UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

On-surface self-assembly and characterization of a macromolecular charge transfer complex by scanning… Capsoni, Martina Carla 2016

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

Item Metadata

Download

Media
24-ubc_2016_september_capsoni_martina.pdf [ 62.79MB ]
Metadata
JSON: 24-1.0305019.json
JSON-LD: 24-1.0305019-ld.json
RDF/XML (Pretty): 24-1.0305019-rdf.xml
RDF/JSON: 24-1.0305019-rdf.json
Turtle: 24-1.0305019-turtle.txt
N-Triples: 24-1.0305019-rdf-ntriples.txt
Original Record: 24-1.0305019-source.json
Full Text
24-1.0305019-fulltext.txt
Citation
24-1.0305019.ris

Full Text

On-surface self-assembly and characterizationof a macromolecular charge transfer complexby scanning tunneling microscopy and spectroscopybyMartina Carla CapsoniM.Sc. in Physics Engineering, Politecnico of Milano, 2011B.Sc. in Physics Engineering, Politecnico of Torino, 2008A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Physics)The University of British Columbia(Vancouver)June 2016c©Martina Carla Capsoni, 2016AbstractOrganic-based technologies have recently attracted significant interest. Characterization of theirstructure and properties at native length scales are essential for their implementation in devices. On-surface self-assembly of metal-organic frameworks is a simple way to fabricate molecular systemswith specific functionalities.In this thesis work, the morphology and electronic structure of self-assembled linearnanochains, featuring a triiron linkage between two bisterpyridine-based ligands on an Ag(111) sur-face, have been investigated with scanning tunneling microscopy and spectroscopy. An in situ, cleanand reliable on-surface preparation technique was developed for thermally-activated self-assemblyof complexes based on the metal-organic motif of dyes used in photovoltaic and catalysis applica-tions. Tunneling spectroscopy on the metal-organic nanostructures obtained suggests the formationof a coordination bond with charge transfer between metal and ligand. Furthermore, the electronicstructure indicates the presence of the desired metal-to-ligand charge transfer optical transitions,characteristic of the related complexes. The unprecedented triiron coordination link has potentialfor being an efficient reaction center for catalysis applications, as well as for having interesting mag-neto, spin, and electronic properties. Each step and aspect of the chains formation process has beencharacterized via scanning tunneling microscopy measurements and growth studies, and the resultsare supported by density functional theory calculations. Additionally, the relevance and influenceof the silver metal substrate on both bare ligands and chains has been investigated. Bare moleculesshow a strong interaction with the substrate, as demonstrated by their specific adsorption configura-tions and an electronic structure which is distinct from when they are electronically decoupled fromthe surface by an NaCl bilayer. When the molecules are in chains the silver plays a key role in thestructure of the coordination link. This work shows the potential of using on-surface self-assemblyand scanning tunneling microscopy and spectroscopy, not only to prepare with high-fidelity cleanand controlled structures but also as a flexible platform to investigate and tailor functional propertiesof different systems for a large variety of applications where a solid support is essential.iiPrefaceOne of the three osmium-terpyridine complexes was synthesized by a group in Taiwan at the Depart-ment of Chemistry and Low-Carbon Energy Research Center in the National Tsing Hua Universityin Hsinchu, while the other two in Mike Wolf’s group at the Chemistry Department of UBC. Thethermogravimetric analysis of these compounds was performed by Mike Wolf’s group.Scanning tunneling microscopy and spectroscopy measurements of the TPPT/Ag(111)and Fe/TPPT/Ag(111) systems were done by Dr. Agustin Schiffrin and me. The ones on theTPPT/NaCl/Ag(111) system were done by me with the help of Tanya Roussy and Katherine Cochrane.I performed all the data analysis.Density functional theory calculations on the Fe-TPPT system on Ag(111) have beenperformed by Chenguang Wang at the Renmin University of China (Beijing), Department of Physicsand Beijing Key Laboratory of Optoelectronic Functional Materials & Micro-nano Devices. Theones on the TPPT molecules in gas-phase were done by Katherine Cochrane.Section 4.1 - 4.5 of chapter 4 and the appendix A on the density functional theory methodsare based on the paper (and supplementary information) “On-surface synthesis of iron terpyridinenanochains featuring a linear tri-iron linkage” by M. Capsoni et al. (in submission).iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Supplementary Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiiGlossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiiiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxvDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxvi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Polypyridine metal-organic complexes . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Depositing osmium-terpyridine on an Ag(111) surface . . . . . . . . . . . . . . . 51.4 On-surface supramolecular chemistry . . . . . . . . . . . . . . . . . . . . . . . . 71.4.1 Supramolecular chemistry of molecular self-assembly . . . . . . . . . . . 71.4.2 On-surface supramolecular coordination chemistry . . . . . . . . . . . . . 81.5 Iron-terpyridine self-assembled nanochains . . . . . . . . . . . . . . . . . . . . . 102 Experimental Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.1 Introduction to scanning tunneling microscopy . . . . . . . . . . . . . . . . . . . 122.2 Scanning tunneling microscopy and spectroscopy theory . . . . . . . . . . . . . . 152.3 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.4 Samples preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282.4.1 Metallic substrate: Ag(111) surface . . . . . . . . . . . . . . . . . . . . . 28iv2.4.2 NaCl/Ag(111) substrate . . . . . . . . . . . . . . . . . . . . . . . . . . . 292.4.3 Molecular deposition: terpyridine-phenyl-phenyl-terpyridine . . . . . . . . 302.4.4 Metal deposition: iron . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303 Bare Ligand: Morphology and Electronic Structure . . . . . . . . . . . . . . . . . . 323.1 Morphology of TPPT ligands on Ag(111) . . . . . . . . . . . . . . . . . . . . . . 323.2 Ligands morphology on NaCl/Ag(111) . . . . . . . . . . . . . . . . . . . . . . . . 373.3 Electronic structure of bare ligands on silver . . . . . . . . . . . . . . . . . . . . . 403.3.1 Single ligand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403.3.2 Two adjacently bound molecules . . . . . . . . . . . . . . . . . . . . . . . 453.3.3 Three adjacently bound molecules . . . . . . . . . . . . . . . . . . . . . . 473.4 Electronic structure of bare ligands on NaCl/Ag(111) . . . . . . . . . . . . . . . . 493.5 Comparison between the electronic structure of ligands on bare Ag(111) and onNaCl/Ag(111) substrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564 Discussion of Chain Formation Process . . . . . . . . . . . . . . . . . . . . . . . . . 584.1 Iron-TPPT chain formation: general . . . . . . . . . . . . . . . . . . . . . . . . . 584.2 Morphology of single Fe-coordinated molecules . . . . . . . . . . . . . . . . . . . 594.3 Morphology of Fe-TPPT chains . . . . . . . . . . . . . . . . . . . . . . . . . . . 634.4 Chains statistics and annealing effects . . . . . . . . . . . . . . . . . . . . . . . . 684.5 Proposed model for the chain formation process . . . . . . . . . . . . . . . . . . . 714.6 Different iron deposition methods . . . . . . . . . . . . . . . . . . . . . . . . . . 714.7 Preliminary attempt to form chains on NaCl/Ag(111) . . . . . . . . . . . . . . . . 734.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735 Electronic Structure of Fe3-TPPT Chains . . . . . . . . . . . . . . . . . . . . . . . . 755.1 “One-molecule chain”: single TPPT coordinated with Fe on both its ends . . . . . 765.2 Four- and five-molecule chains . . . . . . . . . . . . . . . . . . . . . . . . . . . . 805.3 Detail of a five-molecule chain: the three molecules in the middle of the chain . . . 845.4 Detail of three-molecule chain with not coordinated end (bare tpy) . . . . . . . . . 865.5 Detail of a three-molecule chain and one end-molecule coordinated with Fe . . . . 885.6 General comments and interpretations . . . . . . . . . . . . . . . . . . . . . . . . 905.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 916 Conclusions and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 936.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 936.2 Open questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 956.3 Future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98vAppendix A Density Functional Theory Methods for On-Surface Calculations . . . . . 112Appendix B Details on Data Analysis for Scanning Tunneling Spectroscopy Measure-ments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113Appendix C Additional Spectroscopy Measurements Results . . . . . . . . . . . . . . . 117C.1 Single bare ligand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124C.2 Coordinated structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126C.2.1 “One-molecule chain” . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126C.2.2 Long chain: six-molecule chain . . . . . . . . . . . . . . . . . . . . . . . 128C.2.3 Detail of a six-molecule chain: three molecules in the middle of the chain . 130C.2.4 Detail of a three-molecule chain and its end-molecule non-coordinated withFe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132C.2.5 Detail of a four-molecule chain and one end-molecule coordinated with Fe 134viList of TablesTable C.1 Grid measurements on bare TPPT molecules and coordinated structures on Ag(111)performed in 2013. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117Table C.2 Matching of the grid label with the grid filename of the measurements done onbare TPPT molecules and coordinated structures on Ag(111) in 2013. . . . . . . 118Table C.3 Grid measurements on bare TPPT molecules on Ag(111) and NaCl/Ag(111) per-formed in 2014-2015. “*” indicates the grid measurements reported in the maintext, while “@” the ones reported in this appendix. . . . . . . . . . . . . . . . . 119Table C.4 Matching of the grid label with the grid filename of the measurements done onbare TPPT molecules on Ag(111) and NaCl/Ag(111) in 2014/2015. “*” indicatesthe grid measurements reported in the main text, while “@” the ones reported inthis appendix. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120Table C.5 Grid measurements on whole coordinated structure on Ag(111) performed in2014. “*” indicates the grid measurements reported in the main text, while “@”the ones reported in this appendix. . . . . . . . . . . . . . . . . . . . . . . . . 121Table C.6 Matching of the grid label with the grid filename of the measurements done onwhole coordinated structure on Ag(111) in 2014. “*” indicates the grid measure-ments reported in the main text, while “@” the ones reported in this appendix. . 121Table C.7 Grid measurements on different types of coordinated structure on Ag(111) per-formed in 2014. “*” indicates the grid measurements reported in the main text,while “@” the ones reported in this appendix. . . . . . . . . . . . . . . . . . . . 122Table C.8 Matching of the grid label with the grid filename of the measurements done dif-ferent types of coordinated structure on Ag(111) in 2014. “*” indicates the gridmeasurements reported in the main text, while “@” the ones reported in this ap-pendix. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123viiList of FiguresFigure 1.1 Schematic of the most common coordination geometries: octahedral (a), square-planar (b), and tetrahedral (c). Figures reproduced from [32] . . . . . . . . . . 4Figure 1.2 General schematic of the four different types of transitions occurring in metal-ligand complexes. MC are the transitions occurring inside the metal d orbitals;LMCT are the ones between occupied ligand orbitals and unoccupied metalones; MLCT between occupied metal orbitals and unoccupied ligand ones; andL-L transitions are between ligand orbitals. Image adapted from [33]. . . . . . 4Figure 1.3 (a) Absorption spectra of the two different of ruthenium-terpyridine compoundsshown above the spectra [green (TUS-25) and black (black dye) curves] andthe radical R (blue curve) in solution. Figure reproduced from [34]. (b) Solarspectrum as viewed through the atmosphere. Figure reproduced from [35]. . . 5Figure 1.4 (a), (b) and (c) are the chemical structure of the three Os-terpyridine-basedligands. Complex (a) was synthesized at the Department of Chemistry andLow-Carbon Energy Research Center in the National Tsing Hua University inHsinchu (Taiwan) [39], while (b) and (c) have been synthesized in Mike Wolf’sgroup at the Chemistry Department at UBC. Both (b) and (c) complexes have aPF6 counterion. (d) ThermoGravimetric Analysis (TGA) of the (b) compound(measurement performed by Mike Wolf’s group). This shows the percentage ofweight loss of the complex as a function of the increasing temperature. Eachvisible drop should correspond to the loss from the compound of a piece of it.Here, the first drop (solid red line), very likely close to the decomposition tem-perature, is at around 350◦C. (e) Constant-current STM image of unidentifiedadsorbates on a Ag(111) substrate (Vb = +2V; It = 10pA). This image was takenafter trying to deposit the (b) compound at a temperature close to its decompo-sition temperature (about 350◦C). Because of their relatively small dimensionthose adsorbates could be the counterions PF6. . . . . . . . . . . . . . . . . . 6viiiFigure 1.5 (I) STM topographic images of tunable metal-organic honeycomb nanomesheswith designed dicarbonitrile linear linkers on Ag(111). On the lower side ofpanel a schematics of the structures are superimposed on the STM images.Panel b: large-scale STM image. Figure reproduced from [55]. (II) STMconstant-current mode images of L-methionine stripes with molecular resolu-tion. Panel a: Grating of double rows. Panel b: Individual molecules appearas elliptical features. Schematic of the molecules and the interaction bonding(white dashed lines) are superimposed on the STM image. Figure reproducedfrom [57]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9Figure 1.6 Figures reproduced from [41]. (a) Schematic of atoms or molecules diffusingon a surface after being deposited from the vapor phase. F is the depositionrate and D is the diffusion rate. The type of growth is determined by the ratiobetween D and F. If F is bigger than D, the growth is controlled by kineticsresulting in non-equilibrium configurations. Examples are the metallic islandsshown in the two images on top of the second panel from the left of figure1.5(b). If D/F is large, the growth occurs close to the equilibrium and well-defined nanostructures can form. Examples are shown in the first, the second(bottom image) and third panels of figure 1.6(b) [41]. . . . . . . . . . . . . . . 10Figure 1.7 Schematic of the thermally activated self-assembly of TPPT molecules and Featoms (red dots) on a Ag(111) surface. The final structure, characteristics of thecoordination linkage, and electronic properties of the resulting chains are thefocus of this thesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11Figure 2.1 Examples of scanning tunneling microscopy and spectroscopy capabilities. (a)Constant-current STM image of the atomically resolved reconstructed Si(111)- (7 x 7) surface obtained by Binnig in 1983. Figure reproduced from [64]. (b)Three-dimensional view of an STM topography of carbon monoxide (CO) andbenzene coadsorbed on Rh(111). The three-fold ring-like features are benzenemolecules. CO is not resolved. Figure reproduced from [65]. (c) Chemicalimaging using inelastic scanning tunneling spectroscopy. (i) Constant-currentSTM image of a C2H2 molecule (left) and a C2D2 molecule (right). (ii), (iii)and (iv) d2I/dV2 images of the same area as (i) at 358 mV (i), 266 mV (ii)and 311 mV. (ii) reveals only the C2H2 molecule because at that voltage one ofits vibrational modes is excited. (iii) shows only the C2D2 molecule for samereason as (ii). (iv) none of the vibrational mode of any of the two adsorbates areexcited therefore they are not revealed by the d2I/dV2. Figure reproduced from[68]. (d) Example of atomic manipulation: building a quantum coral using ironatoms for electrons confinement on Cu(111). Figure reproduced from [69]. . . 14ixFigure 2.2 Schematic of the tunneling effect. A particle of wavefunction Ψ and energyE, lower than the energy of the potential barrier U0, has a non-zero probabil-ity of tunneling through the barrier of a finite width L. Its wavefunction decaysexponentially within the barrier and on the other side of it the wavefunction am-plitude is reduced depending on the thickness of the barrier. Figure reproducedfrom [74]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15Figure 2.3 Left side: Simplified schematic of a scanning tunneling microscope. Tip andsample are separated by a distance dz of few A˚. In this case a bias voltage Vbiasis applied to the sample. A system of piezos allows scanning the tip over thesurface of the sample (x - y plane) and adjusting the dz separation distance.The feedback loop system regulates dz by comparing the measured amplifiedtunneling current It with the pre-set Iset current value. Right side: Schematicof a tunneling junction. The metallic tip’s Density Of States (DOS) is assumedto be a step-function. The applied bias voltage shifts the tip’s Fermi level (EF )up with respect to the organic samples one. The electrons contributing to thetunneling current are tunneling through the vacuum barrier from the tip to theLUMO of the organic sample. . . . . . . . . . . . . . . . . . . . . . . . . . . 16Figure 2.4 Schematic to visually describe how single point STS curves and spatially re-solved energy distribution maps are related to each other and to the electronicstructure of the sample. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19Figure 2.5 Picture of the whole STM system showing the main consisting parts. . . . . . . 22Figure 2.6 Picture of the preparation chamber of the Omicron system with all its mainfeatures shown. The leak valve and the homebuilt thermal effusion cell are notvisible from this angle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23Figure 2.7 Schematic of the acoustical and environmental isolations of the Omicron STMsystem. Figure reproduced from [86]. . . . . . . . . . . . . . . . . . . . . . . 23Figure 2.8 Detailed image of the STM head with the springs suspension system. Figurereproduced from [87]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24Figure 2.9 Piezos system. (a) Schematic of the single scanner tube. (b) Schematic of thepiezo scanner showing the electronic connections for the x-, y- and z-motions.Figures reproduced from [87]. . . . . . . . . . . . . . . . . . . . . . . . . . . 25Figure 2.10 Picture of an Ag(111) sample mounted on a molybdenum sample plate. . . . . 26Figure 2.11 (a) Picture of the STM head from outside the copper heat shield with the coldtemperature deposition port visible at the bottom left. (b) Picture of the innercopper heat shield screwed to the bottom of the LHe cryostat. . . . . . . . . . 27Figure 2.12 Characteristics of the Ag(111) surface. (a) Constant-current STM image. Ag(111)atomic resolution (Vb = -100mV, It = 100pA). Hexagonal lattice and lattice con-stant are shown. (b) STS spectrum showing the silver surface states at -67 mV(solid red line) [91]. Note: The hump around -0.6V is due to the tip. . . . . . . 28xFigure 2.13 Characteristics of NaCl and NaCl/Ag(111) surface. (a) NaCl 3-dimensionalcrystal structure. Figure reproduced from [98]. (b) and (d) Constant-currenttopographic images of NaCl grown on top of a clean Ag(111) surface. (b)shows the Moire´ pattern of salt on Ag(111) (Vb=2V, It=30pA). (d) NaCl islandsformed on top of the Ag(111) surface. Bi and trilayers are visible (Vb=300mV,It=40pA). (c) Plot showing the dI/dV of NaCl on Ag(111) characterized by theinterface state about +96mV (solid red line). . . . . . . . . . . . . . . . . . . . 29Figure 2.14 STM constant-current images. (a) TPPT molecules on Ag(111) deposited atroom temperature and imaged at ∼4.3K (5 x 5 nm2, Vbias = 200mV, It = 50pA).(b) TPPT molecules and iron adatoms (dots - cyan arrow) on Ag(111). TPPTdeposited at room temperature, Fe at ∼4.3K, and imaged at ∼4.3K (5 x 5 nm2,Vbias = 500mV, It = 10pA). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Figure 3.1 (a) Chemical structure of a terpyridine-phenyl-phenyl-terpyridine molecule ingas-phase. (b) STM constant-current mode topographic image of a bare moleculeon Ag(111) with superimposed chemical structure [Carbon (C) atoms are turquoise,nitrogen (N) blue, and hydrogen (H) white], (Vb = 10mV, It = 1nA). Red arrowpoints towards the small depression close by N atom of the peripheral pyridine.(c) DFT-simulated side-view images of a single TPPT molecule adsorbed on theAg substrate along the most favorable direction [1, -1, 0]. It shows a small twistangle between the two phenyl groups in the middle of the molecule (DFT cal-culation performed by Chenguang Wang). (d) Apparent height intensity profilealong the solid blue line shown in the inset at the top of the plot. Red arrows arepointing to the two depressions position close to the top outer pyr groups. . . . 33Figure 3.2 Constant-current STM images of different isolated molecules on Ag imaged atdifferent bias. (a) Vb= -2.5V, It=25pA. (b) Vb= -1V, It=50pA. (c) Vb= -700mV,It=1nA. (d) Vb= +9mV, It=1nA. (e) Vb= +500mV, It=1nA. (f) Vb= +1.3V, It=1nA.(g) Vb= +2V, It=1nA. (h) Vb= +2.7V, It=50pA. Images dimension: 3.8 x 3.6 nm2.Green arrows in panel (d) and (e) point to the small asymmetry visible along themolecular axis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34xiFigure 3.3 (a) Constant-current STM image of a low coverage RT deposition of TPPTson Ag(111) (Vb = 200mV, It = 50pA). Molecules adsorb following a ∼ ± 16◦(red and yellow dashed ellipses) angle with respect to the <1, -1, 0> crys-tallographic directions (white arrow), resulting in six equivalent orientations(labels 1, 2 and 3). (b) Constant-current STM image of an isolated moleculewith atomic resolution of the Ag(111) substrate (scanning parameters on themolecule: Vb = -200mV, It = 10pA; on the Ag(111): Vb = -10mV, It = 300nA).The molecule’s orientation with respect to the [1, -1, 0] direction is shown. (c)and (d) schematics of the clockwise (∼-16◦) and counter-clockwise (∼+16◦)orientations of a TPPT with respect to the [1, -1, 0] silver axis. . . . . . . . . . 35Figure 3.4 (a) Constant-current STM image of two bare molecules on Ag(111) in the adja-cent configuration (Vb = -10mV, It = 1nA). The red dashed circles represent theattractive proton acceptor/organic ring interaction between the two molecules[107, 108]. (b) Constant-current STM image of high coverage TPPTs RT de-position on Ag(111) (Vb = -500mV, It = 20pA). Green dashed circle: extramolecule in addition to the zigzag pattern. . . . . . . . . . . . . . . . . . . . 36Figure 3.5 Constant-current STM images. (a) After RT deposition of TPPT molecules theNaCl island appears free of adsorbates, while all the ligands are ordered in aclosed-packed manner on the Ag(111) patch (Vb = 300mV, It = 40pA). Note:this image was rotated of 90◦. The vertical lines in the left half of the image arecaused by a series of subsequent tip changes occurring during the scan. In theAg area each of the oblique bright segment is a single TPPT molecule. Uponcloser inspection they appear to be arranged in their characteristic adjacent waydriven by the attractive proton acceptor/organic ring interaction. (b) After 4Kdeposition of the TPPT the ligands are imaged to be equally distributed on bothsilver and salt (Vb=500mV, It=1pA). . . . . . . . . . . . . . . . . . . . . . . . 37Figure 3.6 Constant-current STM image of an isolated molecule on NaCl/Ag(111) with themiddle of the molecular axis twisted as a result of a point spectroscopy measuretaken in that location (Vb = -1V, It = 10pA). Image dimension: 3.6 x 2.5 nm2. . 38Figure 3.7 Constant-current STM images of different isolated molecules on NaCl/Ag(111)at different scanning parameters. Specifically, the bias voltage is indicated oneach panel and the tunneling current values are: (a) and (b) 2pA, (c) 5pA, (d)25pA, (e) and (f) 20pA, and (g) 30pA. Images dimension: 5 x 4.8 nm2. . . . . 39Figure 3.8 Large-area constant-current STM images. (a) TPPT deposited at 4K on Na-Cl/Ag(111) substrate adsorbed on both Ag(111) and NaCl (Vb=500mV, It=1pA).(b) TPPT molecules arrangement on NaCl after half an hour annealing at 40K(Vb=500mV, It=1pA). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39xiiFigure 3.9 Normalized dI/dV curves of an isolated bare molecule obtained by averagingover the regions enclosed by the corresponding colour outlines in the topogra-phy (see inset). Red: whole single molecule; blue: single molecule molecularaxis; dashed black: Ag(111), reference spectrum. Grid set-point parameters:Vbias = -2.50V, It = 50pA. Note: the divergence near zero-bias created by nor-malization is removed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41Figure 3.10 STS energy maps at different biases showing the density of states of a singlebare ligand. From (a) to (j) the biases are: -1.9V, -1V, +0.4V, +0.7V, +1.15V,+1.6V, +1.8V, +2.0V, 2.3V and +2.7V. Grid set-point parameters: Vbias = -2.50V,It = 50pA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Figure 3.11 STS energy maps at Vbias = 1.15V of two different single bare ligand on Ag(111)showing the opposite tilt angle of the axis of symmetry of the two lobes in themiddle of the molecule [(a) ∼+16◦, (b) ∼-16◦]. Grids set-point parameters: (a)Vbias = -2.50V, It = 50pA; (b) Vbias = -2.50V, It = 50pA. . . . . . . . . . . . . . 44Figure 3.12 Normalized dI/dV curves of two adjacently bound molecules obtained by aver-aging over the regions enclosed by the corresponding colour outlines in the to-pography (see inset). Red: whole top single molecule; blue: top single moleculemolecular axis; orange: whole bottom single molecule; cyan: bottom singlemolecule molecular axis; dashed black: Ag(111), reference spectrum. Gridset-point parameters: Vbias = -2.00V, It = 400pA. Note: the divergence nearzero-bias created by normalization is removed. . . . . . . . . . . . . . . . . . 45Figure 3.13 STS energy maps at different biases showing the density of states of two bareligands adjacently bound. From (a) to (h) the biases are: -1.9V, -0.8V, +0.4V,+0.7V, +1.15V, +1.5V, +1.8V and +2V. Grid set-point parameters: Vbias = -2.00V, It = 400pA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Figure 3.14 Normalized dI/dV curves of three molecules adjacently bound obtained by av-eraging over the regions enclosed by the corresponding colour outlines in the to-pography (see inset). Red: whole top single molecule; blue: top single moleculemolecular axis; orange: whole middle single molecule; cyan: middle singlemolecule molecular axis; violet: whole bottom single molecule; pink: bottomsingle molecule molecular axis; dashed black: Ag(111), reference spectrum.Grid set-point parameters: Vbias = -2.50V, It = 50pA. Note: the divergence nearzero-bias created by normalization is removed. . . . . . . . . . . . . . . . . . 47Figure 3.15 STS energy maps at different biases showing the density of states of three bareligands adjacently bound. From (a) to (j) the biases are: -1.6V, -0.9V, +0.4V,+0.7V, +1.15V, +1.6V, +1.8V, +2.0V, 2.3V and +2.7V. Grid set-point parame-ters: Vbias = -2.50V, It = 50pA. . . . . . . . . . . . . . . . . . . . . . . . . . . 48xiiiFigure 3.16 (a) Normalized dI/dV plots of negative and positive bias of an isolated moleculeadsorbed on NaCl/Ag(111). These were obtained from two different singlepoint spectroscopy taken in the same position on the molecule (light blue dot inthe topography - insert) with different set-point parameters and bias range (neg-ative bias set-point: Vbias = -3V, It = 2pA; positive bias set-point: Vbias = +1.40V,It = 5pA). Dashed grey curves: NaCl reference spectra. At positive energies thethree peaks correspond to the LUMO (+1.65V, red bar), the double degener-ate LUMO+1 (+2.2V, green bar) and to the LUMO+2 (+2.9V, orange bar). (b)DFT-calculated normalized LDOS of a single TPPT in gas-phase (DFT per-formed by Katherine Cochrane). To compare the energy position of the peaksbetween experiment and theory, the energy of the first peak at positive bias (theLUMO) in the DFT spectrum was matched in value to the first experimentalone equal to +1.65V. Vertical bars correspond to: black, HOMO; red, LUMO;green, LUMO+1; orange, LUMO+2. . . . . . . . . . . . . . . . . . . . . . . . 50Figure 3.17 STS point spectra taken at twenty equally spaced locations along the long axisof a bare ligand on NaCl/Ag(111). They are presented with a vertical offsetand therefore they do not share the same zero baseline. The inset on top showsthe twenty positions were the measurements were taken. Corresponding colourplots are shown below it. The numbers on the right side are referenced in the text. 52Figure 3.18 (a) Topography of the grid measurement (Vb = -2V, It = 0.6pA) performed onan isolated TPPT on NaCl/Ag(111). (b) STS energy maps showing the electrondistribution at +1.65V of the molecule in panel (a). Grid set-point parameters:Vb = -2.00V, It = 0.6pA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53Figure 3.19 Gas-phase DFT-calculated electron distributions maps of the HOMO, the LUMO,the double degenerate LUMO+1 and LUMO+2 of a single TPPT molecule. Iso-surface level 0.02 electrons/r3Bohr. (DFT performed by Katherine Cochrane). . . 54Figure 3.20 Normalized dI/dV obtained from point spectroscopy taken at the center of themolecule at the position of the blue dots shown in the insets at the top of the twopanels. (a) TPPT on Ag(111) (set-point parameters: Vbias = 0.60V, It = 50pA),(b) on NaCl/Ag(111) (set-point parameters: Vbias = 1.40V, It = 5pA). . . . . . . 54Figure 3.21 STS energy maps showing the electron distributions of (a) a bare ligand onAg(111) at +1.15V (grid set-point parameters: Vbias = -2.50V, It = 50pA); and(b) on NaCl/Ag(111) at +1.65V (grid set-point parameters: Vbias = -2V, It =0.6pA). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55Figure 3.22 Plots of peak intensity profiles as a function of position along the lines shownby the dashed cyan arrows in the insets at the top of the panels. (a) is relative tothe +2.3V peak of a TPPT on Ag(111) and (b) the +2.2V on NaCl/Ag(111). . . 56xivFigure 3.23 Plot of the intensity profile as a function of position along the line shown by thedashed green arrow in the inset. The profile was obtained from the STS map(inset) of an isolated bare TPPT on Ag(111) at +1.8V. . . . . . . . . . . . . . . 57Figure 4.1 Constant-current STM images. (a) Isolated Fe adatoms (solid purple arrows)deposited on TPPT/Ag(111) system at ∼4.3K (Vb = 500mV, It = 10pA). TPPTmolecules still mantain the same orientation as before the addition of iron [fig-ure 3.3(a)]. (b) System of panel (a) after annealing at 323K for 10 minutes (Vb= 200mV, It = 50pA). (c) Fe deposited are RT on the TPPT/Ag(111) system (Vb= -1V, It= 10pA). Cyan arrows point towards iron clusters. (d) System of panel(c) after annealing at 373K for 10 minutes (Vb = -1V, It = 50pA). . . . . . . . . 60Figure 4.2 STM images and corresponding chemical structures of Fe-coordinated TPPT.(a) and (b): Single TPPT with right tpy group coordinated to one Fe adatom (Vb= -200 mV, It = 1 nA). Coordination is mediated by the rotation [red arrow in(a)] of peripheral pyr groups. Non-metalated tpy shows a depression (turquoisearrow) next to the peripheral pyr [panel (c)], which is not observed (green arrow)for metalated tpy [panel (d)]. (c)-(d) same scale. Blue: N; cyan: carbon; white:hydrogen; red: Fe. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61Figure 4.3 Isolated TPPT terminated only on the right-hand side by a single Fe atom. (a)Optimized DFT model on Ag(111) substrate along [1 -1 0] direction and (b)corresponding DFT-simulated STM image. Images calculated by ChenguangWang. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62Figure 4.4 STM images and corresponding chemical structures of Fe-coordinated TPPT.(a) and (b): Single TPPT with both tpy groups coordinated to one Fe adatom(Vb = -500 mV, It = 1 nA). (c) and (d): Adjacent proton acceptor/organic ringinteraction (red dashed lines) between two, singly metalated TPPTs (Vb = -500mV, It = 1 nA). Blue: N; cyan: carbon; white: hydrogen; red: Fe. . . . . . . . . 62Figure 4.5 High-resolution STM and DFT of a 5-molecule Fe-TPPT nanochain on Ag(111).(a) and (b): Negative and positive bias STM images of a 5-molecule Fe-TPPTchain [(a) : Vb = -500 mV ; (b): 100 mV, It = 1 nA]. (c) Difference STM to-pographic map resulting from subtraction of (a) with (b). Profile C along thecoordination center shows two protrusions, fitted with two Gaussians. (e) and(f): DFT-simulated topographic STM images of the optimal Fe-TPPT/Ag(111)configuration [(e): Vb = -500 mV; (f): 100 mV]. (g) Theoretical difference mapresulting from subtraction of (e) and (f). (d) Model of the DFT-optimized metal-organic chain. (h) and (i): Top and side-view of the DFT-optimized coordinationcenter. DFT images done by Chenguang Wang. . . . . . . . . . . . . . . . . . 64xvFigure 4.6 DFT simulated structure for an infinite chain of TPPT molecules linked by aFe-Ag-Fe structure along the [1 -1 0] direction (a). Panels (b) and (c) are thetop and side-view of the Fe-Ag-Fe linkage center, respectively. Images done byChenguang Wang. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66Figure 4.7 STM constant-current image showing bent chains (dashed red circled features)after annealing the TPPT/Ag(111) system at 523K for 30 minutes (Vb = -0.1V,It = 100pA). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66Figure 4.8 DFT-simulated constant current images of an infinite chain of TPPT moleculeslinked by the Fe-Ag-Fe structure shown in figure 4.6 at -500 mV (a), +200 mV(b) and their subtraction (c) = (a) - (b). Images done by Chenguang Wang. . . . 67Figure 4.9 Probability of finding a chain composed of n TPPT molecules after deposition ofFe at RT in continuous flux, before (blue) and after (red) subsequent annealingat 373 K during 10 min. Error bars represent uncertainty in the number of countsand are standard errors weighted by the area of each STM image used. Insert:detail of the longer chain distributions in the dashed box. . . . . . . . . . . . . 69Figure 4.10 Probability of finding a chain composed of n TPPT molecules after deposition ofFe at RT in subsequent very short depositions (10 seconds) separated in time bya fixed interval (5 seconds), before (blue) and after (red) subsequent annealing at373 K during 10 min. Error bars represent uncertainty in the number of countsand are standard errors weighted by the area of each STM image used. Insert:detail of the longer chain distributions in the dashed box. . . . . . . . . . . . . 70Figure 4.11 Constant-current STM images. (a) Fe/TPPT/Ag(111) system after simultaneousdeposition at RT of iron adatoms and TPPT molecules (Vb = -1V, It = 100pA).(b) Fe/TPPT/Ag(111) system after depositing the ligands first and then addingsmall amount of iron in subsequent very short depositions separated in time bya fixed interval (both depositions at RT) (Vb = -1V, It = 50pA). . . . . . . . . . 72Figure 4.12 Constant-current STM topographic images. (a) Fe/TPPT/NaCl/Ag(111) sys-tem: the NaCl island is completely free of adsorbates; the adjacent Ag(111)patch is covered by well-ordered and closed-packed TPPT molecules and ran-domly sparse Fe clusters (bright dots) (Vb=500mV, It=1pA). (b) Zoom in thearea enclosed by the red square in panel (a). It shows ordered TPPTs and Feclusters (dashed purple arrows) (Vb=500mV, It=25pA). (c) Zoom in the area en-close by the green rectangle in panel (b). It shows the TPPTs ordering (Vb=500mV,It=25pA). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73xviFigure 5.1 (a) Normalized dI/dV curves of a “1-molecule chain” obtained by averagingover the regions enclosed by the corresponding colour outlines (see inset). Pur-ple: whole “one-molecule chain”; red: molecular axis; dashed black: Ag(111),reference spectrum. Grid set-point parameters: Vb = -2.0V It = 1nA. Note: thedivergence near zero-bias created by normalization is removed. (b) Positivebias normalized dI/dV curves of bare ligand on silver (green, point-dash line),and on NaCl (grey, point-dash line) and of “one-molecule chain” on Ag(111)(solid, blue line). Grids set-point parameters: bare ligand on Ag(111) Vb =-2.0V It = 1nA; bare ligand on NaCl/Ag(111) Vb = +1.20V It = 0.4pA; coor-dinated molecule Vb = -2.0V It = 1nA. Note: each of these measurements wastaken with a different tip, which is why the curves look so different. . . . . . . 76Figure 5.2 STS energy maps at different biases showing the density of states of a “1-molecule chain” with both the tpy ends coordinated with Fe. From (a) to (j)the biases are: -0.6V, -0.3V, -0.1V, +0.4V, +0.8V, +1.2V, +1.5V, +1.65V, +1.8Vand +2V. Grid set-point parameters: Vb = -2.0V It = 1nA. . . . . . . . . . . . . 78Figure 5.3 STS energy maps showing the electrons distribution at the energy of the firsttunneling resonance of a single: (a) bare TPPT on silver at +1.15V (grid set-point parameters: Vb = -2.50V It = 50pA); (b) bare TPPT on salt at +1.65V(grid set-point parameters: Vb = -2.0V It = 0.6pA); and (c) a TPPT coordinatedwith Fe on both ends at +1.5V (grid set-point parameters: Vb = -2.0V It = 1nA). 79Figure 5.4 Normalized dI/dV curves of a four-molecule chain obtained by averaging overthe regions enclosed by the corresponding colour outlines (see inset). Purple:whole chain; cyan (L1), orange (L2), red (L3) and blue (L4): molecular axesfrom left to right; dashed black: Ag(111), reference spectrum. Bars: green,first resonance of bare ligand on silver; grey, on NaCl. On top of the inset asketch of the chain. Grid set-point parameters: Vb = -2.50V It = 2.5nA. Note:the divergence near zero-bias created by normalization is removed. . . . . . . . 80Figure 5.5 Normalized dI/dV curves of a five-molecule chain obtained by averaging overthe regions enclosed by the corresponding colour outlines (see inset). Purple:whole chain; cyan (L1), orange (L2), violet (L3), red (L4) and blue (L5): molec-ular axes from left to right; dashed black: Ag(111), reference spectrum. Bars:green, first resonance of bare ligand on silver; grey, on NaCl. On top of the inseta sketch of the chain. Grid set-point parameters: Vb = -2.50V It = 1.5nA. Note:the divergence near zero-bias created by normalization is removed. . . . . . . . 81Figure 5.6 STS energy maps at different biases showing the density of states of a four-molecule chain. From (a) to (h) the biases are: -0.4V, -0.1V, +0.4V, +0.8V,+1.2V, +1.5V, +1.65V and +1.8V. Grid set-point parameters: Vb = -2.50V It =2.5nA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82xviiFigure 5.7 STS energy maps at different biases showing the density of states of a five-molecule chain. From (a) to (h) the biases are: -0.4V, -0.1V, +0.4V, +0.8V+1.2V, +1.5V, +1.65V and +1.8V. Grid set-point parameters: Vb = -2.50V It =1.5nA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83Figure 5.8 Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines(see inset). Purple: whole system; blue (L1), violet (L2) and red (L3): molec-ular axes from left to right; dashed black: Ag(111), reference spectrum. Bars:green, first resonance of bare ligand on silver; grey, on NaCl. Grid set-pointparameters: Vb = -2.50V It = 2.5nA. Note: the divergence near zero-bias createdby normalization is removed. . . . . . . . . . . . . . . . . . . . . . . . . . . . 84Figure 5.9 STS energy maps at different biases showing the density of states of two coor-dination sites and three molecular axes. From (a) to (h) the biases are: -0.5V,-0.1V, +0.4V, +0.8V, +1.2V, +1.5V, +1.65V and +1.8V. Grid set-point parame-ters: Vb = -2.50V It = 2.5nA. . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Figure 5.10 Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines(see inset). Purple: whole structure; blue (L1), violet (L2) and red (L3): molec-ular axes from left to right; dashed black: Ag(111), reference spectrum. Bars:green, first resonance of bare ligand on silver; grey, on NaCl. Grid set-pointparameters: Vb = -2.50V It = 1nA. Note: the divergence near zero-bias createdby normalization is removed. . . . . . . . . . . . . . . . . . . . . . . . . . . . 86Figure 5.11 STS energy maps at different biases showing the density of states of two coordi-nation sites, three molecules, one of which is an edge-molecule not coordinatedwith Fe. From (a) to (h) the biases are: -0.6V, -0.1V, +0.3V, +0.9V, +1.2V,+1.5V, +1.65V and +1.8V. Grid set-point parameters: Vb = -2.50V It = 1nA. . . 87Figure 5.12 Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines(see inset). Purple: whole structure; red (L1), violet (L2) and blue (L3): molec-ular axes from left to right; dashed black: Ag(111), reference spectrum. Bars:green, first resonance of bare ligand on silver; grey, on NaCl. Grid set-pointparameters: Vb = -2.50V It = 50pA. Note: the divergence near zero-bias createdby normalization is removed. . . . . . . . . . . . . . . . . . . . . . . . . . . . 88Figure 5.13 STS energy maps at different biases showing the density of states of two coor-dination sites, three molecules, one of which is an edge molecule coordinatedwith Fe. From (a) to (i) the biases are: -0.6V, -0.1V, +0.4V, +0.9V, +1.2V, +1.5V,+1.65V, +1.8V and +2.5V. Grid set-point parameters: Vb = -2.50V It = 50pA. . 89xviiiFigure 5.14 STS energy maps showing the electron distribution of a 4-molecule chain at(a) +1.8V, (b) +1.5V, and (c) -0.1V. (a) and (c) show electron states at the metallocations, while (b) on the ligands’ axis. Grid set-point parameters: Vb = -2.50VIt = 2.5nA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92Figure 6.1 Ligand similar to a TPPT molecule but with three phenyl groups instead of two. 96Figure 6.2 Probable twisted Os-terpyridine structure. . . . . . . . . . . . . . . . . . . . . 96Figure B.1 Comparison of the I(V ) (left), dI/dV (center), (dI/dV )/(I/V ) (right) curvesobtained from raw (top) and smoothed data with span equal to 11 (bottom).Areas circled with cyan and green dashed lines are artifacts of the data analysisprocess. Note: The divergence in the smoothed data plot is bigger than in theraw data one. This is a result of the smoothing process. Different values of thesmoothing span results in different amplitudes of the divergence. . . . . . . . . 114Figure B.2 Comparison of the I(V ) (left), dI/dV (center), (dI/dV )/(I/V ) (right) STSmaps of an isolated TPPT on Ag(111) at Vbias = 1.6V, obtained from raw (top)and smoothed data with kernel of size 5 for the x and y directions and of size 11for the bias (bottom). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115Figure B.3 Schematic of the method used to determine the ROI. (a) First, select a rectan-gular area (dashed green rectangle) around the ROI (in this case the ROI is themolecular axis of a single TPPT molecule). (b) Second, set a threshold valuefor the apparent height of the topography (here, 54 pm). (c) Obtain the ROI,such as the all the (x,y) points satisfying the two conditions and discard all theothers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115Figure C.1 Normalized dI/dV curves of a single bare molecule obtained by averaging overthe regions enclosed by the corresponding colour outlines in the topography(see inset). Red: whole single molecule; blue: single molecule molecular axis;dashed black: Ag(111), reference spectrum. Grid set-point parameters: Vbias =-2.50V, It = 50pA. Note: the divergence near zero-bias created by normalizationis removed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124Figure C.2 STS energy maps at different biases showing the density of states of a singlebare ligand. From left to right, from top to bottom the biases are: -1.6V, -0.2V,+0.4V, +0.7V, +1.15V, +1.6V, +1.8V, +2.0V, 2.3V and +2.9V. Grid set-pointparameters: Vbias = -2.50V, It = 50pA. . . . . . . . . . . . . . . . . . . . . . . 125Figure C.3 (a) Normalized dI/dV curves of a “1-molecule chain” obtained by averagingover the regions enclosed by the corresponding colour outlines (see inset). Pur-ple: whole “one-molecule chain”; red: molecular axis; dashed black: Ag(111),reference spectrum. Grid set-point parameters: Vb = -2.0V It = 1nA. Note: thedivergence near zero-bias created by normalization is removed. . . . . . . . . . 126xixFigure C.4 STS energy maps at different biases showing the density of states of a “1-molecule chain” with both the tpy ends coordinated with Fe. From left to right,from top to bottom the biases are: -0.5V, -0.1V, +0.4V, +0.8V, +1.2V, +1.5V,+1.65V, +1.8V and +2V. Grid set-point parameters: Vb = -2.0V It = 1nA. . . . . 127Figure C.5 Normalized dI/dV curves of a six-molecule chain obtained by averaging overthe regions enclosed by the corresponding colour outlines (see inset). Purple:whole chain; green, cyan, orange, violet, red and blue: molecular axes; dashedblack: Ag(111), reference spectrum. Bars: green, first resonance of bare ligandon silver; grey, on NaCl. On top of the inset a sketch of the chain. Grid set-pointparameters: Vb = -2.50V It = 2.5nA. Note: the divergence near zero-bias createdby normalization is removed. . . . . . . . . . . . . . . . . . . . . . . . . . . . 128Figure C.6 STS energy maps at different biases showing the density of states of a six-molecule chain. From top to bottom the biases are: +0.8V +1.2V, +1.5V,+1.65V and +2V. Grid set-point parameters: Vb = -2.50V It = 2.5nA. . . . . . . 129Figure C.7 Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines(see inset). Purple: whole system; blue, violet and red: molecular axes; dashedblack: Ag(111), reference spectrum. Bars: green, first resonance of bare ligandon silver; grey, on NaCl. Grid set-point parameters: Vb = -2.50V It = 100pA.Note: the divergence near zero-bias created by normalization is removed. . . . 130Figure C.8 STS energy maps at different biases showing the density of states of two coordi-nation sites and three molecular axes. From left to right, from top to bottom thebiases are: -0.1V, +0.4V, +0.8V, +1.2V, +1.5V, +1.65V, +1.8V, +2V, and +2.5V.Grid set-point parameters: Vb = -2.50V It = 100pA. . . . . . . . . . . . . . . . 131Figure C.9 Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines(see inset). Purple: whole structure; red, violet and blue: molecular axes;dashed black: Ag(111), reference spectrum. Bars: green, first resonance ofbare ligand on silver; grey, on NaCl. Grid set-point parameters: Vb = -2.50VIt = 200pA. Note: the divergence near zero-bias created by normalization isremoved. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132Figure C.10 STS energy maps at different biases showing the density of states of of twocoordination sites, three molecules, one of which is an edge-molecule not co-ordinated with Fe. From left to right, from top to bottom the biases are: -0.1V,+0.4V, +0.8V, +1.2V, +1.5V, +1.65V, +1.8V, +2V, and +2.5V. Grid set-pointparameters: Vb = -2.5V It = 200pA. . . . . . . . . . . . . . . . . . . . . . . . 133xxFigure C.11 Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines(see inset). Purple: whole structure; blue, violet and red: molecular axes;dashed black: Ag(111), reference spectrum. Bars: green, first resonance ofbare ligand on silver; grey, on NaCl. Grid set-point parameters: Vb = -2.0V It =800pA. Note: the divergence near zero-bias created by normalization is removed. 134Figure C.12 STS energy maps at different biases showing the density of states of two coordi-nation sites, three molecules, one of which is an edge molecule coordinated withFe. From left to right, from top to bottom the biases are: -0.6V, -0.1V, +0.4V,+0.8V, +1.2V, +1.5V, +1.65V, +1.8V and +2V. Grid set-point parameters: Vb =-2.0V It = 800pA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135xxiList of Supplementary MaterialsVideo 1 Density of states of an isolated bare ligand: http://hdl.handle.net/2429/58286Video 2 Density of states of a four-molecule chain: http://hdl.handle.net/2429/58286xxiiGlossaryAi area of the considered imageAmin minimum size image of all imagesAFM Atomic Force MicroscopyD diffusion ratedz separation distance between tip and sampleDFT Density Functional TheoryDOS Density Of StatesDSSC Dye Sensitized Solar Cell∆V potential differencee electron chargeE energyEF Fermi energyEµ , Eν energies relative to the states ψµ and ψν , respectivelyF deposition ratef (Eµ), f (Eν) Fermi-Dirac distributionsφS, φt sample and tip work functions, respectivelyGGA General Gradient Approximationh¯ Plank constantHAS Helium Atom ScatteringHOMO Highest Unoccupied Molecular OrbitalHUV UltraHigh VacuumIt tunneling currentIR Infra-Redκ inverse decay lengthL-L Ligand-LigandLDOS Local Density Of StatesLEED Low-Energy Electron DiffractionLHe Liquid HeliumLMCT Ligand to Metal Charge TransferxxiiiLN2 Liquid NitrogenLUMO Lowest Unoccupied Molecular OrbitalLT Low-Temperaturem electron massMµν tunneling matrix elementMC Metal ChargeMLCT Metal to Ligand Charge TransferMO Molecular OrbitalMOF Metal Organic FrameworkOFET Organic Field Emission TransistorOLED Organic Light Emitting DiodeOMBE Organic Molecular Beam EpitaxyOPV Organic Photovoltaicψµ , ψν non-orthogonal sample and tip statesPAW Projector Augmented Wavesph phenylpyr pyridineρS local density of states of the sampleρt local density of states of the tiprt tip positionRPBE Revised Perdew-Burke-ErnzerhofRT Room TemperatureROI Region Of InterestSEM Scanning Electron MicroscopySTM Scanning Tunneling Microscopy/MicroscopeSTS Scanning Tunneling SpectroscopyT transmission probabilityTGA ThermoGravimetric AnalysisTPPT Terpyridine-Phenyl-Phenyl-Terpyridinet py terpyridineTSP Titanium Sublimation PumpUV Ultra-VioletVbias bias voltageVz voltage applied to the z-piezoVASP Vienna Ab-Initio Simulation Packagewi weighting factorWKB Wentzel-Kramers-BrillouinXRD X-Ray Diffractionz separation distance between tip and samplexxivAcknowledgmentsI want to acknowledge everyone who helped and supported me during my Ph.D., as well as the fol-lowing funding agencies: Natural Sciences and Engineering Research Council, Canadian Founda-tion for Innovation, British Columbia Knowledge Development Fund, the Max Planck-UBC Centrefor Quantum Materials, and the University of British Columbia.Un agradecimiento muy especial para Agustin Schiffrin por ser un mentor excelente, por su ayuda,y por su presencia y apoyo constante a lo largo de mi doctorado.Un ringraziamento speciale va anche al mio amatissimo papa’ senza l’aiuto del quale questa tesinon avrebbe visto la luce del sole.xxvDedicationAn Stephanie gewidmet...xxviChapter 1Introduction1.1 MotivationCurrently almost all commercial electronic devices are made of inorganic materials. Since the be-ginning of the Silicon Era (which drastically changed the electronics and computing industries)and thanks to the advent of the transistor [1], we have experienced an unprecedented technologi-cal progress based on technological research and development in the semiconductor industry. Thecontinual improvement of the performance of digital systems is described by Moore’s Law [2]: anexponential growth of the density of transistors and memory bits together with a miniaturizationof these functional units. As a result of this exponential improvement very compact, high-quality,high-speed devices are commercially available at considerably lower prices compared to a decadeago. However, in more recent years, an interest in organic materials and nanostructures has beenincreasing to facilitate further advances in computing and other electronic devices. This interesthas led to the development and implementation of electronic and optoelectronic devices such asOrganic Light Emitting Diodes (OLED), Organic Field Effect Transistors (OFET) and OrganicPhotovoltaics (OPV). Some of the advantages of using and commercializing organic devices are:low-cost, lightweight, mechanical flexibility, versatility in terms of fabrication and performance,ease of manufacture, and reduced environmental impact [3]. OPV systems are a good example oforganic-based devices that present numerous advantages compared to the silicon-based solar cells(i.e. limited cost for materials and manufacturing, thin, flexible, lightweight and transparent, avail-able in different colours, easy and fast to process, and environmentally friendly). Two are two maintypes of OPV systems: donor-acceptor and Dye Sensitized Solar Cell (DSSC). The donor-acceptortype consists of a heterostructure of donor-acceptor blends of photoactive materials, placed betweentwo electrodes [4]. When the sunlight is absorbed a bound electron-hole pair (exciton) forms in thebulk of one of the photoactive materials. The driving force provided by the donor-acceptor interfaceallows to overcome the exciton binding such that the pair can be separated (the hole is transportedby donor material, the electron by the acceptor), and lead to electric current. The structure and theworking mechanism of a DSSC is quite different. A DSSC consists of a TiO2 nanoparticles substrate1on which organic dyes are adsorbed, anchored by a functional group. The organic dyes are needed inorder to have absorption of the sunlight in the visible range. When the light is absorbed by the dye,the excited electron is injected into the substrate and it will percolate through it, reach the electrodeand contribute to the current production. At the same time, the oxidized dye is regenerated by theelectrolyte solution in which is embedded. Even though the internal mechanism of the two types ofOPV devices is different, they both have interfaces between diverse materials that play a key role.In both these types of organic devices, the biggest limitation is the efficiency, which is so far too lowto compete with the inorganic-based ones. For example, a commercial silicon-based photovoltaicpanel has greater than 20% efficiency (SunPower e20 modules [5]), while the best performing mar-ket organic solar cells (Heliatek [6]), including DSSCs, can achieve at most 12% [7]. One of themain challenges in improving the performance of organic devices lies in understanding the elec-tronic and optoelectronic processes involved at molecular interfaces in order to gain control overthem at the molecular scale. For example, in DSSCs the main limitations to the efficiency, are thecharge injection from the excited state of the adsorbed molecules into the substrate, and the electronrecombination [8–12]. These processes are affected by the interface’s structure, such as adsorptiongeometry of the organic compounds on the solid support and energy level alignment between ad-sorbate and substrate [13], and by the local environment, which organic materials are very sensitiveto [14, 15]. All these aspects vary significantly at the molecular scale and therefore, comprehensivebasic research investigations assessing the opto electronic structure, and control of the adsorptionconfiguration at this native scale is the key to improve and develop organic nanostructured devices,designed to implement more effective internal processes and higher efficiency.The main focus of this thesis was on studying suitable systems for organic photovoltaicand catalysis applications with particular interest in metal-organic complexes of the polypyridine1family. This class of compounds exhibits a wide range of opto-electronic, spin, and chemical prop-erties that can be exploited in various technological applications. Examples include DSSCs [16,17], catalysis [18–20], molecular electronics [21–23], molecular magnetism [24] and biomedicine[25, 26]. Metal-organic complexes consisting of group 8 transition metals (iron, ruthenium, andosmium) coordinated with bis- and terpyridine ligands have received large attention due to theirefficient broadband absorption, which ranges from ultraviolet to near-infrared. These optical transi-tions include intramolecular photo-induced Metal-to-Ligand and Ligand-to-Metal Charge Transfer(MLCT and LMCT), aiding the photo-induced charge separation sought for photovoltaics [8, 27],and photocatalysis [28].In photovoltaic applications, the efficiency depends on processes at a length scale of theorder < 10nm, on the details of the interfaces at the molecular scale, and on the local environ-ment. For example, scanning tunneling microscopy and spectroscopy studies on single N3 dyes(Ru-based bipyridine complexes) on TiO2 anatase (101) have shown that the level alignment be-tween the Lowest Occupied Molecular Orbital (LUMO) of the molecule and the conduction band1A pyridine (C5H5N) is an heterocyclic organic compound with one methine group (=CH-) replaced by a nitrogenatom.2of the substrate, and the associated driving force for photoelectron injection into the substrate areinfluenced by the adsorption configuration of the dye and their surrounding environment [13]. Inparticular, the molecules’ LUMO position was observed to vary between 1.3V and 2.3V, dependingon the dyes’ different binding modes with the surface, if they are adsorbed on a terrace or on astep edge, and also if they are alone or in a pair. This finding highlights the importance of havingcontrol over adsorbates-substrate systems and the surrounding environment at the molecular level.Moreover, it shows the capability of a Scanning Tunneling Microscope (STM) to provide struc-tural, electronic, and optoelectronic information at the atomic/molecular scale by relying on localtip-sample interactions. This study demonstrates that the STM technique is a very powerful probefor local investigation and characterization of the structure and the local electronic properties ofclean, well-defined surface systems at the critical length scale (molecular level). However, an STMis not as good for investigating solution-processed materials, and it cannot give information aboutthe bulk of a sample. The other big limitation of an STM is the temporal resolution, which is limitedby the instrument’s electronics. Nonetheless, with STM one can obtain very useful information oncarefully designed model systems that can help improving and implementing actual devices.1.2 Polypyridine metal-organic complexesCoordination complexes such as the deep, blue pigment called Prussian blue, have been used sincethe ancient Egyptian civilization. However, their nature was elucidated only with the studies ofChristian Blomstrand (father of the ion chain theory) in the late 1800s, followed by the Danishscientist Sophus Jorgensen and the Noble laureate Alfred Werner [29]. A coordination complex,or metal-organic complex, is formed by a central atom or ion (coordination center) surrounded byan array of bound molecules or ions, called ligands [30] [figure 1.1(a)]. The term ligand comesfrom Latin and it means to tie or bind. Typically, coordination centers are metals atoms includingtransition metals. The atom of the ligand bonded to the coordination center is called the donor as itshares its lone pair electrons with the metal. This sharing of charges creates an electrostatic attrac-tion that stabilizes the complex. Generally, coordination systems are composed of more than oneligand that can be either of the same species or not. Bonds between coordination centers and sur-rounding ligands are usually quite strong and the number of coordination bonds with the metal coredetermines the coordination number of the whole complex (most common numbers are 6, 4 and 2).The number of bonds depends on geometry, charge and electron configuration of both coordinationcenter and surrounding ligands. In particular, since metal ions have usually more than one possiblecoordination number they can arrange in different ways depending on the surrounding environment.The most common geometries are the octahedral [figure 1.1(a)], the square-planar [figure 1.1(b)],and the tetrahedral [figure 1.1(c)] ones. In transition metal atom complexes the orbitals involved arethe σ and pi of the ligand and the d, s and p of metal ion. It is the overlap of the ligand and metalorbitals, and the ligand-ligand repulsion that lead to specific coordination geometries and influencethe final electronic structure, by determining the occupied and unoccupied, metal and ligands energy3levels positions, and as a consequence, their alternation. Because of spatial constraint in a plane,square-planar complexes present an electronic structure much more complicated than the octahe-dral ones (3-directional geometry). Unlike the latter case, in the former one there are two distinctlydifferent sets of potential pi-bonding orbitals, the ones parallel to the molecular plane and the onesperpendicular to it [31].Figure 1.1: Schematic of the most common coordination geometries: octahedral (a), square-planar (b), and tetrahedral (c). Figures reproduced from [32]Figure 1.2: General schematic of the four different types of transitions occurring in metal-ligand complexes. MC are the transitions occurring inside the metal d orbitals; LMCT arethe ones between occupied ligand orbitals and unoccupied metal ones; MLCT betweenoccupied metal orbitals and unoccupied ligand ones; and L-L transitions are betweenligand orbitals. Image adapted from [33].The properties of these compounds, such as light absorption (colour), magnetism, and reactivity,are dictated by their electronic structure. Magnetism and reactivity are mainly related to unpaired4Figure 1.3: (a) Absorption spectra of the two different of ruthenium-terpyridine compoundsshown above the spectra [green (TUS-25) and black (black dye) curves] and the radicalR (blue curve) in solution. Figure reproduced from [34]. (b) Solar spectrum as viewedthrough the atmosphere. Figure reproduced from [35].electrons or half-filled orbitals, whereas light absorption depends on the possible electronic transi-tions occurring between occupied and unoccupied states. The electrons sharing between ligands andcoordination center results in the d orbitals of the metal to split in energy and lose their degeneracy.The characteristic of this splitting is determined by different factors: the nature of the coordinationcenter and the ligands, the metal oxidation state, and the arrangement of the ligands around thecentral atom. In coordination complexes, there are four types of possible electronic transitions: 1)Ligand-Ligand (L-L) between a piL (occupied) orbital of the ligand and a pi∗L (unoccupied) one; 2)d-d transitions (MC) where electrons are excited from an occupied metal orbital (piM) to anothermetal unoccupied one (σ∗M); 3) LMCT transitions where electrons are excited from a piL ligand or-bital to an unoccupied metal one (σ∗M); and 4) MLCT transitions from an occupied metal state (piM)to a ligand pi∗L one [figure 1.2]. Due to the large energy difference between different initial and finalstates high energy excitations can also occur in these compounds. They range from the near-UVregion of the L-L transitions to the visible or near-IR of the LMCT and MLCT ones [figure 1.3].Polypyridine complexes are examples of coordination structures consisting of pyridinebased ligands like bipyridine and terpyridine, coordinated to transition metal atoms such as ruthe-nium [figure 1.3(a)], osmium, iron, platinum, etc. Because of the robustness and electronic structureresulting from the metal-ligand coordination they are widely used in photovoltaic [17, 36] and catal-ysis [18–20] applications.1.3 Depositing osmium-terpyridine on an Ag(111) surfaceRuthenium (Ru) and osmium (Os) polypyridine complexes are commonly used in working DSSCs[8–11]. In these devices the dyes are anchored onto the TiO2 nanoparticles substrate in a disorderedmanner, covering the majority of the surface (about a monolayer). The first experiments were tar-geted to study Os-terpyridine compounds on an Ag(111) surface. Os-terpyridine complexes werechosen over the Ru ones mainly for two reasons [37, 38]. The MLCT/LMCT transitions occur at5longer wavelengths, and therefore the adsorption range is broader; and also these compounds showstronger coordination bonds, which should result in greater stability at higher temperatures. Theprimary challenge was the in situ deposition of the whole complex in ultrahigh vacuum via thermalevaporation. This type of deposition method consists of gradually heating a quartz crucible contain-ing the molecular solid up to its sublimation temperature. Around that temperature a molecular fluxis formed and the deposition begins. Its success depends on different factors. The most importantrequirement is that the decomposition temperature must be higher than the temperature where sig-nificant sublimation and deposition occurs. If that is not the case, some of the chemical bonds ofthe complex can break due to the high temperature.Figure 1.4: (a), (b) and (c) are the chemical structure of the three Os-terpyridine-based ligands.Complex (a) was synthesized at the Department of Chemistry and Low-Carbon EnergyResearch Center in the National Tsing Hua University in Hsinchu (Taiwan) [39], while(b) and (c) have been synthesized in Mike Wolf’s group at the Chemistry Department atUBC. Both (b) and (c) complexes have a PF6 counterion. (d) ThermoGravimetric Anal-ysis (TGA) of the (b) compound (measurement performed by Mike Wolf’s group). Thisshows the percentage of weight loss of the complex as a function of the increasing tem-perature. Each visible drop should correspond to the loss from the compound of a pieceof it. Here, the first drop (solid red line), very likely close to the decomposition tempera-ture, is at around 350◦C. (e) Constant-current STM image of unidentified adsorbates on aAg(111) substrate (Vb = +2V; It = 10pA). This image was taken after trying to deposit the(b) compound at a temperature close to its decomposition temperature (about 350◦C).Because of their relatively small dimension those adsorbates could be the counterionsPF6.Three different Os-terpyridine complexes were tried in this experiment [figure 1.4(a)-(c)]. The depositions of each was performed by gradually increasing the temperature of the cruciblecontaining the molecular powder, up to the decomposition temperature (∼350◦C) but unfortunatelynone of them were successful. The only adsorbates imaged afterwards had dimensions too small tobe the entire compound and they were associated with broken parts of the whole, larger complex6[see figure 1.4(e)]. The reasons for the failure were attributed to the large size of the molecule, theweak metal-ligand coordination bond, and charged species with counterion (molecular salt) of allthese compounds.On realizing that in situ deposition of the whole molecular complexes onto a surface bystandard methods was unlikely to be reliable, we decided to pursue an on-surface supramolecu-lar chemistry approach to fabricate a clean, well-defined and controlled photoactive metal-organicnanostructure on the inorganic surface.1.4 On-surface supramolecular chemistry1.4.1 Supramolecular chemistry of molecular self-assemblySupramolecular chemistry is the domain of chemistry that studies systems composed of a discretenumber of assembled compounds bound together via weak and reversible intermolecular interac-tions [40]. One of the main concepts explained by supramolecular chemistry is molecular self-assembly. The self-assembly of a molecular system is a process where the constituents adopt adefined arrangement without intervention [41]. There are two kinds of self-assembly: intermolec-ular and intramolecular. The latter indicates the natural folding of a molecular complex in orderto minimize its own energy, while the former refers instead to the actual self-assembly of differentatoms and molecules held together by hydrogen bonds, metal coordination, van der Waals forces,pi−pi interactions, and electrostatic effects.Single building blocks can assemble through directed interactions via non-covalent bondsto give higher order self-organized structures. Lower temperatures can be problematic because self-assemblies might require distortions of the molecules into a more energetically favorable confor-mation and in a cold environment this would be slowed down. Because of the control over theresulting structures and the different functionalities shown by these types of assemblies they areused in a wide variety of applications such as energy storage [42], catalysis [43], sensors [44], andphotovoltaic systems [16].Metal-Organic Frameworks (MOFs) are a large class of porous crystalline materials, witha very high surface area consisting of self-assembled organic and inorganic components [45–48].They can extend over large areas (up to centimeters) and in multiple directions. They are very flexi-ble in their geometry, size — without changing topology —, and functionalities. For these reasons,MOFs are exploited in different fields of applications, like gas storage and separation, heterogeneouscatalysis, membranes, high-capacity adsorbents, biomedicine, fuel cells, supercapacitors and so on.From a careful choice of the basic building motifs and control over their connectivity at the vertices,it is possible to engineer the final structure of the frameworks and their functions and properties.71.4.2 On-surface supramolecular coordination chemistryThe functional motifs of self-assembled nanostructures and coordination complexes can be extendedto surface-bound supramolecular chemistry [49–51]. This offers practical advantages in terms ofboth synthesis and control for applications where a solid support is essential such as photovoltaicsand heterogeneous catalysis. The presence of a surface introduces both challenges and opportuni-ties in the design of bottom-up nanostructures [41, 49, 52, 53]. The interplay with the substratecan modify the desired design, but can also allow fabrication of unique systems presenting coordi-nation symmetries, metal oxidation states, and polyatomic metal centers [51, 54, 55] that are verydifferent from those obtained in three-dimensions (e.g. solution). It is the balance between metal-ligand, intermolecular, and surface interactions that determines the formation, the morphology, andthe chemical and electronic properties of a structure. Metal atoms can covalently bond on the sur-face to more energetically favorable adsorption sites and therefore they might have less freedomin their spatial arrangement [49]. On the other hand, molecules interact with weak and directionalnon-covalent bonds and their adsorption is determined by a competition of intermolecular hydro-gen bonds, pi−pi and van der Waals interactions, and adsorption interaction with the substrate [56].STM studies have also revealed that the formation of metal-ligand nanostructures is usually directedby both metal and ligands [53]. For these reasons, several factors must be considered in order tofabricate, with high-fidelity, robust, low-dimensional coordination systems on a surface with de-sired functionalities. Moreover, the different components, i.e. organic linkers and metal atoms [41],must be deposited in a controlled manner [figure 1.5]. Their mobility on the substrate can differby several orders of magnitude depending on the temperature [58]. Once the adsorbates are on thesurface their behavior is determined by adsorbate-adsorbate and adsorbate-substrate interactions,and on macroscopic parameters such as deposition rate F and temperature. The final arrangement isdefined by phenomena including surface diffusion, desorption, non-covalent binding, self-assembly,chemical reactions induced by the substrate, and binding to specific sites such as step-edges. Thebalance between all these processes is determined by the interplay between thermodynamics andkinetics of the adsorbates-substrate system [41]. The adsorbates’ diffusion must be thermally acti-vated because there are potential barriers on the surface that need to be overcome in order to findthe most stable and energetically favorable adsorption configuration. The key parameter for defin-ing the kinetics of the growth is the ration D/F, where D is the adsorbate’s diffusivity obeying theArrhenius law [41, 59]. At a constant deposition rate, D/F defines the average distance travelled byan adsorbate to find another one and form a new aggregate or attach to a pre-existing one. For largevalues of D/F (D faster than F), the adsorbates can explore the surface and the growing structurescan find a configuration of minimum energy. Here, the assembly process happens near conditions ofequilibrium [60]. If D/F is small (F is too fast) then the system cannot relax into a thermodynamicequilibrium but it reaches a metastable non-equilibrium configuration depending on the balance be-tween adsorbate-substrate and lateral adsorbate-adsorbate interactions. In this case, the growth ofthe assembly is determined by kinetics [figure 1.6] [61]. A given set of initial building blocks canself-assemble via different energetic pathways depending on macroscopic parameters like tempera-8Figure 1.5: (I) STM topographic images of tunable metal-organic honeycomb nanomesheswith designed dicarbonitrile linear linkers on Ag(111). On the lower side of panel aschematics of the structures are superimposed on the STM images. Panel b: large-scaleSTM image. Figure reproduced from [55]. (II) STM constant-current mode images ofL-methionine stripes with molecular resolution. Panel a: Grating of double rows. Panelb: Individual molecules appear as elliptical features. Schematic of the molecules and theinteraction bonding (white dashed lines) are superimposed on the STM image. Figurereproduced from [57].ture, and local environmental conditions [62]. Self-assembly pathways can be distinguished in theones near-equilibrium conditions and the ones far-from-equilibrium. In the former case the system’sevolution is mainly governed by thermodynamics, while in the latter one the dynamic effects de-termine the resulting structure. Hence, since the same initial components can form different stableassemblies depending on the self-assembly pathway followed, it is possible to control the growth ofstructures and obtain systems with specific characteristics by tuning the thermodynamic and kinetickey parameters.In conclusion, on-surface supramolecular chemistry is a clean surface science preparationtechnique of robust and well-ordered nanostructures and frameworks with defined functionalities.By carefully choosing and designing the different components and controlling the growth param-eters (deposition rate and temperature) it is possible to construct atomically perfect systems witha good control over structure and properties [41, 49, 52, 53]. This method also provides a flexi-ble platform for investigating the influence of different ligands and metal combinations, giving the9Figure 1.6: Figures reproduced from [41]. (a) Schematic of atoms or molecules diffusing ona surface after being deposited from the vapor phase. F is the deposition rate and Dis the diffusion rate. The type of growth is determined by the ratio between D and F.If F is bigger than D, the growth is controlled by kinetics resulting in non-equilibriumconfigurations. Examples are the metallic islands shown in the two images on top of thesecond panel from the left of figure 1.5(b). If D/F is large, the growth occurs close to theequilibrium and well-defined nanostructures can form. Examples are shown in the first,the second (bottom image) and third panels of figure 1.6(b) [41].opportunity to explore metal charge and spin effects, coupling, and ligand steric and vibrationaleffects.1.5 Iron-terpyridine self-assembled nanochainsOn-surface self-assembly offers control at the single molecule level over structure and propertiesof a large variety of systems and allows fabrication of high-fidelity structures from carefully cho-sen components. Scanning tunneling microscopy is an unbeatable surface science technique forlocal investigation and characterization of clean systems at the nanoscale. In this work, we used thecombination of the two to characterize the structural and electronic properties of an on-surface self-assembled macromolecular complex. The choice of the molecular and metallic components wasmade in order to fabricate a metal-ligand complex presenting suitable properties for photovoltaic10and catalysis applications. The molecules used were bis-terpyridine ligands (Terpyridine-Phenyl-Phenyl-Terpyridine (TPPT)) coordinated with iron (Fe) metal centers. Ligands and metal atomswere subsequently deposited in situ, in ultrahigh vacuum, onto a clean Ag(111) surface and theywere observed to form thermally activated self-assembled nanochains. This thesis discusses themorphology, formation steps, and electronic structure of the resulting surface-synthesized macro-molecular structures [figure 1.7].Figure 1.7: Schematic of the thermally activated self-assembly of TPPT molecules and Featoms (red dots) on a Ag(111) surface. The final structure, characteristics of the coor-dination linkage, and electronic properties of the resulting chains are the focus of thisthesis.11Chapter 2Experimental MethodsThis chapter is devoted to the introduction and description of the scanning tunneling microscopytechnique, the equipment used in the experiments and the sample preparation methods. In particular,scanning tunneling microscopy and the associated spectroscopic counterpart Scanning TunnelingSpectroscopy (STS) are introduced in section 2.1, and the related theories in section 2.2. A detaileddescription of the experimental setup is given in section 2.3, and the sample preparation methods aredescribed in section 2.4. Properties and preparation characteristics of Ag(111) and NaCl/Ag(111)surfaces are introduced and discussed in sections 2.4.1 and 2.4.2, respectively. The molecular andmetal depositions are described in sections 2.4.3 and 2.4.4.2.1 Introduction to scanning tunneling microscopyThe advent of STM has been a major breakthrough in studying and understanding the propertiesof different types of surfaces, and in controlling and building nanoscale devices and nanostructureswith specific functionalities starting from single atoms and molecules. The first scanning tunnelingmicroscope was developed in 1982 at IBM by Binnig and Rohrer [63] as a versatile tool for localinvestigations of metallic and semiconducting surfaces, including adsorbed atoms and molecules.The operational principle which the STM is based on is the quantum mechanical tunneling effectwhereby a particle can tunnel from one electrode to another (tip and sample) when they are sepa-rated by a small distance of the order of 5 -10 A˚. Measured tunneling currents are of the order of1pA - 10nA, and they decay exponentially with the distance in the vacuum barrier that separatesthe two electrodes. By scanning a metallic tip over the sample and measuring the tunneling current,one can obtain topographical information about the surface and adsorbates with very high resolu-tion. However, in interpreting STM images, one should be aware that electronic information aboutthe surface is convolved with the physical structure meaning that care must be taken when lookingat them. Compared to surface science techniques like Low-Energy Electron Diffraction (LEED),X-Ray Diffraction (XRD), and Helium Atom Scattering (HAS) that only allow mapping from av-eraged areas of the reciprocal space of samples with a periodic structure, STM measurements givedirect, real-space, local observations of conducting/semiconducting surfaces and adsorbates such12as molecules. Common employed techniques for real-space structural and morphological samplescharacterization are Scanning Electron Microscopy (SEM) and ambient Atomic Force Microscopy(AFM). SEM can image areas ranging from approximately 1 centimeter to 5 micron with a reso-lution between less than 1 nanometer and 20 nanometers. Ambient AFM can scan areas of 100microns in XY directions with molecular resolution. However, even though both these techniquescan resolve structural characteristics of a sample at the molecular level, Low-Temperature ScanningTunneling Microscopy (LT-STM) in Ultra-High Vacuum (UHV) can also provide information onthe electronic structure of a system with very high energy resolution. Demonstrating the power ofthis technique in a striking way, in 1983, Binnig et al. reported atomic resolution of the Si(111) (7 x7) surface reconstruction [64] [figure 2.1(a)]. Adsorbates like single atoms and molecules, but alsobigger compounds, can be deposited and studied directly on conductive substrates. The first organicspecies imaged with STM were benzene molecules co-adsorbed with carbon monoxide (CO) ona Rh(111) surface [65] [figure 2.1(b)]. STM experiments have also shown that electrons can stilltunnel through thin insulating films (for example a few monolayers of salt (NaCl)) grown on topof a conductive substrate, enabling a probe of adsorbates at least partially electronically decoupledfrom the surface [66]. Moreover, the metallic tip of the microscope serves not only as one of theelectrodes but it can also be used as a nanoprobe to manipulate atoms and molecules on the surfacein a controllable way and with atomic precision [figure 2.1(d)]. Eigler et al. demonstrated such amanipulation procedure with Xe atoms on Ni(110) for the first time in 1990 [67].An STM can be used as a microscope to obtain structural information. It can accesstopographic information of a sample by recording the tip-sample distance adjustments required tomaintain the measured tunneling current at a constant pre-set value (constant-current STM). Thanksto the sub-atomic spatial resolution of the position of the electrons obtainable with STM, it is possi-ble to achieve atomic resolution of systems with well-localized orbitals. Since the tunneling currentdepends on the electronic structure of the sample (and also of the tip - see section 2.2) constant-current STM images are influenced by the integrated Local Density Of States (LDOS) as will beshown below.An STM is also a powerful spectroscopic tool to learn about the electronic propertiesof systems with atomic resolution. Information about the LDOS of a sample is contained in thefirst derivative of the tunneling current (conductance of the tunneling junction), see equation (2.4)in the next section [70–72]. In principle, STM can access chemical mapping of molecular speciesadsorbed on the surface since the tunneling electrons can excite vibrational modes due to inelas-tic scattering [68]. The energy loss of the tunneling electrons is quite sensitive to the the excitedmodes’ chemical nature and so the measured current may contain information of the system’s chem-istry much like IR and Raman spectroscopy [figure 2.1(c)]. Ho et al. [68] have presented a methodto obtain the chemical information of adsorbates via the second derivative of the tunneling signal.Lastly, with an STM one can obtain information on the average or effective work function of tipand sample from the inverse decay length of the exponentially decaying tunneling current (see nextsection) [73].13Figure 2.1: Examples of scanning tunneling microscopy and spectroscopy capabilities. (a)Constant-current STM image of the atomically resolved reconstructed Si(111) - (7 x 7)surface obtained by Binnig in 1983. Figure reproduced from [64]. (b) Three-dimensionalview of an STM topography of carbon monoxide (CO) and benzene coadsorbed onRh(111). The three-fold ring-like features are benzene molecules. CO is not resolved.Figure reproduced from [65]. (c) Chemical imaging using inelastic scanning tunnelingspectroscopy. (i) Constant-current STM image of a C2H2 molecule (left) and a C2D2molecule (right). (ii), (iii) and (iv) d2I/dV2 images of the same area as (i) at 358 mV(i), 266 mV (ii) and 311 mV. (ii) reveals only the C2H2 molecule because at that voltageone of its vibrational modes is excited. (iii) shows only the C2D2 molecule for samereason as (ii). (iv) none of the vibrational mode of any of the two adsorbates are excitedtherefore they are not revealed by the d2I/dV2. Figure reproduced from [68]. (d) Ex-ample of atomic manipulation: building a quantum coral using iron atoms for electronsconfinement on Cu(111). Figure reproduced from [69].In conclusion, STM is a unique technique that allows accessing information on the topog-raphy and electronic structure, below and above the Fermi level, of surfaces and adsorbed specieswith atomic resolution, and from there determining their relevant physical and chemical properties.142.2 Scanning tunneling microscopy and spectroscopy theoryThe quantum mechanical effect of tunneling, the principle upon which scanning tunneling micro-scopes are based, states that there is a finite probability that an electron can penetrate through apotential barrier with energy higher than its own [figure 2.2]. The larger the width of the barrierdz, the lower the tunneling probability. This event is possible because the wavefunction of theparticles/electrons within the barrier region is non-zero unlike in classical mechanics but it decaysexponentially with the distance. Thus, electrons from a conductive sample have a non-zero prob-ability to cross a potential barrier of finite width and tunnel into another conductive electrode, i.e.the tip, keeping their energy unchanged (the same is valid for electrons tunneling from the tip tothe sample). The overall measured tunneling current It is zero unless a bias voltage Vbias is appliedFigure 2.2: Schematic of the tunneling effect. A particle of wavefunction Ψ and energy E,lower than the energy of the potential barrier U0, has a non-zero probability of tunnelingthrough the barrier of a finite width L. Its wavefunction decays exponentially within thebarrier and on the other side of it the wavefunction amplitude is reduced depending onthe thickness of the barrier. Figure reproduced from [74].to one of the electrodes in order to shift the two Fermi levels (EF ) with respect to each other. Inthis way, electrons with energy between EF and (EF - eVbias) (with e being the electron charge) con-tribute to the non-zero measured It . Figure 2.3 (right side) shows a tunneling junction where a biasvoltage is applied and electrons can flow from the tip to the LUMO of an organic species adsorbedon the surface. Since the electron probability density decays exponentially through the barrier thetunneling current is exponentially sensitive to the tip-sample distance and thus current variationscontain information of the sample topography with a sub-A˚ vertical resolution (e.g. a distance vari-ation of ∼ 1 A˚ corresponds to an It variation of one order of magnitude [75]). In the low bias range,the tunneling current It can be approximated by:It ∝ exp(−2κz), (2.1)15where z is the separation between substrate and tip, and κ is the inverse decay length [75]. Gen-erally, the lateral resolution is mainly limited by the geometry of the tip (i.e. only the last atom isconsidered because of the exponential dependence of the tunneling current). In real-space imagingideal tips with a single atom at the apex could reach lateral resolution on the order of 1 A˚. It is thanksto such high lateral and vertical resolutions that an STM can achieve atomic imaging.Figure 2.3: Left side: Simplified schematic of a scanning tunneling microscope. Tip and sam-ple are separated by a distance dz of few A˚. In this case a bias voltage Vbias is applied tothe sample. A system of piezos allows scanning the tip over the surface of the sample (x- y plane) and adjusting the dz separation distance. The feedback loop system regulatesdz by comparing the measured amplified tunneling current It with the pre-set Iset currentvalue. Right side: Schematic of a tunneling junction. The metallic tip’s Density Of States(DOS) is assumed to be a step-function. The applied bias voltage shifts the tip’s Fermilevel (EF ) up with respect to the organic samples one. The electrons contributing to thetunneling current are tunneling through the vacuum barrier from the tip to the LUMO ofthe organic sample.When an STM is used in constant-current mode two-dimensional topographic images are obtained.In this case, the tip is scanned over the surface of the sample and the It current is measured. Dur-ing the scan the tip-sample separation dz is controlled by a feedback loop and it is adjusted withpicometer precision, in order to keep the measured current constant and equal to the pre-set It value.The other pre-set parameter is the bias voltage applied between tip and sample (conventionally thesample bias is reported with respect to the tip) that shifts the Fermi levels with respect to each otherand determines which electrons can contribute to the tunneling current (only the ones with an energybetween EF and (EF −eVbias)). The tip is moved in the x-, y- and z- directions by a system of piezo-electric actuators [figure 2.3 (left side)]. The voltage applied to the z-piezo (Vz), which is the onecontrolling the tip-sample distance, is recorded at each position (x,y) in a matrix Vz(x,y) from whichthe two-dimensional surface topography of the sample is obtained. It is important to note that themeasured tunneling current depends on the electronic structure of both tip and sample; therefore, anSTM topographic map is the result of the convolution of the geometrical and electronic properties16of the tip-sample system. This is a key factor to take into account in the data interpretation.The expression for the It current given by the equation (2.1) can also be written in a formthat gives more insights into the influence of the electronic structure of the sample. According toTersoff and Hamman [70, 71, 75], by applying first order perturbation theory, the tunneling currentIt can be approximated as:It =2pieh¯ ∑µ,ν[ f (Eν)− f (Eµ)]|Mµν |2δ (Eµ + eVbias−Eν), (2.2)where f (Eν) and f (Eµ) are the Fermi-Dirac distributions of tip and sample respectively, e is theelectron charge, h¯ is the Planck constant, Vbias is the bias voltage applied to the sample, and Mµνis the tunneling matrix element between the non-orthogonal sample and tip states ψµ and ψν withenergies Eµ and Eν , respectively. The tunneling matrix element Mµν can be written as:Mµν =h¯2m∫S(ψ∗µ∇ψν −ψ∗ν∇ψµ)dσ , (2.3)where S is any surface lying within the barrier region and m is the free electron mass. The expressionin parentheses is the probability current density operator. The above formula was introduced byBardeen in 1961 [76] to describe the tunneling current between two metallic electrodes separatedby an insulating oxide layer.In order to obtain a more practical expression for It several assumptions must be made.1. The ψν and ψµ are assumed to be independent from each other, even though this is valid onlyfor large tip-sample separations.2. For periodic systems the substrate states can be considered to be two-dimensional Bloch statesdecaying exponentially into the potential barrier. This is not valid for non-periodic systemswhere the sum in equation (2.4) is replaced by an integral.3. The tip wavefunctions are considered to be s-wave in nature.4. The two Fermi-Dirac distributions, f (Eν) and f (Eµ), are approximated by step-functionsassuming a very low temperature.Under these assumptions and for bias voltages approximately below 1V, the expression for It be-comes:It ∝∑µ|ψµ(rt)|2δ (Eµ −EF). (2.4)A more general expression for the tunneling current, valid also for larger values of Vbias, was givenby Selloni et al. [77] using the Wentzel-Kramers-Brillouin (WKB) approximation:It(Vbias) ∝∫ eVbias0ρS(rt,E)T (E,Vbias,z)dE. (2.5)17Here, E is the energy considered with respect to the Fermi level, ρS(rt,E) is the local density ofstates of the sample at the energy E and at the tip position rt, ρt (density of states of the tip) wasassumed to be constant in the energy range [EF ; (EF -eVbias)], and T (E,Vbias,z) is the tunnelingtransmission probability. T (E,Vbias,z) is obtained from Bardeen’s tunneling matrix Mµν assumingtwo-dimensional Bloch functions for the sample states, and both tip and sample states decayingexponentially in the potential barrier. Under these assumptions, the expression for the tunnelingtransmission probability is:T (E,Vbias,z) = exp(−2z mh¯2√φS+φt + eVbias−2E), (2.6)where φS and φt are the sample and tip work functions, respectively. Typical values for metalswork function are of the order of 5 eV (e.g. φAg(111) = 4.74eV [78]). Equation (2.6) gives theexponential dependence previously shown in equation (2.1) and also relates the inverse decay lengthκ to physical quantities. Both the Tersoff-Hamann approach and the more general expression ofSelloni link the tunneling current to the sample topographic and electronic properties. Even thoughin both approaches a number of quite strong assumptions are made (s-wave nature of the tip states,tip and sample states independence, independence of the transmission probability term from the x, yposition, and simplified dependence from the energy E), both these expressions are valid and largelyused [79, 80].Equation (2.5) shows that It carries information about the sample electronic density ofstates ρS(rt,E). For large biases, by taking the first derivative of It with respect to Vbias one obtains:dI(Vbias)dVbias∝ eρS(rt,eVbias)T (eVbias,Vbias,z)+ e∫ EF+eVbiasEFρS(rt,E)dT (E,eVbias,z)dVbiasdE. (2.7)Here, ρS(rt,E) is convoluted with the derivative of the transmission factor T with respect to Vbias,which is characterized by having exponential behavior. If one divides dIt/dVbias by It/Vbias oneobtains:dI(Vbias)/dVbiasIt(Vbias)/Vbias=dIt/dVbiasd(lnVbias)/Vbias/d(ln It)/dIt=d(ln It)d(lnVbias)=ρS(eVbias)+∫ eVbias0ρS(E)T (eVbias,eVbias)dd(eVbias)(T (E,eVbias))dE1eVbias∫ eVbias0 ρS(E)T (E,eVbias)T (eVbias,eVbias)dE.(2.8)The first term in the numerator of this equation is the LDOS of the sample, while the second oneis a background term characterized by an exponential behavior due to the transmission factor T.Since the background term is found to be significantly smaller than the first term, this normalizationprocess (of dividing the first derivative dIt/dVbias by It/Vbias) is very useful in eliminating the strongexponential contribution due to T [81–84]. At small voltages (i.e. eVbias << φS), the transmission18factor can be considered constant, and therefore one can directly assume:dI(Vbias)dVbias∝ ρS(rt,eVbias). (2.9)The technique for extracting spectroscopic information from the samples’ electronic configurationby using tunneling current measurements at different biases is known as scanning tunneling spec-troscopy. The STS experimental procedure consists of placing the tip at a specific location rt on thesample, defining the tip-sample separation distance via pre-set It and Vbias values (set-point parame-ters), opening the feedback loop, sweeping the applied bias over a desired range of values (but beingcareful to stay in the tunneling regime and avoid field emission of the electrons) and recording theIt for each bias value. During such a measurement, the tip-sample separation remains constant andit is determined by the set-point parameters. This type of measurement can be repeated on a seriesof equally spaced points in the x and y directions to cover a large area (on the order of tens of nm2in size) containing, for example, an adsorbate of interest. Then, by evaluating the first derivative ofIt with respect to Vbias at each (x, y) point one can:1. learn about the density of the electronic states at specific points (x, y);2. spatially resolve electronic features (position of the electronic states in space) of the wholearea considered by looking at the dIt/dVbias value (or normalized dIt/dVbias, as appropriate)per each (x, y) at a specific Vbias.Hereinafter, I will refer to this measuring method as grid measurements. The dIt/dVbias and the(dIt/dVbias)/(It/Vbias) versus bias plots will be generally called STS spectra or point spectroscopy;and the spatial distributions of normalized dIt/dVbias at a specific Vbias will be referred as normal-ized dIt/dVbias, STS energy maps, or electron density distributions [figure 2.4]. Details on the dataanalysis for STS measurements are discussed in appendix B.Figure 2.4: Schematic to visually describe how single point STS curves and spatially resolvedenergy distribution maps are related to each other and to the electronic structure of thesample.STS measurements are extremely useful for investigating the energy position of occupied and unoc-cupied states/orbitals of systems with complex electronic structures. This is because the measured19It current depends mostly on the sample’s available states to tunnel from or to (if one assumes thetip’s DOS is flat and as an infinite reservoir of electrons at any energy). Depending on the appliedbias polarity one can investigate unoccupied and occupied electronic states. Since by convention thesample bias is reported relative to the tip, the negative bias range in STS measurements correspondsto the occupied states of the sample, while the positive range to the unoccupied ones.202.3 Experimental setupThe commercial LT-STM and AFM used to perform the STM and STS measurements was pur-chased from Omicron Nanotechnology (Germany) [85]. The setup consist of two main chambersseparated by a gate valve [figure 2.5]. Both are kept in UHV at pressures below ∼7x10−11 mbar(lower limit of the gauges). The pressure in the two separate chambers is measured independentlyvia ion gauges. In order to achieve and keep such a clean environment, a vacuum system consistingof one turbo pump backed pumped by a rough pump, two ion pumps, and two Titanium SublimationPumps (TSP) is used. Moreover, the cold cryostat in the LT-STM side serves also as a cryo-pump.After being open to atmosphere, a bake-out of the whole instrument at 150◦C for 108 hours isdone in order to attain such a low pressure. One of the chambers (preparation or prep chamber) isdedicated to in situ samples preparation while the other one, the LT-STM side, is where the mea-surements are performed. The preparation chamber is equipped with an ion gun for Ar+ sputtering,an annealing system, a leak valve to control the introduction of gasses into the chamber, an OrganicMolecular Beam Epitaxy (OMBE) cell for molecular deposition, a homebuilt thermal effusion cellfor salt evaporation, and an electron-beam (e-beam) metal evaporator for metals deposition fromrods or crucibles [figure 2.6]. Another OMBE cell is available on the LT side for low-temperaturemolecular depositions. Samples can be easily transferred from one chamber to the other using atranslational manipulator. In order to minimize the acoustic and environmental vibrations, whichstrongly affect the quality of the measurements, the instrument is located in an acoustically insu-lated room, surrounded by 25 cm thick concrete walls, with a concrete inertia block supported bypneumatic isolators that serves as the floor. Another series of isolators legs are available for theSTM itself but during the experiments discussed here they were never used since not necessary toimprove the measurements quality [figure 2.7] [86]. Since the tunneling junction is exponentiallysensitive to tip-sample distance variations, it is very important to isolate this part of the system fromany mechanical vibration. During the measurements the LT-STM stage is suspended by three softsprings [figure 2.8]. Their resonance frequency is quite low (about 2 Hz) so that the high frequencyvibrations are damped. The vibrations of the suspended system are reduced by using an eddy currentdamping mechanism.The microscope has a maximum scan range of (1.8 x 1.8) µm2 at 5 K. The tip is scannedover the sample by means of a piezo system made of a single tube that controls the motion of thetip in the x-, y-, and z- (tip-sample separation) directions [figure 2.9(a)]. For the tip movements inx and y two electrodes per each direction of movement are attached on the tip holder. They inducebending of the tube as a result of an applied voltage [figure 2.9(b)]. The tip holder is anchored tothe z-piezo tube by means of a magnet [figure 2.9(a)]. At 5 K the piezo sensitivity in x and y is 3.6nm/V, while in the z direction is 1.2 nm/V.In constant-current mode operation, the topography is measured using a feedback loopthat monitors the tunneling current and adjusts the tip height at each point (x,y) by applying a volt-age to the z-piezo (Vz) to maintain a constant current. The constant current value is pre-set to adesired tip height to maximize the signal, but reduce unintended interaction with the surface. The21Figure 2.5: Picture of the whole STM system showing the main consisting parts.Vz(x,y) signal is recorded and converted to a distance, generating a topographic map. An I/V cur-rent pre-amplifier located just outside the LT-STM chamber, amplifies the very low tunneling currentand converts it into a voltage signal. The I/V converter features two ranges (3nA/330nA).22Figure 2.6: Picture of the preparation chamber of the Omicron system with all its main featuresshown. The leak valve and the homebuilt thermal effusion cell are not visible from thisangle.Figure 2.7: Schematic of the acoustical and environmental isolations of the Omicron STMsystem. Figure reproduced from [86].23Figure 2.8: Detailed image of the STM head with the springs suspension system. Figure re-produced from [87].24Figure 2.9: Piezos system. (a) Schematic of the single scanner tube. (b) Schematic of thepiezo scanner showing the electronic connections for the x-, y- and z-motions. Figuresreproduced from [87].25Substrates are mounted on molybdenum sample plates [figure 2.10]. During the mea-surements they are grounded and the bias voltage is applied to the tip (although the sample bias isreported throughout). The tips used in the experiments were cut Ag-terminated platinum iridium(Pt90/Ir10) tips (Goodfellow [88]).Figure 2.10: Picture of an Ag(111) sample mounted on a molybdenum sample plate.Detailed images of the STM head are shown in figures 2.8 and 2.11(a). The STM head ispositioned below the LHe cryostat inside of two sets of copper (Cu) heat shields that protect it fromouter thermal radiation (necessary to maintain the low temperature) [figure 2.11(b)]. Each set ofheat shields has a rotating door that allows access to the sample and depositions onto cold substrateswhen all the parts are properly aligned [figure 2.11(a)]. The STM head is in thermal contact withthe cryostat and therefore they stay at the same temperature.The cryogenic system consists of two concentric cryostat baths. The outer one is filledwith Liquid Nitrogen (LN2) and serves as a thermal shield. The inner one can be filled either withLN2 or LHe and serves to keep microscope and sample cold. A silicon (Si) diode is attached to thesample stage and allows monitoring the temperature in a range from 1.4 to 500 K. During all themeasurements presented in this work, the temperature of the sample was stable at∼4.3 K. There areseveral reasons for measuring at this low temperature. First, the diffusion of adsorbates is greatlyreduced: at low temperature, adatoms and molecules are frozen at defined positions and can beatomically resolved more easily. Moreover, the tip structure is more stable. Secondly, drift of thepiezoelectric system due to thermal relaxation are reduced. And lastly, thermal broadening of theelectronic features in spectroscopy measurements can be assumed to be negligible for the scale ofmolecular states we are interested in. Since some experiments may require variations of the sampletemperature, a heater element is fitted under the sample stage together with the silicon diode temper-ature sensor. The sample temperature during the counter heating can be monitored and maintainedstable by a close-loop feedback controller. This assembly gives a very precise way to control thetemperature above the temperature of the bath at the expense of an increased consumption of thecryo-liquid. The achievable temperature depends on the cryo-liquid used (liquid He or N2), due to26Figure 2.11: (a) Picture of the STM head from outside the copper heat shield with the coldtemperature deposition port visible at the bottom left. (b) Picture of the inner copperheat shield screwed to the bottom of the LHe cryostat.different base temperature (5K and 77K, respectively) and latent heat of evaporation. With LHe, themaximum achievable stable temperature is about 100K, while with LN2 one could go above 250K.Also, taking into account the thermal drift of the piezos, and with other proper precautions, it ispossible to scan while counter heating.272.4 Samples preparation2.4.1 Metallic substrate: Ag(111) surfaceIn all the experiments done in this work the substrate used was a chemo-mechanically polishedmonocrystalline Ag(111) sample (MaTeck [89]). This is characterized by a hexagonal lattice of2.88 A˚ of lattice constant [90] [figure 2.12(a)], low chemical reactivity, small corrugation, and close-packed, threefold symmetry. Its surface also features a Shockley-type surface state at -67 mV (nearthe Fermi level) [91] [figure 2.12(b)].Figure 2.12: Characteristics of the Ag(111) surface. (a) Constant-current STM image.Ag(111) atomic resolution (Vb = -100mV, It = 100pA). Hexagonal lattice and latticeconstant are shown. (b) STS spectrum showing the silver surface states at -67 mV(solid red line) [91]. Note: The hump around -0.6V is due to the tip.Preparation. Before any deposition, the Ag(111) surface was cleaned using repeated Ar+sputtering cycles, at energies between 0.7 - 1 keV and currents of typically 3-4 µA, followed by sam-ple annealing at temperature of 790 K for about 30 minutes. Constant-current STM images of thebare clean surface showed a low concentration of water molecules (0.78± 0.32 molecules/100nm2)and a negligible amount of CO. These two types of contaminants can be easily distinguished in STMimages because up to -93mV water molecules are imaged as protrusions [92] while CO is imagedas a depression at every bias.282.4.2 NaCl/Ag(111) substrateSalt (NaCl) is commonly used in STM studies as an ultra-thin insulating film to reduce the electronicinteraction of adsorbates with underlying metal substrates [66] and also to mimic a more bulk-likebehavior, while still allowing tunneling, and therefore STM and STS measurements. NaCl can beeasily grown on top of different metallic surfaces (e.g. Ag(111) [93], Cu(111) [66], Ag(100) [94]).On Ag(111), NaCl forms (100)-terminated islands that can extend up to several microns in diameter[95]. The first stable layer is the bilayer with non-polar edges, where sodium (Na) and chlorine(Cl) atoms alternate. The silver steps act as nucleation sites for the first and the following layers,and from there islands grow symmetrically on both sites of the steps [94]. The face-centered cubicstructure of a salt crystal with a lattice constant of 5.64 A˚ is shown in figure 2.13(a). Because of thelattice mismatch with the silver substrate a Moire´ pattern can be observed as in figure 2.13(b) [96].Figure 2.13(b) and (d) are constant-current STM images of the NaCl/Ag(111) substrate used in theexperiments, while panel (c) shows the salt characteristic NaCl on Ag(111) interface state with anonset of ∼96mV [97].Figure 2.13: Characteristics of NaCl and NaCl/Ag(111) surface. (a) NaCl 3-dimensional crys-tal structure. Figure reproduced from [98]. (b) and (d) Constant-current topographicimages of NaCl grown on top of a clean Ag(111) surface. (b) shows the Moire´ patternof salt on Ag(111) (Vb=2V, It=30pA). (d) NaCl islands formed on top of the Ag(111)surface. Bi and trilayers are visible (Vb=300mV, It=40pA). (c) Plot showing the dI/dVof NaCl on Ag(111) characterized by the interface state about +96mV (solid red line).29Preparation. The growth of a few monolayers (mostly bilayers but also some trilayers)of salt on top of the cleaned Ag(111) surface was done with the substrate at a temperature between333 K and 353 K. The NaCl was evaporated from the homebuilt thermal effusion cell heated at 813K. The deposition time was of about 5 minutes in order to get a salt bilayers coverage of about 40%.2.4.3 Molecular deposition: terpyridine-phenyl-phenyl-terpyridineThe Terpyridine-Phenyl-Phenyl-Terpyridine (TPPT) molecules (HetCat Switzerland [99]) were sub-limated from a glass crucible of an OMBE cell (Kentax [100]), heated to a temperature of 550 K,onto the bare Ag(111) [figure 2.14(a)] or NaCl/Ag(111) surfaces at either room temperature or∼4.3K, for about 2 minutes. Typical deposition rate was of ∼ 4 x 10−4 molecules/(nm2 ∗ s). Thisvalue was determined from statistical counting of the TPPT molecules in large-scale images.2.4.4 Metal deposition: ironIron (Fe) adatoms were evaporated onto the TPPT/Ag(111) and TPPT/NaCl/Ag(111) systems from apure (99.99+ purity) iron rod (2mm in diameter) purchased from Goodfellow [88] using an electron-beam metal evaporator (Omicron Nanotechnology [85]). When the iron deposition was subsequentto the molecular one, its duration (of the order of few minutes) depended on the number of TPPTsobserved on the surface. In particular, the evaporation time was determined by the number ofterpyridine (tpy) groups to be coordinated (1 TPPT = 2 tpy groups). A 1:1 (1tpy : 1Fe) ratio wasaimed for. The Fe deposition rate used as a reference was ∼ 4 x 10−4 Fe atoms/(nm2 ∗ s). Thiswas obtained with a voltage of 800V and a flux current of ∼7.8 nA and it was determined fromstatistical counts after 4K deposition in large-scale images. The flux current gives a reasonablyprecise and consistent measurement of the deposition rate. This allowed estimation of the depositionrate and therefore deposition time for different sample preparations. For depositions performed inthe preparation chamber, the rate was scaled by the appropriate geometrical factor to account for adifferent distance. In those sample preparations, where the metal was deposited at the same time asthe TPPTs, the duration of the metal deposition was determined according to previous experiments’results. The substrate was held at either room temperature or at ∼4.3 K [figure 2.14(b)] duringdeposition depending on the experiment.30Figure 2.14: STM constant-current images. (a) TPPT molecules on Ag(111) deposited atroom temperature and imaged at ∼4.3K (5 x 5 nm2, Vbias = 200mV, It = 50pA). (b)TPPT molecules and iron adatoms (dots - cyan arrow) on Ag(111). TPPT deposited atroom temperature, Fe at ∼4.3K, and imaged at ∼4.3K (5 x 5 nm2, Vbias = 500mV, It =10pA).31Chapter 3Bare Ligand: Morphology andElectronic StructureThe next three chapters are dedicated to presenting the experimental data and their analyses. Theoverall goal is to investigate the structural and electronic properties of each component (bare ligands,single coordinated ligands, molecules at the edge and inside chains, and whole chains) of the self-assembled linear nanochains consisting of bis-terpyridine ligands coordinated with Fe. From this,we determine whether or not we obtained a metal-organic complex with the desired properties,characteristics, and functionalities.This chapter focuses on describing and discussing the morphology and the electronicstructure of the Terpyridine-Phenyl-Phenyl-Terpyridine (TPPT) ligands, which form the organicpart of the metal-organic nano-structures presented later in this thesis. These molecules under studywere deposited onto two different substrates: a bare Ag(111) surface and salt islands (bilayersand trilayers) grown on top of the Ag(111). This second type of substrate was chosen in order toinvestigate the properties of the bare ligand electronically decoupled from the metal and understandwhat effects have the interaction with the silver substrate on the TPPT molecules. Morphology of theligand adsorbed on the bare Ag(111) (section 3.1) and on the NaCl/Ag(111) (section 3.2) substratesis discussed first and then their electronic structure (section 3.3 for the bare Ag and section 3.4for salt on silver) is analyzed. The last section (3.5) is dedicated to the comparison of the resultsobtained on the ligand adsorbed on Ag(111) versus on salt islands. Hereinafter, most of the timesthe Ag(111) substrate will be referred to as silver, Ag or bare Ag, while NaCl/Ag(111) as salt.3.1 Morphology of TPPT ligands on Ag(111)The chemical structure of a bare TPPT ligand in gas-phase is shown in figure 3.1(a). This is char-acterized by two terpyridine1 groups at the two ends of the molecule connected together by two1Three pyridine rings connected together.32phenyl2 (ph) rings. In gas-phase, the nitrogen (N) atoms of the peripheral pyridines (pyr) pointtowards the exterior of the molecule due to N-N electrostatic repulsion between them and the oneof the middle ring [figure 3.1(a)] [101].The deposition of the TPPTs onto the clean Ag surface wasalways performed with the substrate at room temperature (RT), while during the STM measure-ments the sample was at ∼4.3K. In constant-current mode TPPT molecules are generally imagedas dog-bone shape features as shown in figure 3.1(b). They adsorb planar to the surface becauseFigure 3.1: (a) Chemical structure of a terpyridine-phenyl-phenyl-terpyridine molecule in gas-phase. (b) STM constant-current mode topographic image of a bare molecule on Ag(111)with superimposed chemical structure [Carbon (C) atoms are turquoise, nitrogen (N)blue, and hydrogen (H) white], (Vb = 10mV, It = 1nA). Red arrow points towards thesmall depression close by N atom of the peripheral pyridine. (c) DFT-simulated side-view images of a single TPPT molecule adsorbed on the Ag substrate along the mostfavorable direction [1, -1, 0]. It shows a small twist angle between the two phenyl groupsin the middle of the molecule (DFT calculation performed by Chenguang Wang). (d)Apparent height intensity profile along the solid blue line shown in the inset at the top ofthe plot. Red arrows are pointing to the two depressions position close to the top outerpyr groups.of a relatively strong interaction between the underlying noble metal support and the phenyl [102]2A phenyl group or ring (chemical formula C6H5) is a cyclic group of six carbon atoms linked together in a hexagonalplanar geometry.33and pyridine [103, 104] groups, similar to other tpy containing molecules [105, 106]. However,there is a small twist angle between the two phenyls in the middle of the molecule, as one can seein the side-view Density Functional Theory (DFT)-simulated image of figure 3.1(c)3. Close to theouter pyr rings of the tpy groups [red arrows in figure 3.1(b) and insert 3.1(d)] small depressions areobserved. These are indicative of a reduction in the silver local density of states likely caused by anelectrostatic repulsion of the electrons on the silver surface due to the N atoms. The pyr orientationof bare TPPT molecules adsorbed on the Ag substrate seems then to be consistent with the gas-phaseone [101]. This statement is not always true; generally, gas-phase and on-surface configurations candiffer because of the two-dimensional constraint due to the substrate and the interaction with it.Figure 3.2 shows isolated molecules imaged at different biases. Between -2.5V and+500mV [panels (a) - (e)] the ligand maintains the characteristic dog-bone shape and the inten-sity is overall quite homogenous. The main difference observed is the width of the molecular axisthat is notably smaller at low positive bias [panels (d) and (e)]. Moreover, at low bias [panels (d)and (e)] a small asymmetry along the molecular axis is visible [green arrows in panel (d) and (e)],probably due to the twisting of the two phenyl rings seen in the result of the DFT simulation [figure3.1(c)]. At +1.3V [panel (f)] the dog-bone shape is still visible but the molecular axis is considerablylarger and brighter (than the tpy groups), indicating a higher number of states to tunnel through inthat region. At +2V [panel (g)], the intensity on the molecular axis is slightly higher than at the ex-tremities of the ligand. The imaging at +2.7V [panel (h)] is very different. Here, the molecule doesnot resemble a dog-bone anymore but it looks more like an elongated oval and the characteristic“>” and “<”-shapes of the tpy groups are no longer distinguishable.Figure 3.2: Constant-current STM images of different isolated molecules on Ag imaged atdifferent bias. (a) Vb= -2.5V, It=25pA. (b) Vb= -1V, It=50pA. (c) Vb= -700mV, It=1nA.(d) Vb= +9mV, It=1nA. (e) Vb= +500mV, It=1nA. (f) Vb= +1.3V, It=1nA. (g) Vb= +2V,It=1nA. (h) Vb= +2.7V, It=50pA. Images dimension: 3.8 x 3.6 nm2. Green arrows inpanel (d) and (e) point to the small asymmetry visible along the molecular axis.3DFT calculation performed by Chenguang Wang. Details of the formalism are reported in appendix A.34Atomic resolution images of the Ag alongside isolated TPPT molecules [figure 3.3(b)]show that ligands’ adsorption occurs with their long axis oriented at an angle of +(16◦±1◦) and-(16◦±1◦) off the [1, −1, 0] direction of the Ag surface [figure 3.3(b), (c) and (d)]. Given the three-fold symmetry of the silver substrate six geometrically equivalent orientations are expected andobserved in large-scale images [figure 3.3(a)]. This well-defined adsorption geometry indicates arelatively strong molecule-substrate interaction, as expected for pyr containing molecules on noblemetals [105, 106].Figure 3.3: (a) Constant-current STM image of a low coverage RT deposition of TPPTs onAg(111) (Vb = 200mV, It = 50pA). Molecules adsorb following a∼± 16◦ (red and yellowdashed ellipses) angle with respect to the <1, -1, 0> crystallographic directions (whitearrow), resulting in six equivalent orientations (labels 1, 2 and 3). (b) Constant-currentSTM image of an isolated molecule with atomic resolution of the Ag(111) substrate(scanning parameters on the molecule: Vb = -200mV, It = 10pA; on the Ag(111): Vb =-10mV, It = 300nA). The molecule’s orientation with respect to the [1, -1, 0] direction isshown. (c) and (d) schematics of the clockwise (∼-16◦) and counter-clockwise (∼+16◦)orientations of a TPPT with respect to the [1, -1, 0] silver axis.Under low coverage conditions (much less than one monolayer) TPPT molecules areobserved either by themselves or adjacently bonded following staggered rows or zigzag patterns[figures 3.3(a) and 3.4(a)]. The intermolecular origin of this lateral packing can be explained by anattractive proton acceptor/organic ring interaction [107, 108] between the N of the peripheral pyri-35dine groups of one molecule and the hydrogen (H) of the adjacent one [red dashed circles in figure3.4(a)]. This provides further evidence supporting the outward position of the nitrogen atoms ofthe outer pyr rings. When the density of the molecules on the surface increases the staggered rowsand zigzag patterns start growing into islands formed by well-ordered dog-bone shape tiles [figure3.4(b)]. The growth begins with the addition of extra molecules to the previously formed TPPTmotifs [see top right corner of figure 3.4(b): an extra molecule (green dashed circle) is added to thezigzag pattern]. The ligand’s arrangement within these large ordered domains seems to be drivenby the same attractive proton acceptor/organic ring interaction observed in low coverage conditionsbetween adjacently bound TPPTs [figure 3.3(a)].Figure 3.4: (a) Constant-current STM image of two bare molecules on Ag(111) in the adja-cent configuration (Vb = -10mV, It = 1nA). The red dashed circles represent the attrac-tive proton acceptor/organic ring interaction between the two molecules [107, 108]. (b)Constant-current STM image of high coverage TPPTs RT deposition on Ag(111) (Vb= -500mV, It = 20pA). Green dashed circle: extra molecule in addition to the zigzagpattern.The bare terpyridine groups of these molecules and their conformation on the silver sub-strate are observed to play a key role in both the molecular adsorption configuration and in theinteraction with other ligands and with the substrate. As anticipated, to further investigate the effectof the metal substrate on the TPPT molecules the same ligands were deposited onto a few mono-layers of salt grown on top of the Ag to obtain at least some electronic decoupling between the two(see next section).363.2 Ligands morphology on NaCl/Ag(111)To study the morphology and electronic structure of TPPT molecules at least partially electronicallydecoupled from the silver substrate a few (two/three) insulating layers of salt (NaCl) were grown ontop of the metal. Ultrathin insulating layers such as NaCl are used to reduce the electronic interactionof molecules with the underlying metal substrate [66] and also to mimic a more bulk-like behavior,while still allowing tunneling, and therefore STM and STS measurements.The NaCl/Ag(111) surface was prepared as described in section 2.4.2. The moleculardeposition was first tried with the sample at RT. Constant-current STM images after depositionshow no molecules adsorbed on the salt islands, while they are found to be well-aligned and orderedon the bare silver areas [figure 3.5(a)]. This indicates that at RT the interaction between TPPTsand salt is not strong enough for adsorption to occur. The subsequent depositions were performedwith the substrate at ∼4.3K [figure 3.5(b)]. At this low temperature, the mobility of the adsorbatesis reduced and the molecules remain on the salt. However, they are observed to be quite unstablesince they can be easily disturbed by the scanning tip. When the tip was positioned over the middleof the long axis (between the two phenyl groups) or over the outer pyridine rings, the moleculeswere found to be least stable. Scanning in constant-current mode with high values of Vbias andIt could easily induce twists of the phenyl groups or undesired movements and position changesof the molecules. Similarly, these effects were observed after taking point spectroscopy or gridmeasurements. In order to avoid this, the scanning parameters used were It between 2 and 10 pAand Vbias between ± 500mV. An example of a distorted molecule is shown in figure 3.6, where theligand shows a clear twist in its center (green arrow) that was accidentally induced by the tip duringa spectroscopy measure.Figure 3.5: Constant-current STM images. (a) After RT deposition of TPPT molecules theNaCl island appears free of adsorbates, while all the ligands are ordered in a closed-packed manner on the Ag(111) patch (Vb = 300mV, It = 40pA). Note: this image wasrotated of 90◦. The vertical lines in the left half of the image are caused by a series ofsubsequent tip changes occurring during the scan. In the Ag area each of the obliquebright segment is a single TPPT molecule. Upon closer inspection they appear to bearranged in their characteristic adjacent way driven by the attractive proton acceptor/or-ganic ring interaction. (b) After 4K deposition of the TPPT the ligands are imaged to beequally distributed on both silver and salt (Vb=500mV, It=1pA).37Figure 3.6: Constant-current STM image of an isolated molecule on NaCl/Ag(111) with themiddle of the molecular axis twisted as a result of a point spectroscopy measure taken inthat location (Vb = -1V, It = 10pA). Image dimension: 3.6 x 2.5 nm2.As on bare Ag, on salt TPPT molecules seem to adsorb planar to the surface but hereno specific orientations were found. This observation results from the comparison of the ligands’adsorption directions on salt with, first, the orientations followed by the molecules adsorbed onthe bare silver patches, and after by looking for new, common directions (not found) followed bythe TPPTs on salt. The absence of specific adsorption orientations can be explained by the weakadsorption of the TPPTs on salt or/and by the limited diffusion of the molecules due to the lowtemperature (∼4.3K) of the substrate during the deposition. At this temperature TPPTs are notexpected to diffuse [109, 110].Figure 3.7 shows different single ligands at different biases. Up to +500mV [panels (a)-(d)] TPPTs show the characteristic dog-bone shape with more intensity on the tpy groups below-500mV [panels (a) and (b)], while above there is a quite homogenous brightness over the wholemolecule [panels (c) and (d)]. At and above ∼+1.7V [panels (e) to (g)] the ligand is imaged asa large oval with two “>”-shape features at the left (>) and right (<) side of this. As the biasincreases the oval shape becomes larger and the two “>”-shape features are more prominent. Theshape of a TPPT on salt above ∼1.7V [figure 3.7(e)-(g)] looks quite different from the one on silver[figure 3.2(g) and (h)]. This is a first indication that the electronic structure of a single bare ligandis different in the two cases (silver and salt).Generally, molecular clusters are more stable than isolated molecules alone. This in-creased stability allows for performing STM and STS measurements with more freedom in thechoice of the scanning parameters to improve the signal to noise ratio and also investigate the elec-tronic structure at higher energies. A higher molecular stability was observed already in TPPTsseparated by very small distances (less than one nanometer). Examples of that are panels (e)-(g)of figure 3.7, in which those “high” bias images were possible due to the proximity of anothermolecule, which is partially visible at the top right corner. In order to form TPPT aggregates oneneeds to induce some molecular mobility and thus provide the chance of clustering from the meet-ing of two or more ligands. To do that, counter heating experiments were carried out. In theseexperiments, the substrate was heated up using the heating element underneath the sample stage38Figure 3.7: Constant-current STM images of different isolated molecules on NaCl/Ag(111) atdifferent scanning parameters. Specifically, the bias voltage is indicated on each paneland the tunneling current values are: (a) and (b) 2pA, (c) 5pA, (d) 25pA, (e) and (f)20pA, and (g) 30pA. Images dimension: 5 x 4.8 nm2.and held constant for a given period of time. Different trials at various temperatures (15K, 25K and40K) were performed each for 30 minutes. STM images taken after each annealing process did notshow any molecular clustering, as was expected in case of a significant diffusion [figure 3.8 - 40Kannealing]. This means that even though the molecule-salt interaction is quite weak, annealing at40K is not sufficient to induce molecular diffusion on salt.Figure 3.8: Large-area constant-current STM images. (a) TPPT deposited at 4K on Na-Cl/Ag(111) substrate adsorbed on both Ag(111) and NaCl (Vb=500mV, It=1pA). (b)TPPT molecules arrangement on NaCl after half an hour annealing at 40K (Vb=500mV,It=1pA).Despite the limited resolution due to the low stability of the molecules a large amountof information about the morphology of TPPTs on salt was obtained from STM measurements. Inparticular, the topographic images at different biases have revealed that TPPTs on salt have signif-icantly different electronic structure than when adsorbed on bare silver. This aspect was furtherinvestigated with STS measurements and the results are presented in section 3.4.393.3 Electronic structure of bare ligands on silverIn this section the electronic structure of a single ligand, as well as clusters of two and three ad-jacently bonded TPPT’s on bare Ag(111), are discussed. The data presented were obtained fromsingle point spectroscopy at different locations on the molecules or from grid measurements. Ifnot otherwise specified, spectroscopy plots represent the normalized dI/dV obtained from a gridmeasurements by averaging over the regions enclosed by the same colour curves shown in the to-pographies on top of the plots. Thanks to the normalization (equation 2.8) the influence of theset-point is minimized and therefore different normalized tunneling conductances ((dI/dV )/(I/V ))can be compared. In every figure showing the normalized dI/dV curves the dashed black one (greyin the case on salt, figure 3.16(a)) corresponds to the substrate spectrum. This is considered thereference spectrum and therefore deviations from it in the other curves of the same plot must beinterpreted as being the result of proper features of the considered adsorbate/adsorbates. This isvalid only if the spectra were all taken with the same tip. Since around zero bias all the normalizeddI/dV curves present a divergence (a byproduct of the normalization of the data (equation 2.8)) thatregion is omitted in the graphs. The STS maps represent the spatially resolved electronic distribu-tions. Overlaid on each of them is a dashed contour of the molecule considered, obtained from thecorresponding topography images at the set-point bias voltage of the grid. Additional details aboutthe data analysis are reported in appendix B.3.3.1 Single ligandFigure 3.9 shows STS on a single TPPT molecule. Two spectroscopy curves are presented: the redone encompassing the whole molecule and the blue one on the long molecular axis. Both show twostrong peaks at +1.15 V and +2.3 V corresponding to two tunneling resonances relative to unoccu-pied states. In the negative bias range4, even though several humps are present, no obvious featuresrelated to occupied orbitals are visible. These humps are also present in the reference spectrum andthis means that they are due to the tip LDOS5. The absence of any visible feature at negative biascould be explained by a probable strong hybridization between the molecule and the silver substrateor the absence of molecular states over the range of energies probed.4For convention, the negative bias range in STS plots and maps shows the occupied states of the sample (below theFermi level), while the positive one, the unoccupied states (above Fermi).5Ideally the Ag(111) spectrum should be practically flat with only the silver states onset around -67meV. Divergencesfrom this ideal behavior are mainly due to intrinsic electronic states of the tip which, in reality, never has a featurelessspectrum as simplified theories assume. Moreover, even after normalization, the exponential background observed at bothnegative and positive bias is due to the transmission factor introduced in equation (2.5).40Figure 3.9: Normalized dI/dV curves of an isolated bare molecule obtained by averaging overthe regions enclosed by the corresponding colour outlines in the topography (see in-set). Red: whole single molecule; blue: single molecule molecular axis; dashed black:Ag(111), reference spectrum. Grid set-point parameters: Vbias = -2.50V, It = 50pA. Note:the divergence near zero-bias created by normalization is removed.41The two negative bias STS maps in figure 3.10 at -1.9V (a) and -1V (b) show distinctintensity localized on the molecule. This might indicate states associated with the ligand at theseenergies that are not visible in the STS curves, or it might be the result of tunneling matrix effects.At −1V (b) the intensity is homogeneously distributed on the ligand, which could be indication ofa molecule-metal hybrid state or an interface state [111]. More interesting features are observedat positive energies. At +0.4V (c) the higher intensity of the normalized dI/dV is spread over theentire molecule, while around +0.7V (d) it is concentrated mainly on the tpy groups [figure 3.10]. Atthe energy corresponding to the first tunneling resonance (+1.15V, (e)), it is localized in the middleof the molecule on two very closed but clearly separate lobes. These are observed to be symmetricabout an axis tilted of ∼+16◦ with respect to the horizontal. In other grid measurements of singlemolecules the tilt was found to be ∼-16◦ [figure 3.11(b)]. These inclination angles are the sameas the ones found for the ligand’s orientations with respect to the silver axis. Therefore, a possibleexplanation for this breaking of the symmetry could be the molecule’s adsorption on the metal or itcould also be partially due to the small twist angle of the two ph rings. In the bias range between+1.6V and +2V the higher intensity is found to be mainly around the tpy groups and less (at +1.6V,(f) and +2V, (h)) or none (at +1.8V, (g)) in the center of the ligand. At +2.3V (i), which is the energyof the second visible resonance, this is at the far edges of the ligand as well as in the middle of themolecular axis. Lastly, at +2.7V (j) the intensity pattern resembles the shape of a 6-point star withmore intense regions sitting on its points. Often, and not only in this case, the higher intensity isfound to be outside the molecules’ overlaid contour. A possible reason for this could be that theobserved electronic states are not only molecular states but also molecular-substrate states resultingfrom a probable hybridization of the orbitals. To see the entire data set on the single bare ligand, seeaccompanying video 1: Video 1 - bare ligand.42Figure 3.10: STS energy maps at different biases showing the density of states of a single bareligand. From (a) to (j) the biases are: -1.9V, -1V, +0.4V, +0.7V, +1.15V, +1.6V, +1.8V,+2.0V, 2.3V and +2.7V. Grid set-point parameters: Vbias = -2.50V, It = 50pA.43Figure 3.11: STS energy maps at Vbias = 1.15V of two different single bare ligand on Ag(111)showing the opposite tilt angle of the axis of symmetry of the two lobes in the middleof the molecule [(a) ∼+16◦, (b) ∼-16◦]. Grids set-point parameters: (a) Vbias = -2.50V,It = 50pA; (b) Vbias = -2.50V, It = 50pA.443.3.2 Two adjacently bound moleculesFigure 3.12 shows the normalized dI/dV of two molecules adjacently bound via attractive protonacceptor/organic ring interaction. As in the single molecule case, no specific features are visible inthe negative bias range and both the TPPTs show a molecular orbital (MO) at +1.15V. This is thesame energy of the first resonance of a single molecule [green bar in figure 3.12], implying that theadjacent binding does not affect the overall electronic structure of the two interacting ligands, i.e.that they are not electronically coupled. Moreover, this suggests that the adjacent bonding betweenTPPTs is not driven by a covalent/hybridization interaction.Figure 3.12: Normalized dI/dV curves of two adjacently bound molecules obtained by aver-aging over the regions enclosed by the corresponding colour outlines in the topography(see inset). Red: whole top single molecule; blue: top single molecule molecular axis;orange: whole bottom single molecule; cyan: bottom single molecule molecular axis;dashed black: Ag(111), reference spectrum. Grid set-point parameters: Vbias = -2.00V,It = 400pA. Note: the divergence near zero-bias created by normalization is removed.45The STS maps of figure 3.13 show basically the same intensity pattern described previ-ously for the single molecule. The only difference is in the negative energy (-0.8V versus -1V of thesingle molecule) at which some intensity is observed to be homogeneously distributed on both theTPPTs.Figure 3.13: STS energy maps at different biases showing the density of states of two bareligands adjacently bound. From (a) to (h) the biases are: -1.9V, -0.8V, +0.4V, +0.7V,+1.15V, +1.5V, +1.8V and +2V. Grid set-point parameters: Vbias = -2.00V, It = 400pA.463.3.3 Three adjacently bound moleculesThe electronic structure of three adjacently bound molecules is the same as the case discussed above.Figure 3.14 shows the STS curves of the three bound molecules (green bars: single TPPT resonancepeaks positions). In the negative bias range no peaks are visible. The two strong resonances atpositive bias for all the three TPPTs are located at ∼+1.15V and ∼+2.3V and no appreciable shiftsfrom the single molecule case are observed. This suggests, once again, that the three TPPTs areelectronically independent.Figure 3.14: Normalized dI/dV curves of three molecules adjacently bound obtained by aver-aging over the regions enclosed by the corresponding colour outlines in the topography(see inset). Red: whole top single molecule; blue: top single molecule molecular axis;orange: whole middle single molecule; cyan: middle single molecule molecular axis;violet: whole bottom single molecule; pink: bottom single molecule molecular axis;dashed black: Ag(111), reference spectrum. Grid set-point parameters: Vbias = -2.50V,It = 50pA. Note: the divergence near zero-bias created by normalization is removed.Figure 3.15 shows the STS maps. Here also one can observe a very similar intensity pat-tern to the ones of a single ligand or two-adjacent molecules on Ag at both negative and positive bias.The only difference is in the negative energies at which some intensity is visible on the molecules.Here, they are at -1.6V and -0.9V instead of -1.9V and -1V or -0.8V. These small differences mightbe related to slightly different molecule-substate interactions. The similarity of the energy mapsbetween the single ligand and the two or three adjacently bound ones supports the hypothesis thatthe interacting molecules are non-electronically coupled.47Figure 3.15: STS energy maps at different biases showing the density of states of three bareligands adjacently bound. From (a) to (j) the biases are: -1.6V, -0.9V, +0.4V, +0.7V,+1.15V, +1.6V, +1.8V, +2.0V, 2.3V and +2.7V. Grid set-point parameters: Vbias = -2.50V, It = 50pA.48In conclusion, TPPT molecules - even if adjacently bound to one another - do not seemto electronically couple. Also, there is evidence that suggests an important interaction betweensingle ligands and the metal. Therefore, in order to further investigate this aspect, spectroscopymeasurements on bare ligand on salt were performed.3.4 Electronic structure of bare ligands on NaCl/Ag(111)Despite the low stability of the TPPT molecules on NaCl bilayers and the elevated chances of con-formational deformations requiring extremely low tunneling currents, spectroscopic measurementswere nevertheless performed on single TPPT’s and relevant information about their electronic struc-ture was obtained.Single point spectroscopy measurements were taken at different positions on the molecule.Interesting results were found from the spectroscopy at the center of the molecule [figure 3.16(a)].Due to the remarkably good agreement of the number of peaks and the energy difference betweenthe peaks at positive energies between theoretical6 and experimental results, a peak at negative biascorresponding to the Highest Occupied Molecular Orbital (HOMO) was expected around -2.3V[figure 3.16]. However, no molecular features between -3V and -0.5V were observed. It is unlikelythat this could be due to a too low tunneling transmission, and therefore the HOMO is probably at anenergy lower than -3V. In the positive energy range (between +1.4V and +3V) three clear peaks cor-responding to unoccupied Molecular Orbitals (MOs) of the molecule are visible at +1.65V (LowestUnoccupied Molecular Orbital (LUMO) - red), +2.2V (double degenerate LUMO+1 - green) and+2.9V (LUMO+2 - orange)7. The experimental energy differences between these peaks are in verygood agreement with the theoretical ones (considering that the substrate was not included in thecalculations): 0.5V versus 0.4V between the first two, and 0.7V versus 0.9V between the next two.6DFT calculations were performed by Katherine Cochrane using Gaussian on a single fully relaxed TPPT moleculein gas-phase with the bases set B3LYP/6-31g. The NaCl/Ag(111) surface was not included.7LUMO, LUMO+1, LUMO+2... are the conventional names for the first, second, third... unoccupied MOs, from lowto high energy, and each corresponding to a peak in the STS plot.49Figure 3.16: (a) Normalized dI/dV plots of negative and positive bias of an isolated moleculeadsorbed on NaCl/Ag(111). These were obtained from two different single point spec-troscopy taken in the same position on the molecule (light blue dot in the topogra-phy - insert) with different set-point parameters and bias range (negative bias set-point:Vbias = -3V, It = 2pA; positive bias set-point: Vbias = +1.40V, It = 5pA). Dashed greycurves: NaCl reference spectra. At positive energies the three peaks correspond to theLUMO (+1.65V, red bar), the double degenerate LUMO+1 (+2.2V, green bar) and tothe LUMO+2 (+2.9V, orange bar). (b) DFT-calculated normalized LDOS of a singleTPPT in gas-phase (DFT performed by Katherine Cochrane). To compare the energyposition of the peaks between experiment and theory, the energy of the first peak atpositive bias (the LUMO) in the DFT spectrum was matched in value to the first exper-imental one equal to +1.65V. Vertical bars correspond to: black, HOMO; red, LUMO;green, LUMO+1; orange, LUMO+2.50More details about the spatial distribution of the LDOS of a single molecule on NaClwere obtained from a line of STS taken along the long axis of the ligand. Twenty point spectra,evenly distributed in space, were performed at the positions (colourful dots) shown in the topog-raphy inset of figure 3.17. The resulting spectra are presented in the waterfall plot in figure 3.17.For convenience, the spectra are numbered to be referenced in this text and there is a matching inthe colour between the curve and the position (dot) on the molecule where the measure was taken.The first resonance peak (Vbias = ∼+1.7V) is observed to be higher in intensity in the middle of theTPPT than at the extremities of the ligand (curves 6 - 15). The behavior of the second peak (Vbias =∼+2.2V) is different: the intensity is higher over both the two tpy groups (curves 3 - 8 and 12 - 18),and is not evident at the center of the molecule (curve 11). From these observations we concludethat around +2.2 V electrons are mostly localized on the tpy groups, while around +1.7 V they areconcentrated in the middle of the ligand. The position of the maximum of both peaks is observedto slightly and symmetrically (with respect to the middle of the molecule - curve 11) shift in energydepending on the position. The first one moves from +1.78V (curve 4) down in energy to +1.6V(curve 11 - middle of the molecule) and then up again to +1.78V (curve 17). Curves 4 and 17 arerelative to the same symmetric locations at the two extremities of the molecule. The second peakinstead shifts over a smaller energy range from +2.21V (curves 4 and 17 - molecule’s extremities)to +2.25V (curves 9 and 12 - positions almost symmetric slightly off the middle of the molecule).51Figure 3.17: STS point spectra taken at twenty equally spaced locations along the long axis ofa bare ligand on NaCl/Ag(111). They are presented with a vertical offset and thereforethey do not share the same zero baseline. The inset on top shows the twenty positionswere the measurements were taken. Corresponding colour plots are shown below it.The numbers on the right side are referenced in the text.52Figure 3.18(b) shows the energy map obtained from a grid measurement at the energyof the LUMO peak (Vbias = +1.65V). The poor contrast is due to the very low tunneling currentneeded to not perturb the molecule. Here, the higher intensity is localized in the middle of themolecule, which is consistent to that found from the line of STS taken along the long axis of theligand [figure 3.17] and similar to what observed for TPPTs on Ag at +1.15V [energy correspondingto the first resonance peak - figure 3.10]. The asymmetry visible in the map could be due to a slightlyasymmetric tip or a possible small drift of the piezos during the measurement.Figure 3.18: (a) Topography of the grid measurement (Vb = -2V, It = 0.6pA) performed on anisolated TPPT on NaCl/Ag(111). (b) STS energy maps showing the electron distributionat +1.65V of the molecule in panel (a). Grid set-point parameters: Vb = -2.00V, It =0.6pA.In figure 3.19 the DFT-calculated electron distributions of the HOMO, the LUMO, thedouble degenerate LUMO+1, and LUMO+2 are shown8. In the HOMO and LUMO the electronicstates are localized on the long molecular axis, which is consistent with the experimental observa-tions [figures 3.17 and 3.18(b)]. In the LUMO+1, the states are on the tpy groups, as also indicatedby the line of STS measurements [figure 3.17]. Finally, in the LUMO+2 they are mainly on thetpy with some also on the ligand’s axis. Unfortunately, the limitations on the measuring parametersdid not allow obtaining the experimental energy maps corresponding to the LUMO+1 and +2 tocompare with the theory.In conclusion, the observed electronic structure of a single TPPT on NaCl has somesimilarities as well as some striking differences to a single TPPT on Ag. This will be analyzed anddiscussed in detail in the next section.3.5 Comparison between the electronic structure of ligands on bareAg(111) and on NaCl/Ag(111) substratesTo further understand the electronic structure of the TPPT molecules, here the results on silver andon salt are compared and related to the gas-phase DFT calculations.The adsorption configuration of the TPPT molecules on both types of surfaces looks quitesimilar, since in both cases they are imaged planar to the substrate. However, we do not have anyexperimental or theoretical evidence to state something about a possible twist of the two ph groups8DFT performed by Katherine Cochrane.53Figure 3.19: Gas-phase DFT-calculated electron distributions maps of the HOMO, the LUMO,the double degenerate LUMO+1 and LUMO+2 of a single TPPT molecule. Isosurfacelevel 0.02 electrons/r3Bohr. (DFT performed by Katherine Cochrane).on salt. The adsorbate-substrate interaction on silver appears to be quite strong, as indicated by thepreferential adsorption orientations [figure 3.3(a)]. On the other hand, ligands on salt are observedto be quite unstable even at 4K, suggesting that the coupling here is much weaker. Moreover, therather remarkable agreement with gas-phase DFT calculations, not including the substrate, furtherindicates that the molecules are well decoupled from the metal on bilayer NaCl.The main differences between TPPTs on silver and on salt were found in the electronicstructure.Figure 3.20: Normalized dI/dV obtained from point spectroscopy taken at the center of themolecule at the position of the blue dots shown in the insets at the top of the two pan-els. (a) TPPT on Ag(111) (set-point parameters: Vbias = 0.60V, It = 50pA), (b) onNaCl/Ag(111) (set-point parameters: Vbias = 1.40V, It = 5pA).54Figure 3.20 shows the normalized dI/dV relative to the center of the molecular axis ofan isolated TPPT adsorbed on bare silver (a) and on salt (b). Both the number of peaks and theirenergy positions are different. For the TPPT on Ag only two peaks, separated in energy of 1.15V, areobserved in the bias range between 0.6V and 2.6V. On salt between 1.4V and 3V there are insteadthree peaks, spaced in energy of 0.55V (between the first and the second) and 0.70V (between thesecond and the third).Both the energy maps corresponding to the first resonance (at +1.15V in the case ofsilver, and at +1.65V in the one on salt) show electronic states well-localized in the middle of themolecule [figure 3.21]. The only clear difference is that on salt [panel (b)] the electrons are localizedon a single circular area in the middle of the molecule, while on Ag [panel (a)] the same circulararea seems to be divided into two distinct lobes. The similarity in the distribution of the electronicstates suggests that these two resonances correspond to the same MO. Moreover, since the DFT-calculated energy distribution of the LUMO for an isolated TPPT also presents electrons states onthe ligand’s axis, it is very likely that both the +1.15V and +1.65V peaks correspond to the LUMOof the molecule. Most likely both the energy shift of ∼0.5V and the breaking of the symmetry inthe LDOS of a TPPT on Ag are due to the interaction between ligand and metal.Figure 3.21: STS energy maps showing the electron distributions of (a) a bare ligand onAg(111) at +1.15V (grid set-point parameters: Vbias = -2.50V, It = 50pA); and (b) onNaCl/Ag(111) at +1.65V (grid set-point parameters: Vbias = -2V, It = 0.6pA).In order to understand if the +2.3V peak of the silver case corresponds to the +2.2Vor the +2.9V peak of the salt case, intensity profiles of the +2.3V (silver) and +2.2V (salt) peaksas a function of position along a straight line going through the middle of the molecule in thedirection of its long axis, were examined [figure 3.22]. The intensity profile of the +2.2V peakon NaCl [panel (b)] never goes to zero and shows two symmetric humps about the middle of themolecule. Panel (a), which corresponds to the intensity profile of the +2.3V peak on Ag(111), showszero intensity at the edges of the molecule and in two other regions symmetric about the middle.Between the zero-intensity parts, the profile shows three high-intensity humps. It seems then clearthat these two profiles do not represent the same spatial electronic distribution and therefore thehypothesis is that these two tunneling resonances do not correspond to the same MO. It is thenpossible that the +2.3V peak on Ag corresponds to the +2.9V one on NaCl. This issue cannot55Figure 3.22: Plots of peak intensity profiles as a function of position along the lines shown bythe dashed cyan arrows in the insets at the top of the panels. (a) is relative to the +2.3Vpeak of a TPPT on Ag(111) and (b) the +2.2V on NaCl/Ag(111).be experimentally proven because the +2.9V peak was inaccessible due to the instability of themolecules when probed at this bias. However, DFT-calculated LDOS of the LUMO+2 [figure 3.19]seems to support this assumption, as follows. The intensity profile that can be obtained by followingthe electron distribution along a straight line going through the center of the molecule in the directionof the long axis looks very similar to the one shown in figure 3.22(a): two higher humps at the edges(on the tpy groups) and a lower one in the middle of the ligand. Additional evidence comes fromthe energy difference between the peaks: the ∆V = 1.25V between the +1.65V and the +2.9Vpeaks on salt is much closer to ∆V = 1.15V between the +1.15V and the +2.3V peaks on Ag thanthe ∆V = 0.55V between the +1.65V and the +2.2V on salt. From a comparison of the intensityprofile behavior of +2.2V peak on NaCl [figure 3.21(b)] with those obtainable in the same way(i.e. considering the intensity profile along the same direction) from the energy maps of a bareligand on Ag [see figure 3.10] it seems that the most similar one is at the energy of +1.8V [figure3.23]. If that is the case, the energy difference from the first resonance at +1.15V is 0.65V, whichis very close to the 0.55V found on NaCl. In this scenario one could then assume that the MOcorresponding to the +2.2V peak hybridizes with the silver substrate and is substantially broadenedin energy and therefore not visible in the point spectroscopy. Further, this state corresponds toelectron density on the tpy groups [figure 3.19], which are expected to interact strongly with metalsurfaces [103, 104] and are therefore the most likely to show strong hybridization with the silversubstrate. As a consequence of all these observation, it seems quite clear that the Ag(111) substratehas important effects on the morphology and the electronic structure of the TPPT molecules.3.6 ConclusionIn this chapter, the morphology and the electronic structure of TPPT ligands were described and dis-cussed. The dog-bone shaped molecules were found to interact strongly with the Ag(111) substrate56Figure 3.23: Plot of the intensity profile as a function of position along the line shown by thedashed green arrow in the inset. The profile was obtained from the STS map (inset) ofan isolated bare TPPT on Ag(111) at +1.8V.via their terpyridine groups. This aspect was deduced from the observation of specific orientationsof the molecules on the bare Ag and also from a quite different electronic structure than when theyare electronically decoupled from the metal via the NaCl layers. The interaction with the silver mostlikely produces a shift of ∼0.5V towards lower energies of the LUMO of the molecule. Moreover,it seems very likely that a hybridization of at least one unoccupied molecular orbitals of the ligandwith those of the metal occurs (the +2.2V peak present in the spectrum for salt but not in Ag).The presence of small depressions close to the outer pyr rings of the tpy groups andthe adjacent lateral binding between ligands via an attractive proton acceptor/organic ring interac-tion supports the hypothesis of an outward-facing orientation of nitrogen atoms of the outer pyridinerings. The N-H interactions between adjacent molecules appear to determine the molecular arrange-ment on the surface under high-coverage conditions but to not lead to electronic coupling betweenthem.57Chapter 4Discussion of Chain Formation ProcessThe main focus of this work was to fabricate and characterize the structure and electronic proper-ties of metal-ligand structures on a solid support that present properties sought for photovoltaic andcatalysis applications. In the previous chapter, the organic part of the metal-organic complex (bareTPPT molecules) in study was investigated and its characteristics analyzed in detail. This chapteris dedicated to the description and discussion of the on-surface synthesis and morphology of lineariron-TPPT nanochains. Upon addition of iron (Fe) atoms onto the TPPT/Ag(111) system (describedin the previous chapter), thermally activated self-assembled nanostructures are formed. A generaldescription of their formation is described in section 4.1. Subsequently, a detailed study of the mor-phology of single coordinated molecules and chains is presented (sections 4.2 and 4.3, respectively),followed by a statistical analysis of the chains’ distribution length, and the effects of the annealingon the sample (section 4.4). Section 4.5 describes a possible model for the chain formation process.Characteristics of sample preparations under different conditions are discussed in section 4.6, whilesection 4.7 describes a preliminary attempt to form chains on NaCl/Ag(111).Sections 4.1 - 4.5 of this chapter are based on text from: Martina Capsoni, AgustinSchiffrin, Chen-Guang Wang, Tanya S. Roussy, Katherine A. Cochrane, Adam Q. Shaw, Wei Jiand Sarah A. Burke “On-surface synthesis of iron terpyridine nanochains featuring a linear tri-ironlinkage” (in submission) [112]. All the DFT calculations and DFT images presented in this chapterwere performed by Chenguang Wang.4.1 Iron-TPPT chain formation: generalTo form the Fe-TPPT nano-chains, the bare ligand and the iron adatoms were sequentially depositedon the clean Ag(111) surface, in that order. After room temperature deposition of the molecules, Fewas deposited by e-beam evaporation with the substrate held either at ∼4.3K or at RT. In the firstscenario the metal atoms, imaged as bright dots in figure 4.1(a) (see purple arrows), are observed tobe randomly distributed on the surface and do not appear to perturb the preferential orientation andmotifs of the molecules seen before their addition [figure 3.3(a)]. At this low temperature, transitionmetal atoms [58, 113], as well as nitrogen-containing aromatic molecules [109] are immobile on58close-packed metal surfaces so it is not surprising that no interaction occurs. After annealing at 323KSTM images showed some of the atomic and molecular species arranged into linear nanostructures,where TPPT molecules no longer bind laterally to each other but instead in a head-to-head motif[figure 4.1(b)]. STM images taken after RT Fe deposition showed similar linear nanostructures[figure 4.1(c)] indicating that the energy available at room temperature provides sufficient mobilityto form chains. In both scenarios (annealing after 4K Fe deposition and after RT Fe deposition)chains were observed to be not very long and possess a relatively high number of defects suchas iron clusters at the coordination sites [blue arrows in figure 4.1(c)]. This was true even if theamount of Fe was tuned in order to have roughly one Fe per bare tpy groups (two Fe each TPPT1).The diffusion of both atoms and molecules on the silver substrate at RT deduced by the formationof chains, is supported by previous work [58, 109, 110, 113] where it was also found that the Femobility [58, 113] is at least an order of magnitude larger than for nitrogen-containing aromaticmolecules [109, 110]. This high mobility could explain the presence of iron clusters, since thechance of Fe-Fe interaction would be higher than the Fe-tpy one. It was also observed that theannealing at 373K of samples, where chains were already formed, resulted in an increase of theaverage chain length [figure 4.1(d)]. This matter will be discussed in section 4.4.From a careful analysis of the chain orientations, performed by taking the bare ligandsthat are always present in small numbers in large-scale images as a reference [figure 4.1(b)-(d)], itwas found that they do not follow any specific orientation (neither corresponding to the six orienta-tions observed for the bare ligand, nor another consistent orientation). Additionally, some of themwere even observed to be not completely straight. Therefore, the coordinated molecule-substrateinteraction seems clearly to be strongly suppressed. This loss in preferential orientation is probablydue to the tpy groups of the molecules now actively involved in the coordination bond and so nolonger interacting with the silver.As the head-to-head binding between TPPTs was not observed in the TPPT/Ag(111)system without Fe, nor in the Fe/TPPT/Ag(111) system upon metal deposition at ∼4.3K, one canconclude that chain formation results from the interaction between the diffusing Fe adatoms and theorganic species. Details about the coordination, the morphology of the coordinated structures, andtheir formation process are discussed in the following sections.4.2 Morphology of single Fe-coordinated moleculesIn large-scale images, together with linear chains of TPPTs, single molecules terminated on one orboth ends by Fe were observed. These are referred to as “1-molecule chains” and are consideredeffectively as monomers for the growth of chains. To investigate their configuration, high-resolutionSTM images were taken [figures 4.2 - 4.4]. Figure 4.2(b) shows an isolated TPPT molecule wherethe left-hand side shows the distinct “>”-shape, as seen before Fe deposition [figure 3.1(b)], whereasthe right-hand side appears brighter and with a central protrusion. As the latter was not observed1The number of molecules was determined from statistical counts after each TPPT deposition and before the additionof Fe.59Figure 4.1: Constant-current STM images. (a) Isolated Fe adatoms (solid purple arrows) de-posited on TPPT/Ag(111) system at ∼4.3K (Vb = 500mV, It = 10pA). TPPT moleculesstill mantain the same orientation as before the addition of iron [figure 3.3(a)]. (b) Sys-tem of panel (a) after annealing at 323K for 10 minutes (Vb = 200mV, It = 50pA). (c) Fedeposited are RT on the TPPT/Ag(111) system (Vb = -1V, It= 10pA). Cyan arrows pointtowards iron clusters. (d) System of panel (c) after annealing at 373K for 10 minutes (Vb= -1V, It = 50pA).60prior to Fe deposition one deduces that this is the result of an Fe adatom that has interacted with thetpy group. This was also confirmed by DFT simulations of the STM imaging2 [figure 4.3], allowingfor identification of Fe adatoms attached to TPPT molecules [figures 4.2 - 4.4 show structural modelsidentifying Fe locations from the images]. All the tpy groups of the TPPT molecules show one ofthese two characteristics, further implying that the termination consists of either zero or one Fe atom,but not more, as this would appear distinct in STM imaging. Closer examination of the negative biasSTM images shows small depressions in the substrate [blue arrows in figure 4.2(b), zoom (c), andblue box in figure 4.4(d)] adjacent to the outer pyridine rings of the bare tpy groups, likely causedby the repulsion of the silver electrons due to outward pointing N atoms, as discussed in section 3.1.Conversely, these depressions are not observed where Fe is present [green arrows in figure 4.2(b),zoom (d), and green box in figure 4.4(b) and (d)] indicating a rotation of the pyridines, which canthen participate in forming coordination bonds with Fe, again consistent with the DFT results [figure4.3].Figure 4.2: STM images and corresponding chemical structures of Fe-coordinated TPPT. (a)and (b): Single TPPT with right tpy group coordinated to one Fe adatom (Vb = -200 mV,It = 1 nA). Coordination is mediated by the rotation [red arrow in (a)] of peripheral pyrgroups. Non-metalated tpy shows a depression (turquoise arrow) next to the peripheralpyr [panel (c)], which is not observed (green arrow) for metalated tpy [panel (d)]. (c)-(d)same scale. Blue: N; cyan: carbon; white: hydrogen; red: Fe.The molecular configuration with all three pyr oriented with the N atoms towards the Fehas two implications for the adsorption geometries observed. The coordination with Fe, involvingall three pyr (which previously strongly interacted with the substrate), reduces the interaction of themolecules with the Ag(111). This is consistent with the observed loss of preferential orientationof the chains with respect to the crystalline axes of the substrate [figure 4.1(b)-(d)], as well as with2Details on the DFT simulations performed by Chenguang Wang are discussed in appendix A.61Figure 4.3: Isolated TPPT terminated only on the right-hand side by a single Fe atom. (a) Op-timized DFT model on Ag(111) substrate along [1 -1 0] direction and (b) correspondingDFT-simulated STM image. Images calculated by Chenguang Wang.Figure 4.4: STM images and corresponding chemical structures of Fe-coordinated TPPT. (a)and (b): Single TPPT with both tpy groups coordinated to one Fe adatom (Vb = -500 mV,It = 1 nA). (c) and (d): Adjacent proton acceptor/organic ring interaction (red dashedlines) between two, singly metalated TPPTs (Vb = -500 mV, It = 1 nA). Blue: N; cyan:carbon; white: hydrogen; red: Fe.occasional tip-induced deformation of chains observed during STM imaging under some parame-ters. Also, the staggered intermolecular configuration, observed for bare [figures 3.3(a), 3.4 and4.1(a)] and singly coordinated [figure 4.4(d)] ligands, characterized by lateral interlocking betweentpy groups of adjacent molecules [red dashed lines in figure 4.4(c)], was not seen for tpy groups co-62ordinated with Fe. This provides further support for the outward-facing N configuration of the distalpyridines of bare tpy groups, as well as explaining the intermolecular arrangement in the zig-zagand staggered packing observed in the absence of Fe. Hence, this thermally activated configurationchange of the tpy both enables Fe coordination and alters intermolecular and molecule-substrateinteractions.4.3 Morphology of Fe-TPPT chainsStructural details of the metal-organic chains were studied via high-resolution STM images at dif-ferent biases. Figure 4.5(a) and (b) show STM topographs of a chain composed of five TPPTmolecules, with a periodicity of (2.3± 0.2) nm. STM images at negative biases [figure 4.5(a)] showtaller features at the coordination centers where the Fe should be, with two subtle protrusions. Atpositive biases [figure 4.5(b)], these regions are imaged as slight depressions compared to the sur-rounding ligand. The difference in contrast over the Fe is emphasized in figure 4.5(c), correspondingto the subtraction of the topographic maps in figure 4.5(a) and (b). At each coordination center, twoclear protrusions (fitted with two Gaussians) appear separated by a distance of (5 ± 1) A˚. In orderto better understand the structure of these chains, Fe-TPPT chains were modeled by DFT3 usingperiodic boundary conditions along the [1, -1, 0] axis of Ag(111) and a spacing of at least 8.7 A˚between two neighboring chains along the [1, 1, -2] direction. Coordination linkages consisting ofone, two or three Fe adatoms were considered. The structure with three linearly-arranged Fe atomsbetween the flat-lying tpy groups of the TPPT ligands [figure 4.5(d), (h), and (i)] was found to bemost energetically favorable by 0.57 eV, with a periodicity of 2.32 nm, in excellent agreement withthe experiment. A coordination configuration involving one or two adatoms is hindered by stericrepulsion between the flat-lying tpy. STM images simulated from the DFT model [figure 4.5(e)-(f)]reproduce the experimental data well [figure 4.5(a)-(c)]. Importantly, the 3-Fe coordination config-uration appears as two bright protrusions in the simulated topographic map subtraction separatedby a distance of (4.2 ± 0.1) A˚ [figure 4.5(g)], in reasonable agreement with the experimental result,and dominated by the electronic states from the two tpy-bound Fe [A and C in figure 4.5(h) and(i)] that are structurally 0.74 A˚ higher than the central Fe [B in figure 4.5(h) and (i)] in the model.The agreement between the simulated and measured STM images, including these two protrusions,strongly supports the model that the inter-ligand binding is mediated by a tpy-triiron-tpy coordi-nation scheme. This linear triiron cluster between tpy groups has not been previously observed,and presents a highly unusual and unexpected structure compared to related solution-synthesizedcompounds. Here, the two-dimensional constraint imposed by the surface during coordination, thestabilization provided by the interaction with the solid support, and the ultra-high vacuum environ-ment where low oxidation states can persist, present vastly different conditions than synthesis insolution. A similar linear tri-metallic linkage was observed by Wang et al. [114] for tpy containingmetal-organic chains utilizing Cu adatoms from a Cu(111) substrate but these do not involve the co-3Simulations details are discussed in appendix A.63V = +100 mVb(b)(d)[1 -1 0]CC2.2 Å0V = -500 mVb2.3 ± 0.2 nm(a)(c)ΔZ1 nm5 ± 1 ÅV = -500 mVbΔZV = +100 mVb2.32 nm4.2 ± 0.1 Å(g)(f)(e)0.74 Å(i)(h)FeBFeAFeCFigure 4.5: High-resolution STM and DFT of a 5-molecule Fe-TPPT nanochain on Ag(111).(a) and (b): Negative and positive bias STM images of a 5-molecule Fe-TPPT chain[(a) : Vb = -500 mV ; (b): 100 mV, It = 1 nA]. (c) Difference STM topographic mapresulting from subtraction of (a) with (b). Profile C along the coordination center showstwo protrusions, fitted with two Gaussians. (e) and (f): DFT-simulated topographic STMimages of the optimal Fe-TPPT/Ag(111) configuration [(e): Vb = -500 mV; (f): 100 mV].(g) Theoretical difference map resulting from subtraction of (e) and (f). (d) Model of theDFT-optimized metal-organic chain. (h) and (i): Top and side-view of the DFT-optimizedcoordination center. DFT images done by Chenguang Wang.64ordination of the outer pyridine groups of the tpy. While Cu is known to form two-fold coordinationbonding that could yield linear structures this has not previously been observed for Fe. Given thisunusual triiron arrangement, we also considered other types of linkage structures involving bridg-ing ligands incorporated from small quantities of residual gasses (CO and H2O) present even underUHV conditions, or Ag adatoms from the substrate. The density of water molecules on the samplesurface before iron deposition [(0.78 ± 0.32) molecules/100nm2] was significantly less than thedensity of the links [(2.27 ± 0.41) links/100nm2 before annealing and (3.82 ± 0.66) links/100nm2after annealing]. This indicates that there would be fewer than one H2O molecule per linkage atmost, while a hydroxyl model would require at least one water molecule per Fe-Fe bond [115]. Weexpect that linkages involving a bridging ligand would be imaged differently in STM than thosewithout one due to a substantially different electronic structure. As all the coordination linkages areimaged similarly, we do not see evidence for an H2O mediated structure, even in small proportionsand especially given the good agreement between the experiments and the model imaging providingsupport to the Fe3 linkage model. Molecular structures where Fe bonds to the carbon (C) atoms of aCO molecule have been reported in literature [116, 117]. However, the density of CO on the surfacebefore the iron deposition is negligibly small, leading to the elimination of any possible coordinationmodel involving CO, which would require six CO per linkage to create a bridged structure similarto triiron dodecacarbonyl. The other possible alternative to the Fe-Fe-Fe linkage model is Fe-Ag-Fe[figure 4.6], where a mobile Ag adatom is incorporated from the Ag(111) substrate. However, thereis strong experimental evidence and supporting DFT theory calculations that contradict this possiblestructure. Annealing experiments were performed on the TPPT/Ag(111) system without the pres-ence of Fe. Coordinated molecules and chains mediated by Ag adatoms from the substrate wereobserved after annealing at 523K [figure 4.7]. The chain morphology looks distinctly different fromthose obtained after the addition of Fe since they are clearly not linear and have a triangular shape atthe linkage. Since these coordinated molecules and chains were not observed at room temperature,one concludes that there is similarly insufficient mobility of Ag at RT to contribute to mixed Ag-Fecoordination centers during the deposition of Fe, and the linkages formed must only contain Fe asthe metal. A different scenario could occur after the annealing of the Fe/TPPT/Ag(111) system at373K. However, all of the linear nanostructures before and after annealing are always imaged in thesame way, indicating that even newly formed or extended chains are electronically and structurallyequivalent. Additionally, no chains exhibiting the morphology of Ag-mediated linkages (as in fig-ure 4.7: bent with a triangular centre) were observed, again indicating that there is not sufficientmobility of Ag adatoms even at 373K to contribute to chain formation. As previously discussed, atnegative bias two tall features are observed at the coordination center, while at positive bias theseare imaged as slight depressions. DFT calculations of an infinite chain of TPPTs linked by Fe-Fe-Feshow exactly the same behavior [figure 4.5(e) and (f)]. DFT simulations of the Fe-Ag-Fe link showthat both negative [figure 4.8(a)] and positive [figure 4.8(b)] bias STM images exhibit two clear tallfeatures at the coordination site and their subtraction [figure 4.8(c)] does not show the two verystriking protrusions as for the triiron case. Therefore, the imaging characteristic of an Fe-Ag-Fe65Figure 4.6: DFT simulated structure for an infinite chain of TPPT molecules linked by a Fe-Ag-Fe structure along the [1 -1 0] direction (a). Panels (b) and (c) are the top and side-view of the Fe-Ag-Fe linkage center, respectively. Images done by Chenguang Wang.Figure 4.7: STM constant-current image showing bent chains (dashed red circled features) af-ter annealing the TPPT/Ag(111) system at 523K for 30 minutes (Vb = -0.1V, It = 100pA).66mixed linkage is not consistent with what was experimentally observed. In conclusion, given theFigure 4.8: DFT-simulated constant current images of an infinite chain of TPPT moleculeslinked by the Fe-Ag-Fe structure shown in figure 4.6 at -500 mV (a), +200 mV (b) andtheir subtraction (c) = (a) - (b). Images done by Chenguang Wang.quantities of possible bridging ligands available in ultra-high vacuum, and the differences in imag-ing for any of these other structures, the linear triiron linkage seems to be the most likely structuralcandidate.674.4 Chains statistics and annealing effectsTo quantify aspects of the chain formation process, statistics of the observed chain lengths wereobtained for different sample preparations and different deposition methods (see details section 4.6).Figure 4.9 shows the probability of finding a chain composed of n molecules upon Fe deposition atroom temperature, before (blue) and after (red) annealing at 373 K, without additional depositionof Fe. To account for different image sizes for the length histograms, average chain lengths andstandard deviations were weighted by a factor:wi =AiAmin, (4.1)where Ai is the area of the image under consideration, and Amin is the minimum size image of all theones considered. This places greater importance on results from large area images where more rep-resentative statistics can be gathered than on smaller ones. Evidence for the validity of this methodwas obtained by comparison to results from a randomly generated test sample, and permutation test-ing to ensure that the uncertainties in the underlying distribution were properly represented by thisparticular normalization. Chains that intersect the boundaries of images also skew the distributiontowards lower chain length. Again, we tested the effect of counting the edge truncated chains asthe length visible in the image, versus discarding these chains, on a test distribution sampled bydifferent image sizes. Although the underlying distribution is not directly recoverable in either case,the effect on the recovered distribution of counting the edge truncated chains as the visible lengthwas less than discarding these chains from the statistics. Furthermore, as long as the sampled imagesizes were similar, the resulting statistics recovered a comparable distribution. Therefore, in chainlength statistics, these edge truncated chains were counted as the number of visible units. Sincethe image sizes sampled before and after annealing were similar, these distributions should be com-parable, not playing a significant role in the observed differences between pre- and post-annealinglength distributions, although both distributions may be skewed toward shorter chain lengths.The rise in the average chain length for the different sample preparations and differentdeposition methods is of about 19%. In samples prepared by depositing the Fe atoms in continu-ous flux, the experimental average number of molecules per chain before thermal annealing is 1.36,whereas after it is 1.65 [figure 4.9]. In samples where the iron was added in subsequent very shortdepositions separated in time by a fixed interval, the experimental average number of molecules perchain before thermal annealing is 1.31, whereas after it is 1.57 [figure 4.10], which is comparablewith the continuous flux preparation. This increase in average chain length accompanied by a de-crease in the number of “1-molecule chains” indicates that these serve as monomers that can diffuseand attach to other monomers to nucleate a new chain, or grow existing ones. Most notably, despitea small increase in the average chain length, the population remains dominated by monomers (“1-molecule chains”), even with an excess of Fe, indicating that the reservoir of reactants is not readilyconsumed and implying that the probability of successful chaining events is low. This may eitherindicate that the equilibrium does not favor extended chains due to a high probability of linkages68dissociating, or that the reaction is kinetically limited by slow diffusion or by the necessity of a thirdFe atom interacting with the monomers to extend the chains.Before and after annealing statistics counts were also done for the number and dimen-sions of the iron clusters but they were both found to be not statistically significant.Figure 4.9: Probability of finding a chain composed of n TPPT molecules after deposition ofFe at RT in continuous flux, before (blue) and after (red) subsequent annealing at 373 Kduring 10 min. Error bars represent uncertainty in the number of counts and are standarderrors weighted by the area of each STM image used. Insert: detail of the longer chaindistributions in the dashed box.69Figure 4.10: Probability of finding a chain composed of n TPPT molecules after depositionof Fe at RT in subsequent very short depositions (10 seconds) separated in time by afixed interval (5 seconds), before (blue) and after (red) subsequent annealing at 373 Kduring 10 min. Error bars represent uncertainty in the number of counts and are standarderrors weighted by the area of each STM image used. Insert: detail of the longer chaindistributions in the dashed box.704.5 Proposed model for the chain formation processBased on a TPPT-Fe3-TPPT coordination model, the chain formation mechanism can be describedas a multi-step process. Upon the addition of Fe atoms to the TPPT/Ag(111) system, thermal acti-vation induces atomic and molecular diffusion, as well as rotation of the distal pyr rings, allowingthe tpy groups of TPPT to rapidly coordinate with Fe iron adatoms. As these parameters are influ-enced by the substrate, the choice of surface plays an important role in the kinetics, as well as in thestability of the final structure. Chain nucleation occurs when two metalated tpy’s interact with anadditional Fe atom. Subsequent tpy-Fe3-tpy binding events involving both a metalated monomer,and an additional Fe atom, yield the formation of longer chains. This multi-step scenario mightexplain the low chaining yield [figure 4.9 - 4.10] and the observed defects such as Fe clusters [figure4.1(c)]. Further annealing resulted in an increase in the average chain length and decrease in thereservoir of metalated monomers (“1-molecule chains”) implying that the enhanced mobility of theFe-TPPT units, along with detachment of Fe atoms from small clusters [118, 119], allows furthergrowth of the chains. This shift towards longer chains implies that the growth is kinetically limitedat room temperature, rather than limited by detachment rates that would tend towards shorter chainlengths.4.6 Different iron deposition methodsSamples prepared by depositing the Fe atoms in continuous flux always showed a large number ofmetal clusters [cyan arrows in figure 4.1(c)] and “1-molecule chains”, and relatively short chains(max 6 molecules as in figure 4.9). This scenario seems to imply that the Fe arrives and diffuses onthe surface too fast to find a bare TPPT end or participate in the nucleation and growth of pre-existingchains before meeting another Fe. In order to see how the Fe deposition affects the characteristicsof the sample, two other different Fe deposition methods were tried. In the first method, ligands andmetal were deposited simultaneously on the substrate at RT [figure 4.11(a)]. The idea was to try tohave a similar number of bare tpy’s and Fe atoms simultaneously diffusing on the surface in such away that the probability for an Fe to find another one and cluster would be close to the probabilityfor an Fe to find a TPPT end and coordinate. However, since Fe diffuses faster than TPPTs, and atany time the number of molecules available on the surface for coordination is limited (with respectto the previous method) this resulted in a higher overall number of clusters, a lower amount ofcoordinated molecules and fewer, very short chains. Conversely, since the molecules diffuse slowlyrelative to Fe, the second method was an attempt to reduce the metal deposition rate to allow Fe toattach to a tpy group before finding another Fe atom to nucleate or join a cluster. This scenario issimilar to what was described in section 1.4.2 but where the diffusion of the ligands and depositionrate of Fe are the important parameters. Since the evaporator used cannot maintain a constant flux,in order to slow down the metal deposition the iron was deposited in subsequent smaller amounts(for about 5 to 10 seconds) separated in time by a (fixed) waiting period (of about 10 seconds)on the TPPT/Ag(111) system [figure 4.11(b)]. Before the addition of Fe, statistical counts of the71number of adsorbed TPPT were performed in order to estimate the number of bare tpy groups tobe coordinated (1 TPPT = 2 tpy groups). With this second method, the number of clusters wasreduced by ∼60%, the number of coordinated single ligands increased by ∼15%, and the averagechain length was found to be similar to the one when the metal was added in continuous flux [figure4.1(c)]. This demonstrates that with less iron diffusing on the surface there is a lower probability ofcluster formation (Fe atoms meeting each other) and a higher probability that the adatoms bind tobare terpyridines. Nonetheless, it seems to not have any significant effect on the formation of longerchains, probably because of the complexity of the Fe3 linkage.Figure 4.11: Constant-current STM images. (a) Fe/TPPT/Ag(111) system after simultane-ous deposition at RT of iron adatoms and TPPT molecules (Vb = -1V, It = 100pA).(b) Fe/TPPT/Ag(111) system after depositing the ligands first and then adding smallamount of iron in subsequent very short depositions separated in time by a fixed inter-val (both depositions at RT) (Vb = -1V, It = 50pA).724.7 Preliminary attempt to form chains on NaCl/Ag(111)As in the case of the bare ligand, we wanted to study the linear metal-organic nanostructures de-coupled from the metal, in order to reduce possible interactions with the substrate as well as toattempt optical experiments [120] without quenching effects due to the metal [121]. Since TPPTmolecules diffuse off the NaCl islands at RT, one cannot sequentially deposit TPPT and then Fe.Instead, a simultaneous deposition at RT was tried. After deposition, STM images of the sampleshowed no coordinated molecules or chains on NaCl islands [figure 4.12(a)]. Moreover, no singleFe or Fe-clusters were found. The Ag(111) areas adjacent to salt islands were observed to be fullycovered by well-ordered TPPTs arranged as they were seen under high-coverage condition togetherwith random sparse iron clusters (bright dots) [figure 4.12 (b) and (c)].Figure 4.12: Constant-current STM topographic images. (a) Fe/TPPT/NaCl/Ag(111) system:the NaCl island is completely free of adsorbates; the adjacent Ag(111) patch is coveredby well-ordered and closed-packed TPPT molecules and randomly sparse Fe clusters(bright dots) (Vb=500mV, It=1pA). (b) Zoom in the area enclosed by the red square inpanel (a). It shows ordered TPPTs and Fe clusters (dashed purple arrows) (Vb=500mV,It=25pA). (c) Zoom in the area enclose by the green rectangle in panel (b). It shows theTPPTs ordering (Vb=500mV, It=25pA).As diffusion of both Fe and TPPT appears to be too fast on NaCl for the formation ofchains at RT, future attempts should focus on temperature range [40K - 300K]. This lower limitcomes from the result of the counter heating experiment at 40K showing that TPPT molecules donot diffuse significantly (section 3.2).4.8 ConclusionIn this chapter, the formation and morphology of linear, self-assembled nanostructures, consistingof TPPT molecules coordinated with iron adatoms on an Ag(111) surface, were discussed in detail.STM measurements, supported by DFT results, show that the on-surface self-assemblyof the two species (TPPT and Fe) actually results in the formation of metal-ligand complexes. Thechaining process starts with the binding of the bare tpy groups of the molecules to single Fe by73means of all the three N atoms. In order for this coordination to occur, one needs mobility of bothmetal and organic species (observed at or above RT) and thermally activated rotation of the outerpyridines of the tpy groups (otherwise pointing towards the exterior of the molecule). Next, twofacing Fe-coordinated tpy need to meet in presence of a single free Fe in order to nucleate a newchain or grow an existing one by forming a unique triiron linkage. The complexity of the formationprocess seems to explain the low chaining yield, and the relatively short length of the chains (onlyup to 9 molecules in the best scenario).Finally, a preliminary attempt to form chains on salt islands with the substrate at RT,during the simultaneous depositions of ligand and metal, was presented. The diffusion of bothspecies appears to be too fast at room temperature for the formation of any coordinated structures.Future attempts will focus on properly tuning the substrate temperature during deposition.In conclusion, in this chapter a reliable and controlled fabrication of identical metal-organic complexes using diverse preparation methods and different initial conditions has been pre-sented. It is also been introduced the prospect of controlling the formation of similar structures on adifferent substrate by tuning the key growth parameters.74Chapter 5Electronic Structure of Fe3-TPPTChainsIn the previous chapter the synthesis and morphology of identical metal-organic complexes featur-ing a three iron linkage between two facing TPPT molecules, has been described and discussedin details. It has also been shown the reliability and control on the fabrication of such complexesusing different preparation methods and initial conditions. In this chapter, the electronic structureof different Fe-coordinated TPPT structures is analyzed in order to determine if these metal-ligandsystems exhibit the desired properties like an absorption band that goes up to near-IR region, soughtfor photovoltaic applications. Section 5.1 looks at a single molecule coordinated with Fe on both itstpy groups. Section 5.2 discusses two examples of long (4- and 5-molecule) chains. And, sections5.3 - 5.5 compare three types of sites within a chain: three molecules in the middle of the chain(5.3), at the end with the edge-molecule without Fe (5.4), and coordinated with Fe (5.5). Represen-tative data is shown in this chapter, and an overview of the data collected and additional examplesare reported in appendix C. The last section of this chapter is dedicated to general comments and in-terpretations of the results, which are also considered in comparison to the bare ligands on Ag(111)and NaCl/Ag(111). Here, some probable conclusions regarding the final product of this on-surfacesynthesis of the Fe-TPPT complex are drawn, as well.All of the following results were obtained from different grid measurements. If not oth-erwise specified, as described in appendix B, spectroscopy plots represent the normalized dI/dVobtained from a grid measurement by averaging over the regions enclosed by the same colour out-lines shown in the topographies on top of the plots. The STS maps represent the spatially resolvednormalized dI/dV at a given energy. Overlaid on top of each there is a black, dashed contour of thestructure considered, obtained from the corresponding topography image at the set-point bias volt-age of the grid. For each case, STS maps at biases corresponding to the most interesting featuresare shown. In addition, the bias +1.2V and +1.65V are specifically considered, to be compared tothe bare ligand on Ag(111) (+1.2V) and on salt (+1.65V) cases (these are the biases corresponding75to the first resonance peak of each of the two cases).5.1 “One-molecule chain”: single TPPT coordinated with Fe on bothits endsFigure 5.1(a) shows normalized STS curves obtained from a grid measurement on a single TPPT,with both its tpy groups coordinated with Fe. In the negative bias range, no obvious features cor-responding to occupied states are visible. At positive energies, both the STS curves from the entiremolecular axis [red curve in figure 5.1(a)] and the smaller region at the center of the ligand [bluecurve in figure 5.1(b), corresponding to the blue dot in the topography of figure 5.1(a)] show ashoulder at ∼+1.5V. When compared to the first resonance peak of a bare ligand on silver this isshifted towards higher energies, while when compared to the one on salt the energy shift is smaller[figure 5.1(b)].Figure 5.1: (a) Normalized dI/dV curves of a “1-molecule chain” obtained by averaging overthe regions enclosed by the corresponding colour outlines (see inset). Purple: whole“one-molecule chain”; red: molecular axis; dashed black: Ag(111), reference spectrum.Grid set-point parameters: Vb = -2.0V It = 1nA. Note: the divergence near zero-biascreated by normalization is removed. (b) Positive bias normalized dI/dV curves of bareligand on silver (green, point-dash line), and on NaCl (grey, point-dash line) and of “one-molecule chain” on Ag(111) (solid, blue line). Grids set-point parameters: bare ligandon Ag(111) Vb = -2.0V It = 1nA; bare ligand on NaCl/Ag(111) Vb = +1.20V It = 0.4pA;coordinated molecule Vb = -2.0V It = 1nA. Note: each of these measurements was takenwith a different tip, which is why the curves look so different.76Even though the STS curves do not show clear features for negative bias, the STS maps(from grids) do. STS maps at -0.6V and -0.3V [figure 5.2 (a) and (b)] do show some intensity con-fined within the molecule about the molecular axis. At -0.1V [figure 5.2 (c)] the intensity is observedto be at the edges of the coordinated ligand, close to where the Fe should reside, which is consistentwith what observed from constant-current topographic images at negative bias [figure 4.5]. At pos-itive energies a specific pattern is observed [figure 5.2 (d) - (j)], which was found to be the same forall the other coordinated structures analyzed and discussed below. The small differences in someenergy values between the different measurements of the different structures are probably due tobroadening effects. At +0.4V (d), the higher intensity of the normalized conductance is localizedabout the long molecular axis. As the energy increases, it moves towards the tpy groups and is con-centrated on four clearly separate locations, symmetric with respect to the molecule’s center (+0.8V,(e)). Then, at +1.2V (f), it asymmetrically merges back to the middle. At +1.5V (g) (close to thefirst resonance), it is mainly on the ligand. At +1.65V (h), some high intensity starts appearing onthe Fe, where it is exclusively observed to be at +1.8V (i). Finally, at +2V (j), it starts to spread backtowards the interior of the ligand, even though it is still mainly at the two far edges of the molecule.As it was observed for a bare ligand on both silver and salt, the energy of the first resonant peakcorresponds to the situation where the LDOS is localized in the middle of the molecule [figure 5.3].77Figure 5.2: STS energy maps at different biases showing the density of states of a “1-moleculechain” with both the tpy ends coordinated with Fe. From (a) to (j) the biases are:-0.6V, -0.3V, -0.1V, +0.4V, +0.8V, +1.2V, +1.5V, +1.65V, +1.8V and +2V. Grid set-pointparameters: Vb = -2.0V It = 1nA.78Figure 5.3: STS energy maps showing the electrons distribution at the energy of the first tun-neling resonance of a single: (a) bare TPPT on silver at +1.15V (grid set-point parame-ters: Vb = -2.50V It = 50pA); (b) bare TPPT on salt at +1.65V (grid set-point parameters:Vb = -2.0V It = 0.6pA); and (c) a TPPT coordinated with Fe on both ends at +1.5V (gridset-point parameters: Vb = -2.0V It = 1nA).795.2 Four- and five-molecule chainsFigures 5.4 and 5.5 show the normalized STS of a four- and a five-molecule chain respectively, asobtained from grid measurements. The four-molecule chain has both of its edge molecules coordi-Figure 5.4: Normalized dI/dV curves of a four-molecule chain obtained by averaging over theregions enclosed by the corresponding colour outlines (see inset). Purple: whole chain;cyan (L1), orange (L2), red (L3) and blue (L4): molecular axes from left to right; dashedblack: Ag(111), reference spectrum. Bars: green, first resonance of bare ligand on silver;grey, on NaCl. On top of the inset a sketch of the chain. Grid set-point parameters: Vb= -2.50V It = 2.5nA. Note: the divergence near zero-bias created by normalization isremoved.nated with Fe and, therefore, every molecule of the structure has both its tpy groups linked to Fe. Asa result, all the STS spectra on the long axis of the molecules present a peak at∼+1.5V, even thoughthe ones of the two end-molecules (cyan and blue) peak at a slightly higher bias. The five-moleculechain has neither of its end-molecules coordinated and, therefore, these two have a different struc-tural (tpy point outwards) and electronic configuration compared to the others inside the chain. Thisis reflected in their spectroscopy that shows the first tunneling resonance peak at ∼+1.2V (cyan andblue), which is∼0.3V lower than the peak position of the ligands inside the chain, but quite close tothe one of a bare TPPT on Ag. Also, the peak corresponding to the central TPPT (violet) is slightly80Figure 5.5: Normalized dI/dV curves of a five-molecule chain obtained by averaging over theregions enclosed by the corresponding colour outlines (see inset). Purple: whole chain;cyan (L1), orange (L2), violet (L3), red (L4) and blue (L5): molecular axes from left toright; dashed black: Ag(111), reference spectrum. Bars: green, first resonance of bareligand on silver; grey, on NaCl. On top of the inset a sketch of the chain. Grid set-point parameters: Vb = -2.50V It = 1.5nA. Note: the divergence near zero-bias created bynormalization is removed.shifted towards an higher energy with respect to the two neighboring ones (orange and red). The+1.5V peak of the fully coordinated ligands is ∼0.35V higher than a bare molecule on silver (greenbar) and ∼0.15V lower than the one on salt (grey bar). For both chains the negative side of eachSTS spectrum does not present any obvious feature.81The negative energy maps of the four-molecule chain [figure 5.6 (a) and (b)] show ahigher intensity on the molecular axis at -0.4V (a), whereas at -0.1V (b) about the metal center.At positive biases the intensity pattern is the same as the one observed for the “1-molecule chain.”In particular, at +0.4V (c) the higher intensity of the normalized dI/dV is on the long molecularaxis; around +0.8V (d) it is on four symmetric spots close to the tpy; and at +1.2V (e) these fourspots merge together. At the first resonance, occurring at +1.5V (f), the higher conductance is inthe middle of the molecule; at +1.65V (g) the intensity is quite delocalized over the entire chain;while at +1.8V (h) is on the metal. To view the full grid measurement of this four-molecule seeaccompanying video: Video 2 - four-molecule chain.Figure 5.6: STS energy maps at different biases showing the density of states of a four-molecule chain. From (a) to (h) the biases are: -0.4V, -0.1V, +0.4V, +0.8V, +1.2V, +1.5V,+1.65V and +1.8V. Grid set-point parameters: Vb = -2.50V It = 2.5nA.82Figure 5.7 shows the STS maps of the five-molecule chain. Throughout the energy rangebetween -0.4V and 1.8V the intensity pattern of all the TPPTs inside the chain, such as the ones withboth the tpy’s coordinated, is the same as the one described for the four-molecule chain. Exceptionsare observed for the two edge-molecules that are not coordinated on one end. Their peripheral pyrgroups have a different structural configuration, since the N atoms are pointing towards the exteriorof the molecule and their electronic structure is expected to differ since the end tpy groups are notcoordinated with Fe. This influences the normalized conductance as can be observed by looking atthe +0.8V energy map (d). Here, only two of the four symmetric spots are visible, and these onesare the ones by the coordinate side. This is consistent with what was observed for fully coordinatedligands. The non-coordinated side of the edge molecules instead shows the same intensity pattern ofa bare ligand on Ag(111) [see figure 3.10]. At +0.4V [panel (c)] a different intensity on the molec-ular axis of the three molecules inside the chain is visible. And, at +1.65V [panel (g)] the intensityis highly delocalized with the strongest intensity in the middle of the whole chain.Figure 5.7: STS energy maps at different biases showing the density of states of a five-molecule chain. From (a) to (h) the biases are: -0.4V, -0.1V, +0.4V, +0.8V +1.2V, +1.5V,+1.65V and +1.8V. Grid set-point parameters: Vb = -2.50V It = 1.5nA.835.3 Detail of a five-molecule chain: the three molecules in the middleof the chainFigure 5.8 shows the STS curves (from grid) of the three internal molecules of a 5-molecule chain.Once again, no obvious features are visible in the negative bias range. The first tunneling resonanceof each one of the three TPPTs is at about +1.5V, but the one of the middle molecule (violet) isslightly higher, which is consistent with that shown in the previous section.Figure 5.8: Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines (see inset).Purple: whole system; blue (L1), violet (L2) and red (L3): molecular axes from left toright; dashed black: Ag(111), reference spectrum. Bars: green, first resonance of bareligand on silver; grey, on NaCl. Grid set-point parameters: Vb = -2.50V It = 2.5nA. Note:the divergence near zero-bias created by normalization is removed.84The energy maps for both negative [panels (a) and (b)] and positive [panels (c) - (h)]energy are shown in figure 5.9 and they present the same intensity pattern described previously, alsoconfirming the results of the other measurement on a different 5-molecule chain.Figure 5.9: STS energy maps at different biases showing the density of states of two coordi-nation sites and three molecular axes. From (a) to (h) the biases are: -0.5V, -0.1V, +0.4V,+0.8V, +1.2V, +1.5V, +1.65V and +1.8V. Grid set-point parameters: Vb = -2.50V It =2.5nA.855.4 Detail of three-molecule chain with not coordinated end (baretpy)The normalized dI/dV ’s obtained from a grid measurement of a 3-molecule chain with one endmolecule not coordinated with Fe (right side - red) and the other one coordinated (left side - blue),are shown in figure 5.10. The spectra of the first two TPPTs from the left — both with both ends co-ordinated with Fe — (blue and violet) have a peak at∼+1.5V, but the spectrum of the Fe-coordinatedend-molecule (blue) peaks at a slightly higher energy than the spectrum of the TPPT inside the chain(violet). The non-coordinated end-TPPT spectrum (red) has a peak at about +1.2V, consistent withthe end-molecules of the 5-molecule chain, as previously discussed.Figure 5.10: Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines (see inset).Purple: whole structure; blue (L1), violet (L2) and red (L3): molecular axes from left toright; dashed black: Ag(111), reference spectrum. Bars: green, first resonance of bareligand on silver; grey, on NaCl. Grid set-point parameters: Vb = -2.50V It = 1nA. Note:the divergence near zero-bias created by normalization is removed.86The energy maps of this system for both negative and positive bias are shown in figure5.11. The normalized conductance pattern is the same as was observed and described in the pre-vious sections. In particular, the first two molecules from the left behave like any other moleculecoordinated on both ends, while the edge non-coordinated molecule presents the same characteris-tics discussed in section 5.2 for the end-molecules of the 5-molecule chain. The two coordinationcenters do not show the same intensity at the same bias, as one can mainly observe in panels (b),(g), and (h).Figure 5.11: STS energy maps at different biases showing the density of states of two coor-dination sites, three molecules, one of which is an edge-molecule not coordinated withFe. From (a) to (h) the biases are: -0.6V, -0.1V, +0.3V, +0.9V, +1.2V, +1.5V, +1.65Vand +1.8V. Grid set-point parameters: Vb = -2.50V It = 1nA.875.5 Detail of a three-molecule chain and one end-moleculecoordinated with FeThe last case to be discussed is the structure shown in the sketch in figure 5.12, a 3-molecule chainwith both the edge-molecules coordinated with Fe (right end is not considered). The STS curves(from grid) and maps are shown in figures 5.12 and 5.13, respectively. No additional details can beadded from the observation of the spectra and the energy maps up to +1.8V. At +2.5V [figure 5.13(i)] one can observe that the higher intensity of the normalized conductance is clearly moving fromthe Fe-tpy coordination areas, where it is found at +1.8V (h), back towards the ligand.Figure 5.12: Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines (see inset).Purple: whole structure; red (L1), violet (L2) and blue (L3): molecular axes from left toright; dashed black: Ag(111), reference spectrum. Bars: green, first resonance of bareligand on silver; grey, on NaCl. Grid set-point parameters: Vb = -2.50V It = 50pA. Note:the divergence near zero-bias created by normalization is removed.88Figure 5.13: STS energy maps at different biases showing the density of states of two coor-dination sites, three molecules, one of which is an edge molecule coordinated with Fe.From (a) to (i) the biases are: -0.6V, -0.1V, +0.4V, +0.9V, +1.2V, +1.5V, +1.65V, +1.8Vand +2.5V. Grid set-point parameters: Vb = -2.50V It = 50pA.895.6 General comments and interpretationsIn all the cases discussed above, the STS spectra of the whole coordinated structures (purple curvesof figures 5.1, 5.4, 5.5, 5.8, 5.10, and 5.12) never show clear peaks, but instead look more like abroad rise over a range between +0.5V and +1.8V with a shoulder at ∼+1.5V. The STS spectrarelative to the coordination regions, i.e. the areas between two facing tpy groups, were not shownsince these curves do not present any prominent feature at either negative or positive biases.Every TPPT coordinated with iron at either end (“1-molecule chain”, molecules inside achain, and edge coordinated molecules) shows a first resonance peak at ∼+1.5V [figures 5.1, 5.4,5.5, 5.8, 5.10, and 5.12], which is ∼0.35V higher than that of a bare ligand on silver but onlyslightly lower than that on salt. This could be either the result of the formation of a coordinationbond with charge transfer from the metal to the ligand, and/or the fact that the coordination itselfdrastically reduces the interaction of the TPPT molecules with the silver — as observed from theloss in preferential orientations of the chains — and therefore coordinated TPPTs behave morelike “free” molecules. Without a detailed knowledge about the electronic structure of the Fe3-TPPT chains electronically decoupled from the metal substrate, neither of these two explanationscan be excluded. Depending on the position occupied by a molecule in a chain and depending onits configuration (one or either tpy groups coordinated), consistent shifts in the energy of the firstresonance peak are observed throughout all the measurements [figures 5.4, 5.5, 5.8, 5.10, and 5.12].In fully-coordinated molecules these are found to be symmetric with respect to the middle of thechain. For example, in the four-molecule chain [figure 5.4], the two coordinated end-molecules(cyan and blue) have a similar STS spectrum, as do the two molecules inside the chain (orange andred). But the two end-molecules’ curves peak at a slightly higher bias than the other two. In thefive-molecule chain [figure 5.5] the central TPPT spectrum (violet) has the peak at slightly highervoltage still. From these observations, one can conclude that in the chains there is a symmetric (withrespect to the center) alternation of slightly higher and slightly lower values of the first resonancepeak, always around +1.5V. When a ligand has instead one bare and one linked tpy, a differentscenario is observed [figures 5.5 and 5.10]. Here, the peak energy is at ∼1.2V, which is closer inenergy to the resonance of a bare ligand on Ag (only ∼+0.05V higher), and lower than the one ofa TPPT on salt by ∼0.45V. The fact that its resonance is closer to the one of a bare ligand on Agcould be explained if this molecule had a stronger interaction with the substrate via its bare tpyand therefore it behaves more like a TPPT adsorbed on silver than one electronically decoupled.Another possible explanation is that, in this case, there is less charge transfer to the ligand sinceonly one tpy end is Fe-coordinated.At negative bias none of the STS spectra show obvious features related to occupied states[figures 5.1, 5.4, 5.5, 5.8, 5.10, and 5.12]. However, grid measurements show some intensity lo-calized on the ligand between -0.6V and -0.3V (depending on the measurement), and on the metalat -0.1V [figures 5.2, 5.6, 5.7, 5.9, 5.11, and 5.13]. This could be due to either a tunneling matrixeffect and so not be related to the electronic structure at all, or to a strong hybridization of states ofthe chain with the silver surface.90The electronic distribution of a “one-molecule chain” [figure 5.2] is observed to be quitedifferent from the one of a bare ligand on Ag(111) [figure 3.10]. The main difference is that theelectronic states of a coordinated TPPT are usually localized either on the long axis and close toits middle or on the metal, whereas the states of a bare TPPT at positive energies are generallydistributed in different, well-separated locations over the entire molecule. At the energy of the firstresonance peak, the higher density of states is localized in the middle of the molecular axis, as wasalso observed for bare TPPTs on both silver and salt [figure 5.3]. This seems to indicate that thisfirst peak visible in all the STS spectra should correspond to a molecular state, and more specificallyto the LUMO of a TPPT molecule.The energy maps of all the cases presented in this chapter show consistent patterns inthe spatial electronic distribution [figures 5.2, 5.6, 5.7, 5.9, 5.11, and 5.13]. In particular, an alter-nation between ligand (from -0.6V to -0.3V and from +0.4V to +1.7V) and metal intensity (-0.1Vand +1.8V) is observed. This could result in MLCT and LMCT optical transitions and, therefore,new optical excitations between metal and ligand states could occur. For example, an electron fromthe -0.1V metal state could be excited to the LUMO of the molecule at +1.5V by a photon withan energy of at least 1.6V corresponding to ∼775nm [figure 5.14]. This transition would be inthe near-infrared region. As mentioned in the introduction terpyridine-based ligands coordinated totransition metal atoms generally present adsorption bands up to the infrared region. It follows thatthe TPPT-Fe3-TPPT nanochains could then be considered for photovoltaic studies and applications.The alternating of metal-ligand-metal states is different from the ligand-metal-ligand one expectedfor an octahedrally coordinated tpy-Fe-tpy complex [122]. But, the TPPT-Fe3-TPPT systems do nothave an octahedral geometry and also the three metal atoms interact with each other resulting in aquite different geometry. The energy maps of the coordinated systems at +1.65V [figures 5.2, 5.6,5.7, 5.9, 5.11, and 5.13] show a high density of states spread over the whole structures, indicating adelocalization of the electrons at that energy. This characteristic could be interesting for molecularelectronics studies where these nanochains could serve as semiconductor nanowires.5.7 ConclusionIn this chapter the electronic structure of different cases of coordinated molecules was presented andinvestigated in detail. The analysis shows that coordinated systems present a specific spectroscopy,very different from the one of a bare ligand, that seems to prove the formation of a coordinationbond with charge transfer from metal to ligand and the chance of new optical transitions betweenmetal and ligand states (MLCT/LMCT transitions). This is what we set out to achieve with ourchoice of a bis-terpyridine based ligand (TPPT molecule) and the transition metal atom iron.91Figure 5.14: STS energy maps showing the electron distribution of a 4-molecule chain at (a)+1.8V, (b) +1.5V, and (c) -0.1V. (a) and (c) show electron states at the metal locations,while (b) on the ligands’ axis. Grid set-point parameters: Vb = -2.50V It = 2.5nA.92Chapter 6Conclusions and Outlook6.1 ConclusionsIn this thesis, a detailed study of the fabrication, morphology, and electronic structure of self-assembled iron-bis-terpyridine nanochains on a Ag(111) surface was presented. By borrowing ametal-organic coordination motif of successful dyes used in photovoltaic and catalysis applications— such as Ru or Os-polypyridine complexes — an in situ, on-surface method for the fabrication ofa well-defined, photoactive, metal-organic structure has been developed. Forming self-assembledcomplexes on a surface in ultra-high vacuum is not only a useful technique to prepare clean andcontrolled structures with high-fidelity: it also provides a flexible platform for investigating and tai-loring functional properties of diverse systems. Scanning tunneling microscopy and spectroscopywere used as measuring techniques to elucidate the structure of the chains and to learn about theirlocal electronic properties with sub-molecular resolution.The first step in the fabrication process of the macromolecular complexes was to de-posit the organic component, the Terpyridine-Phenyl-Phenyl-Terpyridine (TPPT) molecules ontothe Ag(111) substrate. Their structure and electronic states were investigated in detail. In STMconstant-current mode these molecules were imaged as dog-bone shaped features planar to the sur-face. The six geometrically equivalent adsorption orientations observed with respect to the silver[1, -1, 0] direction suggested a strong interaction with the substrate underneath, most likely via theterpyridine groups. This strong interaction was further supported by clear changes of the ligandelectronic structure decoupled from the metal by a NaCl bilayer showing significant differences inthe STS. In particular, an evident energy shift (∼0.5eV) of the LUMO and a possible hybridizationof the LUMO+1 of the molecules on Ag were found. The observed strong interplay is also consistentwith other studies on pyridine containing molecules deposited on noble metals surfaces [103, 104].The bare ligands on the Ag(111) substrate are characterized by having the N atoms of their pe-ripheral pyridine rings pointing towards the exterior of the molecule, due to the N-N electrostaticrepulsion between them and the N of the middle ring. This specific N position allows the adjacentbinding of bare TPPTs by means of an attractive acceptor/organic ring interaction [107, 108], which93also leads to the formation of staggered rows and zigzag patterns of molecules, and the growth ofwell-ordered islands in high-coverage conditions.The second step of the fabrication process was to deposit iron (Fe) adatoms onto theTPPT/Ag(111) system. Thermally activated self-assembled linear chains formed when the substratetemperature was at or above room temperature. Some thermal energy was found to be required toinitiate the self-assembly in order to activate diffusion of both the ligands and metal atoms, as wellas to allow a configuration change of the ligand (distal pyridine rotation). STM experimental results,together with density functional theory calculations revealed that these self-assembled nanochainsfeature an unusual linear triiron linkage between facing terpyridine groups. Here, all three N of a tpygroup are bound to the nearest Fe due to the distal pyridine rotation and the silver substrate plays akey role by interacting with the middle iron stabilizing a previously inaccessible complex (e.g. fromsynthesis in solution). The chain formation mechanism appears to follow a multi-step process. First,the bare tpy groups of the TPPTs must coordinate with a single iron atom by means of the distalpyridines rotation forming “one-molecule chains”. Then, these can participate in the nucleation andgrowth of longer chains. Since their tpy groups are now involved in the coordination bond, theinteraction with the substrate is weaker and their diffusion should be enhanced. The binding of twofacing Fe-coordinated tpy groups can occur only if they meet in the presence of an additional freeiron atom. The low chaining yield and the relatively short maximum length of the chains are likelyexplained by low probability of combining simultaneously three units: TPPT-Fe, Fe, and Fe-TPPT.To determine the characteristics and properties of the self-assembled metal-ligand com-plexes, the electronic structure of the chains was investigated by STS. The Fe-TPPT coordinationwas observed to visibly modify the ligand’s electronic structure. Fe-coordinated molecules (“one-molecule chains”, TPPTs inside a chain or with only one tpy group coordinated) were all foundto present a different LDOS from a bare ligand. In particular, the LUMO peak was observed tobe shifted towards a higher energy (∼0.35eV) because of a probable charge transfer between themetal and the ligand. Moreover, and more importantly, an alternation of metal and ligand stateswas observed in the energy maps suggesting the possible occurrence of MLCT/LMCT optical tran-sitions. Since the energy difference between the closest metal state observed below Fermi and thefirst ligand state above Fermi is of about 1.6eV (∼775nm), which corresponds to a transition in thenear-IR region, these metal-organic structures should then present an adsorption range at least upto the near-IR, as was sought to achieve with this particular choice of molecular and metallic com-ponents. Finally, the density of states of the coordinated systems at +1.65V shows a delocalizationof the electronic states over the whole nanostructures. This could be an interesting property to beinvestigated in molecular electronics studies where these nanochains could serve as semiconductornanowires.In conclusion, these self-assembled thermally activated Fe-bis-terpyridine nanochainsfabricated on a Ag(111) surface provide: 1) a metal-organic nanostructure anchored to a solidmetallic support, characterized by a highly probable broad adsorption up to the near-IR, soughtin photovoltaics applications; and 2) an unexpected triiron linkage, which might show interesting94electro, magnetic and spin properties, as well as being a good candidate for catalytic reactions.6.2 Open questionsThese results, of the bare TPPT molecules and the Fe-TPPT chains on Ag(111) leave open two mainquestions.1. Can we obtain a more complete understanding of the interaction between bare ligands andsilver substrate, e.g. which orbitals if any are hybridized?2. Can we confirm the charge transfer between Fe and TPPTs due to the coordination bond andthe formation of MLCT/LMCT optical transitions in the coordinated structures?Since chains could not be formed on insulating layers to date, in situ optical experimentsremain inaccessible to confirm 2. However, Density Functional Theory (DFT) calculations of theelectronic structure of bare ligands on Ag and on salt and chain on Ag should be able to provideinsight about:1. The bare TPPT and silver substrate interaction. More specifically, if an hybridization occurswhat effects do this have on the adsorbed species?2. The nature of the MOs observed in the STS of the bare ligands and the chains, such as if astate is molecular (TPPT), metallic (Fe), or Ag-TPPT due to hybridization.3. The effects of the Fe-tpy coordination, e.g. charge transfer from the metal to the ligand, andformation of MLCT/LMCT optical transitions.4. And, the characteristics of the triiron linkage, e.g. electro, magnetic and spin properties.6.3 Future directionsThe work presented in this thesis opens up the road to a series of other possible experiments withdifferent goals.• A possible first experiment is to substitute the TPPT molecules with a ligand with the sametermination end (terpyridine group) but with a longer or different body [123, 124] (for exam-ple three phenyl groups instead of two, figure 6.1). Here, one could study the changes in theelectronic structure of both bare and Fe-coordinated molecules, as well as observe differentmorphologies. For example, if the new ligand shows a higher flexibility in its body, twistsin the structure of both bare and coordinated molecules could be found, almost certainly alsoinfluencing the electronic properties.• By choosing a different metal center, one could study the influence on the conformation andthe electronic properties of the coordinated structure. A possible choice is osmium (Os).95Figure 6.1: Ligand similar to a TPPT molecule but with three phenyl groups instead of two.This is a bigger atom than Fe (same column but two rows down in the periodic table), whichmeans that it will occupy more space between two facing tpy groups, and therefore a singleOs atom might be enough to fill up the available space between two TPPTs. Additionally, themost probable coordination number for Os is 6, and requires a coordination structure with anoctahedral geometry [125]. In order to form this type of bond the TPPTs might have to twistand, as a consequence of that, the chains would no longer be as planar as the ones with Fe[figure 6.2].Figure 6.2: Probable twisted Os-terpyridine structure.• Another interesting experiment to try would be the synthesis and the study of the Fe-TPPTchains on an NaCl/Ag(111) substrate. Here, one can investigate the chains electronicallydecoupled from the metal, and also perform optical experiments [120] to prove or disprovethe formation of MLCT/LMCT optical transitions. This second type of experiments cannot beperformed with the chains directly adsorbed on the bare metal because the strong electroniccoupling with the metal would quench the fluorescence of the molecules [121], which is whatone could measure. The major challenge in achieving the formation of Fe-TPPT chains onsalt is finding the right temperature at which to heat the NaCl/Ag(111) substrate up to after4K depositions of both ligand and metal. The temperature must be high enough to have bothspecies diffusing and TPPT molecules able to rotate their peripheral pyridines. On the otherhand, it must be not too high in order to avoid having the adsorbates diffuse off the salt. Fromprevious experiments, it was determined that at 40K TPPTs do not diffuse, while at RT theydo not stay on salt islands. Hence, a good temperature (if any) must be found between 40K96and 300K. An additional challenge is the big difference in the mobility between Fe and ligand[58, 109, 110, 113]. If the iron atoms start diffusing too early with respect to the molecules,they will either cluster or diffuse off the salt before they can coordinate to a terpyridine groupand no chains will form. As an additional note, since the silver surface was determined tobe highly involved in the coordination linkage, it will not be surprising that in the absence ofit, if chains form, they will have a different morphology and, of course, a different electronicstructure.• The TPPT-Fe3-TPPT chains might present very interesting functionalities like electro, mag-netic [126] and spin properties and/or catalytic functions linked to the unusual linear 3-Felinkage. By combining theoretical DFT calculations and STM experiments, like the oneslisted below, one can investigate these properties.– STM studies of Fe-Pc molecules on Au(111) have demonstrated a relatively high Kondotemperature in these complexes [127, 128]. Therefore, STM measurements at low tem-perature and very low bias (in the order of meV) on the Fe3-TPPT chains should allowthe observation of a possible Kondo effect to determine whether the Fe3 clusters have anet spin.– Spin-polarized STM measurement performed on individual iron and chromium atomson cobalt nanoislands allowed the observation of spin polarized states of the single metalatoms [129]. With this technique, one should be able to determine the spin state of eachiron atom in the chain linkage.– Iron is known to be a low cost, low toxicity, powerful catalyzer for large-scale synthesisof fine chemicals [130, 131], also when part of metal-ligand complexes [132–134]. STMexperiments should be able to resolve possible catalytic reactions occurring on differenttypes of gas and molecules introduced in the chamber and deposited [135–137] on theFe/TPPT/Ag(111) sample.97Bibliography[1] W. S. Gorton. The genesis of the transistor. Proceedings of the IEEE, 86(1):50–52, 1998.ISSN 0018-9219. doi:10.1109/JPROC.1998.658758. → pages 1[2] Gordon E. Moore. Cramming more components onto integrated circuits, reprinted fromelectronics, volume 38, number 8, april 19, 1965, pp.114 ff. Solid-State Circuits SocietyNewsletter, IEEE, 11(5):33–35, 2006. ISSN 1098-4232.doi:10.1109/N-SSC.2006.4785860. → pages 1[3] Jack C. Wells Anatoli Korkin, Predrag S. Krstic´. Nanotechnology for Electronics,Photonics, and Renewable Energy. Springer, 2010. → pages 1[4] Jenny Nelson. Organic photovoltaic films. Current Opinion in Solid State and MaterialsScience, 6(1):87 – 95, 2002. ISSN 1359-0286.doi:http://dx.doi.org/10.1016/S1359-0286(02)00006-2. URLhttp://www.sciencedirect.com/science/article/pii/S1359028602000062. → pages 1[5] URL http://us.sunpower.com. Accessed: 2016-03-03. → pages 2[6] URL http://www.heliatek.com/en/. Accessed: 2016-03-03. → pages 2[7] URL http://www.nrel.gov/ncpv/images/efficiency chart.jpg. Accessed: 2016-03-14. →pages 2[8] Jean Pierre Sauvage, Jean Paul Collin, Jean Claude Chambron, Stephane Guillerez,Christophe Coudret, Vincenzo Balzani, Francesco Barigelletti, Luisa De Cola, and LuciaFlamigni. Ruthenium(ii) and osmium(ii) bis(terpyridine) complexes in covalently-linkedmulticomponent systems: Synthesis, electrochemical behavior, absorption spectra, andphotochemical and photophysical properties. Chemical Reviews, 94(4):993–1019, 1994.ISSN 0009-2665. doi:10.1021/cr00028a006. URL http://dx.doi.org/10.1021/cr00028a006.→ pages 2, 5[9] Guocan Li, Paolo G. Bomben, Kiyoshi C. D. Robson, Serge I. Gorelsky, Curtis P.Berlinguette, and Michael Shatruk. Ru complexes of thienyl-functionalized dipyrrins asncs-free sensitizers for the dye-sensitized solar cell. Chemical Communications, 48(70):8790–8792, 2012. ISSN 1359-7345. doi:10.1039/C2CC34311H. URLhttp://dx.doi.org/10.1039/C2CC34311H. → pages[10] Takashi Funaki, Hiromi Funakoshi, Osamu Kitao, Nobuko Onozawa-Komatsuzaki,Kazuyuki Kasuga, Kazuhiro Sayama, and Hideki Sugihara. Cyclometalated ruthenium(ii)complexes as near-ir sensitizers for high efficiency dye-sensitized solar cells. Angewandte98Chemie International Edition, 51(30):7528–7531, 2012. ISSN 1521-3773.doi:10.1002/anie.201108738. URL http://dx.doi.org/10.1002/anie.201108738. → pages[11] Philippe P. Laine´, Sebastiano Campagna, and Fre´de´rique Loiseau. Conformationally gatedphotoinduced processes within photosensitizer–acceptor dyads based on ruthenium(ii) andosmium(ii) polypyridyl complexes with an appended pyridinium group. CoordinationChemistry Reviews, 252(23–24):2552–2571, 2008. ISSN 0010-8545.doi:http://dx.doi.org/10.1016/j.ccr.2008.05.007. URLhttp://www.sciencedirect.com/science/article/pii/S001085450800101X. → pages 5[12] Keita Omata, Shota Kuwahara, Kenji Katayama, Shen Qing, Taro Toyoda, Kun-Mu Lee, andChun-Guey Wu. The cause for the low efficiency of dye sensitized solar cells with acombination of ruthenium dyes and cobalt redox. Phys. Chem. Chem. Phys., 17:10170–10175, 2015. doi:10.1039/C4CP05981F. URLhttp://dx.doi.org/10.1039/C4CP05981F. → pages 2[13] Christopher S. Kley, Christian Dette, Gordon Rinke, Christopher E. Patrick, Jan Cˇechal,Soon Jung Jung, Markus Baur, Michael Du¨rr, Stephan Rauschenbach, Feliciano Giustino,Sebastian Stepanow, and Klaus Kern. Atomic-scale observation of multiconformationalbinding and energy level alignment of ruthenium-based photosensitizers on TiO2 anatase.Nano Letters, 14(2):563–569, 2014. ISSN 1530-6984. doi:10.1021/nl403717d. URLhttp://dx.doi.org/10.1021/nl403717d. → pages 2, 3[14] Bernard Wenger, Michael Gra¨tzel, and Jacques-E. Moser. Rationale for kineticheterogeneity of ultrafast light-induced electron transfer from Ru(II) complex sensitizers tonanocrystalline TiO2, journal = Journal of the American Chemical Society, volume = 127,number = 35, pages = 12150-12151, year = 2005, doi = 10.1021/ja042141x, note =PMID:16131154, url = http://dx.doi.org/10.1021/ja042141x, eprint =http://dx.doi.org/10.1021/ja042141x. → pages 2[15] C. Uhlmann, I. Swart, and J. Repp. Controlling the orbital sequence in individualcu-phthalocyanine molecules. Nano Letters, 13(2):777–780, 2013. doi:10.1021/nl304483h.URL http://dx.doi.org/10.1021/nl304483h. PMID: 23356959. → pages 2[16] Anders Hagfeldt, Gerrit Boschloo, Licheng Sun, Lars Kloo, and Henrik Pettersson.Dye-sensitized solar cells. Chemical Reviews, 110(11):6595–6663, 2010. ISSN 0009-2665.doi:10.1021/cr900356p. URL http://dx.doi.org/10.1021/cr900356p. → pages 2, 7[17] Brian E. Hardin, Henry J. Snaith, and Michael D. McGehee. The renaissance ofdye-sensitized solar cells. Nat Photon, 6(3):162–169, 2012. ISSN 1749-4885. URLhttp://dx.doi.org/10.1038/nphoton.2012.22. → pages 2, 5[18] Andreas Winter, George R. Newkome, and Ulrich S. Schubert. Catalytic applications ofterpyridines and their transition metal complexes. ChemCatChem, 3(9):1384–1406, 2011.ISSN 1867-3899. doi:10.1002/cctc.201100118. URLhttp://dx.doi.org/10.1002/cctc.201100118. → pages 2, 5[19] James H. Alstrum-Acevedo, M. Kyle Brennaman, and Thomas J. Meyer. Chemicalapproaches to artificial photosynthesis. 2. Inorganic Chemistry, 44(20):6802–6827, 2005.ISSN 0020-1669. doi:10.1021/ic050904r. URL http://dx.doi.org/10.1021/ic050904r. →pages99[20] Christopher K. Prier, Danica A. Rankic, and David W. C. MacMillan. Visible lightphotoredox catalysis with transition metal complexes: Applications in organic synthesis.Chemical Reviews, 113(7):5322–5363, 2013. ISSN 0009-2665. doi:10.1021/cr300503r.URL http://dx.doi.org/10.1021/cr300503r. → pages 2, 5[21] Nunzio Tuccitto, Violetta Ferri, Marco Cavazzini, Silvio Quici, Genady Zhavnerko,Antonino Licciardello, and Maria Anita Rampi. Highly conductive ∼40-nm-long molecularwires assembled by stepwise incorporation of metal centres. Nat Mater, 8(1):41–46, 2009.ISSN 1476-1122.doi:http://www.nature.com/nmat/journal/v8/n1/suppinfo/nmat2332 S1.html. URLhttp://dx.doi.org/10.1038/nmat2332. → pages 2[22] Ryota Sakamoto, Shunsuke Katagiri, Hiroaki Maeda, and Hiroshi Nishihara.Bis(terpyridine) metal complex wires: Excellent long-range electron transfer ability andcontrollable intrawire redox conduction on silicon electrode. Coordination ChemistryReviews, 257(9–10):1493–1506, 2013. ISSN 0010-8545.doi:http://dx.doi.org/10.1016/j.ccr.2012.08.025. URLhttp://www.sciencedirect.com/science/article/pii/S0010854512002275. → pages[23] Wenjie Liang, Matthew P. Shores, Marc Bockrath, Jeffrey R. Long, and Hongkun Park.Kondo resonance in a single-molecule transistor. Nature, 417(6890):725–729, 2002. ISSN0028-0836. URL http://dx.doi.org/10.1038/nature00790. → pages 2[24] Jose Antonio Real, Ana Belen Gaspar, and M. Carmen Munoz. Thermal, pressure and lightswitchable spin-crossover materials. Dalton Transactions, (12):2062–2079, 2005. ISSN1477-9226. doi:10.1039/B501491C. URL http://dx.doi.org/10.1039/B501491C. → pages 2[25] Martin R. Gill and Jim A. Thomas. Ruthenium(ii) polypyridyl complexes and DNA-fromstructural probes to cellular imaging and therapeutics. Chemical Society Reviews, 41(8):3179–3192, 2012. ISSN 0306-0012. doi:10.1039/C2CS15299A. URLhttp://dx.doi.org/10.1039/C2CS15299A. → pages 2[26] A. Winter, M. Gottschaldt, G. R. Newkome, and U. S. Schubert. Terpyridines and theircomplexes with first row transition metal ions: cytotoxicity, nuclease activity andself-assembly of biomacromolecules. Curr Top Med Chem, 12(3):158–75, 2012. ISSN1873-4294 (Electronic) 1568-0266 (Linking). URLhttp://www.ncbi.nlm.nih.gov/pubmed/22236160. → pages 2[27] David N. Bowman, James H. Blew, Takashi Tsuchiya, and Elena Jakubikova. Elucidatingband-selective sensitization in Iron(II) Polypyridine-TiO2 assemblies. Inorganic Chemistry,52(15):8621–8628, 2013. ISSN 0020-1669. doi:10.1021/ic4007839. URLhttp://dx.doi.org/10.1021/ic4007839. → pages 2[28] James K. Hurst. Water oxidation catalyzed by dimeric -oxo bridged ruthenium diiminecomplexes. Coordination Chemistry Reviews, 249(3–4):313–328, 2005. ISSN 0010-8545.doi:http://dx.doi.org/10.1016/j.ccr.2004.06.017. URLhttp://www.sciencedirect.com/science/article/pii/S001085450400164X. → pages 2[29] Edwin C. Constable and Catherine E. Housecroft. Coordination chemistry: the scientificlegacy of alfred werner. Chemical Society Reviews, 42(4):1429–1439, 2013. ISSN1000306-0012. doi:10.1039/C2CS35428D. URL http://dx.doi.org/10.1039/C2CS35428D. →pages 3[30] G.A. Lawrance. Introduction to coordination chemistry. Wiley, 2010. → pages 3[31] G. L. Miessler and D. A. Tarr. Inorganic Chemistry. Pearson Prentice Hall, 3rd edition,2004. → pages 4[32] URL https://en.wikipedia.org/wiki/Coordination geometry. Accessed: 2016-05-22. → pagesviii, 4[33] V. Balzani and F. Scandola. Supramolecular Photochemistry. Ellis Harwood, England,1991. → pages viii, 4[34] Hironobu Ozawa, Yasuyuki Yamamoto, Kei Fukushima, Satoshi Yamashita, and HironoriArakawa. Synthesis and characterization of a novel ruthenium sensitizer with ahexylthiophene-functionalized terpyridine ligand for dye-sensitized solar cells. ChemistryLetters, 42(8):897–899, 2013. doi:10.1246/cl.130217. → pages viii, 5[35] URL http://koreabd.com/?c=37/42&p=5&uid=526. Accessed: 2016-03-03. → pages viii, 5[36] Anders Hagfeldt, Gerrit Boschloo, Licheng Sun, Lars Kloo, and Henrik Pettersson.Dye-sensitized solar cells. Chemical Reviews, 110(11):6595–6663, 2010. ISSN 0009-2665.doi:10.1021/cr900356p. URL http://dx.doi.org/10.1021/cr900356p. → pages 5[37] S. Altobello, R. Argazzi, S. Caramori, C. Contado, S. Da Fre´, P. Rubino, C. Chone´,G. Larramona, and C.A. Bignozzi. Sensitization of nanocrystalline TiO2 with blackabsorbers based on Os and Ru polypyridine complexes. Journal of the American ChemicalSociety, 127(44):15342–15343, 2005. doi:10.1021/ja053438d. URLhttp://dx.doi.org/10.1021/ja053438d. PMID: 16262377. → pages 5[38] Lingamallu Giribabu, Ravi Kumar Kanaparthi, and Veerapandian Velkannan. Molecularengineering of sensitizers for dye-sensitized solar cell applications. The Chemical Record,12(3):306–328, 2012. ISSN 1528-0691. doi:10.1002/tcr.201100044. URLhttp://dx.doi.org/10.1002/tcr.201100044. → pages 5[39] Kuan-Lin Wu, Shu-Te Ho, Chun-Cheng Chou, Yuh-Chia Chang, Hsiao-An Pan, Yun Chi,and Pi-Tai Chou. Engineering of osmium(ii)-based light absorbers for dye-sensitized solarcells. Angewandte Chemie International Edition, 51(23):5642–5646, 2012. ISSN1521-3773. doi:10.1002/anie.201200071. URL http://dx.doi.org/10.1002/anie.201200071.→ pages viii, 6[40] J.L. Steed, J.W.; Atwood. Supramolecular chemistry. 2000. → pages 7[41] Johannes V. Barth, Giovanni Costantini, and Klaus Kern. Engineering atomic and molecularnanostructures at surfaces. Nature, 437(7059):671–679, 2005. ISSN 0028-0836. URLhttp://dx.doi.org/10.1038/nature04166. → pages ix, 7, 8, 9, 10[42] Dirk M. Guldi, Hiroshi Imahori, Koichi Tamaki, Yukiyasu Kashiwagi, Hiroko Yamada,Yoshiteru Sakata, and Shunichi Fukuzumi. A molecular tetrad allowing efficient energystorage for 1.6 s at 163 K. The Journal of Physical Chemistry A, 108(4):541–548, 2004.ISSN 1089-5639. doi:10.1021/jp036382n. URL http://dx.doi.org/10.1021/jp036382n. →pages 7101[43] Yoshihiro Miyake, Kazunari Nakajima, Kouitsu Sasaki, Ryuichi Saito, Haruyuki Nakanishi,and Yoshiaki Nishibayashi. Design and synthesis of diphosphine ligands bearing anosmium(ii) bis(terpyridyl) moiety as a light-harvesting unit: Application to photocatalyticproduction of dihydrogen. Organometallics, 28(17):5240–5243, 2009. ISSN 0276-7333.doi:10.1021/om900489d. URL http://dx.doi.org/10.1021/om900489d. → pages 7[44] Vaishali Vajpayee, Hyunuk Kim, Anurag Mishra, Partha Sarathi Mukherjee, Peter J. Stang,Min Hyung Lee, Hwan Kyu Kim, and Ki-Whan Chi. Self-assembled molecular squarescontaining metal-based donor: synthesis and application in the sensing of nitro-aromatics.Dalton Transactions, 40(13):3112–3115, 2011. ISSN 1477-9226.doi:10.1039/C0DT01481H. URL http://dx.doi.org/10.1039/C0DT01481H. → pages 7[45] H. Furukawa, K. E. Cordova, M. OKeeffe, and O.M. Yaghi. The chemistry and applicationsof metal-organic frameworks. Science, 341:974–988, 2013.doi:DOI:10.1126/science.1230444. → pages 7[46] Stuart L. James. Metal-organic frameworks. Chem. Soc. Rev., 32:276–288, 2003.doi:10.1039/B200393G. → pages[47] Hong-Cai Zhou, Jeffrey R. Long, and Omar M. Yaghi. Introduction to metalorganicframeworks. Chemical Reviews, 112(2):673–674, 2012. doi:10.1021/cr300014x. URLhttp://dx.doi.org/10.1021/cr300014x. PMID: 22280456. → pages[48] Timothy R Cook, Yao-Rong Zheng, and Peter J Stang. Metal-organic frameworks andself-assembled supramolecular coordination complexes: Comparing and contrasting thedesign, synthesis and functionality of metal-organic materials. Chemical reviews, 113(1):734–777, 01 2013. doi:10.1021/cr3002824. URLhttp://www.ncbi.nlm.nih.gov/pmc/articles/PMC3764682/. → pages 7[49] Johannes V. Barth. Molecular architectonic on metal surfaces. Annual Review of PhysicalChemistry, 58(1):375–407, 2007. doi:doi:10.1146/annurev.physchem.56.092503.141259.URLhttp://www.annualreviews.org/doi/abs/10.1146/annurev.physchem.56.092503.141259. →pages 8, 9[50] Stepanow Sebastian, Lin Nian, and V. Barth Johannes. Modular assembly oflow-dimensional coordination architectures on metal surfaces. Journal of Physics:Condensed Matter, 20(18):184002, 2008. ISSN 0953-8984. URLhttp://stacks.iop.org/0953-8984/20/i=18/a=184002. → pages[51] Sebastian Stepanow, Nian Lin, Dietmar Payer, Uta Schlickum, Florian Klappenberger,Giorgio Zoppellaro, Mario Ruben, Harald Brune, JohannesV Barth, and Klaus Kern.Surface-assisted assembly of 2D metal–organic networks that exhibit unusual threefoldcoordination symmetry. Angewandte Chemie International Edition, 46(5):710–713, 2007.ISSN 1521-3773. doi:10.1002/anie.200603644. URLhttp://dx.doi.org/10.1002/anie.200603644. → pages 8[52] Johannes V. Barth. Fresh perspectives for surface coordination chemistry. Surface Science,603(10–12):1533–1541, 2009. ISSN 0039-6028.doi:http://dx.doi.org/10.1016/j.susc.2008.09.049. URLhttp://www.sciencedirect.com/science/article/pii/S0039602809000569. → pages 8, 9102[53] N. Lin, S. Stepanow, M. Ruben, and J. V. Barth. Surface-confined supramolecularcoordination chemistry. Top Curr Chem, 287:1–44, 2009. ISSN 0340-1022 (Print)0340-1022 (Linking). doi:10.1007/128 2008 150. URLhttp://www.ncbi.nlm.nih.gov/pubmed/23604441. → pages 8, 9[54] Sebastian Stepanow, Magali Lingenfelder, Alexandre Dmitriev, Hannes Spillmann, ErikDelvigne, Nian Lin, Xiaobin Deng, Chengzhi Cai, Johannes V. Barth, and Klaus Kern.Steering molecular organization and host-guest interactions using two-dimensionalnanoporous coordination systems. Nat Mater, 3(4):229–233, 2004. ISSN 1476-1122. URLhttp://dx.doi.org/10.1038/nmat1088. → pages 8[55] U. Schlickum, R. Decker, F. Klappenberger, G. Zoppellaro, S. Klyatskaya, M. Ruben,I. Silanes, A. Arnau, K. Kern, H. Brune, and J. V. Barth. Metalorganic honeycombnanomeshes with tunable cavity size. Nano Letters, 7(12):3813–3817, 2007. ISSN1530-6984. doi:10.1021/nl072466m. URL http://dx.doi.org/10.1021/nl072466m. → pagesix, 8, 9[56] Ludwig Bartels. Tailoring molecular layers at metal surfaces. Nature chemistry, 2(2):87–95,2010. → pages 8[57] Agustin Schiffrin, Andreas Riemann, Willi Auwa¨rter, Yan Pennec, Alex Weber-Bargioni,Dean Cvetko, Albano Cossaro, Alberto Morgante, and Johannes V. Barth. Zwitterionicself-assembly of l-methionine nanogratings on the Ag(111) surface. Proceedings of theNational Academy of Sciences, 104(13):5279–5284, 2007. doi:10.1073/pnas.0607867104.URL http://www.pnas.org/content/104/13/5279.abstract. → pages ix, 9[58] Agustin Schiffrin, Joachim Reichert, Willi Auwa¨rter, Gesine Jahnz, Yan Pennec, AlexanderWeber-Bargioni, Valeri S. Stepanyuk, Larissa Niebergall, Patrick Bruno, and Johannes V.Barth. Self-aligning atomic strings in surface-supported biomolecular gratings. PhysicalReview B, 78(3):035424, 2008. URL http://link.aps.org/doi/10.1103/PhysRevB.78.035424.→ pages 8, 58, 59, 97[59] J.A. Venables. Introduction to Surface and Thin Film Processes.Cambridge University Press, 2000. → pages 8[60] Holger Roder, Elmar Hahn, Harald Brune, Jean-Pierre Bucher, and Klaus Kern. Buildingone- and two-dimensional nanostructures by diffusion-controlled aggregation at surfaces.Nature, 366(6451):141–143, 11 1993. URL http://dx.doi.org/10.1038/366141a0. → pages 8[61] Johannes V. Barth, Jens Weckesser, Chengzhi Cai, Peter Gnter, Lukas Brgi, OlivierJeandupeux, and Klaus Kern. Building supramolecular nanostructures at surfaces byhydrogen bonding. Angewandte Chemie International Edition, 39(7):1230–1234, 2000.ISSN 1521-3773.doi:10.1002/(SICI)1521-3773(20000403)39:7〈1230::AID-ANIE1230〉3.0.CO;2-I. URL http://dx.doi.org/10.1002/(SICI)1521-3773(20000403)39:7〈1230::AID-ANIE1230〉3.0.CO;2-I.→ pages 8[62] S. Whitelam and R. L. Jack. The statistical mechanics of dynamic pathways toself-assembly. Annual Review of Physical Chemistry, 66:143–163, 2015.doi:DOI:10.1146/annurev-physchem-040214-121215. → pages 9103[63] G. Binnig and H. Rohrer. Scanning tunneling microscopy. Helvetica Physica Acta, 55(6):726–735, 1982. ISSN 0018-0238. URL 〈GotoISI〉://WOS:A1982QR54200011. → pages 12[64] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel. 7x7 reconstruction on Si(111) resolved inreal space. Physical Review Letters, 50(2):120–123, 1983. ISSN 0031-9007.doi:DOI10.1103/PhysRevLett.50.120. URL 〈GotoISI〉://WOS:A1983PW72600010. →pages ix, 13, 14[65] H. Ohtani, R. J. Wilson, S. Chiang, and C. M. Mate. Scanning tunneling microscopyobservations of benzene molecules on the Rh(111)-(3 x 3) (C6H6+2CO) surface. Phys RevLett, 60(23):2398–2401, 1988. ISSN 1079-7114 (Electronic) 0031-9007 (Linking). URLhttp://www.ncbi.nlm.nih.gov/pubmed/10038341. → pages ix, 13, 14[66] Jascha Repp, Gerhard Meyer, Sladjana M. Stojkovic´, Andre´ Gourdon, and ChristianJoachim. Molecules on insulating films: Scanning-tunneling microscopy imaging ofindividual molecular orbitals. Physical Review Letters, 94(2):026803, 2005. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.94.026803. → pages 13, 29, 37[67] D. M. Eigler and E. K. Schweizer. Positioning single atoms with a scanning tunnelingmicroscope. Nature, 344(6266):524–526, 1990. ISSN 0028-0836.doi:DOI10.1038/344524a0. URL 〈GotoISI〉://WOS:A1990CX59500059. → pages 13[68] B. C. Stipe, M. A. Rezaei, and W. Ho. Single-molecule vibrational spectroscopy andmicroscopy. Science, 280(5370):1732–1735, 1998. ISSN 0036-8075.doi:DOI10.1126/science.280.5370.1732. URL 〈GotoISI〉://WOS:000074197800035. →pages ix, 13, 14[69] M. F. Crommie, C. P. Lutz, and D. M. Eigler. Confinement of electrons to quantum corralson a metal surface. Science, 262(5131):218–220, 1993. doi:10.1126/science.262.5131.218.URL http://www.sciencemag.org/content/262/5131/218.abstract. → pages ix, 14[70] J. Tersoff and D. R. Hamann. Theory and application for the scanning tunnelingmicroscope. Physical Review Letters, 50(25):1998–2001, 1983. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.50.1998. → pages 13, 17[71] J. Tersoff and D. R. Hamann. Theory of the scanning tunneling microscope. PhysicalReview B, 31(2):805–813, 1985. URL http://link.aps.org/doi/10.1103/PhysRevB.31.805. →pages 17, 112[72] H. J. W. Zandvliet and A. van Houselt. Scanning tunneling spectroscopy. Annual Review ofAnalytical Chemistry, 2:37–55, 2009. ISSN 1936-1327.doi:DOI10.1146/annurev-anchem-060908-155213. URL〈GotoISI〉://WOS:000268069300003. → pages 13[73] R. J. de Vries, A. Saedi, D. Kockmann, A. van Houselt, B. Poelsema, and H. J. W. Zandvliet.Spatial mapping of the inverse decay length using scanning tunneling microscopy. AppliedPhysics Letters, 92(17), 2008. ISSN 0003-6951. doi:Artn174101Doi10.1063/1.2917716.URL 〈GotoISI〉://WOS:000255524000094. → pages 13[74] URL http://www.ntmdt.com/spm-basics/view/tunneling-effect. Accessed: 2016-03-03. →pages x, 15104[75] C. J. Chen. Introduction to Scanning Tunneling Microscopy. Oxford University Press, 2ndedition, 2008. → pages 15, 16, 17[76] J. Bardeen. Tunnelling from a many-particle point of view. Physical Review Letters, 6(2):57–59, 1961. URL http://link.aps.org/doi/10.1103/PhysRevLett.6.57. → pages 17[77] A. Selloni, P. Carnevali, E. Tosatti, and C. D. Chen. Voltage-dependent scanning-tunnelingmicroscopy of a crystal surface: Graphite. Physical Review B, 31(4):2602–2605, 1985.URL http://link.aps.org/doi/10.1103/PhysRevB.31.2602. → pages 17[78] C. Kittel. Introduction to Solid State Physics. John Wiley and Sons Inc., 7th edition, 1996.→ pages 18[79] Werner A. Hofer, Adam S. Foster, and Alexander L. Shluger. Theories of scanning probemicroscopes at the atomic scale. Rev. Mod. Phys., 75:1287–1331, Oct 2003.doi:10.1103/RevModPhys.75.1287. URLhttp://link.aps.org/doi/10.1103/RevModPhys.75.1287. → pages 18[80] Flemming Besenbacher. Scanning tunnelling microscopy studies of metal surfaces. Reportson Progress in Physics, 59(12):1737, 1996. → pages 18[81] R. M. Feenstra and P. Ma˚rtensson. Fermi-level pinning at the Sb/GaAs(110) surface studiedby scanning tunneling spectroscopy. Physical Review Letters, 61(4):447–450, 1988. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.61.447. → pages 18, 113[82] Joseph A. Stroscio, R. M. Feenstra, and A. P. Fein. Electronic structure of the Si(111)2 x 1surface by scanning-tunneling microscopy. Phys. Rev. Lett., 57:2579–2582, Nov 1986.doi:10.1103/PhysRevLett.57.2579. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.57.2579. → pages[83] J. A. Kubby, J. E. Griffith, R. S. Becker, and J. S. Vickers. Tunneling microscopy ofGe(001). Physical Review B, 36(11):6079–6093, 1987. URLhttp://link.aps.org/doi/10.1103/PhysRevB.36.6079. → pages[84] R. J. Hamers, Ph Avouris, and F. Bozso. Imaging chemical-bond formation with thescanning tunneling microscope: NH3 dissociation on Si(001). Physical Review Letters, 59(18):2071–2074, 1987. URL http://link.aps.org/doi/10.1103/PhysRevLett.59.2071. → pages18, 113[85] URL http://www.scientaomicron.com. Accessed: 2016-03-03. → pages 21, 30[86] Ben MacLeod. An ultra-low-vibration facility for housing a dilution temperature scanningtunneling microscope. Master’s thesis, University of British Columbia, 2015. → pages x,21, 23[87] Omicron Nanotechnology. LT STM User’s Guide. Oxford Instruments, 4.1 edition,December 2013. → pages x, 24, 25[88] URL http://www.goodfellow.com. Accessed: 2016-03-03. → pages 26, 30[89] URL http://www.mateck.com. Accessed: 2016-03-16. → pages 28105[90] S. Orisaka, T. Minobe, T. Uchihashi, Y. Sugawara, and S. Morita. The atomic resolutionimaging of metallic Ag(111) surface by noncontact atomic force microscope. AppliedSurface Science, 140(3–4):243–246, 1999. ISSN 0169-4332.doi:http://dx.doi.org/10.1016/S0169-4332(98)00534-0. URLhttp://www.sciencedirect.com/science/article/pii/S0169433298005340. → pages 28[91] H. Hovel, B. Grimm, and B. Reihl. Modification of the Shockley-type surface state onAg(111) by an adsorbed xenon layer. Surface Science, 477(1):43 – 49, 2001. ISSN0039-6028. doi:http://dx.doi.org/10.1016/S0039-6028(01)00704-X. URLhttp://www.sciencedirect.com/science/article/pii/S003960280100704X. → pages x, 28[92] Karina Morgenstern and Jouko Nieminen. Imaging water on Ag(111): Field inducedreorientation and contrast inversion. The Journal of Chemical Physics, 120(22):10786–10791, 2004. doi:http://dx.doi.org/10.1063/1.1737366. URLhttp://scitation.aip.org/content/aip/journal/jcp/120/22/10.1063/1.1737366. → pages 28[93] K. A. Cochrane, A. Schiffrin, T. S. Roussy, M. Capsoni, and S. A. Burke. Pronouncedpolarization-induced energy level shifts at boundaries of organic semiconductornanostructures. Nat Commun, 6, 2015. doi:10.1038/ncomms9312. URLhttp://dx.doi.org/10.1038/ncomms9312. → pages 29[94] Marina Pivetta, Franc¸ois Patthey, Massimiliano Stengel, Alfonso Baldereschi, andWolf-Dieter Schneider. Local work function Moire´ pattern on ultrathin ionic films:NaCl on Ag(100). Physical Review B, 72(11):115404, 2005. URLhttp://link.aps.org/doi/10.1103/PhysRevB.72.115404. → pages 29[95] Jascha Repp, Gerhard Meyer, and Karl-Heinz Rieder. Snell’s law for surface electrons:Refraction of an electron gas imaged in real space. Physical Review Letters, 92(3):036803,2004. URL http://link.aps.org/doi/10.1103/PhysRevLett.92.036803. → pages 29[96] Qinmin Guo, Zhihui Qin, Cunding Liu, Kan Zang, Yinghui Yu, and Gengyu Cao.Bias dependence of apparent layer thickness and Moire´ pattern on NaCl/Cu(001). SurfaceScience, 604(19–20):1820–1824, 2010. ISSN 0039-6028.doi:http://dx.doi.org/10.1016/j.susc.2010.07.013. URLhttp://www.sciencedirect.com/science/article/pii/S0039602810002992. → pages 29[97] S. Heidorn, C. Bertram, J. Koch, K. Boom, F. Matthaei, A. Safiei, J. Henzl, andK. Morgenstern. Influence of substrate surface-induced defects on the interface statebetween NaCl(100) and Ag(111). The Journal of Physical Chemistry C, 117(31):16095–16103, 2013. ISSN 1932-7447. doi:10.1021/jp405297h. URLhttp://dx.doi.org/10.1021/jp405297h. → pages 29[98] URL http://www.metafysica.nl/turing/preparation 3dim 2.html. Accessed: 2016-03-03. →pages xi, 29[99] URL http://www.hetcat.com. Accessed: 2016-03-03. → pages 30[100] URL http://www.kentax.de. Accessed: 2016-03-03. → pages 30[101] Sujoy Karan, Christian Hamann, Hao Tang, Artur R. Stefankiewicz, Jean-Marie Lehn, andRichard Berndt. Surface trapping and STM observation of conformational isomers of a106bis(terpyridine) ligand from metallosupramolecular grids. ChemPhysChem, 16(7):1370–1373, 2015. ISSN 1439-7641. doi:10.1002/cphc.201500100. URLhttp://dx.doi.org/10.1002/cphc.201500100. → pages 33, 34[102] Wei Liu, Javier Carrasco, Biswajit Santra, Angelos Michaelides, Matthias Scheffler, andAlexandre Tkatchenko. Benzene adsorbed on metals: Concerted effect of covalency and vander waals bonding. Physical Review B, 86(24):245405, 2012. URLhttp://link.aps.org/doi/10.1103/PhysRevB.86.245405. → pages 33[103] Jin Qiang Hou, Hong Seok Kang, Ki Wan Kim, and Jae Ryang Hahn. Bindingcharacteristics of pyridine on Ag(110). The Journal of Chemical Physics, 128(13):134707,2008. doi:doi:http://dx.doi.org/10.1063/1.2888932. URLhttp://scitation.aip.org/content/aip/journal/jcp/128/13/10.1063/1.2888932. → pages 34, 56,93[104] M. Bader, J. Haase, K. H. Frank, A. Puschmann, and A. Otto. Orientational phase transitionin the system pyridine/Ag(111): A near-edge X-ray-absorption fine-structure study.Physical Review Letters, 56(18):1921–1924, 1986. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.56.1921. → pages 34, 56, 93[105] Xingqiang Shi, Rui Qin Zhang, Christian Minot, K. Hermann, Michel A. Van Hove, WeihuaWang, and Nian Lin. Complex molecules on a flat metal surface: Large distortions inducedby chemisorption can make physisorption energetically more favorable. The Journal ofPhysical Chemistry Letters, 1(19):2974–2979, 2010. ISSN 1948-7185.doi:10.1021/jz1011753. URL http://dx.doi.org/10.1021/jz1011753. → pages 34, 35[106] X. Q. Shi, W. H. Wang, S. Y. Wang, N. Lin, and M. A. Van Hove. Thermally activatedtransformation of the adsorption configurations of a complex molecule on a Cu(111)surface. Catalysis Today, 177(1):50–54, 2011. ISSN 0920-5861.doi:http://dx.doi.org/10.1016/j.cattod.2011.03.072. URLhttp://www.sciencedirect.com/science/article/pii/S0920586111003634. → pages 34, 35[107] Emmanuel Arras, Ari Paavo Seitsonen, Florian Klappenberger, and Johannes V. Barth.Nature of the attractive interaction between proton acceptors and organic ring systems.Physical Chemistry Chemical Physics, 14(46):15995–16001, 2012. ISSN 1463-9076.doi:10.1039/C2CP42293J. URL http://dx.doi.org/10.1039/C2CP42293J. → pages xii, 35,36, 93[108] Manuela Mura and Fabien Silly. Experimental and theoretical analysis of hydrogen bondingin two-dimensional chiral 4,4-(1,4-phenylene)bis(2,2:6,2-terpyridine) self-assemblednanoarchitecture. The Journal of Physical Chemistry C, 119(48):27125–27130, 2015.doi:10.1021/acs.jpcc.5b07231. URL http://dx.doi.org/10.1021/acs.jpcc.5b07231. → pagesxii, 35, 36, 93[109] M. Eichberger, M. Marschall, J. Reichert, A. Weber-Bargioni, W. Auwa¨rter, R. L. C. Wang,H. J. Kreuzer, Y. Pennec, A. Schiffrin, and J. V. Barth. Dimerization boosts one-dimensionalmobility of conformationally adapted porphyrins on a hexagonal surface atomic lattice.Nano Letters, 8(12):4608–4613, 2008. ISSN 1530-6984. doi:10.1021/nl802995u. URLhttp://dx.doi.org/10.1021/nl802995u. → pages 38, 58, 59, 97107[110] Dirk Ku¨hne, Florian Klappenberger, Re´gis Decker, Uta Schlickum, Harald Brune, SvetlanaKlyatskaya, Mario Ruben, and Johannes V. Barth. Self-assembly of nanoporous chiralnetworks with varying symmetry from sexiphenyl-dicarbonitrile on Ag(111). The Journal ofPhysical Chemistry C, 113(41):17851–17859, 2009. ISSN 1932-7447.doi:10.1021/jp9041217. URL http://dx.doi.org/10.1021/jp9041217. → pages 38, 59, 97[111] Michael Rohlfing, Ruslan Temirov, and Frank Stefan Tautz. Adsorption structure andscanning tunneling data of a prototype organic-inorganic interface: PTCDA on Ag(111).Phys. Rev. B, 76:115421, Sep 2007. doi:10.1103/PhysRevB.76.115421. URLhttp://link.aps.org/doi/10.1103/PhysRevB.76.115421. → pages 42[112] M. Capsoni, A. Schiffrin, C. Wang, T.S. Roussy, K.A. Cochrane, A.Q. Shaw, W. Ji, and S.A.Burke. On-surface synthesis of iron terpyridine nanochains featuring a linear tri-ironlinkage. in submission, 2016. → pages 58[113] N. Knorr, H. Brune, M. Epple, A. Hirstein, M. A. Schneider, and K. Kern. Long-rangeadsorbate interactions mediated by a two-dimensional electron gas. Physical Review B, 65(11):115420, 2002. URL http://link.aps.org/doi/10.1103/PhysRevB.65.115420. → pages58, 59, 97[114] Weihua Wang, Yuning Hong, Xingqiang Shi, Christian Minot, Michel A. Van Hove,Ben Zhong Tang, and Nian Lin. Inspecting metal-coordination-induced perturbation ofmolecular ligand orbitals at a submolecular resolution. The Journal of Physical ChemistryLetters, 1(15):2295–2298, 2010. ISSN 1948-7185. doi:10.1021/jz100855k. URLhttp://dx.doi.org/10.1021/jz100855k. → pages 63[115] Lisa M. Brines, Michael K. Coggins, Penny Chaau Yan Poon, Santiago Toledo, WernerKaminsky, Martin L. Kirk, and Julie A. Kovacs. Water-soluble Fe(II)–H2O complex with aweak O–H bond transfers a hydrogen atom via an observable monomeric Fe(III)–OH.Journal of the American Chemical Society, 137(6):2253–2264, 2015. ISSN 0002-7863.doi:10.1021/ja5068405. URL http://dx.doi.org/10.1021/ja5068405. → pages 65[116] F. Albert Cotton and Jan M. Troup. Further refinement of the molecular structure of triirondodecacarbonyl. Journal of the American Chemical Society, 96(13):4155–4159, 1974.ISSN 0002-7863. doi:10.1021/ja00820a016. URL http://dx.doi.org/10.1021/ja00820a016.→ pages 65[117] F. Albert Cotton and Jan M. Troup. Accurate determination of a classic structure in themetal carbonyl field: nonacarbonyldi-iron. Journal of the Chemical Society, DaltonTransactions, (8):800–802, 1974. ISSN 0300-9246. doi:10.1039/DT9740000800. URLhttp://dx.doi.org/10.1039/DT9740000800. → pages 65[118] J. Me´ndez, R. Caillard, G. Otero, N. Nicoara, and J. A Martı´n-Gago. Nanostructuredorganic material: From molecular chains to organic nanodots. Advanced Materials, 18(15):2048–2052, 2006. ISSN 1521-4095. doi:10.1002/adma.200502115. URLhttp://dx.doi.org/10.1002/adma.200502115. → pages 71[119] A´lvarez Lucı´a, Pela´ez Samuel, Caillard Renaud, A. Serena Pedro, A. Martı´n-Gago Jose´, andMe´ndez Javier. Metal-organic extended 2D structures: Fe-PTCDA on Au(111).Nanonotechnology, 21(30):305703, 2010. ISSN 0957-4484. URLhttp://stacks.iop.org/0957-4484/21/i=30/a=305703. → pages 71108[120] Chi Chen, C. A. Bobisch, and W. Ho. Visualization of fermis golden rule through imagingof light emission from atomic silver chains. Science, 325(5943):981–985, 2009.doi:10.1126/science.1174592. URLhttp://www.sciencemag.org/content/325/5943/981.abstract. → pages 73, 96[121] E. Dulkeith, A. C. Morteani, T. Niedereichholz, T. A. Klar, J. Feldmann, S. A. Levi, F. C.J. M. van Veggel, D. N. Reinhoudt, M. Mo¨ller, and D. I. Gittins. Fluorescence quenching ofdye molecules near gold nanoparticles: Radiative and nonradiative effects. Phys. Rev. Lett.,89:203002, Oct 2002. doi:10.1103/PhysRevLett.89.203002. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.89.203002. → pages 73, 96[122] Yohan D. M. Champouret, Jean-Didier Marechal, Rajinder K. Chaggar, John Fawcett,Kuldip Singh, Feliu Maseras, and Gregory A. Solan. Factors affecting imine coordination in(iminoterpyridine)MX2 (M = Fe, Co, Ni, Zn): synthesis, structures, DFT calculations andethylene oligomerisation studies. New J. Chem., 31:75–85, 2007. doi:10.1039/B610562A.URL http://dx.doi.org/10.1039/B610562A. → pages 91[123] P.R. Andres and U.S. Schubert. New functional polymers and materials based on2,2:6,2-terpyridine metal complexes. Advanced Materials, 16(13):1043–1068, 2004. ISSN1521-4095. doi:10.1002/adma.200306518. URLhttp://dx.doi.org/10.1002/adma.200306518. → pages 95[124] Harald Hofmeier and Ulrich S. Schubert. Recent developments in the supramolecularchemistry of terpyridine-metal complexes. Chem. Soc. Rev., 33:373–399, 2004.doi:10.1039/B400653B. URL http://dx.doi.org/10.1039/B400653B. → pages 95[125] Jrme Fortage, Fausto Puntoriero, Fabien Tuyras, Grgory Dupeyre, Antonino Arrigo, IlariaCiofini, Philippe P. Lain, and Sebastiano Campagna. Photoinduced electron transfer inOs(terpyridine)-biphenylene-(bi)pyridinium assemblies. Inorganic Chemistry, 51(9):5342–5352, 2012. doi:10.1021/ic300297p. URL http://dx.doi.org/10.1021/ic300297p.PMID: 22524304. → pages 96[126] Thomas Classen, Guido Fratesi, Giovanni Costantini, Stefano Fabris, Frank Louis Stadler,Cheolkyu Kim, Stefano de Gironcoli, Stefano Baroni, and Klaus Kern. Templated growth ofmetal–organic coordination chains at surfaces. Angewandte Chemie, 117(38):6298–6301,2005. ISSN 1521-3757. doi:10.1002/ange.200502007. URLhttp://dx.doi.org/10.1002/ange.200502007. → pages 97[127] L. Gao, W. Ji, Y. B. Hu, Z. H. Cheng, Z. T. Deng, Q. Liu, N. Jiang, X. Lin, W. Guo, S. X.Du, W. A. Hofer, X. C. Xie, and H.-J. Gao. Site-specific Kondo effect at ambienttemperatures in iron-based molecules. Phys. Rev. Lett., 99:106402, Sep 2007.doi:10.1103/PhysRevLett.99.106402. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.99.106402. → pages 97[128] Emi Minamitani, Noriyuki Tsukahara, Daisuke Matsunaka, Yousoo Kim, Noriaki Takagi,and Maki Kawai. Symmetry-driven novel Kondo effect in a molecule. Phys. Rev. Lett., 109:086602, Aug 2012. doi:10.1103/PhysRevLett.109.086602. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.109.086602. → pages 97[129] Y. Yayon, V. W. Brar, L. Senapati, S. C. Erwin, and M. F. Crommie. Observing spinpolarization of individual magnetic adatoms. Phys. Rev. Lett., 99:067202, Aug 2007.109doi:10.1103/PhysRevLett.99.067202. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.99.067202. → pages 97[130] David J Michaelis. Iron as a powerful catalysts for transition metal-catalyzed reactions.Angew. Chem. Int. Ed, 47:3317–3321, 2008. → pages 97[131] Ingmar Bauer and Hans-Joachim Kno¨lker. Iron catalysis in organic synthesis. ChemicalReviews, 115(9):3170–3387, 2015. doi:10.1021/cr500425u. URLhttp://dx.doi.org/10.1021/cr500425u. → pages 97[132] JeongYong Lee, Omar K. Farha, John Roberts, Karl A. Scheidt, SonBinh T. Nguyen, andJoseph T. Hupp. Metal-organic framework materials as catalysts. Chem. Soc. Rev., 38:1450–1459, 2009. doi:10.1039/B807080F. URL http://dx.doi.org/10.1039/B807080F. →pages 97[133] Lianpeng Tong, Ruifa Zong, Rongwei Zhou, Nattawut Kaveevivitchai, Gang Zhang, andRandolph P. Thummel. Ruthenium catalysts for water oxidation involving tetradentatepolypyridine-type ligands. Faraday Discuss., pages –, 2015. doi:10.1039/C5FD00051C.URL http://dx.doi.org/10.1039/C5FD00051C. → pages[134] Prashant Kumar, Ashish Kumar Singh, Rampal Pandey, and Daya Shankar Pandey.Bio-catalysts and catalysts based on ruthenium(ii) polypyridyl complexes impartingdiphenyl-(2-pyridyl)-phosphine as a co-ligand. Journal of Organometallic Chemistry, 696(22):3454 – 3464, 2011. ISSN 0022-328X.doi:http://dx.doi.org/10.1016/j.jorganchem.2011.06.031. URLhttp://www.sciencedirect.com/science/article/pii/S0022328X11004153. → pages 97[135] Bas Hulsken, Richard Van Hameren, Jan W. Gerritsen, Tony Khoury, Pall Thordarson,Maxwell J. Crossley, Alan E. Rowan, Roeland J. M. Nolte, Johannes A. A. W. Elemans, andSylvia Speller. Real-time single-molecule imaging of oxidation catalysis at a liquid-solidinterface. Nat Nano, 2(5):285–289, 05 2007. URLhttp://dx.doi.org/10.1038/nnano.2007.106. → pages 97[136] Flemming Besenbacher, Jeppe V. Lauritsen, and Stefan Wendt. STM studies of modelcatalysts. Nano Today, 2(4):30 – 39, 2007. ISSN 1748-0132.doi:http://dx.doi.org/10.1016/S1748-0132(07)70115-9. URLhttp://www.sciencedirect.com/science/article/pii/S1748013207701159. → pages[137] M. Rler, P. Geng, and J. Wintterlin. A high-pressure scanning tunneling microscope forstudying heterogeneous catalysis. Review of Scientific Instruments, 76(2):023705, 2005.doi:http://dx.doi.org/10.1063/1.1841951. URLhttp://scitation.aip.org/content/aip/journal/rsi/76/2/10.1063/1.1841951. → pages 97[138] John P. Perdew, Kieron Burke, and Matthias Ernzerhof. Generalized gradient approximationmade simple. Physical Review Letters, 77(18):3865–3868, 1996. URLhttp://link.aps.org/doi/10.1103/PhysRevLett.77.3865. PRL. → pages 112[139] P. E. Blo¨chl. Projector augmented-wave method. Physical Review B, 50(24):17953–17979,1994. URL http://link.aps.org/doi/10.1103/PhysRevB.50.17953. PRB. → pages 112110[140] G. Kresse and D. Joubert. From ultrasoft pseudopotentials to the projector augmented-wavemethod. Physical Review B, 59(3):1758–1775, 1999. URLhttp://link.aps.org/doi/10.1103/PhysRevB.59.1758. PRB. → pages 112[141] G. Kresse and J. Furthmu¨ller. Efficient iterative schemes for ab initio total-energycalculations using a plane-wave basis set. Physical Review B, 54(16):11169–11186, 1996.URL http://link.aps.org/doi/10.1103/PhysRevB.54.11169. PRB. → pages 112[142] G. Kresse and J. Furthmu¨ller. Efficiency of ab-initio total energy calculations for metals andsemiconductors using a plane-wave basis set. Computational Materials Science, 6(1):15–50,1996. ISSN 0927-0256. doi:http://dx.doi.org/10.1016/0927-0256(96)00008-0. URLhttp://www.sciencedirect.com/science/article/pii/0927025696000080. → pages 112[143] B. Hammer, L. B. Hansen, and J. K. Nørskov. Improved adsorption energetics withindensity-functional theory using revised Perdew-Burke-Ernzerhof functionals. PhysicalReview B, 59(11):7413–7421, 1999. URLhttp://link.aps.org/doi/10.1103/PhysRevB.59.7413. PRB. → pages 112[144] Stefan Grimme. Semiempirical gga-type density functional constructed with a long-rangedispersion correction. Journal of Computational Chemistry, 27(15):1787–1799, 2006. ISSN1096-987X. doi:10.1002/jcc.20495. URL http://dx.doi.org/10.1002/jcc.20495. → pages 112[145] Zhi-Xin Hu, Haiping Lan, and Wei Ji. Role of the dispersion force in modeling theinterfacial properties of molecule-metal interfaces: adsorption of thiophene on coppersurfaces. Scientific Reports, 4:5036, 2014. ISSN 2045-2322. doi:10.1038/srep05036. URLhttp://www.ncbi.nlm.nih.gov/pmc/articles/PMC4030267/. 24849493[pmid] Sci Rep. →pages 112[146] KW Hipps. Scanning tunneling spectroscopy (STS). Springer, 2006. → pages 113111Appendix ADensity Functional Theory Methods forOn-Surface CalculationsDensity Functional Theory (DFT) simulations on the Fe-TPPT coordinated structures including theinfluence of the Ag(111) surface were performed by our collaborators Chenguang Wang and WeiJi. These calculations were carried out using the General Gradient Approximation (GGA) for theexchange-correlation potentials [138], the Projector Augmented Waves (PAW) method [139, 140],and a plane-wave basis set as implemented in the Vienna Ab-initio Simulation Package (VASP)[141, 142]. Revised Perdew-Burke-Ernzerhof (RPBE) [143], with the second version of Grimme’sdispersion corrections [144], was adopted throughout all calculations. For the method above, van derWaals interactions were considered by the DFT-D2 level, which is known to give a better descriptionof geometries and corresponding energies than those from the standard DFT [145]. The kineticenergy cutoff for the plane wave basis was set to 400 eV in configuration optimizations and increasedto 600 eV for energy calculations. A supercell (8 x 7) consisting of 224 Ag atoms in 4 layers withat least 15 A˚ vacuum region was employed to model the chain configuration on the Ag(111) surfaceand an (11 x 7) one consisting of 308 Ag atoms in 4-layers for isolated molecules. The surfacesBrillouin zones were sampled using the gamma point only in geometry optimizations and (3 x 3x1) in energy calculations. In geometry optimizations, all atoms except those at the bottom twoAg layers were fully relaxed until the residual force per atom was less than 0.02 eV/A˚. Tersoff-Hamann approximations [71] were employed to execute STM simulations in VASP. Isosurface levelused 10−5 electrons/r3Bohr. All simulated STM figures used biases consistent with the experimentalsettings.112Appendix BDetails on Data Analysis for ScanningTunneling Spectroscopy MeasurementsScanning Tunneling Spectroscopy (STS) measurements were performed in two ways, as describedin detail in section 2.2. One consisted of doing STS measurements in single locations on the sample(single point spectroscopy), and the other one of performing grid measurements on a large area (onthe order of tens of nm2), with a spatial resolution of ∼1.9pm/pixel for single molecules and of∼7.8pm/pixel for long chains.Single point spectroscopy measurements were performed by sweeping the bias forwardand backward several times consecutively (usually around 10 times) in order to be able to aver-age over many repetitions of the same, identical measurement. Depending on the size of the biasrange the resolution in energy varied from about 7.8mV to 10.3mV, well above the thermal limit of∼2.2mV at 4K [146]. Both forward and backward I(V ) curves should be zero at zero bias voltage.However, when that was not the case because of instrumental errors, both the forward and backwardI(V ) were corrected by adding (or subtracting) independently a current offset such that they wouldboth be zero at zero bias. Once this offset correction was done, the data were ready to be dulyanalyzed using my own Matlab codes (also the offset correction was done in Matlab). First, the av-eraged I(V ) was smoothed in order to obtain cleaner spectra, but paying particular care not to looseor alter the information contained in the measurements during the smoothing process. For smooth-ing the data I used the Matlab function “smooth” that works as a lowpass filter with filter coefficientequal to the reciprocal of the span, which was chosen to be 11. Since molecular features are usuallyquite broad in energy, they were not significantly altered by this choice of the size of the span [figureB.1]. Then, dI/dV and (dI/dV )/(I/V ) (normalized dI/dV ) were obtained by taking the numericalderivative (and numerically dividing by (I/V ) to normalize) of the corrected and smoothed I(V ).The normalization on the dI/dV (see equation (2.8)) was applied in order to reduce the strong ex-ponential contribution due to the transmission factor (see section 2.2 for more details) [figure B.1][81–84]. A byproduct of the normalization process is a divergence of the (dI/dV )/(I/V ) aroundzero bias because of the division by zero. Since this is an artifact of the normalization in every113normalized STS spectrum the region around zero bias [circled cyan dashed area in figure B.1] wasremoved and not considered.Figure B.1: Comparison of the I(V ) (left), dI/dV (center), (dI/dV )/(I/V ) (right) curves ob-tained from raw (top) and smoothed data with span equal to 11 (bottom). Areas circledwith cyan and green dashed lines are artifacts of the data analysis process. Note: Thedivergence in the smoothed data plot is bigger than in the raw data one. This is a resultof the smoothing process. Different values of the smoothing span results in differentamplitudes of the divergence.When grid measurements were performed the bias was swept back and forth over thebias range of interest only once on each point (because of a software limitation). The result of a gridmeasurement is a matrix containing forward and backward tunneling current values I(x,y,Vbias)for each Vbias at each (x,y) point in space. These data were treated similarly to the single pointspectroscopy ones. First, forward and backward I(V ) values were offset-corrected from possibleinstrumental errors, then averaged, and finally they were smoothed. Here, for the smoothing I usedthe Matlab function “smooth3” with a box as convolution kernel of size 5 for the x and y directionsand of size 11 for the bias [figure B.2]. Finally, dI/dV and (dI/dV )/(I/V ) (normalized dI/dV )were obtained by taking the numerical derivative (and numerically dividing by (I/V ) to normalize)of the corrected and smoothed I(x,y,Vbias) [figure B.2].The normalized dI/dV curves shown in chapters 3 and 5 were obtained mostly from gridmeasurements as a result of spatially averaging over the I values contained in specific regions (theones enclosed by the corresponding colour outlines in the topography images) [figure B.3(c)]. Theseregions, containing all the I(x,y) values to be considered, were determined from the topographyimage corresponding to the grid measurement. This was done by selecting first a rectangular areaaround the Region Of Interest (ROI) [e.g. a single molecule, a whole chain, a small area on the baresilver, or a molecular axis like in figure B.3(a)] and then by setting a minimum threshold value for the114Figure B.2: Comparison of the I(V ) (left), dI/dV (center), (dI/dV )/(I/V ) (right) STS mapsof an isolated TPPT on Ag(111) at Vbias = 1.6V, obtained from raw (top) and smootheddata with kernel of size 5 for the x and y directions and of size 11 for the bias (bottom).apparent height of the topography [figure B.3(b)]. This threshold was used to make sure to consideronly the points (x,y) in the selected rectangular region that were actually of interest (e.g. only pointsbelonging to the TPPT molecular axis and no contributions from the surrounding substrate). Then,all the current values corresponding to the (x,y) points satisfying the two above conditions, wereadded together and the total divided by the number of considered points in order to get a spatiallyaveraged value of I for each bias voltage. Depending on the target of interest the number of averagedpoints varied, from about 460 to 4230 for whole complexes (such as whole chains or molecules) andfrom about 770 to 3700 for the regions around the molecular axis of the molecules (generally, gridmeasurements consisted of more than 16000 points). Numerical dI/dV and (dI/dV )/(I/V ) (theFigure B.3: Schematic of the method used to determine the ROI. (a) First, select a rectangulararea (dashed green rectangle) around the ROI (in this case the ROI is the molecular axisof a single TPPT molecule). (b) Second, set a threshold value for the apparent heightof the topography (here, 54 pm). (c) Obtain the ROI, such as the all the (x,y) pointssatisfying the two conditions and discard all the others.115ones shown in this thesis) were then obtained from these average values. Energy maps were insteadobtained from the smoothed I(x,y,Vbias) matrices by numerically calculating the dI/dV (x,y,Vbias)and the normalized counterpart (dI/dV (x,y,Vbias))/(I/V ) (the ones presented in this thesis), andthen plotting the values for a specific bias (Vbias) as a function of (x,y).Both single point spectroscopy and grid measurements on similar targets (e.g. single bareTPPT, single coordinated molecules, long chains) were repeated several times, in different samplepreparations, and with different tips in order to corroborate the results [see tables C.1-C.8 in ap-pendix C]. In the main text of this thesis the most relevant results are shown as representatives ofall the others. Additional spectroscopy measurement results consistent with the ones shown in themain text, are reported in appendix C.116Appendix CAdditional Spectroscopy MeasurementsResultsIn this appendix additional examples of spectroscopy measurements on a single bare ligand anddifferent coordinated structures are reported.The tables below contain lists of all the grid measurements (and corresponding filenames)performed on the different structures between 2013 [tables C.1 and C.2] and 2014-2015 [tables C.3- C.8].Table C.1: Grid measurements on bare TPPT molecules and coordinated structures onAg(111) performed in 2013.Single bare molecule on Ag(111)Grid label Set-point parameters Bias range No. of pixel Size Comments(Vbias, It) [V] [nm2]1b2013 -2V, 50pA -2 - +2.8 512 x 512 5 x 5 Huge drift2b2013 -2.5V, 50pA -2.5 - +2.7 512 x 512 5 x 5 Drift3b2013 -2.5V, 50pA -2.5 - +2.8 512 x 512 5 x 5 Some drift but okayCoordinated structures - two coordination centers at the end of a chain1c2013 -2.5V, 50pA -2.5 - +2.5 512 x 512 6 x 6 Drift2c2013 -2.5V, 500pA -2.5 - +1 512 x 512 6 x 6 Initial tip change3c2013 -2V, 3nA -2 - +2 512 x 512 6 x 6 Drift4c2013 -2V, 20nA -2 - +2.5 512 x 512 6 x 6 Tip change5c2013 -3V, 1nA -3 - +3 512 x 512 6 x 6 Chain ”broke”117Table C.2: Matching of the grid label with the grid filename of the measurements done on bareTPPT molecules and coordinated structures on Ag(111) in 2013.Grid label Grid filename1b2013 20130425-Z53-IV252b2013 20130428-Z14-IV63b2013 20130501-Z74-IV1291c2013 20130524-Z51-IV642c2013 20130525-Z61-IV933c2013 20130527-Z32-IV1004c2013 20130528-Z54-IV2295c2013 20130530-Z66-IV86118Table C.3: Grid measurements on bare TPPT molecules on Ag(111) and NaCl/Ag(111) per-formed in 2014-2015. “*” indicates the grid measurements reported in the main text,while “@” the ones reported in this appendix.Single bare molecule on Ag(111)Grid label Set-point parameters Bias range No. of pixel Size Comments(Vbias, It) [V] [nm2]1b2014 -0.2V, 5pA -0.2 - +2.8 512 x 512 5 x 5 Switched2b2014 -0.5V, 10pA -0.5 - +2.8 512 x 512 5 x 5 Switched, tilted3b2014 -2.5V, 50pA -2.5 - +2.8 512 x 512 5 x 5 Switched, tilted5b2014 +2.7V, 100pA +2.7 - -2.5 512 x 512 5 x 5 Tilted6b2014 -2.5V, 50pA -2.5 - +2.8 512 x 512 5 x 5 Good7b2014* -2.5V, 50pA -2.5 - +2.8 512 x 512 5 x 5 Good8b2014 -2.5V, 50pA -2.5 - +2.8 512 x 512 5 x 5 Good onswitched TPPT9b2014 +2.7V, 50pA +2.7 - -2.5 512 x 512 5 x 5 Good10b2014@ -2.5V, 50pA -2.5 - +2.7 512 x 512 5 x 5 Good11b2014 +2.7V, 50pA +2.7 - -2.5 512 x 512 5 x 5 Good12b2014 +2.7V, 50pA +2.7 - -2.5 512 x 512 5 x 5 Good15b2014 -2.5V, 1nA -2.5 - +2 512 x 512 4 x 4 Good (slight creep)16b2014 +2V, 1nA +2 - -2.5 512 x 512 4 x 4 Good17b2014 -2.5V, 25pA -2.5 - +2 512 x 512 4 x 4 Good, low current18b2014 -2V, 500pA -2 - +2 512 x 512 4.5 x 4.5 Good (slight creep)Two-adjacent molecules on Ag(111)4b2014 -2.5V, 50pA -2.5 - +2.8 512 x 512 6 x 6 Top TPPTswitched, tilted19b2014* -2V, 400pA -2 - +2 512 x 512 6 x 6 GoodThree-adjacent molecules on Ag(111)21b2014 -2.5V, 50pA -2.5 - +2.8 512 x 512 6 x 6 Good13b2014* -2.5V, 50pA -2.5 - +2.7 512 x 512 6 x 6 Good14b2014 +2.7V, 50pA +2.7 - -2.5 512 x 512 6 x 6 GoodSingle bare molecule on NaCl/Ag(111)1bs2015 -2V, 0.6pA -2 - +2 512 x 512 4 x 4 Very low current2bs2015* -2V, 0.6pA -2 - +2 512 x 512 4 x 4 Good119Table C.4: Matching of the grid label with the grid filename of the measurements done onbare TPPT molecules on Ag(111) and NaCl/Ag(111) in 2014/2015. “*” indicates the gridmeasurements reported in the main text, while “@” the ones reported in this appendix.Grid label Grid filename1b2014 20140130-Z35-IV462b2014 20140131-Z7-IV1303b2014 20140203-Z18-IV254b2014 20140206-Z112-IV3075b2014 20140209-Z35-IV326b2014 20140211-Z48-IV30-17b2014* 20140212-Z58-IV388b2014 20140212-Z63-IV459b2014 20140213-Z12-IV11-1/210b2014@ 20140214-Z31-IV2011b2014 20140215-Z10-IV912b2014 20140216-Z12-IV613b2014* 20140218-Z49-IV3214b2014 20140219-Z8-IV515b2014 20140402-Z46-IV3516b2014 20140403-Z24-IV1017b2014 20140404-Z48-IV30-118b2014 20140405-Z57-IV5219b2014* 20140406-Z65-IV8221b2014 20140211-Z48-IV301bs2015 20150923-Z37-IV1062bs2015* 20150924-Z105-IV318120Table C.5: Grid measurements on whole coordinated structure on Ag(111) performed in 2014.“*” indicates the grid measurements reported in the main text, while “@” the ones reportedin this appendix.“One-molecule chain” with both ends coordinatedGrid name Set-point parameters Bias range No. of pixel Size Comments(Vbias, It) [V] [nm2]13c2014 -2.5V, 200pA -2.5 - +2.7 512 x 512 5 x 5 Switched14c2014 -2.5V, 200pA -2.5 - +2.5 512 x 512 5 x 5 Good16c2014 -2.5V, 100pA -2.5 - +2.5 512 x 512 4.5 x 4.5 Switched17c2014@ -2V, 1nA -2 - +2 512 x 512 4.5 x 4.5 Good18c2014* -2V, 1nA -2 - +2 512 x 512 4 x 4 GoodTwo-molecule chain1c2014 -2.5V, 300pA -2.5 - +2.5 512 x 512 7 x 7 SwitchedThree-molecule chain7c2014 -2.5V, 1nA -2.5 - +2 512 x 512 9 x 9 GoodFour-molecule chain26c2014* -2.5V, 2.5nA -2.5 - +2 512 x 128 12 x 3 GoodFive-molecule chain28c2014* -2.5V, 1.5nA -2.5 - +2 512 x 128 16 x 4 Tip not good29c2014 -2.5V, 3nA -2.5 - +2 512 x 128 16 x 4 Good30c2014 -2.5V, 25pA -2.5 - +2.5 512 x 128 16 x 4 GoodSix-molecule chain25c2014@ -2.5V, 2.5nA -2.5 - +2 512 x 128 16 x 4 GoodTable C.6: Matching of the grid label with the grid filename of the measurements done onwhole coordinated structure on Ag(111) in 2014. “*” indicates the grid measurementsreported in the main text, while “@” the ones reported in this appendix.Grid label Grid filename1c2014 20140313-Z14-IV177c2014 20140320-Z74-IV8513c2014 20140326-Z68-IV13214c2014 20140327-Z17-IV2316c2014 20140329-Z29-IV917c2014@ 20140401-Z72-IV4618c2014* 20140410-Z110-IV16525c2014@ 20140417-Z131-IV14026c2014* 20140419-Z50-IV7628c2014* 20140425-Z57-IV5229c2014 20140426-Z98-IV16630c2014 20140427-Z102-IV170121Table C.7: Grid measurements on different types of coordinated structure on Ag(111) per-formed in 2014. “*” indicates the grid measurements reported in the main text, while “@”the ones reported in this appendix.Three molecules and two coordination centers inside of a chainGrid name Set-point parameters Bias range No. of pixel Size Comments(Vbias, It) [V] [nm2]2c2014 -2.5V, 50pA -2.5 - +2.7 512 x 512 7 x 7 Switched6c2014 +2.5V, 20pA +2.5 - -2.5 512 x 512 7 x 7 Switched8c2014 -2.5V, 1nA -2.5 - +2 512 x 256 5 x 3 Tip-change15c2014@ -2.5V, 100pA -2.5 - +2.5 512 x 512 5 x 5 1 TPPT switched21c2014 -2V, 800pA -2 - +2 512 x 256 6 x 3 Good22c2014 -2.5V, 100pA -2.5 - +2.5 512 x 256 6 x 3 Switched23c2014 -2.5V, 25pA -2.5 - +2.5 512 x 256 6 x 3 Good27c2014* -2.5V, 2.5nA -2.5 - +2 512 x 512 5 x 5 1 TPPT in chain, goodThree molecules and two coordination centers at the end of a chain - coordinated end3c2014* -2.5V, 50pA -2.5 - +2.5 512 x 512 7 x 7 2 Switched4c2014 +2.5V, 10pA +2.5 - -2.5 512 x 512 7 x 7 Switched5c2014 +2.5V, 5pA +2.5 - -2.5 512 x 512 7 x 7 Good but low signal9c2014 +2V, 100pA +2 - -2.5 512 x 512 7 x 7 2 Chain broken19c2014@ -2V, 800pA -2 - +2 512 x 512 7 x 7 2 GoodThree molecules and two coordination centers at the end of a chain - non-coordinated end10c2014* -2.5V, 1nA -2.5 - +2 512 x 512 7 x 7 Good11c2014 -2.5V, 200pA -2.5 - +2 512 x 512 7 x 7 Small switch?12c2014@ -2.5V, 200pA -2.5 - +2.5 512 x 512 7 x 7 2 Good122Table C.8: Matching of the grid label with the grid filename of the measurements done differ-ent types of coordinated structure on Ag(111) in 2014. “*” indicates the grid measure-ments reported in the main text, while “@” the ones reported in this appendix.Grid label Grid filename2c2014 20140314-Z32-IV233c2014* 20140315-Z15-IV33b2014 20140203-Z18-IV254c2014 20140316-Z42-IV115c2014 20140317-Z12-IV36c2014 20140318-Z97-IV288c2014 20140321-Z96-IV1699c2014 20140321-Z101-IV17210c2014* 20140324-Z25-IV7511c2014 20140324-Z29-IV9512c2014@ 20140325-Z36-IV9615c2014@ 20140328-Z20-IV519c2014@ 20140411-Z32-IV6121c2014 20140414-Z27-IV6822c2014 20140415-Z36-IV8623c2014 20140416-Z35-IV327c2014* 20140421-Z67-IV90123C.1 Single bare ligandFigure C.1 shows normalized dI/dV curves of a single bare ligand. Both curves (whole moleculeand molecular axis) present two clear peaks at +1.15V and +2.3V but no clear features at negativebias.Figure C.1: Normalized dI/dV curves of a single bare molecule obtained by averaging overthe regions enclosed by the corresponding colour outlines in the topography (see in-set). Red: whole single molecule; blue: single molecule molecular axis; dashed black:Ag(111), reference spectrum. Grid set-point parameters: Vbias = -2.50V, It = 50pA. Note:the divergence near zero-bias created by normalization is removed.124Figure C.2 shows the STS energy maps at different biases of the same single bare ligandas figure C.1. The pattern of the electron density distribution observed is consistent with what wasdescribed previously in section 3.3.1 for another single TPPT molecule. Generally, the electronicstates are located on different well-separated positions on the molecule.Figure C.2: STS energy maps at different biases showing the density of states of a single bareligand. From left to right, from top to bottom the biases are: -1.6V, -0.2V, +0.4V, +0.7V,+1.15V, +1.6V, +1.8V, +2.0V, 2.3V and +2.9V. Grid set-point parameters: Vbias = -2.50V,It = 50pA.125C.2 Coordinated structuresC.2.1 “One-molecule chain”Figures C.3 and C.4 show the normalized dI/dV curves and the STS energy maps at different biasesof a “1-molecule chain” with both its tpy ends coordinated with Fe, respectively.Both the STS curves in figure C.3 show a shoulder around +1.5V, which is more prominent in thered curve corresponding to the coordinated ligand’s molecular axis. No obvious features can beobserved at negative bias.Figure C.3: (a) Normalized dI/dV curves of a “1-molecule chain” obtained by averaging overthe regions enclosed by the corresponding colour outlines (see inset). Purple: whole“one-molecule chain”; red: molecular axis; dashed black: Ag(111), reference spectrum.Grid set-point parameters: Vb = -2.0V It = 1nA. Note: the divergence near zero-biascreated by normalization is removed.126The electron density distributions of figure C.4 show the same pattern described in section5.1 for another “1-molecule chain” . At -0.1V and between +1.8V and +2V the electronic states areon the metal, whereas at -0.5V and between +0.4V and +1.65V they are located about the ligand’saxis.Figure C.4: STS energy maps at different biases showing the density of states of a “1-moleculechain” with both the tpy ends coordinated with Fe. From left to right, from top to bottomthe biases are: -0.5V, -0.1V, +0.4V, +0.8V, +1.2V, +1.5V, +1.65V, +1.8V and +2V. Gridset-point parameters: Vb = -2.0V It = 1nA.127C.2.2 Long chain: six-molecule chainThe electronic structure of a six-molecule chain with one end-molecule Fe-coordinated (right end)and one end non-coordinated (left end) is presented in this section. Figure C.5 shows the normalizeddI/dV curves relative to the whole chain and the molecular axes of each of the TPPT. The STS of allthe molecules coordinated on both ends with Fe present a shoulder around +1.5V, while the one ofthe left non-coordinated edge-molecule at∼+1.2V. These results are consistent with what discussedpreviously in chapter 5.Figure C.5: Normalized dI/dV curves of a six-molecule chain obtained by averaging overthe regions enclosed by the corresponding colour outlines (see inset). Purple: wholechain; green, cyan, orange, violet, red and blue: molecular axes; dashed black: Ag(111),reference spectrum. Bars: green, first resonance of bare ligand on silver; grey, on NaCl.On top of the inset a sketch of the chain. Grid set-point parameters: Vb = -2.50V It =2.5nA. Note: the divergence near zero-bias created by normalization is removed.128Figure C.6 shows some energy maps of the six-molecule chain. Negative bias maps arenot reported because they do not show any clear feature, probably due to noise or a tip effect. Atpositive energies the observed pattern of the electron distributions is consistent with what was shownpreviously in section 5.2 for the four- and five-molecule chains.Figure C.6: STS energy maps at different biases showing the density of states of a six-molecule chain. From top to bottom the biases are: +0.8V +1.2V, +1.5V, +1.65V and+2V. Grid set-point parameters: Vb = -2.50V It = 2.5nA.129C.2.3 Detail of a six-molecule chain: three molecules in the middle of the chainIn this section three TPPT molecules inside of a six-molecule chain are analyzed. In figure C.7their normalized STS are presented. The topography image in figure C.7 shows that two tip changesoccurred during the grid measurement, one at the bottom of the chain and the other one close the top.Each STS curve presents a first visible resonance ∼+1.5V, while no clear features were observed atnegative bias. The periodic jumps visible in all the STS between -1V and +1.2V are probably dueto noise or to an unstable tip.Figure C.7: Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines (see inset).Purple: whole system; blue, violet and red: molecular axes; dashed black: Ag(111),reference spectrum. Bars: green, first resonance of bare ligand on silver; grey, on NaCl.Grid set-point parameters: Vb = -2.50V It = 100pA. Note: the divergence near zero-biascreated by normalization is removed.130Figure C.8 shows the energy maps of the three molecules within the six-molecule chain.Their electron distribution pattern is consistent with previous results, including at +2.5V where thedensity of states from the metal spread back towards the ligand. Below the Fermi level no clearfeatures were observed neither on the metal nor on the molecular axes. This is probably due to a tipeffect or a noisy measurement, as observed from the STS in figure C.7.Figure C.8: STS energy maps at different biases showing the density of states of two coordi-nation sites and three molecular axes. From left to right, from top to bottom the biasesare: -0.1V, +0.4V, +0.8V, +1.2V, +1.5V, +1.65V, +1.8V, +2V, and +2.5V. Grid set-pointparameters: Vb = -2.50V It = 100pA.131C.2.4 Detail of a three-molecule chain and its end-molecule non-coordinated withFeIn this section the electronic structure [figures C.9 and C.10] of a three-molecule chain with one end-molecule non-coordinated with Fe and the other end coordinated but non considered, is presented.The STS of the non-coordinated end-molecule shows a first resonance at about +1.2V, while theones of the other two TPPTs have a peak at ∼+1.5V, slightly shifted with respect to each other[figure C.9]. In all the STS between -0.8V and +0.8V a series of periodic jumps are visible. Theseare probably due to noise or an unstable tip.Figure C.9: Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines (see inset).Purple: whole structure; red, violet and blue: molecular axes; dashed black: Ag(111),reference spectrum. Bars: green, first resonance of bare ligand on silver; grey, on NaCl.Grid set-point parameters: Vb = -2.50V It = 200pA. Note: the divergence near zero-biascreated by normalization is removed.132The energy maps at positive biases of figure C.10 show the same pattern described insection 5.4. Fully coordinated molecules behave as any TPPT molecule with both ends coordinatedwith Fe, and the edge non-coordinated molecule shows an electron distribution similar to a bareligand at its left non-coordinated side, and similar to a coordinated TPPT at its right coordinatedend. At -0.1V the signal about the metal centers is clearly weak, which is not consistent with whatshown in chapter 5. This could be the result of a tip effect or it could also be due to noise.Figure C.10: STS energy maps at different biases showing the density of states of of twocoordination sites, three molecules, one of which is an edge-molecule not coordinatedwith Fe. From left to right, from top to bottom the biases are: -0.1V, +0.4V, +0.8V,+1.2V, +1.5V, +1.65V, +1.8V, +2V, and +2.5V. Grid set-point parameters: Vb = -2.5VIt = 200pA.133C.2.5 Detail of a four-molecule chain and one end-molecule coordinated with FeThe case presented in this section is similar to the one discussed in section 5.5: three fully coor-dinated molecules part of a four-molecule chain. All the normalized dI/dV curves in figure C.11show a shoulder around +1.5V and no obvious features at negative bias.Figure C.11: Normalized dI/dV curves of the structure sketched on top of the inset obtainedby averaging over the regions enclosed by the corresponding colour outlines (see inset).Purple: whole structure; blue, violet and red: molecular axes; dashed black: Ag(111),reference spectrum. Bars: green, first resonance of bare ligand on silver; grey, on NaCl.Grid set-point parameters: Vb = -2.0V It = 800pA. Note: the divergence near zero-biascreated by normalization is removed.134The density of states pattern at different biases shown in figure C.12 is found to be gener-ally pretty consistent with previous results. However, the two metal linkages do not show the exactsame behavior, occurrence which was rarely observed. This could be caused by an electronic effectdue to the interaction with the substrate; a geometric effect due to the adsorption and orientation ofthe chain on the surface; or even due to a CO or water molecule adsorbed on the silver substrateunderneath one of the linkages and maybe interacting with the Fe itself.Figure C.12: STS energy maps at different biases showing the density of states of two coor-dination sites, three molecules, one of which is an edge molecule coordinated with Fe.From left to right, from top to bottom the biases are: -0.6V, -0.1V, +0.4V, +0.8V, +1.2V,+1.5V, +1.65V, +1.8V and +2V. Grid set-point parameters: Vb = -2.0V It = 800pA.135

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:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0305019/manifest

Comment

Related Items