Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Site specific heterogeneous catalysis : CO and C₂H₄ bonding with Fe-Terpyridine on a Ag(111) surface… DeJong, Miriam Dieneke 2019

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2019_may_dejong_miriam.pdf [ 12.86MB ]
Metadata
JSON: 24-1.0378374.json
JSON-LD: 24-1.0378374-ld.json
RDF/XML (Pretty): 24-1.0378374-rdf.xml
RDF/JSON: 24-1.0378374-rdf.json
Turtle: 24-1.0378374-turtle.txt
N-Triples: 24-1.0378374-rdf-ntriples.txt
Original Record: 24-1.0378374-source.json
Full Text
24-1.0378374-fulltext.txt
Citation
24-1.0378374.ris

Full Text

Site Specific Heterogeneous Catalysis:CO and C2H4 Bonding with Fe-Terpyridine on a Ag(111)Surface Studied by Scanning Probe MicroscopybyMiriam Dieneke DeJongB.Sc., University of Colorado Colorado Springs, 2015A Thesis Submitted in Partial Fulfillment of the Requirements forthe Degree ofMaster of ScienceinThe Faculty of Graduate and Postdoctoral Studies(Physics)The University of British Columbia(Vancouver)April 2019c○ Miriam Dieneke DeJong, 2019The following individuals certify that they have read, and recommend to the Facultyof Graduate and Postdoctoral Studies for acceptance, the thesis entitled:Site Specific Heterogeneous Catalysis: CO and C2H4 Bonding with Fe-Terpyridineon a Ag(111) Surface Studied by Scanning Probe Microscopysubmitted by Miriam Dieneke DeJong in partial fulfillment of the requirementsfor the degree of Master of Sciencein PhysicsExamining Committee:Sarah Burke, Physics & Astronomy, ChemistrySupervisorDouglas Bonn, Physics & AstronomyAdditional ExamineriiAbstractHeterogeneous catalysis is a key process in the manufacturing of chemicals, cleanfuels, plastics, and pharmaceuticals, and often involves gaseous reactants catalyzedby solid surfaces producing gaseous products. At the solid-gas interface, a varietyof potential nucleation sites and reaction pathways for chemical transformationsexist. To track these different pathways and distinguish the main reaction fromside reactions, it is essential to explore the surface site-by-site, at the atomicscale. Experimental techniques offering the ability to probe surfaces on a site-specific, atomic scale are limited, and rigorous knowledge of reaction mechanismsin many heterogeneous catalysis processes is lacking. In this work, Scanning ProbeMicroscopy (SPM) techniques including Scanning Tunneling Microscopy (STM),Scanning Tunneling Spectroscopy (STS), and non contact Atomic Force Microscopy(ncAFM) were utilized to examine the site-by-site surface transformation of smallgaseous adsorbates at the single-molecule and submolecular scales.SPM techniques are able to access site-specific information from surfaces on theatomic scale, which is vital to elucidating reaction mechanisms in heterogeneouscatalysis. For this work, an Fe-terpyridine coordination complex supported on aAg(111) surface was tested as a catalyst for the transformation of gaseous carbonmonoxide (CO) and ethylene (C2H4). Using SPM techniques, bonding was measuredbetween CO and Fe-terpyridine active sites, and between C2H4 and the active sitesat temperatures of T ≤ 30K. Physical structural changes in ncAFM and STMimages as well as shifts in electronic structure obtained through STS measurementswere used to characterize the bonds, and both Fe-CO and Fe-C2H4 bonds werepreceded by metastable reaction intermediates. The measured bonding activitypresented in this work demonstrates the potential of surface-supported Fe-terpyridinecomplexes as catalysts in the transformation of small hydrocarbons and oxides, andthe fundamental mechanistic insight obtained through this SPM investigation opensthe door to future optimization of these heterogeneous transformations.iiiLay SummaryBillions of dollars worth of consumable goods rely on surface chemistry for production.The manufacturing of pharmaceuticals, clean fuels, plastics, and biomaterials allinvolve chemical reactions that occur at the interface between two different phasesof matter, such as a solid catalyst surface with gaseous reactants. Despite theirpervasive presence in manufacturing processes, many reaction mechanisms in theseheterogeneous environments remain largely unknown, which makes optimizationchallenging. Research efforts are dedicated to elucidating reaction mechanisms inheterogeneous systems in order to develop more energy efficient manufacturingprocesses with less negative environmental impacts. The focus of this thesis was toprovide fundamental insight to the reaction mechanism of a surface catalyst in thepolymerization of ethylene, which is a crucial transformation for plastic production.ivPrefaceThe majority of data collection for this thesis was performed by me with the help ofErik Mårsell and Gary Tom. The code used for data analysis was written by me,Erik Mårsell, and Gary Tom. I performed all the data analysis for this work. Thiswork is original and unpublished at the time of thesis submission.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxi1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Heterogeneous Catalysis . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Probing the Surface with SPM . . . . . . . . . . . . . . . . . . . . . 31.3 Metal-Organic Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . 41.4 Fe-Terpyridine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.5 Overview of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62. Experimental Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 82.1 Scanning Tunneling Microscopy . . . . . . . . . . . . . . . . . . . . . 82.2 Tunneling Current Theory . . . . . . . . . . . . . . . . . . . . . . . . 122.3 Scanning Tunneling Spectroscopy . . . . . . . . . . . . . . . . . . . . 162.4 Non-Contact Atomic Force Microscopy . . . . . . . . . . . . . . . . . 182.5 The Microscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.6 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.6.1 Ag(111) Substrate . . . . . . . . . . . . . . . . . . . . . . . . 232.6.2 Depositing TPT and Fe . . . . . . . . . . . . . . . . . . . . . 242.6.3 Dosing CO and C2H4 . . . . . . . . . . . . . . . . . . . . . . 253. Comparison to TPPT-Fe . . . . . . . . . . . . . . . . . . . . . . . . 273.1 Adsorption Geometries . . . . . . . . . . . . . . . . . . . . . . . . . . 273.2 Coordinated Structures . . . . . . . . . . . . . . . . . . . . . . . . . 313.3 Attempted TPT Nanochain Growth . . . . . . . . . . . . . . . . . . 33vi3.4 Electronic Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . 384. Reaction with CO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424.1 CO Adsorption and Diffusion on Ag(111) . . . . . . . . . . . . . . . 424.2 CO Interactions with Fe-tpy: Visual Data . . . . . . . . . . . . . . . 444.3 CO Prebond Instability . . . . . . . . . . . . . . . . . . . . . . . . . 494.4 CO Interactions with Fe-tpy: Statistical Data . . . . . . . . . . . . . 524.5 STS on Fe-CO Bond Configuration . . . . . . . . . . . . . . . . . . . 554.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575. Reaction with C2H4 . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.1 C2H4 Adsorption and Diffusion on Ag(111) . . . . . . . . . . . . . . 585.2 C2H4 Interactions with Fe-tpy: Visual Data . . . . . . . . . . . . . . 615.3 Dissociation of Fe-C2H4 Bonds . . . . . . . . . . . . . . . . . . . . . 645.4 C2H4 Interactions with Fe-tpy: Statistics . . . . . . . . . . . . . . . 685.5 STS on Fe-C2H4 Bond Configuration . . . . . . . . . . . . . . . . . . 696. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726.2 Open Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726.3 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78A. Catalog of Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89B. Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92C. Making CO-Terminated Tips . . . . . . . . . . . . . . . . . . . . . . 96D. Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99D.1 Hand-Counted Statistics . . . . . . . . . . . . . . . . . . . . . . . . . 99D.2 Machine Learning Statistics . . . . . . . . . . . . . . . . . . . . . . . 102D.3 Applying Machine Learning to SPM . . . . . . . . . . . . . . . . . . 104viiList of TablesA.1 Proposed bond constituents for all repeatedly observed nodes. the‘-’ indicates a bonding interaction, and the ‘...’ indicates a weakerattractive interaction with no significant reordering of electronic struc-ture. The ‘n’ in 1.4 represents a range of Fe adatoms present at chaincenters with the likely value n = 3 based on previous studies. [76] It isalso possible that other molecules are bridging the chain centers withFe. The proposed bond for 4.1, ‘*’, is only based on STM topographicappearance and should be investigated more thoroughly to decipherthe bond constituents with high confidence. . . . . . . . . . . . . . . 89A.2 Descriptions for visually identifying different nodes in STM topographs. 90viiiList of Figures1.1 Simple example of heterogeneous catalysis where gaseous hydrogenand ethylene adsorb on the catalyst surface (a), hydrogen dissociatesand diffuses on the surface (b), hydrogenation of ethylene results inthe formation of ethane which desorbs from the surface (c). . . . . . 11.2 Schematic of a surface on the atomic scale showing the diversity ofpotential nucleation sites for a chemical reaction. Figure based onRef. [18] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3 Chemical structure of the organic molecule used for the metal-organiccoordinated catalyst, terpyridine-phenyl-terpyridine (TPT). . . . . . 62.1 General schematic of a scanning tunneling microscope. Piezo motorsmove the tip with picometer precision towards and around the sample,the pre-amplifier increases the It signal for feedback and data acquisi-tion. Vb is the bias applied between sample and tip, which initiatesthe exponentially sensitive tunneling current, It. κ is related to theenergy of the tunneling electrons, and d is the tip sample separation. 92.2 Example of a STM topograph with select features labeled. . . . . . . 102.3 Black solid lines trace the variations in tip sample separation (z) asthe tip scans over two different molecular adsorbates while operatingin constant current mode. Scanning over the ethylene molecule (C2H4)in (a) causes the tip to withdraw from the sample, while scanning overthe carbon monoxide molecule (CO) in (b) induces a decrease in tipsample separation. Example topographs illustrating the movement ofthe tip outlined in (a) and (b). Topograph of a single C2H4 molecule(c) and a single CO molecule (d) adsorbed on Ag(111). Althoughboth molecules are physically adsorbed on top of the substrate atoms,C2H4 images as protruding from the surface, whereas CO imagesas a depression in the topography. Vb = 20mV , It = 20pA (a) andVb = 20mV , It = 50pA (b). . . . . . . . . . . . . . . . . . . . . . . . 11ix2.4 Schematic of the one-dimensional tunneling problem applied to sep-arate tip plus barrier and sample plus barrier subsystems.  is theenergy of electrons, and z is normal to the surface with z = 0 at thetip termination and z = d at the sample plane. Oscillatory solutions(ψt and ψs) are shown in the sample and tip regions where V = 0,and exponentially decaying tails for both tip and sample extend intothe vacuum barrier region under the constant potential V = V0. zsis the position in the barrier where the two systems can be stitchedtogether, and where both ψt and ψs are nonzero. Figure based onRef. [25]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.5 Schematic showing how the tunneling process depends on the Vbapplied to the tunneling junction. Ef,t and Ef,s are the Fermi energiesof tip and sample, respectively. ρt and ρs are the density of states ofthe tip and sample, respectively.  is the energy of tunneling electrons,and Vb is the bias voltage applied to the tunneling junction. In (a),Vb = 0V and therefore no tunneling can occur. In (b), a positive biasis applied to the sample making it favorable for electrons to tunnelfrom the tip occupied states into the lowest unoccupied molecularorbital (LUMO) of the sample. A negative bias applied to the samplein (c) causes electrons to flow from the highest occupied molecularorbital (HOMO) of the sample into empty tip states. . . . . . . . . . 162.6 Line spectroscopy measurement (a) and bias slice from a grid (b) atV = 1.05V , both displaying normalized dI/dV. The topography insetin (a) shows the enclosed areas that were averaged over to generate theplot (orange=phenyl, magenta=tpy). The phenyl shows a tunnelingresonance at V = 1.05V in the line STS (a), and the correspondingbias slice (b) shows the spatially distributed local density of states isconcentrated around the phenyl of the TPT at V = 1.05V . . . . . . 172.7 General schematic of a AFM with a qPlus sensor setup for ncAFMmeasurements. A quartz tuning fork oscillates at resonance frequency,fr, and is scanned along the xy plane of the sample while held atconstant tip sample separation, d. Changes in the forces betweentip and sample cause the tuning fork to deviate from resonancefrequency. These deviations comprise the fundamental measuredquantity in ncAFM mode, ∆f . When the tip apex is terminatedby a CO molecule, as pictured here, submolecular resolution can beachieved. The gold wire collects the It signal separate from the ∆fsignal, avoiding crosstalk and allowing simultaneous data acquisitionof the It and ∆f signals. . . . . . . . . . . . . . . . . . . . . . . . . . 19x2.8 Theoretical Lennard-Jones force profile, FLJ vs. z (a), and an exper-imentally observed frequency shift curve, ∆f vs. ∆z (b). In (a), zcorresponds to the absolute tip sample separation, while in (b), ∆zcorresponds to a change in tip sample separation from the setpointposition ∆z = 0.0nm at Vb = 20mV and It = 10pA. As the tipsweeps through forward and backward curves in (b), moving towardsand then away from the sample, the ∆f curve traces a similar shapeto the short-range portion of the Lennard-Jones curve in (a). . . . . 212.9 STS on clean Ag(111) with the surface state highlighted at Vb =−67mV (a), and a topograph the Ag(111) surface after sputteringand annealing cycles (b) that shows three H2O molecules adsorbed.Vb = 20mV , It = 20pA (b). . . . . . . . . . . . . . . . . . . . . . . . 232.10 Sample preparation after TPT deposition (a) and subsequent Fedeposition (b). Fe-terpyridine nodes appear brighter in the topography(red arrows) than uncoordinated terpyridines (blue arrow). Vb =20mV , It = 20pA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242.11 Topograph after dosing C2H4 (a), and after dosing CO (b) on differentsamples. C2H4 appears as a protrusion on the substrate (teal arrows),while CO images as a depression (black arrows). Vb = 20mV (a,b),It = 20pA (a), and It = 50pA (b). . . . . . . . . . . . . . . . . . . . 263.1 Chemical structure of TPT and TPPT. Both molecules have identicaltpy groups separated by a single phenyl for TPT and two phenylspacers for TPPT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.2 Constant-height ncAFM image of an isolated TPT molecule (a) im-aged with a CO-terminated tip at Avib = 10mV and ∆z = −40pmfrom the set point Vb = 20mV and It = 10pA on Ag(111). Topograph(b) shows the same molecule in (a) imaged with the same tip andset point parameters as (a). Gas-phase chemical structure of TPT issuperimposed on topograph (b). . . . . . . . . . . . . . . . . . . . . 283.3 Adsorption angles of TPT relative to Ag(111) crystallographic direc-tions. (b) consists of two separate topographs spliced together thatwere acquired one after another without re-positioning the piezomo-tors in between images. Drift was minimized by allowing the tipto settle in position for t > 1hr before imaging. The upper half of(b) shows an isolated TPT molecule with the long molecular axislabeled, (i), and the lower half shows atomic resolution of Ag(111)(FCC lattice superimposed) with the [0 1 -1] direction labeled (grayarrow). The long molecular axis of TPT measures ∼ 15° from the [0 1-1] direction. (a) is an overview topograph encompassing the area of(b), and showing TPT molecules all adsorb at ±15° to <111>. Thestep edge was removed from (a) for clarity. Vb = −20mV , It = 30pA(a). Vb = −20mV , It = 50pA (b, upper). Vb = 5mV , It = 100pA (b,lower). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29xi3.4 Topography images after TPT depositions show molecules assemblein hydrogen-bonded rows. Overview of molecular ordering at highsurface coverage (b), and close up of two H bonded TPT moleculeswith chemical structure superimposed (a) shows the C-H...N hydrogenbonds (blue dashed lines). Vb = 20mV , It = 20pA (b). . . . . . . . . 303.5 Comparison of non-metalated (a, d, g), singly-metalated (b, e, h),and doubly-metalated (c, f, i) TPT molecules. Topographs (d, e, f)were imaged with CO-terminated tips and have chemical structuressuperimposed. Line profiles (a, b, c) show a difference in apparentheight of ∼ 0.2Å between metalated and non-metalated sides of TPTmolecules at Vb = 20mV , with the metalated side higher. Blue arrowsindicate positions of N atoms and red arrows positions of Fe atoms.Upon coordination with Fe, the tpy groups switch from the trans,transconformation (peripheral N atoms face outward) to cis,cis (N atomsface inward towards coordination center). Vb = 20mV It = 10pA (d)and (f), Vb = 20mV It = 20pA (e), Avib = 10mV ∆ z = −0.04nm(g), Avib = 50mV ∆ z = −0.04nm (h), Avib = 50mV ∆ z = 0nm (i),set point Vb = 20mV It = 10pA on Ag(111) for (g,h,i). . . . . . . . . 323.6 Three different sample preparations attempting to grow TPT-Fe3-TPTnanochains, all resulting in node diversity. White arrows in (a) and(b) point towards Fe3 nodes, which have symmetric intensity aboutthe coordination center and are identifiable by image comparison withRefs. [76] and [24]. The Fe3 node yield in (b) is ∼ 50%. Topograph(c) imaged with a CO-terminated tip shows two chain centers closeup. Laplace-filtered constant-height ncAFM image (d) of the samemolecules in (c) imaged with a CO-terminated tip, Avib = 30mV ,Vb = −.95mV , and ∆ z = −0.03nm from the set point Vb = 20mVand It = 10pA on Ag(111). The different nodes in (c) also havedifferent apparent bond lengths of 4.0Åand 3.4Åin (d). Vb = −200mVIt = 20pA (a), Vb = 20mV It = 20pA (b), and Vb = 20mV It = 10pA(c). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.7 TPT deposition on the same sample performed in three incremen-tal steps. All depositions were performed on a room temperaturesubstrate as described in Section 2.6. The sample is imaged after de-positing molecules on the substrate for t = 13s (a), after an additionalt = 13s deposition (b), and after an additional t = 2s deposition (c).Terrace coverage dramatically increases between (b) and (c) with onlyan additional t = 2s deposition, supporting the idea that moleculessaturated step edge sites before occupying terraces. A single stepedge vacancy is indicated with a white arrow in (a). Vb = 20mV ,It = 50pA for (a, b, c). . . . . . . . . . . . . . . . . . . . . . . . . . . 36xii3.8 Two TPT molecules most likely coordinate-bonded with one Feadatom mimicking the solution chemistry minimum energy configura-tion, and distorted by 2D surface constraints. Laplace-filtered ncAFMimage (b) emphasizes the twisted tpy groups where the brightestregions correspond to one pyridine from each molecule sitting ontop of the adjacent molecule’s pyridine. Vb = 20mV It = 10pA (a),Avib = 30mV ∆ z = 0.05nm (b). . . . . . . . . . . . . . . . . . . . . 373.9 Averaged normalized dI/dV of unoccupied electronic states of anisolated TPT molecule. Topography inset in (d) outlines the tpy andphenyl areas of the molecule averaged over to generate the spectrafor (d). Solid curves in (d) are the tpy and phenyl spectra with theAg(111) background spectra subtracted. Normalized bias slices (a, b,c) show the spatial distribution of the TPT LUMO states. . . . . . . 393.10 Morphology of the LUMO phenyl state upon metalation (d) showsthe LUMO increases in energy with increased number of Fe-tpy bonds.STS slices of the LUMO resonance on an isolated molecule (a), singly-coordinated TPT (b), and doubly-coordinated TPT (c) illustrate thespatial distribution of the phenyl states. Red arrows point towardsFe coordination sites. Dashed lines in (d) are Ag(111) backgroundspectra for each of the three grids (a,b,c). . . . . . . . . . . . . . . . 403.11 Normalized bias slices at Vb = −0.09V for singly-coordinated (a) anddoubly-coordinated (b) TPTs show the occupied Fe state (indicatedby red arrows). Average dI/dV for singly-metalated (c) and doubly-metalated (d) TPTs near the Fermi energy. Solid backgrounds in (c)and (d) represent the Fe/phenyl spectra, to emphasize the preciselocation of the Fe state. . . . . . . . . . . . . . . . . . . . . . . . . . 414.1 Top-view (a) and side-view (b) of the potential adsorption sites ofCO on a Ag(111) surface (Ag gray, C black, and O red). The topsite consists of the carbon atom of CO bonded to one Ag atom, whilethe bridge and hcp/fcc (hexagonal close-packed/face-centered cubic)sites have CO bonded with two and three Ag atoms, respectively.Frontier orbital diagrams of a free CO molecule (c) show the electrondensity is concentrated on the C atom for both the 5σ HOMO and2pi∗ antibonding LUMO. . . . . . . . . . . . . . . . . . . . . . . . . . 434.2 Topographs before dosing CO (a) and a similar area after dosingCO (b) both scanned with Vb = 20mV and It = 50pA. Singly anddoubly-coordinated active sites of (a) morph into the CO prebondnodes or the Fe-CO bond nodes (i) and (ii) all labeled in (b). COisolated on the surface or weakly bound to a distal, non-metalatedpyridine is also apparent after dosing. The substrate temperaturereached Tmax = 11.6K during CO dosing at PLT = 1.6E − 9mbar fort = 3min. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45xiii4.3 Comparison of the same singly Fe-CO terminated molecules (a,c,e)and doubly Fe-CO terminated molecules (b,d,f) at different biases.Vb = 20mV and It = 20pA for (a,b) and Vb = −200mV , It = 200pAfor (e,f). Line profiles (c) and (d) correspond to lines drawn ontopographs (a) and (b), respectively. Vertical solid lines in (c) and(d) separate the max and min apparent heights indicated by blueplus signs in (a) and (b), respectively. At Vb = 20mV (a,b), thesingle-CO-terminated (a) and double-CO-terminated (b) moleculesimage differently, with (d) showing a larger maximum apparent heightfor the double-termination compared with the maximum apparentheight shown in (c) of the single-termination. . . . . . . . . . . . . . 464.4 Comparison of CO prebond (a,b,c) and Fe-CO bond (d,e,f) configu-rations. STM topographs for the prebond (a) and bond (d) motifswere imaged with CO-terminated tips at Vb = 20mV and It = 10pA.Constant height ncAFM images (Laplace-filtered) for the prebond(b) and bond (e) configurations show a distinct difference in thelength of lines at the right side of the molecules, which is the loca-tion of the CO. The proposed chemical structures for the prebondand bond motifs are shown in (c) and (f), respectively. Parame-ters for (b): Vb = −0.95mV , Avib = 50mV , ∆ z = −0.01nm fromsetpoint Vb = 20mV and It = 10pA on Ag(111). Parameters for(e): Vb = −0.95mV , Avib = 20mV , ∆ z = −0.06nm from setpointVb = 20mV and It = 10pA on Ag(111). . . . . . . . . . . . . . . . . . 484.5 Direct comparison between CO prebond and bond motifs on differentends of the same TPT molecule. Topograph (a) imaged at Vb = 20mVand It = 10pA with a CO-terminated tip (scale bar 6.0Å). Laplace-filtered constant-height ncAFM image (b) taken with a CO-terminatedtip (scale bar 4.0Å). Avib = 50mV , and ∆ z = −0.01nm from thesetpoint parameters Vb = 20mV and It = 10pA on Ag(111) for (b). . 494.6 Topographs of the same area on a sample consisting of active sites (redarrows), CO prebond motifs (gray arrows), and CO bond motifs (blackarrows) imaged at setpoint parameters of Vb = 20mV , It = 20pA (a),and Vb = −200mV , It = 200pA (b). Active sites image uniformlywithin (a) and (b), Fe-CO bonds image uniformly within (b), andCO prebond motifs image non-uniformly in both (a) and (b). Thesample shown in (a) and (b) was prepped with two CO doses bothat PLT ∼ 2 · 10−9mbar for t = 3min with the substrate reachingTmax = 11.6K, followed by annealing at T = 25K for t = 5min, andannealing at T = 35K for t = 5min. More information on the ‘higherorder 4’ node is presented in the Conclusion and Appendix A. . . . . 50xiv4.7 The same molecule scanned up (a) and down (b) without movingpiezo motors or changing scanning parameters of Vb = −25mV andIt = 50pA between images. At some point during up and down scans,the CO weakly bound to the N atom of the peripheral pyridine in (a)moves to the CO prebond position imaged in (b). . . . . . . . . . . . 514.8 Summary of observed reversible (a to b) and irreversible (b to c)interactions of CO with coordinated TPTs. . . . . . . . . . . . . . . 524.9 Total CO coverage in three different experiments (a,b,c) as a functionof substrate annealing temperature. Below T = 11.6K, the maximumsubstrate temperature reached during typical gas dosages, CO was notdetected on the surface. CO was dosed at T = 11.6K once in experi-ment (a) and twice in (b). CO was not dosed in experiment (c). COcoverage positively correlates with substrate annealing temperaturein all three experiments. . . . . . . . . . . . . . . . . . . . . . . . . . 534.10 Counts of CO prebond and Fe-CO bond motifs at different experi-mental stages for three different samples (a,b,c). ‘Prepare AS’ refersto samples only consisting of active sites. ‘Dose’ stages consisted ofdosing the respective gas at PLT ∼ 2 · 10−9mbar for t = 3min. Inexperiment (c), only C2H4 was dosed on the sample. The * in (c)indicates a software reset occurred at that stage, and a different regionof the sample, several millimeters away from the previous region ofthe sample, was probed after the reset. The surface density of COwas much less than the density of active sites in the experimentalstages presented in (a-c) and competition for active sites between COmolecules was assumed negligible. . . . . . . . . . . . . . . . . . . . . 544.11 Averaged dI/dV of the bare singly-coordinated active site showinga resonance at Vb = −0.09V (a, ‘Fe SC’), and the Fe-CO bondconfiguration showing a tunneling resonance at Vb = −0.165V (c).Upon formation of the Fe-CO bond, the Fe state previously at Vb =−0.090V shifts to Vb = −0.165V . Normalized STS map of an Fe-CObonded molecule at Vb = −0.165V (d), and a singly-coordinated TPTat Vb = −0.09V (b). All dI/dV spectra were scaled by their respectivegrid setpoint parameters. Grid setpoint parameters were It = 1nA,Vb = −280mV for the Fe-CO bonded molecule (c,d), and It = 1nA,Vb = −2V for the singly-coordinated TPT (a,b). . . . . . . . . . . . 565.1 Top-view (a) and side-view (b) showing most-probable adsorptionconfigurations for C2H4 on Ag(111) (C=black, H=teal, Ag=gray, notto scale). The di−σ configuration is the preferred adsorption site andconsists of a σ bond between each C atom of the ethylene with oneAg atom of the substrate. The slightly-higher energy pi configurationconsists of pi bonding between one Ag substrate atom and the C2H4molecule. The pi HOMO and antibonding pi∗ LUMO are drawn in (c).(i) Ref [124], (ii) Ref [127]. . . . . . . . . . . . . . . . . . . . . . . . . 59xv5.2 Topographs mapping the same area of a sample containing TPT, Fe,C2H4, and trace CO after C2H4 deposition at PLT ∼ 8.4x10−9mbarfor t = 2.5min with Tmax = 11.1K (a), after annealing at T = 20Kfor t = 5min (b), and after annealing at T = 25K for t = 5min(c). Difference topographs, (a-b) and (b-c), highlight small adsorbatemovement between images at different annealing steps. The minimumand maximum intensities of the difference topographs show slightC2H4 movement in (a-b), and increased movement in (b-c). Thediffusion length of C2H4 on Ag(111) is large relative to the surfacedensity of active sites between 20K < T ≤ 25K for t = 5minsample annealing. Arrows track similarities and differences betweenthe frames with a key in (a-b). Vb = 20mV and It = 20pA (a,b,c). . 605.3 Topograph after dosing sample with ethylene at PLT ∼ 8.4x10−9mbarfor t = 2.5min with Tmax = 11.1K (a) and topograph after annealingthe same sample at T = 30K for t = 5min (b). Arrows tracksimilarities and differences to the active sites pre and post anneal.The difference topograph (b-a) highlights these changes. Topograph(a) is exactly the same as topograph (a) in Fig 5.2. Bare tpy-Feactive sites (red arrows, a) morph into C2H4 prebond (light purplearrows, b), Fe-C2H4 bond (dark purple arrow, b), and CO prebond(additional gray arrow, b) motifs after the anneal. Additionally, freeethylene molecules on the surface (teal arrows, a) are attracted tonitrogens of uncoordinated tpy groups (teal arrows, b) after annealing.An interaction potentially consisting of Fe-CO and C2H4 (pink arrow,b) is also labeled. Vb = 20mV , It = 20pA (a,b). . . . . . . . . . . . . 625.4 Topographs of the C2H4 prebond (a) and bond (d) configurationsimaged at Vb = 20mV and It = 10pA. Laplace-filtered ncAFMimages of the C2H4 prebond (b) and bond (e) motifs imaged atconstant height with Avib = 50mV from a setpoint of Vb = 20mV andIt = 10pA on Ag(111). ∆z = −0.01nm for (b), and ∆z = −0.022nmfor (e). (a,d,b,e) were all imaged with the same CO-terminated tip.The proposed chemical structures are drawn for the prebond (c) andbond (f) configurations where it is postulated that ethylene is firstclose to the tpy-Fe center, but still bonded with the substrate in theprebond configuration, and then, after overcoming an energy barrier,ethylene binds with the metal center in the bond configuration. Sheartransformation performed on (b) to compensate for drift. . . . . . . 63xvi5.5 Series of topographs (a,b,c,d,e,f,g) acquired consecutively at the sameregion of a sample with setpoint parameters Vb = −25mV and It =50pA. Frames (a)-(e) were subjected to a voltage pulse after imageacquisition was complete. The pulse was applied approximately atthe location marked by the yellow ‘x’ in each relevant frame, and allpulses were Vb = −0.7V for t = 200ms. No pulse was applied afterframe (f) acquisition, and the same image (d) is repeated in bothcolumns for ease of comparison to both (d-c) and (e-d). Differencetopographs (b-a,c-b,d-c,e-d,f-e,g-f) highlight the changes from frameto frame. By applying pulses at different locations on the samplenode structure changes from bond to prebond to bare active siteswere observed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655.6 Topographs before (a) and after (b) a pulse at Vb = 2.0V for t = 20mswas applied in the upper left corner of (a) near the yellow ‘x’. ThreeFe-C2H4 bonding nodes in (a) dissociate in response to the bias pulseresulting in three additional bare tpy-Fe nodes and three additionalfree C2H4 molecules visible on the surface in (b). The differencetopograph (b-a) emphasizes these pulse induced changes. Vb = 20mV ,It = 20pA (a,b). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 675.7 Counted C2H4 prebond and Fe-C2H4 bond motifs at different stagesin a reaction. Gray bars in the background show the max substratetemperature reached corresponding to each experimental stage. The *indicates a software reset and new location on the sample being probed.As annealing temperatures increase, a shift in node terminations fromprebond to bond is observed. . . . . . . . . . . . . . . . . . . . . . . 685.8 dI/dV averaged over the bare tpy-Fe active site (a) and averaged overthe C2H4 prebond and C2H4 bond motifs (c). Background Ag(111)spectra are included for baseline comparisons. STS at different set-point conditions were rescaled by the tunnelling resistance at thesetpoint: dI/dVIt/Vb . Inset topographs show the areas averaged over ingenerating the line STS for the tpy-Fe signal (red circle, a), C2H4prebond signal (light purple circle, b), and Fe-C2H4 bond signal(dark purple circle, b, CO-terminated tip). The Fe state resonance atVb = 0− 0.09V (a) shifts upon bonding with ethylene to a resonanceat Vb = −0.31V (dark purple, b). The C2H4 prebond configuration(light purple, b) does not transmit a tunneling resonance within themeasured bias range. Normalized dI/dV maps show the spatial elec-tronic distribution of the bare active site at the Fe state energy (b),as well as the Fe-C2H4 bond motif at the Fe-C2H4 resonance energy(d). Since the prebond motif does not emit a characteristic signal, no(dI/dV)/(I/V) map of this species is included. . . . . . . . . . . . . . 70xvii6.1 Topographs of four repeatedly observed higher order structures de-noted higher order 1 through higher order 4 (a through d). COmolecules weakly bound to the uncoordinated sides of the moleculesin 2 and 4 (to the leftmost pyridine in 2 and the rightmost pyridinein 4) are dispensable in defining the coordinated-side node structures.Vb = 20mV , It = 20pA (a-d). . . . . . . . . . . . . . . . . . . . . . . 736.2 STM and ncAFM images of a higher order 3 interaction. Overviewtopograph taken with a Ag-terminated tip (a) with the white boxedregion enclosing the higher order 3 node. Close-up topograph (b)acquired with a CO-terminated tip of the same enclosed node from (a).The white arrow in (b) points toward a noisy region in the topography,characteristic of this higher order 3 motif. A CO molecule is visibleon the lower, rightmost pyridine in (b). ncAFM image (c) of the samemolecule shown in (a) and (b). The constant height offset began atthe bottom of (c) at ∆z = −20pm from Vb = 20mV and It = 10pAon Ag(111) and was manually adjusted at the horizontal line in (c)to ∆z = +20pm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746.3 Topographs (a-c) show a sample after annealing to Tmax = 40K andbefore annealing at room temperature. The sample after annealing atroom temperature T ∼ 298K for t = 10min is shown in (d). Beforethe room temperature anneal, the sample contained many higherorder nodes, as pictured and labeled with arrows in (a), (b), and (c).The labeling key is inset in (d). After the anneal, new structures,potentially oligomers, were imaged on the sample and are labeledwith magenta arrows. These new structures are similar in appearanceto ethylene, but have apparent heights twice as large. . . . . . . . . 756.4 Topographs comparing an ethylene molecule (a) and a potentialoligomer (b). Several oligomers were observed on a sample thatcontained many higher order interactions (see Appendix A 4.1-4.4),and was subsequently annealed at room temperature, T ∼ 298K, fort = 10min. The height profiles of the ethylene (c) and oligomer (d)show the oligomer is more than twice the apparent height than theethylene under the same bias voltage, Vb = 20mV , It = 50pA (a,b). . 76A.1 Nodes observed throughout experiments performed for this thesis.Node 4.2 and 4.4 are shown with a single CO molecule weakly bondedto the pyridine of an uncoordinated tpy group. The CO moleculesare not defining elements of the node structures shown in 4.2 and 4.4. 91xviiiB.1 Fourier transforms of averaged dI/dV at the phenyl LUMO of a non-metalated TPT before filtering (a), and after lowpass filtering (b).Average dI/dV of the same data without filtering (c) and with lowpassfiltering (d). The dashed vertical line in (c) and (d) is located atVb = 1.05V and shows the phenyl LUMO remains unshifted by thefiltering process. No smoothing has been applied to any data in thisfigure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93B.2 Comparison of the effects of gaussian smoothing over 15 bias pixels atdifferent stages in data processing. Data is averaged over the phenylstate of a bare TPT molecule grid and lowpass filtered at a passbandof 0.45 normalized frequency (a). Then the data is gaussian smoothedeither before normalization only (b), after normalization only (c), orboth before and after normalization (d). Dashed line in (a), (b), (c),and (d) marks the phenyl state at 1.05V and shows the state does notchange in energy with smoothing when compared to the raw, filtereddata (a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95C.1 Frequency shift with respect to changing z tip sample separation. z =0 defines the tip sample separation at the setpoint scanning parametersVb = 20mV and It = 10pA. Forward spectroscopy measurementsare represented with black lines and the backward measurements areshown with red lines in (a) and (b). In (a) the tip starts away fromthe setpoint at zmax = 5.0Å and moves closer to the sample byz = 6.0Å to zmin = −1.0Å, and back again. In (b) the tip sweepsfrom zmax = 5.0Å to zmin = −1.5Å and back. The curve in (b)begins to resemble the minimum in the Lennard Jones potential shownin Fig 2.8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97C.2 Frequency shift as a function of z spectroscopy with z = 0 definingthe tip sample separation at the setpoint parameters Vb = 20mV andIt = 15pA. A difference in forward and backward scans of a Df(z)measurement taken over a CO molecule is an indication that the COmolecule has been picked up by the tip. . . . . . . . . . . . . . . . . 98C.3 Topograph (a) and corresponding change in tunneling current signal(b) of two CO molecules imaged with a CO-terminated tip. A well-centered CO-terminated tip will image CO molecules on the surfacewith a small symmetric bump at their center, as shown in both (a)and (b). Vb = 20mV , It = 10pA. If the bump is not centered, ncAFMimaging should not be attempted with that tip. . . . . . . . . . . . . 98D.1 Uncertainty decreases by increasing the total area of individual images(a), and by averaging count over several images (b). . . . . . . . . . 100xixD.2 Surface coverage of active sites and chain centers throughout exper-imental stages in four different experiments (a-d). Active sites andchain centers are quantities that should be conserved throughoutexperimental stages within each experiment. In (d) the active sitesand chain centers were not conserved, thus the statistical data fromthis experiment was not presented in this thesis. . . . . . . . . . . . 101D.3 STM topograph (a) and segmentation (b) of the topograph generatedby training the pixel classification tool in ilastik. Phenyls, bareterpyridine groups, Fe-terpyridine active sites, chain centers, and theAg(111) surface are separately identified in the segmented image. . . 102D.4 Four separate parts (a-d) of the segmentation shown in Fig D.3 (b).Individual terpyridine groups are shown in (a), phenyls in (b), activeFe-tpy sites in (c), and chain centers in (d). Each individual image(a-d) of the segmentation was fed through an algorithm that countsand records the number of items contained within. . . . . . . . . . . 103xxList of Abbreviations∆f Frequency shift∆z Constant height offsetAFM Atomic Force Microscope/MicroscopyAvib Vibration amplitudeAg(111) face-centered cubic silver surfaceBE Binding energyC2H4 EthyleneCO2 Carbon dioxidedI/dV Differential conductance(dI/dV)/(I/V) Normalized differential conductanceDC Doubly-coordinatedEf Fermi energyfr Resonance frequencyH2 Hydrogen gasHOMO Highest occupied molecular orbitalIt Tunneling currentLUMO Lowest unoccupied molecular orbitalLT Low temperaturencAFM non-contact Atomic Force MicroscopyNO Nitric oxideO2 OxygenPrep Preparation chamberρ Local Density of StatesSC Singly-coordinatedSPM Scanning Probe Microscope/MicroscopySTM Scanning Tunneling Microscope/MicroscopySTS Scanning Tunneling SpectroscopyT Temperaturet timeTPT Terpyridine-Phenyl-TerpyridineTPPT Terpyridine-Phenyl-Phenyl-Terpyridinetpy TerpyridineUHV Ultra high vacuumVb Bias voltagexxiChapter 1Introduction1.1 Heterogeneous CatalysisHeterogeneous catalysts play a vital role in the manufacturing of chemicals, cleanfuels, and pharmaceuticals. [1–4] They are key components in many pollutionreduction technologies, [5–10] used in the growth of semiconductor devices, [11] andare central in biomaterials engineering. [12] Despite their ubiquitous presence on anindustrial scale, fundamental understanding of reaction mechanisms in heterogeneouscatalysis is often lacking. As the industry strives to develop processes and technologieswith increased energy efficiency and decreased negative environmental impacts, [7,13] the need for fundamental understanding of heterogeneous catalysis grows.The basic definition of heterogeneous catalysis is a reaction in which the catalystand reactants are in different phases of matter, such as a solid catalyst surface andgaseous reactants. A simplified example of a heterogeneous reaction is presented inFig 1.1 where the hydrogenation of ethylene to ethane on a metal catalyst surface isshown. In Fig 1.1, first gaseous hydrogen (H2) and ethylene (C2H4) are adsorbedon the surface (a), next H2 dissociates upon adsorption (b), and finally individualH atoms bind to C2H4 forming ethane (C2H6) which desorbs from the surface (c).Figure 1.1: Simple example of heterogeneous catalysis where gaseous hydrogen andethylene adsorb on the catalyst surface (a), hydrogen dissociates and diffuses onthe surface (b), hydrogenation of ethylene results in the formation of ethane whichdesorbs from the surface (c).11.1. HETEROGENEOUS CATALYSISHaving the catalyst and products in different phases of matter is a key benefit toheterogeneous catalysis as it affords easy separability. [1, 3] Additionally, tuning thecatalyst properties has led to enhanced selectivity in reactions. [14–17]Figure 1.2: Schematic of a surface on the atomic scale showing the diversity ofpotential nucleation sites for a chemical reaction. Figure based on Ref. [18]Although Fig 1.1 depicts a surface reaction in a simplistic manner, elucidatingdetailed reaction mechanisms in practice is quite complicated due to the complexnature of the surface on the atomic scale. Fig 1.2 shows some of the many featurespresent on a nominally atomically flat crystalline surface including step edges, ahole, and a screw dislocation; all of which are potential nucleation sites for reactivity.Non-uniformity of the surface often leads to undesired side reactions and decreasedefficiency making it difficult to control behavior at the active sites. [3] Since the1920s, it has been discussed that different features on a catalyst surface influencereactivity, and that only certain features comprising ‘active sites’ are responsiblefor catalyzing the reaction. [19] Lots of experimental evidence indicates increasedcatalytic activity at step edges over terraces and that adsorbates often bind morereadily in the presence of defects on the surface, however, it is often unknownexactly which sites on a surface are the active sites. [20, 21] Thus, probing thesurface on a site-specific, atomic scale is essential to elucidate reaction mechanismsin heterogeneous catalysis. An experimental technique particularly well equippedfor this task is Scanning Probe Microscopy (SPM).21.2. PROBING THE SURFACE WITH SPM1.2 Probing the Surface with SPMSPM techniques are often employed in surface catalysis investigations because theyoffer the ability to image surfaces at the atomic scale, track diffusion patterns ofreactants in real-time, probe the local electronic structure relevant to bond formation,and image single inter- and intra-molecular bonds of reactants, catalysts, andproducts. [22–26] Other surface averaging techniques have been shown to inaccuratelydescribe behavior when interested in site-specific mechanistic information, [27]however, the SPM is a local probe that innately measures data on a site-specificbasis at the atomic scale.To probe the atomic landscape with high precision, many SPM setups operateat very low temperatures, around T ∼ 5K, which essentially freezes out any surfaceactivity, and has the advantage of stabilizing highly reactive intermediates with shortlifetimes. [28] Custom SPM systems can be designed to operate in the high pressureand temperature regimes common to many industrial catalytic processes, however,these systems are rare as many drawbacks such as increased noise often outweighthe gains. [29, 30] In low temperature systems, reactions can be incrementallyinduced with controlled in situ heating and variable temperature SPM setups. [6,30] Additionally, the probe itself in an SPM system can induce all the basic stepsin a chemical reaction at low temperatures where reactions would not typicallyoccur. Processes including voltage pulse-induced molecular dissociation, [31] scan-induced diffusion of adsorbates into nanoclusters, [23] electron-induced moleculardesorption, [32] and lateral-manipulation causing bond-formation [33] are someexamples of tip-induced chemical transformations performed in SPM experiments.These transformations proceed along energy minimums mimicking the naturalprocesses and allowing exceptional control over reaction progression. [6, 34, 35]More information on probe-induced chemical transformations as well as details onthe SPM setup used for this thesis will be presented in Chapter 2.Since their advent in the 1980s, SPM techniques have successfully elucidatedreaction mechanisms for the adsorption, diffusion, oxidation, epoxidation, and disso-ciation of a variety of small hydrocarbons and oxides on active catalyst surfaces. [31,36–40] Many chemical transformations have been induced on metallic substrates andmetal-oxides by annealing at high temperatures [4, 41] including the polymerizationof small hydrocarbons. [42, 43] Although valuable mechanistic information has beengained from studying chemical transformations on metal and metal-oxide catalystsurfaces, these transformations often exhibit poor selectivity and low efficiency. [43,44] A new class of catalysts consisting of surface-supported metal-organic complexesmay offer the selectivity necessary to improve a variety of industrial-scale catalyticprocesses.31.3. METAL-ORGANIC CATALYSTS1.3 Metal-Organic CatalystsCurrently, most industrial manufacturing processes use supported inorganic nanopar-ticles as heterogeneous catalysts. [14] These nanoparticles often vary in size andactivity making it challenging to tune reactivity and selectivity. Undesirable sidereactions are often present, which lead to catalyst degradation and decreased effi-ciency. Metal particle size has been found to directly link with reactivity where smallparticle sizes are associated with increased catalytic activity. [3, 45] This suggeststhat creating uniform, low-coordination metal sites may lead to increased efficiencyand selectivity for heterogeneous catalysts.For many years, solution-based metal-organic complexes have been successfullyemployed as homogeneous catalysts in a variety of chemical transformations, [46–49]including the selective polymerization of ethylene, [50] which was the target reactionfor the research comprising this thesis. Solution-based catalysts are easy to tuneby varying the metal or ligand constituents, and are more easily understood thanheterogeneous catalysts since each unit in a homogeneous reaction is identical andin a nominally identical environment under reaction conditions. Although solution-based catalysts are better understood, easily tunable, and effective, they do not offerease of separability and catalyst regeneration like heterogeneous catalysts do.Recently, the idea to combine metal-organic catalysts with a surface supporthas emerged as an active field of research. [51, 52] Surface-supported metal-organiccatalysts are promising candidates for high selectivity due to their ease of fabrication,uniformity, low-coordination, and tunability. By changing the metal atom in metal-organic structures, the order of molecular orbitals and spin-state of the complex canbe tuned. [53] Additionally, changing the organic ligand has been shown to controlthe oxidation state of the active metal sites in metal-organic structures. [14, 54]Oxidation state, spin structure, and bond order all influence reactivity.Self-assembly processes are often employed to grow surface-supported metal-organic complexes whereby metal and organic constituents are deposited separatelyon the surface support, and assemble into highly uniform, coordination bondedstructures. The self-assembly process for metal-organic complexes involves a delicateinterplay of deposition and diffusion rates, along with metal-organic and adsorbate-substrate interactions. [55–58] In the past decade, these underlying growth mecha-nisms have become well understood allowing the reproducible on-surface synthesisof metal-organic structures with well defined coordination geometries. [59, 60] Theprocedure used in this thesis for forming self-assembled metal-organic active sites ispresented in Section 2.6.Surface-supported metal-organic complexes have already shown promise as cata-lysts in a variety of chemical transformations. The first reported chemical transfor-mation by an active metal site in a metal-organic surface-supported complex was thedissociation of oxygen (O2) through cooperative catalytic action of two Fe2+ ions ina carboxylate bridged coordination network on Cu(001). [61] Around the same time,the binding of nitric oxide (NO) to Fe and Co metal centers in porphyrins on Ag(111)was explored and reported a noticeable weakening of the chemical bond between41.4. FE-TERPYRIDINEthe coordinated metal ion and substrate upon NO bond formation with the metalion. [62] Chemical transformation of the NO molecule was not observed and thebonding was reversed between T = 500−600K. Additionally, several studies focusedon optimizing the hydrogen evolution reaction have employed surface-supportedmetal-organic structures, [63–65] and one particular study engineered a bimetallicporphyrin network that exhibited catalytic activity two orders of magnitude higherthan single metallo-porphyrin species. [65]The capture and reduction of carbon dioxide (CO2) has also received attentionin the surface-supported metal-organic catalysis community. A network of Fe dimersbridged by BDBA1 molecules on Ag(001) exhibited a variety of binding sites for CO2near the Fe nodes, but upon binding with CO2 a structural weakening and eventualcollapse of the entire metal-organic network occurred. [66] Au coordinated phenylenediisocyanide chains on Au(111) successfully captured CO2 and self-catalyzed theaggregation of CO2 clusters into monolayers and eventually a bilayer, but did notreport any chemical conversion of CO2. [67, 68] More recent studies exhibiting O2dissociation include research on Mn coordinated porphyrins supported on Ag(111)that reported O2 cleavage and oxidation of the Mn centers from MnII to MnIII ,[69] and another study that reported O2 cleavage at V metal centers in V-DPTZ2chains on Au(001). [15]Although the current body of work shows promise for metal-organic supportedcatalysts in dissociative transformations, they are yet to be tested as catalysts inpolymerization reactions. The work in this thesis aims at testing a self-assembledmetal-organic complex supported on Ag(111) as a catalyst in the polymerizationof ethylene, and the following section provides information on the specific complexchosen for this work.1.4 Fe-TerpyridineMany metal-organic structures are known to accelerate organic transformationsin solution. [51, 70–72] Of these metal-organic complexes, terpyridine (tpy) isone of the most commonly used organic ligands due to its ability to bond with alarge number of metals forming a variety of oxidation states. [73] Terpyridine-Fecoordination complexes in particular have successfully catalyzed the polymerizationof small hydrocarbons in wet chemistry, [74] which makes the tpy-Fe complex aprime candidate for the study of ethylene polymerization on a surface support.Additionally, Fe is a less expensive alternative to other metal active sites appliedin terpyridine coordination chemistry, such as Os and Ru, [75] making Fe-basedcomplexes more attractive for scalability.In previous SPM studies, self-assembled Fe-tpy complexes have been reproduciblygrown on a Ag(111) substrate. [24, 58, 76] In those studies, an organic moleculecontaining two terpyridine end groups separated by two phenyl spacers was used as1BDBA is short for the organic molecule 4,4’-di(1,4-buta-1,3-diynyl)-benzoic acid.2DPTZ is short for the organic molecule dipyridinyltetrazine.51.5. OVERVIEW OF THESISFigure 1.3: Chemical structure of the organic molecule used for the metal-organiccoordinated catalyst, terpyridine-phenyl-terpyridine (TPT).the ligand. Upon self-assembly with Fe adatoms, linear nanochains formed consistingof organic molecules connected by Fe nodes in a head-to-head configuration. Thelinear structure of these particular tpy-Fe complexes are attractive as polymerizationcatalysts because they contain more space between chains for polymers to growthan two-dimensional metal-organic analogs. Additionally, the nanochains were heldtogether by multimetallic (Fe3) centers which should be very reactive and are onlystabilized by the surface. These multimetallic nodes are an interesting catalyst motifsince they exhibit cooperative effects, interesting spin structure, and low stability,which implies high reactivity. For all these reasons, a terpyridine-containing organicmolecule coordination bonded with Fe adatoms was chosen for the active sites inthe study of ethylene polymerization for this thesis.The organic molecule used for experiments in this thesis is shown in Fig 1.3.Three pyridine rings comprise one terpyridine group located at each end of themolecule with a single phenyl spacer in between. The organic molecule, 4’,4” ”-(1,4-Phenylene)bis(2,2’:6’,2”-terpyridine), was ordered from Sigma Aldrich and is referredto as terpyridine-phenyl-terpyridine (TPT) for short. The TPT molecule used forthis thesis is similar to that from the aforementioned studies. [24, 58, 76] Previousstudies used a molecule with two phenyl spacers whereas the molecule employed forthis thesis has only one phenyl spacer between tpy groups. A detailed comparisonbetween the two molecules will be presented in Chapter 3. Additionally, details onthe self-assembly of the TPT-Fe coordinated structures employed for this work willbe presented with sample preparation procedures in Section 2.6.1.5 Overview of ThesisThe remainder of this thesis is organized as follows: Chapter 2 details the experimen-tal apparatus, methods, and theory employed for data collection in this thesis, alongwith sample preparation procedures. Chapter 3 provides a comparison between themetal-organic catalyst used for this work and the similar complex studied by ourgroup previously for optoelectronic properties and applications. Chapter 4 providesexperimental evidence for bonding of Fe-tpy active sites with carbon monoxide,61.5. OVERVIEW OF THESISand chapter 5 presents evidence for ethylene bonding with the active sites. Finally,the conclusion in Chapter 6 contains a brief summary of the main findings fromthis thesis along with an in-depth account of open questions regarding evidence ofethylene polymerization detected in one experiment performed during this thesisresearch.7Chapter 2Experimental MethodsThe scanning tunneling microscope (STM) was invented by Gerd Binning andHeinrich Rohrer in the early 1980s at IBM. [77] Unlike optical microscopes, whichare limited to the diffraction limit of visible light, a STM employs quantum tunnelingto generate topography images of surfaces and molecules on the atomic scale. Soonafter the invention of the STM, extensions of the technique including scanningtunneling spectroscopy (STS) and atomic force microscopy (AFM) were introduced.[78] Since their advent, scanning probe techniques have proven to be powerful toolsin elucidating surface chemistry phenomena at the atomic and molecular scales.Some example studies that have furthered understanding of surface reactions includesubmolecular imaging of reaction intermediates and products, [4, 79, 80] trackingsingle molecule diffusion in real-time on surfaces, [22, 81] determining the bond orderin adsorbed molecular species, [82] and creating [83] or dissociating [84] single bondsthrough tip-adsorbate interactions. Several of these techniques were applied to thesystem studied for this thesis and those will be discussed in detail throughout thebody of text.This chapter begins with a description of a STM in Section 2.1 followed by atheoretical explanation of the tunneling current in Section 2.2. Section 2.3 and2.4 discuss STS and ncAFM techniques, respectively. A brief description of themicroscope used throughout experimentation is presented in Section 2.5, and therecipe for sample preparation is discussed in Section 2.6.2.1 Scanning Tunneling MicroscopyA simplified schematic of a STM is drawn in Fig 2.1. The basic operation ofthe microscope involves an atomically sharp metal tip approached very near toan atomically flat conductive or semi-conductive sample. The tip is driven bypiezoelectric motors with picometer precision, and typically operates at 5− 10Å tip-sample separation. When a bias voltage (Vb) is applied between the two electrodes, aresulting current flows. This current, referred to as the tunneling current (It), is the82.1. SCANNING TUNNELING MICROSCOPYFigure 2.1: General schematic of a scanning tunneling microscope. Piezo motorsmove the tip with picometer precision towards and around the sample, the pre-amplifier increases the It signal for feedback and data acquisition. Vb is the biasapplied between sample and tip, which initiates the exponentially sensitive tunnelingcurrent, It. κ is related to the energy of the tunneling electrons, and d is the tipsample separation.fundamental measured quantity in a STM, and is generated by electrons tunnelingthrough the vacuum barrier between tip and sample. It is a weak signal, on theorder of pico to nano amps, and it decreases exponentially with increasing tip sampleseparation. A pre-amplifier is an essential component in the STM design, whichamplifies the It signal so it may be transmitted and recorded by computer software.As shown in Fig 2.1, the tip of the microscope is terminated by a single atom. Thesingle atom termination combined with the picometer-precise piezoelectric motors,and the exponential dependence of the It allows the STM to achieve sub-Ångstromresolution imaging.The most common operational mode of the STM is constant current mode, andthis method of imaging was employed throughout measurements for this thesis. Inconstant current mode, the It and Vb are set as constant parameters while the tipis raster scanned along the xy plane of the sample. When the tip encounters aprotrusion, defect, or other feature in the xy plane while scanning, variation in the Itoccurs, and a feedback circuit adjusts the z position of the tip in real time to maintainthe setpoint current. This variation in z distance is recorded and used to generatetopography images of the atomic landscape. Fig 2.2 shows an example topograph92.1. SCANNING TUNNELING MICROSCOPYwith select features from the atomic landscape labeled. The dogbone-shaped organicmolecules appear as the brightest features on the surface. The uniform background isthe Ag(111) surface, and the dark triangular feature in the topograph is a depressionin the surface, with several molecules adsorbed at the edges.Figure 2.2: Example of a STM topograph with select features labeled.Intuitively, one can understand that It variations are induced by physical featuresbetween the sample and tip, however, it is important to realize that these It variationsare also directly related to the electronic structure of the tip and the sample. Inparticular, the local density of states of the sample is directly related to the tunnelingcurrent, and will be discussed in detail in Section 2.2. This electronic influence isillustrated schematically in Fig 2.3 where the change in tip sample separation (z) istraced by the black solid line as the tip scans over two different molecular adsorbateson a surface. When the tip scans over ethylene (C2H4) the It increases thus zincreases (a), but when the tip scans over carbon monoxide (CO) the It decreasesand thus z decreases (b). Example topographs (c) and (d) illustrate the tip profilesshown in (a) and (b), respectively. Even though both molecules are adsorbed abovethe surface plane, the ethylene molecule (c) appears as a protrusion, and the COmolecule (d) appears as a depression due to differences in the local density of statesof the two adsorbates. Fig 2.3 illustrates the complexity of the It, and hints that thesignal is a convolution of both physical and electronic structure between the tip andsample. This both complicates interpretation of STM topographs, as well as provides102.1. SCANNING TUNNELING MICROSCOPYFigure 2.3: Black solid lines trace the variations in tip sample separation (z) as thetip scans over two different molecular adsorbates while operating in constant currentmode. Scanning over the ethylene molecule (C2H4) in (a) causes the tip to withdrawfrom the sample, while scanning over the carbon monoxide molecule (CO) in (b)induces a decrease in tip sample separation. Example topographs illustrating themovement of the tip outlined in (a) and (b). Topograph of a single C2H4 molecule(c) and a single CO molecule (d) adsorbed on Ag(111). Although both molecules arephysically adsorbed on top of the substrate atoms, C2H4 images as protruding fromthe surface, whereas CO images as a depression in the topography. Vb = 20mV ,It = 20pA (a) and Vb = 20mV , It = 50pA (b).112.2. TUNNELING CURRENT THEORYdirect insight to electronic properties through spectroscopy measurements, as willbe shown at the end of Section 2.2. To further understand the tunneling currentconvolution, a theoretical model of the signal will be presented in the followingsection.2.2 Tunneling Current TheoryLong before the invention of the STM, planar junctions with insulating barrierswere used to investigate material properties. Theory for electron transfer through ametal-insulator-metal gap had been developed by Bardeen in the 1960s. [85] Afterthe STM’s invention, this theory was quickly applied to the tip sample tunnelingjunction by Tersoff and Hamann. [86, 87] There are different approaches to thetunneling problem utilizing approximations valid under varying conditions. All ofthese approaches generally lead to the same conclusion, that the It is related tothe local density of states (ρ) of the sample and the tip. One derivation of thisrelationship will be presented here, namely, the energy-dependent Bardeen modelutilizing a one-dimensional approximation. [25, 85] The interested reader may findother detailed descriptions, and derivations in Refs. [88], [89], and [58], among manyothers.Figure 2.4: Schematic of the one-dimensional tunneling problem applied to separatetip plus barrier and sample plus barrier subsystems.  is the energy of electrons, andz is normal to the surface with z = 0 at the tip termination and z = d at the sampleplane. Oscillatory solutions (ψt and ψs) are shown in the sample and tip regionswhere V = 0, and exponentially decaying tails for both tip and sample extend intothe vacuum barrier region under the constant potential V = V0. zs is the positionin the barrier where the two systems can be stitched together, and where both ψtand ψs are nonzero. Figure based on Ref. [25].The energy-dependent Bardeen model applied to the tunneling junction requiresmany assumptions in order to gain a fundamental picture of the It. Initial assump-tions include considering the tip plus barrier and sample plus barrier as separate,non-interacting systems, assuming a constant potential V0, and approaching thetunneling problem in one-dimension only. These assumptions are illustrated inFig 2.4. ψt and ψs are the wave functions of the tip and sample, respectively, is the energy of tunneling electrons, z is perpendicular to the sample, and zs is a122.2. TUNNELING CURRENT THEORYlocation in the vacuum barrier region where the separate ψt and ψs systems will becombined later.These initial assumptions allow one to solve the time-independent Schrödingerequation for the two separate systems resulting in oscillatory wave solutions in thetip and sample regions, with exponentially decaying tails into the barrier region.[90] These exponentially decaying solutions can be stitched together by applyingcontinuous boundary conditions resulting in known solutions in all regions of thecombined tip/barrier/sample system. Equation 2.1 and 2.2 show the exponentiallydecaying solutions of the tip and the sample inside the barrier regions withψt = ψt(z = 0)e−κz, (2.1)andψs = ψs(z = d)e−κ(z−d). (2.2)The tip sample separation is d, and κ holds the energy dependence withκ =√2m(V0 − )~(2.3)where m is the mass of a single electron, and ~ is Planck’s constant. These solutions(Equation 2.1 and 2.2) will be used to further simplify the problem later on.Subsequently, applying time-dependent perturbation theory to the one-dimensionalproblem while considering a bias applied to the system, the transition rates betweeninitial and final states can be calculated using Fermi’s Golden Rule. [91] Thiscaptures the physics of tunneling electrons between tip and sample. An expressionfor the current can be drawn from the transition rate, leading to the followingexpression for the tunneling current:It(, Vb, d, T ) =4pie~∫ ∞−∞ρt(− eVb)ρs()|M(, Vb, d)|2{f(− eVb, T )− f(, T )}d.(2.4)With ρt and ρs equal to the density of states of the tip and sample, respectively, eis the charge of an electron, and T is the temperature. The absolute value of thematrix element squared, |M(, Vb, d)|2, in Equation 2.4 arises from applying Fermi’sGolden Rule, and is related to the transition probability of a tunneling event tooccur. Finally, f(, Vb, T ) is the Fermi function wheref(, Vb, T ) =11 + exp[(− eVb)/kBT ] , (2.5)and kB is the Boltzmann constant. Equation 2.5 captures the temperature dependentpopulation distribution of the available states for tunneling.From Equation 2.4, one can see that the ρt and ρs enter into the equationsymmetrically, implying that the tunneling current is a convolution of the density ofstates of the tip and the sample. Additionally, the applied bias (Vb), temperature(T ), and tip sample separation (d) influence the It signal.132.2. TUNNELING CURRENT THEORYFurther simplifications to Equation 2.4 may be applied. If the low temperaturelimit is taken, the Fermi function becomes a step function of zero or one outside orinside the bias window, respectively, and the Fermi functions drop out of Equation 2.4.The infinite integration limits also become bounded by the window of applied biasin the low temperature approximation because tunneling transitions induced bythermal fluctuations will no longer be excited outside the bias window when T → 0.The low temperature approximation is quite valid for experiments presented in thisthesis (all performed around T = 4.3K), and results in the simplified expressionIt(, Vb, d) =4pie~∫ eVb0ρt(− eVb)ρs()|M(, Vb, d)|2d. (2.6)The tunneling current is still a function of , Vb, and d. Further simplification of Itrequires addressing the tunneling matrix elements presented in Equation 2.7.M(, V0, z) =~22m∫z=zs[ψt(, z)∂ψs(, z)∂z− ψs(, z)∂ψt(, z)∂z]dS (2.7)To derive Equation 2.7, explicit dependence on Vb was replaced by the simpleconstant potential approximation V = V0, energy conservation between tip sampletunneling transitions was assumed, only one dimensional tunneling was allowed,and the integration was set up over a separation surface S at z = zs. All of thesefeatures are products of the one dimensional tunneling problem set up in Fig 2.4.Inserting the one dimensional wave solutions from inside the barrier (Equation 2.1 andEquation 2.2) into Equation 2.7 allows evaluation and simplification of Equation 2.7resulting inM(, V0, d) =~2mκψt(0)ψs(d)Ae−2κd, (2.8)where A equals the constant area of the tunneling electrodes, the wave functions (ψtand ψs) are evaluated at constant z positions, and the energy dependence is stillheld in κ. The energy dependence dominates in the exponential factor sinceκ e−2κd, (2.9)and therefore, Equation 2.8 can be approximated asM(, V0, d) ∝ e−2κd = exp[−2d√2m(V0 − )~]. (2.10)Evaluating the matrix elements using the wave solutions to the one-dimensionalbarrier problem reduces the matrix elements, M(, Vb, T ), to a transmission function,T (, Vb, d). Additionally, the square potential energy barrier (V0) can be moreaccurately described in terms of the work functions of the sample φs and tip φtin relation to the bias Vb applied between them. Combining these factors gives anexpression for the transmission function:M(, V0, d)→ T (, Vb, d) ∝ exp[−2d√2m~2(φt + φs2+eVb2− )], (2.11)142.2. TUNNELING CURRENT THEORYwhich increases exponentially with increases bias voltage. Substituting Equation 2.11into the expression for the tunneling current (Equation 2.6) gives the further simpli-fied formIt(, Vb, d) =4pie~∫ eVb0ρt(− eVb)ρs()T (, Vb, d)d. (2.12)Here it is clear that the tunneling current is related to the local density of statesof the tip and the sample, along with the transmission factor. As mentioned inFig 2.1, It ∝ e−2κz, which aides in the sub-Ångstrom resolution achievable by theSTM, and arises from the transmission factor.To further simplify Equation 2.12, ρt can be assumed constant since the tipelectronic structure should not be changing significantly during measurements.1This implies that small contributions to the tunneling current are the product ofthe transmission factor and the density of states of the sample asdIt ∝ ρs()T (, Vb, d)d. (2.13)In the limit of small bias voltages where T (, Vb, d)→constant, and the elastic limitof tunneling electrons2 a relationship between changes in current, dIt, with respectto changing bias, Vb, can be simplified in many cases to the density of states of thesample asdItdVb∝ ρs(Vb). (2.14)Equation 2.14 carries important implications for SPM techniques. It indicates thatby varying the bias voltage and measuring the responding changes in the It, thelocal density of states of the sample can be probed. This quantity, dI/dVb (alsodenoted dI/dV ) is measured in STS mode, and details of STS measurements will bediscussed in Section 2.3.It is important to remember that Equation 2.14 is an approximation only validin the limit of low temperature, small bias, energy conservation, one dimensionaltunneling, non-interacting tip sample wave functions, and single particle picturefor both tip and sample. Although many approximations were employed to reachEquation 2.14, these approximations are well justified for many systems and allowEquation 2.14 to serve as a useful conceptual picture. Additionally, the same relation-ship was derived by Tersoff and Hamman using a a different set of approximations.[86, 87]At larger biases, the transmission factor does not reduce to a small constant,and Equation 2.14 does not hold. To probe the ρs at larger biases, a normalizationtechnique is often employed in data analysis in order to quench the backgroundeffects of the transmission function. A common normalization approach is:dIdVIV∼ ρs + smaller magnitude terms. (2.15)1The experimentalist must monitor the tip electronic structure to ensure accuracy and consistencyin measurements.2The elastic limit considers only tunneling transitions that are energetically conserved essentiallyresulting in the variable substitution → eVb.152.3. SCANNING TUNNELING SPECTROSCOPYSince the majority of data presented in this thesis falls within the small bias limit,this normalization scheme will not be discussed further here. For more informationon the normalization scheme, see Refs. [89] and [58].2.3 Scanning Tunneling SpectroscopyAs discussed in the previous section, spectroscopy measurements provide access tothe local density of states of the sample (Equation 2.14 and Equation 2.15). Toperform spectroscopy measurements, the tip is held at a fixed distance d from thesample, the constant current feedback is turned off, the Vb is swept through a desiredrange of values, and the responding It is recorded. STS measurements are eitheracquired at a single point, or at several points over an area comprising a grid.Figure 2.5: Schematic showing how the tunneling process depends on the Vb appliedto the tunneling junction. Ef,t and Ef,s are the Fermi energies of tip and sample,respectively. ρt and ρs are the density of states of the tip and sample, respectively. is the energy of tunneling electrons, and Vb is the bias voltage applied to thetunneling junction. In (a), Vb = 0V and therefore no tunneling can occur. In (b),a positive bias is applied to the sample making it favorable for electrons to tunnelfrom the tip occupied states into the lowest unoccupied molecular orbital (LUMO)of the sample. A negative bias applied to the sample in (c) causes electrons to flowfrom the highest occupied molecular orbital (HOMO) of the sample into empty tipstates.Precisely how the It relates to the ρs is pictorially explained in Fig 2.5, where thespecific electrons comprising the It are shown to be dependent on the magnitude andsign of the Vb applied to the tunneling junction as well as the local density of states.In Fig 2.5  is the energy of tunneling electrons, z is normal to the surface, Ef,t is theFermi energy of tip, and Ef,s the Fermi energy of the sample. The shaded regionsat and below the Ef represent occupied states of the tip or sample, ‘LUMO’ labelsthe lowest unoccupied molecular orbital of the sample, and ‘HOMO’ the highestoccupied molecular orbital. The ‘HOMO’ and ‘LUMO’ labeling conventions arisefrom molecular orbital theory in chemistry and are useful in the context of probingmolecular electronic structure.162.3. SCANNING TUNNELING SPECTROSCOPYFig 2.5 (a) shows the tunneling junction before a bias voltage is applied, andno tunneling can occur. After a bias voltage is applied to the tunneling junction(Fig 2.5 b,c), the Fermi levels of sample and tip shift with respect to each otherand proportional to Vb. This shift provides the driving force for electrons to flowand results in a measurable It. The bias voltage is always referenced on the sampleas a convention in STM literature, though it makes no physical difference whethera positive bias is applied to the sample or a negative bias of the same magnitudeis applied to the tip.3 When Vb = + as shown in Fig 2.5 (b), electrons (negativecharge carriers) are attracted to the more positive sample electrode and tunnel fromthe tip into the LUMO states of the molecule. When Vb = − as shown in Fig 2.5 (c),electrons from the HOMO states tunnel out of the more negative sample electrodeinto the empty tip states. In this way, by switching the polarity and magnitudeof the bias, different molecular orbitals can be probed. The HOMO and LUMOpictured are the frontier molecular orbitals, however, additional orbitals further fromFermi may also be probed as long as the tunneling junction remains stable underthe applied Vb.Figure 2.6: Line spectroscopy measurement (a) and bias slice from a grid (b) atV = 1.05V , both displaying normalized dI/dV. The topography inset in (a) showsthe enclosed areas that were averaged over to generate the plot (orange=phenyl,magenta=tpy). The phenyl shows a tunneling resonance at V = 1.05V in the lineSTS (a), and the corresponding bias slice (b) shows the spatially distributed localdensity of states is concentrated around the phenyl of the TPT at V = 1.05V .Spectroscopy measurements are presented as line spectroscopy plots or biasslices in this thesis. Fig 2.6 shows examples of each of these two data types, withthe line spectroscopy plot (a) and bias slice (b) both acquired on the same bareTPT molecule.4 The line spectroscopy plot was generated by averaging spectra3Though it can make a difference in the noise spectrum.4This data is also presented in Section 3.4 in greater detail.172.4. NON-CONTACT ATOMIC FORCE MICROSCOPYover regions in a grid. This reduces noise and identifies spatially specific electronicproperties of the molecule. The regions averaged over (tpy and phenyl) are enclosedin the topography inset in (a), and a clear tunneling resonance at the phenyl isvisible at Vb = 1.05V . The bias slice in (b) shows the spatially distributed electronicdensity of states of the molecule at Vb = 1.05V , and indeed the electronic densityis concentrated around the phenyl at this bias. Bias slices in this thesis are alwayspresented as normalized dI/dV to avoid the introduction of spatial artifacts, whereasline spectroscopy plots are presented as dI/dV when Vb ≤ |400mV | (small biasapproximation), and normalized dI/dV when Vb > |400mV |. As will be discussed inthe following section, different sensors and tip terminations were used to acquiredifferent data sets throughout this thesis, but for all STS measurements, the data wasacquired with a Ag-terminated 80-20 Pt-Ir tip (0.38mm diameter) from Goodfellow.2.4 Non-Contact Atomic Force MicroscopyThe final experimental method employed for this work was non-contact Atomic ForceMicroscopy (ncAFM), which is one member in a large family of AFM techniquesthat image utilizing variations in forces between the tip and sample rather than thetunneling current. One advantage of exploiting forces is the ability to probe insulatingsurfaces. Another advantage to ncAFM for this thesis research in particular, is thecapacity to image with sub-molecular resolution when the tip is terminated with anappropriate molecule, often carbon monoxide. [92] Sub-molecular imaging allowsidentification of the internal bond structure of molecular adsorbates, and is highlyapplicable for studying surface chemical reactions.To perform ncAFM measurements a qPlus sensor was used, [93] and a generalschematic of the experimental setup is shown in Fig 2.7. The qPlus sensor consistsof a piezoelectric quartz tuning fork with a conductive tip attached to one prongand the opposite prong epoxied to a macor base. The qPlus is characterized bya stiff spring constant on the order of k ∼ 1000Nm , a resonance frequency aroundfr ∼ 25kHz, and a high quality factor of Q ∼ 25k in vacuum, which togetherproduce very small oscillation amplitudes when excited. [94] The qPlus designused in our system also consists of a gold wire that collects the It signal, which iselectrically isolated from the oscillating tuning fork signal. This allows simultaneousSTM and ncAFM measurements to be taken without significant crosstalk betweenthe signals. The gold wire is also lightweight and slack enough between connectionsthat it does not significantly influence the amplitude or frequency of oscillations ofthe tuning fork.To perform ncAFM measurements, the tuning fork is self-excited to resonancefrequency, fr, and scanned along the xy plane of the sample while maintainingconstant tip sample separation, d. Small deviations from resonance resulting fromvariations in tip sample force profiles are detected using a phase locked loop. Theseslight deflections from resonance, ∆f , comprise the fundamental measured signalin ncAFM experiments, and generate contrast in ncAFM images. The changes182.4. NON-CONTACT ATOMIC FORCE MICROSCOPYFigure 2.7: General schematic of a AFM with a qPlus sensor setup for ncAFMmeasurements. A quartz tuning fork oscillates at resonance frequency, fr, andis scanned along the xy plane of the sample while held at constant tip sampleseparation, d. Changes in the forces between tip and sample cause the tuning forkto deviate from resonance frequency. These deviations comprise the fundamentalmeasured quantity in ncAFM mode, ∆f . When the tip apex is terminated by a COmolecule, as pictured here, submolecular resolution can be achieved. The gold wirecollects the It signal separate from the ∆f signal, avoiding crosstalk and allowingsimultaneous data acquisition of the It and ∆f signals.192.4. NON-CONTACT ATOMIC FORCE MICROSCOPYin frequency are related to the sum of the forces between tip and sample (Ft,s),and in the limit of small oscillation amplitudes, the change in frequency is directlyproportional to the force gradient in the tip sample junction. [25, 94]∆f ∝ ∂Ft,s∂z(2.16)There are several forces influencing the tip sample junction including long rangeelectrostatic (Fe) and van der Waals (Fv) forces, along with short range chemicalforces (Fc) and other repulsive interactions (Foth). Ft,s from Equation 2.16 can beexpanded toFt,s ∝ Fe + Fv + Fc + Foth. (2.17)The electrostatic interaction is quite long range and arises in the presence ofa potential difference between tip and sample. This potential difference can be acombination of static charges, dipoles, different work functions, and bias voltage.[82] Modeling this system as a capacitor with the tip approximated as a sphere ona cone, [95] and the sample as a semi-infinite solid, the electrostatic force can bewritten asFe =−pi0Rz(Vb − VSP )2 (2.18)where 0 is the permittivity of free space, R is the radius of the tip, z is the distancefrom tip apex to sample, Vb is the applied bias voltage, and VSP is the surfacepotential from any other sources including surface charge distributions. For ncAFMmeasurements, electrostatic interactions are not desirable and the Vb can be tunedsuch that Fe → 0. [25]Another long range interaction known as van der Waals forces arises whenfluctuating dipoles attract one another. If the tip is approximated as a sphere andthe sample again approximated as a semi-infinite solid, the van der Waals force canbe deduced asFv = −HR6z2(2.19)where R is the radius of the tip, z the distance from tip apex to sample, and H theHamacker constant.5 [96] The van der Waals force dominates tip sample interactionswhen d > 1nm.The short range chemical force (Fc) also contributes to the tip sample interaction,and can be attractive or repulsive. Chemical forces arise due to overlap of electronwave functions, and if this overlap acts to reduce the outer shell total energy theinteraction is attractive. [25] Another interaction due to overlap of wave functions isPauli repulsion, labeled in Equation 2.17 as Foth where the ‘oth’ represents ‘other’interactions. Pauli repulsion is not technically a force, and arises due to the quantummechanical Pauli exclusion principle, however, it does highly influence the tip samplejunction at small d separations. This repulsive interaction is responsible for the highcontrast and sub-molecular resolution achievable by ncAFM. [92] Both chemical5The Hamacker constant is a material property related to the atom densities of the tip andsample along with the pair potential between them.202.4. NON-CONTACT ATOMIC FORCE MICROSCOPYforces and Pauli repulsion dominate in the small tip sample separation regime(d < 1nm) and proper theoretical treatment of these interactions requires solvingthe Schrödinger equation for a complex system in which analytic solutions scarcelyexist.Often, all four interactions described above (Equation 2.17) are modeled with asimple Lennard-Jones potential. [25, 82, 96] This potential captures the long rangeattractive forces (Fe and Fv) along with the short range chemical force (Fc), andPauli repulsion (Foth). The Lennard Jones potential is given byULJ = 4U0[(σz)12−(σz)6](2.20)where σ is the distance where ULJ = 0, U0 is the depth of the potential well, and zis the tip sample separation. [25] For practical models, σ, U0, and the exponentscan be fitted parameters. [97] The force can be calculated from the potential bytaking the negative gradient, F = −∇ULJ , resulting inFLJ = −4U0[6σ(σz)7− 12σ(σz)13]. (2.21)Figure 2.8: Theoretical Lennard-Jones force profile, FLJ vs. z (a), and an experi-mentally observed frequency shift curve, ∆f vs. ∆z (b). In (a), z corresponds tothe absolute tip sample separation, while in (b), ∆z corresponds to a change intip sample separation from the setpoint position ∆z = 0.0nm at Vb = 20mV andIt = 10pA. As the tip sweeps through forward and backward curves in (b), movingtowards and then away from the sample, the ∆f curve traces a similar shape to theshort-range portion of the Lennard-Jones curve in (a).Fig 2.8 shows the Lennard-Jones force versus distance curve (FLJ vs. z) (a), andan experimentally acquired frequency shift versus change in tip sample separationplot (∆f vs. ∆z) (b). The FLJ curve has an attractive long-range feature, apotential minimum, and a short-range repulsive feature. As z decreases, tip sampleseparation decreases. The ∆f plot (b) traces a similar shape to the FLJ curve(a). The ∆f curve is measured by starting at ∆z = 0.0nm when Vb = 20mV and212.5. THE MICROSCOPEIt = 10pA, then turning off the feedback and reducing the tip sample separationby ∆z = −0.5Å (forward curve) which measuring ∆f . Finally, ∆z is increasedagain back to ∆z = 0.0nm (backward curve) while measuring ∆f . Both forwardand backward curves comprise the same ladel-like shape as shown in the short-rangeregion of the FLJ curve in (a). The minimum in the ∆f plot is generally chosenas the ∆z offset for constant-height ncAFM imaging, right where Pauli repulsionbegins to dominate the force curve. [92]The high resolution achievable in ncAFM experiments is due to the CO-terminated tip and the procedure for making CO-terminated tips is describedin Appendix C. The orbitals of the CO molecule partially screen the otherwisemetal-metal attractive bonding that would occur without the presence of the CO.CO-terminated tips have the advantage of achieving sub-molecular resolution, butcaution is required in ncAFM image analysis. The CO molecule is flexible on thetip apex; bending and jumping as it scans along the molecule of interest. Thisflexibility has a sharpening effect on bond imaging, [97] but also leads to bondlength distortions and even questionable identification of inter-molecular Hydrogenbonds, in some cases. [98] It is expected that ncAFM measurements performed oncoordination complexes and covalent bonds are more reliable due to the increasedelectron density in these types of bonds, compared with Hydrogen bonds. [99] Forthis thesis, the identification of apparent bonds in ncAFM images is combined withSTS evidence to build a case for bonding. This case will be detailed in Chapter 4and Chapter 5.2.5 The MicroscopeAll experiments for this thesis were performed in an Omicron low temperature (LT),ultra high vacuum (UHV) SPM. The microscope is located in a state-of-the-artlow-vibration facility called the Laboratory for Atomic Imaging Research (LAIR) atUBC. The low-vibration facility has the microscope suspended on a 36 metric tonconcrete block supported by six pneumatic isolators, all on a foundation separatefrom the rest of the building. This design drastically reduces noise in measuredsignals. The microscope consists of two main chambers. The low temperature(LT) chamber, and the preparation chamber (prep). Samples were cleaned andactive sites prepared in the prep chamber where the base pressure was typicallyaround Pprep ∼ 2 · 10−10mbar. The LT chamber housed the microscope head, whichwas thermally coupled to a liquid helium bath meaning all experimental data wasobtained at T ∼ 4.3K. The base pressure in the LT chamber was typically aroundPLT ∼ 3 · 10−12mbar during data collection. Although this LT, UHV environmentprovides the conditions to maintain high purity surfaces for weeks at a time, thereare always traces of residual gases in the microscope such as hydrogen (H2), water(H2O), and carbon monoxide (CO). Gaseous C2H4 and CO were also dosed on thesample in the LT chamber, and this procedure along with preparation of active siteswill be discussed in more detail in the following section. For more information on222.6. SAMPLE PREPARATIONthe low-vibration facility and microscope setup, including sample plate design, seeRefs. [89, 100].2.6 Sample PreparationSample preparation for experiments performed in this thesis involved three mainparts: cleaning the substrate, preparing the metal-organic active sites, and dosinggaseous CO and C2H4 onto the active sites. This section will describe these threeparts in detail.2.6.1 Ag(111) SubstrateA Ag(111) substrate was used for all experiments performed for this thesis, whichwas cleaned by repeated cycles of Ar+ sputtering at UHV = 1kV for t ∼ 20 minfollowed by annealing at T = 420°C for t ∼ 20 min. The clean Ag(111) sampleshows a characteristic step in the otherwise flat tunneling signal at Vb = −67mV[101] highlighted in the line spectroscopy plot of Fig 2.9 (a). Before taking STSmeasurements on molecules, point STS measurements were always made on thesurface to check for this clear and otherwise flat surface state signal. The flat andclear signal indicated that no additional artifact features from the tip were present.Figure 2.9: STS on clean Ag(111) with the surface state highlighted at Vb = −67mV(a), and a topograph the Ag(111) surface after sputtering and annealing cycles (b)that shows three H2O molecules adsorbed. Vb = 20mV , It = 20pA (b).The topograph in Fig 2.9 (b) shows the Ag(111) surface after several cleaningcycles, and three H2O molecules are visible on the surface.6 The 2D electron gas6H20 was identified by switching the Vb and comparing topographs to the contrast shift observedfor H20 on Ag(111) in Ref. [102], along with size comparisons.232.6. SAMPLE PREPARATIONof the surface produces ripple-like features in (b), which are commonly observedthroughout topographs in this thesis imaged at small biases (|Vb| ≤ 50mV ).Residual H2O was always present on samples, due to a small partial pressure ofH2O in the vacuum chamber at all times. On average there was 1 H2O/(40nm)2on the substrate after sputtering and annealing cycles were complete, which wasdetermined from counting adsorbates in large scale images. Identifying potentialcontaminants in the vacuum chamber is essential in order to properly decipherreactants and products in a chemical reaction. For the experiments performed inthis thesis, the amount of residual H2O in the chamber was much smaller than theaverage surface density of catalytic sites,7 and therefore, it was not expected thatresidual H2O interfered appreciably with the experiments.2.6.2 Depositing TPT and FeAfter the Ag(111) was cleaned and checked in the LT, the sample was moved tothe Prep and TPT was deposited on the substrate by subliming TPT molecules atT = 220°C for t = 30sec from an Organic Molecular Beam Epitaxy cell aimed atthe substrate.8Figure 2.10: Sample preparation after TPT deposition (a) and subsequent Fedeposition (b). Fe-terpyridine nodes appear brighter in the topography (red arrows)than uncoordinated terpyridines (blue arrow). Vb = 20mV , It = 20pA.The sample was then brought into the LT to check and calculate the surfacecoverage of TPT molecules. Fig 2.10 (a) shows a topograph of the Ag(111) surface7A catalytic site is defined as a Fe-TPT coordinated node, and the average surface density was350 active sites/(100nm)28A range of deposition temperatures and times (T = 220− 240°C and t = 5− 90secs) were usedto prepare samples, however, the most common and consistent depositions used the parameterslisted in the main text.242.6. SAMPLE PREPARATIONafter TPT deposition, with four TPT molecules visible. Next, the sample was movedback to the Prep and brought up to room temperature prior to Fe deposition. Thesample was either annealed at T ∼ 32°C for t ∼ 2min and left for an additionalt = 20min, or left in the prep for t > 2hr to reach room temperature. It was necessaryfor the TPT/Ag(111) sample to be at room temperature upon Fe deposition in orderto form TPT-Fe coordinated structures. If the sample was too cold, clusters of Fewould form with very few TPT coordinated complexes.Next, Fe was deposited on the room temperature TPT/Ag(111) sample. Feadatoms were evaporated from a 99.99%+ purity Fe rod using an electron beam metalevaporator set to UHV = 800V and Iflux = 2.3nA. Fe was deposited in short burstsby alternating opening the shutter for t = 5 sec and closing for t = 15 sec, whichhelped minimize Fe cluster formation. Total open shutter time varied depending onTPT coverage with a 1Fe:1TPT desired ratio. Fe deposition time was calculatedbased on the Fe deposition rate of Q ∼ 9 · 10−4 Fe/s·nm2 at Iflux = 2.3nA.9 For atypical preparation, TPT surface coverage of ∼100 TPT/(50nm)2 required an Fedeposition time of t ∼ 30 sec shutter open at Iflux = 2.3nA for the 1Fe:1TPT ratioto be reached.Fig 2.10 shows a sample after depositing TPT (a), and after subsequent depositionof Fe (b). When no Fe is present on the sample TPT molecules predominatelyassemble in staggered rows as shown in (a), and after Fe deposition, the moleculesare imaged separately on the surface. The staggered rows are stabilized by hydrogenbonds and will be discussed in Section 3.1 in detail. The Fe coordination bondis imaged brighter in the topography (red arrows) than the bare tpy group (bluearrow), and molecules were either doubly-coordinated, singly-coordinated, or non-coordinated. Coordination structures will be described in detail in Section 3.2.2.6.3 Dosing CO and C2H4After the metal-organic complex was prepared on the sample, gaseous CO or C2H4,depending on the experiment, was dosed onto the sample while it was held in themicroscope head. Gas was leaked into the LT chamber to a pressure of PLT ∼2 · 10−9mbar, the heat shields to the microscope were opened for t = 3 min,and the maximum temperature during dosing was recorded (Tmax = 11.6K fort = 3min depositions at the listed pressure). A variety of higher pressures andshorter deposition times were explored throughout this thesis in order to optimizecoverage while minimizing Tmax during depositions, but the main dosing parameterswere those listed above. 10Fig 2.11 shows a topograph after C2H4 (a) and CO (b) depositions on differentsamples. C2H4 images as a protrusion in the topography (teal arrow, a) and CO as a9Fe deposition rate was calculated by counting TPTs before Fe depositions and coordinatebonded TPTs after Fe depositions, with 1Fe counted per coordination bond. This did not take intoaccount activity at step edges or Fe clusters, but modeled the coordination activity on terraces ofsimilar surface density well.10Pressure up to PLT ∼ 2 · 10−8mbar with a deposition time of just t = 3 sec kept the maximumtemperature to Tmax ∼ 7K and resulted in reasonably high surface coverage of gaseous adsorbates.252.6. SAMPLE PREPARATIONFigure 2.11: Topograph after dosing C2H4 (a), and after dosing CO (b) on differentsamples. C2H4 appears as a protrusion on the substrate (teal arrows), while COimages as a depression (black arrows). Vb = 20mV (a,b), It = 20pA (a), andIt = 50pA (b).depression (black arrow, b) at Vb = 20mV . After dosing, the samples were annealedat a variety of temperatures up to T = 40K for t = 5 min to induce bonding betweenthe active sites and gaseous adsorbates. All annealing steps were performed fort = 5min. Upon annealing, the heater generally would overshoot the temperature byT = 1− 2K, so a reported annealing temperature of T = 40K reached a maximumtemperature of around T ∼ 42K recorded by the temperature controller for part ofthe total t = 5min annealing time. The Details on the new structures formed uponannealing samples containing active sites and gaseous adsorbates will be discussedin Chapter 4 and Chapter 5.26Chapter 3Comparison to TPPT-FePrior to the work completed for this thesis on tpy-Fe reactivity, the Laboratory forAtomic Imaging Research (LAIR), along with collaborators at Monash University,researched properties of a similar system: terpyridine-phenyl-phenyl-terpyridine(TPPT) coordinated with Fe on a Ag(111) substrate. Fig 3.1 compares the chemicalstructure of TPT and TPPT, with the only structural difference being an additionalphenyl separating the tpy ends in TPPT where TPT has only one phenyl.Figure 3.1: Chemical structure of TPT and TPPT. Both molecules have identicaltpy groups separated by a single phenyl for TPT and two phenyl spacers for TPPT.Previous works on TPPT-Fe characterized the electronic and physical propertiesof thermally-activated, self-assembled, one-dimensional nano chains consisting ofTPPTs connected with tri-iron nodes on a Ag(111) substrate. [24, 58, 76, 103]The physical and electronic properties associated with this subtle structural changebetween TPT and TPPT will be discussed in the following subsections.3.1 Adsorption GeometriesSPM studies involving terpyridine containing organic molecules report flat or quasi-flat physisorption on noble metal surfaces, [104–106] and TPT on Ag(111) is no273.1. ADSORPTION GEOMETRIESexception. Fig 3.2 shows an isolated TPT imaged using constant height ncAFM (a),and constant current STM (b). The bright regions in (a) represent chemical bondswithin the molecule [107] and closely resemble the gas-phase chemical structureof TPT superimposed on topograph (b), which confirms flat physisorption. Theouter pyridine rings in (a) appear distorted similarly to TPPT, however, the centralphenyl is imaged symmetrically elongated rather than twisted out of the molecularplane as reported for TPPT. [24] This suggests TPT lies flatter on the substratecompared with TPPT, and is reasonable since the twisted conformation arises fromsteric repulsion between two phenyl groups. [108]Figure 3.2: Constant-height ncAFM image of an isolated TPT molecule (a) imagedwith a CO-terminated tip at Avib = 10mV and ∆z = −40pm from the set pointVb = 20mV and It = 10pA on Ag(111). Topograph (b) shows the same molecule in(a) imaged with the same tip and set point parameters as (a). Gas-phase chemicalstructure of TPT is superimposed on topograph (b).In addition to flat physisorption, TPT adsorbs at a particular angle with respectto the <111> directions of the Ag(111) surface. The adsorption angle for TPTis shown in Fig 3.3. Atomic resolution in Fig 3.3 (b, lower) serves as a guide todetermine the adsorption angle of TPT relative to the surface directions. (b) showsthe long molecular axis of a single TPT molecule is +15° from the [0 1 −1] direction.1In a (100nm)2 range neighboring the atomic resolution in (b), partly shown in (a),adsorption angles of all molecules were measured as ∼ ±15° from the three-foldsymmetric <111>. The gray arrow in (a) marks the [0 -1 1] direction, and the blackand white arrows mark +15° and −15°, respectively, from <111>. In addition tothe directly measured adsorption angles, a maximum of six adsorption angles wereconsistently present in topography images acquired after depositing TPT,2 and thisordering indicates a strong molecule-substrate interaction.1There are six equivalent low index crystallographic directions on the FCC Ag(111) surface. Thelabeling choice of these directions is arbitrary and only used to distinguish directions in Fig 3.3.2Adsorption angles relative to the substrate are considered on terraces only. Step edges aredisregarded as these are influenced by different interactions present at a step edge.283.1. ADSORPTION GEOMETRIESFigure 3.3: Adsorption angles of TPT relative to Ag(111) crystallographic directions.(b) consists of two separate topographs spliced together that were acquired oneafter another without re-positioning the piezomotors in between images. Drift wasminimized by allowing the tip to settle in position for t > 1hr before imaging. Theupper half of (b) shows an isolated TPT molecule with the long molecular axislabeled, (i), and the lower half shows atomic resolution of Ag(111) (FCC latticesuperimposed) with the [0 1 -1] direction labeled (gray arrow). The long molecularaxis of TPT measures ∼ 15° from the [0 1 -1] direction. (a) is an overview topographencompassing the area of (b), and showing TPT molecules all adsorb at ±15° to<111>. The step edge was removed from (a) for clarity. Vb = −20mV , It = 30pA(a). Vb = −20mV , It = 50pA (b, upper). Vb = 5mV , It = 100pA (b, lower).293.1. ADSORPTION GEOMETRIESThe same approximate adsorption angle of ∼ ±15° from the molecule’s long axisto <111> was reported for TPT and TPPT. [76, 103] It was previously reported thattpy groups strongly hybridize with the substrate compared to phenyl spacers, [103]thus, it is probable that the tpy groups dominate the adsorption configuration forboth TPT and TPPT. Due to the strong tpy interaction with the substrate, alongwith the relatively small periodicity of Ag surface atoms compared to the size of thetpy groups, it is reasonable that TPT adopts a similar adsorption configuration toTPPT on Ag(111) with respect to <111>.Figure 3.4: Topography images after TPT depositions show molecules assemble inhydrogen-bonded rows. Overview of molecular ordering at high surface coverage (b),and close up of two H bonded TPT molecules with chemical structure superimposed(a) shows the C-H...N hydrogen bonds (blue dashed lines). Vb = 20mV , It = 20pA(b).In Fig 3.3 (a), most TPT molecules are imaged adjacently on the surface, attractedto each other through weak hydrogen (H) bonding. Similar H bonding was reportedfor inter-molecular TPPT interactions as well. [58, 76] The weak hydrogen bondsbetween C-H...N of adjacent pyridines are drawn in Fig 3.4 (a) with blue dashed lines.[109] At higher surface coverages, TPT molecules were almost exclusively imaged inH bonded rows as shown in Fig 3.4 (b), opposed to having unbound molecules on thesurface as was observed for TPPT at roughly the same surface coverage (see Fig 1(a)in [76]). Additionally, the H bonded rows of TPT molecules are highly linear, unlikethe more commonly observed zipper configurations for TPPT (see Fig 1(a) in [76]).These observations indicate that TPT has different inter-molecular interactionsand/or molecule-substrate interactions than the TPPT/Ag(111) system. Furtherdifferences between the TPT and TPPT systems were observed upon Fe coordination,and those suggest TPT may have a much higher diffusion coefficient than TPPT on303.2. COORDINATED STRUCTURESAg(111). These differences will be discussed in more detail in Section 3.3.3.2 Coordinated StructuresAfter characterizing bare TPT molecules, Fe was added to the substrate at roomtemperature, as described in Section 2.6, and TPT-Fe coordination complexes self-assembled on the surface. Upon coordination with Fe, a preferred molecular adsorp-tion angle with respect to the substrate was no longer observed, and Fe-coordinatedTPT molecules no longer appeared in H bonded rows. These observations, consistentwith previous studies, [24, 58, 76] indicate a reduced molecule-substrate interactionupon tpy-metalation, and that the coordination bond is energetically favorable overH bonding between adjacent TPT molecules when Fe is present on the sample. Ingeneral, coordination bonds are much stronger than H bonds. [110] Details of thecoordination bond are shown in Fig 3.5.Fig 3.5 compares isolated bare (a, d, g), singly-coordinated (SC) (b, e, h), anddoubly-coordinated (DC) (c, f, i) TPT molecules. Red arrows indicate locations ofFe adatoms coordinate-bonded with terpyridine groups. STM topographs (d, e, f)were imaged with a CO-terminated tip and have chemical structures superimposed.ncAFM images (g, h, i) were measured at constant height, and Laplace-filtered. Bluearrows indicate the different position of N atoms in peripheral pyridine groups beforemetalation (outward facing N atoms) and after metalation (inwards facing N atomstowards coordination center). The change in N atom position upon metalation isthermally activated, the most common tpy-metal coordination motif observed ingeneral, [111] and the only motif observed in the work for this thesis.Line profile (a) depicted in topograph (d) is symmetric, showing ∼ 0.8Å maximumapparent heights on the bare tpy groups relative to the substrate at Vb = 20mV . Asobserved for TPPT, two minima on either side of the profile center (blue arrows)depict a N-substrate electrostatic interaction where the electro-negative N atomsrepel the substrate conduction electrons, resulting in a reduction of the local densityof states at the N positions. [58, 76] The ncAFM image (g) shows a subtle decreasein intensity at N atom positions compared to C positions on the pyridine rings,suggesting the N atoms may be closer to the substrate than the C atoms.The doubly-coordinated TPT (c, f, i) exhibits a symmetric line profile (c) withmaximum height peaks ∼ 1.2Å at Vb = 20mV relative to the substrate at themetalated sides of the molecule. The center of the profile no longer displays apparentheight minima as with Fig 3.5 (a), which supports the conformational change ofall three N atoms inward facing and bonding with the Fe center upon coordination.The precise Fe atom positions, red arrows in (f), are not explicitly evident in ncAFMimages (i) and (h), which is consistent with previous work on TPPT-Fe. [24] Themetal coordination center is not always pronounced in ncAFM imaging because themetal adatom can interact strongly with the metal substrate causing it’s verticalposition to lie below the molecular plane. [105, 107] The yellow arrow in (f) indicatesa noisy region in the ncAFM imaging, and this type of noise is typical of tip-induced313.2. COORDINATED STRUCTURESFigure 3.5: Comparison of non-metalated (a, d, g), singly-metalated (b, e, h),and doubly-metalated (c, f, i) TPT molecules. Topographs (d, e, f) were imagedwith CO-terminated tips and have chemical structures superimposed. Line profiles(a, b, c) show a difference in apparent height of ∼ 0.2Å between metalated andnon-metalated sides of TPT molecules at Vb = 20mV , with the metalated sidehigher. Blue arrows indicate positions of N atoms and red arrows positions of Featoms. Upon coordination with Fe, the tpy groups switch from the trans,transconformation (peripheral N atoms face outward) to cis,cis (N atoms face inwardtowards coordination center). Vb = 20mV It = 10pA (d) and (f), Vb = 20mVIt = 20pA (e), Avib = 10mV ∆ z = −0.04nm (g), Avib = 50mV ∆ z = −0.04nm(h), Avib = 50mV ∆ z = 0nm (i), set point Vb = 20mV It = 10pA on Ag(111) for(g,h,i).323.3. ATTEMPTED TPT NANOCHAIN GROWTHmotion of adsorbates. This noise may have been ruled an anomaly, except that it wasmeasured at the same location on different molecules with different tips. Noise inthe doubly-coordinated STS data also appears at this same region and may indicatethe doubly-coordinated motif does not sit entirely flat on the surface.The singly-coordinated TPT in Fig 3.5 (b, e, h) exhibits asymmetric features.The profile (b) shows an apparent height difference of ∼ 0.2Å at Vb = 20mV betweenmetalated and non-metalated tpy groups, and only one local density of states minimanear the profile center corresponding to the right side outward facing N atom of theuncoordinated tpy group. The different apparent heights relative to the substratebetween metalated and non-metalated tpys in (d), (e), and (f) are caused by differentimaging set point parameters, but all show the same trend of an increased apparentheight in metalated tpy groups. Along with structural changes in the molecule uponcoordination, electronic changes were also probed and will be discussed in detail inSection 3.4. First, a discussion of attempts to grow Fe-coordinated TPT nanochainswill be presented in the following section.3.3 Attempted TPT Nanochain GrowthPrior SPM studies successfully established conditions for the self-assembly of TPPTmolecules arranged in linear chains connected by three Fe adatoms. [24, 58, 76]Metal-metal bonds in this tri-iron configuration may hold promise as catalysts, [112]and initially a goal of this research was to compare reactivity of tri-iron nodes inTPT nanochains with activity at single-iron TPT coordination sites. Substratepreparation as well as metal and organic molecule depositions were carried outfollowing the same recipe from previous work with TPPT and Fe (see ref. [24]‘Sample Preparation’ for details), as well as testing slight variations on the recipe.Upon growing nanochains with TPT, instead of uniform Fe3 linkages betweenadjacent molecules, a diversity of coordination nodes persisted in multiple samplepreparations.3 Varying the substrate temperature during and after Fe depositionfrom 50° C to 150° C was explored and yielded no notable improvements in nodeuniformity. In fact, annealing at higher temperatures after Fe deposition increasedchain non-linearity and diversity. Non-linear chains were reported for TPPT-Fe afterannealing the sample at T = 250°C where the coordination bond reportedly involvedAg substrate atoms. [58] However, the non-linear TPPT-Ag coordinated chains donot resemble the crooked TPT chains formed at elevated substrate temperatures inthis work.Examples of three different sample preparations resulting in node diversity areshown in Fig 3.6. Tri-iron nodes in (a) and (b), labeled by white arrows, areidentifiable based on image comparison with TPPT-Fe3-TPPT from previous works,[24, 76] and have a uniform and symmetric intensity around the coordination center.Some Fe3 nodes are visible in (a) and (b), along with many other nodes of varying312 different sample preparations using the same temperature-dependent recipe, but slightlydifferent molecule and Fe coverages were attempted without growing uniform tri-iron-linked chains.333.3. ATTEMPTED TPT NANOCHAIN GROWTHFigure 3.6: Three different sample preparations attempting to grow TPT-Fe3-TPTnanochains, all resulting in node diversity. White arrows in (a) and (b) point towardsFe3 nodes, which have symmetric intensity about the coordination center and areidentifiable by image comparison with Refs. [76] and [24]. The Fe3 node yield in(b) is ∼ 50%. Topograph (c) imaged with a CO-terminated tip shows two chaincenters close up. Laplace-filtered constant-height ncAFM image (d) of the samemolecules in (c) imaged with a CO-terminated tip, Avib = 30mV , Vb = −.95mV ,and ∆ z = −0.03nm from the set point Vb = 20mV and It = 10pA on Ag(111). Thedifferent nodes in (c) also have different apparent bond lengths of 4.0Åand 3.4Åin(d). Vb = −200mV It = 20pA (a), Vb = 20mV It = 20pA (b), and Vb = 20mVIt = 10pA (c).343.3. ATTEMPTED TPT NANOCHAIN GROWTHintensity and symmetry. The three molecule chain in (c) and (d) exhibits twodifferent nodes, which have measurably different apparent bond lengths in (d) anddifferent intensities in (c).In all sample preparations performed for this thesis, uniform chain centers werenot realized. Nanochain growth with higher tri-iron node yield is presented inFig 3.6 (b) with approximately ∼ 50% of nodes exhibiting the uniform intensitycharacteristic of Fe3 motifs, and the rest unidentified. With this large amount ofnode variety, adding gases and identifying changes to the chain centers becameunreliable, and the main focus of this research shifted to bonding at single-ironactive sites with CO and C2H4, as will be discussed in Chapter 4 and Chapter 5,respectively. Activity at chain centers was observed during experiments performedfor this thesis, but consistent analysis was intractable. Exploring the reactivity atuniform Fe3 chain centers could be an interesting topic for future studies.A possible explanation for the variety of chain linkages could be a changein molecule-substrate interactions between TPT/Ag(111) when compared toTPPT/Ag(111). This was already hinted at in reference to the H bonding behaviorof TPT versus TPPT presented in Section 3.1. A change in molecule-substrateinteractions could result in increased surface mobility of the lighter organic moleculeTPT, and even a subtle change in diffusion coefficient can significantly influencegrowth and self-assembly properties. [57] Evidence supporting a possibly significantlyhigher diffusion coefficient for TPT than TPPT, is presented in Fig 3.7 and Fig 3.4.Fig 3.7 shows the deposition of TPT, usually t = 30s in total, broken down inthree parts in order to elucidate the details of molecular ordering on the substrate asa function of surface coverage. TPT was deposited on a room temperature substratesfollowing the normal sample preparation recipe presented in Section 2.6. Moleculargrowth proceeded as follows: molecules first occupied step edges (a), next saturatedstep edges and began to order on terraces (b), and continued to fill terraces at anaccelerated rate while step edges remained saturated (c). In (a), molecules bindto the step edge at an angle such that one peripheral pyridine sits on the upperterrace and the diagonally opposite pyridine sits on the lower terrace. A singlestep edge vacancy is indicated with a white arrow in (a). In a (120nm)2 area nomolecules were present in their entirety on terraces after the first deposition. Afteran additional t = 13s molecular deposition, the sample was imaged and is shown in(b). The step edges are saturated with molecules, and low coverage appears on theterrace. Depositing for only t = 2s more at the same temperature resulted in a largeincrease in molecule surface density on the terrace shown in (c). It is common forstep edges to create nucleation sites for molecules adsorbed to noble metal surfaces[113], however, the observed step saturation before terrace occupation of TPT upondeposition was not observed for TPPT deposited under the same conditions,4 andsupports the idea of a higher surface diffusion rate for TPT compared with TPPT.Final evidence suggesting a change in molecule-substrate interactions of the4TPPT was deposited in the same preparation chamber of the Omicron LT SPM, also on aroom temperature substrate.353.3. ATTEMPTED TPT NANOCHAIN GROWTHFigure 3.7: TPT deposition on the same sample performed in three incrementalsteps. All depositions were performed on a room temperature substrate as describedin Section 2.6. The sample is imaged after depositing molecules on the substrate fort = 13s (a), after an additional t = 13s deposition (b), and after an additional t = 2sdeposition (c). Terrace coverage dramatically increases between (b) and (c) withonly an additional t = 2s deposition, supporting the idea that molecules saturatedstep edge sites before occupying terraces. A single step edge vacancy is indicatedwith a white arrow in (a). Vb = 20mV , It = 50pA for (a, b, c).363.3. ATTEMPTED TPT NANOCHAIN GROWTHFigure 3.8: Two TPT molecules most likely coordinate-bonded with one Fe adatommimicking the solution chemistry minimum energy configuration, and distorted by2D surface constraints. Laplace-filtered ncAFM image (b) emphasizes the twisted tpygroups where the brightest regions correspond to one pyridine from each moleculesitting on top of the adjacent molecule’s pyridine. Vb = 20mV It = 10pA (a),Avib = 30mV ∆ z = 0.05nm (b).373.4. ELECTRONIC STRUCTURESsingle-phenyl molecule, is presented in Fig 3.8, where a twisted node configurationis shown, which has not been reported for TPPT-Fe nanochain assembly. Basedon its resemblance to the most commonly encountered, solution stable tpy-M-tpymotif with D2d symmetry, [111] the twisted node of Fig 3.8 presumably consists ofa single Fe adatom coordinated to six N atoms from adjacent tpy groups. Insteadof perpendicular tpy groups of D2d symmetry stabilizing the node, the tpys aretwisted into the plane due to 2D surface constraints. The ability for pyridines totwist out of the molecular plane in order to stabilize the node imaged in Fig 3.8 wasnot observed for TPPT, and is indicative of reduced molecule-substrate interactions.It is important to note that although reduced molecule-substrate interactionsleading to an increased surface diffusion rate may have played a role in the diversity ofcoordination nodes observed for TPT, there are a few other considerations requiringfurther investigation in order to rule out their possible influence upon experiments.A peak in the residual gas analyzer (RGA) spectra at mass 19 (likely Fluorine),comparable with the magnitude of the hydrogen (H2) peak, was detected consistentlyin the STM chamber even after using the titanium sublimation pump (TSP) inthe prep on a four hour cycle for several days. Fluorine is highly reactive anddifficult to remove from a vacuum chamber, and its’ presence in the Omicron is notunreasonable since previous experiments incorporated Fluorinated molecules. ThisFluorine contamination could have been facilitating the diversity of coordinationnodes observed upon TPT/Fe self-assembly.3.4 Electronic StructuresBefore discussing the electronic properties of the coordination bond, a brief com-parison between the bare TPT molecule to previous studies on the bare TPPTmolecule will be made. As with studies on TPPT, no occupied electronic states weremeasurable within a stable bias range. [58, 103]Fig 3.9 (d) shows averaged normalized dI/dV5 of the unoccupied electronic statesof an isolated TPT molecule. The lowest unoccupied molecular orbital (LUMO) atVb = 1.05V (a) is located at the phenyl, the LUMO +1/+2 at Vb = 1.95V (b) isconcentrated at the tpy groups, and the LUMO +4 at Vb = 2.55V (c) is distributedover the phenyl and tpy groups of the molecule. The spatial morphology of theLUMO states follows the same pattern for TPPT, however, the reported energiesare different by up to ∼ 25mV when compared to previous experiments (see Fig.6 in ref. [103]). Additionally, the LUMO +1 is more pronounced for TPT whencompared to the same state on TPPT, which may be another indication of a reducedinteraction between the Ag substrate atoms and tpy groups for TPT.The electronic structure of TPT changes upon metalation with Fe. Fig 3.10shows the shift in energy of the phenyl LUMO state as the TPT changes fromuncoordinated, to singly-coordinated, and doubly-coordinated. As the number ofFe-tpy bonds increases, the energy of the LUMO increases from Vb = 1.05V on5Methods for STS data analysis are presented in Appendix B.383.4. ELECTRONIC STRUCTURESthe bare, isolated molecule (a), to Vb = 1.25V on a singly metalated TPT (b), toVb = 1.60V on a doubly-metalated TPT (c). The bias slices (a, b, c) illustratethe spatial distribution of the LUMO, with a noticeable asymmetry in the singly-coordinated species, (b), where the phenyl state is spatially shifted towards thenon-metalated side of the molecule.Figure 3.9: Averaged normalized dI/dV of unoccupied electronic states of an isolatedTPT molecule. Topography inset in (d) outlines the tpy and phenyl areas of themolecule averaged over to generate the spectra for (d). Solid curves in (d) are the tpyand phenyl spectra with the Ag(111) background spectra subtracted. Normalizedbias slices (a, b, c) show the spatial distribution of the TPT LUMO states.Upon metalation, singly and doubly-coordinated TPTs exhibited an occupiedelectronic state at Vb = −0.09V , consistent with the Fe-tpy state measured forTPPT at Vb = −0.09V . [76] Fig 3.11 compares the SC and DC Fe states. SpatialSTS maps show the Fe state (red arrows) at maximum intensity for singly (a) anddoubly-coordinated (b) molecules. Comparing tpy (blue arrow) and Fe-tpy (red393.4. ELECTRONIC STRUCTURESarrow) sides of (a), it is evident the Vb = −0.09V is associated with the coordinationbond, and not the tpy.Figure 3.10: Morphology of the LUMO phenyl state upon metalation (d) shows theLUMO increases in energy with increased number of Fe-tpy bonds. STS slices ofthe LUMO resonance on an isolated molecule (a), singly-coordinated TPT (b), anddoubly-coordinated TPT (c) illustrate the spatial distribution of the phenyl states.Red arrows point towards Fe coordination sites. Dashed lines in (d) are Ag(111)background spectra for each of the three grids (a,b,c).Additionally, a state at Vb = −0.09V was not measured for the non-metalatedmolecule. Average dI/dV in near the Fermi energy compares molecular states for thesingly (c) and doubly-coordinated (d) molecule. Solid background curves in (c) and(d) are Fe spectra divided by phenyl spectra for the respective SC or DC molecule,which assists in locating the position of the Fe states since the Fe state partiallyoverlaps with the Ag(111) surface state at Vb = −0.69mV .403.4. ELECTRONIC STRUCTURESFigure 3.11: Normalized bias slices at Vb = −0.09V for singly-coordinated (a) anddoubly-coordinated (b) TPTs show the occupied Fe state (indicated by red arrows).Average dI/dV for singly-metalated (c) and doubly-metalated (d) TPTs near theFermi energy. Solid backgrounds in (c) and (d) represent the Fe/phenyl spectra, toemphasize the precise location of the Fe state.Although TPT exhibits similar adsorption properties and electronic structure toTPPT, there are notable differences between the two adsorbate substrate systems.In particular, TPT exhibits a weakened interaction with the Ag(111) substratecompared with TPPT. This is evident in the prominent LUMO +1 signal at thetpy groups for TPT, the persistent noisy region near the tpy group in ncAFMimaging for the doubly-coordinated TPT molecule, and the observed twisted nodepresented in Fig 3.8 that was not observed for TPPT. As discussed in Fig 5.2, TPTdisplays a higher surface diffusion coefficient than TPPT resulting from the weakenedadsorbate-substrate interaction. Upon coordination with Fe, both TPPT and TPTexhibit the same Fe state at Vb = −0.09V , and this Fe state comprises the activesites for bonding that will be explored in subsequent chapters.41Chapter 4Reaction with COResidual gases are always present in vacuum chambers, even in the controlledenvironment of UHV. As discussed in Chapter 2, CO is a common trace gas presentin these systems. In addition to its residual presence, CO was actively dosed in theOmicron system for tip functionalization in multiple ncAFM experiments performedpreviously and for this thesis. For these reasons, CO was always detected on samplesduring experiments for this work, even when only ethylene was dosed in the system.In order to isolate interactions between CO and active sites, experiments wereperformed where only CO was dosed on samples the results of which are presented inthis chapter. The interaction between CO and active Fe-tpy sites was confined to thetwo dimensional surface and was induced by annealing samples consisting of activesites and adsorbed CO at low temperatures, T ≤ 40K. Annealing mobilized the COwhere it was able to migrate towards active sites, and two resulting species wereobserved. One metastable ‘prebond’ interaction, and a stable ‘bond’ configuration.These motifs will be discussed throughout this chapter. The chapter is dividedinto five sections. Section 4.1 discusses the adsorption and diffusion of CO onthe substrate, Section 4.2 details the experimentally observed prebond and bondconfigurations, Section 4.3 discusses evidence for the instability of the CO prebondconfiguration, Section 4.4 presents statistical data on the observed configurationstracked through different stages in the reaction, and Section 4.5 contains STS dataon the bond motif.4.1 CO Adsorption and Diffusion on Ag(111)Key to understanding surface reactivity, is first understanding adsorption anddiffusion behavior of individual reactants on the substrate of interest. CO has beenstudied on a variety of transition and noble metal substrates as a model molecule-substrate system both experimentally and theoretically, and the consensus amongstudies is that the Ag-CO bond is weak compared with CO bonding with other metalsubstrates, which means CO desorbs and diffuses relatively easily on the Ag(111)424.1. CO ADSORPTION AND DIFFUSION ON AG(111)substrate. [114–117]The possible adsorption configurations of CO on Ag(111) are drawn in Fig 4.1(a) top view and (b) side view with Ag atoms gray, C atoms black, and O atomsred. The top site bonding consists of one Ag atom bonded with the C of the COmolecule and this is the most energetically favorable adsorption configuration witha calculated binding energy (BE) of BE = −0.32eV . [114] The top site has alsobeen experimentally observed as the preferred adsorption site for CO on Ag(111)using HREELS. [118] Other possible adsorption sites include the bridge, where Cis bonded with two Ag atoms with a BE = −0.29eV , or the hcp/fcc (hexagonalclose-packed/face-centered cubic) sites, where C is bonded with three Ag atomswith a BE = −0.28eV .1 [114] Experimentally, using surface potential isotherms, aBE = −0.28eV for CO on Ag(111) was measured, [119] which agrees well with theaforementioned theoretical values.Figure 4.1: Top-view (a) and side-view (b) of the potential adsorption sites of COon a Ag(111) surface (Ag gray, C black, and O red). The top site consists of thecarbon atom of CO bonded to one Ag atom, while the bridge and hcp/fcc (hexagonalclose-packed/face-centered cubic) sites have CO bonded with two and three Agatoms, respectively. Frontier orbital diagrams of a free CO molecule (c) show theelectron density is concentrated on the C atom for both the 5σ HOMO and 2pi∗antibonding LUMO.1The difference between hcp and fcc is in the orientation of the layers of atoms below the surfacelayer where hcp has two alternating layer positions and fcc has three alternating layer positions.For a more detailed discussion see Chapter 1 of Kittel’s Introduction to Solid State Physics 8thedition.434.2. CO INTERACTIONS WITH FE-TPY: VISUAL DATAFor adsorption sites on metal surfaces, CO always bonds perpendicular to thesurface with the C atom bonded to the metal, as depicted in Fig 4.1 (a,b). This isattributed to the fact that the majority of electron density resides around the C atomin both frontier molecular orbitals of CO, [120, 121] the 5σ HOMO and 2pi∗ LUMOdrawn in Fig 4.1 (c). Generally, is it assumed that the frontier orbitals are primarilyinvolved in binding molecular adsorbates to surfaces, [116] and a popular modelof this bonding behavior was first presented by Blyholder in 1964. The Blyholdermodel proposes a combination of electron donation from occupied CO 5σ HOMOstates into empty surface metal orbitals with back-donation from unoccupied CO2pi∗ LUMO states into occupied surface orbitals. [122] This model does not explainall bonding behavior between CO and various substrates, [116, 123] and it has beenargued by some that CO binding to Ag(111) also involves strong contributions fromthe lower pi orbitals of the CO. [27] However, the Blyholder model does provide asimplistic starting point for generally understanding the bonding behavior betweenCO and a variety of metal substrates from a molecular orbital theory perspective.Regardless of the specific orbitals involved in binding CO to the substrate, CO isalso observed bound perpendicular to metal substrates. [117]When CO was dosed and adsorbed on the Ag(111) sample containing tpy-Fe active sites, the substrate temperature reached Tmax ∼ 11.6K during typicaldepositions. When the surface was probed post-deposition, a variety of changes topreviously bare tpy-Fe nodes were observed, along with free CO on the surface, andare shown in Fig 4.2 The large number of observed interactions between CO andFe-tpy nodes suggested the CO was at least slightly diffusing on the substrate at thedeposition temperature of T = 11.6K. Prior STM studies probing a CO on Ag(111)system reported the onset of CO diffusion at T ∼ 12K, [23] which agrees well withexperimental observations from this thesis. The diffusion barrier was also estimatedin calculations performed by Chen et al. as significantly lower than that of manyother small molecules and atoms adsorbed on Ag surfaces. [114] Since CO diffuseseasily on the Ag(111) substrate, it is not surprising a variety of interactions withthe active Fe-tpy nodes were observed. These interactions will be discussed in detailin the following section.4.2 CO Interactions with Fe-tpy: Visual DataAs mentioned in the previous section, different motifs were observed after dosingsamples with CO. In particular, three different motifs were observed post CO dose,which will be referred to as CO surface, CO prebond, and Fe-CO bond configurations.These three types are presented in Fig 4.2 where roughly the same area of a sampleis shown before dosing (a) and after dosing CO (b) at PLT = 1.6 ∗ 10−9mbar fort = 3min with the substrate reaching Tmax = 11.6K during dosing.The active sites are labeled in Fig 4.2 (a) and consist of a tpy-Fe coordinationbond. In (b), arrows indicate the three observed motifs post CO dose. The surfaceCO is observed as either free on the Ag(111) surface, or weakly H-bonded to the444.2. CO INTERACTIONS WITH FE-TPY: VISUAL DATAdistal pyridine of a non-metalated tpy group. The CO prebond configuration isidentifiable by a dark termination where the Fe-tpy node was previously bright.The Fe-CO bond is observed in two different ways, labeled (i) and (ii) in (b). Thecategorization of the Fe-CO bond as either type (i) or (ii) depends on whetherthe opposite termination of the molecule is metalated or not. The bond is imagedas a localized bright termination (ii) when the opposite tpy is metalated, or theFe-CO bond is imaged as a half-ring-shaped termination (i) when the opposite tpy isnon-metalated. The bond motif (ii) is consistently observed as long as the oppositetpy is minimally coordination-bonded with Fe; meaning the opposite tpy may consistof a simple tpy-Fe node, or a more complex tpy-Fe-X node.2 The Fe-CO bonds(i) and (ii) appear differently in topograph (b) of Fig 4.2 imaged at Vb = 20mV .However, at larger | Vb |, such as that used to generate topographs (e) and (f) inFig 4.3, the difference between type (i) and (ii) Fe-CO bonds is negligible, which iswhy they were both categorized as Fe-CO bonds.Figure 4.2: Topographs before dosing CO (a) and a similar area after dosing CO(b) both scanned with Vb = 20mV and It = 50pA. Singly and doubly-coordinatedactive sites of (a) morph into the CO prebond nodes or the Fe-CO bond nodes (i)and (ii) all labeled in (b). CO isolated on the surface or weakly bound to a distal,non-metalated pyridine is also apparent after dosing. The substrate temperaturereached Tmax = 11.6K during CO dosing at PLT = 1.6E − 9mbar for t = 3min.As discussed in Chapter 2, STM constant current topography is influenced bythe local density of states of the sample. Thus the observed differences in nodeterminations are directly related to changes in local electronic structure, which isalso influenced by the applied bias voltage. To clarify this more explicitly, Fig 4.3compares close up topographs of the Fe-CO bond type (i) in (a,c,e) and type (ii)2Where X=CO, X=C2H4 (discussed in Chapter 5), or X equals a ‘higher order’ interaction asshown in Appendix A.454.2. CO INTERACTIONS WITH FE-TPY: VISUAL DATAFigure 4.3: Comparison of the same singly Fe-CO terminated molecules (a,c,e) anddoubly Fe-CO terminated molecules (b,d,f) at different biases. Vb = 20mV andIt = 20pA for (a,b) and Vb = −200mV , It = 200pA for (e,f). Line profiles (c) and(d) correspond to lines drawn on topographs (a) and (b), respectively. Vertical solidlines in (c) and (d) separate the max and min apparent heights indicated by blue plussigns in (a) and (b), respectively. At Vb = 20mV (a,b), the single-CO-terminated(a) and double-CO-terminated (b) molecules image differently, with (d) showinga larger maximum apparent height for the double-termination compared with themaximum apparent height shown in (c) of the single-termination.464.2. CO INTERACTIONS WITH FE-TPY: VISUAL DATAin (b,d,f) at two different biases. Vb = 20mV for (a,b), and Vb = −200mV for(e,f). Comparing (a) and (b) in Fig 4.3, both imaged at Vb = 20mV , the differencein terminations between different types of Fe-CO bonds is apparent: (b) shows abright, localized termination on both ends of the molecule, and (a) shows a dimmerhalf-ring termination, same as was observed in Fig 4.2 (b). The line profiles (c) and(d) show apparent heights corresponding to the blue lines overlaying topographs (a)and (b), respectively. The blue plus signs in (a) and (b) correspond to the black,solid vertical lines in the line profiles (c) and (d), respectively, and these vertical linesbound the maximum and minimum apparent heights measured in the topographs.In comparing the maximum apparent heights between (c) and (d), it is clear that(b) has a stronger tunneling current at the Fe-CO terminations at Vb = 20mV . Also,there is a distinct shoulder around ∼ 1.2nm in (c) that is almost undetectable in(d). Topographs (e) and (f) depict the same molecules as (a) and (b), respectively,only scanned at the more negative bias Vb = −200mV . When comparing the twotypes of Fe-CO bonds depicted in (e) and (f) at the larger | Vb |, the differences interminations are minimal with only a slightly increased intensity at the terminationsin (f) compared to (e). This similarity is why both type (i) and (ii) nodes areconsidered to be comprised of the same Fe-CO bond motif. Additionally, all Fe-CObond motifs show a dark region extending onto the surface away from the nodetermination, regardless of bias or opposite coordination of the tpy group.3The distinct difference between type (i) and type (ii) Fe-CO bonds was consistentlyobserved throughout topographs acquired at Vb = 20mV in the data collected forthis thesis, and could be the result of bias dependent electronic coupling between themolecule ends. However, there was no further evidence detected that this potentialcoupling influenced reactivity. Additionally, no shift in the Fe state resonancemeasured at Vb = −0.09V was detected between singly and doubly-coordinatedTPTs (see Fig 3.11). Therefore, in the sections of this thesis devoted to reactionstatistics (Section 4.4 and Section 5.4), each end of the TPT molecule was countedas an independent active site.A more detailed comparison of the CO prebond and Fe-CO bond configurationsare shown in Fig 4.4. Topographs imaged with CO terminated tips reveal ingreater detail the dark termination of the prebond motif (a) and the bright/half-ringtermination of the bond motif (d). ncAFM images also acquired with CO terminatedtips reveal submolecular details of the prebond (b) and bond (e) configurations.The termination in (e) is distinctly longer than (b), suggesting a change in COorientation with respect to the substrate and active site between prebond and bondstructures. The proposed chemical structure for the prebond motif (c) shows a COmolecule nearby the tpy-Fe node, but still bound upright on the Ag substrate. Thesuggested chemical structure for the bonded configuration (f) has the CO moleculelying parallel to the Ag substrate and linearly bound to the Fe-tpy node. The changein CO orientation from perpendicularly bound to the substrate to lying parallel tothe substrate and bound to the Fe would explain the difference in measured lengths3See Appendix A for more information on visual identification of the Fe-CO bond configuration.474.2. CO INTERACTIONS WITH FE-TPY: VISUAL DATAFigure 4.4: Comparison of CO prebond (a,b,c) and Fe-CO bond (d,e,f) configurations.STM topographs for the prebond (a) and bond (d) motifs were imaged with CO-terminated tips at Vb = 20mV and It = 10pA. Constant height ncAFM images(Laplace-filtered) for the prebond (b) and bond (e) configurations show a distinctdifference in the length of lines at the right side of the molecules, which is thelocation of the CO. The proposed chemical structures for the prebond and bondmotifs are shown in (c) and (f), respectively. Parameters for (b): Vb = −0.95mV ,Avib = 50mV , ∆ z = −0.01nm from setpoint Vb = 20mV and It = 10pA on Ag(111).Parameters for (e): Vb = −0.95mV , Avib = 20mV , ∆ z = −0.06nm from setpointVb = 20mV and It = 10pA on Ag(111).of the terminations in (b) and (e).Fig 4.5 provides a direct comparison of the prebond and bond configurationsterminating the ends of the same molecule. This dual termination allows theopportunity to make a comparison with the exact same CO-terminated tip. Theprebond configuration is on the right side of Fig 4.5 (a) and (b), and the bondconfiguration on the left of (a) and (b). The same bright (left) and dark (right)characteristic terminations of the bond and prebond configurations, respectively, arevisible in topograph (a), and the increase in length of the Fe-CO bond termination(left) in the ncAFM image (b) is apparent when compared directly with the shorterCO prebond termination (right).484.3. CO PREBOND INSTABILITYFigure 4.5: Direct comparison between CO prebond and bond motifs on differentends of the same TPT molecule. Topograph (a) imaged at Vb = 20mV and It = 10pAwith a CO-terminated tip (scale bar 6.0Å). Laplace-filtered constant-height ncAFMimage (b) taken with a CO-terminated tip (scale bar 4.0Å). Avib = 50mV , and∆ z = −0.01nm from the setpoint parameters Vb = 20mV and It = 10pA on Ag(111)for (b).4.3 CO Prebond InstabilityUnlike identification of the Fe-CO bond configuration, which changes dependingon bias, the CO prebond configuration was identifiable at all probed biases by adark termination, albeit with slight differences in appearance from node to node.Fig 4.6 shows an overview topograph of the same area imaged at different biasesof Vb = 20mV (a), and Vb = −200mV (b), which highlights some of the prebondvariations. The sample was imaged after two CO doses at PLT ∼ 2 · 10−9mbar fort = 3min and annealing at Tmax = 35K for t = 5min. Fe-tpy active sites (redarrows), CO prebond motifs (gray arrows), Fe-CO bond motifs (black arrows), andone higher order interaction (orange arrow) are all labeled.4 The prebond motifs allexhibit slightly non-uniform, dark terminations.4See the Conclusion and Appendix A for more information on higher order interactions.494.3. CO PREBOND INSTABILITYFigure 4.6: Topographs of the same area on a sample consisting of active sites (redarrows), CO prebond motifs (gray arrows), and CO bond motifs (black arrows)imaged at setpoint parameters of Vb = 20mV , It = 20pA (a), and Vb = −200mV ,It = 200pA (b). Active sites image uniformly within (a) and (b), Fe-CO bondsimage uniformly within (b), and CO prebond motifs image non-uniformly in both(a) and (b). The sample shown in (a) and (b) was prepped with two CO doses bothat PLT ∼ 2 · 10−9mbar for t = 3min with the substrate reaching Tmax = 11.6K,followed by annealing at T = 25K for t = 5min, and annealing at T = 35Kfor t = 5min. More information on the ‘higher order 4’ node is presented in theConclusion and Appendix A.504.3. CO PREBOND INSTABILITYFor an explicit example highlighting the slight differences between prebondconfigurations, see the two boxes in Fig 4.6 (a) and (b). The left-boxed CO prebondconfiguration resembles the bare active site (red arrows) with possibly a surfaceCO nearby, however, the right-boxed CO prebond motif seems to have the darktermination overpowering any remnants of the previously bright single coordination.If the CO prebond configuration is only loosely attracted to the tpy-Fe node, butstill weakly bound to the Ag(111) substrate, this may explain why the intensityof the dark termination in the prebond motif varies from one node to the next.Additionally, upon coordination TPTs lose their preferential adsorption configurationon the substrate meaning that adjacent CO binding sites on Ag would be expectedat a variety of distances from the coordination center. During experiments, it wascommon to observe a variety of CO prebond terminations such as those presented inFig 4.6, and these observations support the idea that the CO prebond configurationis a metastable intermediate to the Fe-CO bond motif.Figure 4.7: The same molecule scanned up (a) and down (b) without moving piezomotors or changing scanning parameters of Vb = −25mV and It = 50pA betweenimages. At some point during up and down scans, the CO weakly bound to the Natom of the peripheral pyridine in (a) moves to the CO prebond position imaged in(b).Fig 4.7 presents further evidence that the CO prebond configuration is a weakinteraction. The same area is imaged in Fig 4.7 without changing scanning parametersor moving piezo motors between scans (a) up and (b) back down. The CO imagedin the scans changes positions at some point between (a) and (b) moving from theweakly bound position on a distal uncoordinated pyridine (a), to the active siteforming a CO prebond motif (b). There was no impulse for adsorbate motion outsideof the presence of the tip, and this movement from the pyridine bound position to514.4. CO INTERACTIONS WITH FE-TPY: STATISTICAL DATAprebond position was commonly observed while scanning many topographs for thisthesis.In addition to CO molecules switching from the pyridine bound position to COprebond motifs while scanning, a switch in the reverse direction was also observedmany times while scanning at modest parameters of Vb ∼ 20mV and Vb ∼ 20pA.However, a change to or from the Fe-CO bond configuration was never observedwhile scanning at modest parameters during experiments.5 These interactions aresummarized in Fig 4.8. CO movement on a Ag(111) surface induced by the STMtip has previously been experimentally observed at T = 5K while scanning atIt = 5pA and Vb = 5mV , [23] and tip induced CO motion seems reasonable sincethe interaction between CO and Ag(111) is quite weak. [114] However, since theprocess from Fe-CO bond back to prebond or bare active site (and vice versa) wasnot observed upon scanning, the strength of the Fe-CO bond is determined to bestronger than the prebond interaction energy.Figure 4.8: Summary of observed reversible (a to b) and irreversible (b to c)interactions of CO with coordinated TPTs.4.4 CO Interactions with Fe-tpy: Statistical DataTo further understand the interactions of CO with Fe-tpy active sites, large scaleimages acquired at different stages in experiments were scrutinized and differentmotifs were identified and tracked. Stages of the experiments included dosing gasesand annealing at temperatures up to T = 40K to induce reactions. The countingtechnique and error propagation employed for the statistical analysis are discussedin detail in Appendix D.3.As discussed at the beginning of this chapter, residual CO was present inexperiments whether CO was actively dosed in the chamber or not. Fig 4.9 showsthe number of CO molecules per (10nm)2 area as a function of temperature. BelowT = 11.6K,6 CO was not detected on samples in these experiments. CO wasdosed on the sample at T = 11.6K for experiment (a) and (b), and dosed twicefor experiment (b) since the first dose resulted in very low surface coverage. Inexperiment (c), CO was not dosed in the system, only ethylene was dosed. Ethylene5The change of a Fe-CO bond motif to a bare active site was observed after applying large biaspulses of Vb = 1V .6As described in Section 2.6.3, the substrate temperature during typical gas dosages reachedT = 11.6K. Thus T = 11.6K is the lowest temperature where CO was recorded on the surface forthe experiments presented in Fig 4.9, which corresponds to Tmax during gas dosage.524.4. CO INTERACTIONS WITH FE-TPY: STATISTICAL DATAwas dosed twice at T = 11.6K on the surface and annealing at T = 30K was alsorepeated in experiment (c), which explains the separate data points for the sametemperatures at T = 11.6K and T = 30K. For all three experiments, CO coveragepositively correlated with annealing temperature, meaning more CO was introducedon samples as experiments progressed to higher annealing temperatures.Figure 4.9: Total CO coverage in three different experiments (a,b,c) as a functionof substrate annealing temperature. Below T = 11.6K, the maximum substratetemperature reached during typical gas dosages, CO was not detected on the surface.CO was dosed at T = 11.6K once in experiment (a) and twice in (b). CO was notdosed in experiment (c). CO coverage positively correlates with substrate annealingtemperature in all three experiments.The statistical counts for CO prebond and Fe-CO bond configurations from thesame three experiments evaluated in Fig 4.9 are presented in Fig 4.10, where a generaltrend of increasing CO prebond and Fe-CO bond motifs were observed as annealingtemperatures increased. Annealing times were t = 5min for all experiments, and gasdoses for the respective type were all performed at PLT ∼ 2 ·10−9mbar for t = 3min.The x-axis in (a,b,c) shows the experimental step performed prior to counting, andthe y-axis shows the percentage of active sites occupied by either prebond or bondmotifs. Maximum temperature corresponding to each experimental stage is indicatedby the gray background bars in (a,b,c). The * indicates a software reset occurredand new area of the sample was probed after that point.7 For all three experimentsshown in Fig 4.10, the coverage of gaseous adsorbates, either ethylene or CO, wasalways less than the coverage of active sites, and competition for active sites wasassumed negligible.In experiment (a) and (c), as the samples were annealed to higher temperatures,7The software reset was caused by a crash of the Matrix SPM controller, which caused a tipcrash and required coarse motion to a new area of the sample. The structures observed after thecrash were similar to the initial starting conditions after dosing, but prior to annealing.534.4. CO INTERACTIONS WITH FE-TPY: STATISTICAL DATAFigure 4.10: Counts of CO prebond and Fe-CO bond motifs at different experimentalstages for three different samples (a,b,c). ‘Prepare AS’ refers to samples onlyconsisting of active sites. ‘Dose’ stages consisted of dosing the respective gas atPLT ∼ 2 · 10−9mbar for t = 3min. In experiment (c), only C2H4 was dosed onthe sample. The * in (c) indicates a software reset occurred at that stage, and adifferent region of the sample, several millimeters away from the previous region ofthe sample, was probed after the reset. The surface density of CO was much lessthan the density of active sites in the experimental stages presented in (a-c) andcompetition for active sites between CO molecules was assumed negligible.544.5. STS ON FE-CO BOND CONFIGURATIONthe counts of Fe-CO bonds increased over CO prebonds. For experiment (b), thecounts of prebond configurations are higher than bond motifs at the maximumannealed temperature of T = 30K. Based on the metastable nature of the COprebond motif, it was expected that the more stable Fe-CO bond motif, whichlikely requires increased activation energy to form compared to the prebond, wouldbe observed with more frequency than the prebond at higher sample annealingtemperatures. However, experiment (b) does not exhibit that trend within theprobed temperature range. This difference between experiments may be caused byslight differences in sample preparations, such as initial concentrations of activessites and reactants, or the sampled areas could be too limited to properly deducetrends between experiments. Since activated thermal processes have a probabilitydistribution proportional to exp[ EakBT ],8 there is nonzero probability that a higherenergy process can occur at lower temperatures, which could explain the increasedpresence of Fe-CO bonds in experiments (a) and (c) at higher annealing temperatures.SPM experiments on surface chemistry are probing a very localized region of areaction, compared to analytic solution chemistry techniques, and the sampling areamay need to be increased to gain more insight from statistical data. Suggestionsto improve future statistical SPM experiments on surface chemical reactions arepresented in Appendix D.3.4.5 STS on Fe-CO Bond ConfigurationTo further elucidate bonding behavior between CO and active sites, STS measure-ments were attempted on the bare CO molecule, the CO prebond configuration,and the CO bond configuration. Of these attempts, no grids were successfullymeasured on the bare CO or the CO prebond. At | Vb |= 100mV the CO prebondconfiguration was not stable and switched from prebond to a singly-coordinatedmotif during measurement several times. The difficultly in probing free surface COand prebond motifs is attributed to the low surface diffusion barrier of CO adsorbedon Ag(111), [114] and the instability of the prebond configuration as observed inSection 4.3. Successful STS grid measurements were performed on the Fe-CO bondconfiguration, the results of which are presented in Fig 4.11 and show a clear shiftin electronic structure, indicative of bonding.Fig 4.11 (a) and (c) compares dI/dV averaged and scaled9 over the Fe-CO bondand bare tpy of one molecule (c) with the dI/dV averaged over the Fe state ona different singly-coordinated molecule (a), along with the Ag(111) backgroundspectrum for each molecule. Insets in (a) and (c) circle the regions averaged over togenerate the dI/dV line spectra. Upon formation of the Fe-CO bond, the previous Festate at Vb = −0.09V shifts to Vb = −0.165V . Fig 4.11 (d) shows a normalized dI/dVbias slice at the Fe-CO resonance of Vb = −0.165V , and (b) shows a normalized slice8kB is the Boltzmann constant, Ea is the activation energy, and T is the temperature.9Since spectra on two different molecules with different grid setpoint parameters are displayed,dividing the dI/dV by the setpoint I/V appropriately scales the spectra for comparison.554.5. STS ON FE-CO BOND CONFIGURATIONFigure 4.11: Averaged dI/dV of the bare singly-coordinated active site showing aresonance at Vb = −0.09V (a, ‘Fe SC’), and the Fe-CO bond configuration showinga tunneling resonance at Vb = −0.165V (c). Upon formation of the Fe-CO bond, theFe state previously at Vb = −0.090V shifts to Vb = −0.165V . Normalized STS mapof an Fe-CO bonded molecule at Vb = −0.165V (d), and a singly-coordinated TPTat Vb = −0.09V (b). All dI/dV spectra were scaled by their respective grid setpointparameters. Grid setpoint parameters were It = 1nA, Vb = −280mV for the Fe-CObonded molecule (c,d), and It = 1nA, Vb = −2V for the singly-coordinated TPT(a,b).564.6. SUMMARYat Vb = −0.09V of a singly-coordinated TPT for comparison. The singly-coordinatedFe state (red arrow) is quite localized compared with the larger spatial distributionof Fe-CO bond state (black arrow). This change in spatial distribution of the localdensity of states along with the clear shift in electronic resonance is indicative ofbonding, and supports the claim that Fe-CO bonding has been observed.4.6 SummaryTwo distinct interactions between CO and active Fe-tpy sites were observed through-out experiments performed for this thesis. One was the metastable CO prebondconfiguration followed by the Fe-CO bond motif. Although the statistical analysispresented in this chapter was not entirely conclusive, STS and ncAFM measurementsstrongly suggests an energy activated progression from CO prebond to Fe-CO bondmotifs. Appendix D.3 presents information on how to improve statistical SPMexperiments. The next chapter presents evidence for bonding between ethylene andthe active sites that follows a similar progression from metastable prebond motif toa stable bonding motif as was reported for CO in this chapter.57Chapter 5Reaction with C2H4The experiments in this section were performed by dosing active sites with gaseousethylene, and then alternating characterizing the species present and annealing thesample up to T = 40K while recording observed structural and electronic changesto the active sites during the characterization phase. Similar to interactions betweenCO and Fe-tpy nodes, C2H4 appears to adopt ‘prebond’ and ‘bond’ configurations.As discussed in Chapter 4, CO was always present in experiments performed withC2H4, and thus some data presented in this chapter will contain information onboth CO and C2H4 interactions with Fe-tpy active sites. Observations indicatingthe presence of both molecules at one active site were also made, and potentiallyhigher order bonds consisting of multiple small molecules of one or both types wereobserved as well. These species are referred to as ‘higher order’ and require furtherefforts to identify all species formed. Higher order nodes will be briefly mentionedin this chapter and discussed in further detail in the Conclusion and Appendix A.This chapter is divided as follows: Section 5.1 discusses the adsorption and diffusionof C2H4 on Ag(111). Section 5.2 presents the visual topographs and ncAFM imageson the prebond and bond configurations of C2H4 with Fe-tpy. Section 5.3 showsdata dissociating the Fe-C2H4 bond, and confirming the constituents. Section 5.4consists of statistical data on C2H4 prebond and bond motifs observed on samplesupon annealing, and Section 5.5 details the electronic morphology of the Fe-C2H4bonded configuration.5.1 C2H4 Adsorption and Diffusion on Ag(111)Adsorption and diffusion of C2H4 on Ag(111) forms a foundation for understandingbonding mechanisms between C2H4 and the surface-bound catalyst. Fig 5.1 showsthe most likely adsorption configurations of ethylene on Ag(111) top-view (a) andside-view (b) along with the frontier molecular orbitals of ethylene (c), which areprimarily involved in binding ethylene to the surface.Based on previous studies at low temperatures, C2H4 adsorbs molecularly on585.1. C2H4 ADSORPTION AND DIFFUSION ON AG(111)Ag(111) with the C-C bond lying parallel to the substrate [124] either in the piposition, bonded to one Ag atom, or the di− σ position, bonded to two Ag atoms.1However, there are discrepancies between studies in the literature regarding preferredadsorption positions and energies for C2H4 on Ag(111) with binding energies reportedfrom E = +0.065 in the pi position, [124] to E = −0.25eV for di−σ, [127] along withthe di− σ value reported in Fig 5.1. Molecular adsorption of ethylene on Ag(111)is accompanied by a stretch in the C-C bond length from the gas phase length of1.10Å to 1.34Å [17] or 1.43Å [127]. The C-C stretch is a result of the dominantelectronic interaction between substrate and adsorbate originating from the C-Cbond of the ethylene. Desorption was not observed at the low temperatures probedfor this thesis, but has been experimentally reported at T = 128K elsewhere. [124]Figure 5.1: Top-view (a) and side-view (b) showing most-probable adsorptionconfigurations for C2H4 on Ag(111) (C=black, H=teal, Ag=gray, not to scale). Thedi−σ configuration is the preferred adsorption site and consists of a σ bond betweeneach C atom of the ethylene with one Ag atom of the substrate. The slightly-higherenergy pi configuration consists of pi bonding between one Ag substrate atom andthe C2H4 molecule. The pi HOMO and antibonding pi∗ LUMO are drawn in (c). (i)Ref [124], (ii) Ref [127].A simplistic view of C2H4 binding to Ag(111) is provided by the Dewar-Chatt-Duncanson (DCD) model, which describes a combination of electron donation fromthe C2H4 pi HOMO to unoccupied metal d-states with back-donation from occupied1At higher temperatures and on more chemically active substrates, dissociation upon adsorptionof ethylene has also been observed. [125, 126]595.1. C2H4 ADSORPTION AND DIFFUSION ON AG(111)Figure 5.2: Topographs mapping the same area of a sample containing TPT, Fe,C2H4, and trace CO after C2H4 deposition at PLT ∼ 8.4x10−9mbar for t = 2.5minwith Tmax = 11.1K (a), after annealing at T = 20K for t = 5min (b), and afterannealing at T = 25K for t = 5min (c). Difference topographs, (a-b) and (b-c),highlight small adsorbate movement between images at different annealing steps.The minimum and maximum intensities of the difference topographs show slightC2H4 movement in (a-b), and increased movement in (b-c). The diffusion lengthof C2H4 on Ag(111) is large relative to the surface density of active sites between20K < T ≤ 25K for t = 5min sample annealing. Arrows track similarities anddifferences between the frames with a key in (a-b). Vb = 20mV and It = 20pA(a,b,c).605.2. C2H4 INTERACTIONS WITH FE-TPY: VISUAL DATAmetal d-states into the unoccupied C2H4 pi∗ LUMO as stabilizing the physisorption.[128, 129] This model was shown to describe the bonding of ethylene to Cu and Nisurfaces quite well, [123, 130] and provides a starting point for understanding thebonding mechanism between metals and alkenes in general. The binding energy ofethylene to Ag substrates is quite weak compared with other more reactive substratessuch as Pt, Ru, [17] Pd, [127] or Cu [131] and this weaker binding energy enhances theprobability for bond-forming reactions to occur with other adsorbates on the surface.[127] This makes Ag an ideal substrate for studying adsorbate-adsorbate rather thansubstrate-adsorbate reactions. Additionally a weak binding energy implies a lowdiffusion barrier, consistent with the high mobility observed experimentally herebetween 20K<T<25K (Fig 5.2).Fig 5.2 shows three topographs (a,b,c) of the same area of a sample afterdifferent experimental stages; after depositing ethylene at PLT ∼ 8.4x10−9mbarfor t = 2.5min with Tmax = 11.1K (a), after annealing the sample from (a) fort = 5min at T = 20K (b), and after annealing the sample from (b) for t = 5minat T = 25K (c). The difference topographs (a-b) and (b-c) highlight the changesbetween the experimental stages of the reaction (a,b,c). The outline of moleculesthat have not actually changed between (a), (b), and (c) are slightly visible in thedifference topographs due to subtle changes in the STM tip between images. In(a-b) the maximum contrast difference highlights slight movement of C2H4 on thesubstrate, whereas in (b-c) there are significant changes in the positions of C2H4molecules pre and post annealing at T = 25K.In Fig 5.2, significant mobility of C2H4 on Ag(111) is apparent at T ≥ 20K.Additionally, a diffusion length of C2H4 on Ag(111) at T = 25K large enough forethylene to physically interact with the active sites at the surface coverage presentedin topographs (a,b,c) was observed. The colored arrows in Fig 5.2 show differentmolecule terminations along with free molecules on the surface throughout stagesof the experiment, with a key shown in the upper right of the figure. CO prebondmotifs (gray arrows) are present in all topographs, as expected due to low COdiffusion barrier of T ∼ 11K and residual presence of CO from ncAFM experiments.From (b) to (c) one tpy-Fe node (red arrow, b) changes to a CO prebond node(gray arrow, c), and tpy-Fe interactions with CO are the only changes to activesites detected when the sample was annealed to Tmax = 25K. In order to inducean interaction between tpy-Fe active sites and free C2H4 on the surface, sampleannealing temperatures of T ∼ 30K were required.5.2 C2H4 Interactions with Fe-tpy: Visual DataInteractions between bare tpy-Fe active sites and C2H4 are displayed in Fig 5.3,which includes topographs of the same sample area before (a) and after (b) annealinga sample decorated with ethylene and active sites at T = 30K for t = 5min. Thedifference topograph (b-a), also included in Fig 5.3, highlights changes to the samplepre and post anneal.615.2. C2H4 INTERACTIONS WITH FE-TPY: VISUAL DATAFigure 5.3: Topograph after dosing sample with ethylene at PLT ∼ 8.4x10−9mbarfor t = 2.5min with Tmax = 11.1K (a) and topograph after annealing the samesample at T = 30K for t = 5min (b). Arrows track similarities and differencesto the active sites pre and post anneal. The difference topograph (b-a) highlightsthese changes. Topograph (a) is exactly the same as topograph (a) in Fig 5.2. Baretpy-Fe active sites (red arrows, a) morph into C2H4 prebond (light purple arrows,b), Fe-C2H4 bond (dark purple arrow, b), and CO prebond (additional gray arrow,b) motifs after the anneal. Additionally, free ethylene molecules on the surface (tealarrows, a) are attracted to nitrogens of uncoordinated tpy groups (teal arrows, b)after annealing. An interaction potentially consisting of Fe-CO and C2H4 (pinkarrow, b) is also labeled. Vb = 20mV , It = 20pA (a,b).Topograph (a) in Fig 5.3 corresponds to a sample dosed with ethylene andannealed up to Tmax = 25K prior to imaging.2 Between 25K < T ≤ 30K, achange in structure of the bare active sites is evident, deduced as the interactionbetween C2H4 and tpy-Fe. Namely, two different node terminations are observedcontaining ethylene; the ‘C2H4 prebond’ (light purple arrows, b), and the ‘Fe-C2H4bond’ (dark purple arrow, b). Both of these terminations are rounded comparedto the bare tpy-Fe sites (red arrows), and the Fe-C2H4 bond configuration (darkpurple arrow) is characterized by an intensity maxima centered on the node, whereasthe C2H4 prebond motif (light purple arrows) has a more uniform distribution ofintensity on the node. The difference topograph (b-a), clearly shows the localized2Fig 5.3 (a) was cropped from the same image as Fig 5.2 (c).625.2. C2H4 INTERACTIONS WITH FE-TPY: VISUAL DATAbright region on the Fe-C2H4 bond motif. At the C2H4 prebond motifs in thedifference topograph, the nodes are terminated by bright spots comparable in size toa single ethylene molecule. These bright regions for the C2H4 prebond are slightlyoutward from the tpy-Fe site suggesting, along with evidence to follow, that theC2H4 prebond motif consists of a single C2H4 very near the active site, but notstrongly bonded to the metal center.Figure 5.4: Topographs of the C2H4 prebond (a) and bond (d) configurations imagedat Vb = 20mV and It = 10pA. Laplace-filtered ncAFM images of the C2H4 prebond(b) and bond (e) motifs imaged at constant height with Avib = 50mV from asetpoint of Vb = 20mV and It = 10pA on Ag(111). ∆z = −0.01nm for (b), and∆z = −0.022nm for (e). (a,d,b,e) were all imaged with the same CO-terminatedtip. The proposed chemical structures are drawn for the prebond (c) and bond (f)configurations where it is postulated that ethylene is first close to the tpy-Fe center,but still bonded with the substrate in the prebond configuration, and then, afterovercoming an energy barrier, ethylene binds with the metal center in the bondconfiguration. Shear transformation performed on (b) to compensate for drift.Fig 5.3 also shows an interaction between C2H4 (teal arrows) and nonmetalatedtpy groups (blue arrows) where C2H4 molecules are imaged free on the Ag(111)surface prior to annealing (teal arrow, a), and C2H4 molecules are imaged near baretpy groups after annealing (teal arrow, b). The nitrogen lone pair on the pyridineligand of bare tpy groups can capture C2H4 molecules through weak hydrogenbonding forming a C-H2...N conformation. In the difference topograph, these Hbonded C2H4 molecules appear as bright spots similar to those observed terminatingthe C2H4 prebond motifs. A CO prebond node (gray arrow) is also present intopography images (a) and (b), as expected. Finally, an interaction potentiallybetween an Fe-CO bond configuration with an additional C2H4 molecule is marked(pink arrow, b), this is a ‘higher order’ interaction. This Fe-CO-C2H4 structure wasobserved repeatedly throughout experiments, but requires additional investigationto confirm its composition and bonding configuration. This unconfirmed, higher635.3. DISSOCIATION OF FE-C2H4 BONDSorder node, along with other higher order interactions are discussed further in theConclusion.To elucidate internal structures of C2H4 prebond and bond motifs, ncAFMmeasurements on the two species were performed. Fig 5.4 presents these ncAFMimages along with close up STM topographs and proposed chemical structures ofthe prebond and bond motifs. As mentioned previously, the prebond motif likelyconsists of an C2H4 molecule close to the tpy-Fe center, but still bonded to theAg substrate (c), whereas the bond motif consists of the C2H4 molecule bonded tothe tpy-Fe center (f). Topographs (a) and (d) show the C2H4 prebond and C2H4bond structures, respectively, both imaged with the same CO-terminated tip. Thelocalized bright region for the bond configuration is visible, and the slightly outwardposition of the ethylene for the prebond is also apparent in the topographs.The difference between prebond and bond motifs in the ncAFM images is subtleat first glance. Weak Ag...C2H4 bonds (prebond configuration) and stronger Fe-C2H4bonds (bond configuration) both primarily involve a combination of donation andback-donation of electron density around the C=C bond of the ethylene molecule.Thus both configurations would likely maintain geometry with the C=C bond parallelto the substrate plane. This explains why a large difference in node terminations isnot evident between the prebond and bond configurations, as was the case with thenodes involving CO (see Fig 4.4 (b,e)).The small differences evident between the ncAFM images of C2H4 prebond andbond motifs provide valuable information on the geometry of these interactions. Thebond termination (e) consists of a uniformly-dark, mostly-vertical line, whereas theprebond termination (b) is more arched outward and shows a more stark variation incontrast between the middle of the termination, where the ethylene is located, andthe top and bottom of the terpyridine group. Upon bonding with Fe, the ethylene isexpected to move closer to the metal center, both inward and downward, becausethis stronger interaction should consist of a shorter bond length,3 This expecteddecrease in separation of the ethylene from the active Fe center upon bonding isconsistent with observations between (b) and (e) and support the account of asurface-mediated C2H4 prebond step preceding the Fe-C2H4 bond configuration.5.3 Dissociation of Fe-C2H4 BondsTo confirm the presence of ethylene in the prebond and bond motifs voltage pulseswere applied near the altered active sites and changes including dissociation ofFe-C2H4 bonds were observed. This technique of using the bias voltage to dissociateindividual molecular bonds was discovered and implemented soon after the inventionof the STM, and is a powerful tool in surface chemistry analysis. [31, 84]3Henze et al. [132] observed shorter vertical bond distances for PTCDA in the chemisorbed statethan physisorbed state. Liu et al. [133] concluded that weak physisorption of organic moleculesresults in longer vertical bond lengths using benzene as a calculated example.645.3. DISSOCIATION OF FE-C2H4 BONDSFigure 5.5: Series of topographs (a,b,c,d,e,f,g) acquired consecutively at the sameregion of a sample with setpoint parameters Vb = −25mV and It = 50pA. Frames(a)-(e) were subjected to a voltage pulse after image acquisition was complete. Thepulse was applied approximately at the location marked by the yellow ‘x’ in eachrelevant frame, and all pulses were Vb = −0.7V for t = 200ms. No pulse was appliedafter frame (f) acquisition, and the same image (d) is repeated in both columns forease of comparison to both (d-c) and (e-d). Difference topographs (b-a,c-b,d-c,e-d,f-e,g-f) highlight the changes from frame to frame. By applying pulses at differentlocations on the sample node structure changes from bond to prebond to bare activesites were observed.655.3. DISSOCIATION OF FE-C2H4 BONDSFig 5.5 shows the same area of a sample exposed to a series of bias pulses alongwith the resulting changes to node structure. For topographs (a)-(e) in Fig 5.5, atthe end of each image acquisition a bias pulse of Vb = −0.7V for t = 200ms wasapplied approximately at the yellow ‘x’ in the frame. From (a) to (b) the lowernode changed from Fe-C2H4 bond to C2H4 prebond motif, clearly visible in thedifference topograph (b-a). Applying a voltage pulse at the same location afterframe (b) acquisition, resulted in the prebond node changing back to the bondconfiguration (compare frames (b) and (c) along with the difference (c-b)). Afterframe (c) acquisition, the pulse was applied in closer proximity to the lower Fe-C2H4bond node, which resulted in complete dissociation of the Fe-C2H4 bond leaving abare tpy-Fe active site and additional free C2H4 on the surface in frame (d) (compareframes (c) and (d) along with the difference (d-c)).Next a pulse was applied near the upper Fe-C2H4 bond node, after frame (d)acquisition was complete, which resulted in dissociation of that upper node as well asthe lower tpy-Fe picking up a CO molecule to form the CO prebond motif (compareframes (d) and (e) along with the difference (e-d)). Because of the low diffusionbarrier of CO on the surface, along with the changes in potential energy landscapeinduced by the voltage pulse, it is not surprising a CO molecule makes an appearanceduring this experiment. Finally, after frame (e) acquisition a pulse was applied inthe same position as frame (d), which caused the CO prebond motif to change backto the bare tpy-Fe node and slight movements in the free C2H4 molecules were alsoobserved (compare frames (e) and (f) along with (f-e)). After obtaining frame (f),no pulse was applied, the microscope scanned frame (g) immediately, and a changein the lower node from bare tpy-Fe to CO prebond was observed (compare frames(f) and (g) along with the difference (g-f)). This node change induced by scanningonly (no pulse) again affirms the ease of mobility of the CO molecules as well as theinstability of the CO prebond motif that was previously discussed in Fig 4.7.Pulsing experiments dissociating Fe-C2H4 bond motifs were readily reproducible,and the results of one other pulse experiment is shown in Fig 5.6. In Fig 5.6 (a),three Fe-C2H4 bond configurations are imaged on the sample (dark purple arrows),and after image (a) acquisition was complete a pulse of Vb = 2.0V for t = 20mswas applied near the upper left corner of (a). The same area of the sample imagedafter the pulse is shown in topograph (b). Here the three previous Fe-C2H4 bondmotifs have dissociated leaving three additional bare tpy-Fe active sites and threeadditional free C2H4 molecules on the surface. Difference topograph (b-a) in Fig 5.6emphasizes the changes induced by the bias pulse.Although Fe-C2H4 bond dissociation was easily reproduced, Fe-C2H4 bondformation was never observed during pulse experiments. Also the Fe-C2H4 bondto prebond shift, or vice versa, was infrequently observed. Ethylene motion on thesubstrate as a result of pulsing was prevalent throughout the pulse experiments,but the direction and magnitude of the movement proved difficult to tune. Onecould possibly investigate pulse induced bond formation further as the direction ofpulse-driven molecular motion has been precisely controlled in other systems, [105]and pulse-induced bond formation has also been observed in other systems. [83]665.3. DISSOCIATION OF FE-C2H4 BONDSFigure 5.6: Topographs before (a) and after (b) a pulse at Vb = 2.0V for t = 20mswas applied in the upper left corner of (a) near the yellow ‘x’. Three Fe-C2H4bonding nodes in (a) dissociate in response to the bias pulse resulting in threeadditional bare tpy-Fe nodes and three additional free C2H4 molecules visible onthe surface in (b). The difference topograph (b-a) emphasizes these pulse inducedchanges. Vb = 20mV , It = 20pA (a,b).675.4. C2H4 INTERACTIONS WITH FE-TPY: STATISTICS5.4 C2H4 Interactions with Fe-tpy: StatisticsThe observation of the C2H4 prebond motif as a surface-supported intermediate tothe Fe-C2H4 bond is supported by reaction statistics presented in this section. Toobtain the reaction statistics, samples containing ethylene and active sites4 wereprobed and annealed at temperatures up to T = 40K, and the number of differentmotifs were counted over large sample areas at each temperature increment, thesame procedure as for CO reaction statistics. Details on the statistical analysis arepresented in Appendix D.3.Figure 5.7: Counted C2H4 prebond and Fe-C2H4 bond motifs at different stagesin a reaction. Gray bars in the background show the max substrate temperaturereached corresponding to each experimental stage. The * indicates a software resetand new location on the sample being probed. As annealing temperatures increase,a shift in node terminations from prebond to bond is observed.Fig 5.7 summarizes the statistical findings for ethylene prebond and bond con-figurations, where the percentage of active sites occupied by the C2H4 prebond orFe-C2H4 bond motifs were tracked throughout experimental stages of dosing andannealing. The data presented here was acquired at the same time as the data forexperiment (c) in Fig 4.10 for the CO prebond and bond configurations. Maximumsubstrate temperatures reached at each experimental step are indicated by thebackground gray bars in Fig 5.7. The * indicates a software reset occurred and newarea of the sample was probed after that point. 5 In Fig 5.7, below T = 30K, theC2H4 prebond and bond counts are both quite low. At T = 30K, there is a notable4Residual CO was also present on these samples. The surface density of CO and C2H4 combinedwas much less than the density of active sites and competition for active sites between CO andC2H4 was assumed negligible.5The software reset was caused by a crash of the Matrix SPM controller, which caused a tipcrash and required coarse motion to a new area of the sample. The structures observed after thecrash were similar to the initial starting conditions after dosing, but prior to annealing.685.5. STS ON FE-C2H4 BOND CONFIGURATIONincrease in the percentage of active sites occupied by C2H4. At T = 35K, thenumber of prebond configurations drops significantly while the bond configurationcount increases compared to the previous annealing temperature at T = 30K. Thetrend of prebond counts decreasing and bond configurations increasing continuesat T = 40K. This statistical analysis of prebond and bond configurations supportthe notion of the C2H4 prebond being a metastable intermediate to the more stableFe-C2H4 bond configuration. As temperature on the substrate increases, ampleenergy is available for the Fe-C2H4 bond to form and its stability likely preventsswitching back to C2H4 prebond configurations.5.5 STS on Fe-C2H4 Bond ConfigurationThe final evidence supporting the observed trend from C2H4 prebond to bondmotifs is shown in spectroscopy data presented in this section. In comparing STSmeasurements on bare tpy-Fe active sites with measurements over C2H4 prebondand C2H4 bond configurations, differences in the electronic structure between thespecies were illuminated. Fig 5.8 details these differences; the most remarkablefeature of which is a clear shift in the Fe state of the bare active site upon bondingwith ethylene to form the Fe-C2H4 bond motif.Fig 5.8 (a) shows dI/dV averaged over the bare tpy-Fe active site of a singly-coordinated TPT (see topography inset for exact region averaged over to generateFe-tpy active site spectra). As presented in Chapter 3.4, the Fe state energy signatureis located at Vb = −0.09V . The line STS in (c) shows dI/dV averaged over theC2H4 prebond (light purple) and Fe-C2H4 bond (dark purple) configurations on twodifferent molecules (see topography insets for precise areas averaged over)6. Alongwith those two spectra, the background Ag(111) spectrum from the bond motifmeasurement is included for a baseline comparison. The C2H4 prebond signal doesnot show any tunneling resonances within the measured bias range, however, the Fe-C2H4 bond spectrum shows a clear resonance at Vb = −0.31V . This new electronicsignature of the Fe-C2H4 bond motif is a clear indication that bonding between theactive site and the ethylene molecule has occurred. This shift in electronic structureis emphasized by the black arrow between (a) and (c). Additionally, the Fe-C2H4resonance at Vb = −0.31V presented here is different from the Fe-CO resonance atVb = −0.165V presented in Fig 4.11; the electronic signatures are unique as theconstituents of the bonds are unique.Fig 5.8 (b) and (d) display normalized dI/dV maps of a singly-coordinated TPTat the Fe resonance, and a Fe-C2H4 bond motif at the Fe-C2H4 bonding resonance,respectively. The distribution of electronic structure around the Fe-C2H4 bond in(d) is broad compared to the localized Fe state resonance in (b). This broad spatialdistribution of the Fe-C2H4 resonance could be influenced by the presence of the6The area averaged over to generate the C2H4 prebond spectrum (light purple circle, b) isopposite a Fe-CO bond termination on the same molecule. The Fe-CO signal from this moleculewas compared to a lone Fe-CO termination signal, and they matched. Thus, having differentterminations on the same molecule does not significantly impact the electronic signature.695.5. STS ON FE-C2H4 BOND CONFIGURATIONFigure 5.8: dI/dV averaged over the bare tpy-Fe active site (a) and averaged overthe C2H4 prebond and C2H4 bond motifs (c). Background Ag(111) spectra areincluded for baseline comparisons. STS at different setpoint conditions were rescaledby the tunnelling resistance at the setpoint: dI/dVIt/Vb . Inset topographs show the areasaveraged over in generating the line STS for the tpy-Fe signal (red circle, a), C2H4prebond signal (light purple circle, b), and Fe-C2H4 bond signal (dark purple circle,b, CO-terminated tip). The Fe state resonance at Vb = 0− 0.09V (a) shifts uponbonding with ethylene to a resonance at Vb = −0.31V (dark purple, b). The C2H4prebond configuration (light purple, b) does not transmit a tunneling resonancewithin the measured bias range. Normalized dI/dV maps show the spatial electronicdistribution of the bare active site at the Fe state energy (b), as well as the Fe-C2H4bond motif at the Fe-C2H4 resonance energy (d). Since the prebond motif does notemit a characteristic signal, no (dI/dV)/(I/V) map of this species is included.705.5. STS ON FE-C2H4 BOND CONFIGURATIONC=C bond likely lying parallel to the Fe center and vertically oriented with respectto image (d).As was discussed in detail in Chapter 2, the dI/dV is proportional to the localdensity of states of the sample, and the local density of states of organic moleculesnear the Fermi energy consists of frontier molecular orbitals that are primarilyinvolved in bonding. Since the Fe-C2H4 system is being probed around the Fermienergy, the electronic shift indicated in Fig 5.8 is strong evidence that bondingbetween the active metal center and the ethylene molecule has occurred. Additionally,the lack of signal for the C2H4 prebond configuration likely indicates no binding tothe metal site has occurred for the prebond motif.71Chapter 6Conclusion6.1 ConclusionsBonding was observed between small gaseous adsorbates, CO and C2H4, and activeFe-tpy sites supported on a Ag(111) surface at T ≤ 40K for this thesis. Bondingbetween Fe and the gaseous adsorbates was confirmed through an electronic signaturechange in the STS data, and a physical structure change observed in the ncAFMdata. Both of these measured bonds, Fe-CO and Fe-C2H4, were preceded by anintermediate prebond configuration where the adsorbate interacted with the Fecenter, but was still bonded with the surface. The prebond motif was characterizedby no measurable change in the electronic structure in the STS data, and observedstructural changes in STM topographs and ncAFM images. Statistical data supportsthe observed trend from prebond to bond configurations for the interaction betweenFe-tpy with C2H4 as temperature increased. At higher annealing temperatures andwith more reactants dosed on the sample, higher order structures appeared withincreased frequency, which indicates that the prebond and bond configurations areintermediates along a reaction pathway.6.2 Open QuestionsThe observed Fe-CO and Fe-C2H4 bonds appear to be the beginning stage in areaction involving more than one small-molecule reactant bonding to the activeFe-tpy sites. These higher order node interactions are shown in Fig 6.1, and wereobserved predominantly at increased surface coverage of CO and C2H4 along withhigher annealing temperatures (30K < T < 45K). The close-up topographs inFig 6.1 highlight distinguishing features of the higher order nodes. The higher ordernode 1 (Fig 6.1(a)) has a half-ring termination, similar to the Fe-CO bond, withwhat appears to be an additional ethylene molecule overlapping the ring. This ismerely an educated guess based on topographic features, but definitive informationon physical structure of this node, and other higher order nodes, would require726.2. OPEN QUESTIONSadditional evidence (such as structural ncAFM data or electronic STS data) since thetopography is a convolution of physical and electronic structure. Node 2 (Fig 6.1(b))has several round regions near the termination similar in size to individual ethylenemolecules. Node 3 (Fig 6.1(c)) and node 4 (Fig 6.1(d)) show a sizable intensityincrease at the node center where (c) consists of a circular region asymmetricallyset on the node, and (d) has a two-lobed structure centered on the node. Whilescanning node 3, noise in the It near the bright central region was often apparentindicating an unstable or mobile feature under the tip. The dark regions in thelower half of (b) and (d) are CO molecules bound weakly to peripheral pyridines,and are superfluous to defining the node structure. Due to the higher surfacecoverage of gaseous adsorbates, CO molecules were commonly observed bonded tonon-coordinated sides of molecules displaying higher order interactions. Ethylenemolecules, however, were not typically observed at the pyridine position on sampleswith many higher order nodes.Figure 6.1: Topographs of four repeatedly observed higher order structures denotedhigher order 1 through higher order 4 (a through d). CO molecules weakly boundto the uncoordinated sides of the molecules in 2 and 4 (to the leftmost pyridine in 2and the rightmost pyridine in 4) are dispensable in defining the coordinated-sidenode structures. Vb = 20mV , It = 20pA (a-d).Additional data on the higher order node 3 is presented in Fig 6.2. The whitebox in overview topograph (a) of Fig 6.2 encloses a molecule with a higher ordernode 3. The same molecule is imaged with a CO-terminated tip in the close-uptopograph (b), where the white arrows point towards the characteristic noisy region.A CO molecule is also present bonded to the lower right pyridine in (b). The ncAFMimage (c) shows additional internal structure at the higher order node 3. This imagewas acquired at two different height offsets: ∆z = −20pm from Vb = 20mV andIt = 10pA on Ag(111) in the bottom half of the image, and ∆z = +20pm from thesame setpoint in the upper half of the image. It was necessary to move the tip awayfrom the molecule in the upper half of the ncAFM image to avoid tip crash, andstill resolve internal structure. The internal structure of the higher order node 3 is736.2. OPEN QUESTIONSpronounced near the center, with connecting features mostly towards the right side,and a circular ring enclosing the entire node. Since the CO tip is flexible, and thenoise at the node is a common signature of an instability, it is difficult to deriveinsight from this data at this point. Chemical identification of this node along withthe other three higher order nodes is an open area for future investigations.Figure 6.2: STM and ncAFM images of a higher order 3 interaction. Overviewtopograph taken with a Ag-terminated tip (a) with the white boxed region enclosingthe higher order 3 node. Close-up topograph (b) acquired with a CO-terminated tipof the same enclosed node from (a). The white arrow in (b) points toward a noisyregion in the topography, characteristic of this higher order 3 motif. A CO moleculeis visible on the lower, rightmost pyridine in (b). ncAFM image (c) of the samemolecule shown in (a) and (b). The constant height offset began at the bottom of(c) at ∆z = −20pm from Vb = 20mV and It = 10pA on Ag(111) and was manuallyadjusted at the horizontal line in (c) to ∆z = +20pm.Beyond deciphering the structure of higher order nodes, an annealing experimentperformed on a sample containing many higher order nodes yielded a new form onthe surface separate from the active sites, which may be of greater interest. Thatparticular experiment started with a sample that had previously been exposed to a746.2. OPEN QUESTIONShigh pressure dose of ethylene (PLT = 3 · 10−8mbar with Tmax = 6.7K for t = 5sec)and sample annealing up to Tmax = 30K for t = 5min. That sample is shown inFig 6.3 (a-c), and contained many higher order nodes labeled with arrows withinthe figure. The labeling key for all motifs in this figure is inset in Fig 6.3 topograph(d). Subsequently, that sample was annealed at room temperature T ∼ 298K byremoving it from the STM head into the LT chamber and returning it to the headafter t ∼ 10min. The sample after the room temperature anneal is shown in Fig 6.3(d).Figure 6.3: Topographs (a-c) show a sample after annealing to Tmax = 40K andbefore annealing at room temperature. The sample after annealing at room tem-perature T ∼ 298K for t = 10min is shown in (d). Before the room temperatureanneal, the sample contained many higher order nodes, as pictured and labeled witharrows in (a), (b), and (c). The labeling key is inset in (d). After the anneal, newstructures, potentially oligomers, were imaged on the sample and are labeled withmagenta arrows. These new structures are similar in appearance to ethylene, buthave apparent heights twice as large.After annealing the sample at room temperature, new forms are visible on thesample and labeled with magenta arrows in Fig 6.3 (d). These new forms arepotentially oligomers of more than one C2H4 and/or CO molecule. The potentialoligomers image similar to ethylene on the surface, but have an apparent heightaround twice that of ethylene when imaged at the same bias, which will be discussedlater in this section in Fig 6.4. After annealing, several 80nm2 images were acquiredwith all but one containing two or three of the potential oligomer structures. It756.2. OPEN QUESTIONSFigure 6.4: Topographs comparing an ethylene molecule (a) and a potential oligomer(b). Several oligomers were observed on a sample that contained many higher orderinteractions (see Appendix A 4.1-4.4), and was subsequently annealed at roomtemperature, T ∼ 298K, for t = 10min. The height profiles of the ethylene (c) andoligomer (d) show the oligomer is more than twice the apparent height than theethylene under the same bias voltage, Vb = 20mV , It = 50pA (a,b).766.3. FUTURE DIRECTIONSis possible that these potential oligomers are small Fe clusters that have becomeuncoordinated from tpy groups, however, this does not seem likely as Fe typicallyforms larger, nonuniform clusters on the surface. This experiment was repeated ona different sample containing mainly prebond and bond configurations of CO andC2H4, and no oligomer-like structures were observed after the room temperatureanneal. This is surprising since there should be ample energy at room temperatureto quickly progress through all reaction steps observed at low temperature, andresult in the same product being formed. However, the surface coverages of COand C2H4 for the repeated experiment were much lower than those on the samplepictured in Fig 6.3, which could simply result in less product being formed andperhaps not enough of the surface was sampled to detect the products formed.Fig 6.4 directly compares topographs and profiles of ethylene (a,c) and theobserved oligomer structure (b,d). The topographs (a) and (b) were both obtainedat the same scanning parameters of Vb = 20mV and It = 50pA, and appear quitesimilar, except for the apparent height maximum in (a) is only z = 30pm whereasfor (b) it is z = 70pm. The light blue lines in topographs (a) and (b) trace theprofiles shown in (c) and (d), respectively. The oligomer structure in (d) has anapparent height more than twice that of ethylene in (c), and this larger apparentheight was consistently measured in all oligomer structures observed on the sampleafter the room temperature anneal. If the structure in (b) is in fact a stable oligomer,such as C4H8, this would verify that Fe-tpy complexes are robust surface-supportedcatalysts for hydrocarbon oligomerization. In turn, the fundamental understandingof the reaction mechanism for this oligomerization of ethylene using Fe-tpy catalystscould aide in optimizing many large-scale industrial technologies and processes inthe future.6.3 Future DirectionsSpecific directions future projects may take include:• Determining the composition of the higher order nodes presented in this chapter• Reproducing the potential oligomerization reaction described in this chapterand characterizing the reaction mechanism and products• Probing the reactivity of Fe-tpy with other gaseous adsorbates• Differentiating reactivity of singly-coordinated versus doubly-coordinated TPTmolecules• Developing machine learning algorithms to properly categorize and trackstatistical data for surface reactions (see Appendix D.3 for more details)77Bibliography[1] E. Groppo, C. Lamberti, S. Bordiga, G. Spoto, and A. Zecchina. The structureof active centers and the Ethylene Polymerization mechanism on the Cr/SiO2catalyst: A frontier for the characterization methods. Chemical Reviews 105.1(2005), pp. 115–183. doi: 10.1021/cr040083s.[2] M. Konsolakis. Surface Chemistry and Catalysis. Vol. 6. 102. Shu-Kun Lin,2016.[3] X.-F. Yang et al. Single-Atom Catalysts: A New Frontier in HeterogeneousCatalysis. Accounts of Chemical Research 46.8 (2013), pp. 1740–1748. doi:10.1021/ar300361m.[4] A. Riss et al. Imaging single-molecule reaction intermediates stabilized bysurface dissipation and entropy. Nature Chemistry 8.7 (2016), pp. 678–683.doi: 10.1038/nchem.2506.[5] F. Abild-Pedersen et al. The nature of the active site in heterogeneousmetal catalysis. Chemical Society Reviews 37 (2008), pp. 2163–2171. doi:10.1039/b800260f.[6] J. K. Nørskov. Catalysis frozen in time. Nature 414.6862 (2001), pp. 405–406.doi: 10.1038/35106674.[7] B. K. Min and C. M. Friend. Heterogeneous gold-based catalysis for greenchemistry: Low-temperature CO oxidation and propene oxidation. ChemicalReviews 107.6 (2007), pp. 2709–2724. doi: 10.1021/cr050954d.[8] K. C. Taylor. Automobile Catalytic Converters. Berlin: Springer-Verlag, 1984.doi: 10.1007/978-3-642-69486-8.[9] P. Bagot. Fundamental surface science studies of automobile exhaust catalysis.Materials Science and Technology 20.6 (2004), pp. 679–694. doi: 10.1179/026708304225016789.[10] K. H. Kim and S. K. Ihm. Heterogeneous catalytic wet air oxidation ofrefractory organic pollutants in industrial wastewaters: A review. Journalof Hazardous Materials 186.1 (2011), pp. 16–34. doi: 10.1016/j.jhazmat.2010.11.011.78BIBLIOGRAPHY[11] H. Amano, N. Sawaki, I. Akasaki, and Y. Toyoda. Metalorganic vapor phaseepitaxial growth of a high quality GaN film using an AlN buffer layer. AppliedPhysics Letters 48.5 (1986), pp. 353–355. doi: 10.1063/1.96549.[12] P. D’Orazio. Biosensors in clinical chemistry - 2011 update. Clinica ChimicaActa 412.19-20 (2011), pp. 1749–1761. doi: 10.1016/j.cca.2011.06.025.[13] C. M. Friend and B. Xu. Heterogeneous catalysis: A central science for asustainable future. Accounts of Chemical Research 50.3 (2017), pp. 517–521.doi: 10.1021/acs.accounts.6b00510.[14] D. Skomski, C. D. Tempas, K. A. Smith, and S. L. Tait. Redox-active on-surface assembly of metal-organic chains with single-site Pt(II). Journal ofthe American Chemical Society 136.28 (2014), pp. 9862–9865. doi: 10.1021/ja504850f.[15] C. D. Tempas et al. Redox-Active ligand controlled selectivity of vanadiumoxidation on Au(100). Chemical Science 9.6 (2018), pp. 1674–1685. doi:10.1039/c7sc04752e.[16] Q. Li et al. RSC Advances Impact of surface arrangement and compositionon ethylene adsorption over Pd – Ag surface alloys : a computational study †.70932 (2016), pp. 70932–70942. doi: 10.1039/c6ra12912a.[17] R. Ren, Y. Lü, X. Pang, and G. Wang. Metal catalyzed ethylene epoxidation:A comparative density functional theory study. Journal of Natural GasChemistry 20.3 (2011), pp. 303–310. doi: 10.1016/S1003-9953(10)60176-4.[18] G. E. Brown et al. Metal oxide surfaces and their interactions with aqueoussolutions and microbial organisms. Chemical Reviews 99.1 (1999), pp. 77–174.doi: 10.1021/cr980011z.[19] H. Taylor. A Theory of the Catalytic Surface. Proc. R. Soc. 108.3 (1925),pp. 105–111.[20] A. Kokalj, A. Dal Corso, S. De Gironcoli, and S. Baroni. The interactionof ethylene with perfect and defective Ag(001) surfaces. Journal of PhysicalChemistry B 106.38 (2002), pp. 9839–9846. doi: 10.1021/jp025823k.[21] G. A. Somorjai, K. R. McCrea, and J. Zhu. Active sites in heterogeneouscatalysis: Development of molecular concepts and future challenges. Topicsin Catalysis 18.3-4 (2002), pp. 157–166. doi: 10.1023/A:1013874202404.[22] A.-K. Henß et al. Density fluctuations as door-opener for diffusion on crowdedsurfaces. Science 363.6428 (2019), pp. 715–718. doi: 10.1126/science.aav4143.[23] M. Kulawik, H.-P. Rust, N. Nilius, M. Heyde, and H.-J. Freund. STM studiesof ordered (√31×√31) R9° CO islands on Ag(111). Physical Review B 71.15(2005), p. 153405. doi: 10.1103/PhysRevB.71.153405.79BIBLIOGRAPHY[24] C. Krull et al. Iron-based trinuclear metal-organic nanostructures on a surfacewith local charge accumulation. Nature Communications 9.1 (2018). doi:10.1038/s41467-018-05543-4.[25] B. Voigtländer. Scanning Probe Microscopy. 2015, p. 375. doi: 10.1007/978-3-662-45240-0.[26] C. J. Chen. Introduction to Scanning Tunneling Microscopy Second Edition.Oxford University Press (2008), (2008). doi: 10.1002/jemt.1070280110.[27] M. Kulawik et al. Interaction of CO molecules with surface state electronson Ag(1 1 1). Surface Science 590.2-3 (2005). doi: 10.1016/j.susc.2005.05.068.[28] Q. Shen, H. Y. Gao, and H. Fuchs. Frontiers of on-surface synthesis: Fromprinciples to applications. Nano Today 13 (2017), pp. 77–96. doi: 10.1016/j.nantod.2017.02.007.[29] V. Navarro, M. A. Van Spronsen, and J. W. Frenken. In situ observation ofself-assembled hydrocarbon Fischer-Tropsch products on a cobalt catalyst.Nature Chemistry 8.10 (2016), pp. 929–934. doi: 10.1038/nchem.2613.[30] M. Bowker and P. R. Davies, eds. Scanning Tunneling Microscopy in SurfaceScience, Nanoscience and Catalysis. Weinheim, Germany: Wiley-VCH VerlagGmbH & Co. KGaA, 2010.[31] B. C. Stipe et al. Single-molecule dissociation by tunneling electrons. PhysicalReview Letters 78.23 (1997), pp. 4410–4413. doi: 10.1103/PhysRevLett.78.4410.[32] L. Bartels, G. Meyer, and K. Rieder. Dynamics of Electron-Induced Manipu-lation of Individual CO Molecules on Cu(111). Physical Review Letters 80.9(1998), pp. 2004–2007.[33] S.-w. Hla, L. Bartels, G. Meyer, and K.-h. Rieder. Inducing All Steps of aChemical Reaction with the Scanning Tunneling Microscope Tip: TowardsSingle Molecule Engineering. Physical Review Letters 85.13 (2000), pp. 2777–2780.[34] S.-W. Hla and K.-H. Rieder. STM Control of Chemical Reactions: Single-Molecule Synthesis. Annual Review of Physical Chemistry 54.1 (2003),pp. 307–330. doi: 10.1146/annurev.physchem.54.011002.103852.[35] E. Meyer, H. J. Hug, and R. Bennewitz. Scanning Probe Microscopy: The Labon a Tip. Berlin Heidelberg: Springer-Verlag, 2004. doi: 10.1016/j.elspec.2004.02.002.[36] P. Gambardella et al. Oxygen dissociation at pt step. Physical Review Letters87.5 (2001), pp. 56103–1–56103–4. doi: 10.1103/PhysRevLett.87.056103.[37] J. P. Hahn and W. Ho. Oxidation of a single carbon monoxide moleculemanipulated and induced with a scanning tunneling microscope. PhysicalReview Letters 87.16 (2001), pp. 1–4. doi: 10 . 1103 / PhysRevLett . 87 .166102.80BIBLIOGRAPHY[38] T. Zambelli, J. Wintterlin, J. Trost, and G. Ertl. Identification of the "ActiveSites" of a Surface-Catalyzed Reaction. Science 273.5282 (1996), pp. 1688–1690. doi: 10.1126/science.273.5282.1688.[39] M. L. Bocquet et al. New insights into ethene epoxidation on two oxidizedAg{111} surfaces. Journal of the American Chemical Society 125.19 (2003),pp. 5620–5621. doi: 10.1021/ja0297741.[40] B. C. Stipe. Single-Molecule Vibrational Spectroscopy and Microscopy. Sci-ence (New York, N.Y.) 280.5370 (1998), pp. 1732–1735. doi: 10.1126/science.280.5370.1732.[41] Z. Shi and N. Lin. Self-assembly of a two-dimensional bimetallic coordinationframework and dynamic control of reversible conversions to homo-metallichydrogen-bond arrays. ChemPhysChem 11.1 (2010), pp. 97–100. doi: 10.1002/cphc.200900756.[42] D. Zhong et al. Linear Alkane Polymerization on a Gold Surface. Science 334(2011), pp. 213–216.[43] G. A. Somorjai and A. L. Marsh. Active sites and states in the heterogeneouscatalysis of carbon-hydrogen bonds. Philosophical Transactions of the RoyalSociety A: Mathematical, Physical and Engineering Sciences 363.1829 (2005),pp. 879–900. doi: 10.1098/rsta.2004.1543.[44] J. H. Lunsford. Catalytic conversion of methane to more useful chemicalsand fuels: A challenge for the 21st century. Catalysis Today 63.2-4 (2000),pp. 165–174. doi: 10.1016/S0920-5861(00)00456-9.[45] S. Schauermann and H. J. Freund. Model Approach in Heterogeneous Catal-ysis: Kinetics and Thermodynamics of Surface Reactions. Accounts of Chem-ical Research 48.10 (2015), pp. 2775–2782. doi: 10.1021/acs.accounts.5b00237.[46] X. Chen, K. M. Engle, D. H. Wang, and Y. Jin-Quan. Palladium(II)-cataIyzedC-H aetivation/C-C cross-coupling reactions: Versatility and practicality.Angewandte Chemie - International Edition 48.28 (2009), pp. 5094–5115. doi:10.1002/anie.200806273.[47] A. S. Goldman and K. I. Goldberg. Organometallic C-H Bond Activation :An Introduction (2004).[48] V. Ritleng, C. Sirlin, and M. Pfeffer. Ru-, Rh-, and Pd-Catalyzed C-C BondFormation Involving C-H Activation and Addition on Unsaturated Substrates:Reactions and Mechanistic Aspects. Chemical Reviews 102.5 (2002), pp. 1731–1770. doi: 10.1021/cr0104330.[49] B. L. Small and M. Brookhart. Iron-Based Catalysts with Exceptionally HighActivities and Selectivities for Oligomerization of Ethylene to Linear α-Olefins.Journal of the American Chemical Society 120.28 (1998), pp. 7143–7144. doi:10.1021/ja981317q.81BIBLIOGRAPHY[50] B. L. Small, M. Brookhart, and A. M. A. Bennett. Highly Active Iron andCobalt Catalysts for the Polymerization of Ethylene. Journal of the AmericanChemical Society 120.16 (1998), pp. 4049–4050. doi: 10.1021/ja9802100.[51] A. Winter, G. R. Newkome, and U. S. Schubert. Catalytic applications ofterpyridines and their transition metal complexes. ChemCatChem 3.9 (2011),pp. 1384–1406. doi: 10.1002/cctc.201100118.[52] D. Farrusseng, S. Aguado, and C. Pinel. Metal-organic frameworks: Opportu-nities for catalysis. Angewandte Chemie - International Edition 48.41 (2009),pp. 7502–7513. doi: 10.1002/anie.200806063.[53] P. Liljeroth, I. Swart, S. Paavilainen, J. Repp, and G. Meyer. Single-moleculesynthesis and characterization of metal-ligand complexes by low-temperatureSTM. Nano Letters 10.7 (2010), pp. 2475–2479. doi: 10.1021/nl100834v.[54] D. Skomski et al. Two- and Three-Electron Oxidation of Single-Site VanadiumCenters at Surfaces by Ligand Design. Journal of the American ChemicalSociety 137.24 (2015), pp. 7898–7902. doi: 10.1021/jacs.5b03706.[55] M. Roos et al. Intermolecular vs molecule-substrate interactions: A combinedSTM and theoretical study of supramolecular phases on graphene/Ru(0001).Beilstein Journal of Nanotechnology 2.1 (2011), pp. 365–373. doi: 10.3762/bjnano.2.42.[56] L. Bartels. Tailoring molecular layers at metal surfaces. Nature Chemistry2.2 (2010), pp. 87–95. doi: 10.1038/nchem.517.[57] F. Rosei et al. Properties of large organic molecules on metal surfaces. Vol. 71.5-8 SPEC. 2003, pp. 95–146. doi: 10.1016/S0079-6816(03)00004-2.[58] M. C. Capsoni. On-surface self-assembly and characterization of a macro-molecular charge transfer complex by scanning tunneling microscopy andspectroscopy. Text. University of British Columbia, 2016. doi: 10.14288/1.0305019.[59] R. Gutzler, S. Stepanow, D. Grumelli, M. Lingenfelder, and K. Kern. Mim-icking Enzymatic Active Sites on Surfaces for Energy Conversion Chemistry.Accounts of Chemical Research 48.7 (2015), pp. 2132–2139. doi: 10.1021/acs.accounts.5b00172.[60] S. Stepanow, N. Lin, and J. V. Barth. Modular assembly of low-dimensionalcoordination architectures on metal surfaces. Journal of Physics CondensedMatter 20.18 (2008). doi: 10.1088/0953-8984/20/18/184002.[61] S. Fabris et al. Oxygen dissociation by concerted action of di-iron centers inmetal-organic coordination networks at surfaces: Modeling non-heme iron en-zymes. Nano Letters 11.12 (2011), pp. 5414–5420. doi: 10.1021/nl2031713.[62] W. Hieringer et al. The surface trans effect: Influence of axial ligands onthe surface chemical bonds of adsorbed metalloporphyrins. Journal of theAmerican Chemical Society 133.16 (2011), pp. 6206–6222. doi: 10.1021/ja1093502.82BIBLIOGRAPHY[63] D. Grumelli, B. Wurster, S. Stepanow, and K. Kern. Bio-inspired nanocata-lysts for the oxygen reduction reaction. Nature Communications 4 (2013),pp. 1–6. doi: 10.1038/ncomms3904.[64] A. J. Clough, J. W. Yoo, M. H. Mecklenburg, and S. C. Marinescu. Two-dimensional metal-organic surfaces for efficient hydrogen evolution from water.Journal of the American Chemical Society 137.1 (2015), pp. 118–121. doi:10.1021/ja5116937.[65] B. Wurster, D. Grumelli, D. Hötger, R. Gutzler, and K. Kern. Drivingthe Oxygen Evolution Reaction by Nonlinear Cooperativity in BimetallicCoordination Catalysts. Journal of the American Chemical Society 138.11(2016), pp. 3623–3626. doi: 10.1021/jacs.5b10484.[66] J. Čechal et al. CO2Binding and Induced Structural Collapse of a Surface-Supported Metal-Organic Network. Journal of Physical Chemistry C 120.33(2016), pp. 18622–18630. doi: 10.1021/acs.jpcc.6b05582.[67] M. Feng et al. Cooperative Chemisorption-Induced Physisorption of CO2Molecules by Metal-Organic Chains. ACS Nano 9.12 (2015), pp. 12124–12136.doi: 10.1021/acsnano.5b05222.[68] M. Feng, H. Sun, J. Zhao, and H. Petek. Self-catalyzed carbon dioxideadsorption by metal-organic chains on gold surfaces. ACS Nano 8.8 (2014),pp. 8644–8652. doi: 10.1021/nn5035026.[69] B. E. Murphy et al. Homolytic Cleavage of Molecular Oxygen by ManganesePorphyrins Supported on Ag(111). ACS Nano 8.5 (2014), pp. 5190–5198. doi:10.1021/nn501240j.[70] F. Fache, E. Schulz, M. L. Tommasino, and M. Lemaire. Nitrogen-containingligands for asymmetric homogeneous and heterogeneous catalysis. ChemicalReviews 100.6 (2000), pp. 2159–2231. doi: 10.1021/cr9902897.[71] G. Rothenberg. Catalysis - Concepts and Green Applications. Wiley-VCH,2008. doi: 10.1002/aoc.1408.[72] D. Astruc. Organometallic Chemistry and Catalysis. Grenoble Sciences, 2007.doi: 10.1007/978-3-540-46129-6.[73] C. Haensch et al. Reversible supramolecular functionalization of surfaces:Terpyridine ligands as versatile building blocks for noncovalent architectures.Langmuir 24.22 (2008), pp. 12981–12985. doi: 10.1021/la8026682.[74] Y. Nakayama, Y. Baba, H. Yasuda, K. Kawakita, and N. Ueyama. Stere-ospecific Polymerizations of Conjugated Dienes by Single Site Iron Com-plexes having Chelating N,N,N-Donor Ligands. Macromolecules 36.21 (2003),pp. 7953–7958. doi: 10.1021/ma0300802.[75] H. Hofmeier and U. S. Schubert. Recent developments in the supramolecularchemistry of terpyridine–metal complexes. Chem. Soc. Rev. 33.6 (2004),pp. 373–399. doi: 10.1039/B400653B.83BIBLIOGRAPHY[76] A. Schiffrin et al. Designing Optoelectronic Properties by On-Surface Synthe-sis: Formation and Electronic Structure of an Iron–Terpyridine Macromolecu-lar Complex. ACS Nano (2018), acsnano.8b01026. doi: 10.1021/acsnano.8b01026.[77] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel. Tunneling through a con-trollable vacuum gap. Applied Physics Letters 40.2 (1982), pp. 178–180. doi:10.1063/1.92999.[78] G. Binnig and C. F. Quate. Atomic Force Microscope. Physical Review Letters56.9 (1986), pp. 930–933. doi: 10.1103/PhysRevLett.56.930.[79] N. Kocić et al. Control of reactivity and regioselectivity for on-surface de-hydrogenative aryl-aryl bond formation. Journal of the American ChemicalSociety 138.17 (2016), pp. 5585–5593. doi: 10.1021/jacs.5b13461.[80] S. Kawai et al. Thermal control of sequential on-surface transformation of ahydrocarbon molecule on a copper surface. Nature Communications 7 (2016),pp. 1–3. doi: 10.1038/ncomms12711.[81] T. Komeda, Y. Kim, M. Kawai, B. N. Persson, and H. Ueba. Lateral hopping ofmolecules induced by excitation of internal vibration mode. Science 295.5562(2002), pp. 2055–2058. doi: 10.1126/science.1069016.[82] S. Morita, F. J. Giessibl, E. Meyer, and R. Wiesendanger, eds. NoncontactAtomic Force Microscopy. Vol. 3. Springer, 2015.[83] N. Pavliček et al. Polyyne formation via skeletal rearrangement inducedby atomic manipulation. Nature Chemistry 10.8 (2018), pp. 853–858. doi:10.1038/s41557-018-0067-y.[84] G. Dujardin, R. Walkup, and P. Avouris. Dissociation of Individual Moleculeswith Electrons from the Tip of a Scanning Tunneling Microscope. 255.5049(1992), pp. 1232–1235. doi: 10.1126/science.255.5049.1232.[85] J. Bardeen. Tunnelling from a many-particle point of view. Physical ReviewLetters 6.2 (1961), p. 57. doi: 10.1103/PhysRevLett.41.1825.[86] J. Tersoff and D. Hamann. Theory and application for the scanning tunnelingmicroscope. Physical review letters 50.25 (1983), p. 1998.[87] J. Tersoff and D. R. Hamann. Theory of the scanning tunneling microscope.Physical Review B 31.2 (1985), pp. 805–813. doi: 10.1103/PhysRevB.31.805.[88] S. N. Magonov and M.-h. Whangbo. Surface Analysis with STM and AFM.2007. doi: 10.1002/9783527615117.[89] K. A. Cochrane. Single Molecule Perspectives of Model Organic Semiconduc-tors: Energy Level Mapping by High-Resolution Scanning Probe Microscopy.PhD thesis. University of British Columbia, 2017.[90] D. J. Griffiths. Introduction to Quantum Mechanics. 2nd. New York: Cam-bridge University Press, 2017.84BIBLIOGRAPHY[91] A. Goswami. Quantum Mechanics. 2nd ed. 1997.[92] L. Gross, F. Mohn, N. Moll, P. Liljeroth, and G. Meyer. The ChemicalStructure of a Molecule Resolved by Atomic Force Microscopy. Science325.5944 (2009), pp. 1110–1114. doi: 10.1126/science.1176210.[93] F. J. Giessibl. qPlus Sensor. 1998.[94] F. J. Giessibl. Atomic resolution on Si(111)-(7×7) by noncontact atomic forcemicroscopy with a force sensor based on a quartz tuning fork. Applied PhysicsLetters 76.11 (2000), pp. 1470–1472. doi: 10.1063/1.126067.[95] M. Saint Jean, S. Hudlet, C. Guthmann, and J. Berger. Van der Waals andcapacitive forces in atomic force microscopies. Journal of Applied Physics86.9 (1999), pp. 5245–5248. doi: 10.1063/1.371506.[96] S. Sadewasser and T. Glatzel, eds. Kelvin Probe Force Microscopy: Measuringand Compensating Electrostatic Forces. Springer, 2012. doi: 10.1007/978-3-642-22566-6.[97] N. Moll, L. Gross, F. Mohn, A. Curioni, and G. Meyer. The mechanismsunderlying the enhanced resolution of atomic force microscopy with func-tionalized tips. New Journal of Physics 12 (2010). doi: 10.1088/1367-2630/12/12/125020.[98] P. Hapala et al. Mechanism of high-resolution STM/AFM imaging withfunctionalized tips. Physical Review B 90.8 (2014), pp. 1–20. doi: 10.1103/PhysRevB.90.085421.[99] S. P. Jarvis. Resolving intra- and inter-molecular structure with non-contactatomic force microscopy. International Journal of Molecular Sciences 16.8(2015), pp. 19936–19959. doi: 10.3390/ijms160819936.[100] B. MacLeod. An Ultra-Low-Vibration Facility for Housing a Dilution Temper-ature Scanning Tunneling Microscope. PhD thesis. The University of BritishColumbia, 2015.[101] H. Hövel, B. Grimm, and B. Reihl. Modification of the Shockley-type surfacestate on Ag(111) by an adsorbed xenon layer. Surface Science 477.1 (2001),pp. 43–49. doi: 10.1016/S0039-6028(01)00704-X.[102] K. Morgenstern and J. Nieminen. Imaging water on Ag(111): Field inducedreorientation and contrast inversion. Journal of Chemical Physics 120.22(2004), pp. 10786–10791. doi: 10.1063/1.1737366.[103] M. Capsoni et al. Selective Hybridization of a Terpyridine-Based Moleculewith a Noble Metal. Journal of Physical Chemistry C 121.42 (2017), pp. 23574–23581. doi: 10.1021/acs.jpcc.7b08576.[104] W. Wang et al. Manipulating localized molecular orbitals by single-atomcontacts. Physical Review Letters 105.12 (2010), pp. 2–5. doi: 10.1103/PhysRevLett.105.126801.85BIBLIOGRAPHY[105] R. Pawlak et al. Design and Characterization of an Electrically PoweredSingle Molecule on Gold. ACS Nano 11.10 (2017), pp. 9930–9940. doi: 10.1021/acsnano.7b03955.[106] H. E. Hoster et al. Structure formation in bis(terpyridine) derivative adlayers:Molecule - Substrate versus molecule - Molecule interactions. Langmuir 23.23(2007), pp. 11570–11579. doi: 10.1021/la701382n.[107] S. Zint et al. Imaging Successive Intermediate States of the On-SurfaceUllmann Reaction on Cu(111): Role of the Metal Coordination. ACS Nano11.4 (2017), pp. 4183–4190. doi: 10.1021/acsnano.7b01109.[108] J. Jia, H. S. Wu, Z. Chen, and Y. Mo. Elucidation of the forces governingthe stereochemistry of biphenyl. European Journal of Organic Chemistry 3(2013), pp. 611–616. doi: 10.1002/ejoc.201201273.[109] E. Bosch, N. P. Bowling, and J. Darko. The power of nonconventional phenylC-H...N hydrogen bonds: Supportive crystal-packing force and dominantsupramolecular engineering force. Crystal Growth and Design 15.4 (2015),pp. 1634–1641. doi: 10.1021/cg5014076.[110] G. Pawin et al. A surface coordination network based on substrate-derivedmetal adatoms with local charge excess. Angewandte Chemie - InternationalEdition 47.44 (2008), pp. 8442–8445. doi: 10.1002/anie.200802543.[111] E. C. Constable. 2,2’:6’,2"-terpyridines: From chemical obscurity to commonsupramolecular motifs. Chemical Society Reviews 36.2 (2007), pp. 246–253.doi: 10.1039/b601166g.[112] I. G. Powers and C. Uyeda. Metal-Metal Bonds in Catalysis. ACS Catalysis7.2 (2017), pp. 936–958. doi: 10.1021/acscatal.6b02692.[113] M. ( o. C. S. U. Kamna, S. ( Stranick, M. D. I. of Standards, Technology),and P. ( o. C. S. U. Weiss. Imaging Substrate-Mediated Interactions. Science274.5284 (1996), pp. 118–119. doi: 10.1126/science.274.5284.118.[114] B. W. J. Chen, D. Kirvassilis, Y. Bai, and M. Mavrikakis. Atomic andMolecular Adsorption on Ag(111). The Journal of Physical Chemistry C (Jan.2018). doi: 10.1021/acs.jpcc.7b11629.[115] B. Pascucci, G. S. Otero, P. G. Belelli, F. Illas, and M. M. Branda. Compara-tive density functional theory based study of the reactivity of Cu, Ag, and Aunanoparticles and of (111) surfaces toward CO oxidation and NO2reduction.Journal of Molecular Modeling 20.9 (2014), pp. 1–11. doi: 10.1007/s00894-014-2448-5.[116] M. Gajdoš, A. Eichler, and J. Hafner. CO adsorption on close-packed transi-tion and noble metal surfaces: Trends from ab initio calculations. Journal ofPhysics Condensed Matter 16.8 (2004), pp. 1141–1164. doi: 10.1088/0953-8984/16/8/001.86BIBLIOGRAPHY[117] F. Abild-Pedersen and M. P. Andersson. CO adsorption energies on metalswith correction for high coordination adsorption sites - A density functionalstudy. Surface Science 601.7 (2007), pp. 1747–1753. doi: 10.1016/j.susc.2007.01.052.[118] K. Jacobi, M. Bertolo, and W. Hansen. HREELS investigation of physisorbedmolecules on single-crystalline surfaces through dipole and resonant electron-scattering mechanism. Journal of Electron Spectroscopy and Related Phenom-ena 54-55(C) (1990), pp. 529–540.[119] G. McElhiney, H. Papp, and J. Pritchard. The adsorption of Xe and COon Ag(111). Surface Science 54.3 (1976), pp. 617–634. doi: 10.1016/0039-6028(76)90209-0.[120] W. M. Huo. Electronic Structure of CO and BF. The Journal of ChemicalPhysics 43.2 (1965), pp. 624–647. doi: 10.1063/1.1696786.[121] S. S. Sung and R. Hoffmann. How Carbon Monoxide Bonds to Metal Surfaces.Journal of the American Chemical Society 107.3 (1985), pp. 578–584. doi:10.1021/ja00289a009.[122] G. Blyholder. Molecular Orbital View of Chemisorbed Carbon Monoxide.The Journal of Physical Chemistry 68.10 (1964), pp. 2772–2777. doi: 10.1021/j100792a006.[123] A. Nilsson and L. G. M. Pettersson. Chemical bonding on surfaces probedby X-ray emission spectroscopy and density functional theory. 2004. doi:10.1016/j.surfrep.2004.06.002.[124] M. L. Bocquet, A. M. Rappe, and H. L. Dai. A density functional theorystudy of adsorbate-induced work function change and binding energy: Olefinson Ag(111). Molecular Physics 103.6-8 (2005), pp. 883–890. doi: 10.1080/00268970412331333609.[125] L. V. Moskaleva et al. Ethylene conversion to ethylidyne over Pd(111): Re-visiting the mechanism with first-principles calculations. Journal of PhysicalChemistry C 113.6 (2009), pp. 2512–2520. doi: 10.1021/jp8082562.[126] S. Booyens, L. Gilbert, D. Willock, and M. Bowker. The adsorption of etheneon Fe(1 1 1) and surface carbide formation. Catalysis Today 244 (2015),pp. 122–129. doi: 10.1016/j.cattod.2014.06.025.[127] P. A. Sheth, M. Neurock, and C. M. Smith. First-principles analysis of theeffects of alloying Pd with Ag for the catalytic hydrogenation of acetylene-ethylene mixtures. Journal of Physical Chemistry B 109.25 (2005), pp. 12449–12466. doi: 10.1021/jp050194a.[128] M. Dewar. No Title. Bull. Soc. Chim. Fr. 18.79 (1951).[129] J. Chatt and L. Duncanson. Olefin Co-ordination Compounds. Part III. * Infra-red Spectra and Structure: Attempted Preparation of Acetylene Complexes.Journal of the Chemical Society 0 (1953), pp. 2939–2947. doi: 10.1039/JR9530002939.87BIBLIOGRAPHY[130] H. Öström et al. Ethylene on Cu(1 1 0) and Ni(1 1 0): Electronic structureand bonding derived from X-ray spectroscopy and theory. Surface Science559.2-3 (2004), pp. 85–99. doi: 10.1016/j.susc.2004.04.041.[131] C. G. Bernardo and J. A. Gomes. The adsorption of ethylene on the (110)surfaces of copper, silver and platinum: A DFT study. Journal of MolecularStructure: THEOCHEM 582.1-3 (2002), pp. 159–169. doi: 10.1016/S0166-1280(01)00771-0.[132] S. K. Henze, O. Bauer, T. L. Lee, M. Sokolowski, and F. S. Tautz. Verticalbonding distances of PTCDA on Au(1 1 1) and Ag(1 1 1): Relation to thebonding type. Surface Science 601.6 (2007), pp. 1566–1573. doi: 10.1016/j.susc.2007.01.020.[133] W. Liu, A. Tkatchenko, and M. Scheffler. Modeling adsorption and reactionsof organic molecules at metal surfaces. Accounts of Chemical Research 47.11(2014), pp. 3369–3377. doi: 10.1021/ar500118y.[134] C. Sommer, C. Strähle, U. Köthe, and F. A. Hamprecht. ilastik: InteractiveLearning and Segmentation Toolkit. Eighth IEEE International Symposiumon Biomedical Imaging (ISBI 2011).Proceedings. 2011, pp. 230–233. doi:10.1109/ISBI.2011.5872394.88Appendix ACatalog of SpeciesThe different nodes repeatedly observed throughout experiments performed for thisthesis are reported in this Appendix. Table A.1 lists the nodes with proposedbond constituents. Table A.2 lists the observed nodes with visual descriptions foridentification in STM topographs. Fig A.1 displays STM topographs of the observednodes acquired with Ag-terminated tips at Vb = 20mV and It = 20 − 50pA. Themotifs in Fig A.1 were all tracked and counted separately for statistical analysis ofreacted species.Label Common Name Proposed Bond1.1 TPT n/a1.2 Singly-coordinated tpy-Fe1.3 Doubly-coordinated Fe-TPT-Fe1.4 Chain center TPT-Fen-TPT2.1 CO prebond tpy-Fe...CO2.2 C2H4 prebond tpy-Fe...C2H43.1 Fe-CO bond tpy-Fe-CO3.2 Fe-C2H4 bond tpy-Fe-C2H44.1 Higher order 1 tpy-Fe-CO...C2H4*4.2 Higher order 2 unknown4.3 Higher order 3 unknown4.4 Higher order 4 unknownTable A.1: Proposed bond constituents for all repeatedly observed nodes. the ‘-’indicates a bonding interaction, and the ‘...’ indicates a weaker attractive interactionwith no significant reordering of electronic structure. The ‘n’ in 1.4 represents arange of Fe adatoms present at chain centers with the likely value n = 3 basedon previous studies. [76] It is also possible that other molecules are bridging thechain centers with Fe. The proposed bond for 4.1, ‘*’, is only based on STMtopographic appearance and should be investigated more thoroughly to decipher thebond constituents with high confidence.89Label Common Name Visual Description1.1 TPT uniform intensity throughout molecule1.2 Singly-coordinated coordinated end appears brighter andwith a flattened termination1.3 Doubly-coordinated both ends are bright and flattened1.4 Chain center node connecting multiple TPTs in chainsthese have many different appearancesSee Fig 3.6 for more examples2.1 CO prebond dark, localized termination similar inshape to a free CO molecule on thesurface2.2 C2H4 prebond bright rounded termination with uniformintensity3.1 Fe-CO bond half ring-shaped termination with darkcenter and dark region extending onthe surface ∼ 1nm away from termination3.2 Fe-C2H4 bond rounded termination with localizedbright center4.1 Higher order 1 same as Fe-CO bond, but with anadditional ethylene attached4.2 Higher order 2 several small bright, round featurestrailing the termination, similar inshape to bare ethylene on the surface4.3 Higher order 3 rounded node with high intensity center,often exhibits noisy region near centerwhile scanning4.4 Higher order 4 two high intensity circular regionsTable A.2: Descriptions for visually identifying different nodes in STM topographs.90Figure A.1: Nodes observed throughout experiments performed for this thesis. Node4.2 and 4.4 are shown with a single CO molecule weakly bonded to the pyridine ofan uncoordinated tpy group. The CO molecules are not defining elements of thenode structures shown in 4.2 and 4.4.91Appendix BData AnalysisLine spectroscopy plots in this thesis were presented as dI/dV when analyzingelectronic states close to the Fermi energy (|Vb| ≤ 400mV ), or normalized dI/dVwhen showing electronic states away from Fermi (|Vb| > 400mV ). Two-dimensionalbias slices from grid measurements were always normalized to minimize spatialartifacts introduced by the different height offsets at each pixel from the setpointparameters It and Vb.The first step in STS grid data processing was to average the forward andbackward I vs. V curves measured at each spatial pixel in the grid. Then a smallconstant correction was added to the I(V) curve at every spatial pixel such thatwhen Vb = 0V , It = 0V . This correction was necessary likely due to current leakagein the pre-amplifier.[89]After that, the numerical derivative dI/dV was calculated. Then a circular orsquare area of the grid was selected to average dI/dV and I(V) over; the area wasselected to encompass the entire resonance curve at a particular energy. This spatialaveraging reduced the noise of the signal and typically involved averaging over∼ 100− 800 spatial pixels, depending on the distribution of the electronic state andthe spatial resolution of the grid. Typical spatial resolution for grids was 0.3Å perpixel for raw data. When bias slices were analyzed and displayed in this thesis, themaps were smoothed using the ‘smoothdata’ Matlab function specifying ‘gaussian’over 5 pixels. After smoothing the spatial resolution was reduced to ∼0.8Åper pixel.After plotting and examining the dI/dV, it was evident that a high-frequency,periodic spiky-noise was present in our data and this needed to be removed (seeFig B.1 (c) unfiltered inset compared with (d) filtered inset). dI/dV was averagedover ∼ 600 spatial pixels in a circular area over the phenyl of the bare TPT moleculefor Fig B.1. The Fourier transform of this data is plotted in (a) and shows ahigh-frequency peak at the normalized frequency 0.5. Since molecular states arelow-frequency in nature, this high-frequency peak is likely artificial and generatedin the instrumentation.1 The noise peak changes frequency based on the set point1The exact nature and origin of this noise is currently a mystery to be investigated by the92Figure B.1: Fourier transforms of averaged dI/dV at the phenyl LUMO of a non-metalated TPT before filtering (a), and after lowpass filtering (b). Average dI/dVof the same data without filtering (c) and with lowpass filtering (d). The dashedvertical line in (c) and (d) is located at Vb = 1.05V and shows the phenyl LUMOremains unshifted by the filtering process. No smoothing has been applied to anydata in this figure.parameters and sampling rates chosen for each particular experiment, thus forevery data set the peak must be identified separately. To filter this high-frequencynoise, the average dI/dV was treated with the ’lowpass’ Matlab filter using thenormalized passband frequency of 0.45 for the data shown in Fig B.1 (a), andspecifying ‘steepness’ as 0.95 for all filtered data sets. Since the low-frequencyregion of interest is preserved, molecular states are unaltered by the lowpass filteringprocess; comparing Fig B.1 (c) and (d) at the phenyl state explicitly illustrates this.The dashed lines in (c) and (d) are located at Vb = 1.05V , the LUMO resonance,and show the data is not energetically shifted between filtered and raw plots, itis simply less noisy. The Fourier transform of the I(V) data does not contain anmanufacturers of the microscope in the near future.93obvious high-frequency noise peak, however, for consistency in data processing, theI(V) data was lowpass filtered same as the dI/dV data.When states close to Fermi were being analyzed, the averaged, filtered dI/dV wasgaussian smoothed using the built in Matlab function ’smoothdata’ and specifying’gaussian’ with a convolution window of 5 bias pixels (px). Smoothed and rawdata were always compared to ensure no visible alteration in energy to the statesoccurred upon smoothing. Bias resolution varied throughout this thesis because avariety of bias windows were measured in collecting data based on the particularexperiment. For example, raw data ranges from ±400mV to ±2V split over 512pixels gives raw bias resolutions of ∼1.5mV/px (just above the thermal resolutionlimit of ∼1.3mV2), and ∼8mV/px, respectively. After smoothing over 5 pixels, thebias resolution decreases by a factor of 2.5, which resulted in bias resolutions of∼ 20mV/px for the ±400mV range and ∼8mV/px for the ±2V range.When states away from Fermi were analyzed, the averaged, filtered dI/dV andI(V) were both smoothed over the same number of pixels before normalization, andthe normalized dI/dV was smoothed after numeric division as well. This correspondsto Fig B.2 (d). Smoothing the I(V) and dI/dV over different amounts of pixels beforenormalization introduced artifacts into the data. Fig B.2 shows how smoothingreduced the noise in the signal, without altering the phenyl LUMO at Vb = 1.05V .Smoothing only before normalization results in an artifact at Vb = 1.05V in the datashown in (b). Again, smoothed and raw normalized data were always compared toensure no significant changes to the states occurred upon analysis.2This value was obtained from a calculation performed by Shun Chi, a former member of theLAIR team.94Figure B.2: Comparison of the effects of gaussian smoothing over 15 bias pixels atdifferent stages in data processing. Data is averaged over the phenyl state of a bareTPT molecule grid and lowpass filtered at a passband of 0.45 normalized frequency(a). Then the data is gaussian smoothed either before normalization only (b), afternormalization only (c), or both before and after normalization (d). Dashed line in(a), (b), (c), and (d) marks the phenyl state at 1.05V and shows the state does notchange in energy with smoothing when compared to the raw, filtered data (a).95Appendix CMaking CO-Terminated TipsThis Appendix provides instructions for terminating the tip with a CO moleculefor ncAFM measurements. Before beginning, the qPlus sensor must be excited atresonance frequency (usually fr ∼ 25KHz) with the phase locked loop (PLL) on, andthe tip approached to the sample. ncAFM measurements should not be attemptedwith a Q < 15K otherwise poor quality images will result. Higher values such asQ = 50K are desirable. Additionally, the sample must be dosed with CO. Typicalscanning parameters for samples containing CO are Vb = 20mV and It = 10− 20pA.The bias voltage must be small and close to |Vb = 20mV | otherwise the CO will notappear in topographs.How to make a CO-terminated tip:1. Locate a CO molecule on the surface2. Zoom in on the CO molecule to ∼ (20nm)2 scanning area or less and wait fordrift to settle3. Park the tip centered above the CO molecule4. Turn on z spectroscopy option5. Run z spectroscopy moving the tip slightly away from and then towards thesample. Example parameters are from zmax = +5.0Å to zmin = −1.0Å whenVb = 20mV and It = 10pA,1 which generated the example frequency shiftcurve shown in Fig C.1 (a)6. Change the z range so the tip moves slightly closer to the substrate as shownin Fig C.1 (b) with zmin = −1.5Å. The curve shape will begin to resemble aLennard-Jones potential1The z spectroscopy ranges suggested in Step 5 are useful as general starting points, but mayrequire tuning.967. Repeat Step 6 changing the zmin value until forward and backward curves inDf(z) spectroscopy no longer align. This indicates a potential CO pickup andis shown in Fig C.228. Scan over a different CO molecule on the surface checking for a well-definedcentral bump at the CO in the STM topograph as shown in Fig C.39. If the central bump is present and symmetric, the CO-termination on the tipis successful10. Otherwise, clean the tip and begin again from Step 1Figure C.1: Frequency shift with respect to changing z tip sample separation. z = 0defines the tip sample separation at the setpoint scanning parameters Vb = 20mVand It = 10pA. Forward spectroscopy measurements are represented with blacklines and the backward measurements are shown with red lines in (a) and (b). In(a) the tip starts away from the setpoint at zmax = 5.0Å and moves closer to thesample by z = 6.0Å to zmin = −1.0Å, and back again. In (b) the tip sweeps fromzmax = 5.0Å to zmin = −1.5Å and back. The curve in (b) begins to resemble theminimum in the Lennard Jones potential shown in Fig 2.8.2The Df(z) spectrum shown in Fig C.2 was acquired on a different sample with a different tipand different setpoint parameters from the spectra shown in Fig C.1.97Figure C.2: Frequency shift as a function of z spectroscopy with z = 0 defining thetip sample separation at the setpoint parameters Vb = 20mV and It = 15pA. Adifference in forward and backward scans of a Df(z) measurement taken over a COmolecule is an indication that the CO molecule has been picked up by the tip.Figure C.3: Topograph (a) and corresponding change in tunneling current signal(b) of two CO molecules imaged with a CO-terminated tip. A well-centered CO-terminated tip will image CO molecules on the surface with a small symmetric bumpat their center, as shown in both (a) and (b). Vb = 20mV , It = 10pA. If the bumpis not centered, ncAFM imaging should not be attempted with that tip.98Appendix DStatistical AnalysisD.1 Hand-Counted StatisticsFor the statistical analysis in this thesis, images were scrutinized and repeatedlyobserved motifs were hand-counted at different stages in different experiments. Thecounted nodes are shown in the catalog of species in Appendix A, and in addition tothose motifs, free and H bonded ethylene and CO molecules were tracked as well. 14different categories of visually identifiable interactions were defined for all countedimages as well as one category counted for uncertainty. The 14 visually identifiedcategories were:1. Active sites (Fe-tpy coordination bonds)2. Chain centers3. Free CO on the surface4. CO weakly bonded to a non-metalated tpy group5. Free C2H4 on the surface6. C2H4 hydrogen bonded to a non-metalated tpy group7. CO prebond8. CO bond9. C2H4 prebond10. C2H4 bond11. Higher order 1 node12. Higher order 2 node13. Higher order 3 node99D.1. HAND-COUNTED STATISTICS14. Higher order 4 nodeThe experimental stages separated in Fig 4.10 and Fig 5.7 were defined by anychange applied to the substrate such as annealing at temperatures from T = 15−40K,or dosing different gases on the substrate where the substrate reached a maximumtemperature up to Tmax = 11.6K. Any of these changes to the sample could influencereactivity.The uncertainty in hand-counting was tracked by a separate category for eachimage. Any node that did not fit into any of the other 14 tracked categories wasput into the ‘uncertain’ category. This gives a one-sided error distribution as isshown for the statistical data presented in Fig 4.10 for CO and Fig 5.7 for C2H4.The uncertain nodes were either nodes that were partially cut off by the edge of theimage, and thus undefinable, or nodes that simply were not observed repeatedly,which happened quite rarely. If an image contained uncertain nodes due to poor tipquality, those images were discarded from the analysis. The chain center categorydid consist of a variety of visually different nodes due to the non uniformity of chaincenters as discussed in Chapter 3.Figure D.1: Uncertainty decreases by increasing the total area of individual images(a), and by averaging count over several images (b).Fig D.1 shows the general trend that uncertainty in counting decreased as morearea of the sample was counted. Fig D.1 (a) plots the uncertainty versus the area ofan individual image and (b) plots the uncertainty versus the total number of imagesaveraged in an experimental stage. These plots show that by increasing the area ofindividual images counted and by increasing the number of images averaged overfor statistics at each experimental stage, the uncertainty decreases. A larger imagesize does imply more identifiable nodes versus nodes cut off by the edges, however,images that are too large with a poor quality tip or not enough pixels containmany unidentifiable nodes. The three points in (b) with the largest uncertaintiescorrespond to images with small total area. There is an interplay between individualimage size and total number of images counted at each experimental stage thatinfluences the uncertainty in counting.100D.1. HAND-COUNTED STATISTICSUncertainty depends on the size and quality of images, the total number ofimages counted at each experimental stage, and the surface coverage of active sitesand gaseous adsorbates in each image. To ensure that enough of the sample wasbeing counted for statistical significance, a check was developed (see Fig D.2). Thenumber of active sites and chain centers should be relatively constant throughoutexperimental stages at low temperatures for the same sample, and if these quantitiesremained relatively constant throughout experimental stages, it was determined thatenough area of the sample had been counted to gain a statistical understanding ofsurface activity.1Figure D.2: Surface coverage of active sites and chain centers throughout experi-mental stages in four different experiments (a-d). Active sites and chain centers arequantities that should be conserved throughout experimental stages within eachexperiment. In (d) the active sites and chain centers were not conserved, thus thestatistical data from this experiment was not presented in this thesis.Fig D.2 shows these conserved quantities, active sites and chain centers, over theexperimental stages in four different experiments. (a), (b), and (c) from Fig D.2correspond to the same experiments in (a), (b), and (c) from Fig 4.4, respectively.The horizontal axes for Fig D.2 (a-d) correspond to experimental stages (genericallynumbered) of the reactions and the vertical axes correspond to the counted surface1Diffusion of coordinated TPTs was observed after annealing the sample at T = 35K, which couldpotentially change ‘active sites’ to ‘chain centers’. The observed diffusion lengths of coordinatedTPTs were not large (∼ 5nm) at T = 35K. Since larger diffusion lengths of small adsorbates wereobserved at lower temperature (see Fig 5.2), the influence of TPT diffusion on counted categorieswas assumed negligible.101D.2. MACHINE LEARNING STATISTICScoverage of active sites and chain centers. (d) shows an experiment where the activesites and chain centers were not conserved and thus the data from this experimentwas not presented in this thesis. Experiment (d) had very low surface coverage ofactive sites and chain centers compared to the other experiments, and thus a muchlarger area of the sample probably needed to be counted for proper sampling of thesurface activity. Experimental stage 6 in (c) corresponds to the Matrix crash asdiscussed in the statistics sections of Chapter 4 and Chapter 5 where a differentregion of the sample was probed after the crash, millimeters away from the previousarea.D.2 Machine Learning StatisticsMachine learning applied to SPM surface reactions has the potential to drasticallyreduce image analysis time by automatically counting different reacted species inimages fed through properly trained machine learning algorithms. Machine-countingfor statistical analysis was attempted for this work using the pixel classificationpackage of ilastik. [134] STM topographs were fed into the pixel classification tool,and different features were marked to train the software to distinguish those features.The images were then segmented by the classifier into constituent parts. Fig D.3shows an STM topograph (a) and the segmented image of the topograph (b) outputby ilastik after training the pixel classification tool.Figure D.3: STM topograph (a) and segmentation (b) of the topograph generatedby training the pixel classification tool in ilastik. Phenyls, bare terpyridine groups,Fe-terpyridine active sites, chain centers, and the Ag(111) surface are separatelyidentified in the segmented image.The segmented image from Fig D.3 (b) consists of four separately classified102D.2. MACHINE LEARNING STATISTICSparts, which are shown in Fig D.4. The four parts shown separately in Fig D.4 arebare terpyridine groups (a), phenyls (b), Fe-tpy active sites (c), and chain centers(d). These individual images were fed through a counting algorithm in Matlab andthe number of items in each image was recorded. This process worked well forindividually trained images, but when trying to use the batch processing feature ofilastik to segment untrained images, many identification errors by the classifier tooloccurred.Figure D.4: Four separate parts (a-d) of the segmentation shown in Fig D.3 (b).Individual terpyridine groups are shown in (a), phenyls in (b), active Fe-tpy sites in(c), and chain centers in (d). Each individual image (a-d) of the segmentation wasfed through an algorithm that counts and records the number of items containedwithin.In principle, machine learning software is very attractive for counting reactedspecies in surface chemistry SPM experiments, but in practice there are difficultiesassociated with training the pixel classifier. Hundreds of images are usually required103D.3. APPLYING MACHINE LEARNING TO SPMfor proper training, which were not available at each stage in the reaction forexperiments performed for this thesis. Additionally, the classifier had difficulty withdrawing similarities between SPM images of different size and slightly different quality.When more adsorbates and higher order nodes were present on the samples, moreclassifications were needed by the pixel classification tool, which drastically increasedthe required computation time for training. In order to apply machine learning toSPM surface chemistry statistical analysis successfully, consistency in images andexperiments is necessary. Additionally, other methods to reduce computation andtraining time are needed. The following section provides suggestions for improvingmachine learning approaches to statistical SPM surface chemistry studies.D.3 Applying Machine Learning to SPMHand-counting statistics for SPM surface reactions is tedious, but useful in deter-mining reaction pathways, and machine learning could provide a tool to drasticallyspeed up the statistical analysis of SPM surface chemistry data with proper training.Following are suggestions and things to keep in mind for data collection practices thatcould help generate data better suited for analysis by machine learning techniques.• Consistency in sample cleaning procedures is key. Changing the amount ofterraces versus step edges and defects on a surface introduces new potential re-action pathways, and minimizing these differences between sample preparationswill assist in identifying reaction pathways.• Consistency in active site preparation is also important. The number ofchain centers versus active sites changes depending on surface coverage ofTPT molecules, surface coverage of Fe adatoms, how the Fe was depositedon the sample (short bursts with shutter open versus longer shutter opentime), the temperature of the sample upon Fe deposition, the temperature ofthe sample post Fe deposition, how long the sample was left at a particulartemperature after depositing Fe, and whether the sample was left in the prepor LT chamber after Fe deposition. Different surface concentrations of activesites in different experiments makes it difficult to compare statistical resultsbetween experiments in order to draw conclusions.• Reduction in the number of chain centers upon active site preparation couldreduce uncertainty in experimental outcomes since the activity at non uniformchain centers was intractable for the system studied in this thesis.• Consistency in dosage of C2H4 and CO could also make comparison betweenexperiments and stages within experiments more straightforward. Dosing athigher pressures for shorter times reduces the maximum substrate temperaturereached during gas dosages, which reduces adsorbate diffusion upon depositionand could increase consistency between depositions. For example, dosing gases104D.3. APPLYING MACHINE LEARNING TO SPMat PLT ∼ 2 · 10−8mbar for t ∼ 3sec with the heat shields open can keep themaximum substrate temperature below Tmax < 7K.• Collecting images at each experimental stage of uniform size and quality (samenumber of points and lines, and similar tip sharpness) could reduce trainingtime for the pixel classification tool in ilastik.• Collecting images using the same scanning parameters at each experimentalstage is necessary since topograph appearance changes as a function of biasvoltage applied.• Collecting large quantities of images at each experimental stage is necessaryto properly train the classifier.• Training the classifier to properly handle uncertainties could reduce requiredcomputation time. In organic surface chemistry, there is a degree of nonuniformity in node appearance and allowing the classifier to group nodes witha certain threshold of uncertainty into an ‘uncertain’ category could reduceclassification computation time significantly.105

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0378374/manifest

Comment

Related Items