditions are close for most steps

indicating that some competition may be expected between the pathways at moderately

102

Figure 4.4: Potential dependent rate constants at 20°C for (a,c,e,g) FF to THFA and (b,d,f,h)
4-PP to PC on the four surface facets.

103

Figure 4.5: Effect of potential on the zeroth-order rate constant of the rate limiting steps at
20C in the FF pathways (solid lines) and 4-PP (dashed lines) pathways on the four surface
facets. Equilibrium for each rate limiting step is marked as a point.

negative applied potentials. At any given potential, the rate constants are larger for FF

reactions, and may be sufficient to overcome the competition with 4-PP reactions near

equilibrium conditions.

The favorability of a given net reaction will be dictated by the rate limiting step (RLS)

within that mechanism. For the FF and 4-PP pathways on a given facet, the rate limiting

steps are the transition states that exhibit the lowest rate constant or slowest reaction. The

rate constants of rate limiting steps for the FF (solid lines) and 4-PP (dashed lines) pathways

on each facet are compared in Figure 4.5. On all of the facets, the FF pathway RLS are faster

than those of the 4-PP pathway by 10 to 20 orders of magnitude. If these reactions were

experimentally zeroth-order, this would indicate that, across a range of applied potentials,

the FF pathway is kinetically favored over 4-PP reactions.

104

4.4 Conclusions

The thermodynamic investigation on the inhibitionary effect of furfural presence on 4-

propylphenol ECH was further probed here by addition of activation barrier calculation.

Linear scaling relationships were utilized to make rapid predictions of the reaction barriers

in the hydrogenation of FF to tetrahydrofurfuryl alcohol and the hydrodeoxygenation of

4-PP to propylcyclohexane. The 4-PP pathway exhibits large barriers and positive reaction

energies. While the FF to THFA also has some positive reaction energies, the barriers are

lower of that than the 4-PP path. Additionally, the FF barriers generally decrease along

the reaction, whereas the 4-PP path barriers stay relatively consistent. Due to the lower

barriers and more favorable reaction energies, the FF to THFA reaction is predicted to be

more thermodynamically favorable than the 4-PP to PC reaction on the Pt/Ru catalyst.

Thermodynamic analysis was complemented by an analysis of kinetic rate constants.

Here, the zeroth order rate constants corresponding to each transition state were calculated

from the respective activation free energy. Generally, the 4-PP reactions exhibit smaller rate

constants than the FF reactions. The difference in rate constants between the two pathways
are as large as 20 orders of magnitude with those of the FF path being closer to 1010 s−1
and the 4-PP path constants are lower than 10−10 s−1. Therefore, the FF to THFA reaction

is predicted to also be more kinetically favorable than the 4-PP to PC pathway.

Through calculation of equilibrium potentials for each reaction step, potential dependent

rate constants and rate limiting steps (RLS) are determined. Most equilibrium potentials

occur near the range of 0 VRHE and -0.5 VRHE for the FF reactions. Equilibrium potentials

of 4-PP reactions are generally similar, but slightly more negative. For a given pathway on

a given facet, the rate limiting step is the transition state with the slowest rate constant

across the potential range of -2 VRHE to 0.6 VRHE. For both FF and 4-PP pathways, the

rate limiting step is either the first or second reaction step. Comparing RLS of FF and

4-PP pathways on a given facet, the FF pathways exhibit faster rates by up to 20 orders

of magnitude. This further suggests the kinetic favorability of FF reduction to THFA over

4-PP ECH.

This work lays the groundwork toward a dynamic model of the inhibition of 4-PP ECH

105

by FF ECH. The predicted potential dependent rate constants may be applied to a potential

dependent dynamic micro-kinetic model. To further scale up the model, dimensionality

due to non-zeroth order reactions would need to be considered. In line with this ambition,

reaction rates should be validated through experimental methods.

106

Chapter 5

Toward Validation of a New Bioink

for 3D Printed Cell-Laden Tissue

Scaffolds

5.1

Introduction1

Engineered biomaterials are a current focus in medical applications to enhance the success

of the treatments. One approach is in the development of high-performance materials to
treat bone degradation resulting from osteoporosis. 106 Traditional methods for bone tis-

sue generation take healthy tissue from elsewhere in the body to artificially produce new
tissue.179 This leaves the patient with further physical trauma in multiple locations. Alter-

natively, biomaterials may be developed that promote tissue generation without requiring
additional extraction from the patient’s bone.106 Suitable materials must meet several cri-

teria to be viable options for these applications, especially if the material will be implanted

in humans. Tissue-regenerative materials must exhibit biocompatibility, maintain stability

in physiological conditions, mimic the extracellular matrix with structure tunability, and
promote cell viability. 107–110 Collagen, and the cheaper alternative, gelatin, meet most of
the requirements as a biomaterial.108

While initially attractive, gelatin degrades at physiological temperatures (˜37C). 111 To

overcome stability constraints, gelatin methacryloyl (GelMA) was developed to enhance ma-

1This work was completed in collaboration with Victoria Toomajian, Anthony Tundo, Nureddin Asham-

makhi, Christopher Contag

107

terial stability by bonding methacrylate to the gelatin backbone in place of lysine groups.

Chemical modification of gelatin increases the melting point above physiological temper-
atures.111 GelMA is stable at physiological temperatures (37C)111, whereas gelatin melts
around 32C.180 Hydrogels of GelMA also take longer to experience the same degree of mass
loss as gelatin. 107,110 To address cell viability, GelMA was further modified by chemical link-
age to silver-doped bioactive glass (Ag-BG).112 Bioactive glass is a traditional component
in tissue-regenerative materials with appreciable cell viability properties.106 The addition
of silver has also been shown to increase the antibacterial properties of biomaterials.109,112

Chemically linked GelMA-AgBG (GAB) is therefore hypothesized to have superior cell via-

bility properties to GelMA alone.

In tissue engineering, a three-dimensional (3D) template, termed “scaffolds”, can help
guide and support the growth of tissue. 106 Additive manufacturing enables customized scaf-

fold design - making this solution more versatile. Some of these methods are fused filament
fabrication (FFF)109, polymer foam, and extrusion printing.106 Currently, extrusion print-
ing of bio inks is a prevalent approach to obtaining cell-laden scaffolds.181–185 Extrusion

printing enables control of scaffold structure while also allowing for an environment capa-
ble of cell viability in terms of media presence through the extrusion process. 183,186,187 In

extrusion printing applications, some substrates can also be further strengthened through

photo polymerization, or UV cross linking. In these cases, the polymer structures become

linked by a photo-initiator which creates a more complex matrix structure and a scaffold
with enhanced mechanical properties. 188

Previously, GAB was developed for 3D printing applications to leverage the photopoly-
merization capabilities of GelMA and antibacterial properties of Ag-BG. 110,188 In 2021,

Marsh et al. reported rheological and degredation properties of the GAB bioink in compar-
ison to inks of GelMA alone and a nanocomposite of GelMA and Ag-BG nanoparticles. 110

The viscosity, and storage and loss moduli determined for GAB suggest that the material is

a viable option for a bioink in 3D printing applications. The rehological parameters of GAB

were also higher than that of GelMA alone, suggesting that GAB will be more resistant to

shear forces and better suited to protect the cells than GelMA. Further, the GAB was shown

to swell less and retain more mass over time than the GelMA or nanocomposite. These char-

108

acteristics lead to the hypothesis that GAB will be a superior polymer for 3D bioprinting

applications. However, 3D bioprinting with a cell-laden GAB bioink was not reported.

Beyond generation of stable scaffolds, biomaterials must be assessed for biocompatibil-

ity. This is accomplished by tracking the viability of cells incorporated in to the bioinks

through extrusion printing and in the time following. In bone tissue regeneration applica-

tions, fibroblast (3T3) cells can be incorporated for validation purposes. Fibroblasts are

useful because they are very stable and have been engineered to express biolumiesence while

alive. Specifically, 3T3 cells express firefly bioluminescence (fLuc2), in which live cells will
emit light at a wave length of approximately 560 nm. 189,190 Luminesence is the result of the
fLuc2 catalyzed D-Luciferin oxidation in the presence of ATP from the cells and oxygen. 189

The D-Luciferin and cell media are added to the well of a printed scaffold immediately af-

ter printing to initiate the reaction. Radience measurements are collected at multiple time

points spanning from a few minutes post print to multiple days post print to assess the time

dependence of cell viability. The magnitude of the radiance indicates the number of live cells

within a scaffold as a direct correlation to the amount of ATP present.

This work aims to extend that of Marsh et al. through comparison of cell viability in

extrusion printing of GelMA and GAB against a control of blended GelMA + AgBG bio-

inks. GelMA and GAB materials were first synthesized and characterized as reported in

the literature. Then, extrusion printing parameters for the three bio-inks were optimized.

Upon printing parameter optimization, cell viability studies were executed with 3T3 cells

in GelMA, GAB, and blend bio-inks. GAB bio-ink is hypothesized to exhibit enhanced

cell viability during extrusion printing compared to GelMA and blend bio-inks. Enhanced

properties of GAB will exemplify its superiority as a novel biomaterial. Here, however, a

combination of rheological inconsistencies and material acidity resulted in a non-viable cellu-

lar environment across the GelMA, GAB, and Blend inks. Future work requires improvement

of material synthesis protocols to improve the ability to be reconstituted into an ink with

a pH closer to 7. Nevertheless, the work herein reports optimization of printing parameters

and cell viability studies for the prepared materials. Additional methods utilized in efforts

to overcome the challenges of the current materials are also reported.

109

5.2 Methods

5.2.1 Material Synthesis

Scaffold generation by 3D bio-printing (3DBP) uses two synthesized materials, GelMA and

GAB. The benefits of chemically linking AgBG to GelMA are elucidated by a standard

blend of GelMA and AgBG ink. GelMA and GAB are synthesized following previously
established protocols.110 Briefly, GelMA was synthesized by the addition of methacrylic

anhydride (MAA) to Type A, 300G bloom strength gelatin dissolved in DMSO. GelMA was

precipitated from DMSO using toluene in an amount of three times the DMSO volume. The

precipitate was rinsed three times and then dissolved in one liter of DI water for 24 hours.

Water and any remaining solvent were removed by lyophilization (i.e. freeze-drying). We
deviate from reported protocols110 by using a -84C lyophilizer instead of operating at -50C

for the materials used in these studies.

GAB synthesis adds Ag-BG sol steps to the GelMA protocol. Ag-BG sol is comprised of

methanol (MeOH), tetraethyl orthosilicate (TEOS), and triethyl phosphate (TEP) solvents.

Upon combination of solvent, solids are added in order: aluminum nitrate nonahydrate

(AlNT), silver nitrate (AgNT), and calcium nitrate tetrahydrate (CaNT). GelMA synthesis

through the addition of MAA is done separately during the Ag-BG sol process. Chemical

linker (3-Glycidyloxypropyl)trimethoxysilane (GPTMS) is added to the GelMA solution after

MAA is allowed ample time to react with gelatin. Ag-BG sol is added with water to the

GelMA solution, allowing a day for the reaction. The resulting GAB solution is precipitated,

rinsed, and lyophilized as previously described. The overall scheme is shown in Fig 5.1.

5.2.2 Material Characterization

GelMA and GAB materials are characterized via several means. The chemical structures of

each material are characterized by Fourier-transformed infrared spectroscopy (FTIR; Jasco

FT/IR-4600) and proton nuclear magnetic resonance (H-NMR; Agilent DirectDrive2 500
MHz NMR spectrometer)110. FTIR spectra were collected on reflected chemical bonds

present in the materials. H-NMR is primarily used to observe the degree of substitution of

110

Figure 5.1: GAB Synthesis by reaction of GelMA and GPTMS with AgBG solution.

lysine by methacryloyl groups. To collect the spectra, 50 mg of each material was dissolved

in 1 mL of deuterium oxide samples.

Microscopic morphology and distribution of elements are assessed by scanning electron

microscopy and energy dispersive spectroscopy (SEM-EDS). Samples were prepared for imag-
ing, as reported previously 110. Briefly, the solid materials first underwent graded ethanol

dehydration. Then, the solids were further dried by critical point liquid CO2. Once de-

hydration was complete, the solids were mounted and coated with Osmium. SEM (JEOL

6610LV) images were collected using a 5 kV beam voltage. EDS (Oxford EDS) spectra were

collected at a 20 kV beam voltage.

5.2.3 3D Printing and Bio-Printing

This project’s printing consists of two phases using an Allevi 3 printer. Phase 1 includes

parameter optimization for printing synthesized GelMA, GAB and a control blend of GelMA

with Ag-BG nanoparticles. To enhance the mechanical properties of the scaffolds, photo-

polymerization of the substrate material was introduced via a photo-initiator component

and applied UV light during the printing process. Allevi recommends lithium phenyl (2,4,6-

trimethyl benzoyl) phosphonate (LAP) as a photo-initiator, so Allevi GelMA is also printed

111

with LAP for validation purposes.1,191 Synthesized materials are printed with VA-086 photo
initiator, as in previous work.110 Phase two consists of incorporating cells into inks and

executing cell viability tests on printed scaffolds.

Bioinks were initially prepared as Allevi recommends.1 Briefly, 0.1 g/mL GelMA is sus-

pended in 0.1 M N-[2-Hydroxyethyl]piperazine-N-[2-ethanesulfonic acid] (HEPES) buffer

with 5 mg/mL photoinitiator. Upon consistent suspension, inks were collected in a capped

syringe and left to rest either at room temperature for at least 2 hours or placed on ice for 15

minutes. Before extruding, the syringes were given a minimum of 15 minutes to equilibrate

at 22C in the printer extruder. Initially, inks were printed using recommended and reported

values for details of the extrusion mechanism and print. Regarding the extrusion process,

tunable parameters include the velocity of the tip during printing, or print speed, temper-

ature of the extruder body which is held constant during prints, and the pressure applied

to extrude the ink. Print specific parameters include the height of each print layer, and

UV crosslinking intensity and time. As trials proceeded, other optimum parameters were

determined. Recommended, reported, and optimized parameters are compared in Table 5.1.

Table 5.1: 3D Printing Parameter Comparison.

Parameter

Allevi Recommendation 1

Reported 5

Optimized Parameters

Layer Height (mm)

Print Speed (mm/s)

Extruder Temperature (C)

Pneumatic Pressure (psi)

0.45

5

22

18-28

UV Crosslinking Conditions

15% intensity for 10s

Nozzle Gauge

Bioink Concentration (g/mL)

27

0.1

0.5

4

27

12

120s

27

0.1

0.5

10

22-27

5-15

45% intensity for 60s

27

0.15

To mimic an extracellular matrix, scaffolds are printed as 10 mm by 10 mm grids. In this

work, only one layer is printed with a height of 0.5 mm, as defined in Table 5.1. However,
many layers of the grid may be stacked to produce a 1 cm3 cube scaffold. The template in

Fig 5.2 was provided to the Allevi software as an additional input for printing.

In 3DBP trials, Fibroblast (3T3) cells were incorporated into the bio-inks in a quantity of
approximately 1.5×106 cells per mL of bio-ink. The solvent was 0.1 M HEPES. Cell viability

was assessed via bioluminescence measurements. The radiance was read via a Perkin-Elmer

112

Figure 5.2: Matrix template used for 3DP and 3DBP.

in vitro imaging system (IVIS). A the light was filtered to collect luminesence readings around
560 nm.189 Samples were imaged 5 and 10 minutes after D-luciferin addition immediately

after printing and at the 24-hour mark.

5.3 Results and Discussion

5.3.1 Material Characterization

Chemical structures of gelatin, GelMA, and GAB were characterized by FTIR and H-
NMR.110 NMR Spectra are provided in Figure 5.3a. The figure inset zooms in on the

location of the lysine peak around 2.8 ppm. The absence of the lysine peak around a chem-
ical shift of 2.8 ppm indicates complete substitution.110 Both GelMA and GAB lack lysine

peaks, indicating nearly 100% substitution. Additionally, a peak just above 3 ppm appears

in gelatin and is reduced in the synthesized materials. This peak is suspected to be proline
and hydroxyproline amino acids, which may also be involved in the reaction.192,193

Characterization of the materials by FTIR (Figure 5.3b) determines if solvents were re-

moved, confirms that GAB maintains the base GelMA structure, and confirms characteristic
peaks of Ag-BG sol in GAB.110,113 Some characteristics of DMSO and MAA appear most
prominently in both GelMA and GAB. Namely, two additional peaks around 1015 cm−1 and

113

Figure 5.3: (a) NMR Spectra of synthesized GelMA and GAB, and gelatin.

950 cm−1 appear in both materials that were not reported in the literature.110. This may

be indicative of insufficient solvent extraction during toluene precipitation and lyophiliza-

tion. Ideally, any residual solvent not extracted during precipitation would evaporate while

dissolving the precipitate in water to prepare for lyophilization. The presence of DMSO and

MAA peaks indicates that insufficient solvent extraction occurred during these syntheses.

It is, however, promising that the GAB has maintained the characteristic amide I-III peaks
near 1650 cm−1, 1500 cm−1, and 1450 cm−1, which indicates that the GelMA structure is
preserved.110 Additionally, there are hints of a broad peak around 800 cm−1 for GAB. That
peak represents a Si-O bond and is indicative of Ag-BG.113

The surface morphology and distribution of elements within the GelMA and GAB ma-

terials were assessed by SEM-EDS. As shown in Fig 5.4a, GelMA has pores in the material

visible at the micron level. The presence of the pores is crucial for dissolving or suspending

the finished solid material to produce a bioink. Conversely, the GAB does not have visible

pores at the 50 or 5 µm scales. The lack of pores may prevent the complete dissolution of

the GAB because pores enable more rapid incorporation of solvent into the bulk material.

Instead, a suspension of GAB particles in the HEPES buffer could be expected.

The EDS analyses of GelMA (Fig 5.4c&e) and GAB (Fig 5.4d&f) suggest that, in both

materials, elements are homogenously distributed to the micron level. The GelMA has a

homogenous distribution of C, N, and O, which is maintained with similar relative counts in

the GAB. Additionally, the GAB has uniformly distributed Si and Ag with noteable counts

114

Figure 5.4: SEM (a,b), EDS imaging (c,d), and EDS spectra (e,f) of synthesized GelMA and
GAB.

115

Figure 5.5: 1cm by 1cm squares 3D printed with optimized prrinting parameters and 0.1
mg/mL concentrations for (a) Allevi GelMA and (b) Synthesized GelMA bioinks.

of Si, specifically. This indicates a homogenous linkage of AgBG to the GelMA backbone.

5.3.2 Optimization of 3D Printing Parameters

Optimization of biomaterials in extrusion printing included trials of purchased and synthe-

sized GelMA at concentrations of 10 wt/v%. During printing trials, suspensions had low

viscosities which caused immediate release of syringe contents with applied pressures greater

than 10 psi instead of shaped prints. To compensate for liquid behavior, specific deviations

from printing parameters in the literature were made. Namely, a combination of low pres-

sures and faster printing speeds produced a more controlled shape print. Sample prints of

Allevi GelMA and synthesized GelMA at specified parameters are shown in Figure 5.5.

Printed Allevi GelMA served as initial validation on a handful of parameters such as

print speed, pneumatic pressure, and extruder temperature. However, the inks of synthesized

materials were less viscous than the purchased GelMA ink which resulted in low resolution

of prints. While the print in Fig 5.5a exhibits some spreading of the ink, the shape is overall

consistent with comparable wall widths. The wall widths of the synthesized GelMA print in

Fig 5.5b are not as consistent marked by the jagged inner square shape. Some synthesized

GelMA ink was also prematurely released in the center of the square shape, further indicating

116

low viscosity of the ink. To obtain greater control over the print resolution, concentration

and UV Crosslinking conditions were further optimized. Additionally, the purchased GelMA

produced a transparent bioink whereas the synthesized GelMA was more opaque. This

indicates that the synthesized materials do not completely dissolve, but are a suspension of

polymer particles in the HEPES buffer.

As shown in Fig 5.6a, a matrix was achieved by increasing ink concentration to 15 mg/mL

and UV intensity to 45%. These optimized parameters allowed the synthesized GelMA and

the GelMA + AgBG Blend control to print a mesh successfully. However, when the printing

temperature was also increased to 27C, the GAB material was not printable and continuously

clogged the needle tip due to inconsistent suspension. To overcome this, GAB inks were

produced with a 1:1 ratio of GAB and GelMA with the intent of improving the printability

of GAB without substantially affecting the cell viability properties. While extrusion of ink

was made possible for GAB in this mixture, these scaffolds did not have mesh resolution like

GelMA alone or the Blend control. Imaged of the most optimal prints of GelMA, Blend,

and GAB+GelMA are provided in Fig 5.6.

The printing issues for GAB bio-inks resulted from an inability of the material to be

reconstituted into the HEPES buffer. The remaining solids clogged the tip during printing.

The material’s pores can collapse when the frozen solid melts during the lyophilization pro-

cess. This results in a stiff plastic that becomes difficult to dissolve. Addition of HCl in

amounts between 10 and 48 µL per mL solvent was used to improve GAB dissolution. The

acid helped to some extent in dissolving the GAB, but was not a complete solution. A more

careful lyophilization of a fresh synthesis of GAB could produce a porous structure, but was

not possible within the time frame of this project. Printing and bioprinting proceeded with

the parameters detailed in Table 5.1 for GelMA, Blend, and GAB+GelMA mixture inks.

5.3.3 3D Bioprinting

3D Bioprinting (3DBP) describes conditions in which cell-laden inks are extruded. These

trials incorporated fibroblast (3T3) cells into the bio-inks, and initial 3DBP trials utilized

otherwise unmodified bio-inks. The HEPES buffer doubles as a pH indicator: A pink color

indicates approximately neutral conditions, but it turns yellow when the media is acidic.

117

Figure 5.6: 3D prints under 15 mg/mL concentration and 45% UV Intesity conditions of (a)
GelMA, (b) GelMA + AgBG Blend, and (c) GAB + GelMA Blend.

118

Figure 5.7: Initial 3DBP Samples of 3T3 Cells in a GelMA + AgBG Blend bio ink.

The media must be maintained around neutral conditions for cell viability. When acidic

media were detected, the inks were neutralized by adding sodium bicarbonate.

5.3.3.1 Initial 3DBP with Blend Bioink

Initial trials of 3DBP were done with the GelMA + AgBG blend bioink. The scaffolds

shown in Fig 5.7 are noticeably yellow, indicating acidic media. In these trials, the blend,

GelMA, and GAB+GelMA were all measured to have a pH of approximately 6. The 3T3
cells prefer an environment pH in the range of 7.0 - 7.6.194 No viability was recorded among

these samples, likely because of ink acidity.

5.3.3.2 3DBP with Neutralized Bioinks

To return the inks to a neutral pH, 1 to 5 mg increments of sodium bicarbonate were

added before cell incorporation. The pH was measured after each addition. Neutralization

continued until pH 7 was achieved, as measured by pH strips. For the GelMA and blend

inks, approximately 25 mg of sodium bicarbonate was required to achieve neutral conditions.

The GAB+GelMA ink required up to 50 mg of sodium bicarbonate. Cells were incorporated

upon neutralization, and the inks were prepared for printing.

Sample prints of neutralized inks are shown in Fig 5.8. As shown in Fig 5.8a&b, the

119

Figure 5.8: 3DBP Samples of 3T3 cells in neutralized (a) GelMA, (b) GelMA + AgBG
Blend, and (c) GAB + GelMA Blend.

120

GelMA and GelMA + AgBG blend prints are pink, but the mesh resolution was lost. There

is a finite window of time and temperature in which the gelling of the inks is optimal for
printing.1,110 The prrinting window may have been surpassed due to the processing required

for neutralization, cell incorporation, and cooling. This would result in a low-viscosity fluid

that would print like the samples shown. The GAB+GelMA blend performs consistently with

the acellular GAB trials in which the GAB does not suspend completely and clogs prevent

consistent printing. Like the GelMA and control blend, the GAB+GelMA was less viscous

after neutralization. The GAB+GelMA ink also still had a yellow hue, which indicates

a non-neutral pH. Although initial pH strip readings suggested a pH of approximately 7,

double the amount of sodium bicarbonate was required for neutralization compared to the

GelMA and blend inks. The buffing capacity of HEPES may have been overwhelmed by the

excess sodium bicarbonate once equilibrium was achieved.

5.3.4 Cell Viability

The cell-laden samples in neutralized ink were assessed for cell viability immediately after

printing and 24 hours later. The luminescence readings are summarized in Figure 5.9. Only

the GelMA samples express viability immediately after printing (Figure 5.9a-c). Some of the

cells survived the first day, but viability was substantially diminished after 24 hours (Figure

5.9d).

The lack of cell viability in these materials could be due to a combination of factors.

The first, and most apparent, is the acidic nature of the acellular inks. The materials

themselves could also be toxic due to residual toluene present from the synthesis process.

These factors alone can kill cells but also may contribute to non-optimal rheology for printing.

The inks were capable of achieving a gel-like consistency. However, if printing was delayed

or extended over four or so hours from the time that the ink was print ready, the viscosity

was notably decreased. Although the low viscosity of the inks required less than standard

printing pressures, the extrusion still exhibits a large force on the inks and could be an

additional contributor to the cellular demise. Further work would need to tease apart the

interplay of material and process complications. In particular, the synthesis process should

be reevaluated to remove toluene and adjust the pH to a neutral level.

121

Figure 5.9: Luminesence readings of 3T3 cells immediately after printing bioinks of (a)
GelMA, (b) GelMA + AgBG Blend, and (c) GAB + GelMA Blend and after 24 hours (d-f).

5.4 Conclusions

In conclusion, GelMA and GAB were synthesized and enabled some inks that were 3D

printable in acellular trials. Most characteristics of the materials were consistent with those

previously reported, but synthesized GAB did not have visible pores at the micron level.

This is suspected to cause challenges during resuspension as a bio-ink. The materials were

further found to produce acidic inks and may also be toxic due to residual toluene. Both

factors are suspected to have resulted in little to no cell viability

Initial extrusion printing trials of Allevi GelMA and synthesized GelMA suggest that

3D printing is viable. However, cellular printing with synthesized materials exposed several

points of improvement. While GelMA and the Blend were printable in acellular trials, intro-

ducing cellular media resulted in lower viscosity inks that did not allow for mesh resolution.

The rheology of the ink further changed upon neutralization before the incorporation of

cells and resulted in an even lower resolution 3DBP. Rheological issues may be overcome by

ensuring pH-neutral materials and careful lyophilization to enable more porous structures

during synthesis. Revised material synthesis protocols and re-optimizing printing parameters

are required for future work. More recent revisions to the material synthesis and printing

protocols show promising results toward improved cell viability with GAB.

122

Chapter 6

Conclusions

This work takes a multi-scale approach to understanding engineering challenges in producing

and optimizing biomaterials for versatile applications. Biomaterials can be broadly applied

to enable advancements in medical treatments, energy production, and consumer products.

Chapter 2 presents a finite difference model to predict transient behavior in the biocatalytic

reaction cascade of the MDH-CS enzyme complex. Chapter 3 employs a first-principles

thermodynamic analysis to assess the reaction mechanisms and cause of inhibition involved

in the electrochemical upgrading of a binary mixture of bio-oil constituents furfural and 4-

propylphenol. Chapter 4 utilizes linear scaling relationships to determine activation barriers

and kinetic rate constants for the system presented in Chapter 3. Chapter 5 sought to validate

a new gelatin-based bio-ink for the 3D printing of cell-laden scaffolds in tissue regenerative

applications. Each topic demonstrates how we can learn from nature to engineer efficient

solutions for various applications.

6.1 Enzyme-inspired Efficient Reaction Cascades

In Chapter 2, the dynamics of intermediate channeling in the malate dehydrogenase cit-

rate synthase (MDH-CS) complex is modeled using a finite difference approach applied to

a Markov state model. The model predicts lag times for the recombinant and mutant com-

plexes comparable to their experimental counterpart. Surface transport was further assessed

through analysis of specific active sites and paths. The paths on the recombinant com-

plex exhibited comparable lag times, indicating similar efficiencies and insensitivity to path

specifics. The mutant complex, however, exhibited more sensitivity to the reaction path and

was predicted to have more significant variation in path efficiencies.

123

Previous work identified Path 1 from SM 1 to SC1 as the most efficient path, which we

have confirmed by analysis of surface occupancy. A path preference is more distinct in the

mutant than the recombinant, where the surface occupancy and bulk OAA are higher for the

mutant than the recombinant. The greater efficiency of the recombinant than the mutant is

attributed to the recombinant’s preference for Path 1 and a lower amount of OAA leaked to

the bulk.

We conclude this work by briefly investigating the implications of reversible kinetics at

MDH active sites in the recombinant complex. For a fixed concentration of reaction product

NADH, the rate of CoA generation is predicted in channeling and free enzyme systems. The

channeling system was significantly less sensitive to NADH presence than free enzymes. This

is marked by a minimal change in CoA generation rate at high NADH concentration for the

channeling system compared to the free enzymes.

6.1.1 Future Work

The model presented in Chapter 2 can handle many reaction and transport events efficiently

while predicting the transient behavior of MDH-CS. However, some further development

of this system, more broadly, may extend this approach to rapid and dynamic prediction

of reaction cascade behavior. Regarding the MDH-CS system directly, intriguing results

were determined while investigating the implications of the reversible reaction at MDH.

Specifically, a maximum amplification of the CoA generation rate for the channeled system

compared to free enzymes existed around moderate NADH concentrations. This may be

attributed to the interplay of transport and reaction kinetics. Compared to the mutant,

which has different MDH kinetic rates, a more thorough investigation of this aspect may

reveal a reason for the maximum amplification of product generation. The design of efficient

artificial cascades is promoted by understanding how to accurately tune this system to the

highest performance.

Broad application of this technique for studying biocatalytic cascades still requires over-

coming some computational limitations. The connectivity network describing intermediate

transport calculated by a Markov state model enables consideration of transport in the finite

difference model. However, building a MSM is computationally demanding. By applying

124

these methods to other biocatalytic cascades, one may be able to predict the general behavior

of electrostatically channeling enzyme systems. Generalizing transport to overcome the com-

putational demand of MSMs would enable a more rapid prediction of the interplay between
cascade kinetics and surface transport for the design of efficient reaction cascades. 195

6.2 Multiscale Modeling of Electrochemical Conver-

sion of Biomass

In Chapters 3 and 4, computational and experimental methods were employed to assess the

mechanisms that may cause inhibition of the electrocatalytic hydrogenation of 4-propylphenol

to propylcyclohexane in the presence of furfural on a Pt/Ru electrode. Chapter 3 primarily

targets the thermodynamics of adsorption and intermediate reaction states. DFT methods

were used to predict the adsorption energy and reaction free energies of the 4-PP to PC and

FF to THFA pathways stable and radical intermediates on Pt(100), Pt(110), Pt(111), and

Ru(0001). Free energies and heats of adsorption were also determined experimentally for

FF, FA, and MF by blocked hydrogen under potential deposition on a Pt electrocatalyst.

The aqueous phase results are correlated to the gas phase and 1/16th coverage to compare

with the free energies of adsorption predicted by DFT.

Computationally, the electronic adsorption energy of reactants 4-PP and FF are most

competitive on Pt(110) and Ru(0001), and 4-PP always binds more strongly to a given

facet. Where 4-PP and FF were competitive, the free energy of adsorption further favors

4-PP because of the increase in Zero Point Energy for larger molecules. Experimental free

energies of adsorption on Pt support some conclusions determined through the computational

assessment. MF, FF, and FA exhibit comparable binding strengths in the aqueous phase,

but discrepancies arise when extrapolating to the gas phase. The isotherms of FF and FA

are nearly identical and suggest that either molecule would completely block Hupd at higher

organic concentrations than MF. Quantitatively, the experimental and computational results

are dissimilar, which may be due to effects not accounted for in Henry’s Law.

Predicted free energies of reactions along both pathways were further assessed for variable

125

applied potential. Reactions for both paths are most favorable on Pt(110) and competitive

on Pt(100). The FF pathway requires less potential to have favorable reaction energies than

the 4-PP pathway on each of the facets studied. Finally, FF proceeds to THFA via FA on

all facets, as previous experiments suggested. The 4PP pathway also proceeds as expected

on Pt facets but deviates to a propylbenzene derivative instead of the 4-propylcyclohexanol

intermediate on Ru. The pathway reaction free energies and the adsorption energies of FF

and 4-PP are competitive on Ru. Due to this, it is thermodynamically likely that FF would

block the adsorption and reaction of 4-PP on Ru, resulting in a decrease in PC product.

Positive reaction energies suggest that more barriers exist in the 4-PP pathways than

in the FF pathways. In Chapter 4, the activation free energy barriers are assessed. Linear

scaling relationships correlate intermediate state energies to the activation energy to enable

rapid prediction of transition state energy. The 4PP pathway exhibits larger barriers and

more positive reaction energies than the FF to THFA path. The FF to THFA reaction

is predicted to be more thermodynamically favorable than the 4PP to PC reaction on the

Pt/Ru catalyst. The thermodynamic analysis was complemented by an analysis of potential-

dependent kinetic rate constants calculated from the activation free energy of each transition

state. Generally, the 4PP reactions exhibit smaller rate constants than the FF reactions.

Therefore, the FF to THFA reaction is also predicted to be more kinetically favorable than

the 4PP to PC pathway.

6.2.1 Future Work

The work presented in Chapters 3 and 4 sought to understand the complexities of interacting

bio-oil molecules during electrocatalytic hydrogenation processes. Some details have been

elucidated, which indicate competitive activity between furans and phenols. However, real

bio-oil contains a much broader range of molecular species and likely involves several more

interaction effects. In addition to considering activation barriers and solvation effects, the

model needs to be scaled in time or space and in the number of interacting reaction pathways.

Continually considering factors such as potential dependence might enable greater synergy

between computational predictions and experimental results.

In Chapter 4, future research directions were initiated by predicting reaction rate con-

126

stants from activation barriers. This approach was also expanded to include potential depen-

dence, enabling a transient FF and 4PP competitive reactions model. Experimental methods
may also be considered to validate the predicted rate constants.104 However, further consid-

erations may need to compensate for the ability of computational methods to predict radical

formation, whereas experiments target stable intermediates.

Similar studies of other bio-oil constituents would make it possible to predict the behavior

of a real bio-oil, but it would be time-intensive. Predicting adsorption energies and reaction

barriers by high-throughput means would expedite the prediction of reaction rate constants,

enabling larger-scale models of interacting constituent molecules.

6.3 Manufactured Cell-Laden Tissue Scaffolds

In Chapter 5, validation of GAB as a new biomaterial was pursued in anticipation that GAB

bioink has superior cell viability in 3D printed scaffolds for tissue regeneration applications.

Initially, GelMA and GAB were synthesized in a consistent way with the literature. Most

characteristics of the materials were consistent with those previously reported, but synthe-

sized GAB did not have visible pores at the micron level. The lack of micropores in GAB

may have contributed to the challenges faced during resuspension to produce bio-inks. The

materials further resulted in acidic inks that may be additionally toxic because of residual

toluene from the synthesis. These factors are suspected of causing little to no cell viability

within the cell-laden scaffolds.

Initial extrusion printing trials of Allevi GelMA and synthesized GelMA suggest that 3D

printing is a viable approach. Comparable mesh resolution was achieved with synthesized

GelMA compared to the purchased GelMA. Similarly, the control GelMA - AgBG blend

ink yielded mesh resolution. Due to the reconstitution challenges with the GAB material,

some resolution was only achieved when GAB was mixed with GelMA. In acellular trials,

nearly optimal printing parameters were determined for the GelMA, but the GAB presented

challenges that required further optimization. Cellular printing with synthesized materials

exposed several more points of improvement.

127

6.3.1 Future Work

Several aspects of further work are required to validate GAB as a viable and more effective

bio-ink. Many of these items may be related to the synthesis process. The acidic and toxic

nature of the materials was hypothesized to have caused low cell viability. This may be

overcome by incorporating a dialysis process within the synthesis to exchange potentially

problematic residual reagents with water. Further, lyophilization should be executed care-

fully to ensure a more porous structure. A porous and more pH-neutral material may better

enable cellular printing because the neutralization process of Chapter 5 reduced the viscosity,

resulting in an inability to achieve mesh resolution.

While GelMA and the Blend were printable in acellular trials, the introduction of cellular

media also resulted in lower viscosity inks that did not allow for mesh resolution. Using any of

these materials as a bioink requires further optimization of printing parameters. Alternative

scaffold manufacturing techniques may also yield desirable results. An example includes

electrospinning, where a very fine-resolution mesh is built by drawing a thread of the bioink
along an electric field. 196,197 A grounded collector will gather the micron-scale fibers by

applying high voltage at the discharge point of the bioink. This approach may overcome the

rheological challenges of 3D printing while producing a fine mesh.

Beyond cell viability, the longevity of cells within the scaffolds is also an important

parameter to consider. The presence of oxygen during early tissue generation stages may

enable longer cell viability. Oxygen generation techniques, such as incorporating calcium
peroxide198,199 or electrochemical water splitting 200,201, could further aid in cell longevity.

More recent work revised the material synthesis protocol of GAB and GelMA to address

concerns due to pH and the use of toluene. Recent cellular printing also used mesenchymal

stem cells (MSCs) instead of fibroblasts. These adjustments show promising improvements

in printibility and cell viability.

128

6.4 Conclusion

Current and future work strives to advance the understanding of bio-inspired materials for

various engineering challenges and solutions. Systematic, multi-scale studies are employed

to understand and develop bio-inspired materials from medical treatment to chemical and

energy production. First, a finite difference model of TCA cycle enzyme aims to under-

stand mechanisms within a naturally occurring reaction cascade to develop efficient artificial

reaction cascades. Then, thermodynamic and kinetic analysis of electrocatalytic upgrad-

ing of bio-oil constituent mixtures was executed to understand the underlying mechanisms

contributing to the complexities in bio-oil upgrading to commodity chemicals. Finally, val-

idation of a new gelatin-based bio-ink for tissue regeneration was reported to establish a

more bio-compatible engineered solution to tissue regenerative materials.

In conclusion,

nature presents ample opportunities for bio-inspired materials and processing to produce

alternative resources. The work herein established a greater mechanistic understanding of

bioelectrochemical systems for designing efficient solutions in energy, commodity chemicals,

and medical applications.

129

BIBLIOGRAPHY

turning point

[1] World Energy Outlook 2022 shows

the global energy crisis can be a his-
toric
- News.
secure
https://www.iea.org/news/world-energy-outlook-2022-shows-the-global-energy-crisis-
can-be-a-historic-turning-point-towards-a-cleaner-and-more-secure-future,
October
2022.

towards a cleaner and more

future

[2] Ian Wheeldon, Shelley D. Minteer, Scott Banta, Scott Calabrese Barton, Plamen
Atanassov, and Matthew Sigman. Substrate channelling as an approach to cascade
reactions. Nat. Chem., 8(4):299–309, April 2016.

[3] Yoo Seok Lee, Koun Lim, and Shelley D. Minteer. Cascaded Biocatalysis and Bio-
electrocatalysis: Overview and Recent Advances. Annu. Rev. Phys. Chem., January
2021.

[4] Tony Bridgwater. Biomass for energy. Journal of the Science of Food and Agriculture,

86(12):1755–1768, 2006.

[5] Adam Christoph Marsh. DEVELOPMENT OF 3D BIOACTIVE AND ANTIBAC-
TERIAL SILICATE-BASED SCAFFOLDS FOR BONE TISSUE REGENERATION
IN LOAD-BEARING APPLICATIONS. PhD thesis, Michigan State University, 2021.

[6] Kanchan Suklal Chavan and Scott Calabrese Barton.

Simulation of Intermedi-
ate Channeling by Nanoscale Confinement. The Journal of Physical Chemistry C,
122(26):14474–14480, July 2018.

[7] Yuanchao Liu, Ivana Matanovic, David P. Hickey, Shelley D. Minteer, Plamen
Atanassov, and Scott Calabrese Barton. Cascade Kinetics of an Artificial Metabolon
by Molecular Dynamics and Kinetic Monte Carlo. ACS Catalysis, 8(8):7719–7726,
August 2018.

[8] Jia-Wei Shen, Tao Wu, Qi Wang, and Hai-Hua Pan. Molecular simulation of protein
adsorption and desorption on hydroxyapatite surfaces. Biomaterials, 29(5):513–532,
February 2008.

[9] Adrie J. J. Straathof. Transformation of Biomass into Commodity Chemicals Using

Enzymes or Cells. Chem. Rev., 114(3):1871–1908, February 2014.

[10] Su Keun Kuk, Raushan K Singh, Dong Heon Nam, Ranjitha Singh, Jung-Kul Lee, and
Chan Beum Park. Photoelectrochemical Reduction of Carbon Dioxide to Methanol
through a Highly Efficient Enzyme Cascade. Angew Chem Int Ed, 56(14):3827–3832,
March 2017.

[11] Junfa Yin, Tian Xu, Ning Zhang, and Hailin Wang. Three-Enzyme Cascade Biore-
actor for Rapid Digestion of Genomic DNA into Single Nucleosides. Anal. Chem.,
88(15):7730–7737, August 2016.

130

[12] Florika C. Macazo and Shelley D. Minteer. Enzyme cascades in biofuel cells. Curr.

Opin. Electrochem., 5(1):114–120, October 2017.

[13] H. Scott Fogler. Elements of Chemical Reaction Engineering. Prentice Hall, fifth

edition, 2016.

[14] Athel Cornish-Bowden. One hundred years of Michaelis–Menten kinetics. Perspectives

in Science, 4:3–9, March 2015.

[15] Leonor Michaelis and Maud Leonora Menten. Die Kinetik der Invertinwirkung. Bio-

chemische Zeitschrift, 49:333–369, 1913.

[16] Beyza Bulutoglu, Kristen E. Garcia, Fei Wu, Shelley D. Minteer, and Scott Banta.
Direct Evidence for Metabolon Formation and Substrate Channeling in Recombinant
TCA Cycle Enzymes. ACS Chem. Biol., 11(10):2847–2853, 2016.

[17] Yun Wang and Anthony K. Mittermaier. Characterizing Bi-substrate Enzyme Kinetics

at High Resolution by 2D-ITC. Anal. Chem., 93(37):12723–12732, September 2021.

[18] Athel Cornish-Bowden. Chapter 6 - Two-substrate reactions.

In Athel Cornish-
Bowden, editor, Fundamentals of Enzyme Kinetics, pages 99–129. Butterworth-
Heinemann, January 1979.

[19] Matthew J. Kummer, Yoo Seok Lee, Mengwei Yuan, Bassam Alkotaini, John Zhao,
Emmy Blumenthal, and Shelley D. Minteer. Substrate Channeling by a Rationally
Designed Fusion Protein in a Biocatalytic Cascade. JACS Au, 1(8):1187–1197, August
2021.

[20] Adrian H. Elcock, Gary A. Huber, and J. Andrew McCammon. Electrostatic channel-
ing of substrates between enzyme active sites: Comparison of simulation and experi-
ment. Biochem., 36(51):16049–16058, 1997.

[21] Changsun Eun, Peter M Kekenes-Huskey, Vincent T Metzger, and J Andrew Mc-
Cammon. A model study of sequential enzyme reactions and electrostatic channeling.
Journal of Chemical Physics, 140(10):105101, March 2014.

[22] Mario Milani, Alessandra Pesce, Martino Bolognesi, Alessio Bocedi, and Paolo Ascenzi.
Substrate channeling: Molecular bases. Biochemistry and Molecular Biology Education,
31(4):228–233, July 2003.

[23] Setare Mostajabi Sarhangi and Dmitry V. Matyushov. Electron Tunneling in Biology:

When Does it Matter? ACS Omega, 8(30):27355–27365, August 2023.

[24] Yuanchao Liu, David P. Hickey, Shelley D. Minteer, Alex Dickson, and Scott Cal-
abrese Barton. Markov-State Transition Path Analysis of Electrostatic Channeling. J.
Phys. Chem. C, 123(24):15284–15292, June 2019.

131

[25] Konstantin Shatalin, Sandrine Lebreton, Magali Rault-Leonardon, Christian V´elot,
and Paul A. Srere. Electrostatic Channeling of Oxaloacetate in a Fusion Protein of
Porcine Citrate Synthase and Porcine Mitochondrial Malate Dehydrogenase. Biochem.,
38(3):881–889, January 1999.

[26] Yan Xie, Shelley D. Minteer, Scott Banta, and Scott Calabrese Barton. Markov State
Study of Electrostatic Channeling within the Tricarboxylic Acid Cycle Supercomplex.
ACS Nanosci. Au, 2(5):414–421, October 2022.

[27] Yan Xie and Scott Calabrese Barton. Infrequent metadynamics study of rare-event
electrostatic channeling. Phys. Chem. Chem. Phys., 23(23):13381–13388, 2021.

[28] V. ´Asgeirsson, H. J´onsson, and K. T. Wikfeldt. Long-Time Scale Simulations of
Tunneling-Assisted Diffusion of Hydrogen on Ice Surfaces at Low Temperature. J.
Phys. Chem. C, 121(3):1648–1657, January 2017.

[29] Kanchan Suklal Chavan and Scott Calabrese Barton. Confinement and Diffusion of
Small Molecules in a Molecular-Scale Tunnel. J. Electrochem. Soc., 167(2):023505,
February 2020.

[30] Yubo Fan, Liliya Lund, Qiang Shao, Yi-Qin Gao, and Frank M. Raushel. A Combined
Theoretical and Experimental Study of the Ammonia Tunnel in Carbamoyl Phosphate
Synthetase. J. Am. Chem. Soc., 131(29):10211–10219, July 2009.

[31] Frank M. Raushel, James B. Thoden, and Hazel M. Holden. Enzymes with Molecular

Tunnels. Acc. Chem. Res., 36(7):539–548, July 2003.

[32] Yuanchao Liu, David P. Hickey, Jing-Yao Guo, Erica Earl, Sofiene Abdellaoui, Ross D.
Milton, Matthew S. Sigman, Shelley D. Minteer, and Scott Calabrese Barton. Sub-
strate Channeling in an Artificial Metabolon: A Molecular Dynamics Blueprint for an
Experimental Peptide Bridge. ACS Catalysis, 7(4):2486–2493, April 2017.

[33] Erica Earl and Scott Calabrese Barton. Simulation of intermediate transport in
nanoscale scaffolds for multistep catalytic reactions. Phys. Chem. Chem. Phys.,
19(23):15463–15470, 2017.

[34] Da-Jiang Liu, Jing Wang, David M. Ackerman, Igor I. Slowing, Marek Pruski, Hung-
Interplay between Anomalous
Ting Chen, Victor S.-Y. Lin, and James W. Evans.
Transport and Catalytic Reaction Kinetics in Single-File Nanoporous Systems. ACS
Catal., 1(7):751–763, July 2011.

[35] Yu-ming M. Huang, Gary A. Huber, Nuo Wang, Shelley D. Minteer, and J. Andrew
McCammon. Brownian dynamic study of an enzyme metabolon in the TCA cycle:
Substrate kinetics and channeling. Protein Sci., 27(2):463–471, 2018.

[36] Xue Bin and Peter D. Pawelek. Evidence of isochorismate channeling between the
Escherichia coli enterobactin biosynthetic enzymes EntC and EntB. Protein Science,
33(8):e5122, 2024.

132

[37] Sylvie Ouellette, Paknoosh Pakarian, Xue Bin, and Peter D. Pawelek. Evidence of
an intracellular interaction between the Escherichia coli enzymes EntC and EntB and
identification of a potential electrostatic channeling surface. Biochimie, 202:159–165,
November 2022.

[38] Asha Kumari. Chapter 2 - Citric Acid Cycle. In Asha Kumari, editor, Sweet Biochem-

istry, pages 7–11. Academic Press, January 2018.

[39] Arieh Warshel. Molecular Dynamics Simulations of Biological Reactions. Acc. Chem.

Res., 35(6):385–395, June 2002.

[40] Mie Andersen, Chiara Panosetti, and Karsten Reuter. A Practical Guide to Surface

Kinetic Monte Carlo Simulations. Front. Chem., 7, 2019.

[41] Abhijit Chatterjee and Dionisios G. Vlachos. An overview of spatial microscopic and
accelerated kinetic Monte Carlo methods. J Computer-Aided Mater Des, 14(2):253–
308, July 2007.

[42] Arti Bhoutekar, Susmita Ghosh, Swati Bhattacharya, and Abhijit Chatterjee. A new
class of enhanced kinetic sampling methods for building Markov state models. J. Chem.
Phys., 147(15):152702, June 2017.

[43] Yang Cao, Linda R. Petzold, Muruhan Rathinam, and Daniel T. Gillespie. The nu-
merical stability of leaping methods for stochastic simulation of chemically reacting
systems. J. Chem. Phys., 121(24):12169, 2004.

[44] Sherly Rusli, Janna Grabowski, Anja Drews, and Matthias Kraume. A Multi-Scale
Approach to Modeling the Interfacial Reaction Kinetics of Lipases with Emphasis on
Enzyme Adsorption at Water-Oil Interfaces. Processes, 8(9):1082, 2020.

[45] H`ector Prats, Francesc Illas, and Ram´on Say´os. General concepts, assumptions, draw-
backs, and misuses in kinetic Monte Carlo and microkinetic modeling simulations ap-
plied to computational heterogeneous catalysis. Int. J. Quantum Chem., 118(9):e25518,
2018.

[46] Christina M. Wark, Yan Xie, and Scott Calabrese Barton. Finite difference model of
electrostatic channeling in TCA cycle enzymes. Electrochim. Acta, 528:146131, July
2025.

[47] Fayaz A. Malla, Suhaib A. Bandh, Shahid A. Wani, Anh Tuan Hoang, and Nazir Ah-
mad Sofi. Biofuels: Potential Alternatives to Fossil Fuels. In Suhaib A. Bandh and
Fayaz A. Malla, editors, Biofuels in Circular Economy, pages 1–15. Springer Nature,
Singapore, 2022.

[48] Guangcan Su, Hwai Chyuan Ong, M. Mofijur, T. M. Indra Mahlia, and Yong Sik Ok.
Pyrolysis of waste oils for the production of biofuels: A critical review. Journal of
Hazardous Materials, 424:127396, February 2022.

133

[49] Manali S. Dhawan, Ganapati D. Yadav, and Scott Calabrese Barton.

Zinc-
electrocatalyzed hydrogenation of furfural in near-neutral electrolytes. Sustainable
Energy Fuels, 5(11):2972–2984, 2021.

Hao.

[50] Pengchao
Dimeric,
Chemistry, Mechanistic,
https://www.proquest.com/docview/2036629929/fulltextPDF/B2E002C3B2074166PQ/1?accountid=12598.

of Monomeric,
and Polymeric Lignin Model Compounds With Raney Nickel:
ProQuest.
Product Toxicity

Electrocatalytic

Hydrogenation

Studies

and

-

[51] Chun Ho Lam, Christy B. Lowe, Zhenglong Li, Kelsey N. Longe, Jordan T. Rayburn,
Michael A. Caldwell, Carly E. Houdek, Jack B. Maguire, Christopher M. Saffron,
Dennis J. Miller, and James E. Jackson. Electrocatalytic upgrading of model lignin
monomers with earth abundant metal electrodes. Green Chem., 17(1):601–609, 2015.

[52] Dekui Shen, Wei Jin, Jun Hu, Rui Xiao, and Kaihong Luo. An overview on fast
pyrolysis of the main constituents in lignocellulosic biomass to valued-added chemicals:
Structures, pathways and interactions. Renewable and Sustainable Energy Reviews,
51:761–774, November 2015.

[53] Furkan H. Isikgor and C. Remzi Becer. Lignocellulosic biomass: A sustainable plat-
form for the production of bio-based chemicals and polymers. Polymer Chemistry,
6(25):4497–4559, 2015.

[54] Maria Temnikova, Jury Medvedev, Xenia Medvedeva, Nyhenflore H. Delva, Evgeniia
Khairullina, Elena Krivoshapkina, and Anna Klinkova. Electrochemical Hydrodimer-
ization of Furfural in Organic Media as an Efficient Route to Jet Fuel Precursor.
ChemElectroChem, 10(2):e202200865, 2023.

[55] Meheryar R. Kasad, James E. Jackson, and Christopher M. Saffron. Electrocatalytic
hydrogenation of the formyl group and heteroaromatic ring in furfural on activated
carbon cloth-supported ruthenium. RSC Sustain., 2(10):3001–3013, October 2024.

[56] Eveliina M¨akel¨a, Jos´e Luis Gonz´alez Escobedo, Jouni Neuvonen, Jouko Lahtinen, Ma-
rina Lindblad, Ulla Lassi, Reetta Karinen, and Riikka L. Puurunen. Liquid-phase
Hydrodeoxygenation of 4-Propylphenol to Propylbenzene: Reducible Supports for Pt
Catalysts. ChemCatChem, 12(16):4090–4104, August 2020.

[57] Jeffrey R. Page, Amol Pophali, Taejin Kim, Juan A. Lopez-Ruiz, Stoyan Bliznakov,
and Julia A. Valla. Effect of Pt and Ru-based catalysts on the electrochemical hy-
drodeoxygenation of phenol to cyclohexane. Catal. Sci. Technol., 14(19):5559–5573,
2024.

[58] Shyam Deo and Michael J. Janik. Predicting an optimal oxide/metal catalytic inter-
face for hydrodeoxygenation chemistry of biomass derivatives. Catal. Sci. Technol.,
11(16):5606–5618, 2021.

[59] Meheryar R. Kasad, James E. Jackson, and Christopher M. Saffron. Electrocatalytic
conversion of biomass-derived oxygenated aromatics to cycloalkanes. Sustainable En-
ergy Fuels, November 2024.

134

[60] Tao Peng, Wenbin Zhang, Baiyao Liang, Guanwu Lian, Yun Zhang, and Wei Zhao.
Electrocatalytic valorization of lignocellulose-derived aromatics at industrial-scale cur-
rent densities. Nat Commun, 14(1):7229, November 2023.

[61] Mudasir Akbar Shah, Wasif Farooq, Tasrin Shahnaz, and Muthumariappan Aki-
larasan. Bioenergy and Value-Added Chemicals Derived Through Electrocatalytic
Upgradation of Biomass: A Critical Review. Bioenerg. Res., August 2024.

[62] Yuting Zhou, Grace E. Klinger, Eric L. Hegg, Christopher M. Saffron, and James E.
Jackson. Multiple Mechanisms Mapped in Aryl Alkyl Ether Cleavage via Aqueous
Electrocatalytic Hydrogenation over Skeletal Nickel. J. Am. Chem. Soc., 142(8):4037–
4050, February 2020.

[63] Jeffrey R. Page, Zachary Manfredi, Stoyan Bliznakov, and Julia A. Valla. Recent
Progress in Electrochemical Upgrading of Bio-Oil Model Compounds and Bio-Oils to
Renewable Fuels and Platform Chemicals. Materials (Basel), 16(1):394, January 2023.

[64] Allen J Bard, Larry R Faulkner, New York, Chichester @bullet, Weinheim Brisbane,
and Singapore E Toronto. ELECTROCHEMICAL METHODS Fundamentals and Ap-
plications. Wiley, 1944.

[65] Martin Z. Bazant. Unified quantum theory of electrochemical kinetics by coupled

ion–electron transfer. Faraday Discuss., 246:60–124, 2023.

[66] A. Shayeghi, S. Kr¨ahling, P. H¨ortz, R. L. Johnston, C. J. Heard, and R. Sch¨afer. Ad-
sorption of Acetonitrile, Benzene, and Benzonitrile on Pt(111): Single Crystal Adsorp-
tion Calorimetry and Density Functional Theory. J. Phys. Chem. C, 121(39):21354–
21363, October 2017.

[67] Yun Shi, Yulei Zhu, Yong Yang, Yong-Wang Li, and Haijun Jiao. Exploring Furfural
Catalytic Conversion on Cu(111) from Computation. ACS Catal., 5(7):4020–4032,
July 2015.

[68] Yang Song, Udishnu Sanyal, Dhananjai Pangotra, Jamie D. Holladay, Donald M. Ca-
maioni, Oliver Y. Guti´errez, and Johannes A. Lercher. Hydrogenation of benzalde-
hyde via electrocatalysis and thermal catalysis on carbon-supported metals. Journal
of Catalysis, 359:68–75, March 2018.

[69] Juan A. Lopez-Ruiz, Udishnu Sanyal, Jonathan Egbert, Oliver Y. Guti´errez, and Jamie
Holladay. Kinetic Investigation of the Sustainable Electrocatalytic Hydrogenation of
Benzaldehyde on Pd/C: Effect of Electrolyte Composition and Half-Cell Potentials.
ACS Sustainable Chem. Eng., 6(12):16073–16085, December 2018.

[70] Andrzej Lasia. Mechanism and kinetics of the hydrogen evolution reaction. Interna-

tional Journal of Hydrogen Energy, 44(36):19484–19518, July 2019.

[71] Anthony R. Kucernak and Christopher Zalitis. General Models for the Electrochemical
Hydrogen Oxidation and Hydrogen Evolution Reactions: Theoretical Derivation and

135

Experimental Results under Near Mass-Transport Free Conditions. J. Phys. Chem. C,
120(20):10721–10745, May 2016.

[72] Ana L. Santos, Maria-Jo˜ao Cebola, and Diogo M. F. Santos. Towards the Hydrogen
Economy—A Review of the Parameters That Influence the Efficiency of Alkaline Water
Electrolyzers. Energies, 14(11):3193, May 2021.

[73] Renato Seeber, Chiara Zanardi, and Gy¨orgy Inzelt. Links between electrochemical

thermodynamics and kinetics. ChemTexts, 1(4):18, December 2015.

[74] David S. Sholl and Janice Steckel. Frontmatter. In Density Functional Theory, pages

i–xii. John Wiley & Sons, Ltd, 2009.

[75] W. Kohn and L. J. Sham. Self-Consistent Equations Including Exchange and Corre-

lation Effects. Phys. Rev., 140(4A):A1133–A1138, November 1965.

[76] Shyue Ping Ong, William Davidson Richards, Anubhav Jain, Geoffroy Hautier, Michael
Kocher, Shreyas Cholia, Dan Gunter, Vincent L. Chevrier, Kristin A. Persson, and
Gerbrand Ceder. Python Materials Genomics (pymatgen): A robust, open-source
python library for materials analysis. Computational Materials Science, 68:314–319,
February 2013.

[77] J¨urgen Hafner. Ab-initio simulations of materials using VASP: Density-functional
theory and beyond. Journal of Computational Chemistry, 29(13):2044–2078, 2008.

[78] Markus Bursch, Jan-Michael Mewes, Andreas Hansen, and Stefan Grimme. Best-
Practice DFT Protocols for Basic Molecular Computational Chemistry**. Angew
Chem Int Ed, 61(42):e202205735, October 2022.

[79] Renqin Zhang, Alyssa J. Hensley, Jean-Sabin McEwen, Sandra Wickert, Erik Dar-
latt, Kristina Fischer, Matthias Sch¨oppke, Reinhard Denecke, Regine Streber, Michael
Lorenz, Christian Papp, and Hans-Peter Steinr¨uck.
Integrated X-ray photoelectron
spectroscopy and DFT characterization of benzene adsorption on Pt(111), Pt(355) and
Pt(322) surfaces. Physical Chemistry Chemical Physics, 15(47):20662–20671, 2013.

[80] Karsten Reuter. Ab Initio Thermodynamics and First-Principles Microkinetics for
Surface Catalysis. In Joost Frenken and Irene Groot, editors, Operando Research in
Heterogeneous Catalysis, Springer Series in Chemical Physics, pages 151–188. Springer
International Publishing, Cham, 2017.

[81] Sunghwan Kim, Jie Chen, Tiejun Cheng, Asta Gindulyte, Jia He, Siqian He, Qingliang
Li, Benjamin A Shoemaker, Paul A Thiessen, Bo Yu, Leonid Zaslavsky, Jian Zhang,
and Evan E Bolton. PubChem 2025 update. Nucleic Acids Research, 53(D1):D1516–
D1525, January 2025.

[82] Gabriel A. Bramley, Manh-Thuong Nguyen, Vassiliki-Alexandra Glezakou, Roger
Rousseau, and Chris-Kriton Skylaris. Understanding Adsorption of Organics on

136

Pt(111) in the Aqueous Phase: Insights from DFT Based Implicit Solvent and Sta-
tistical Thermodynamics Models. J. Chem. Theory Comput., 18(3):1849–1861, March
2022.

[83] Cheng-chau Chiu, Alexander Genest, Armando Borgna, and Notker R¨osch. Hy-
drodeoxygenation of Guaiacol over Ru(0001): A DFT Study. ACS Catal., 4(11):4178–
4188, November 2014.

[84] L. Delle Site, A. Alavi, and C. F. Abrams. Adsorption energies and geometries of
phenol on the (111) surface of nickel: An ab initio study. Phys. Rev. B, 67(19):193406,
May 2003.

[85] Kyungtae Lee, Geun Ho Gu, Charles A. Mullen, Akwasi A. Boateng, and Dionisios G.
Vlachos. Guaiacol Hydrodeoxygenation Mechanism on Pt(111): Insights from Density
Functional Theory and Linear Free Energy Relations. ChemSusChem, 8(2):315–322,
January 2015.

[86] David Loffreda. Theoretical insight of adsorption thermodynamics of multifunctional

molecules on metal surfaces. Surface Science, 600(10):2103–2112, May 2006.

[87] Michelle J.S. Spencer and Graeme L. Nyberg. DFT modelling of hydrogen on Cu(110)-

and (111)-type clusters. Molecular Simulation, 28(8-9):807–825, August 2002.

[88] Ask Hjorth Larsen, Jens Jørgen Mortensen, Jakob Blomqvist, Ivano E. Castelli, Rune
Christensen, Marcin Du(cid:32)lak, Jesper Friis, Michael N. Groves, Bjørk Hammer, Cory
Hargus, Eric D. Hermes, Paul C. Jennings, Peter Bjerre Jensen, James Kermode,
John R. Kitchin, Esben Leonhard Kolsbjerg, Joseph Kubal, Kristen Kaasbjerg, Steen
Lysgaard, J´on Bergmann Maronsson, Tristan Maxson, Thomas Olsen, Lars Pastewka,
Andrew Peterson, Carsten Rostgaard, Jakob Schiøtz, Ole Sch¨utt, Mikkel Strange,
Kristian S. Thygesen, Tejs Vegge, Lasse Vilhelmsen, Michael Walter, Zhenhua Zeng,
and Karsten W. Jacobsen. The atomic simulation environment-a Python library for
working with atoms. J Phys Condens Matter, 29(27):273002, July 2017.

[89] Joseph W Ochterski. Thermochemistry in Gaussian, 2000.

[90] J. K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J. R. Kitchin, T. Bligaard,
and H. J´onsson. Origin of the Overpotential for Oxygen Reduction at a Fuel-Cell
Cathode. J. Phys. Chem. B, 108(46):17886–17892, November 2004.

[91] Robert E. Warburton, Alexander V. Soudackov, and Sharon Hammes-Schiffer. The-
oretical Modeling of Electrochemical Proton-Coupled Electron Transfer. Chem. Rev.,
122(12):10599–10650, June 2022.

[92] David R. Weinberg, Christopher J. Gagliardi, Jonathan F. Hull, Christine Fe-
cenko Murphy, Caleb A. Kent, Brittany C. Westlake, Amit Paul, Daniel H. Ess,
Dewey Granville McCafferty, and Thomas J. Meyer. Proton-Coupled Electron Trans-
fer. Chem. Rev., 112(7):4016–4093, July 2012.

137

[93] Sneha A. Akhade, Nicole J. Bernstein, Monica R. Esopi, Michael J. Regula, and
Michael J. Janik. A simple method to approximate electrode potential-dependent
activation energies using density functional theory. Catalysis Today, 288:63–73, June
2017.

[94] Ruslan M. Mensharapov, Dmitry D. Spasov, Nataliya A. Ivanova, Adelina A. Za-
sypkina, Sergey A. Smirnov, and Sergey A. Grigoriev. Screening of Carbon-Supported
Platinum Electrocatalysts Using Frumkin Adsorption Isotherms. Inorganics, 11(3):103,
March 2023.

[95] Nirala Singh, Udishnu Sanyal, John L. Fulton, Oliver Y. Guti´errez, Johannes A.
Lercher, and Charles T. Campbell. Quantifying Adsorption of Organic Molecules on
Platinum in Aqueous Phase by Hydrogen Site Blocking and in Situ X-ray Absorption
Spectroscopy. ACS Catal., 9(8):6869–6881, August 2019.

[96] Jiˇr´ı Klimeˇs, David R Bowler, and Angelos Michaelides. A critical assessment of theo-
retical methods for finding reaction pathways and transition states of surface processes.
J. Phys.: Condens. Matter, 22(7):074203, February 2010.

[97] Graeme Henkelman and Hannes J´onsson. A dimer method for finding saddle points
on high dimensional potential surfaces using only first derivatives. The Journal of
Chemical Physics, 111(15):7010–7022, October 1999.

[98] Jonathan E. Sutton and Dionisios G. Vlachos. A Theoretical and Computational
Analysis of Linear Free Energy Relations for the Estimation of Activation Energies.
ACS Catal., 2(8):1624–1634, August 2012.

[99] Paolo Giannozzi, Stefano Baroni, Nicola Bonini, Matteo Calandra, Roberto Car, Carlo
Cavazzoni, Davide Ceresoli, Guido L. Chiarotti, Matteo Cococcioni, Ismaila Dabo,
Andrea Dal Corso, Stefano de Gironcoli, Stefano Fabris, Guido Fratesi, Ralph Gebauer,
Uwe Gerstmann, Christos Gougoussis, Anton Kokalj, Michele Lazzeri, Layla Martin-
Samos, Nicola Marzari, Francesco Mauri, Riccardo Mazzarello, Stefano Paolini, Alfredo
Pasquarello, Lorenzo Paulatto, Carlo Sbraccia, Sandro Scandolo, Gabriele Sclauzero,
Ari P. Seitsonen, Alexander Smogunov, Paolo Umari, and Renata M. Wentzcovitch.
QUANTUM ESPRESSO: A modular and open-source software project for quantum
simulations of materials. J. Phys.: Condens. Matter, 21(39):395502, September 2009.

[100] P. Giannozzi, O. Andreussi, T. Brumme, O. Bunau, M. Buongiorno Nardelli, M. Calan-
dra, R. Car, C. Cavazzoni, D. Ceresoli, M. Cococcioni, N. Colonna, I. Carnimeo, A. Dal
Corso, S. de Gironcoli, P. Delugas, R. A. DiStasio, A. Ferretti, A. Floris, G. Fratesi,
G. Fugallo, R. Gebauer, U. Gerstmann, F. Giustino, T. Gorni, J. Jia, M. Kawamura,
H.-Y. Ko, A. Kokalj, E. K¨u¸c¨ukbenli, M. Lazzeri, M. Marsili, N. Marzari, F. Mauri,
N. L. Nguyen, H.-V. Nguyen, A. Otero-de-la-Roza, L. Paulatto, S. Ponc´e, D. Rocca,
R. Sabatini, B. Santra, M. Schlipf, A. P. Seitsonen, A. Smogunov, I. Timrov, T. Thon-
hauser, P. Umari, N. Vast, X. Wu, and S. Baroni. Advanced capabilities for materials
modelling with Quantum ESPRESSO. J. Phys.: Condens. Matter, 29(46):465901,
October 2017.

138

[101] Dipika Rajendra Kanchan and Arghya Banerjee. Linear Scaling Relationships for Furan
Hydrodeoxygenation over Transition Metal and Bimetallic Surfaces. ChemSusChem,
16(18):e202300491, 2023.

[102] Bin Liu, Lei Cheng, Larry Curtiss, and Jeffrey Greeley. Effects of van der Waals
density functional corrections on trends in furfural adsorption and hydrogenation on
close-packed transition metal surfaces. Surface Science, 622:51–59, April 2014.

[103] Shanjun Mao, Zhe Wang, Zhirong Chen, Kejun Wu, Kaichao Zhang, Qichuan Li,
Huihuan Yan, Guofeng L¨u, Guodong Huang, and Yong Wang. Towards the selectivity
distinction of phenol hydrogenation on noble metal catalysts. Nano Materials Science,
5(1):91–100, March 2023.

[104] Heine A. Hansen, Venkatasubramanian Viswanathan, and Jens K. Nørskov. Unifying
Kinetic and Thermodynamic Analysis of 2 e– and 4 e– Reduction of Oxygen on Metal
Surfaces. J. Phys. Chem. C, 118(13):6706–6718, April 2014.

[105] Kelth J Laidler’ and M Chrlstlne Klng. The Development of Transition-State Theory.

J Phys Chem, 1983.

[106] Julian R. Jones. Review of bioactive glass: From Hench to hybrids. Acta Biomaterialia,

9(1):4457–4486, January 2013.

[107] Jean-Pierre Draye, Bernard Delaey, Andr´e Van de Voorde, An Van Den Bulcke, Bepke
De Reu, and Etienne Schacht. In vitro and in vivo biocompatibility of dextran dialde-
hyde cross-linked gelatin hydrogel films. Biomaterials, 19(18):1677–1687, September
1998.

[108] K. Panduranga Rao. Recent developments of collagen-based materials for medical
applications and drug delivery systems. Journal of Biomaterials Science, Polymer
Edition, 7(7):623–645, January 1996.

[109] Adam C. Marsh, Yaozhong Zhang, Lucrezia Poli, Neal Hammer, Aljoscha Roch, Martin
Crimp, and Xanthippi Chatzistavrou. 3D printed bioactive and antibacterial silicate
glass-ceramic scaffold by fused filament fabrication. Materials Science and Engineering:
C, 118:111516, January 2021.

[110] Adam C. Marsh, Ehsanul Hoque Apu, Marcus Bunn, Christopher H. Contag, Nureddin
Ashammakhi, and Xanthippi Chatzistavrou. A New Bioink for Improved 3D Bioprint-
ing of Bone-Like Constructs. Preprint, Bioengineering, November 2021.

[111] An I. Van Den Bulcke, Bogdan Bogdanov, Nadine De Rooze, Etienne H. Schacht,
Maria Cornelissen, and Hugo Berghmans. Structural and Rheological Properties of
Methacrylamide Modified Gelatin Hydrogels. Biomacromolecules, 1(1):31–38, March
2000.

139

[112] Adam C. Marsh, Nathan P. Mellott, Martin Crimp, Anthony Wren, Neal Hammer,
and Xanthippi Chatzistavrou. Ag-doped Bioactive Glass-Ceramic 3D Scaffolds: Mi-
crostructural, Antibacterial, and Biological Properties. Journal of the European Ce-
ramic Society, 41(6):3717–3730, June 2021.

[113] Adam C. Marsh, Nathan P. Mellott, Natalia Pajares-Chamorro, Martin Crimp, An-
thony Wren, Neal D. Hammer, and Xanthippi Chatzistavrou. Fabrication and multi-
scale characterization of 3D silver containing bioactive glass-ceramic scaffolds. Bioac-
tive Materials, 4:215–223, December 2019.

[114] Qing Tang, Sujiao Cao, Tian Ma, Xi Xiang, Hongrong Luo, Pavel Borovskikh,
Raul D. Rodriguez, Quanyi Guo, Li Qiu, and Chong Cheng. Engineering Biofunc-
tional Enzyme-Mimics for Catalytic Therapeutics and Diagnostics. Adv. Funct. Mater.,
31(7):2007475, 2021.

[115] Alisdair R Fernie, Fernando Carrari, and Lee J Sweetlove. Respiratory metabolism:
Glycolysis, the TCA cycle and mitochondrial electron transport. Curr. Opin. Plant
Biol., 7(3):254–261, June 2004.

[116] Fei Wu and Shelley Minteer. Krebs Cycle Metabolon: Structural Evidence of Substrate
Channeling Revealed by Cross-Linking and Mass Spectrometry. Angew. Chem. Int.
Ed., 54(6):1851–1854, 2015.

[117] Davide M. Ferraris, Ralf Spallek, Wulf Oehlmann, Mahavir Singh, and Menico Rizzi.
Structures of citrate synthase and malate dehydrogenase of Mycobacterium tubercu-
losis. Proteins, 83(2):389–394, 2015.

[118] Joy Omini, Izabela Wojciechowska, Aleksandra Skirycz, Hideaki Moriyama, and Toshi-
hiro Obata. Association of the malate dehydrogenase-citrate synthase metabolon is
modulated by intermediates of the Krebs tricarboxylic acid cycle. Sci Rep, 11(1):18770,
September 2021.

[119] Robert W. Guynn, Harris J. Gelberg, and Richard L. Veech. Equilibrium Constants of
the Malate Dehydrogenase, Citrate Synthase, Citrate Lyase, and Acetyl Coenzyme A
Hydrolysis Reactions under Physiological Conditions. J. Biol. Chem., 248(20):6957–
6965, October 1973.

[120] Elizabeth Jurrus, Dave Engel, Keith Star, Kyle Monson, Juan Brandi, Lisa E. Felberg,
David H. Brookes, Leighton Wilson, Jiahui Chen, Karina Liles, Minju Chun, Peter
Li, David W. Gohara, Todd Dolinsky, Robert Konecny, David R. Koes, Jens Erik
Nielsen, Teresa Head-Gordon, Weihua Geng, Robert Krasny, Guo-Wei Wei, Michael J.
Holst, J. Andrew McCammon, and Nathan A. Baker.
Improvements to the APBS
biomolecular solvation software suite. Protein Sci., 27(1):112–128, 2018.

[121] Michail Stamatakis and Dionisios G. Vlachos. A graph-theoretical kinetic Monte Carlo
framework for on-lattice chemical kinetics. J. Chem. Phys., 134(21):214115, June 2011.

140

[122] Michail Stamatakis and Dionisios G. Vlachos. Unraveling the complexity of catalytic
reactions via kinetic monte carlo simulation: Current status and frontiers. ACS Catal-
ysis, 2(12):2648–2663, 2012.

[123] Luca Messina, Nicolas Castin, Christophe Domain, and P¨ar Olsson. Introducing ab
initio based neural networks for transition-rate prediction in kinetic Monte Carlo sim-
ulations. Phys. Rev. B, 95(6):064112, February 2017.

[124] Haobo Li, Yan Jiao, Kenneth Davey, and Shi-Zhang Qiao. Data-Driven Machine
Learning for Understanding Surface Structures of Heterogeneous Catalysts. Angew.
Chem. Int. Ed., 62(9):e202216383, 2023.

[125] Ruiqi Zou, Hongxia Li, Junxiao Shi, Chunyan Sun, Geyu Lu, and Xu Yan. Dual-
enhanced enzyme cascade hybrid hydrogel for the construction of optical biosensor.
Biosens. Bioelectron., 263:116613, November 2024.

[126] Bernard Philippe, Youcef Saad, and William J. Stewart. Numerical Methods in Markov

Chain Modeling. Oper. Res., 40(6):1156–1179, 1992.

[127] Pauli Virtanen, Ralf Gommers, Travis E. Oliphant, Matt Haberland, Tyler Reddy,
David Cournapeau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan
Bright, St´efan J. van der Walt, Matthew Brett, Joshua Wilson, K. Jarrod Millman,
Nikolay Mayorov, Andrew R. J. Nelson, Eric Jones, Robert Kern, Eric Larson, C. J.
Carey, ˙Ilhan Polat, Yu Feng, Eric W. Moore, Jake VanderPlas, Denis Laxalde, Josef
Perktold, Robert Cimrman, Ian Henriksen, E. A. Quintero, Charles R. Harris, Anne M.
Archibald, Antˆonio H. Ribeiro, Fabian Pedregosa, and Paul van Mulbregt. SciPy 1.0:
Fundamental algorithms for scientific computing in Python. Nat Methods, 17(3):261–
272, March 2020.

[128] Martin K. Scherer, Benjamin Trendelkamp-Schroer, Fabian Paul, Guillermo P´erez-
Hern´andez, Moritz Hoffmann, Nuria Plattner, Christoph Wehmeyer, Jan-Hendrik
Prinz, and Frank No´e. PyEMMA 2: A Software Package for Estimation, Valida-
tion, and Analysis of Markov Models. J. Chem. Theory Comput., 11(11):5525–5542,
November 2015.

[129] Aric A Hagberg, Daniel A Schult, and Pieter J Swart. Exploring Network Structure,
Dynamics, and Function using NetworkX. Proceedings of the 7th Python in Science
Conference, 2008.

[130] Giancarlo Perrone, Jose Unpingco, and Haw-minn Lu. Network visualizations with

Pyvis and VisJS, June 2020.

[131] Christina M. Wark and Scott Calabrese Barton. Electrostatic Channeling MDH-CS.

https://github.com/scbgroup/Electrostatic Channeling MDH-CS, 2024.

[132] Jia Wu, Jinli Chen, Tianqi Yu, Zhixiang Zhai, Yumei Zhu, Xizi Wu, and Shibin Yin.
Boosting Electrochemical Kinetics of NiCo2 via MoO2 Modification for Biomass Up-
grading Assisted Hydrogen Evolution. ACS Catal., 13(20):13257–13266, October 2023.

141

[133] Mahlet Garedew, Chun Ho Lam, Laurene Petitjean, Shuquan Huang, Bing Song, Fang
Lin, James E. Jackson, Christopher M. Saffron, and Paul T. Anastas. Electrochemical
upgrading of depolymerized lignin: A review of model compound studies. Green Chem.,
23(8):2868–2899, 2021.

[134] Tedd E. Lister, Luis A. Diaz, Michael A. Lilga, Asanga B. Padmaperuma, YuPo
Lin, Varada Menon Palakkal, and Christopher G. Arges. Low-Temperature Electro-
chemical Upgrading of Bio-oils Using Polymer Electrolyte Membranes. Energy Fuels,
32(5):5944–5950, May 2018.

[135] Wei Liu, Wenqin You, Yutao Gong, and Yulin Deng. High-efficiency electrochemical
hydrodeoxygenation of bio-phenols to hydrocarbon fuels by a superacid-noble metal
particle dual-catalyst system. Energy Environ. Sci., 13(3):917–927, 2020.

[136] Francisco W. S. Lucas, R. Gary Grim, Sean A. Tacey, Courtney A. Downes, Joseph
Hasse, Alex M. Roman, Carrie A. Farberow, Joshua A. Schaidle, and Adam Holewin-
ski. Electrochemical Routes for the Valorization of Biomass-Derived Feedstocks: From
Chemistry to Application. ACS Energy Lett., 6(4):1205–1270, April 2021.

[137] Peter Nilges and Uwe Schr¨oder. Electrochemistry for biofuel generation: Production of
furans by electrocatalytic hydrogenation of furfurals. Energy Environ. Sci., 6(10):2925,
2013.

[138] Alex Rom´an. Electrochemical Upgrading of Biomass to Sustainable Chemicals: Mech-
anistic Insights into the Electrocatalytic Oxidation of Furfural and Related Compounds
- ProQuest. PhD thesis, University of Colorado, 2020.

[139] Mia D. Stankovic, Jessica F. Sperryn, Roxanna S. Delima, Connor C. Rupnow,
Michael B. Rooney, Monika Stolar, and Curtis P. Berlinguette. Electrochemical pro-
duction of methyltetrahydrofuran, a biofuel for diesel engines. Energy Environ. Sci.,
16(8):3453–3461, 2023.

[140] Peng Zhou, Si-Xuan Guo, Linbo Li, Tadaharu Ueda, Yoshinori Nishiwaki, Liang
Huang, Zehui Zhang, and Jie Zhang. Selective Electrochemical Hydrogenation of
Phenol with Earth-abundant Ni-MoO2 Heterostructured Catalysts: Effect of Oxy-
gen Vacancy on Product Selectivity. Angewandte Chemie International Edition,
62(8):e202214881, 2023.

[141] Sabyasachi Das, James E. Anderson, Robert De Kleine, Timothy J. Wallington,
James E. Jackson, and Christopher M. Saffron. Technoeconomic analysis of corn stover
conversion by decentralized pyrolysis and electrocatalysis. Sustainable Energy Fuels,
6(11):2823–2834, 2022.

[142] Yingying Li, Gen Huang, Yimin Jiang, Chongyang Ma, Yuxuan Lu, Shuangyin Wang,
and Yuqin Zou. Theoretical Insights into Facet-Dependent Activity and Selectivity of
Cu Catalysts in Electrochemical Furfural Reduction. J. Phys. Chem. C, 127(45):21989–
21998, November 2023.

142

[143] Hengzhou Liu, Deep M. Patel, Yifu Chen, Jungkuk Lee, Ting-Han Lee, Sarah D.
Cady, Eric W. Cochran, Luke T. Roling, and Wenzhen Li. Unraveling Electroreduc-
tive Mechanisms of Biomass-Derived Aldehydes via Tailoring Interfacial Environments.
ACS Catal., 12(22):14072–14085, November 2022.

[144] Amani O. Alghamdi, Abdesslem Jedidi, Saadullah G. Aziz, Kazuhiro Takanabe, and
Luigi Cavallo. Theoretical insights into dehydrogenative chemisorption of alkylaromat-
ics on Pt(1 0 0) and Ni(1 0 0). Journal of Catalysis, 363:197–203, July 2018.

[145] Arghya Banerjee and Samir H. Mushrif. Reaction Pathways for the Deoxygenation
of Biomass-Pyrolysis-Derived Bio-oil on Ru: A DFT Study using Furfural as a Model
Compound. ChemCatChem, 9(14):2828–2838, July 2017.

[146] Cecil N. M. Ouma, Phillimon M. Modisha, and Dmitri Bessarabov. Catalytic dehydro-
genation of the liquid organic hydrogen carrier octahydroindole on Pt (1 1 1) surface:
Ab initio insights from density functional theory calculations. Applied Surface Science,
471:1034–1040, March 2019.

[147] Christopher D. Taylor, Sally A. Wasileski, Jean-Sebastien Filhol, and Matthew Neu-
rock. First principles reaction modeling of the electrochemical interface: Consideration
and calculation of a tunable surface potential from atomic and electronic structure.
Phys. Rev. B, 73(16):165402, April 2006.

[148] Shengguang Wang, Vassili Vorotnikov, Jonathan E. Sutton, and Dionisios G. Vlachos.
Brønsted–Evans–Polanyi and Transition State Scaling Relations of Furan Derivatives
on Pd(111) and Their Relation to Those of Small Molecules. ACS Catal., 4(2):604–612,
February 2014.

[149] Meheryar R. Kasad, James E. Jackson, Scott Calabrese Barton, Calvin Mukarakate,
Uwe Schr¨oder, and Christopher M. Saffron. Electrocatalytic hydrotreatment of bio-oil:
Exploring interactions between functional groups. In Preparation, 2025.

[150] Meng-Yuan Chen, Yao-Bing Huang, Huan Pang, Xin-Xin Liu, and Yao Fu. Hy-
drodeoxygenation of lignin-derived phenols into alkanes over carbon nanotube sup-
ported Ru catalysts in biphasic systems. Green Chemistry, 17(3):1710–1717, 2015.

[151] C Mirodatos. Steady-state and isotopic transient kinetics of benzene hydrogenation

on nickel catalysts. Journal of Catalysis, 105(2):405–415, June 1987.

[152] F Montilla, F Huerta, E Morallon, and J. L Vazquez. Electrochemical behaviour of
benzene on platinum electrodes. Electrochimica Acta, 45(25):4271–4277, August 2000.

[153] Yeohoon Yoon, Roger Rousseau, Robert S. Weber, Donghai Mei, and Johannes A.
Lercher. First-Principles Study of Phenol Hydrogenation on Pt and Ni Catalysts in
Aqueous Phase. J. Am. Chem. Soc., 136(29):10287–10298, July 2014.

[154] Jacob Anibal. UNDERSTANDING THE ELECTROCHEMICAL REDUCTION AND
COUPLING OF BIOMASS DERIVED CARBONYL SPECIES. PhD thesis, Univer-
sity of Delaware, 2021.

143

[155] Qiuju Fu, Wenpeng Xie, Lingzhi Yang, Liting Yan, and Xuebo Zhao. Recent advances
in electrocatalytic upgrading of biomass-derived furfural. Chemical Engineering Jour-
nal, 485:150083, April 2024.

[156] Fahime Khouini and Morteza Vahedpour. New Platinum Bimetallic Surfaces,
M/Pt(111) and Pt/M/Pt(111) (M = Mo, Zn, and Co), for Furfural Adsorption. J.
Phys. Chem. C, 128(22):8961–8973, June 2024.

[157] Martin J. Taylor, Li Jiang, Joachim Reichert, Anthoula C. Papageorgiou, Simon K.
Beaumont, Karen Wilson, Adam F. Lee, Johannes V. Barth, and Georgios Kyriakou.
Catalytic Hydrogenation and Hydrodeoxygenation of Furfural over Pt(111): A Model
System for the Rational Design and Operation of Practical Biomass Conversion Cat-
alysts. J. Phys. Chem. C, 121(15):8490–8497, April 2017.

[158] M. Soledad Zanuttini, Martin Gross, Gustavo Marchetti, and Carlos Querini. Furfural
hydrodeoxygenation on iron and platinum catalysts. Applied Catalysis A: General,
587:117217, October 2019.

[159] Bo zhao, Mengyuan Chen, Qingxiang Guo, and Yao Fu. Electrocatalytic hydrogenation
of furfural to furfuryl alcohol using platinum supported on activated carbon fibers.
Electrochimica Acta, 135:139–146, July 2014.

[160] James Akinola, Isaiah Barth, Bryan R. Goldsmith, and Nirala Singh. Adsorption
Energies of Oxygenated Aromatics and Organics on Rhodium and Platinum in Aqueous
Phase. ACS Catal., 10(9):4929–4941, May 2020.

[161] S¸. C. B˘adescu, P. Salo, T. Ala-Nissila, S. C. Ying, K. Jacobi, Y. Wang, K. Bed¨urftig,
and G. Ertl. Energetics and Vibrational States for Hydrogen on Pt(111). Phys. Rev.
Lett., 88(13):136101, March 2002.

[162] K Christmann, G Ertl, and T Pignet. Adsorption of hydrogen on a Pt(111) surface.

Surface Science, 54(2):365–392, February 1976.

[163] Griffin Ruehl, S. Elizabeth Harman, L´ıney ´Arnad´ottir, and Charles T. Campbell. Ace-
tonitrile Adsorption and Adhesion Energies onto the Pt(111) Surface by Calorimetry.
ACS Catal., 12(1):156–163, January 2022.

[164] Giacomo Marchioro, Samuel Young, Gvina Bulgarit, and Raz-Hemo. Giacomomar-
chioro/PyEnergyDiagrams: New features and pypi. Zenodo, September 2022.

[165] Gianluca Prandini, Antimo Marrazzo, Ivano E. Castelli, Nicolas Mounet, and Nicola
Marzari. Precision and efficiency in solid-state pseudopotential calculations. npj Com-
put Mater, 4(1):1–13, December 2018.

[166] Stefan Grimme, Jens Antony, Stephan Ehrlich, and Helge Krieg. A Consistent and Ac-
curate Ab Initio Parametrization of Density Functional Dispersion Correction (DFT-D)
for the 94 Elements H-Pu. The Journal of chemical physics, 132:154104, April 2010.

144

[167] E.A. Owen, L. Pickup, and J.O. Roberts. Lattice constants of five elements possess-
ing hexagonal structure. Zeitschrift fuer Kristallographie, Kristallgeometrie, Kristall-
physik, Kristallchemie (-144,1977), 91:70–76, 1935.

[168] E.A. Owen and E.W. Roberts. The crystal parameters of osmium and ruthenium at
different temperatures. Zeitschrift fuer Kristallographie, Kristallgeometrie, Kristall-
physik, Kristallchemie (-144,1977), 96:497–498, 1937.

[169] Y. Urashima, T. Wakabayashi, T. Masaki, and Y. Teresaki. Ruthenium, a new mineral
from horokanai, hokkaido, japan. Mineralogical Journal (Japan), 7:438–444, 1974.

[170] Charles Kittel. Introduction to Solid State Physics. Wiley, Hoboken, NJ, 8. ed., [repr.]

edition, 20.

[171] M. Methfessel and A. T. Paxton. High-precision sampling for Brillouin-zone integration

in metals. Phys. Rev. B, 40(6):3616–3621, August 1989.

[172] Rolf Sander. Compilation of Henry’s law constants (version 5.0.0) for water as solvent.

Atmospheric Chemistry and Physics, 23(19):10901–12440, October 2023.

[173] Nina Ni and Samuel H. Yalkowsky. Prediction of Setschenow constants. International

Journal of Pharmaceutics, 254(2):167–172, March 2003.

[174] S. Weisenberger and A. Schumpe. Estimation of gas solubilities in salt solutions at

temperatures from 273 K to 363 K. AIChE Journal, 42(1):298–300, January 1996.

[175] Chemeo - chemical and physical properties. https://www.chemeo.com.

[176] Christina M. Wark

and
https://github.com/scbgroup/CMW diss code.git, 2025.

Scott Calabrese Barton.

CMW diss code.

[177] Colin F. Dickens, Charlotte Kirk, and Jens K. Nørskov. Insights into the Electrochemi-
cal Oxygen Evolution Reaction with ab Initio Calculations and Microkinetic Modeling:
Beyond the Limiting Potential Volcano. J. Phys. Chem. C, 123(31):18960–18977, Au-
gust 2019.

[178] Kaiyue Zhao, Ningyao Xiang, Yu-Qi Wang, Jinyu Ye, Zihan Jin, Linke Fu, Xiaoxia
Chang, Dong Wang, Hai Xiao, and Bingjun Xu. A molecular design strategy to enhance
hydrogen evolution on platinum electrocatalysts. Nat Energy, April 2025.

[179] Tomas Kucera, Karel Urban, and Stavroula Ragkou. Healing of cavitary bone defects.

Eur J Orthop Surg Traumatol, 22(2):123–128, February 2012.

[180] Sulaiman Mad-Ali, Soottawat Benjakul, Thummanoon Prodpran, and Sajid Maqsood.
Characteristics and gelling properties of gelatin from goat skin as affected by drying
methods. J Food Sci Technol, 54(6):1646–1654, May 2017.

145

[181] Batzaya Byambaa, Nasim Annabi, Kan Yue, Grissel Trujillo-de Santiago, Mario Mois´es
Alvarez, Weitao Jia, Mehdi Kazemzadeh-Narbat, Su Ryon Shin, Ali Tamayol, and Ali
Khademhosseini. Bioprinted Osteogenic and Vasculogenic Patterns for Engineering 3D
Bone Tissue. Adv. Healthcare Mater., 6(16):1700015, August 2017.

[182] Zhen Jiang, Broden Diggle, Ming Li Tan, Jekaterina Viktorova, Christopher W Ben-
nett, and Luke A. Connal. Extrusion 3D Printing of Polymeric Materials with Ad-
vanced Properties. Advanced Science, 7(17):2001379, 2020.

[183] Stephanie Knowlton, Chu Hsiang Yu, Fulya Ersoy, Sharareh Emadi, Ali Khademhos-
seini, and Savas Tasoglu. 3D-printed microfluidic chips with patterned, cell-laden
hydrogel constructs. Biofabrication, 8(2):025019, June 2016.

[184] Sean V Murphy and Anthony Atala. 3D bioprinting of tissues and organs. Nat Biotech-

nol, 32(8):773–785, August 2014.

[185] Guoliang Ying, Nan Jiang, Cunjiang Yu, and Yu Shrike Zhang. Three-dimensional bio-
printing of gelatin methacryloyl (GelMA). Bio-des. Manuf., 1(4):215–224, December
2018.

[186] Ahmet Erdem, Mohammad Ali Darabi, Rohollah Nasiri, Sivakoti Sangabathuni,
Yavuz Nuri Ertas, Halima Alem, Vahid Hosseini, Amir Shamloo, Ali S. Nasr, Samad
Ahadian, Mehmet R. Dokmeci, Ali Khademhosseini, and Nureddin Ashammakhi. 3D
Bioprinting of Oxygenated Cell-Laden Gelatin Methacryloyl Constructs. Adv. Health-
care Mater., 9(15):1901794, August 2020.

[187] Jingyi Liu, Liang Li, Hairui Suo, Mengling Yan, Jun Yin, and Jianzhong Fu. 3D
printing of biomimetic multi-layered GelMA/nHA scaffold for osteochondral defect
repair. Materials & Design, 171:107708, June 2019.

[188] Paola Occhetta, Roberta Visone, Laura Russo, Laura Cipolla, Matteo Moretti, and
Marco Rasponi. VA-086 methacrylate gelatine photopolymerizable hydrogels: A para-
metric study for highly biocompatible 3D cell embedding: VA-086 GelMA Photopoly-
merizable Hydrogels. J. Biomed. Mater. Res., 103(6):2109–2117, June 2015.

[189] Bioluminescent Firefly Luciferase Assays. https://www.sigmaaldrich.com/US/en/technical-
documents/technical-article/cell-culture-and-cell-culture-analysis/imaging-analysis-
and-live-cell-imaging/firefly-luciferase-assays, 2023.

[190] Claudia Coronnello, Rosalia Bus`a, Luca Cicero, Albert Comelli, and Ester Badami. A
Radioactive-Free Method for the Thorough Analysis of the Kinetics of Cell Cytotoxi-
city. J Imaging, 7(11):222, October 2021.

[191] Alexander K. Nguyen, Peter L. Goering, Rosalie K. Elespuru, Srilekha Sarkar Das,
and Roger J. Narayan. The Photoinitiator Lithium Phenyl (2,4,6-Trimethylbenzoyl)
Phosphinate with Exposure to 405 nm Light Is Cytotoxic to Mammalian Cells but Not
Mutagenic in Bacterial Reverse Mutation Assays. Polymers (Basel), 12(7):1489, July
2020.

146

[192] Gregory R. Fulmer, Alexander J. M. Miller, Nathaniel H. Sherden, Hugo E. Got-
tlieb, Abraham Nudelman, Brian M. Stoltz, John E. Bercaw, and Karen I. Gold-
berg. NMR Chemical Shifts of Trace Impurities: Common Laboratory Solvents, Or-
ganics, and Gases in Deuterated Solvents Relevant to the Organometallic Chemist.
Organometallics, 29(9):2176–2179, May 2010.

[193] John Josse and William F. Harrington. Role of pyrrolidine residues in the structure
and stabilization of collagen. Journal of Molecular Biology, 9(2):269–287, August 1964.

[194] NIH/3T3 - CRL-1658 | ATCC. https://www.atcc.org/products/crl-1658.

[195] Aashutosh Mistry, Alejandro A. Franco, Samuel J. Cooper, Scott A. Roberts, and
Venkatasubramanian Viswanathan. How Machine Learning Will Revolutionize Elec-
trochemical Sciences. ACS Energy Lett., pages 1422–1431, March 2021.

[196] Annabel L. Butcher, Ching Theng Koh, and Michelle L. Oyen. Systematic mechan-
ical evaluation of electrospun gelatin meshes. Journal of the Mechanical Behavior of
Biomedical Materials, 69:412–419, May 2017.

[197] Yanzhong Zhang, Hongwei Ouyang, Chwee Teck Lim, Seeram Ramakrishna, and
Zheng-Ming Huang. Electrospinning of gelatin fibers and gelatin/PCL composite fi-
brous scaffolds. Journal of Biomedical Materials Research Part B: Applied Biomateri-
als, 72B(1):156–165, 2005.

[198] Neslihan Alemdar. Oxygen-Generating Photocrosslinkable Hydrogel. In Peter Ertl and
Mario Rothbauer, editors, Cell-Based Microarrays: Methods and Protocols, Methods
in Molecular Biology, pages 241–250. Springer, New York, NY, 2018.

[199] Hanna Hanna, Lluis M. Mir, and Franck M. Andre. In vitro osteoblastic differentiation
of mesenchymal stem cells generates cell layers with distinct properties. Stem Cell
Research & Therapy, 9(1):203, July 2018.

[200] Francesca M. Toma, Andrea Sartorel, Matteo Iurlo, Mauro Carraro, Pietro Parisse,
Chiara Maccato, Stefania Rapino, Benito Rodriguez Gonzalez, Heinz Amenitsch, Ta-
tiana Da Ros, Loredana Casalis, Andrea Goldoni, Massimo Marcaccio, Gianfranco
Scorrano, Giacinto Scoles, Francesco Paolucci, Maurizio Prato, and Marcella Bonchio.
Efficient water oxidation at carbon nanotube–polyoxometalate electrocatalytic inter-
faces. Nature Chem, 2(10):826–831, October 2010.

[201] Chao Wei and Zhichuan J. Xu. The Comprehensive Understanding of as an Evaluation
Parameter for Electrochemical Water Splitting. Small Methods, 2(11):1800168, 2018.

[202] G´abor I. Csonka, Adrienn Ruzsinszky, and John P. Perdew. Estimation, Computation,
and Experimental Correction of Molecular Zero-Point Vibrational Energies. J. Phys.
Chem. A, 109(30):6779–6789, August 2005.

147

APPENDIX A

SUPPLEMENTARY INFORMATION FOR CHAPTER 2: FINITE

DIFFERENCE MODEL OF ELECTROSTATIC CHANNELING IN TCA

CYCLE ENZYMES

A.1 Rate Constants

Kinetic parameters and composition are taken from experimental work by Bulutoglu, et al 16.

The reaction rate constants for the enzymes are shown in Table A.1. Substrate concentrations

are shown in Table A.2.

Table A.1: Reaction Rate Constants. 16

kcat(s−1)

KiA(µM )

KA(µM )

KB(µM )

MDH Forward

MDH Reverse

CS Recombinant

CS Mutant

31

870

88

44

0.42

15

4.6

1.7

0.13

87

7.9

1.1

0.83

33

21

7.8

For MDH forward, substrate A is NAD+ and substrate B is L-malate. For MDH reverse,

substrate A is NADH and substrate B is OAA. At CS, substrate A is OAA and substrate B

is a-CoA.

Table A.2: Substrate and Complex Concentrations. 16

Species

Concentration (mM)

aCoA

NAD+

L-malate

NADH

MDH-CS

0.1

2

1

0

1×10−4

A.2 Free Standing Enzymes

In the case where no channeling occurs, rates at each enzyme active site can be calculated

from the kinetic expressions with no reference to the transition state matrix. This calculation
is comparable to the model result of Bulutoglu et al.16, but in this case we extend the

148

simulation time and consider the possibility of reverse reaction at MDH by a variable NADH

concentration, initially set to 100 µM.

Concentration profiles for product Coenzyme A and intermediate OAA are shown in

Figure A1, for both the reversible and irreversible cases of MDH kinetics. CoA generation

is slightly faster for the mutant than for the recombinant free enzymes, consistent with

Bulutoglu et al., who only modeled the irreversible case. As expected, inclusion of the reverse

reaction suppresses CoA generation due to decreased reaction rate at MDH. The decreased

reversible rate also suppresses intermediate OAA concentrations (Fig. A.1b). In contrast

to Fig. A.1a, however, mutant OAA concentrations are higher than for the recombinant

enzymes, again due to faster mutant CS kinetics.

Lag time is estimated by fitting a straight line to the product concentration profiles of

Fig. A.1a, in the 10–20 s time frame where the generation rate is approximately steady-state.

Note that true steady state would require constant OAA concentration, which as shown in

Fig. A.1b remains transient in the 10–20 s time frame.

Table A.3: Lag Times for the Free Enzyme Cascade.

Reversible

Irreversible

Bulutoglu et al. 16

Recombinant Lag Time ( tℓ / s)

Mutant Lag Time (tℓ / s)

1.1 ± 0.008
5.4 ± 0.003

2.5

0.49 ± 0.004
3.2 ± 0.002

0.9

Due to extrapolation procedure, lag times are highly sensitive to the details of the cal-

culation. However, all of the lag times in Table A.3 are of order 1s. Also, the effect of

mutation appears to be ˜50% decrease in lag time and is consistent with the free enzyme
calculations of Bulutoglu 16. Irreversible kinetics at MDH also lead to substantially longer

lag times than reversible kinetics, due to the higher bulk OAA concentration in the case of

irreversible kinetics.

A.3 Effect of Time Scale on Lag Time

The transport time constant, τ , is an assumed parameter in the model. Xie et al. estimated
τ = 10 ps in building the Markov state model 26. The time scale of transport ultimately

149

Figure A.1: Concentration Profiles of Coenzyme A and OAA in a 20 second Simulation of
Free Enzymes.

150

Figure A.2: Effect of Transport Time Scale on Lag Time.

determines whether the system is limited by intermediate transport or by kinetics. To

identify a feasible range of transport time scales, τ was varied, from 10 ns to 10 ms, for

both recombinant and mutant complexes under MSM network model conditions, as shown

in Figure A.2.

Analysis of the change in lag time with transport time constant shows that for both
complexes, lag time is virtually unaffected by transport time scales below 1×10−7 s. Larger

time scales lead to an exponential increase in lag time. This observation holds true for both

complexes, but the recombinant complex is more sensitive to larger transport time constants.

This is attributed to larger difference between the turnover frequency at MDH and at CS for

the recombinant, making the recombinant complex more sensitive to diffusion. In either case,

large differences in rates cause stiffness in the problem. Reducing the magnitude difference

between time scales can increase the computational efficiency by decreasing the stiffness of

the system of equations. For this reason, τ = 100 ns was chosen for the finite difference

model.

151

Figure A.3: Recombinant and Mutant Path Dominant Trajectory Occupancy.

A.4 Path Dominant Trajectory Occupancy

Each path has a corresponding series of nodes that exist on a dominant trajectory. The

dominant trajectories were identified by Xie et al. to display the highest relative probability
compared to the other trajectories in the same path26. Path to path comparison against the

overall model provides insight to the preferences of OAA transport on the complex. Steady

state occupancy of each path’s dominant trajectory and the overall surface occupancy are

given in Fig A.3.

The recombinant complex has a clear preference for Paths 1 and 2 with the respective

dominant trajectories having more than triple the occupancy of OAA than Paths 3 or 4. For

the recombinant complex, combination of the 4 dominant trajectories make up most of the

overall surface occupancy, consistent with the previously reported deep free energy basins
along the dominant trajectories 26. Conversely, the mutant complex did not display as deep

of free energy basins as the recombinant, so the dominant trajectories do not contribute as

strongly to the overall occupancy. The mutant complex sees comparable occupancy of Paths

1 and 3 dominant trajectories that are higher than the occupancy of Paths 2 or 4. This

observation is consistent with the lag time observations in the path analysis (Fig. 2.8).

152

Whereas the recombinant primarily utilizes dominant and efficient paths, low relative

probabilities of the mutant’s dominant trajectories compared to the overall occupancy sug-

gests that the mutant complex utilizes less efficient means of transport. It is possible that

OAA traverses non-dominant paths on the surface of the mutant complex in addition to

those studied, but increased desorption is likely the cause of lower dominant trajectory oc-

cupancy. The increased likelihood of taking less efficient paths, especially through the bulk

instead of channeling, is the driving force behind the lowered efficiency of the mutant complex

compared to the recombinant complex.

153

APPENDIX B

SUPPLEMENTARY INFORMATION FOR CHAPTER 3: QUANTUM

CHEMICAL INSIGHT INTO ELECTROCHEMICAL UPGRADING OF

B.1 Adsorption Configurations

BIO-OIL

To establish possible adsorption configurations, we consider those in which the molecule ring

is parallel to the surface. At low coverages, furfural and guaiacol have been shown to prefer
these “ring down” configurations 83,85,102,145. On each of the platinum and ruthenium facets,

configurations are initially classified by the FF or 4-PP ring center placement. The relaxed

configurations of FF and 4PP on each surface are shown in Figs B.1-B.4 (FF) and Figs B.5-

B.8 (4-PP). The configurations in these figures are labeled by their starting configuration.

Both FF and 4-PP on Pt and Ru generally relax to configurations where the ring is either in

a bridge or hollow. Nearly all configurations maintain a ring-down configuration where the

ring is bound to the surface and the substituent group(s) turn up away from the surface.

B.2 Reaction Pathways

For each hydrogenation step in the two pathways, possible intermediate states were explored.

A decision tree was utilized to decide the proceeding step. For example, three possible in-

termediate steps exist for the first hydrogenation of FF to FA. Of those three options, each

relaxed on the surface, the state with the lowest energy was deemed the most probable

intermediate. For steps involving multiple hydrogenation steps, the potential second inter-

mediate states are only those from the lowest energy first hydrogenation intermediate. As

such, the possibilities for the third hydrogenation are determined from the lowest energy

second intermediate state, and so on. The relaxed configurations in the pathway of FF to

THFA via FA on each surface are provided in Figs B.9-B.12. The relaxed pathway states of

4PP to 4PCOL on each surface are shown in Figs B.13-B.16.

154

Figure B.1: Relaxed positions of Furfural on Pt(100) by starting configuration label.

155

Figure B.2: Relaxed positions of Furfural on Pt(110) by starting configuration label.

156

Figure B.3: Relaxed positions of Furfural on Pt(111) by starting configuration label.

157

Figure B.4: Relaxed positions of Furfural on Ru by starting configuration label.

158

Figure B.5: Relaxed positions of 4-Propylphenol on Pt(100) by starting configuration label.

159

Figure B.6: Relaxed positions of 4-Propylphenol on Pt(110) by starting configuration label.

160

Figure B.7: Relaxed positions of 4-Propylphenol on Pt(111) by starting configuration label.

161

Figure B.8: Relaxed positions of 4-Propylphenol on Ru by starting configuration label.

Figure B.9: Reaction path intermediate states in the transformation of FF to THFA via FA
on Pt(100).

162

Figure B.10: Reaction path intermediate states in the transformation of FF to THFA via
FA on Pt(110).

Figure B.11: Reaction path intermediate states in the transformation of FF to THFA via
FA on Pt(111).

163

Figure B.12: Reaction path intermediate states in the transformation of FF to THFA via
FA on Ru.

Figure B.13: Reaction path intermediate states in thee transformation of 4PP to 4PCOL on
Pt(100).

164

Figure B.14: Reaction path intermediate states in the transformation of 4PP to 4PCOL on
Pt(110).

Figure B.15: Reaction path intermediate states in the transformation of 4PP to 4PCOL on
Pt(111).

165

Figure B.16: Reaction path intermediate states in the transformation of 4PP to 4PCOL on
Ru.

B.2.1 Energies of all Reaction Intermediates

To determine the reaction pathways, the energies of various possible states for a given hydro-

genation step were compared. The one with the lowest energy was chosen as the intermediate

state for a given facet. Summaries of the data and SMILES label of all intermediate states

considered are provided in Table B.1 for FF to FA, Table B.2 for FA to THFA, and Table

B.3 for 4-PP to PC. An example of the decision tree for the FF to THFA pathway and 4PP

to 4PCOL pathway on Pt(110) are provided in Figures B.17 and Figure B.18.

Table B.1: Intermediate State Energies for the 2 step hydrogenation of FF to FA.

H

#

0

1

1

1

State

Pt(100) Energy -

Pt(110) Energy -

Pt(111) Energy -

Ru(0001) Energy -

#

FF

1

2

3

SMILES

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

O=Cc1ccco1-44472.75

[O-

-16.38

]c1ccco1

OCc1ccco1 -16.52

[O-

-16.59

]c1ccco1

-44452.81

-15.62

-16.60

-16.56

-44482.82

-16.13

-16.54

-16.65

-130481.84

-15.88

-15.96

-15.84

166

Figure B.17: Reaction Intermediate Decision Tree for FF to THFA on Pt(110).

Figure B.18: Reaction Intermediate Decision Tree for 4PP to 4PCOL on Pt(110).

167

H

#

2

3

3

3

3

4

4

4

4

4

4

5

5

5

5

6

0

1

1

Table B.2: Intermediate State Energies for the 4 step hydrogenation of FA to THFA.

Pt(100) Energy -

Pt(110) Energy -

Pt(111) Energy -

Ru(0001) Energy -

SMILES

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

-44485.09

-15.94

-44515.93

-15.54

-130513.70

-16.43

-16.63

-15.63

-16.19

-32.59

-15.94

-16.20

-15.46

-32.18

OCc1ccco1

-44505.45

OCC1=C[CH-

-15.89

]CO1

OCC1=CC[CH-

-15.94

]O1

OC[C-

]1CC=CO1

-15.80

OCC1[CH-

-15.14

]C=CO1

OCC1=CCCO1 -32.10

OC[C-

]1C[CH-

]CO1

OCC1[CH-

][CH-]CO1

State

#

FA

1

2

3

4

5

6

7

8

9

OC[C-

-32.18

-32.74

-31.70

]1CC[CH-

]O1

OCC1[CH-

-31.65

-32.47

]C[CH-]O1

OCC1CC=CO1

OC[C-

-48.36

10

11

]1CCCO1

12

OCC1[CH-

]CCO1

13

OCC1C[CH-

]CO1

-47.83

-47.83

-32.16

-48.36

-48.28

14

OCC1CC[CH-

-48.23

-48.43

]O1

THFA OCC1CCCO1

-16.19

-16.13

-16.18

-32.34

-31.53

-31.66

-48.72

-48.60

Table B.3: Intermediate State Energies for the hydrogenation of 4-PP to 4-PCOL and PC.

State

Pt100 Energy -

Pt110 Energy -

Pt111 Energy -

Ru0001 Energy -

H #

#

SMILES

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

4-PP

CCCc1ccc(O)cc1 -44839.58

CCCC1=C[CH-

-15.59

]C(O)C=C1

1

2

-44819.38

-15.80

-44849.03

-16.14

-130848.12

-16.52

CCCC1=CCC-

-15.08

-15.65

Not stable

-16.26

C=C1

168

Table B.3: (cont’d)

State

Pt100 Energy -

Pt110 Energy -

Pt111 Energy -

Ru0001 Energy -

H #

#

SMILES

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

1

1

1

1

2

2

2

2

2

2

2

2

2

2

2

2

2

2

2

3

-15.93

-15.95

-15.28

-15.76

-14.70

-16.15

-15.90

-15.65

Not stable

-15.71

Not stable

-15.27

CCC[C-

-15.32

]1C=CC(O)=CC1

-15.79
CCCC1C=CC(O)=C[CH-

]1

CCCC1=CC=C(O)[CH-

-15.16

]C1

-15.23
CCCC1=CC=C(O)C[CH-

]1

CCCC1=CCC(O)C=C1

CCC[C-

]1C=CC(O)[CH-

]C1

-31.63

-32.39

-31.70

-31.77

-31.56

CCCC1C=CC(O)[CH-

-31.48

-31.76

-32.21

][CH-]1

CCCC1=C[CH-

]C(O)[CH-]C1

CCCC1=C[CH-

]C(O)C[CH-]1

CCC[C-

]1C=CC-CC1

CCCC1C=CC-

-31.64

Not stable

-32.64

C[CH-]1

CCCC1=CCC-

[CH-]C1

CCCC1=CCC-

C[CH-]1

CCCC1C=CC(O)=CC1-31.57

-31.63

-31.81

CCC[C-

]1CC=C(O)[CH-

]C1

18

CCC[C-

]1CC=C(O)C[CH-

]1

19

20

21

22

CCCC1C[CH-

-31.51

-31.60

-32.56

]C(O)=C[CH-]1

CCCC1[CH-

-31.14

Not stable

Not stable

]C=C(O)C[CH-

]1

CCCC1=CC=C(O)CC1

CCCC1C=CC(O)C[CH-

-47.30

-47.46

-48.61

]1

169

Table B.3: (cont’d)

State

Pt100 Energy -

Pt110 Energy -

Pt111 Energy -

Ru0001 Energy -

H #

#

SMILES

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

3

3

3

3

3

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

CCCC1C=CC-

-47.67

CC1

CCCC1C[CH-

-47.67

][C-](O)C[CH-

]1

-48.79

-48.63

CCCC1[CH-

-47.56

Not stable

Not stable

-47.45

]CC-C[CH-]1

CCCC1C[CH-

]C(O)[CH-][CH-

]1

CCCC1CC=C(O)[CH-

]C1

CCCC1CCC(O)=C[CH-

]1

CCCC1[CH-

]CC(O)[CH-

][CH-]1

CCCC1CC=C(O)C[CH-

]1

CCC[C-

]1C=CC(O)CC1

CCCC1=CCC(O)[CH-

]C1

CCCC1=CCC(O)C[CH-

]1

CCCC1C=CC(O)[CH-

Not stable

]C1

CCC[C-]1C[CH-

]C(O)[CH-]C1

CCC[C-]1C[CH-

]C(O)C[CH-]1

CCCC1=C[CH-

]C(O)CC1

CCCC1C=CC(O)CC1-63.63

-63.81

CCCC1CCC-

-63.57

[CH-]C1

CCCC1CCC-

-62.65

C[CH-]1

CCCC1C[CH-

]C(O)[CH-]C1

CCCC1CC=C(O)CC1

170

-64.61

-64.97

Not stable

-47.58

-47.69

-46.46

-48.10

Table B.3: (cont’d)

State

Pt100 Energy -

Pt110 Energy -

Pt111 Energy -

Ru0001 Energy -

H #

#

SMILES

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

E(State 0) / eV

4

4

4

4

4

4

4

5

5

5

5

5

5

6

7

46

47

48

49

50

54

55

43

44

45

51

52

53

4-

CCCC1CCC(O)[CH-

][CH-]1

CCCC1CCC-

C[CH-]1

CCC[C-

]1CCC(O)C[CH-

]1

CCCC1C[CH-

]C(O)C[CH-]1

CCC[C-

]1CCC(O)[CH-

]C1

CCCC1C[CH-

]C(O)C[CH-]1

CCCC1[CH-

]CC(O)C[CH-]1

-62.94

-63.39

-63.30

Not stable

-62.82

CCCC1CCC(O)[CH-

-79.34

Not stable

-81.05

]C1

CCCC1CCC(O)C[CH-

-78.74

-79.60

-78.77

]1

CCCC1CCC-

CC1

CCCC1CCC(O)C[CH-

]1

CCCC1CCC(O)[CH-

]C1

CCC[C-

]1CCC(O)CC1

CCCC1CCC(O)CC1

-80.98

-79.91

PCOL

CCCC1CCCCC1

B.3 Gibbs Correction Terms

Gibbs free energy correction terms were calculated for each gas phase and the preferred

adsorbed states, including intermediate states. The total correction includes the ZPE, an

entropic contribution, and an enthalpic contribution. The following tables provide the correc-

171

tional terms for the gas phases of all stable intermediates and hydrogen, and the corrections

for each reaction state on each surface at 20°C.

Each correctional term is calculated through some combination of all of the vibrational

energies. The Gibbs free energies reported here utilize Atomic Simulation Environment’s
(ASE) Thermochemistry package to calculate the correctional terms 88. Gas phase molecule

corrections are determined by the IdealGasThermo function and corrections of an adsorbate

on some surface are calculated with the HarmonicThermo function. The zero point energy
for both types of calculations is simply a summation of vibrational frequencies 202.

EZPE =

h
2

(cid:88)

νi

i

(B.1)

Where h is Planck’s constant in eV/Hz and νi is the vibrational frequency of vibrational

mode i in Hz.

The remaining terms are more complex for gas phase calculations than adsorbed state cal-

culations due to a greater number of degrees of freedom (DOF). For the gas phase, assuming
ideal gas, the entropy term at some temperature, T , and some pressure, P , is 88,89:

T S = T (Strans + Srot + Selec + Svib)

(B.2)

Where the entropy is a combination of translational, rotational, electronic, and vibrational

terms, as defined below, and a term to account for the change in pressure from the vacuum

calculation.

(cid:32)

Strans = kB

ln

(cid:34)(cid:18) 2πmkB
h2

(cid:19)3/2 kBT
P

(cid:35)

(cid:33)

+ 5/2

The Boltzmann constant is kB and m is the mass.

Srot =





(cid:34)

kB

ln

(cid:34) √

πIAIBIC
σ

(cid:18) 8π2IkBT
σh2
(cid:19)3/2(cid:35)

(cid:20)

kB
ln
(cid:18) 8π2IkBT
σh2

(cid:19)

(cid:21)

+ 1

if

linear

(cid:35)

+ 3/2

if nonlinear

(B.3)

(B.4)

The moment(s) of inertia are represented by I, IA, IB, and IC. These are determined within

172

the analysis package based on the provided symmetry number of the molecule, σ.

Svib = kB

vibDOF
(cid:88)





i

(cid:16)

kBT

ϵi

eϵi/kBT − 1

(cid:17) − ln

(cid:16)

1 − e−ϵi/kBT (cid:17)





The vibrational entropy contribution is a function of the vibrational energies, defined by

ϵi = hνi.

Selec = kBln

(cid:104)

1 + 2

(cid:88)

(cid:105)
(spin)

The electronic entropy contribution accounts for the total spin of the molecule. Here, the

spin is either 0 if all electrons are paired, 0.5 if there is a single unpaired electron, or 1 if

there are two unpaired electrons.

The additional enthalpic correction is88,89:

H0→T =

(cid:90) T

0

CP dT =

(cid:90) T

0

(kB + CV,trans + CV,rot + CV,vib + CV,elec)dT

Where CV,trans, the translational heat capacity, is equal to 3
2 kB , CV,rot is the rotational
heat capacity equal to kB for linear gas molecules or 3
2 kB for non-linear gas molecules, the
electronic heat capacity CV,elec is assumed to be zero, and the vibrational heat capacity,

CV,vib, is a function of vibrational frequencies.

(cid:90) T

0

CV,vibdT =

vib DOF
(cid:88)

i

ϵi
eϵi/kBT − 1

Adsorbed state vibrations and free energy corrections are treated somewhat differently due

to the reduced degrees of freedom in the adsorbate. For these analyses, the vibrations

are treated harmonically and the enthalpy is assumed to be equivalent to the internal en-

ergy in order to predict the Gibbs free energy from the Helmholtz energy calculated by
HarmonicThermo88. Additionally, by considering the adsorbed state free energy correction

as a combined correction of the adsorbate and the surface, the vibrations of the surface can

be ignored because they will cancel out in any reaction change in free energy calculation.

The entropic and enthalpic contributions are calculated for the adsorbate in a similar

173

way to those calculated for the gas phase. The primary difference is that translational and

rotational contributions are assumed to be negligible, so only the vibrational contributions

are considered. The Helmholtz energy is calculated from the internal energy and entropic
vibrational contribution for a given adsorbed state, j 88.

Fj = Ej + EZPE +

(cid:90) T

0

CV,vibdT − T Svib

The Gibbs free energy is then a function of the Helmholtz energy and a pressure-volume

term. The volume is assumed to be the cell size from the geometric optimization and the

pressure can be any arbitrary value.

Gj = Fj + P V

B.3.1 Gas Phase Corrections

Table B.4: Gibbs Corrections of Gas Phase Molecules.

Molecule

EZPE / eV

TS / eV ∆H0→T / eV

Total Correction / eV

FF

FA

THFA

4PP

4PCOL

PC

H2

2.10

2.71

3.93

5.00

6.85

6.74

0.27

0.98

1.04

1.10

1.27

1.34

1.25

0.40

0.18

0.21

0.23

0.30

0.34

0.30

0.09

1.29

1.87

3.06

4.03

5.85

5.79

-0.05

B.3.2 Adsorbed H2(g) and H Corrections

Table B.5: Gibbs Corrections of H on each surface.

Surface

EZPE / eV

TS / eV

∆H0→T / eV

Total Correction / eV

Pt(100)

Pt(110)

Pt(111)

0.73

0.34

0.31

Ru(0001)

0.73

8.70e-09

8.27e-09

1.68e-04

1.49e-04

2.90e-04

2.53e-04

8.70e-09

8.27e-09

174

0.73

0.34

0.31

0.73

Table B.6: Gibbs Corrections of H2(g) on each surface.

Surface

EZPE / eV

TS / eV

∆H0→T / eV

Total Correction / eV

Pt(100)

Pt(110)

Pt(111)

1.07

0.63

0.50

Ru(0001)

1.07

1.31e-04

1.17e-04

3.70e-03

2.99e-03

7.07e-03

5.52e-03

1.31e-04

1.17e-04

1.07

0.63

0.50

1.07

B.3.3 Adsorbed State Corrections of the Furfural to THFA path-

way

Table B.7: Gibbs Corrections of FF to THFA States on Pt(100).

Adsorbate

EZPE / eV

FF

IS

FA

IS1

IS2

IS3

THFA

1.54

1.79

1.94

2.00

2.19

2.29

2.33

TS /

eV

0.70

0.77

0.96

1.21

1.24

1.51

1.92

∆H0→T / eV

Total Correction / eV

0.30

0.31

0.35

0.42

0.43

0.48

0.56

1.13

1.33

1.34

1.21

1.39

1.27

0.96

Table B.8: Gibbs Corrections of FF to THFA States on Pt(110).

Adsorbate

EZPE / eV

FF

IS

FA

IS1

IS2

IS3

THFA

1.62

1.82

1.89

2.09

2.25

2.28

2.40

TS /

eV

0.70

0.84

1.09

1.09

1.21

1.57

1.77

∆H0→T / eV

Total Correction / eV

1.22

1.32

1.19

1.40

1.47

1.21

1.18

0.29

0.33

0.39

0.40

0.43

0.50

0.55

175

Table B.9: Gibbs Corrections of FF to THFA States on Pt(111).

Adsorbate

EZPE / eV

FF

IS

FA

IS1

IS2

IS3

THFA

1.58

1.82

1.97

2.13

2.22

2.40

2.49

TS /

eV

0.75

0.83

1.01

1.04

1.22

1.67

1.80

∆H0→T / eV

Total Correction / eV

0.30

0.33

0.36

0.38

0.43

0.49

0.52

1.14

1.31

1.32

1.48

1.44

1.22

1.21

Table B.10: Gibbs Corrections of FF to THFA States on Ru(0001).

Adsorbate

EZPE / eV

FF

IS

FA

IS1

IS2

IS3

THFA

3.11

3.35

3.22

3.49

3.46

3.64

3.49

TS /

eV

0.32

0.39

0.69

0.64

0.92

0.94

1.52

∆H0→T / eV

Total Correction / eV

0.17

0.19

0.28

0.28

0.36

0.38

0.48

2.96

3.15

2.82

3.14

2.90

3.08

2.45

B.3.4 Adsorbed State Corrections of the 4-PP to 4-PCOL and PC

pathways

Table B.11: Gibbs Corrections of 4-PP to 4-PCOL States on Pt(100).

Adsorbate

EZPE / eV

4-PP

IS1

IS2

IS3

IS4

IS5

4-PCOL

3.13

3.22

3.37

3.44

3.54

3.63

3.72

TS /

eV

2.17

2.51

2.56

3.01

3.16

3.47

3.87

∆H0→T / eV

Total Correction / eV

1.64

1.45

1.56

1.26

1.24

1.07

0.82

0.67

0.73

0.75

0.83

0.86

0.92

0.97

176

Table B.12: Gibbs Corrections of 4-PP to 4-PCOL States on Pt(110).

Adsorbate

EZPE / eV

4-PP

IS1

IS2

IS3

IS4

IS5

4-PCOL

3.07

3.21

3.36

3.45

3.55

3.65

3.66

TS /

eV

2.26

2.42

2.41

2.67

3.03

3.29

3.79

∆H0→T / eV

Total Correction / eV

0.70

0.73

0.74

0.80

0.85

0.90

0.98

1.51

1.52

1.70

1.58

1.38

1.26

0.85

Table B.13: Gibbs Corrections of 4-PP to 4-PCOL States on Pt(111).

Adsorbate

EZPE / eV

4-PP

IS1

IS2

IS3

IS4

IS5

4-PCOL

3.02

3.27

3.43

3.49

3.62

3.76

3.83

TS /

eV

2.66

2.53

2.60

2.99

3.24

3.25

3.64

∆H0→T / eV

Total Correction / eV

0.74

0.73

0.75

0.82

0.86

0.87

0.93

1.11

1.48

1.58

1.32

1.24

1.38

1.12

Table B.14: Gibbs Corrections of 4-PP to PC States on Ru(0001).

Adsorbate

EZPE / eV

4-PP

IS1

IS2

IS3

IS4

IS5

PC + OH

4.82

5.20

5.29

5.45

4.75

5.05

5.73

TS /

eV

1.61

1.37

1.52

1.63

2.90

2.93

3.39

∆H0→T / eV

Total Correction / eV

3.74

4.35

4.33

4.41

2.66

2.94

3.24

0.54

0.52

0.56

0.59

0.81

0.82

0.90

177

Figure B.19: Initial CVs taken before each Hupd experiment (a) and the LSVs acquired
during addition of varied concentrations of (b) furfural, (c) furfuryl Alcohol, and (d) methyl
furan.

B.4 Organic-blocked Hydrogen Under Potential Depo-

sition

For Hupd experiments of FF, FA, andMF, an initial CV was collected before the addition

of any organic material. The initial CVs to understand unihbitied hydrogen deposition are

shown in Figure B.19a. The linear sweeps of potential with each concentration of organic

added are shown in Fig. B.19b-d.

The LSV data shown corresponds to the data used in analyzing adsorption parameters.

The window is set to focus on the area of Hupd.

178