Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Single molecule perspectives of model organic semiconductors : energy level mapping by high-resolution… Cochrane, Katherine Anne 2017

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

Item Metadata


24-ubc_2017_november_cochrane_katherine.pdf [ 112.07MB ]
JSON: 24-1.0354979.json
JSON-LD: 24-1.0354979-ld.json
RDF/XML (Pretty): 24-1.0354979-rdf.xml
RDF/JSON: 24-1.0354979-rdf.json
Turtle: 24-1.0354979-turtle.txt
N-Triples: 24-1.0354979-rdf-ntriples.txt
Original Record: 24-1.0354979-source.json
Full Text

Full Text

Single Molecule Perspectives of ModelOrganic Semiconductors:Energy Level Mapping by High-Resolution ScanningProbe MicroscopybyKatherine Anne CochraneB.Sc., McGill University, 2010A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinThe Faculty of Graduate and Postdoctoral Studies(Chemistry)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)August 2017© Katherine Anne Cochrane 2017AbstractOrganic semiconductors are a promising class of materials for many appli-cations such as photovoltaics, light emitting diodes, and field-effect tran-sistors. As these devices rely on the movement of charge at and near in-terfaces, understanding energy level alignment at these boundaries is essen-tial to improve device performance. Differences in the local environmentand surrounding molecular geometry have the potential to cause signif-icant energy level shifts occurring on single molecule length scales, thusaffecting device properties. Scanning Probe Microscopy is a family of tech-niques that allows investigation of materials on the molecular and submolec-ular level. Scanning Tunneling Spectroscopy (STS) allows for the map-ping of electronic states with spatial and energetic resolution. ElectrostaticForce Spectroscopic (EFS) mapping investigates the local charge distribu-tion of surfaces even down to submolecular resolution. We utilize these tech-niques to investigate the prototypical semiconductors PTCDA and CuPc onNaCl(2ML)/Ag(111).Nanoislands of PTCDA were examined with STS, revealing strong elec-tronic differences between molecules at the edges and those in the center,with energy level shifts of up to 400 meV. We attribute this to the change inelectrostatic environment at the boundaries of clusters, namely via polariza-tion of neighboring molecules. To further investigate the local electrostatics,we use EFS to probe the effect of adding charge to PTCDA molecules, bothisolated and within clusters. We found that the charging energy dependson the initial local charge distribution by spatially resolving the chargingevents with sub-molecular resolution.In order to investigate the influence of interface geometry, we use pixel-by-pixel STS of the prototypical acceptor/donor system PTCDA/CuPc. Weobserve shifting of the donor and acceptor states in opposite directions, in-dicating an equilibrium charge transfer between the two. Further, we findthat the spatial location of electronic states of both acceptor and donor isstrongly dependent on the relative positioning of both molecules in largerclusters. The observation of these strong shifts illustrates a crucial issue:interfacial energy level alignment can differ substantially from the bulk elec-iiAbstracttronic structure in organic materials. This has significant implications fordevice design, where energy level alignment strongly correlates to deviceperformance.iiiLay SummarySolar power is a promising alternative energy source. A new class of materi-als used to generate power is organic photovoltaics (OPVs). These materialshave several advantages over traditional solar panels made from silicon: theycan be flexible, semi-transparent, light-weight, and easier to make. However,current OPV devices are much less efficient than those made from silicon.In order to optimize the conversion of light into electricity, it is importantto understand the properties of these materials on the nanoscopic level.Scanning probe microscopy is a technique that allows us to “see” individ-ual atoms and molecules and learn which characteristics are key to deviceperformance. With this technique, we can also control the exact positionof molecules with respect to each other and see how different interactionscorrespond to different properties. Knowing what molecular properties willmaximize device efficiency will help scientists design better materials for thefuture.ivPrefaceA version of chapter 3 has been published in Nature Communications. Themeasurements were taken mostly by me, with help from Agustin Schiffrin,Tanya Roussy and Martina Capsoni. I analyzed the data and performedthe microelectrostatic calculations. I co-wrote the manuscript with SarahBurke and Agustin Schiffrin. (K.A. Cochrane, A. Schiffrin, T.S. Roussy, M.Capsoni, S.A. Burke. Pronounced polarization-induced energy level shiftsat boundaries of organic semiconductor nanostructures. Nat. Commun.6:8312 (2015).A version of chapter 4 is in preparation for submission. The data forchapter 4 was collected by myself and Bingkai Yuan as well as Tanya Roussy.I analyzed the data and wrote the preliminary manuscript with help fromSarah Burke.A version of chapter 5 has been submitted for publication. The data wascollected by myself and Tanya Roussy. I wrote the code used to analyze thescript with the help of Tanya and Erik Ma˚rsell. I wrote the manuscript.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixGlossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiii1 Introduction and Background . . . . . . . . . . . . . . . . . . 11.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Charge generation in organic semiconductors . . . . . . . . . 21.3 The bulk heterojunction . . . . . . . . . . . . . . . . . . . . 51.4 Small molecules for organic photovoltaics . . . . . . . . . . . 61.5 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.1 Principles of scanning probe microscopy . . . . . . . . . . . . 102.2 Scanning tunneling microscopy and spectroscopy (STM/STS) 132.2.1 Tunneling theory . . . . . . . . . . . . . . . . . . . . 132.2.2 Scanning tunneling microscopy (STM) . . . . . . . . 162.2.3 Scanning tunneling spectroscopy (STS) . . . . . . . . 172.3 Non-contact atomic force microscopy (NC-AFM) . . . . . . . 182.3.1 Tuning fork NC-AFM . . . . . . . . . . . . . . . . . . 212.3.2 High resolution NC-AFM with functionalized tips . . 232.3.3 Electrostatic force spectroscopy (EFS) . . . . . . . . 252.4 Description of apparatus . . . . . . . . . . . . . . . . . . . . 272.4.1 Preparation chamber . . . . . . . . . . . . . . . . . . 29viTable of Contents2.4.2 Imaging chamber . . . . . . . . . . . . . . . . . . . . 302.5 Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.5.1 Metallic substrates . . . . . . . . . . . . . . . . . . . 332.5.2 NaCl deposition and the salt/metal interface . . . . . 362.5.3 Organic molecules . . . . . . . . . . . . . . . . . . . . 393 Polarization Induced Energy Level Shifts in PTCDANanois-lands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 463.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.2.1 Sample preparation . . . . . . . . . . . . . . . . . . . 483.3 Isolated PTCDA on NaCl(2ML)/Ag(111) . . . . . . . . . . . 483.3.1 Adsorption of PTCDA . . . . . . . . . . . . . . . . . 483.3.2 Electronic states of isolated PTCDA on an NaCl bi-layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 503.4 PTCDA nanoislands on NaCl(2ML/Ag(111) . . . . . . . . . 533.4.1 Geometry of nanoclusters on NaCl bilayers . . . . . . 533.4.2 STS mapping of PTCDA nanoislands on NaCl bilayers 603.4.3 Gap maps . . . . . . . . . . . . . . . . . . . . . . . . 653.5 Polarization of PTCDA . . . . . . . . . . . . . . . . . . . . . 673.5.1 Microelectrostatic calculations . . . . . . . . . . . . . 673.5.2 Nanoisland polarization energies . . . . . . . . . . . . 703.6 PTCDA on multilayer NaCl . . . . . . . . . . . . . . . . . . 743.6.1 Isolated PTCDA on multilayer NaCl . . . . . . . . . 743.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 774 Mapping the Hubbard Energy of PTCDA . . . . . . . . . . 784.1 Introduction and background . . . . . . . . . . . . . . . . . . 784.2 Charging of molecules examined with EFS . . . . . . . . . . 804.3 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 814.4 PTCDA clusters . . . . . . . . . . . . . . . . . . . . . . . . . 814.4.1 High resolution NC-AFM . . . . . . . . . . . . . . . . 834.4.2 Spectroscopy of PTCDA nanoislands . . . . . . . . . 844.5 Isolated PTCDA on NaCl(2ML)/Ag(111) . . . . . . . . . . . 914.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 935 Energy Level Alignment of a Bimolecular Heterojunction 955.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 955.2 Isolated molecules . . . . . . . . . . . . . . . . . . . . . . . . 965.2.1 CuPc on NaCl(2ML)/Ag(111) . . . . . . . . . . . . . 96viiTable of Contents5.2.2 PTCDA on NaCl(2ML)/Ag(111) . . . . . . . . . . . . 995.3 Bimolecular system: CuPc and PTCDA . . . . . . . . . . . . 995.3.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . 995.3.2 Electronic structure . . . . . . . . . . . . . . . . . . . 1005.4 Varying molecular geometry . . . . . . . . . . . . . . . . . . 1045.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1086 Conclusions and Outlook . . . . . . . . . . . . . . . . . . . . . 1106.1 Open questions and future work . . . . . . . . . . . . . . . . 1116.1.1 Understanding the molecular states and excitationsobserved . . . . . . . . . . . . . . . . . . . . . . . . . 1116.1.2 Effects of the metal substrate, and stabilizing differentcharge states . . . . . . . . . . . . . . . . . . . . . . . 1126.1.3 Possibility to directly observe charge separation . . . 1136.1.4 Novel materials . . . . . . . . . . . . . . . . . . . . . 1156.1.5 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . 116Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118AppendicesA STS Normalization and Processing . . . . . . . . . . . . . . . 140A.1 Bias offset . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140A.2 Normalization offset . . . . . . . . . . . . . . . . . . . . . . . 140A.3 Smoothing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142A.4 Kappa maps of a dimer on NaCl(2ML)/Ag(111) . . . . . . . 144B EFS Fitting and Modeling . . . . . . . . . . . . . . . . . . . . 145B.1 Background and theory . . . . . . . . . . . . . . . . . . . . . 145B.2 Fitting of VCPD with charging events . . . . . . . . . . . . . 146viiiList of Figures1.1 Examples of niche applications of OPV materials: a. bag(Noon Solar)[10] b. transparent solar cell window coatings(Konakra)[11] c. an organic solar cell transparent within thevisible range (Zhao, et. al [12]) d. patio umbrella (Ham-macker) [13] and e. car windows (Heliatek)[14] . . . . . . . . 11.2 Energy diagram of a donor/acceptor photovoltaic system de-picting the pathway of an electron and a hole upon absorp-tion of a photon from the highest occupied molecular or-bital (HOMO) to the lowest unoccupied molecular orbital(LUMO), forming an exciton. . . . . . . . . . . . . . . . . . . 31.3 Possible pathways of an exciton in a donor/acceptor system. . 41.4 Schematic of the energy levels of an organic material with theionization potential and electron affinity; optical, transport,and band gaps; and exciton binding energy labeled. . . . . . . 51.5 A simplified schematic of a device based on a bulk hetero-junction configuration. . . . . . . . . . . . . . . . . . . . . . . 61.6 A small sampling of the many conjugated organic moleculesused for OPV. Two molecules that form an acceptor/donorjunction, PTCDA (3,4,9,10-perylenetetracarboxylic dianhy-dride) and CuPc (copper phthalocyanine), will be the focusof this thesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.7 Two possible interfaces between a donor and an acceptor withsubtle changes in geometry. How do the energy levels changewith these two interfacial geometries? . . . . . . . . . . . . . 92.1 Scale of various microscopy techniques with correspondingexamples for scale reference. At the smallest scales for each,significant investment along with specialized equipment andexpertise are required to achieve the highest spatial resolution. 112.2 Operation of a scanning tunneling microscope in a. constantheight and b. constant current mode. . . . . . . . . . . . . . 12ixList of Figures2.3 Schematic of the scanning control of an STM. . . . . . . . . . 122.4 Schematic of two wave functions tunneling through a barrier. 132.5 Schematic showing the relation between energy levels andSTS measurements. The third panel indicates the spatiallyresolved pixel-by-pixel STS. The black line represents a pointspectrum occurring at pixel in the (x,y) plane through allenergy slices. . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.6 Relative contributions of the dominant forces between the tipand sample. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.7 Diagram of two main types of AFM: a. beam deflection andb. tuning fork. . . . . . . . . . . . . . . . . . . . . . . . . . . 222.8 Photograph of a quartz tuning fork sensor used in experi-ments in this thesis. This sensor was provided by OmicronNanotechnology, however homebuilt sensors were also used. . 222.9 Schematics of the functionalized tips used in this thesis: a.carbon monoxide b. PTCDA functionalized tip, showing pro-posed molecular geometry.[104] . . . . . . . . . . . . . . . . . 242.10 Example df(z) spectrum of a metal tip functionalized with aCO molecule (∆z = 1 nm). . . . . . . . . . . . . . . . . . . . 242.11 Examples of submolecular imaging of a 4-molecule PTCDAnanoisland with: constant height NC-AFM with a a. CO andb. PTCDA functionalized tips (4 nm x 4 nm, Vb = 0 V, ∆f= −1.3 Hz (a) and −3.5 Hz (b)). c. STM imaging with aPTCDA functionalized tip, (5 nm x 5 nm, It = 15 pA, Vb =0.5 V, ∆z =1.94 A˚). All scale bars are 1 nm. . . . . . . . . . 252.12 Schematic of principle of EFS mapping. Work function dif-ference leads to a parabolic dependence of the frequency shiftsignal. Mapping the maximum of the curve fit to the dataleads to an image of the surface potential between tip andsample. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.13 Omicron LT-SPM in an acoustically and mechanical vibra-tionally isolated pod. . . . . . . . . . . . . . . . . . . . . . . . 282.14 Schematic drawing of Omega pod. Not to scale. . . . . . . . 292.15 Preparation chamber of the microscope, labeled for clarity. . 302.16 Omicron SPM head. . . . . . . . . . . . . . . . . . . . . . . . 312.17 Inside of the imaging chamber showing the carousel, the wob-ble stick and the line for CO dosing. Photo taken with thecryostat taken out, looking down into the chamber from above. 312.18 a. A cut platinum iridium tip. b. A qPlus sensor fromScienta Omicron. . . . . . . . . . . . . . . . . . . . . . . . . . 32xList of Figures2.19 a. A top view of an Ag(111) crystal mounted in a tungstensample plate. b. A side view of an Au(100) crystal an asample plate. c. An Au(111)/mica substrate spot-weldedonto a sample plate. . . . . . . . . . . . . . . . . . . . . . . . 332.20 The silver <111> surface: a. Overview scan (200 nm x 200nm, It = 600 pA, Vb = −50 mV, ∆z = 3 A˚), b. atomicresolution (6 nm x 6 nm, It = 50 nA Vb = 1 V, ∆z = 2.4 A˚).c. Spectroscopy showing the Ag(111) surface state at −65 mV. 342.21 The Au(100) surface: a. Overview scan (80 nm x 80 nm, It= 5 pA, Vb = 0.5 V, ∆z = 9 A˚), b. Tunneling spectra (I/Vand corresponding [dI/dV]/[I/V]), setpoint: It = 15 pA, Vb= −2.5 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352.22 STM of the Au(111) surface. a. Overview constant currentSTM scan showing the herringbone reconstruction and sur-face state scattering from step edges (50 nm x 50 nm, It =800 pA, Vb = −80 mV). b. Atomic resolution of Au(111), theherringbone reconstruction is also observed (30 x 30 nm, It= 600 pA, Vb = −0.9 V, ∆z = 2.5 A˚, first atomic resolutionachieved by the thesis author). c. Scanning tunneling spectraof Au(111), set-point: It = 200 pA, Vb = −1 V. . . . . . . . . 362.23 The NaCl/Ag(111) surface. a. Overview of NaCl tri- andbilayers on Ag(111). Moire´ pattern is observed (70 nm x 70nm, It = 30 pA, Vb = 2 V, ∆z = 3.8 A˚). b. Atomic resolutionof a salt bilayer (6.5 nm x 6.5 nm, It = 1 nA, Vb = 35 mV, ∆z= 1.0 A˚). c. Spectroscopy of NaCl(2ML)/Ag(111) showingthe interface state with an onset at ∼100 mV. . . . . . . . . . 372.24 Homebuilt Knundsen cell for NaCl deposition. Note that thecopper heat shield and shutter assembly are removed for vi-sualization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382.25 The inside of a. a 4-pocket and b. a 3-pocket, water-cooledKentax evaporator. c. A quartz crucible containing CuPc. . . 392.26 (3,4,9,10)–Perylenetetracarboxylic Dianhydride (PTCDA) a..Chemical structure of PTCDA. b. Powdered PTCDA. . . . . 392.27 PTCDA on NaCl(2ML)/Ag(111) with different annealing times(50 x 50 nm, Vb = 0.5 V, ∆z = 4.5 A˚). a. Surface directlyafter deposition at 4.3 K showing isolated molecules (It = 2pA). b. Example of surface after 2–5 minute anneal (It = 15pA). c. Example after 6–9 minute anneal (It = 2 pA). . . . . 402.28 Copper Phthalocyanine (CuPc) a. Chemical structure ofCuPc. b. Powdered, crystalline CuPc. . . . . . . . . . . . . . 41xiList of Figures2.29 STM imaging of CuPc on NaCl(2ML)/Ag(111) a. Overviewimage of CuPc on NaCl(2ML)/Ag(111) after LT depositionand before annealing (50 nm x 50 nm, It = 10 pA, Vb = 0.3 V,∆z = 5.7 A˚) b. – d. STM images of a single CuPc moleculeon NaCl(2ML)/ Ag(111) (3.5 nm x 3.5 nm, b. It = 5 pA, Vb= −3.1 V, c. It = 10 pA, Vb = 0.3 V, and d. It = 10 pA, Vb= 2.25 V). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422.30 STM imaging of subsequently deposited PTCDA and CuPc.a. CuPc and PTCDA with no anneal (35 nm x 35 nm, It = 5pA, Vb = −0.5V V, ∆z = 6.3 A˚) and b. CuPc and PTCDAafter a two-minute “room temperature” anneal (35 nm x 35nm, It = 5 pA, Vb = 5 pA, ∆z = 5.1 A˚). . . . . . . . . . . . . 432.31 CuPc and PTCDA on Ag(111). a. before anneal (75 nm x75 nm, It = 30 pA, Vb = 0.6 V, ∆z = 3.5 A˚) and b. after aone minute room temperature anneal. (40 nm x 40 nm, It =5 pA, Vb = 0.35 V, ∆z = 2.3 A˚). . . . . . . . . . . . . . . . . 442.32 Manipulation of a PTCDA molecule around a PTCDA/CuPcdimer on NaCl(2ML)/Ag(111). Sequential constant currentSTM images (9 nm x 9 nm, Vb=−2V, a. It=3 pA b. It=3pA c. It=185 pA d. It=5 pA e. It=5 pA f. It=100 pAg. It=20 pA h. It=5 pA). Red arrows indicate manipulatedmovement of PTCDA molecules between scans. . . . . . . . . 453.1 STM constant current image of PTCDA on NaCl(2ML)/Ag(111)(30 nm x 30 nm, It = 30 pA, Vb = −2.1 V) demonstratingpreferential orthogonal adsorption with respect to the NaCl(001) plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.2 PTCDA on NaCl(2ML)/Ag(111) a. Adsorption geometryof PTCDA on the NaCl lattice. Electronic density contour(0.0004 eA˚−3) is outlined in orange, calculated using usingdensity functional theory in Gaussian (with basis set B3LYPunrestricted 6-31G). b. Schematic showing the transfer of anelectron from the Ag(111) substrate to a PTCDA molecule,resulting in a negatively charged molecule. . . . . . . . . . . . 493.3 STM constant current topographic images of a single PTCDAon NaCl(2ML)/Ag(111). 4 nm x 4 nm, a. Vb = −0.6 V, It =20 pA, b. Vb = −0.4 V, It = 20 pA, c. Vb = +0.5 V, It = 50pA, d. Vb = +1.8 V, It = 10 pA. . . . . . . . . . . . . . . . . 50xiiList of Figures3.4 STS point spectra of an isolated PTCDA molecule on NaCl(2ML)/Ag(111), (It = 1 pA). The differential conductance(dI/dV) (rescaled by a factor of 13 for It = 1 pA in gray)and normalized differential conductance (dI/dV)/(I/V) areshown (region near zero with divergences with dashed line).The tunneling resonances are identified as O1, U1, and U2. . 513.5 STS point spectra and pixel-by-pixel maps of an isolatedPTCDA on NaCl(2ML)/Ag(111). a. STS averaged over anentire molecule and background NaCl(2ML)/Ag(111) spec-tra, (set-point It=1.5 pA, Vb=-2.1 V). Inset: grid topography(5 nm x 5 nm, It = 1.5 pA, Vb = −2.1 V) with molecularposition overlaid. b. Energy maps of an isolated PTCDAmolecule corresponding to the dotted gray lines in (a), (5 nmx 5nm, at Vb = −0.80 V, −0.70 V, 0.18 V, 0.65 V, 2.12 V,and 2.22 V). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523.6 Electronic states of PTCDA adsorbed on NaCl(2ML)/Ag(111)with tunneling. a. Diagram showing LUMO splitting uponthe addition or removal of an electron b. Predicted corre-sponding electronic resonances . . . . . . . . . . . . . . . . . 533.7 The two dominant polymorphs, α and β, of bulk PTCDA.Rectangle indicates unit cell shift in adjacent layer. . . . . . . 543.8 STM constant current image of PTCDA on NaCl(2ML)/Ag(111)(45 nm x 45 nm, It = 30 pA, Vb = −1.5 V) showing largercluster formation. . . . . . . . . . . . . . . . . . . . . . . . . . 553.9 A “molecular zoo” of PTCDA clusters. STM constant currentimages of a. two geometries of 4 molecule clusters (16 nm x7 nm, It = 15 pA, Vb = 0.5 V), b. a fish shaped island (10nm x 10 nm, It = 30 pA, Vb = 0.55 V), c. a 12 moleculepinwheel (9 nm x 9 nm, It = 30 pA, Vb = 1.0 V), d. a 22molecule island (10 nm x 10 nm, It = 30 pA, Vb = 1.0 V), ande. a larger island showing the bulk “herringbone” structure(20 nm x 20 nm, It = 30 pA, Vb = 1.0 V). . . . . . . . . . . . 563.10 a. Structural model of 12-molecule island showing positionson the NaCl lattice. Molecule identifications (A, B, C) areindicated. b. Model of the hydrogen bond network within thenanoisland. Partial charges (δ) contributing to the hydrogenbonding are shown . . . . . . . . . . . . . . . . . . . . . . . . 57xiiiList of Figures3.11 a. (dI/dV)/(I/V) STS map of a 12-molecule PTCDA is-land at Vb = 1.45V showing the underlying Moire´ pattern(8.5 nm x 8.5 nm, It = 30 pA, set-point bias Vb = −1.5 V).b. NaCl(2ML)/Ag(111) Moire´ pattern obtained by Fourier-filtering the image components in (a) corresponding to thePTCDA island. c. Overlay of (a) and (b). d. STM topogra-phy (Vb = −1.5 V, It = 30 pA ) corresponding to STS mapin (a). e. Adsorption positions of molecules with respectto the Moire´ pattern. f. Geometrically equivalent PTCDAmolecules of type A showing inequivalent adsorption siteswith respect to the Moire´ pattern. g. (dI/dV)/(I/V) spec-tra of the four geometry equivalent A molecules within thecluster located on different sites of the underlying lattice. . . 593.12 STS of a 12-molecule PTCDA island. a. (dI/dV)/(I/V)spectra for molecule types A, B, and C within the cluster(blue, cyan, and green respectively). Thick curves representan average over all equivalent molecules. Thin curves are av-eraged over individual molecules. gray vertical lines denotebias voltages of the STS maps shown below. Inset: STM to-pographic scan taken during grid acquisition with spatiallyaveraged spectra locations represented by colored boxes (8.5nm x 8.5 nm, It = 30 pA, Vb =−1.5 V). b. CorrespondingSTS maps at increasing sample bias (8.5 x 8.5 nm2, Vb =−0.84, −0.52, +0.50, +2.08, +2.19, +2.32, +2.49 V) . . . . . 613.13 STS of two configurations of 4-molecule PTCDA islands. a.(dI/dV)/(I/V) spectra of a “diamond” and a “clover” shapedfour molecule island. Dashed gray lines indicate bias voltagesof maps shown in (b–i). Inset: STM topographic scan (14nm x 7 nm, It = 30 pA, Vb). b.–i. Corresponding STS mapsat increasing sample bias (14 nm x 7 nm, Vb = −0.92 (b),−0.85 (c), −0.67 (d), +0.20 (e), +0.56 (f), +2.14 (g), +2.35(h), +2.47 V (i). . . . . . . . . . . . . . . . . . . . . . . . . . 62xivList of Figures3.14 (dI/dV)/(I/V) spectra of 18-molecule island on NaCl(2ML)/Ag(111). a., b. STM constant current images of an 18-molecule PTCDA island adsorbed on NaCl(2ML)/Ag(111)(9.5 nm x 9.5 nm, It = 30 pA, (a.) Vb = +1.0 V and (b.)Vb = −1.5V). c. Adsorption geometry of 18-molecule islandshowing the location on the NaCl lattice. d. (dI/dV)/(I/V)spectra of all molecules in the 18-molecule cluster with curvesaveraged over individual molecules identified in inset. e. STSshown in d. shown to visualize the occupied states. . . . . . . 643.15 Local energy level alignment within PTCDA islands of dif-ferent sizes. a. For each (dI/dV)/(I/V) spectrum acquiredat a given tip position (x,y), “band edges” are defined asthe voltage onsets ((dI/dV)/((I/V) = 3) of states O1 andU2. The band gap is the voltage difference between theseonsets. (b.–e.) STM topographic images acquired duringspectroscopic measurement for a 4-molecule, 12-molecule, 18-molecule island, and herringbone nanoribbon, respectively (It= 30 pA, Vb = +1.5 V, (b) 6 nm x 6 nm, (c) 8.5 nm x 8.5nm, (d) 9.5 nm x 9.5 nm, (e) 8 nm x 8 nm). (f.–i.) 2D(x,y)-dependent maps of O1 voltage onset for (f) 4-moleculeisland, (g) 12-molecule island, (h) 18-molecule island and (i)herringbone nanoribbon. (j.–m.) Corresponding 2D (x,y)-dependent maps of voltage onset of U2. (n.–q.) Correspond-ing 2D (x,y)-dependent maps of band gaps. Note, in f.– corresponds to no detected onset. . . . . . . . . . . . . 663.16 a. In-plane polarizability tensors of a PTCDA molecule. b.point charge location determined by the center of the PTCDAmolecule on the underlying NaCl lattice, the angles used forthe microelectrostatic calculations are also shown. . . . . . . 683.17 Occupied state of PTCDA islands with polarization energy.The result of the microelectrostatic calculation (crosses) wasplotted for each molecule in three islands: 4-molecule, 12-molecule, and 18-molecule, and compared with the peak ofthe O1 state (dots). The specific molecules within the islandsidentified on the x-axis are labeled in the schematic structureof each cluster above. The vertical axes of O1 and Ep differby a factor of 2.8/e. . . . . . . . . . . . . . . . . . . . . . . . 71xvList of Figures3.18 Vector representation of induced dipole moment in the 12-molecule island with respect to molecule B1. Width of arrowsand transparency of ovals correspond to strength of induceddipole. Relative contribution to Ep respectively from eachmolecule is indicated. . . . . . . . . . . . . . . . . . . . . . . 723.19 Electronic state energies relative to EF derived from the mea-sured spectrum of the isolated molecule (left), applying thecalculated polarization energies for the edge (blue) and center(green) of the 12-molecule cluster (middle), and accountingfor the observed charge transfer as a rigid shift of the energylevels and compared to the experimental STS (right). Redarrows above clusters depict the induced polarization afteraddition of a tunneling electron. . . . . . . . . . . . . . . . . . 733.20 a. and b. STM topography (40 nm width, It = 7 pA andVb = −0.8 V and +0.5 V respectively). Note movement ofa PTCDA molecule from a quad- to a trilayer, indicated byblack arrow in (b). c. Comparison of normalized STS on anisolated PTCDA molecule on NaCl(2ML)/Ag(111) (black),NaCl(3ML)/Ag(111) (maroon), and NaCl(4ML)/Ag(111) (red)(locations of spectra indicated by markers in (b, bilayer spec-tra taken from a different scan). d. Lateral profile of bi-,tri- and quad-layer NaCl layers showing step height (locationindicated by turquoise line in (b). . . . . . . . . . . . . . . . . 753.21 Diagram comparing energy level alignment of PTCDA on2ML, 3ML, and 4ML NaCl on Ag(111). STS is shown forall three systems. The Hubbard gap is indicated by the semi-transparent box in between O1 and U1. . . . . . . . . . . . . 764.1 Isolated PTCDA on NaCl(2ML)/Ag(111) a. STM constantcurrent topograph (3.5 nm x 3.5 nm, It = 10 pA, Vb = 1.8V). b. Electronic states showing splitting upon addition orremoval of an electron during tunneling c. STS acquired witha Pt-Ir STM tip (setpoint: It = 1.5 pA, Vb = -2.1 V) withcorresponding electronic processes depicted. . . . . . . . . . . 794.2 A depiction of a df(V ) curve of a PTCDA molecule goingthrough a charging event at bias VSW with contact potentialdifferences VCPD1 and VCPD2 in the less negative and morenegative charge state respectively. . . . . . . . . . . . . . . . . 80xviList of Figures4.3 STM imaging of PTCDA nanoislands on NaCl(2ML)/Ag(111).a. Overview of a sample showing small PTCDA clusters (50nm x 50 nm, It = 15 pA, Vb = 0.5 V). b. and c. STM imagesof the two 4-molecule islands studied in this experiment: di-amond (left) and clover (right) (14 nm x 6.5 nm, It = 15 pA,Vb = 0.5 V (b.) and −1.5 V (c.). . . . . . . . . . . . . . . . . 824.4 Geometries of two 4-molecule islands positioned on the un-derlying NaCl lattice: a. 2-fold symmetric diamond island b.4-fold symmetric clover island. . . . . . . . . . . . . . . . . . 824.5 Constant height NC-AFM images of the diamond 4-moleculePTCDA nanoisland with: a a. CO and b. PTCDA func-tionalized tips (4 nm x 4 nm, Vb = 0 V). Two different tipfunctionalizations show different distortions of the adsorbedPTCDA carbon rings likely due to the nature of the distor-tions of the tip molecule. . . . . . . . . . . . . . . . . . . . . . 834.6 Constant height NC-AFM image of a clover 4-molecule PTCDAnanoisland with a CO functionalized tip. (4 nm x 4 nm, Vb= 0 V). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 844.7 Simultaneous STS and df(V ) spectra of a “diamond” shaped4-molecule island taken from constant-height pixel-by-pixelgrid with set-point on NaCl(2ML)/Ag(111), It = 2 pA, Vb =0.5 V, oscillation = 60 mV. . . . . . . . . . . . . . . . . . . . 854.8 A depiction of a df(V ) curve of a PTCDA molecule goingthrough three charge states as the bias is swept. . . . . . . . 864.9 a. – c. VCPD maps of a diamond shaped 4-molecule is-land in three different charge states (PTCDA0, PTCDA−,and PTCDA2− respectively) determined from fitting threesegments to df(V ) curves d. – f. lower Hubbard bias map,upper Hubbard bias map and gap from difference respectively(5 nm x 5 nm). . . . . . . . . . . . . . . . . . . . . . . . . . . 874.10 Proposed diagram of electron movement during tunneling into(electron addition) and out of (electron removal) a PTCDAdiamond island. δ+ and δ− representing relative positive andnegative charges respectively. . . . . . . . . . . . . . . . . . . 884.11 Simultaneous STS and df(V ) spectra of a “clover” shaped4-molecule island taken from a constant-height pixel-by-pixelgrid with parameters: set-point on NaCl(2ML)/Ag(111), It= 2 pA, Vb = 0.5 V, oscillation = 45 mV. . . . . . . . . . . . 90xviiList of Figures4.12 a. VCPD maps of a clover shaped 4-molecule island in threedifferent charge states determined from fitting three segmentsto df(V ) curves b. lower Hubbard bias map, upper Hubbardbias map and gap from difference (4.5 nm x 4.5 nm). . . . . 914.13 a. Simultaneous STS and df(V ) spectra of an isolated PTCDAmolecule on NaCl(2ML)/Ag(111), taken from a constant-heightpixel-by-pixel grid with parameters: set-point on NaCl(2ML)/Ag(111), It = 2 pA, Vb = 0.5 V, oscillation = 50 mV. b.and c. VCPD maps of PTCDA− and PTCDA2− respectively.Molecule in maps is in the same orientation as the inset in a,(4 nm x 4 nm). . . . . . . . . . . . . . . . . . . . . . . . . . . 924.14 a. Maps of the LUMO− to LUMO0 transition obtained fromscanning tunneling spectra of the onset of the closest occupiedresonance to EF . (b.) Map of the jumps in bias of the df(V )curves, corresponding to the LUMO− to LUMO2− transition.c. Map of the gap between these two states (all 4 nm x 4 nm). 935.1 STM topography of an isolated CuPc molecule on NaCl(2ML)/Ag(111) (3.5 nm x 3.5 nm; a. It = 5 pA, Vb = −3.1 V, b. It= 10 pA, Vb = 0.3 V, c. It = 40 pA, Vb = 0.75 V, and d. It= 10 pA, Vb = 2.25 V). e. NC-AFM imaging with a PTCDAfunctionalized tip (2.1 nm x 2.1 nm, Vb = 0V). . . . . . . . . 965.2 STS mapping of an isolated CuPc molecule on NaCl(2ML)/Ag(111). a. STM topgraphy (3.5 nm x 3.5 nm, It = 2 pA,Vb = −3.1 V). The locations of the spectra show in c areindicated by light and dark blue dots for the center and edgeof the molecule respectively. b. STS maps of the occupiedstates of CuPc (3.5 nm x 3.5 nm). c. (dI/dV)/(I/V) pointspectra of the inner (dark blue) and outer (light blue) regionof CuPc (set-point: It = 2 pA, Vb = −3.1 V). d. STS mapsof the occupied states of CuPc (3.5 nm x 3.5 nm). . . . . . . 985.3 a. STM image of showing typical cluster sizes and geometryof PTCDA and CuPc on NaCl(2ML)/Ag(111) (40 nm x 40nm, It = 5 pA, Vb = 0.5 V). b,c STM image of PTCDA andCuPc two-molecule cluster (5 nm x 3.5 nm, It = 6 pA, b Vb =−2 V and c Vb = 0.5 V). d. NC-AFM frequency shift imageof a PTCDA/ CuPc two-molecule cluster (3.5 nm x 3.5 nm, 0V, constant z). e. Molecular positioning roughly determinedfrom the NCAFM image d and known adsorption geometryon the underlying NaCl lattice. . . . . . . . . . . . . . . . . . 100xviiiList of Figures5.4 a. STS maps of an isolated PTCDA on NaCl(2ML)/Ag(111)(5 nm x 5 nm, set-point parameters: It = 2.5 pA, Vb = −2V) b. STS maps of a PTCDA/CuPc dimer corresponding toPTCDA, red dashed lines in c, (5 nm x 4 nm, set-point: It =2 pA, Vb = −2.5 V. c. (dI/dV)/(I/V) point spectra of bothisolated CuPc and PTCDA molecules (dashed lines) as wellas PTCDA and CuPc in the dimer (solid lines). d. STS mapsof dimer corresponding to CuPc states, indicated with bluedashed lines in c, (5 nm x 4 nm). e. STS maps of an isolatedCuPc on NaCl(2ML)/Ag(111) (3.5 nm x 3.5 nm, set-point:It = 2 pA, Vb = −3.1 V). The images are rotated to matchthe orientation of the CuPc molecule in d. . . . . . . . . . . . 1025.5 Electronic states of isolated PTCDA (red), isolated CuPc(cyan), dimer PTCDA (maroon), and dimer CuPc (blue) de-termined from STS (also shown). Overall energy resonanceshifts are indicated with dashed lines. Arrows indicate direc-tion of level shift due to charge transfer (ECT) and polariza-tion (EP). NaCl (2ML)/ Ag(111) surface state is indicated ingray. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035.6 Comparison of energy level alignment of PTCDA/CuPc clus-ters of varying geometries. a. – f. Structural models of the 6islands examined. e. – k. STM topographic images of eachisland It = 2 pA (xo: 4 nm x 5 nm, Vb = −2V; xo–x: 6 nm x6 nm, Vb = −2 .1V; ox–o: 5.5 nm x 5.5 nm, Vb = +2.2; oo–x:5 x 5 nm, Vb = −2.1 V; oxo: 6.5 nm x 3.5 nm, Vb = −2V;xox: 5.5 nm x 4 nm, Vb = −2 V. with x = CuPc and o =PTCDA). l. Alignment of CuPc states arranged in increas-ing order of surrounding PTCDA molecules m. Alignmentof PTCDA states arranged in increasing order of surroundingCuPc molecules. Stoichiometry of the island is indicated withrespect to the molecule examined. . . . . . . . . . . . . . . . 1065.7 STS along a line cut (indicated by yellow arrows) of bimolec-ular heterojunctions of islands consisting of varying geome-tries of a. one CuPc and one PTCDA , b. two CuPc andone PTCDA c. one CuPc and two PTCDA. x-axis is distancealong line cut, y-axis is bias, and colormap is (dI/dV)/(I/V)intensity. White dotted line indicates node of line path whereapplicable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108xixList of Figures6.1 STM topography and corresponding frequency shift images ofreversible charging of a PTCDA/CuPc dimer on NaCl(3ML)/Ag(111) (5 nm x 3 nm, It = 2 pA, Vb = 0.5 V . . . . . . . . . 1136.2 PTCDA and CuPc on NaCl(2ML)/Au(100). a. STM topog-raphy (6 nm x 4 nm, It = 5 pA, Vb = 2 V). b. STS maps ofPTCDA unoccupied states (6 nm x 4 nm). c. (dI/dV)/(I/V)point spectra of both isolated CuPc and PTCDA molecules(dashed lines) as well as PTCDA and CuPc in the dimer (solidlines) (set-point: It = 5 pA, Vb = −2.5 V) d. STS maps ofCuPc states (6 nm x 4 nm). . . . . . . . . . . . . . . . . . . 115A.1 I(V) curve of an isolated PTCDA on NaCl(2ML)/Ag(111)showing the raw forward and backward signals and the cor-rected signal after subtraction of an offset. A scaled version(b.) is included to see the zero bias region. . . . . . . . . . . 140A.2 Normalized STS of an isolated PTCDA on NaCl(2ML)/Ag(111)with and without the normalization correction offset C. . . . 141A.3 Raw and n = 3 moving averaged smoothed spectroscopy. a.I(V), b. dI/dV and c., (dI/dV)/(I/V) spectra of raw andsmoothed data of a 12-molecule PTCDA nanoisland (A site)on NaCl(2ML)/Ag(111). We applied an n = 3 boxcar movingaverage filter before computing the numerical dI/dV, yield-ing a bias resolution of ∆V ∗ n/2 = 12 meV. The smoothingprocedure does not alter the location of the peaks or obscureany features, given the widths of the tunneling resonances ob-served. The smoothing aids mostly in the normalized spectrain regions where the tunneling current is near zero. . . . . . 143A.4 Kappa maps of a PTCDA/CuPc dimer on NaCl(2ML)/Ag(111).6 nm x 4 nm, It = 3 pA, a. Vb = −2.1 V, b. −0.75 V, andc. 1.0 V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144B.1 Representation of the analysis used to find the fit parametersand switching bias. a. the raw data, b. a preliminary fit,and c. the difference between the data and the fit. . . . . . . 146B.2 Example spectra of a clover island showing three charge stateswith separate fits. . . . . . . . . . . . . . . . . . . . . . . . . . 147xxGlossary2-D Two dimensionalAC Alternating currentA˚ AngstromAFM Atomic force microscopy/eα Molecular polarizabilityCPD Contact potential differenceCT Charge transferCuPc Copper phthalocyanineDC Direct current∆f , df Frequency shiftDFT Density functional theoryDOS Density of statese Elementary chargeEA Electron affinityEf Fermi energyEFS Electrostatic force spectroscopyEH Hubbard energyEp Polarization energyeV Electron voltsf0 Resonance frequencyFCC Face centered cubich¯ Plank constantH-bond Hydrogen bondHCP Hexagonal close packedHOMO Highest occupied molecular orbitalIP Ionization potential/energyIt Tunneling currentκ inverse decay lengthKPFM Kelvin Probe force microscopyKPFM Kelvin Probe force spectroscopyLCPD Local contact potential differencexxiGlossaryLDOS Local density of statesLHe Liquid HeliumLN2 Liquid nitrogenLT Low temperatureLUMO Lowest unoccupied molecular orbitalMO Molecular orbitalNC-AFM Non-contact AFMOMBE Organic molecular beam epitaxyOPV Organic Photovoltaicsφt, φs work function of the tip and sample respectivelyPTCDA (3,4,9,10)-perylenetetracarboxylic dianhydrideQ Q-factorρs, ρt Density of states of the tip and sample respectivelyRT Room temperatureSOMO Singly occupied molecular orbitalSPM Scanning probe microscopy/eSTM Scanning tunneling microscopy/eSTS Scanning tunneling spectroscopyTSP Titanium sublimation pumpUH Hubbard potentialUHV Ultrahigh vacuumVbias Applied bias between tip and sampleVCPD Contact potential difference biasWKB Wentzel–Kramers–Brillouinz Tip–sample separationxxiiAcknowledgementsFirst and foremost I would like to thank Sarah Burke. You have been somuch more than a supervisor. Your wisdom, inspiration, and support hasbeen crucial to my success, in science and in life. I truly could not havedone this without you.Running a low temperature, ultra high vacuum microscope is best doneby a small village. I would like to thank my fantastic village, the membersof the LAIR (past and present):Doug Bonn, Agustin Schiffrin, Ben MacLoud, Brandon Stuart, GaryTom, Graham Baker, Gelereh Farahi, Martina Capsoni, Miriam Dejong,Sebastian Trembley-Johnson, and Shun Chi. Particularly, I would like toacknowledge Tanya Roussy for being my long term Omicron co-operator,partner in vacuuming cryostats, and fellow dinosaur creator; Bingkai Yuanfor his NC-AFM skills and tip-shaping magic; Erik Ma˚rsell for all of thethesis edits — and snacks; Andrew Macdonald for being the other longterm LAIRite and gummy bear aficionado; and James Day for the walkand talks. I would also like to acknowledge Stephanie Grothe, who alwaysbrought a smile to work and is deeply missed.Thanks to the members of the Grutter group for hosting me while theOmicron was on it’s German vacation, especially Zeno Schumacher who is agreat conference buddy and still helps out with all my AFM/JEOL relatedquestions.It also takes a village to support a PhD student. Janet, Lisa, Heather,Sara, Dylan, Shannon, Katie: thank you, for all of the things. Joanna,Helen, Susan and Natalie: I cannot express how grateful I am for you.Lastly, thanks to Zoe and Oliver. You two put up with my long hours,were always there for me, and always supported my use of lasers.xxiiiChapter 1Introduction andBackground1.1 MotivationAs the continued use of fossil fuels has proven to be detrimental to ourenvironment,[1–3] many are turning to alternative energy, with solar poweras a promising source.[4–6] Current commercial photovoltaic systems relymostly on inorganic semiconductors such as silicon. However, organic pho-tovoltaic (OPV) materials provide a promising choice for next generationsolar cells for many reasons, including their ease of processability, opticalabsorption tunability, transparency, and mechanical flexibility.[7–9] Theseproperties, particularly transparency and flexibility, have lead to widelyavailable commercial devices made with OPV materials, particularly forniche applications (Figure 1.1).Figure 1.1: Examples of niche applications of OPV materials: a. bag (NoonSolar)[10] b. transparent solar cell window coatings (Konakra)[11] c. anorganic solar cell transparent within the visible range (Zhao, et. al [12]) d.patio umbrella (Hammacker) [13] and e. car windows (Heliatek)[14]A major drawback for OPV materials so far has been low power conver-11.2. Charge generation in organic semiconductorssion efficiencies: 12.7% record,[15] with 4-5% being more typical.[16] This isnot yet competitive with silicon based cells, which are commercially avail-able with efficiencies surpassing 20%.[17] In order for a solar cell to be viable,three processes must occur efficiently: the material must absorb light, elec-tric charge must be generated, and the charge must be extracted from thesystem. In inorganic semiconductor based systems, light absorbed by thecell generates unbound electron–hole pairs. The electron and hole can thenfreely move through the conduction and valence band respectively. Themost common semiconductor cells used today are based on p–n junctions.Positive and negative doping of two regions of silicon will direct the flow ofcharge. Separating charge in an organic photovoltaic cell, however, is not assimple. The low dielectric constant of OPV materials results in the forma-tion of tightly bound neutral excitons (a neutral bound electron–hole pair)making charge separation and extraction more difficult than in inorganicsemiconductor materials.[18, 19]1.2 Charge generation in organic semiconductorsOPV systems are typically made up of an electron donating and an elec-tron accepting material.[20, 21] To generate current, the donor material isphoto-excited generating a neutral exciton.[22] In the ideal case, the excitondiffuses through the donor material before recombining, reaching an inter-face between the donor and acceptor. At this junction, the pair can separate;the electron is transferred to the acceptor material, while the hole remainsin the donor material. The coulombically separated electron and hole thenflow to the cathode and anode respectively, generating current (Figure 1.2).21.2. Charge generation in organic semiconductorsFigure 1.2: Energy diagram of a donor/acceptor photovoltaic system depict-ing the pathway of an electron and a hole upon absorption of a photon fromthe highest occupied molecular orbital (HOMO) to the lowest unoccupiedmolecular orbital (LUMO), forming an exciton.There are many different processes that can occur preventing chargefrom being extracted from the active components, thereby reducing the ef-ficiency of the cell. A significant problem is charge recombination, whichcan occur in several different ways (Figure 1.3).[23, 24] If the exciton doesnot arrive at a junction between donor and acceptor within the exciton dif-fusion length, and associated timescale, known as the exciton lifetime theelectron and hole will recombine.[25, 26] The typical diffusion length of anexciton in OPV materials is of the order of 1–30 nm [27, 28] and excitonlifetimes can range from attoseconds to microseconds,[29] depending on thetype, material, and conditions. This difficulty in reaching a junction beforerecombination is termed the “exciton bottleneck”.[30] The electron–hole paircan also recombine at the interface from a charge transfer exciton state, andeven after they have dissociated.[23]31.2. Charge generation in organic semiconductorsFigure 1.3: Possible pathways of an exciton in a donor/acceptor system.For all of these processes, energy level alignment dictates the favorabil-ity of the different electronic pathways the electron and hole take. Severaldifferent energies must be considered (Figure 1.4). The band gap is theequilibrium energy difference between the conduction band and the valenceband, or in the case of organic molecules, the lowest unoccupied molecu-lar orbital (LUMO) and the highest occupied molecular orbital (HOMO).This value can be calculated from density functional theory. The transportgap includes the energy needed to excite a free charge carrier, and is thedifference between the ionization potential and the electron affinity. Thisgap is larger than the optical gap, which is the energy needed by a pho-ton to excite a bound electron-hole pair. The exciton binding energy is thedifference between the calculated band gap and the optical gap. In orderfor charge transfer between the two materials to be favorable, the offset ofthe electron affinity (EA) of the two materials needs to be larger than theexciton binding energy of the donor. In well screened materials such as inor-ganic semiconductors these values are close enough that the differences arenegligible.41.3. The bulk heterojunctionFigure 1.4: Schematic of the energy levels of an organic material with theionization potential and electron affinity; optical, transport, and band gaps;and exciton binding energy labeled.Experimentally, gap energies are measured by electron or optical spec-troscopy yielding the transport and optical gap respectively. Optimizingthe values, differences, and offsets of the interfacial energy levels, and un-derstanding what factors result in charge separation being the most ener-getically favorable process, is a significant field of study.[20, 23, 31–36]1.3 The bulk heterojunctionOne of the OPV systems with the most promise shown so far is the bulkheterojunction (Figure 1.5).[37] This system is composed of a bicontinuousinterpenetrating network of donor and acceptor materials with domains thatare on the order of the exciton diffusion length.[38] Small domains are criticalso that the excitons are likely to reach an acceptor/donor interface beforerecombination occurs. With interpenetrating domains, the area betweendonor and acceptor is maximized and the distance the exciton travels isreduced. The bicontinuous network is needed so that after dissociation,the electron and hole can travel through the donor and acceptor materialsrespectively, to the electrodes. The ideal bulk-heterojunction OPV geometryhas regions of donor close to the anode and acceptor near the cathode, aswell as minimal isolated domains.[23, 39]51.4. Small molecules for organic photovoltaicsFigure 1.5: A simplified schematic of a device based on a bulk heterojunctionconfiguration.Optimizing the molecular geometry of the donor/acceptor interface is awidely studied area currently being examined to optimize charge separation.[40]Macroscopic, microscopic, and nanoscopic properties have all been shown tohave an impact on device efficiency.[24, 41–45] Small changes in molecularstructure as well as cell preparation techniques significantly change chargerecombination and extraction rates.[46–48] Understanding the direct rela-tionship between molecular geometry and the optoelectronic properties ofinterfaces is key to furthering the development of better materials.1.4 Small molecules for organic photovoltaicsFor an organic molecule to be implemented as a photovoltaic material, itmust have bulk semiconducting properties to allow for charge transfer aswell as be absorptive in the range of the solar spectrum that reaches theEarth’s surface. Molecules with significant conjugation often will have bandgaps on the order of 1–3 eV, as is necessary for absorption of visible andnear infrared (NIR) photons. Extensive pi–conjugated systems allows forelectron transport through the materials.[7]OPV materials can be divided into two broad groups: polymers andsmall molecules (see Figure 1.6 for a small sampling). Often a device is made61.4. Small molecules for organic photovoltaicsfrom a combination of the two (e.g. a polymer donor and a fullerene basedacceptor). Small molecule semiconductors have shown significant promisein the OPV market, and have been integrated into functioning devices.[14]Typically small molecules have higher charge-carrier mobilities.[28, 49–51]They are monodisperse, resulting in easier control and ability to design ofthe domains and interfaces.[50] As well, they often have more consistentsynthesis and device preparation.[52]Figure 1.6: A small sampling of the many conjugated organic molecules usedfor OPV. Two molecules that form an acceptor/donor junction, PTCDA(3,4,9,10-perylenetetracarboxylic dianhydride) and CuPc (copper phthalo-cyanine), will be the focus of this thesis.Small molecule components offer several advantages in real devices, andas model systems. Stability of the organic materials is also critical for a de-vice to be viable. These molecules must be stable over time under significantillumination in real-world situations where water and oxygen are present.Designing materials that are resistant to chemical and photo-degradation isalso a significant field of research.[53–57] As well, facile and energy-efficientprocessing is crucial for a device to be marketable. Current mass marketproduction of small molecule devices has involved both the use solutionprocessing[50, 58] and vacuum deposition technology[59]. Each of thesetechniques offer cost and energy efficient device manufacturing. For exam-ple, solution processed cells can be made from printing press technology71.5. Objectivesand vacuum deposition is the same method as used to coat the inside ofchip and other snack bags with metal films. Solubility has often been citedas a drawback to the scalable solution-processing of OPV materials. Oftenthese conjugated molecules are not soluble in solvents that are amenableto mass scale production.[7, 60] One way of overcoming this has been toadd functional groups to the molecules to improve solubility. However thesemodifications have been shown to significantly alter device efficiency.[47]These three features: cell efficiency, ease of processability, and devicestability are essential for OPV materials to be competitive. Processabilityand stability often come at the cost of device efficiency. In order to designbetter materials, it is important to understand what properties lead to moreefficient charge transfer and separation.1.5 ObjectivesMuch research has been carried out regarding acceptor/donor OPV inter-faces and attempting to maximize power conversion, but often at the full de-vice level or on length scales orders of magnitude larger than those relevantfor the key optoelectronic processes. It has been shown that the energy-conversion efficiency (as determined by the open circuit voltage, VOC) oforganic photovoltaic devices is a function of band gap energy of the donormaterial.[61] As well, VOC has been determined to be strongly correlatedwith the energy difference between the HOMO of the donor and the LUMOof the acceptor.[62] In these cases the energy values for the materials areoften taken from bulk measurements. However, charge transfer occurs onthe molecular scale.Graham et. al determined that small changes in molecular design by al-tering functional groups result in significantly different device efficiencies.[47]As well, subtle changes in molecular geometry can drastically change localelectrostatics.[63–66] Since the device physics occurs at the interface where asharp transition in molecular environment occurs, bulk systems may not givean accurate depiction of the interfacial energy level environment. Energyalignment is very sensitive to local nanoscale structure and at interfaces,which are difficult to study and control. However, it is important to under-stand the direct relationship between interfacial geometry and and electronicproperties.To investigate this requires techniques that go beyond the optical limit,as well as using techniques that can simultaneously resolve structure andresulting function. Creation of a model system in ultra high vacuum (UHV)81.5. Objectivesallows for exquisite control and a pristine environment to study interfacialproperties on the energy levels of molecules. Scanning Probe Microscopy(SPM) is a family of techniques that allows for the investigation of theelectronic and structural properties of OPV materials on the molecular scale.The goal of this thesis is to use the tools of scanning probe microscopyto investigate electronic structure that governs charge transfer at organicinterfaces such as in Figure 1.7 on the single molecule scale.Figure 1.7: Two possible interfaces between a donor and an acceptor withsubtle changes in geometry. How do the energy levels change with these twointerfacial geometries?9Chapter 2ExperimentalThis chapter focuses on the experimental techniques used in this thesis.Section 2.1 briefly introduces the scanning probe microscopy (SPM) fam-ily generically, Section 2.2 describes scanning tunneling microscopy (STM)and related tunneling theory, Section 2.3 describes atomic force microscopy(AFM) and corresponding theory. Section 2.4 is a description of the mi-croscope used, an Omicron low-temperature (LT) scanning probe micro-scope. Section 2.5 describes the samples used and their preparation, includ-ing molecular deposition.2.1 Principles of scanning probe microscopyMicroscopy is a technique that allows scientists to investigate a world muchsmaller than what can be seen with the human eye. A traditional com-pound microscope relies on light and optical components, allowing access tothe microscale world. Standard optical microscopes are however diffractionlimited. The first microscope able to resolve beyond this limit was the elec-tron microscope, built in 1933 by Ernst Ruska. Electron microscopy useselectrons in a similar manner that a compound microscope uses light andhas been able to achieve sub-A˚ngstrom resolution.[67] Smaller resolutioninvolves specialized equipment and expertise. Scanning probe microscopy(SPM) is an another technique that can spatially map surfaces with highresolution, but can also give information about electronic states and surfacecharge.Instead of using a beam (of photons or electrons) to scatter off of a sam-ple, SPM measures the interactions between a nanoscopically sharp tip anda surface. The first SPM instrument was the Scanning Tunneling Micro-scope (STM) developed in 1982 by Binnig and Rohrer of IBM.[68] This wasthe first time individual atoms had been directly mapped out in real spacein three dimensions. R. Young initially used quantum tunneling (the flowof electrons through a classically forbidden energy barrier) to map metallicsurfaces, but due to insufficient vibration control, was not able to achieveatomic resolution.[69] STM probes electronic states of surfaces of conducting102.1. Principles of scanning probe microscopyand semiconducting samples. In 1986, the Atomic Force Microscope (AFM)was reported by Binnig, Quate and Gerber.[70] The flow of electrons be-tween tip and sample is not required, allowing for the imaging of insulatingsamples. AFM measures the forces between the tip and the sample yieldingcomplimentary information to STM. The dominant force measured dependson the tip–sample distance and includes chemical, van der Waals, and elec-trostatic interactions.Figure 2.1: Scale of various microscopy techniques with corresponding exam-ples for scale reference. At the smallest scales for each, significant investmentalong with specialized equipment and expertise are required to achieve thehighest spatial resolution.During typical SPM operation, the tip is raster scanned while extremelyclose to the sample (within the range of nanometers to A˚ngstroms). As afunction of position, either the interaction between the tip and sample isrecorded (constant height mode) or, using a feedback loop the height of thetip is adjusted so the interaction is held constant. Figure 2.2 shows thedifferent paths an STM tip can take in constant height or constant currentmode. A piezoelectric scanner controls the x, y, and z position of the tip(Figure 2.3), or in some instruments the sample. Piezoelectric scanners allowfor the A˚ngstrom precision needed for the length scales used in SPM.112.1. Principles of scanning probe microscopyFigure 2.2: Operation of a scanning tunneling microscope in a. constantheight and b. constant current mode.Figure 2.3: Schematic of the scanning control of an STM.AFM and STM provide complementary information about nanoscale sys-tems and when implemented together constitute a powerful tool for probingthe structural, electronic, and electrostatic properties of a surface. Over thepast 35 years SPM has become an essential technique for a diverse group ofscientists, including those in the organic photovoltaic community.[42, 44, 71–73]122.2. Scanning tunneling microscopy and spectroscopy (STM/STS)2.2 Scanning tunneling microscopy andspectroscopy (STM/STS)STM relies on quantum mechanical tunneling of electrons through the clas-sically forbidden region between a tip and a sample. Under application of anexternal bias, this gives a small but measurable net current. The tip and thesample must be extremely close to each other (on the order of A˚ngstroms)such that their electronic wave functions overlap. The current measured isapproximately proportional to the integrated density of states (DOS) anddepends on tip–sample distance.[68] In 1983, Tersoff and Hamann appliedBardeen’s theory of tunneling[74] to the STM junction and established thesimple relationship between tunneling current and DOS (given several ap-proximations) that provides a straight-forward foundation for understandingSTM results in many cases.[75, 76] This, and other approaches, have beendeveloped to address STM more quantitatively and where more complexprocesses are involved. [77–80]2.2.1 Tunneling theoryThe system of two overlapping waves, with wave functions ψµ(r) and ψν(r)tunneling through a potential barrier U with width z, is depicted below inFigure 2.4.Figure 2.4: Schematic of two wave functions tunneling through a barrier.The transmission probability due to the overlap between ψµ and ψν canbe determined by solving the Schrodinger equation for a particle tunnelingthrough a rectangular vacuum barrier (Equation 2.1) with wave function Ψ(Equation 2.2).[81]132.2. Scanning tunneling microscopy and spectroscopy (STM/STS)HˆΨ(r) =[− h¯22m∂2∂r2+ U(r)]Ψ(r) = EΨ(r) (2.1)with a wavefunction in the barrier region that decays exponentially into thebarrier from each side:Ψ(z) = e±κr (2.2)where κ is the inverse decay length, and is dependent on E, the energy ofthe state, and the barrier potential height.κ2 =2mh¯2(U − E) (2.3)Bardeen’s approach solves this system by approximating the region withinthe barrier as a perturbing transfer Hamiltonian which is valid if the twowave functions are assumed to be weakly coupled.[74] The system can beseparated into three independent Hamiltonians: the overlapping wave func-tion’s Hamiltonians, Hˆµ and Hˆν , and the transfer Hamiltonian, HˆT (Equa-tion 2.4).Hˆ = Hˆµ + Hˆν + HˆT (2.4)Fermi’s golden rule is used to approximate the transition rate w12 as theprobability of transition from state µ with energy Eµ to state ν with energyEν coming from the perturbing transfer Hamiltonian HˆT .w12 =(2pih¯)∑ν|M |2 δ(Eµ − Eν) (2.5)where M is the tunneling Matrix element:Mµν =〈ψµ|(Hµ +HT )|ψν〉(2.6)In one dimension, for separable subsystems, Mµν can be simplified to(where dS corresponds to the integral over a surface within the barrier):Mµν =h¯22m∫dS · (ψ∗µ−→∇ψν − ψν−→∇ψ∗µ) (2.7)In the limit of small voltage, summing over all states results in a tunnel-ing current I, where EF corresponds to the Fermi energy.I =2pih¯e2V∑µ,ν|Mµν |2 δ(Eµ − EF )δ(Eν − EF ) (2.8)142.2. Scanning tunneling microscopy and spectroscopy (STM/STS)Tersoff and Hamman applied Bardeen’s theory of tunneling to the tip-sample STM junction to determine an expression for the tunneling current.[76]They assume the tip has an s-like wave function (a reasonable approximationfor most metallic tips), and the Fermi-Dirac distributions are approximatedby step functions (corresponding to the low temperature limit).Following this approach, the tunneling current simplifies to a function ofapplied bias V and tip–sample distance z, and is approximately proportionalto the integral of the density of states of the sample and tip. Here I switchfrom the arbitrary states µ and ν to t and s referring to the tip and samplerespectively.I(z, V ) ∝∫ eV0ρs(E)T (z, V,E)ρt(E − eV )dE (2.9)where T(z,V,E) is the transmission function and is related to the work func-tion of the sample and tip (φs and φt respectively), i.e. the height of thebarrier.T = exp(−2z√2mh¯√φs + φt2+eV2− E)(2.10)This results in the relation:It ∝ e(−2κz) (2.11)This exponential dependence on distance is one of the reasons STM has veryhigh spatial resolution.Equation 2.9 gives a tunnelling current related to the integrated DOS.In order to extract the DOS at a particular energy we differentiate thetunneling current with respect to bias to give:dI/dV ∝ ρs(z, eV )ρt(z, 0)T (eV, eV, z)+∫ eV0ρs(z, E)ρt(z, E − eV )dT (E, eV, z)dVdE (2.12)Often the transmission function is assumed to be bias independent overthe experimental bias voltage range. However, the effect of the bias depen-dence of the transmission function cannot be ignored when scanning largevoltage ranges, such as those needed for studying organic molecules. The152.2. Scanning tunneling microscopy and spectroscopy (STM/STS)effect of the transmission function results in an overall exponential back-ground at large biases. Normalization by dividing dI/dV byI/V results inthe majority of the bias dependent components of the transmission functioncanceling (Equation 2.13).[82, 83]dI/dVI/V∝ρs(z, eV )ρt(z, 0) +∫ eV0ρs(z,E)ρt(z,E−eV )T (eV,eV,z)dT (E,eV,z)dV dE1eV∫ eV0 ρs(z, E)ρt(z, E − eV ) T (E,eV,z)T (eV,eV,z)dE(2.13)In the experiments in this thesis the bias ranges are −3 to 3 V; thenormalizeddI/dVI/V is used as a close representation of the density of states.[82]Various different normalization schemes have been proposed, however mostrequire additional experimental information difficult to acquire, and noneperfectly recover the DOS.[78, 84]We assume that the density of states of the metallic tip is flat over thescanned bias range resulting in ρt(z, E − eV ) being constant.[85] With thisassumption and and ignoring the second integral term (which is typically anorder of magnitude smaller than first term [86, 87]) results in:dI/dVI/V∼ ρs(z, eV ) (2.14)It is important to note this normalization scheme does not work well nearEf (<100 meV) because the quantity diverges as it crosses zero, but is agood approximation for most molecular states.[78] The divergence near zerocan be avoided be adding a small current offset, as described in AppendixA.2.2.2 Scanning tunneling microscopy (STM)Constant current STM imaging gives a topographic image that is a convo-lution of the physical structure and the DOS of the sample. A feedbackloop moves the tip in z so that a constant current is maintained. If thereis a change in height, but not a change in DOS (for example at a surfacestep edge) the tip will move to compensate for this height change. If thereis no change in height, but a change in DOS (such as an underlying defectin the sample) the tip will also move, which could be in either z direction.This results in some surface features appearing as dips and some as protru-sions. Surface adsorbates such as atoms or molecules often result in bothheight and DOS changes. This can also be bias dependent due to energydependent differences of the DOS of adsorbate and sample. For example, a162.2. Scanning tunneling microscopy and spectroscopy (STM/STS)water molecule on Ag(111) will appear as a hole at biases above −70 mVbut a protrusion below −70 mV.[88] If the bias is positive, electrons willflow from the tip to the sample giving information about unoccupied statesof the surface. If the bias is negative, electrons will flow from the sample tothe tip, giving information about the occupied states of the sample.The current signal decays exponentially with tip–sample distance (Equa-tion 2.11), resulting in extremely high x, y and z resolution. This also meansthat generally only the outermost atom on the tip gives rise to a tunnelingsignal to the sample. As a result, the macro- and microscopic structure ofan STM tip does not usually significantly alter typical STM signals on flatsurfaces and only the nanoscopic quality of the tip influences the quality ofthe signal.2.2.3 Scanning tunneling spectroscopy (STS)Scanning tunneling spectroscopy (STS) gives access to the local density ofstates (LDOS, or a quantity that is close to it) as a function of energy andreal space. Features in a spectrum can include frontier molecular orbitals(the lowest unoccupied and highest occupied orbitals that are involved inbonding), surface states and many other electronic features. The normalizedderivative of the current signal gives rise to an expression that is proportionalto the LDOS, Equation 2.14.To obtain this data, we measure the change in current as a function ofapplied voltage with the tip held at a constant height. The I(V) curve may benumerically differentiated or the differential signal is acquired directly withthe use of a lock-in amplifier. For the measurements described in this thesis,the former method is used. Spatially resolved maps are achieved by acquiringthis information as a function of the (x,y) position of the tip (referred to as apixel-by-pixel or grid measurement). The full LDOS(x,y,E) can be obtainedby taking an I(V) (or dI/dV(V)) curve at each (x,y) point in a scan, allowingfor the detailed characterization of the relationship between topography andelectronic states. Full detail of the normalization data processing is includedin Appendix A.172.3. Non-contact atomic force microscopy (NC-AFM)Figure 2.5: Schematic showing the relation between energy levels and STSmeasurements. The third panel indicates the spatially resolved pixel-by-pixel STS. The black line represents a point spectrum occurring at pixel inthe (x,y) plane through all energy slices.It is important to note that STS results in an excitation spectrum. Thestates that are observed correspond to the addition or removal of charge.This is not considered in the above derivation of It and dI/dV which re-lies only on the underlying densities of states, and can significantly al-ter the observed spectrum. However, this also allows for effects such aspolarization,[89] charging[90] and Coulomb repulsion[91] to be observed.2.3 Non-contact atomic force microscopy(NC-AFM)Atomic Force Microscopy (AFM) is an SPM technique that measures theinteraction forces between a tip and a sample,[70] as opposed to electrontunneling. There are three dominant forces that contribute to this interac-tion for all materials: electrostatic, van der Waals (both long range) andchemical (short range) (Figure 2.6).[92] The total force between a tip and asample is the sum of these forces:[92]Ftotal = Felectrostatic + Fchemical + FV DW + Fadditional (2.15)182.3. Non-contact atomic force microscopy (NC-AFM)Figure 2.6: Relative contributions of the dominant forces between the tipand sample.The tip is attached to an oscillating cantilever which responds to thesum of the forces between the tip and the sample. In contact mode, thedeflection of this cantilever is measured as it is raster scanned across thesurface. In non-contact mode (NC-AFM), the cantilever is oscillated atits resonant frequency f0, and as the tip interacts with the surface, theresonant frequency shifts by an amount ∆f .[93] In constant frequency shiftmode, the tip–sample separation is changed to maintain a constant ∆f . Forhigh-resolution sub-molecular resolution imaging, constant height scans aretypically taken. Here, the frequency shift is measured while the tip is heldat a constant height above the surface, z, without the use of a feedback loop.[92, 94, 95]If the amplitude of the oscillation of the cantilever is small comparedwith the range over which the force gradient is constant across the junction(∂F∂z ) at the operating tip–sample distance,1 the measured frequency shift isproportional to the force gradient:[96]∆ffo∝ −∂F∂z(2.16)1This is often ambiguously referred to as “within the small amplitude limit”.192.3. Non-contact atomic force microscopy (NC-AFM)The electrostatic potential, Uelectrostatic, is represented in terms of ca-pacitance, C, and bias across the tip–sample junction, V:Uelectrostatic = −12CV 2 (2.17)This results Felectrostatic depending on the capacitance gradient (∂C∂z ):Felectrostatic =12∂C∂zV 2 (2.18)Here, the bias V is the sum of all sources of electric potential between tipand sample, including the applied bias, contact potential difference (thedifference in work functions of the tip φtip/e and the sample, φsample/e,that generate a potential at the tip), and any local charges, dipoles andfields. These are collectively referred to as VCPD (Although there can beadditional sources of potential difference between tip and sample, in analogyto Kelvin probe measurements where this sum of potential simplifies so asto determine the work functions of metals.)V = Vbias − VCPDVCPD =φtip − φsamplee(2.19)Taking the derivative of Felectrostatic leads to the electrostatic contribu-tion to the measured frequency shift:∆f ∝ −12∂2C∂z2(Vbias − VCPD)2 (2.20)The dependence of electrostatic force on the tip–sample bias allows for theextraction of this contribution from the total force.The electrostatic force can be approximated assuming a spherical tipwith radius R over an infinite plane separated by distance z (with z asthe distance between the plane and the tip atom closest to the plane, andassuming z <<R) with electrostatic potential difference, V:[92, 97]Felectrostatic = −pi0RV2z(2.21)Assuming this simplified sphere/plane interaction, the electric dipolemoments of both the tip (sphere) and sample (plane) atoms result in vander Waals interactions with potential:[92]UV DW = −AHR6z(2.22)202.3. Non-contact atomic force microscopy (NC-AFM)where AH is the Hamaker constant which varies with differing tip and samplematerials, though for most solids is on the order of 1 eV. This leads to aforce:FV DW = −AHR6z2(2.23)The short range chemical forces between tip and sample can be ap-proximated by a Lennard-Jones or Morse potential. These two potentialsdescribe a bonded tip–sample junction with energy Ebond. Where κ is thedecay length of the interaction and σ is the equilibrium distance of the bond.AndUMorse = −Ebond[2e−κ(z−σ) − e−2κ(z−σ)](2.24)orULennard−Jones = −Ebond(2z6σ6− z12σ12)(2.25)Differentiating these potentials leads to the forces:FMorse = −2κEbond[2e−κ(z−σ) − e−2κ(z−σ)](2.26)andFLennard−Jones = −12Ebondσ(2z7σ7− z13σ13)(2.27)These relations oversimplify the tip–sample interactions, but often serve asreasonable models for the data.All of these contributing forces provide a wealth of information about asurface; by altering the tip–sample junction separation and bias, NC-AFMcan resolve both local electrostatics and chemical structure.2.3.1 Tuning fork NC-AFMIn order to simultaneously acquire tunneling current and frequency shift sig-nals for STM and AFM respectively at low temperatures, a quartz tuningfork (qPlus) sensor was used.[98] Many AFMs use a microfabricated can-tilever sensor (typically made from silicon) with the deflection often detectedby reflecting a laser off of the end of the tip and collecting the light with aquadrant photodiode grid (Figure 2.7a).212.3. Non-contact atomic force microscopy (NC-AFM)Figure 2.7: Diagram of two main types of AFM: a. beam deflection and b.tuning fork.Quartz tuning fork sensors (Figure 2.7b and 2.8) have several advan-tages over silicon cantilevers: they are self-sensing, are very stiff, and havehigh quality factors.[92] A self-sensing tip eliminates the need for the opti-cal components of the sensor. The change in frequency is detected by thepiezoelectric quartz so there is no need to align a laser or a photodiode,which is difficult to implement in the constrained space needed to maintainlow temperatures. A large stiffness, kcant, allows for very small amplitudes,though some force sensitivity is sacrificed.[96] The noise of the signal de-pends significantly on the quality factor (Q). A higher Q results in less noiseat small amplitudes.[94]Figure 2.8: Photograph of a quartz tuning fork sensor used in experiments inthis thesis. This sensor was provided by Omicron Nanotechnology, howeverhomebuilt sensors were also used.In order to prevent cross-talk between the frequency shift and tunneling222.3. Non-contact atomic force microscopy (NC-AFM)signal, a separate gold wire that directly connects the tip to the base elec-trode is used to detect the tunneling current.[99] The tuning fork is mountedon a ceramic support, which sits on a gold tripod. The three feet mechani-cally and electrically couple the tip to the scanner head (Figure 2.8).2.3.2 High resolution NC-AFM with functionalized tipsIn 2009, Gross et al. achieved submolecular resolution with NC-AFM byusing a metal tip functionalized by attaching a single CO molecule to thetip apex, allowing for the real space visualization of bonds within a pentacenemolecule.[100] Here, the strongly repulsive Pauli forces that interact with theindividual atoms are believed to be the dominant source of contrast. Thisremarkable technique results in a clear visualization of molecules on surfacesand has allowed for unprecedented identification of chemical structure oforganic molecules.[101] Since then, various different tip apexes have beenused to achieve similar resolution, including atoms such as Xenon [102, 103]as well as larger organic molecules such as PTCDA[104].These images can give insight into various properties of a moleculeon a surface, including: structure[100], bond order[105, 106], adsorptiongeometry[107], and even reaction pathways[108, 109]. Because the image isa convolution of many factors, distinguishing each contribution is not trivial.As well, artifacts due to tip distortions can strongly influence imaging.[110,111] However, there have been significant advances in theory and modeling toaid in the deconvolution of these contributions.[111, 112] The exact mecha-nism of this imaging is still a hotly debated topic in the community[103, 111–116], however it is known that the polar and electrostatic interactions aswell as the tip distortion cannot be ignored.[112] Though quantitative mea-surements are difficult, equivalencies or differences in molecules (chargestate[117], adsorption geometry[107]) can be observed with this technique.232.3. Non-contact atomic force microscopy (NC-AFM)Figure 2.9: Schematics of the functionalized tips used in this thesis: a. car-bon monoxide b. PTCDA functionalized tip, showing proposed moleculargeometry.[104]In practice, the tip is functionalized by bringing the tip closer to thedesired molecule at zero bias, while monitoring the frequency shift signal. Asudden jump in the frequency shift indicates the molecule was either pickedup by the tip or moved laterally. Subsequent scanning in STM mode willindicate which occurred; molecules imaged by STM with a functionalizedtip show more substructure (Figure 2.11c). A scan is then performed inconstant-height mode with the feedback loop off. Frequency shift as a func-tion of tip–sample distance is measured (df(z) spectrum) to find the “turn-around point,” where the tip–sample interaction switches from attractiveto repulsive. This tip–sample distance just beyond this point, where therepulsive interactions start to dominate, is typically where the best contrastfor sub-molecular imaging occurs (Figure 2.10).Figure 2.10: Example df(z) spectrum of a metal tip functionalized with aCO molecule (∆z = 1 nm).242.3. Non-contact atomic force microscopy (NC-AFM)Figures 2.11a and b demonstrate bonds and molecular distortion thatcannot be seen with STM.2Figure 2.11: Examples of submolecular imaging of a 4-molecule PTCDAnanoisland with: constant height NC-AFM with a a. CO and b. PTCDAfunctionalized tips (4 nm x 4 nm, Vb = 0 V, ∆f = −1.3 Hz (a) and −3.5Hz (b)). c. STM imaging with a PTCDA functionalized tip, (5 nm x 5 nm,It = 15 pA, Vb = 0.5 V, ∆z =1.94 A˚). All scale bars are 1 nm.2.3.3 Electrostatic force spectroscopy (EFS)Electrostatic force spectroscopy (EFS) is a method of probing the localcharge in a sample on the macromolecular[118], molecular,[119], atomic,[120] and submolecular[121] scale. The long range electrostatic forces be-tween the tip and sample are often measured by keeping the tip–samplejunction separated by a distance where these long range forces dominate(Figure 2.6). The bias between tip and sample is swept and the frequencyshift measured. There is a parabolic dependence of frequency shift on bias,as seen in Equation 2.20. The measured bias potential with the minimumfrequency shift is the minimum force gradient, corresponding to the totalpotential difference between the tip and the sample. This includes localcharges and the contact potential difference (Equation 2.15). When thissweep is done pixel by pixel, a map of the surface potential between tip andsample is created.2All greyscale images in this thesis are plotted with a continuous black (minimum) towhite (maximum) scale. Where pertinent, ∆z = zmax − zmin (for constant current STMimages) or ∆frange = fmax − fmin (for constant height AFM images) is indicated in thecaption.252.3. Non-contact atomic force microscopy (NC-AFM)This technique has been used to observe the change in submolecularcharge distribution, for example resulting from the geometric switching of atautomer.[121] In order to obtain even higher resolution electrostatic forcemaps, spectra can be obtained with functionalized tips. However, it hasbeen recently determined that the high-resolution electrostatic force mapsobtained with a functionalized tip may not accurately reflect the chargedistribution.[122] In order to obtain resolution on the level of molecularbonds, the tip must be functionalized and close enough to the sample so thatchemical forces strongly contribute to the measured frequency shift. Thisresults in significant tip-sample distance dependent maps, even resulting incontrast inversion.[123] For this reason, all EFS measurements in this thesiswere taken with a metallic tip.Figure 2.12: Schematic of principle of EFS mapping. Work function differ-ence leads to a parabolic dependence of the frequency shift signal. Mappingthe maximum of the curve fit to the data leads to an image of the surfacepotential between tip and sample.The mapping of the VCPD is often referred to as Kelvin Probe ForceMicroscopy (KPFM). Typically in KPFM, an oscillating bias, a lock-in am-plifier detecting at the bias oscillation frequency, and a feedback loop tomaintain the minimum response are used to apply a potential correspond-ing to VCPD while the tip is scanned. Both pixel-by-pixel EFS and KPFMresult in images of surface potentials, though KPFM is much faster as it is ascanning method as opposed to a grid measurement, which often take over24 hours. However, EFS is advantageous for several reasons. Full df(V )spectra can show deviations from parabolic behavior due to charging[124]or polarization effects[118]. In addition, the bias modulation required forKPFM with feedback can have an influence on the measurement, particu-262.4. Description of apparatuslarly with tuning fork sensors. We, and others,3 have found that AC biasesapplied to either the tip or the sample have a tendency to excite unintendedmotion of the cantilever.2.4 Description of apparatusAll experimental data were taken with an Omicron low temperature ultra-high vacuum (UHV) SPM (Figure 2.13). All experimental data in this thesiswere taken with pressures < 5 x 10−12 mbar and at liquid helium temper-ature (∼4.3 K). The system is separated into two main components: thepreparation chamber and the imaging chamber. Both chambers maintainultra high vacuum with an ion pump (Agilent Starcell 300) and periodic ti-tanium sublimation.After every vent of the microscope to atmospheric pres-sure with dry nitrogen gas, the system was baked at 145 ◦C for roughly 3 to4 days to remove gases adsorbed on the walls of the chamber, particularlywater.3Known through discussion within the NC-AFM community272.4. Description of apparatusFigure 2.13: Omicron LT-SPM in an acoustically and mechanical vibra-tionally isolated pod.The microscope is located in a ultra-low vibration facility in order toreduce both mechanical and acoustic sources of noise (Figure 2.14). Themicroscope is on a 36 metric ton concrete inertial mass which is supportedby six pneumatic isolators. This allows for significant reduction of exter-nal vibrational noise. The walls are 25 cm thick concrete with mountedsoft material for acoustic noise reduction. As well, the pod is accessedthrough acoustically isolating doors. Wiring for the microscope as well asgas lines go through feedthroughs that are stuffed with foam. For a moredetailed description of the facility and noise spectra see MacLeod MAScthesis (2015).[125] Most cabling hangs from bungee cords or rests on foamon the floor. The microscope has internal vibration isolation which will bedescribed later. There are also three pneumatic isolators (Newport) on theinertial mass that can directly support the microscope. However these havenot been used while the microscope has been in the basement facility, asthey have not been found to reduce noise levels while the inertial block is282.4. Description of apparatusfloating.Figure 2.14: Schematic drawing of Omega pod. Not to scale.2.4.1 Preparation chamberSamples were prepared for measurement in the preparation chamber (Figure2.15). These two chambers can be isolated by a valve so the measurementchamber can stay clean during sample preparation. Samples were intro-duced into the system through a fast-entry load lock, which was pumpeddown (typically overnight) with a turbomolecular pump (Pfeiffer VacuumHiPace 300) backed by a dry scroll pump (Agilent Technologies, IDP-15).The sample was then transferred to the preparation chamber with the shorttransfer arm. The tips were also introduced into the system this way in atip transfer tool. The metal substrates were cleaned with repeated cyclesof sputtering and annealing. Sputtering occurred by bombardment by Ar+ions at 1–2 kV with a measured sample current of 2–4 µA and a cham-ber pressure of 1–3 x 10−6 mbar. Thin NaCl layers were deposited in thepreparation chamber with a homebuilt Knudsen cell (evaporation details de-scribed later). After preparation, the sample was transferred to a carousel292.4. Description of apparatusin the imaging chamber using the long transfer arm.Figure 2.15: Preparation chamber of the microscope, labeled for clarity.2.4.2 Imaging chamberThe main part of the microscope is the SPM head, which is about the sizeof two hockey pucks stacked on one another (Figure 2.16). The head hangsfrom soft springs and has gold-plated copper fins that magnetically coupleto the heat shield for eddy current damping to reduce vibrational noise.The sample is inserted into the sample stage face down using a wobble stick(Figure 2.17). The entire head is thermally coupled to a liquid helium (LHe)bath, which is in turn surrounded a liquid nitrogen (LN2) bath. This keepsthe scanner head at 4.3 K for 50 hours.302.4. Description of apparatusFigure 2.16: Omicron SPM head.Figure 2.17: Inside of the imaging chamber showing the carousel, the wobblestick and the line for CO dosing. Photo taken with the cryostat taken out,looking down into the chamber from above.The tip sits on the scanner head on three magnetic feet. For solelySTM measurements, cut platinum iridium (Pt-Ir 80-20, 0.38 mm diameter,Goodfellow) tips were used. These tips were prepared by cutting the wire312.5. Samplesat a 45 degree angle while pulling to extrude the end. The cut tip wasthen inserted into the tip holder and the tube crimped to hold the tip inplace (Figure 2.18). qPlus sensors with etched tungsten tips were used forsimultaneous AFM and STM measurements and were either commerciallyprovided by Scienta Omicron or fabricated in-house by Dr. Bingkai Yuan.There is a six parking space carousel (Figure 2.17) to store tips andsamples for fast exchange.Figure 2.18: a. A cut platinum iridium tip. b. A qPlus sensor from ScientaOmicron.Conventionally, the sign of the bias is reported with respect to the sam-ple, and this is referred through this thesis. Experimentally the bias can beapplied to the sample or the tip and different experiments require differentsetups. The AFM preamplifier is in-line with the tip, and therefore floatingat the tip potential (either VB or ground). All experiments with qPlus tipswere performed with the bias applied to the sample so that the bias did notinterfere with the sensor. Experiments using an STM tip were performedwith the bias applied to the tip to reduce electronic noise (determined em-pirically).2.5 SamplesAll experiments were performed on one of three substrates: high purity,single crystal Ag(111) and Au(100) (Matek GmbH), and Au(111) thin filmon mica (SPI Supplies). The two single crystals are top-hat-shaped with apolished face diameter of 8 mm, a thickness of 0.8 mm, and an edge of 1mm. This edge is present to provide space to clamp it to the sample plate(Figure 2.19a, b). The gold film was held onto the sample plate by tantalumfoil spot welded to the plate (Figure 2.19c).322.5. SamplesFigure 2.19: a. A top view of an Ag(111) crystal mounted in a tungstensample plate. b. A side view of an Au(100) crystal an a sample plate. c.An Au(111)/mica substrate spot-welded onto a sample plate.2.5.1 Metallic substratesAg(111)Silver exhibits a face centered cubic (FCC) structure; when cut along the(111) plane, the resulting surface has a hexagonal structure with a latticeconstant of 2.89 A˚(Figure 2.20b). The surface also has a Shockley-typesurface state with an onset at approximately −65 mV, Figure 2.20c.The silver crystal was prepared by repeated cycles of Ar+ sputteringand annealing at approximately 500 ◦C. This resulted in clean surfaces withlarge terraces (100s of nm), Figure 2.20a.332.5. SamplesFigure 2.20: The silver <111> surface: a. Overview scan (200 nm x 200nm, It = 600 pA, Vb = −50 mV, ∆z = 3 A˚), b. atomic resolution (6 nmx 6 nm, It = 50 nA Vb = 1 V, ∆z = 2.4 A˚). c. Spectroscopy showing theAg(111) surface state at −65 mV.Au(100)Gold also exhibits an FCC structure, with a lattice constant of 4.08A˚. Acut along the (100) plane should result in a four-fold symmetric lattice, how-ever strain at the crystalline surface results in a 1 x 5 reconstruction (Figure2.21a), with an atomic arrangement with hexagonal symmetry. There is nosurface state (Figure 2.21b).342.5. SamplesFigure 2.21: The Au(100) surface: a. Overview scan (80 nm x 80 nm, It =5 pA, Vb = 0.5 V, ∆z = 9 A˚), b. Tunneling spectra (I/V and corresponding[dI/dV]/[I/V]), setpoint: It = 15 pA, Vb = −2.5 V.Au(100) was prepared by repeated cycles of sputtering and annealingsimilar to Ag(111), however the annealing temperature was approximately400 ◦C.Au(111)STM constant current images of the Au(111) surface are shown in Figure2.22a, b. Au(111) shows a “herringbone” structure resulting from the (22x√3) reconstruction of the surface into FCC and hexagonal close packed(HCP) regions. The bright stripes are the dislocations between the FCCand the HCP regions.[126] The standing waves seen near step edges anddefects are due to scattering of the surface state. Here, the surface electronsbehave as a 2D electron gas. STS of this surface state is seen in Figure2.22c showing the subtle electronic difference between the FCC and HCPregions.[127]352.5. SamplesFigure 2.22: STM of the Au(111) surface. a. Overview constant currentSTM scan showing the herringbone reconstruction and surface state scatter-ing from step edges (50 nm x 50 nm, It = 800 pA, Vb = −80 mV). b. Atomicresolution of Au(111), the herringbone reconstruction is also observed (30x 30 nm, It = 600 pA, Vb = −0.9 V, ∆z = 2.5 A˚, first atomic resolutionachieved by the thesis author). c. Scanning tunneling spectra of Au(111),set-point: It = 200 pA, Vb = −1 V.Clean Au(111) was prepared by the same method as Au(100).2.5.2 NaCl deposition and the salt/metal interfaceThin salt films deposited on the metal surface were used in all of the ex-periments in this thesis. The thin, insulating film partially electronicallydecouples the molecules from the underlying metallic substrate, while stillallowing for electron tunneling through both the vacuum barrier and in-sulating layer.[128] A molecule directly adsorbed on a metal will hybridize362.5. Sampleswith the metal states and the resulting energy levels will be shifted andbroadened together with additional complex surface interactions.[129, 130]This method has been used extensively to determine electronic propertiesof molecules and allows for simultaneous STM and AFM measurements, asSTM is not possible on a bulk insulator.Figure 2.23: The NaCl/Ag(111) surface. a. Overview of NaCl tri- andbilayers on Ag(111). Moire´ pattern is observed (70 nm x 70 nm, It = 30pA, Vb = 2 V, ∆z = 3.8 A˚). b. Atomic resolution of a salt bilayer (6.5nm x 6.5 nm, It = 1 nA, Vb = 35 mV, ∆z = 1.0 A˚). c. Spectroscopy ofNaCl(2ML)/Ag(111) showing the interface state with an onset at ∼100 mV.The NaCl(100) bilayers grow predominantly in the <112> direction onAg(111) with a lattice mismatch of 11%, resulting in a Moire´ pattern. Thiscorresponds to a spatial modulation of the work function, and therefore ofthe electrostatic potential, visualized as stripes in the islands. An interfacestate also appears with an onset at ∼100 mV, as seen in Figure 2.23c.[131]A homebuilt Knudsen cell (Figure 2.24) was used to deposit NaCl layersby thermal deposition. The thickness and lateral size of the salt layers couldbe controlled by changing the substrate temperature. Depositing NaCl with372.5. Samplesthe substrate held at ∼100 ◦C (done in the preparation chamber) resulted inthe significant presence of trilayers as well as quad-layers. In STM imaging,tunneling through more than a quadlayer of NaCl requires large voltagesand very small tunneling currents, which are at the limits of the experi-mental noise floor. Smaller tunneling currents allow for a larger tip–sampledistance, which is needed if additional insulating layers are present (due tothe additional thickness of the tunneling barrier). With the substrate ata slightly lower temperature, ∼70 ◦C, predominately bilayers were formed,with very few trilayers.Figure 2.24: Homebuilt Knundsen cell for NaCl deposition. Note that thecopper heat shield and shutter assembly are removed for visualization.382.5. Samples2.5.3 Organic moleculesMolecules were thermally deposited using an organic molecular beam epi-taxy (OMBE) cell. The materials were sublimated from quartz crucibles(Figure 2.25c) in a multi-pocket evaporator (Kentax, GmbH, Figure 2.25aand b) into the imaging chamber with the substrate held at 4.3 K.Figure 2.25: The inside of a. a 4-pocket and b. a 3-pocket, water-cooledKentax evaporator. c. A quartz crucible containing CuPc.PTCDA(3,4,9,10)-perylenetetracarboxylic dianhydride (PTCDA) is a perylene deriva-tive organic red dye (Figure 2.26a, b). PTCDA molecules (98%, Alfa Aesar,further purified by degassing in vacuum) were deposited at 280 ◦C withthe substrate held between 4.2 and 4.5 K. Typical deposition times wereapproximately 2 to 3 minutes. Isolated molecules were immobile on bothNaCl bilayers and Ag(111) at 4.2 K under typical scanning conditions, butcould be manipulated by pulsing or scanning with large currents or biases(Figure 2.27a). Typical imaging parameters were between 2 and 100 pA and−2.5 and +3 V.Figure 2.26: (3,4,9,10)–Perylenetetracarboxylic Dianhydride (PTCDA) a..Chemical structure of PTCDA. b. Powdered PTCDA.392.5. SamplesTo form islands, samples were “room-temperature” annealed by remov-ing the sample from the cryostat and holding in the wobble-stick outside theheat shields (Figure 2.17) for varying amounts of time. The increase in tem-perature caused the molecules to diffuse on the surface to form 2-D clusters.Longer annealing times created larger clusters. It was not possible to accu-rately determine the sample temperature during this process. If the samplewas out of the cryostat for longer than ∼30 minutes, all of the PTCDAmolecules diffused to form sparse, very large (∼100s of molecules) islandson the Ag(111) surface. For an Omicron sample plate held in the wobble-stick, 2–5 minutes outside of the cryostat regularly formed 4-molecule clus-ters (Figure 2.27b), while 6–9 minutes formed larger islands (Figure 2.27c).The islands formed were not always stable, but could be manipulated intoa stable conformation. This system is the focus of Chapters 3 and 4.Figure 2.27: PTCDA on NaCl(2ML)/Ag(111) with different annealing times(50 x 50 nm, Vb = 0.5 V, ∆z = 4.5 A˚). a. Surface directly after depositionat 4.3 K showing isolated molecules (It = 2 pA). b. Example of surfaceafter 2–5 minute anneal (It = 15 pA). c. Example after 6–9 minute anneal(It = 2 pA).CuPcCopper phthalocyanine is a blue-purple organic photovoltaic dye (Figure402.5. Samples2.28a, b). CuPc (>99.95%, Aldrich; purified by degassing) was depositedat temperatures from 380 to 420 ◦C onto a substrate held between 4.2 and4.5 K. Deposition times ranged from 30 seconds to 5 minutes depending ondesired coverage and evaporation temperature.Figure 2.28: Copper Phthalocyanine (CuPc) a. Chemical structure of CuPc.b. Powdered, crystalline CuPc.Isolated CuPc molecules were observed both on NaCl bilayers and onbare silver when deposited at 4.3 K (Figure 2.29a). During “room temper-ature” annealing all of the CuPc molecules diffused from the NaCl bilayers,resulting in aggregation at the Ag/NaCl step edge. Gently heating thecryostat stage to 40 K did not cause the molecules to observably diffuse andaggregate.412.5. SamplesFigure 2.29: STM imaging of CuPc on NaCl(2ML)/Ag(111) a. Overviewimage of CuPc on NaCl(2ML)/Ag(111) after LT deposition and before an-nealing (50 nm x 50 nm, It = 10 pA, Vb = 0.3 V, ∆z = 5.7 A˚) b. – d.STM images of a single CuPc molecule on NaCl(2ML)/ Ag(111) (3.5 nm x3.5 nm, b. It = 5 pA, Vb = −3.1 V, c. It = 10 pA, Vb = 0.3 V, and d. It= 10 pA, Vb = 2.25 V).Bi-molecular systemFor some experiments, PTCDA and CuPc molecules were examined to-gether. Typically, PTCDA was deposited first and CuPc subsequently due tothe high mobility of CuPc on NaCl. To form mixed clusters, the sample wasroom temperature annealed after PTCDA deposition and CuPc moleculessubsequently deposited allowing them to diffuse and attach to PTCDA clus-422.5. Samplesters. Another method was the simultaneous deposition of both molecules(Figure 2.30a) and then performing a short room temperature anneal for ap-proximately two minutes (Figure 2.30). As opposed to a surface with onlyCuPc, the CuPc molecules did not diffuse completely off of the NaCl, butclustered around the PTCDA molecules. The sample did not become warmenough for the PTCDA molecules to aggregagte during this short annealtime.Figure 2.30: STM imaging of subsequently deposited PTCDA and CuPc.a. CuPc and PTCDA with no anneal (35 nm x 35 nm, It = 5 pA, Vb =−0.5V V, ∆z = 6.3 A˚) and b. CuPc and PTCDA after a two-minute “roomtemperature” anneal (35 nm x 35 nm, It = 5 pA, Vb = 5 pA, ∆z = 5.1 A˚).On silver, PTCDA forms well-ordered hexagonal herringbone structuresafter room temperature annealing. Subsequent CuPc deposition and anneal-ing forms PTCDA islands decorated by CuPc with some CuPc remainingstacked on the PTCDA islands.432.5. SamplesFigure 2.31: CuPc and PTCDA on Ag(111). a. before anneal (75 nm x 75nm, It = 30 pA, Vb = 0.6 V, ∆z = 3.5 A˚) and b. after a one minute roomtemperature anneal. (40 nm x 40 nm, It = 5 pA, Vb = 0.35 V, ∆z = 2.3 A˚).Molecular manipulationThe self-assembled structures after co-deposition were not always whatwas desired, so occasionally the molecules were manipulated into differentforms. Manipulating molecules on thin insulating films is typically per-formed by tunneling into one of the unoccupied states.[132] We observed thatPTCDA molecules move in one dimension along the NaCl lattice, jumpingbetween lattice sites, though the direction of movement seems to be random.The direction of the CuPc molecules is more easily controlled, though ma-nipulation often results in a molecular rotation. Occasionally, the moleculeis picked up by the tip instead of being manipulated.442.5. SamplesFigure 2.32: Manipulation of a PTCDA molecule around a PTCDA/CuPcdimer on NaCl(2ML)/Ag(111). Sequential constant current STM images (9nm x 9 nm, Vb=−2V, a. It=3 pA b. It=3 pA c. It=185 pA d. It=5 pAe. It=5 pA f. It=100 pA g. It=20 pA h. It=5 pA). Red arrows indicatemanipulated movement of PTCDA molecules between scans.45Chapter 3Polarization Induced EnergyLevel Shifts in PTCDANanoislandsThe following section contains adapted text from:K.A. Cochrane, A. Schiffrin, T.S. Roussy, M. Capsoni, S.A. Burke. Pro-nounced polarization induced energy level shifts at boundaries of organicsemiconductor nanostructures. Nat. Commun. 6:8312 (2015).[133]Figures are altered unless otherwise noted. An expanded introductionand additional data are included.3.1 IntroductionEffects of the local electrostatic environment have been shown to signifi-cantly shift molecular energy levels through polarization of adjacent matter.[63–65, 134–136] Such effects can shift molecular states up to 500meV[64,66, 134] and can contribute of order 1 eV to the transport gap of organicsemiconductors[137, 138] with consequences for both transport and chargetransfer across interfaces.[35] Recent studies using Scanning Tunnelling Mi-croscopy (STM) and Spectroscopy (STS), alongside complementary tech-niques, on mixed donor-acceptor monolayers on metallic and semi-metallicsurfaces have sought to characterize their electronic structure as model in-terfaces for organic photovoltaic materials.[36, 71, 139, 140] In these exam-ples, energy level shifts in the electronic states of both species arise froma combination of intermolecular and molecule-substrate interactions. Al-though these well-ordered systems come close to mimicking photovoltaicdevice materials in a way that is amenable to surface science probes, thesingle-molecule “width” of the interfaces present in these layers, and theinfluence of a metallic substrate, creates an environment that deviates fromactual devices and influences the measured electronic structure.[45, 51]To further our understanding of some of the relevant factors for energy463.2. Experimentallevel changes at interfaces of organic materials, the simplest possible inter-face was examined: that between a molecular cluster and a vacuum. We ex-amine the boundaries of two-dimensional (2D) clusters of PTCDA (3,4,9,10-perylene-tetracarboxylic dianhydride) to study nanoscale lateral interfaces(with the vacuum). PTCDA has been widely used as a prototypical or-ganic semiconductor. It is a promising optoelectronic material[141, 142]with a tendency to order on a wide range of surfaces. This is partially dueto the formation of strong inter-planar hydrogen bonds.[143–146] Studiesof the electronic structure of PTCDA monolayers and thin films on metalsubstrates have shown that molecular energy levels can shift by hundredsof meV due to a variety of effects. This includes small differences in thehydrogen bond lengths between inequivalent adsorption sites,[46, 135, 147–150] and stabilization of charge by polarization of the surroundings[64] as afunction of film thickness and distance from an interface.Using low-temperature STM and STS we have locally probed the struc-ture and electronic states of PTCDA clusters that are decoupled from theAg(111) substrate by a bilayer film of NaCl to allow examination of theintrinsic electronic effects within clusters on the molecular scale.[128] Pixel-by-pixel STS allows us to probe the energy levels of electronic states withsub-molecular spatial resolution, revealing the influence of the abrupt changein the local environment at the edges of nanoislands.3.2 ExperimentalSTM and STS measurements were performed in UHV at ∼4.3 K (OmicronNanotechnology) with an Ag-terminated cut Pt/Ir (Goodfellow) tip. Tipswere verified as metallic on the bare Ag(111) substrate with a flat DOSand the onset of the Ag(111) surface state at −65 mV. STM topographicimages were acquired in constant-current mode. The bias voltage relativeto the sample is reported throughout the text. (dI/dV)/(I/V) curves as afunction of sample bias voltage were obtained by numerically differentiat-ing I(V) data measured for each (x,y) tip position on the surface with thefeedback loop disabled (128 by 128 pixels in x–y, 512 bias points taken forthe voltage range scanned, resulting in ∼8 mV energy resolution for typicalmeasurement conditions, sampling time at each bias was ∼ 6 ms). I/V and(dI/dV)/(I/V) spectra were smoothed with a three-point moving boxcarfilter giving an energy resolution of 12 meV.473.3. Isolated PTCDA on NaCl(2ML)/Ag(111)3.2.1 Sample preparationThe following measurements were performed on a single Ag(111) crystal(Matek GmbH) prepared in UHV by repeated cycles of Ar+ sputtering andannealing at 770 K. NaCl (TraceSELECT ≥ 99.999%, Fluka) was thermallyevaporated at ∼800 K onto the sample held at 370 K, resulting in (001)terminated bilayer islands covering roughly 50% of the surface. PTCDA(98%, Alfa Aesar) was deposited at 550 K with the substrate held between4.2 and 4.5 K in the STM head. Typical deposition times were approximately2 to 3 minutes for coverages in between and 0.1 and 0.01/nm2 (1 – 10molecules per 10 nm x 10 nm image). Isolated molecules were stable onboth NaCl bilayers and Ag(111) at 4 K.Samples were “room-temperature” annealed by removing the samplefrom the cryostat and holding with the wobble-stick for varying amountsof time. The increase in temperature causes the molecules to diffuse on thesurface to form 2D clusters. Longer annealing times created larger clusters.It was not possible to accurately determine the sample temperature duringthis process. If the sample was out of the cryostat for longer than ∼30 min-utes, all of the PTCDA molecules diffused to form sparse, very large (∼100sof molecules) islands on the Ag(111) surface. For an Omicron sample plate,5 minutes outside of the cryostat regularly formed 4 molecule clusters, while7–10 minutes formed larger islands (Figure 3.8). The islands formed werenot always stable, but could be manipulated into a stable conformation bypositioning the tip over the molecule and increasing the bias or scanning at(relatively) high tunneling currents.3.3 Isolated PTCDA on NaCl(2ML)/Ag(111)3.3.1 Adsorption of PTCDAWe found that nearly all isolated PTCDA molecules on NaCl(2ML)/Ag(111)adsorb at a 90o angle with respect to each other, and 45o with respect to the(100) salt island termination determined by STM imaging, as seen in Figure3.1. This is consistent with the adsorption geometry seen previously onbulk NaCl [144, 151], and epitaxial NaCl films[85, 152]. Similar adsorptionis not surprising as the lattice contraction of NaCl(2ML) on Ag(111) issmall compared with bulk NaCl (Cl–Cl distance of 395 ± 6 pm and 390± 8 pm in the two planar orthogonal directions, compared with the bulkCl-Cl distance of 399 pm).[153] PTCDA molecules are adsorbed on the Cl-top site position (the center of the molecule is directly on a Cl- ion, Figure483.3. Isolated PTCDA on NaCl(2ML)/Ag(111)3.2a). Occasionally, a PTCDA molecule was found parallel to the <100>directions (e.g. 45o to the majority of the other PTCDA molecules), alsoobserved by Mohn, et al.[154] Isolated PTCDA molecules on Ag(111) werealso observed, without any clear preferential adsorption angle. Previously,PTCDA on bulk NaCl was determined to have a nearly planar adsorptiongeometry with the anhydride oxygens slightly bent towards the underlyingsubstrate. [151, 155]Figure 3.1: STM constant current image of PTCDA on NaCl(2ML)/Ag(111)(30 nm x 30 nm, It = 30 pA, Vb = −2.1 V) demonstrating preferentialorthogonal adsorption with respect to the NaCl (001) plane.Figure 3.2: PTCDA on NaCl(2ML)/Ag(111) a. Adsorption geometry ofPTCDA on the NaCl lattice. Electronic density contour (0.0004 eA˚−3) isoutlined in orange, calculated using using density functional theory in Gaus-sian (with basis set B3LYP unrestricted 6-31G). b. Schematic showing thetransfer of an electron from the Ag(111) substrate to a PTCDA molecule,resulting in a negatively charged molecule.493.3. Isolated PTCDA on NaCl(2ML)/Ag(111)3.3.2 Electronic states of isolated PTCDA on an NaClbilayerFigure 3.3 shows STM topographic images of isolated PTCDA molecules onNaCl(2ML)/Ag(111) obtained at various biases. At biases less than −0.4V and greater than 1 V, PTCDA on NaCl(2ML)/Ag(111) is imaged bySTM as a double-lobed structure with a single nodal plane along the longaxis, similar to that observed previously.[154, 156, 157] The in-gap states,between −0.4 V and 0.5 V, are imaged as an oblong, single-lobed structurewith subtle substructure.Figure 3.3: STM constant current topographic images of a single PTCDAon NaCl(2ML)/Ag(111). 4 nm x 4 nm, a. Vb = −0.6 V, It = 20 pA, b. Vb= −0.4 V, It = 20 pA, c. Vb = +0.5 V, It = 50 pA, d. Vb = +1.8 V, It =10 pA.STS [dI/dV and (dI/dV)/(I/V)] at different heights over isolated molecules(obtained with different tunneling current set points, Figure 3.4) show threemain features within the experimentally accessible range (−1.5 V to +2.8V). We observe a well-defined tunneling resonance at −0.7 V, which we labelO1, a broad feature at ∼0.6 V (U1), and a strong resonance with an onset at∼2 V (U2). Peak energies were found to be independent of the tip–moleculedistance, indicating there was not an observable effect of the electric fieldin the tunnel junction. [158] The negative differential resistance (a regionof decreasing current with increasing bias) above the U1 state is consis-tent with an increasing bias-dependent barrier height seen previously whenmolecules are separated from a metal surface by an insulator.[158] As thebias is increased after tunneling into a sharp resonance, the tunneling currentbetween the tip and the sample decreases due to orbital mismatch.[65, 159]503.3. Isolated PTCDA on NaCl(2ML)/Ag(111)Figure 3.4: STS point spectra of an isolated PTCDA molecule on NaCl(2ML)/Ag(111), (It = 1 pA). The differential conductance (dI/dV) (rescaledby a factor of 13 for It = 1 pA in gray) and normalized differential con-ductance (dI/dV)/(I/V) are shown (region near zero with divergences withdashed line). The tunneling resonances are identified as O1, U1, and U2.Pixel-by-pixel mapping of an isolated PTCDA was performed in order tospatially resolve the molecular states (Figure 3.5). We observe two differentspatial electronic distributions corresponding to the O1 resonance (at thepeak energy of -0.7 V and the shoulder at -0.8 V), indicating two differentstates. This is not the case with the U1 state, which shows one spatialsymmetry throughout the bias range 0.5 V to 1.5 V. Two features are againseen in U2 at +2.12 V and +2.22 V. As well, there is an area of increasedelectron density surrounding the molecule corresponding to the salt/silverinterface state at +0.18 V. This is similar to other features seen previously inSTS maps of molecules and is attributed to the charging of a molecule.[160]513.3. Isolated PTCDA on NaCl(2ML)/Ag(111)Figure 3.5: STS point spectra and pixel-by-pixel maps of an isolated PTCDAon NaCl(2ML)/Ag(111). a. STS averaged over an entire molecule and back-ground NaCl(2ML)/Ag(111) spectra, (set-point It=1.5 pA, Vb=-2.1 V). In-set: grid topography (5 nm x 5 nm, It = 1.5 pA, Vb = −2.1 V) with molecularposition overlaid. b. Energy maps of an isolated PTCDA molecule corre-sponding to the dotted gray lines in (a), (5 nm x 5nm, at Vb = −0.80 V,−0.70 V, 0.18 V, 0.65 V, 2.12 V, and 2.22 V).PTCDA on NaCl(2ML)/Ag(111) is negatively charged. This is due toelectron transfer from the underlying metallic surface to the lowest unoccu-pied molecular orbital (LUMO). Mohn, et al determined that on NaCl(2ML)/Cu(111), PTCDA is negatively charged,[154] based on the small work func-tion of NaCl (2 ML)/Cu(111)[161] in comparison with the electron affinity ofPTCDA[162]. The work function of NaCl (2 ML) on Ag is even smaller thanthat on Cu.[163, 164] We therefore conclude that, in our case, the degree ofpopulation of the LUMO (and thus the amount of negative charge on themolecule) is the same as on NaCl (2 ML)/Cu(111). This negative chargestate of the molecule is consistent with the observed repulsion and scatter-ing of the NaCl (2 ML)/Ag(111) interface state electrons[153] by PTCDA(Figure 3.5b at 0.18 V).From the STS and the surface induced negative charge of PTCDA, weidentify O1 as the overlapping HOMO and LUMO−1→0 transition, U1 as523.4. PTCDA nanoislands on NaCl(2ML/Ag(111)the LUMO−1→−2 transition, and U2 as the nearly degenerate LUMO+1and LUMO+2.4[165] Unperturbed, the half-occupied LUMO should appearat the Fermi energy (EF ). However, the addition of a charged particle(electron addition or removal) to this system results in the splitting of thestate due to Coulomb repulsion into peaks above and below EF .[166] Theseparation between these states is the Hubbard energy, UH (Figure 3.6).Further studies of the Hubbard states of PTCDA examined by scanningprobe microscopy are detailed in the next chapter.Figure 3.6: Electronic states of PTCDA adsorbed on NaCl(2ML)/Ag(111)with tunneling. a. Diagram showing LUMO splitting upon the addition orremoval of an electron b. Predicted corresponding electronic resonances3.4 PTCDA nanoislands on NaCl(2ML/Ag(111)3.4.1 Geometry of nanoclusters on NaCl bilayersThere are two predominant geometries of bulk crystalline PTCDA, referredto as the α and β polymorphs (Figure 3.7), both exhibiting a “herringbone”structure.[146, 152] However, monolayer and submonolayer PTCDA haveshown many different ordered structures.[85, 143, 147, 150, 151, 165, 167]For submonolayer coverage, PTCDA will self assemble into clusters.[168]The formation of the clusters indicates that the H-bond network formedovercomes the Coulomb repulsion between the negatively charged molecules.The negative charge can be partially screened by the NaCl(2 ML)/Ag(111)interface state and Ag(111) conduction electrons, as has been observed forother polarizable surfaces in previous studies.[168]4In this thesis when superscripts are used to describe charge states, or transitionsbetween charge states, e.g. LUMO LUMO−1→0 refers to the removal of an electron fromthe singly occupied lowest molecular orbital. In–line numbers indicate the specific orbital,e.g. LUMO+1 refers to the second lowest lying molecular orbital.533.4. PTCDA nanoislands on NaCl(2ML/Ag(111)Figure 3.7: The two dominant polymorphs, α and β, of bulk PTCDA. Rect-angle indicates unit cell shift in adjacent layer.Figure 3.8 is an example of a room temperature annealed sample withclusters formed with a longer anneal time (∼ 8 minutes). These nanoislands,exhibiting different sizes and structural arrangements, are well ordered andhave defined edges that serve as a model for an abrupt interface. Largerislands of ∼ 20 molecules or more (e.g. those seen in Figure 3.9d and e) oftenstart to self-assemble into the herringbone structure seen in monolayer andbulk PTCDA. An assortment of islands is shown in Figure 3.9. Occasionally,PTCDA will self-assemble into an ichthyological structure,5 Figure 3.9b.5Ichthyology is the branch of zoology relating to the study of fish.543.4. PTCDA nanoislands on NaCl(2ML/Ag(111)Figure 3.8: STM constant current image of PTCDA on NaCl(2ML)/Ag(111)(45 nm x 45 nm, It = 30 pA, Vb = −1.5 V) showing larger cluster formation.553.4. PTCDA nanoislands on NaCl(2ML/Ag(111)Figure 3.9: A “molecular zoo” of PTCDA clusters. STM constant currentimages of a. two geometries of 4 molecule clusters (16 nm x 7 nm, It = 15pA, Vb = 0.5 V), b. a fish shaped island (10 nm x 10 nm, It = 30 pA, Vb =0.55 V), c. a 12 molecule pinwheel (9 nm x 9 nm, It = 30 pA, Vb = 1.0 V),d. a 22 molecule island (10 nm x 10 nm, It = 30 pA, Vb = 1.0 V), and e. alarger island showing the bulk “herringbone” structure (20 nm x 20 nm, It= 30 pA, Vb = 1.0 V).Here we focus predominantly on small clusters of 4, 12 and 18 molecules.On the basis of the orthogonal orientation of molecules within the 4- and12-molecule clusters (Figure 3.9a, c, and 3.14a), we deduce that the theNaCl adsorption configuration is the same as for isolated PTCDA at low(<12 molecules) PTCDA coverage.Geometry with respect to other moleculesFigure 3.10 shows the molecular positions of a 12-molecule pinwheel islandwith respect to the underlying substrate. Here, we classify molecules intothree categories (A, B, C), based on their position (edge or center) withinthe 12-molecule nanoisland and the degree of interaction with surroundingmolecules. Edge molecules of type A and B differ in the number of nearestneighbors: type A with two, type B with three. Both A and B each have563.4. PTCDA nanoislands on NaCl(2ML/Ag(111)three H-bonds with two neighboring PTCDA molecules. Type A has oneH-donor (proton donating) and two H-acceptors (proton accepting). TypeB has two H-donors and one H-acceptor. Type B is also positioned head-to-head with another PTCDA molecule where the electron rich anhydridegroups face each other. center molecules, C, are fully surrounded by fivePTCDA molecules and form six individual H-bonds, with three H-donorsand three H-acceptors. They also each have a single head-to-head anhydrideinteraction (with B molecules).Figure 3.10: a. Structural model of 12-molecule island showing positions onthe NaCl lattice. Molecule identifications (A, B, C) are indicated. b. Modelof the hydrogen bond network within the nanoisland. Partial charges (δ)contributing to the hydrogen bonding are shownGeometry with respect to the substrateNaCl(100) bilayers on Ag(111) grow predominantly along the <112> direc-tion. The 11% lattice mismatch between the salt and silver lattices results ina Moire´ pattern.[131] This corresponds to a spatial modulation of the workfunction, and therefore of the electrostatic potential, which could influencethe electronic structure and therefore tunneling spectrum. We determinedthe spatial distribution of the Moire´ pattern underneath the 12-moleculeisland (Figure 3.9c) by using a Fourier filter[169] to remove the componentscorresponding to the PTCDA island from the (dI/dV)/(I/V) STS map inFigure 3.11a. That is, Figure 3.11b is a real-space image where the Fouriercomponents related to the Moire´ pattern in Figure 3.11a only were consid-ered. The positions of the PTCDA molecules within the island with respectto this superstructure were deduced by superimposing the molecular model573.4. PTCDA nanoislands on NaCl(2ML/Ag(111)onto the STM topography and the Moire´ interference lattice (Figures 3.11d-f).We observe that molecules with equivalent geometry with respect toother PTCDA molecules that lie on inequivalent sites of the Moire´ pattern[and therefore on inequivalent sites of Ag(111)] show nearly identical STS(the only noticeable difference is the intensities of the U2 peaks with minimalinfluence on peak energy). From this we conclude that the molecular spectro-scopic properties are not significantly influenced by the NaCl(2ML)/Ag(111)electronic structure. This enables the investigation of molecular geometrywith respect to other molecules, as there is not a significant substrate ad-sorption site-dependence. Defects surrounding the island (Figure 3.11a) areconsistent with water clusters.[170] They also do not appear to significantlyperturb the electronic properties of the molecules within the cluster.583.4. PTCDA nanoislands on NaCl(2ML/Ag(111)Figure 3.11: a. (dI/dV)/(I/V) STS map of a 12-molecule PTCDA island atVb = 1.45V showing the underlying Moire´ pattern (8.5 nm x 8.5 nm, It =30 pA, set-point bias Vb = −1.5 V). b. NaCl(2ML)/Ag(111) Moire´ patternobtained by Fourier-filtering the image components in (a) corresponding tothe PTCDA island. c. Overlay of (a) and (b). d. STM topography (Vb= −1.5 V, It = 30 pA ) corresponding to STS map in (a). e. Adsorptionpositions of molecules with respect to the Moire´ pattern. f. Geometricallyequivalent PTCDA molecules of type A showing inequivalent adsorptionsites with respect to the Moire´ pattern. g. (dI/dV)/(I/V) spectra of thefour geometry equivalent A molecules within the cluster located on differentsites of the underlying lattice.593.4. PTCDA nanoislands on NaCl(2ML/Ag(111)3.4.2 STS mapping of PTCDA nanoislands on NaClbilayersTo examine the local electronic structure of PTCDA clusters, we performedpixel-by-pixel STS to construct three dimensional grids of (dI/dV)/(I/V)in (x, y, E) for several different island geometries (examples shown in Fig-ures 3.12, 3.13). The electronic features of the isolated molecule (Figure3.5) significantly differ from those of PTCDA in a nanoisland, such asthe 12-molecule “pinwheel” island introduced in Figure 3.10. Each of thestructurally equivalent positions are also electronically equivalent; all fourmolecules of the same type (A, B, or C) show nearly identical (dI/dV)/(I/V)signatures (thin lines in Figure 3.12 show individual molecules while thicklines are averaged spectra). However, the relative position within the clusterdoes influence the spectroscopic signature for each type of site. The reso-nance U1 is significantly broadened and reduced in intensity compared tothe isolated molecule for all positions and appears spatially delocalized inthe corresponding STS map at Vb = 0.50 V (Figure 3.12). The maximumintensity of U1 also shifts towards EF , consistent with a reduction of theon-site electrostatic repulsion (Hubbard U) due to screening by neighboringmolecules.[171] For the occupied state O1, a clear difference is seen betweencenter and edge molecules: A and B show a shift away from EF with re-spect to a single PTCDA molecule (dashed line), while the center moleculesC show a shift towards EF . These shifts produce a ∼300 meV difference inthe energy of O1 between edge and center molecules. The spatially resolvedSTS maps corresponding to the peak voltages of O1 between edge and cen-ter molecules (Figure 3.5) Vb = −0.84 and −0.53 V) respectively, show theclear spatial separation of the local density of states at these energies. Theunoccupied resonance U2 shows a similar onset for all three molecular po-sitions (∼2 V). However the peak intensities and positions differ for A, B,and C. The corresponding STS maps (Figure 3.5) Vb = 2.08, 2.19, 2.32, and−0.53 V) also clearly show this position dependance.603.4. PTCDA nanoislands on NaCl(2ML/Ag(111)Figure 3.12: STS of a 12-molecule PTCDA island. a. (dI/dV)/(I/V) spectrafor molecule types A, B, and C within the cluster (blue, cyan, and green re-spectively). Thick curves represent an average over all equivalent molecules.Thin curves are averaged over individual molecules. gray vertical lines de-note bias voltages of the STS maps shown below. Inset: STM topographicscan taken during grid acquisition with spatially averaged spectra locationsrepresented by colored boxes (8.5 nm x 8.5 nm, It = 30 pA, Vb =−1.5 V).b. Corresponding STS maps at increasing sample bias (8.5 x 8.5 nm2, Vb =−0.84, −0.52, +0.50, +2.08, +2.19, +2.32, +2.49 V)For comparison, two 4-molecule islands with different geometries (Figure3.13) were also examined. The two islands are identified as “clover” (withfour-fold symmetry) and “diamond” (with two-fold symmetry and the sameA and B-type H-bonding motif as seen in the 12-molecule cluster). Inter-estingly, these two clusters contain all the motifs of the 12 molecule island.Molecules with equivalent geometry are also spectroscopically equivalent.In the diamond island, PTCDA types A and B differ from those in the 12-molecule island only in next-nearest neighbor, whereas these are absent inthe 4-molecule island. Feature O1 for A and B in the diamond island differsby an energy of ∼100 meV. In contrast, O1 is indistinguishable for A and Bin the 12-molecule cluster (within the 12 meV measurement resolution). Theenergy onset of U2 in A and B does not change between the two geometriesof islands, although relative peak intensities of the close-lying states above∼2 eV are influenced by the presence or absence of a next-nearest neighbor(diamond vs 12-molecule island). The influence of next-nearest neighbors613.4. PTCDA nanoislands on NaCl(2ML/Ag(111)on the local electronic structure provides evidence of longer-ranged inter-actions, either via through-space electrostatic effects or via weak electronichybridization within the planar PTCDA structure.Figure 3.13: STS of two configurations of 4-molecule PTCDA islands. a.(dI/dV)/(I/V) spectra of a “diamond” and a “clover” shaped four moleculeisland. Dashed gray lines indicate bias voltages of maps shown in (b–i).Inset: STM topographic scan (14 nm x 7 nm, It = 30 pA, Vb). b.–i. Corre-sponding STS maps at increasing sample bias (14 nm x 7 nm, Vb = −0.92(b), −0.85 (c), −0.67 (d), +0.20 (e), +0.56 (f), +2.14 (g), +2.35 (h), +2.47V (i).An 18-molecule PTCDA island was also examined (Figure 3.14). Edgeand center molecules with the similar bonding configuration as in the 4-and 12-molecule islands were identified, as well as molecules with an inter-mediate level of coordination, which we label D. These molecules have anoccupied resonance, O1 located in between that of the edge and the center623.4. PTCDA nanoislands on NaCl(2ML/Ag(111)molecules. Two molecules do not show the expected behavior. The moleculeS has a spectrum similar to that of an isolated PTCDA molecule. In thepositive bias imaging it appears as a bright spot, which appeared after atip induced motion, and has a protrusion from the side, indicating it couldbe two PTCDA molecules adsorbed on top of each other. For the othermolecule, E, we attribute the unpredicted STS to the molecule lying on adefect of the underlying substrate, which also shifts the adsorption site andenergy levels relative to the adjacent PTCDA. Each of these anomalies wasonly observed once.In all cases, different molecular environments are correlated with mea-surable differences in spectroscopy. The clusters typically show narrowerO1–U2 gaps near the centers (where one can be defined) than at the edges.633.4. PTCDA nanoislands on NaCl(2ML/Ag(111)Figure 3.14: (dI/dV)/(I/V) spectra of 18-molecule island on NaCl(2ML)/Ag(111). a., b. STM constant current images of an 18-molecule PTCDAisland adsorbed on NaCl(2ML)/Ag(111) (9.5 nm x 9.5 nm, It = 30 pA, (a.)Vb = +1.0 V and (b.) Vb = −1.5V). c. Adsorption geometry of 18-moleculeisland showing the location on the NaCl lattice. d. (dI/dV)/(I/V) spectra ofall molecules in the 18-molecule cluster with curves averaged over individualmolecules identified in inset. e. STS shown in d. shown to visualize theoccupied states.643.4. PTCDA nanoislands on NaCl(2ML/Ag(111)3.4.3 Gap mapsTo visualize the resulting energy level alignment within PTCDA clusters,the voltage onsets of O1 and U2 were determined for each pixel in four dif-ferent islands, Figure 3.15 b–e, from the bias at which the onset of the statecrossed a threshold (Figure 3.15a). We define the corresponding “band gap”as the voltage difference between these two onsets (Figure 3.15a). Electronicfeatures O1 and U2 were chosen due to their correspondence with the trans-port gap measured by photoemission[64] and the strong suppression of themid-gap U1 state in clusters. The corresponding gap maps (Figure 3.15 n–q)show a difference of ∼400 meV between the center and edge molecules. Thisis dominated by a downward shift of O1 for the edge molecules relative tothe center molecules, influencing both the gap and energy level alignmentat this interface. Notably, different PTCDA adsorption motifs (for example,herringbone in Figure 3.15e, open square phase in b,c) show similar energyshifts of the electronic features at the edges compared to the cluster centers.653.4. PTCDA nanoislands on NaCl(2ML/Ag(111)Figure 3.15: Local energy level alignment within PTCDA islands of differentsizes. a. For each (dI/dV)/(I/V) spectrum acquired at a given tip position(x,y), “band edges” are defined as the voltage onsets ((dI/dV)/((I/V) =3) of states O1 and U2. The band gap is the voltage difference betweenthese onsets. (b.–e.) STM topographic images acquired during spectro-scopic measurement for a 4-molecule, 12-molecule, 18-molecule island, andherringbone nanoribbon, respectively (It = 30 pA, Vb = +1.5 V, (b) 6 nm x6 nm, (c) 8.5 nm x 8.5 nm, (d) 9.5 nm x 9.5 nm, (e) 8 nm x 8 nm). (f.–i.)2D (x,y)-dependent maps of O1 voltage onset for (f) 4-molecule island, (g)12-molecule island, (h) 18-molecule island and (i) herringbone nanoribbon.(j.–m.) Corresponding 2D (x,y)-dependent maps of voltage onset of U2.(n.–q.) Corresponding 2D (x,y)-dependent maps of band gaps. Note, inf.–m. black corresponds to no detected onset.663.5. Polarization of PTCDA3.5 Polarization of PTCDATo investigate the influence of the local, inhomogeneous electrostatic envi-ronment, we consider the site-specific stabilization of the charge added or re-moved during tunneling spectroscopy by the nearly instantaneous electronicpolarization of all other molecules in the cluster.[171] This stabilization ofthe transient charge that occurs in a single-particle spectroscopy, such asSTS or photoemission, results in a decrease in the measured ionization po-tential and increase in the electron affinity, narrowing the observed transportgap.3.5.1 Microelectrostatic calculationsA simple microelectrostatic model was implemented in order to determinethe energy shifts of electronic levels due to the local polarization of theneighboring molecules for three islands.6Each PTCDA molecule was treatedas a point charge located at the center of the molecule (Figure 3.16), and thepolarization response of all other molecules was considered in computing thestabilization energy of that charge. These calculations assume that there iszero intermolecular overlap. While small overlap is expected in this case,zero overlap is necessary for a practical and accurate approximation of theresponse to an induced field due to molecular polarizability.[19, 172].The position and orientation of the molecules was determined from theknown adsorption of PTCDA on the underlying NaCl lattice. The in-plane,anisotropic polarizability of PTCDA, αMM (short axis, 50.3 A˚3) and αLL(long axis, 88.2 A˚3), previously calculated by density functional theory, wasused.7[173]6Polarization effects occur with a lifetime much longer than the associated tunnelingrate with currents typically used in these STS experiments (2-40 pA[134]).7For the calculations, polarizability was expressed in SI units, with the conversion,(4pi0)A˚3 = (4pi0)10−30m3 = 1.11 ∗ 10−40C2m2J−1673.5. Polarization of PTCDAFigure 3.16: a. In-plane polarizability tensors of a PTCDA molecule. b.point charge location determined by the center of the PTCDA moleculeon the underlying NaCl lattice, the angles used for the microelectrostaticcalculations are also shown.In general, the point charge representing the transient molecular ionformed during tunneling creates an induced dipole moment, ~µ, in each ofthe other molecules:~µ = α · ~E (3.1)with the electric field, ~E, given by:~E =e4pi0R2Rˆ (3.2)where Rˆ is the displacement vector connecting the charged molecule and thepolarized molecule, e is the charge of an electron, and 0 is the permittivityof free space. For convenience we write R for |~R|, and use Rˆ to representthe unit vector along ~R, and use subscripts to denote the components of avector in that direction.To obtain the polarization stabilization energy, Ep:Ep =∣∣∣−~µ · ~E∣∣∣ (3.3)where by convention Ep is positive, and reduces the ionization potentialand increases the electron affinity. For a cluster, a sum of all contributinginduced dipoles from all other molecules in the cluster gives the total EP .For PTCDA, as with many organic molecules, the polarizability α isanisotropic and is represented by a tensor:α =αMM 0 00 αLL 00 0 αNN (3.4)where M , L, and N represent the short, long and normal molecular axesrespectively. Here we only consider the two components αMM and αLL in the683.5. Polarization of PTCDAplane of the molecule as the molecules adsorb nearly flat on the surface[151]and there is minimal electric field component normal to the surface createdby a point charge in-plane.To compute the induced dipole, we must transform the electric fieldvector into the frame of the molecule, which yields a simplified expressionin terms of the angles between the molecular axes. Instead, to work incartesian coordinates, we can write the expression for the induced dipole as:(µxµy)=(cos(−θ) −sin(−θ)sin(−θ) cos(−θ))(αMM 00 αLL)(cos(θ) −sin(θ)sin(θ) cos(θ))(ExEy)(3.5)where θ is the angle of the long axis of the molecule with respect to x, andEx and Ey can be expressed as:Ex =∣∣∣ ~E∣∣∣ sinβ (3.6a)Ey =∣∣∣ ~E∣∣∣ cosβ (3.6b)withsinβ =RxR(3.7a)cosβ =RyR(3.7b)where β is the angle between ~R and the long axis of the molecule.Due to the orthogonal orientation of the PTCDA molecules on NaCl(2ML)/Ag(111), with θ = 0 or 90◦, we can simplify this to:µx = (A · αMM +B · αLL)Ex (3.8a)µx =e4pi0R2(A · αMM +B · αLL)RxR(3.8b)µy = (B · αMM +A · αLL)Ey (3.8c)µy =e4pi0R2(B · αMM +A · αLL)RyR(3.8d)with A = 0 and B = 1 for a molecule aligned with the x-axis and A = 1 andB = 0 for a molecule aligned with the y-axis. This results in a polarizationenergy due to the ith PTCDA molecule as:Ep,i =e216pi220R4i[(Ai · αMM +Bi · αLL)(Ri,x)2|Ri|2++ (Bi · αMM +Ai · αLL)(Ri,y)2|Ri|2](3.9)693.5. Polarization of PTCDAThe contribution from each molecule is summed, and equivalently we canwrite(Ep)total =e216pi220[Nx∑i=1(αLLsin2βi + αMMcos2βi)1R4i++Ny∑j=1(αMMsin2βj + αLLcos2βj)1R4j (3.10)Simplified for the case of orthogonal molecules:(Ep)total =e216pi220N−1∑i=1(αMMsin2βi + αLLcos2βi)1R4i(3.11)3.5.2 Nanoisland polarization energiesThe results of this calculation for three different cluster sizes and geometries(those in Figure 3.15b – d) are shown in Figure 3.17 and compared with themeasured positions of the O1 state.8 The qualitative trends observed in thedata for different types of sites in each of the clusters are generally well repro-duced by the shifts predicted by the polarization energy calculation acrossall three cluster sizes and geometries. For example, for the 12-molecule is-land, the center (C) molecules show an O1 state that is closer to EF thanthe A/B edge sites. Even more strikingly, for the 18-molecule island, nearlyall 18 sites, each unique in geometry, follow the overall trend predicted bythe polarization calculation. Here A, B and C-type positions bear resem-blance to the A, B, and C identified in the 4- and 12-molecule clusters interms of nearest neighbors and H-bonding, while D-type molecules lie at anedge, but are more fully surrounded than either A or B-type sites. The O1states measured for D-type positions were noted to lie between those of thecenter molecules and A/B edge molecules, which is similarly reproduced inthe calculation. Two outliers labeled S and E showed atypical spectroscopicbehavior that we attribute to defects in the cluster or underlying substrate.8The voltage position of O1 was determined from finding the maximum of the peakvisually. This was done by several people multiple times on different days with the stan-dard deviation plotted in Figure 3.17. In most cases, the error bars are not visualizedas they are smaller than the data marker. Assigning error of STS measurements is chal-lenging due to the nature of the states, which are not always single peaks and often shownon-Gaussian behavior. Gaussian fits were attempted, but did not show an improvementover visual identification. This is a corrected version of the figure published in [133]703.5. Polarization of PTCDAFigure 3.17: Occupied state of PTCDA islands with polarization energy.The result of the microelectrostatic calculation (crosses) was plotted for eachmolecule in three islands: 4-molecule, 12-molecule, and 18-molecule, andcompared with the peak of the O1 state (dots). The specific molecules withinthe islands identified on the x-axis are labeled in the schematic structure ofeach cluster above. The vertical axes of O1 and Ep differ by a factor of2.8/e.In these calculations, all molecules in the cluster were considered. Wefind that next-nearest neighbors each still contribute a few percent to thetotal Ep, while next-next nearest neighbors still contribute ∼0.5% each tothe total Ep. The individual contribution towards the polarization energyfor a “B” molecule in the 12-molecule island is shown in Figure 3.18. Whilethe shifts in energy are strongly localized to the edge, the effect arises fromthe interaction with molecules several sites away.713.5. Polarization of PTCDAFigure 3.18: Vector representation of induced dipole moment in the 12-molecule island with respect to molecule B1. Width of arrows and trans-parency of ovals correspond to strength of induced dipole. Relative contri-bution to Ep respectively from each molecule is indicated.Although the polarization energy need not be equal for both the removaland addition of charge,[174] we expect a shift of the unoccupied states dueto polarization as well. The onset of U2 does shift towards Ef for all sitesrelative to the isolated molecule spectrum, however, there is minimal spa-tial variation of the U2 onset. The onset of the O1 state for A/B edgemolecules also lies below the O1 onset for the single-molecule spectrum,implying an overall downward shift of the spectra indicative of a chargetransfer between the cluster and the underlying Ag(111) substrate. As inmixed monolayers,[139] this charge transfer may be spatially varying due tothe different relative shifts of the U1 and U2 states for different sites withan alignment of the unoccupied states, here mediated by the U1 mid-gapstate. Recent theoretical literature has demonstrated site-specific chargetransfer on a molecule/insulator/metal system leading to spatially varyingcharge on molecules within a cluster with an insulating barrier to the metalsubstrate.[175] Assuming equal polarization energies for electron additionand removal, and a site-specific charge transfer that aligns the U2 states asobserved, there is a nearly quantitative agreement between the measuredand predicted position of the O1 state (Figure 3.19).723.5. Polarization of PTCDAFigure 3.19: Electronic state energies relative to EF derived from the mea-sured spectrum of the isolated molecule (left), applying the calculated po-larization energies for the edge (blue) and center (green) of the 12-moleculecluster (middle), and accounting for the observed charge transfer as a rigidshift of the energy levels and compared to the experimental STS (right).Red arrows above clusters depict the induced polarization after addition ofa tunneling electron.The remarkably good agreement between the experimental data and thepredictions of the microelectrostatic model, for clusters of different size andgeometry, provides compelling evidence that the primary mechanism dic-tating the observed site-specific energy levels arises from the stabilizationof charges by the electronic polarization of neighboring PTCDA molecules.The calculation neglects several other possible origins of energy level shiftsincluding: conformational differences due to differing intermolecular interac-tions, in-plane hybridization, and differences in H-bonding. Notably, subtledifferences in hydrogen bonding between adjacent PTCDA molecules havebeen implicated in ongoing discrepancies between experimental and DFT re-sults for PTCDA monolayers on Ag(111) and Ag(100),[149][46], highlightingsome of the computational challenges inherent in interfacial systems, partic-ularly where covalent and non-covalent interactions compete and electroniccorrelations cannot be neglected. These effects, and the inhomogeneousintramolecular charge distribution within the molecule, neglected in the cal-culation, may explain the deviations between the predicted and measuredO1 position.733.6. PTCDA on multilayer NaClThe correspondence between the measured electronic energy level shiftsand the calculated polarization energies is aided by the weak in-plane hy-bridization of PTCDA,[142] as one can adequately describe the injectionand removal of charges by a localized molecular ion. Although Temirov etal.[168] observed a delocalized interface state arising from the interaction ofPTCDA with the metallic substrate, the NaCl bilayer used here suppressesmolecule-substrate hybridization, and our observations are expected to berepresentative of electronic effects that can be attributed to the intrinsicmolecular and intermolecular interactions within the PTCDA clusters. Asthe majority of organic semiconductors are characterized by weak orbitaloverlap (small hopping integrals) and conjugated molecules typically havelarge polarizabilities, these electronic polarization effects are expected to besignificant, if not dominant, relative to other interfacial electronic effects inmany cases. This simple electronic polarization picture adequately capturesthe gross features while being computationally straightforward and tractablefor large systems where ab initio formalisms including van der Waals inter-actions and correlations would be inaccessible. The ability to study largesystems via an analytic model where electronic polarization plays a domi-nant role opens up the possibility of studying realistic interfacial systems,including the effect of disorder, by determining only the position, orientationand anisotropic polarizability of the molecular components.3.6 PTCDA on multilayer NaClPTCDA on tri- and quad-layer NaCl were also examined to observe the effectof further screening of the molecule from the underlying silver substrate.93.6.1 Isolated PTCDA on multilayer NaClFigures 3.20a and b show sequentially acquired STM topographic imagesof PTCDA molecules on a bi-, tri-, and quad-layer NaCl. The moleculeson the quad-layer are unstable and easily moved by the STM tip. Withincreasing insulating film thickness, the distance between the tip and themolecules adsorbed on the sample decreases in order to achieve the set-point tunneling current. The movement of a molecule from the quad- tothe trilayer is observed. STS (Figure 3.20c) shows significant shifts withdifferent film thicknesses, particularly in the U1 and U2 states.9All of the clusters discussed previously were on bilayer NaCl.743.6. PTCDA on multilayer NaClFigure 3.20: a. and b. STM topography (40 nm width, It = 7 pA and Vb= −0.8 V and +0.5 V respectively). Note movement of a PTCDA moleculefrom a quad- to a trilayer, indicated by black arrow in (b). c. Comparisonof normalized STS on an isolated PTCDA molecule on NaCl(2ML)/Ag(111)(black), NaCl(3ML)/Ag(111) (maroon), and NaCl(4ML)/Ag(111) (red) (lo-cations of spectra indicated by markers in (b, bilayer spectra taken froma different scan). d. Lateral profile of bi-, tri- and quad-layer NaCl layersshowing step height (location indicated by turquoise line in (b).Comparison the shifts in energy levels between PTCDA molecules onincreasing NaCl layer thickness are shown in Figure 3.21. A widening ofthe Hubbard gap (UH ' 1.36, 1.61 and 1.65 eV for bi-, tri- and quad-layerrespectively) is observed as molecules are further away from the substrate.This may indicate the molecules on the tri- and quad-layer molecules are lessscreened by the underlying substrate as they are further decoupled by NaCl753.6. PTCDA on multilayer NaCllayers. The Hubbard energy is sensitive to the local screening environmentso this change in UH with distance from the underlying Ag is consistentwith this description of the observed tunneling resonances. There is alsoan overall shift towards higher voltages as molecules are further decoupledfrom the silver, indicating a less negative charge, and therefore less electrontransfer from the underlying substrate with an ionization potential decrease.It is also possible the shifting is due to a tip-induced field effect, howeverno observable shifts due to tip-sample distance were observed on the bilayer(Figure 3.3). Spectra could not be taken at the range of currents necessaryfor tip-height dependence on bi- and tri-layers due to the instability of themolecules and the noise floor of the instrument.Figure 3.21: Diagram comparing energy level alignment of PTCDA on 2ML,3ML, and 4ML NaCl on Ag(111). STS is shown for all three systems. TheHubbard gap is indicated by the semi-transparent box in between O1 andU1.These results show that screening by the underlying Ag is not negligible,however changes in the polarization environment due to the neighboringmolecules still provide contrast on the flat background of the Ag screening.763.7. Summary3.7 SummaryHere, local topographic and spectroscopic measurements performed withsub-molecular resolution on 2D nanoscale clusters of PTCDA have revealeda striking difference between the electronic states of molecules residing atthe edges of these clusters and those in the center. Edge molecules ex-hibit a gap that is up to 400 meV larger than observed for inner molecules(representative of a 2D “bulk”), arising primarily from a shift in the oc-cupied state energies that correspondingly influences level alignment for aboundary region of single-molecule width. These site-specific energy levelshifts measured by STS arise from differences in the local electronic polar-ization environment provided by the cluster that responds instantaneouslyto a transient, localized charge. Electronic polarization effects are expectedto strongly influence hopping-like transport and processes such as photoin-duced charge separation at heterojunction interfaces[176] where a transientmolecular ion is formed. As the polarizability of most organic semiconduc-tors is anisotropic, both the local structure and orientation of molecules atinterfaces will play a significant role in the resulting energy level alignment.Yet, where these effects dominate, as they do in planar arrangements ofPTCDA, an accessible model can be used to predict interfacial energy levelshifts. This model can easily be extended to large and more complex sys-tems to address issues related to interface geometry and disorder, as wellas identify potential optimization paths through careful design of interfaceinteractions.77Chapter 4Mapping the HubbardEnergy of PTCDA4.1 Introduction and backgroundIn developing models of electronic and optoelectronic properties, we oftenstart with the ground state electronic structure. In reality, the properties weare interested in are a result of excitations, which can produce either charged(addition or removal of an electron) or neutral species (e.g. optical excita-tion creating an electron and hole pair).[23, 24, 29, 60] These excitationsmay introduce a perturbation to the ground state electronic structure.[32]In cases where the electronic (dielectric) screening is low, excitations mayinstead give rise to distinct states, e.g. excitons and other quasiparticlesand correlated states. When bands or orbitals are half-filled, the Coulombinteraction gives rise to an energetic cost for the addition of charge (posi-tive or negative).[166, 173] The case of half-filling where charges are poorlyscreened is typically described by an on-site electrostatic repulsion, oftencharacterized by the Hubbard potential (UH).[63]Half-filled orbitals provide one of the simplest examples of how theground state electronic structure does not fully capture the physics behindexcitations of the system needed for transport and other charged excitationsof the system. The addition of an electron to an already charged orbital nat-urally entails overcoming the Coulomb repulsion required to put a secondelectron in the same region of space.[177] In bulk metals and other materialswith a high degree of electronic screening, these charges barely “see” eachother and this interaction can be neglected. However in highly localized or-bitals, such as in molecules and materials with very low dielectric screening,the interaction upon adding an electron to a half-filled band or orbital canbe large. This may dramatically change the expected behavior of a system,e.g. making a predicted metal into an insulator.[166]Previously, we investigated clusters of PTCDA on bilayer NaCl islandson Ag(111) to examine the effects of electronic polarization (Chapter 3 and784.1. Introduction and background[133]). We observed a significant influence of molecular geometry on theelectron addition and removal spectrum measured by Scanning tunnelingspectroscopy (STS, Figure 4.1c).Isolated PTCDA molecules (Figure 4.1a) are singly negatively chargeddue to charge transfer through the thin insulating NaCl layer from the un-derlying Ag(111).[154, 156] This results in the half occupation of the lowestunoccupied molecular orbital (LUMO).[133] Tunneling into this state resultsin an upward shift of the bias at which the LUMO is measured due repulsionfrom the presence of another electron; likewise tunneling out of this stateresults in a downward shift of the LUMO (Figure 4.1)Figure 4.1: Isolated PTCDA on NaCl(2ML)/Ag(111) a. STM constant cur-rent topograph (3.5 nm x 3.5 nm, It = 10 pA, Vb = 1.8 V). b. Electronicstates showing splitting upon addition or removal of an electron during tun-neling c. STS acquired with a Pt-Ir STM tip (setpoint: It = 1.5 pA, Vb =-2.1 V) with corresponding electronic processes depicted.Here, we use Scanning tunneling microscopy (STM) and spectroscopy(STS) simultaneously alongside electrostatic force spectroscopy (EFS) tocharacterize the local charge distributions and charging energies for differ-ent charge states of molecules within clusters and isolated molecules of theprototypical organic semiconductor 3,4,9,10-perylenetetracaboxylic dianhy-dride (PTCDA).794.2. Charging of molecules examined with EFS4.2 Charging of molecules examined with EFSPixel-by-pixel EFS is a method of probing and manipulating local surfacecharges with submolecular resolution.[121] EFS can be utilized to detectcharging events, even due to just one electron. [120] If an atom or a moleculeundergoes a change in charge during the applied bias sweep, a dip or a jumpin the EFS spectrum is observed.[178–181] Evidence for a change in chargestate is shown by the model spectrum in Figure 4.2 where the signal “jumps”between two distinct parabolas. On bulk insulators[117] or thick NaCl films(>5ML)[178], these charges are stabilized and can be controlled. Directlyon metals, momentary charging has been observed occurring as a dip in theEFS curve.[179]Figure 4.2: A depiction of a df(V ) curve of a PTCDA molecule going througha charging event at bias VSW with contact potential differences VCPD1 andVCPD2 in the less negative and more negative charge state respectively.If pixel-by-pixel measurements are performed, the different charge statescan be spatially mapped. This is done by separately fitting the different seg-ments of the curve and spatially plotting the VCPD in each charge state.[117]The bias at which this charging event occurs, VSW , can also be mapped re-sulting in simultaneous energetic and spatial resolution of charging events.The capacitance of the tip–sample junction is assumed to remain constantthroughout the change in charge. This assumes that affects of the addition804.3. Experimentor removal of one electron are negligible compared to the dominant effectof the micro- and macro-scale geometry of the tip–sample junction. Thisreduces the fitting variables to two: the VCPD and the dfmax, which is nec-essary for reliable fits as some of the segments of the parabola are quiteshort (see Appendix B).4.3 ExperimentPTCDA nanoislands on NaCl(2ML)/Ag(111) were obtained by the proce-dure described in detail in Chapter 2 and also used in Chapter 3 (Figure4.3a). The measurements were performed in ultrahigh vacuum (UHV) at4.3 K with a low temperature (LT) scanning probe microscope (SPM). Ahomemade qPlus sensor with a separate gold wire contacting the tip for thetunneling current was used for all measurements, with f0 = 25500 Hz, Q≈ 40,000, and a sensitivity of ∼0.4 pm/mV10. The bias was applied to thesample to prevent coupling to the oscillation of the tuning fork. Constantheight NC-AFM images and electrostatic force spectra were taken by posi-tioning the tip over NaCl at 0.5 V and 2 pA using tunneling feedback, thenturning off the feedback loop. Both carbon monoxide (CO) and PTCDAfunctionalized tips were used for high resolution imaging. EFS grid mea-surements were taken with a metallic tip with acquisition times typicallyaround 100 ms per voltage point and tip oscillations of 40 – 80 mV (∼16–32pm); high resolution imaging was taken using tip oscillations of 20 – 60 mV(∼8–24 pm).4.4 PTCDA clustersTwo 4-molecule PTCDA clusters were studied: a 4-fold symmetric islandreferred to as “clover” shaped, and a 2-fold symmetric island referred toas “diamond” shaped (Figures 4.3b, c, and 4.4). Diamond islands werefound to be more prevalent than clover islands. For both, structures withopposing chirality were found with equal probability and determined to beelectronically equivalent from STS. In the diamond island, there are twodistinct types molecular environments, while in the clover island all of themolecules are adsorbed on equivalent sites (confirmed by STM, STS, andNC-AFM). The electronic equivalence of the molecules within the clover10Determined by changing the oscillation amplitude and observing the current andthe exponential dependence of tunneling current on tip-sample distance. Known as the”distance decay method”.814.4. PTCDA clustersisland suggests equal charge distribution within the island. The differingmolecular sites with distinct spectra in the diamond island allowed for theinvestigation of the impact of initial geometric and electronic inequivalenceon charging energy.Figure 4.3: STM imaging of PTCDA nanoislands on NaCl(2ML)/Ag(111).a. Overview of a sample showing small PTCDA clusters (50 nm x 50 nm, It= 15 pA, Vb = 0.5 V). b. and c. STM images of the two 4-molecule islandsstudied in this experiment: diamond (left) and clover (right) (14 nm x 6.5nm, It = 15 pA, Vb = 0.5 V (b.) and −1.5 V (c.).Figure 4.4: Geometries of two 4-molecule islands positioned on the underly-ing NaCl lattice: a. 2-fold symmetric diamond island b. 4-fold symmetricclover island.824.4. PTCDA clusters4.4.1 High resolution NC-AFMDiamond islandConstant-height sub-molecularly resolved images of the 2-fold symmetric di-amond island are shown in Figure 4.5. Two different tip functionalizationswere used: CO[100] and PTCDA[104]. The PTCDA anhydride oxygens ofthe surface molecule are not observed when the molecule is adsorbed on asalt bilayer.[154] We identify the two distinct sites of PTCDA molecules as Aand B (labeled in Figure 4.4a). A-site molecules have two nearest neighbors,while B-site molecules have three. In addition, the electron rich terminalanhydride groups of B molecules are positioned head-to-head. Differing de-grees of distortion of the benzene rings are observed with the two adsorptionsites of PTCDA molecules. This may indicate inequivalent molecular envi-ronments, either geometric or electrostatic, or is an artifact caused by thefunctionalized tip.Figure 4.5: Constant height NC-AFM images of the diamond 4-moleculePTCDA nanoisland with: a a. CO and b. PTCDA functionalized tips (4nm x 4 nm, Vb = 0 V). Two different tip functionalizations show differentdistortions of the adsorbed PTCDA carbon rings likely due to the nature ofthe distortions of the tip molecule.834.4. PTCDA clustersClover islandFigure 4.6 shows a constant height image of the 4-fold symmetric clover is-land with a CO functionalized tip. A circular feature can be seen in the cen-ter of the island, which we attribute to distortions of the tip CO molecule dueto the close proximity of the surface molecules to each other.[115] Thoughthe nature of these distortions appear in regions where hydrogen bonds(H-bonds) are expected to be seen, the ability to observe H-bonds withNC-AFM is still debated.[114, 182] However, due to the functional groupsinvolved, the apparent intermolecular distances, as well as the planar hy-drogen bonds within PTCDA crystals, we believe the molecules are in factH-bonded together to form this stable cluster. All molecules appear to havesimilar distortion of the benzene rings, indicating symmetric geometric en-vironments and charge states.Figure 4.6: Constant height NC-AFM image of a clover 4-molecule PTCDAnanoisland with a CO functionalized tip. (4 nm x 4 nm, Vb = 0 V).4.4.2 Spectroscopy of PTCDA nanoislandsSimultaneous EFS and STS was performed on both the diamond and theclover shaped islands to examine the effect of adding or removing a chargeon the local electrostatic environment and compare to the electronic statesobserved in STS.844.4. PTCDA clustersDiamond islandFor the 2-fold symmetric diamond island, electrostatic force spectra showsharp jumps in both A and B molecules, indicating sudden changes in lo-cal surface charge with changing bias. This occurs in the negative biasregime (occupied states) for the B molecules and the positive bias regime(unoccupied states) for the A molecules. The simultaneously acquired tun-neling spectra with a qPlus tip show additional features at biases near thesecharging events that are not present with a fixed STM tip, likely due to aback-action effect during tunneling.[183, 184]Figure 4.7: Simultaneous STS and df(V ) spectra of a “diamond” shaped 4-molecule island taken from constant-height pixel-by-pixel grid with set-pointon NaCl(2ML)/Ag(111), It = 2 pA, Vb = 0.5 V, oscillation = 60 mV.From the electrostatic force spectra, we expect that during a bias sweep,tunneling out of or into the half occupied LUMO results in a more positivelyor more negatively charged molecule respectively, leading to a different VCPDin each charge state (Figure 4.8). We do not believe this is a persistent charg-ing of the molecule, such as that seen in molecules on bulk insulators,[117]but rather a transient change in charge state: as electrons tunnel onto/offof the molecule a different average charge is detected by the AFM tip.854.4. PTCDA clustersFigure 4.8: A depiction of a df(V ) curve of a PTCDA molecule going throughthree charge states as the bias is swept.By extracting the biases at which charging events occur, indicated by ajump in the df(V ) curve (Figure 4.8), we can now also examine how the en-ergy required to change charge state varies spatially. These maps show howthe addition and removal of charge differs for the A and B molecules. Fig-ures 4.9a – c show maps of the three charge states: PTCDA0, PTCDA−, andPTCDA2−. In Figure 4.9b, the VCPD map corresponding to PTCDA− (thebias range where there are no electrons tunneling into or out of the molecularorbitals, i.e. the unperturbed condition), we observe that B molecules aremore negatively charged than A. Though the measured value of the VCPDis more positive, this indicates more relative negative charge.[120, 121]. Forthe PTCDA0 map (Figure 4.9a, when electrons are tunneling out of themolecule and into the tip), the A molecules are more negatively charged, in-dicating a loss of electrons primarily from the B molecules. In the PTCDA2−map (Figure 4.9c) the molecules are about equivalent in charge, indicatingthat the charge is predominantly added to the A molecules.864.4. PTCDA clustersFigure 4.9: a. – c. VCPD maps of a diamond shaped 4-molecule islandin three different charge states (PTCDA0, PTCDA−, and PTCDA2− re-spectively) determined from fitting three segments to df(V ) curves d. – f.lower Hubbard bias map, upper Hubbard bias map12 and gap from differencerespectively (5 nm x 5 nm).This addition and removal of local charge from the A and B molecules isalso indicated from the charging maps depicted in Figures 4.9d – f. In themap corresponding to the LUMO− to LUMO0 transition (Figure 4.9d), thecharging only occurs over the B molecules, while in the map correspondingto the LUMO− to LUMO2− transition (Figure 4.9e), the charging occurs12In locations where no jump is found, i.e. in the underlying substrate, the value atthat pixel is set to the extreme of the fit parameter range: −1 for LUMO− to LUMO0and +1 for LUMO− to LUMO2−.874.4. PTCDA clustersover the A molecules. Subtracting these two maps, results in a “Hubbardmap” (Figure 4.9f) which in this case results mostly in features in-betweenthe molecules as charging occurs on opposing sites.From the point spectra, VCPD maps, and Hubbard maps, we postulate apathway of charge transfer and local charge distribution within the diamondisland that differs depending on the electron injection or removal site (i.e.the tip position over molecule A or B). When the tip is positioned over anA molecule at positive biases, electrons are injected into the less negativelycharged species. This results in the acceptance of this charge by A molecules.However, at positive biases with the tip positioned over B, charge will notbe easily accepted into the more negative molecule. Instead of retainingthis charge, the electrons will hop either to the more positive A molecule orcontinue through to the substrate without having altered the net charge stateof B molecules. When negative biases are applied, the electrons tunnelingout of the more negative B molecule will result in a depleted negative surfacecharge. Figure 4.10 depicts this proposed electron pathway.Figure 4.10: Proposed diagram of electron movement during tunneling into(electron addition) and out of (electron removal) a PTCDA diamond island.δ+ and δ− representing relative positive and negative charges respectively.It has been demonstrated theoretically that charge transfer betweenmolecules in small clusters on insulator/metal substrates can result in spa-tially varying charge distributions.[175] The initial charge distribution withinthe island is consistent with the resonances observed in the (dI/dV)/(I/V)spectra (Figures 4.7 and 3.12a). The occupied O1 state in B molecules is fur-884.4. PTCDA clustersther upward (more positive in bias) in the B molecule, indicating a smallerionization potential (IP) and electron affinity (EA) and thus a more neg-atively charge. As well, microelectrostatic calculations of the polarizationenergy, Ep, show that the charges are more stabilized on the B moleculessites: (Ep)A = 0.062 eV and (Ep)B = 0.085 eV (see previous chapter, aswell as [133] for details on the microelectrostatic calculations and discussionof the effects of the polarization energy).It can be seen in Figure 4.7 that the biases of the surface charge switchdo not align with the resonances identified in STS. This indicates an addi-tional energy cost of the charge transfer within the molecule. The upperHubbard state observed in STS sits closer to ∼+0.5 eV. Since that corre-sponds to the on-site Coulomb energy, we would expect a charging event atthis energy when a second electron is injected into the same molecule. How-ever, although B-site molecules have a negative charge (presumably 1−),we should see a positive bias charging event at 0.5 V which corresponds tono A-site electrons, and 2 B-site electrons.Clover islandIn the EFS spectra of the 4-fold symmetric clover island (Figure 4.11), ajump was observed coinciding with the bias of the resonance assigned to theLUMO− to LUMO2− transition in the simultaneously acquired STS. Thisindicates a change in net local charge after tunneling into this orbital. How-ever, the transition corresponding to the the LUMO− to LUMO0 tunnelingresonance (mixed with the nearby HOMO) is seen as a subtle change in cur-vature or not at all with EFS. It is possible that this feature is too subtle tobe visually observed due to the proximity of the charging transition to thepeak of the parabola. It is also possible that the transiently neutral speciesis not present or short–lived, possibly due to backfilling from the substrate.The EA of PTCDA is significantly larger than the work function of bareAg and NaCl(2ML)/Ag(111).[156, 185] There is a strong driving force forcharges from the underlying substrate to re-occupy the “empty” LUMO ata higher rate than electrons tunnel through the vacuum to the tip. Thiswould result in the PTCDA molecules remaining in the nominally −1 state.However, the EFS data ranging from -1.5 V to the onset of the state on theunoccupied side could not accurately be fit with one parabola, indicatingsome change in the average charge. Assuming constant capacitance, theEFS curves could be fit with three separate parabolas corresponding to thedifferent charge states of PTCDA. (See Appendix B)894.4. PTCDA clustersFigure 4.11: Simultaneous STS and df(V ) spectra of a “clover” shaped 4-molecule island taken from a constant-height pixel-by-pixel grid with param-eters: set-point on NaCl(2ML)/Ag(111), It = 2 pA, Vb = 0.5 V, oscillation= 45 mV.The VCPD and Hubbard maps (Figure 4.12) show equivalent surfacecharge distribution and charging energies in each of the molecules, indicatingequal Coulomb repulsion.904.5. Isolated PTCDA on NaCl(2ML)/Ag(111)Figure 4.12: a. VCPD maps of a clover shaped 4-molecule island in three dif-ferent charge states determined from fitting three segments to df(V ) curvesb. lower Hubbard bias map, upper Hubbard bias map and gap from differ-ence (4.5 nm x 4.5 nm).4.5 Isolated PTCDA on NaCl(2ML)/Ag(111)An isolated PTCDA molecule was also examined by simultaneous STS andEFS (Figure 4.13). The df(V ) spectroscopy shows a jump on the unoccupiedbias side, but no observable jump or change in curvature on the occupiedside. As noted above for the clover island, this neutral state may not beaccessible or be short lived due to the strong electron affinity of PTCDArelative to NaCl(2ML)/Ag(111). The VCPD of the PTCDA2− state is shifted∼150 meV upwards (indicating greater negative surface charge) than in thePTCDA− state. From the maps (Figures 4.13b – c), it can be seen that theextra negative charge mostly appears in the “lobes” around the PTCDAmolecule that are also observed in STM imaging.914.5. Isolated PTCDA on NaCl(2ML)/Ag(111)Figure 4.13: a. Simultaneous STS and df(V ) spectra of an isolated PTCDAmolecule on NaCl(2ML)/Ag(111), taken from a constant-height pixel-by-pixel grid with parameters: set-point on NaCl(2ML)/ Ag(111), It = 2 pA,Vb = 0.5 V, oscillation = 50 mV. b. and c. VCPD maps of PTCDA− andPTCDA2− respectively. Molecule in maps is in the same orientation as theinset in a, (4 nm x 4 nm).The inability to observe a charging event on the occupied state preventedthe mapping of the VCPD PTCDA0 charge state or the LUMO− to LUMO0−transition. The map of this transition was created by spatially plottingthe onset of the O1 state extrapolated from the STS (Figure 4.14a) whichwe previously identified as arising from the lower Hubbard state, closelyoverlapping with the HOMO. Most striking is that the lobes, which arefurther from the already negatively charged core, have a lower Hubbard Uthan the middle, indicating that the starting electron density influences themagnitude of the Coulomb repulsion.924.6. ConclusionsFigure 4.14: a. Maps of the LUMO− to LUMO0 transition obtained fromscanning tunneling spectra of the onset of the closest occupied resonance toEF . (b.) Map of the jumps in bias of the df(V ) curves, corresponding tothe LUMO− to LUMO2− transition. c. Map of the gap between these twostates (all 4 nm x 4 nm).144.6 ConclusionsWe use EFS to examine the transient charging behavior of PTCDA clus-ters and an isolated molecule to investigate the influence of initial chargestate and distribution on the charging energy of a singly occupied molecularorbital. Pixel-by-pixel measurements allow us to map out the local chargedistribution in three different charge states, as well as the energy at whichelectron addition and removal occurs. By spatially resolving the chargingevents to generate “Hubbard maps” with sub-molecular resolution, we findthat the initial local charge distribution strongly influences the charge in-jection energy and site. In PTCDA clusters where molecules are initially indifferent charge states, negative charges are injected into the less negativesites, while electrons are removed from the initially more negative sites. Inan island with equivalent initial charge distribution, we observe that chargesare injected equally into all sites, but requiring the full on-site potential. Forthe single molecule, we see that even within a single orbital charge injec-tion depends on the initial charge distribution. Not surprisingly, charges aremore easily added (i.e. a lower energy is required to overcome the Coulombrepulsion) away from the already negatively charged core, and the final 2-charge state shows as an increased intensity on these surrounding extended14The color background color is saturated in Figures 4.14 a and c due to the compara-tively low noise of the STS measurements.934.6. Conclusionslobes.The ability to probe both the charge distribution and charging energy onA˚ngstrom length scales gives new insight into on-site and nearest-neighborinteractions that give rise to strongly correlated systems. Most strikingly,the on-site potential, i.e. the Hubbard U, varies within a single molecule,with charge injection occurring in the most extended regions and avoidingthe regions of highest initial electron density.94Chapter 5Energy Level Alignment of aBimolecular Heterojunction5.1 IntroductionFor organic photovoltaic (OPV) materials, understanding how energy levelalignment depends on molecular structure at an interface and surroundingenvironment is crucial to understanding and improving device performance.[186]As the soft inter- and intra-molecular interactions permit a wide range ofpossible structures and conformations,[43, 187] these interfaces have the po-tential to be highly inhomogeneous on the molecular scale that the excitonsees. Meanwhile, different interfacial structures have the potential to lead todifferent energy level alignment. [35, 36, 47, 133, 187] For example, Graham,et al. demonstrated that device performance could be significantly alteredwith subtle changes in interfacial molecular geometry driven by functionalgroups.[47]Scanning probe microscopy (SPM) offers the ability to resolve molecularand sub-molecular structure as well as local density of states yielding en-ergy level alignment on A˚ngstrom scales. Previous SPM studies on closedmonolayer systems of mixed donor-acceptor prototypical organic semicon-ductors have shown shifts in the energy levels of both components of themixed domains.[36, 37, 71, 139, 140] These shifts were attributed to a com-bination of interactions, including non-covalent intermolecular interactions,as well as intermolecular and molecule-substrate charge transfer. In thesemonolayer studies, the geometry of the interface was shown to be important,however the systems were formed by self-assembly which limits the availablegeometries. As well, often these systems were directly adsorbed on metallicsubstrates where strong interactions with the underlying surface complicateattempts to understand the intermolecular interactions that matter for theorganic–organic interfaces in OPVs.Here we used pixel-by-pixel scanning tunneling microscopy and spec-troscopy (STM/STS, as in Chapter 3) to probe small clusters of acceptor955.2. Isolated moleculesand donor molecules decoupled from the substrate by an insulating film to in-vestigate how energy level alignment is influenced by variations in molecular-scale interface geometry. With the small island size, as well as using molec-ular manipulation, we were able to probe specific geometries that cannot beisolated within closed monolayer systems that showed significant changes inenergy levels with subtle differences in geometry. Copper phthalocyanine(CuPc) and 3,4,9,10-perylenetetracarboxylic dianhydride (PTCDA) are twoprototypical organic semiconductors that are well studied as a prototyp-ical acceptor–donor pair, however the relation between interfacial energylevel alignment and precisely controlled interfacial geometry has not beendetermined. To effectively design new devices, it is crucial to understandthis relationship between molecular structure and electronic properties tomaximize the efficiency of charge transfer and separation.5.2 Isolated molecules5.2.1 CuPc on NaCl(2ML)/Ag(111)Metal phthalocyanines have been extensively studied for a broad range of ap-plications due to their semiconducting, optical and magnetic properties.[188–190] CuPc on NaCl(2ML)/Ag(111) is imaged with STM as a variety of 16-,8-, and 4-fold symmetric structures, with the in-gap states appearing as across (Figure 5.1a – d). A high resolution NC-AFM image with a PTCDAfuncitonalized tip is shown in Figure 5.1e.Figure 5.1: STM topography of an isolated CuPc molecule on NaCl(2ML)/Ag(111) (3.5 nm x 3.5 nm; a. It = 5 pA, Vb = −3.1 V, b. It = 10 pA, Vb= 0.3 V, c. It = 40 pA, Vb = 0.75 V, and d. It = 10 pA, Vb = 2.25 V). e.NC-AFM imaging with a PTCDA functionalized tip (2.1 nm x 2.1 nm, Vb= 0V).Figure 5.2 shows the normalized differential conductance spectra andenergy maps of an isolated CuPc molecule on NaCl(2ML)/Ag(111). Nine965.2. Isolated moleculesresonances are observed within the experimental bias range. The peaksor shoulders of the point STS and distinct shapes in the maps allow for anidentification of states, which we label C–O1 to C–O4 for the occupied statesand C–U1 to C–U5 for the unoccupied states. The resonances are markedwith dashed lines in the point spectra of Figure 5.2c, and the correspondingspatial maps are shown in Figure 5.2b and d. Each resonance is associatedwith an electronic state with the probability density localized in either theinner or outer part of the molecule.975.2. Isolated moleculesFigure 5.2: STS mapping of an isolated CuPc molecule on NaCl(2ML)/Ag(111). a. STM topgraphy (3.5 nm x 3.5 nm, It = 2 pA, Vb = −3.1 V).The locations of the spectra show in c are indicated by light and dark bluedots for the center and edge of the molecule respectively. b. STS maps ofthe occupied states of CuPc (3.5 nm x 3.5 nm). c. (dI/dV)/(I/V) pointspectra of the inner (dark blue) and outer (light blue) region of CuPc (set-point: It = 2 pA, Vb = −3.1 V). d. STS maps of the occupied states ofCuPc (3.5 nm x 3.5 nm).Directly assigning resonances to molecular orbitals of metal phthalocya-nines on insulating surfaces of states from STS resonances has proven tobe nontrivial, requiring formalisms that treat many-body interactions.[191]In CuPc this is partially due to the singly occupied 4s orbital resultingin spin splitting.[192] Several theoretical studies of CuPc have been per-formed, however orbital sequence changes significantly with the type of the-ory used.[190, 192–195] Even the most rigorous calculations published do985.3. Bimolecular system: CuPc and PTCDAnot satisfactorily compare to our data. Because of this, we do not identifythe molecular orbitals specifically, but refer to the states in terms of em-pirically assigned occupied and unoccupied labels, depending on the bias atwhich they occur (positive vs. negative).5.2.2 PTCDA on NaCl(2ML)/Ag(111)15As previously discussed, PTCDA on NaCl(2ML)/Ag(111) is negatively chargeddue to the small work function of NaCl(2ML)/Ag(111) and comparativelylarge electron affinity of PTCDA.[133] This results in the splitting of elec-tron addition and removal transitions, the upper and lower Hubbard states,arising from the singly occupied LUMO.[65, 171]. At biases less than −0.4V or greater than 1 V, PTCDA is imaged with STM as a double lobedstructure on NaCl(2ML)/Ag(111) (Figure 5.3b), while the in-gap states areimaged as a single lobe between −0.4 V and 1 V (Figure 5.3c). PTCDAadsorbs centered on the Cl- top site on NaCl(2ML)/Ag(111).[151, 152]Scanning tunneling spectroscopy of isolated PTCDA (Figure 5.4 c, dashedred line) shows three strong resonances, which we label P–O1, P–U1 andP–U2: P–O1 at −0.70 V with a shoulder at −0.80 V, P–U1 at +0.65 V,and P–U2 with a shoulder at 2.12 V and a peak at 2.22 V. From the STSpoint spectra and the spatial maps of isolated PTCDA, we assign P–O1 asthe overlapping HOMO and LUMO−1→0 (the lower Hubbard state), P-U1as the LUMO−1→−2 (the upper Hubbard state), and P-U2 as the nearlydegenerate LUMO +1/+2.[133] The states seen at +0.18 V around the neg-atively charged PTCDA molecule in the isolated molecule as well as thecoordinated system are due to the scattering of the NaCl(2ML)/Ag(111)interface state.[185]5.3 Bimolecular system: CuPc and PTCDA5.3.1 GeometryOne of the most common structures typically found was the dimer seen inFigures 5.3b – e. STM (Figures 5.3b, c) show the gross intermolecular ar-rangement, but high resolution NC-AFM (Figure 5.3d) gives much clearerdetail without the need for modeling. We determined that the CuPc adsorbson or near the Na+ top-site, similar to CuPc on NaCl bilayers on Cu(100)5.3e.[196]. Two of the CuPc phenyl hydrogens are positioned adjacent to15See Chapters 3 and 4 for an extensive discussion of the scanning probe spectroscopyof PTCDA.995.3. Bimolecular system: CuPc and PTCDAone of the PTCDA anhydride oxygens at a slightly rotated angle, with thePTCDA closer to one of the phenyl rings. NC-AFM can give an indicationof inter-molecular distance, though precise distances cannot be determineddue to distortions caused by the interaction of the PTCDA functionalizedtip with the molecule as well as the slight bending of the molecule due tosubstrate interactions.[113] Nevertheless, the distances between the CuPchydrogens and the PTCDA oxygens suggest that weak hydrogen bonds be-tween the two are likely. The PTCDA molecule can be switched betweentwo equivalent NaCl lattice sites with respect to the dimer by applying avoltage pulse with the tip over the molecule.Figure 5.3: a. STM image of showing typical cluster sizes and geometry ofPTCDA and CuPc on NaCl(2ML)/Ag(111) (40 nm x 40 nm, It = 5 pA, Vb= 0.5 V). b,c STM image of PTCDA and CuPc two-molecule cluster (5 nmx 3.5 nm, It = 6 pA, b Vb = −2 V and c Vb = 0.5 V). d. NC-AFM frequencyshift image of a PTCDA/ CuPc two-molecule cluster (3.5 nm x 3.5 nm, 0 V,constant z). e. Molecular positioning roughly determined from the NCAFMimage d and known adsorption geometry on the underlying NaCl lattice.5.3.2 Electronic structureSTS spectra and maps of the PTCDA/CuPc dimer are shown in Figures5.4b, c and d and compared to the isolated molecules shown in Figures5.4a and e. The resonances in the dimer maps are visually similar to thecorresponding resonances in the isolated molecules. The similar orbital ap-pearance and symmetry confirms the correspondence between of the statesbetween the dimer and isolated molecules. Point spectra of the dimer areshown in Figure 5.4c with the isolated molecule STS shown with dashed1005.3. Bimolecular system: CuPc and PTCDAlines. The direction of the shifts of the resonances are indicated with ar-rows. The fact that the peaks in the dimer do not appear to be significantlybroadened and the orbitals match visually, indicates that there is not signif-icant hybridization between the molecular orbitals with the presence of anadditional molecule.1015.3. Bimolecular system: CuPc and PTCDAFigure 5.4: a. STS maps of an isolated PTCDA on NaCl(2ML)/Ag(111) (5nm x 5 nm, set-point parameters: It = 2.5 pA, Vb = −2 V) b. STS mapsof a PTCDA/CuPc dimer corresponding to PTCDA, red dashed lines in c,(5 nm x 4 nm, set-point: It = 2 pA, Vb = −2.5 V. c. (dI/dV)/(I/V) pointspectra of both isolated CuPc and PTCDA molecules (dashed lines) as wellas PTCDA and CuPc in the dimer (solid lines). d. STS maps of dimercorresponding to CuPc states, indicated with blue dashed lines in c, (5 nmx 4 nm). e. STS maps of an isolated CuPc on NaCl(2ML)/Ag(111) (3.5 nmx 3.5 nm, set-point: It = 2 pA, Vb = −3.1 V). The images are rotated tomatch the orientation of the CuPc molecule in d. 1025.3. Bimolecular system: CuPc and PTCDAThe change in energy level alignment between isolated molecules andthose in the dimer system are compared in more detail Figure 5.5. The STSspectra of a PTCDA molecule adjacent to a CuPc show an overall downwardshift in energy as compared with an isolated PTCDA. Likewise, there isan overall upward shift of the CuPc spectra in the dimer relative to anisolated molecule. Both molecules also show a shift of the states towards Ef .The negative (downward) shift of PTCDA is indicative of a loss of negativecharge. Both the ionization potential (IP) and the electron affinity (EA)have increased. The positive (upward) shift of CuPc indicates the addition ofcharge, with a smaller IP and EA. Though PTCDA is typically thought of asan electron accepting material, the negative charge on the PTCDA moleculeresults in electron donating properties in the CuPc/PTCDA system.Figure 5.5: Electronic states of isolated PTCDA (red), isolated CuPc (cyan),dimer PTCDA (maroon), and dimer CuPc (blue) determined from STS (alsoshown). Overall energy resonance shifts are indicated with dashed lines.Arrows indicate direction of level shift due to charge transfer (ECT) andpolarization (EP). NaCl (2ML)/ Ag(111) surface state is indicated in gray.The inward shift towards EF of the energy levels upon the presenceof an adjacent molecule is related to molecular polarizability. As the ob-served shifts occur in the dimer relative to the isolated molecule on the1035.4. Varying molecular geometrysame substrate, the changes must be related to interactions between thetwo molecules. In Chapter 3, we have shown that intermolecular polar-ization effects can shift energy levels 100’s of mV. [133] Charges added orremoved from each molecule can be stabilized by an induced dipole in theother. Molecular hybridization is an additional factor that could result ina narrowing of the gap, however the lack of peak broadening and change inorbital appearance suggests this is not the case.The polarization energy shift (Ep) due to an addition of an electron wasdetermined for both CuPc and PTCDA by microelectrostatic calculationsusing the same method as described in Chapter 3.[133, 173] The in-planepolarizability tensors were taken from literature values calculated from DFT,with PTCDA short axis α = 50.3 A˚3 and long axis α = 88.2 r A3[173] andCuPc α = 135.8 A˚3 (considered isotropic)[197]. Ep for PTCDA due to anadjacent CuPc molecule was calculated to be 30 meV and the Ep of CuPcdue to PTCDA was 50 meV.The energy shift due to polarization was determined from the measuredgap sizes by taking Ep ≈ (∆isolated −∆dimer)/2, which assumes Ep was thesame for the different states probed. PTCDA was found to match quitewell, with a calculated value of 35 ± 17 meV.16 CuPc was slightly higherthan predicted at ± 17 meV. This is consistent with CuPc showing a slightlylarger inward shift than PTCDA. PTCDA is more polarizable than CuPc,and the polarization is occurring in the molecule(s) other than the one beingmeasured. Also, discrepancies might be due to additional effects of chargetransfer; the asymmetric, broad nature of the peaks requiring fits with toomany degrees of freedom; as well as the fact that Ep for an addition of chargeis not necessarily the same as for removal.[174]5.4 Varying molecular geometrySix different geometries of PTCDA/CuPc interfaces (Figures 5.6a – k) wereinvestigated by STS to determine local changes in energy level alignment.Each molecule was in a different local environment with varying acceptor–donor stoichiometry. For both PTCDA and CuPc, the molecules withinthe clusters were ordered by the number of surrounding acceptor or donormolecules. As a PTCDA molecule was surrounded by more, or more closelyarranged, CuPc molecules, the energy levels shifted down, indicating morenegative charge had been transferred to the PTCDA (Figure 5.6n). ForCuPc, an increasing number of adjacent PTCDA molecules resulted in an16The error arrises from the 12 mV resolution of the STS data.1045.4. Varying molecular geometryincreasing upward shift in energy, indicating a gain in negative charge (Fig-ure 5.6m). In some cases, it was not possible to accurately assign an energyvalue to the PTCDA P–U1 state due to significant reduction in intensityand possible broadening. In addition to stoichiometry, molecular geometrysignificantly affects energy level alignment.1055.4. Varying molecular geometryFigure 5.6: Comparison of energy level alignment of PTCDA/CuPc clustersof varying geometries. a. – f. Structural models of the 6 islands examined.e. – k. STM topographic images of each island It = 2 pA (xo: 4 nm x 5nm, Vb = −2V; xo–x: 6 nm x 6 nm, Vb = −2 .1V; ox–o: 5.5 nm x 5.5 nm, Vb= +2.2; oo–x: 5 x 5 nm, Vb = −2.1 V; oxo: 6.5 nm x 3.5 nm, Vb = −2V; xox:5.5 nm x 4 nm, Vb = −2 V. with x = CuPc and o = PTCDA). l. Alignment ofCuPc states arranged in increasing order of surrounding PTCDA moleculesm. Alignment of PTCDA states arranged in increasing order of surroundingCuPc molecules. Stoichiometry of the island is indicated with respect to themolecule examined.1065.4. Varying molecular geometryFigure 5.7 shows line cuts of the STS taken across the molecular is-land (x-axis) and plotted with respect to bias (y-axis) with the color-scalerepresenting (dI/dV)/(I/V) intensity. This enables a clear visualization ofthe energy level alignment of the heterojunctions. Islands with equivalentnumbers of each molecule but with differing positions also show substantialdifferences. For example, in the two PTCDA:CuPc 1:2 clusters of differentstructures (Figure 5.7b), the PTCDA O1 state is significantly shifted downin the linear island (xox)17 compared with the adjacent island (ox–o). Thisresults in gap energies that are ∼200 meV different for the two differentorientations. It also appears that occupied states in the linear cluster areoverlapping, whereas there are two distinct states observed in the PTCDAof the ox–o cluster. In the PTCDA:CuPc 2:1 clusters (Figure 5.7c) the P–U1state is significantly shifted and broadened in some geometries. This stoi-chiometry also results in the most significant changes in CuPc gap energieswith gap energies changing by ∼150 meV. This indicates that the orienta-tion of the molecules with respect to each other is crucially important fordetermining the electronic structure that influences charge transfer proper-ties. Discrepancies, such as the shifting of states relative to one another,could be due to the distortion of the molecules as well as splitting of nearlydegenerate states.17In this labeling, x represents CuPc molecules and o represents PTCDA molecules,chosen due to the in-gap STM images.1075.5. ConclusionsFigure 5.7: STS along a line cut (indicated by yellow arrows) of bimolecularheterojunctions of islands consisting of varying geometries of a. one CuPcand one PTCDA , b. two CuPc and one PTCDA c. one CuPc and twoPTCDA. x-axis is distance along line cut, y-axis is bias, and colormap is(dI/dV)/(I/V) intensity. White dotted line indicates node of line path whereapplicable.5.5 ConclusionsHere we show that significant changes in electronic structure (and thus en-ergy level alignment) can occur at acceptor–donor interfaces due to localgeometry, especially where there is excess charge present. Varying hetero-molecular islands of PTCDA and CuPc show significant effects on the elec-tronic structure due to both geometry and stoichiometry. As small changesin interfacial molecular geometry have shown to significantly alter deviceefficiency, it is crucial to understand how these subtle changes alter the1085.5. Conclusionsenergetics of these interfaces.Recently, Henneke et. al, have demonstrated that geometric controlover the interface is possible for the PTCDA/CuPc system. Different ratiosof the initial components result in different geometries in self-assembledmonolayers.[198] By changing deposition parameters, a specific structure canbe chosen. Combining self-assembly strategies with the knowledge gainedfrom probing energy level alignment has the potential to create new devicestructures with significant improvements in performance.109Chapter 6Conclusions and OutlookThe interface is the device.-Herbert KroemerThe key active processes that occur in organic devices are influenced bythe respective energy levels of materials at interfaces on the length scalesof single molecules. The measurements presented here show that both sin-gle component and bi-component interfaces are influenced by the dielectricenvironment and charge transfer, and both are sensitive to the local molec-ular scale structure. Investigating PTCDA nanoislands and PTCDA/CuPcnanostrucutures with advanced scanning probe spectroscopic mapping hasdemonstrated that the importance of molecular geometry on the electrostat-ics and energy level alignment in organic semiconducting systems cannot beneglected.Nanoislands of PTCDA on NaCl(2ML)/Ag(111) can be utilized as amodel of a 2D “bulk” interface between a molecular cluster and a vacuum.Examining the difference between edge molecules and center molecules showsshifts in gap energy and level alignment up to 400 meV on the length-scaleof one molecular width. This can be attributed to the change in electrostaticenvironment at the boundaries of clusters, namely via polarization of neigh-boring molecules. Energy levels that significantly change with molecularsurroundings demonstrate the site-specific stabilization of an added or re-moved charge in a anisotropic system. Comparison with microelectrostaticcalculations of the polarization energy, Ep, confirm the significant contribu-tion of molecular polarizability on an organic semiconducting system. Polar-ization contributions must be considered as most conjugated molecules havean anisotropic polarizability, both position and orientation matter. Thiswork shows that polarization effects can have significant influence on singleparticle energy levels. As charge separation involves an initial two-particlestate (the neutral exciton) splitting into two single-particle states (the elec-tron and hole) at an interface, these effects will influence exciton splittingas well as charge transport away from the interface.As PTCDA acquires a negative charge on NaCl(2ML)/Ag(111), nanois-1106.1. Open questions and future worklands can be utilized as a representative molecular system with half filledorbitals. Pixel-by-pixel EFS mapping and point spectra identified the signif-icant dependence of local charge distribution on molecular geometry, whichin turn influences the charging energy. Both on-site and nearest-neighboreffects of Coulomb repulsion were identified, as identified in the spatial vari-ance of the Hubbard energy. Differing initial charge distributions betweenmolecules of an island demonstrate the effects of nearest-neighbor moleculeson charge stabilization by determining the injection site and charging energy.A homogeneous island shows uniform Coulomb repulsion due to equivalentsurface charge distribution. As well, the local charge distribution after theinjection of an electron has been seen with sub-molecular resolution in anisolated PTCDA molecule.Small bimolecular clusters of PTCDA and CuPc molecules show the de-pendence of energy level alignment on molecular geometry in a prototypicalheterojunction. Variations in local molecular geometry and stoichiometrysignificantly impact on the energy alignment at the interface between accep-tor and donor. As this alignment is crucial for charge separation in an OPVdevice, further attention towards the detailed structure and electronic prop-erties directly at the interface is needed. Indeed, small changes in moleculargeometry have been shown to significantly alter device function.[47]The observation of these strong shifts in energy level alignment andcharging energy illustrate a crucial issue: interfacial energy level alignmentcan differ substantially from the bulk electronic structure in organic mate-rials. In order to optimize novel materials, the electronic properties of theinterface must also be considered. With most organic materials, it is impos-sible to completely avoid the complex process of charge separation by way ofa neutral charge species, however understanding and controlling the molecu-lar interfacial environment can yield device performance improvements andhelp understand the possible limits in creating novel materials.6.1 Open questions and future work6.1.1 Understanding the molecular states and excitationsobservedWe have ongoing collaborations with several groups implementing densityfunctional theory (DFT) for both the PTCDA 4-molecule islands as willas the PTCDA/ CuPc heteromolecular dimer. These molecular systemsare extremely complicated to model due to the large number of electronswithin the molecule (particularly CuPc), as well as the impact of the two1116.1. Open questions and future workcomponent surface (insulating layer on top of metallic substrate). It has beenshown that anisotropic charge distribution can occur within small moleculeislands on thin-insulating films on metals.[175] The questions that could beanswered by theory include:• conclusive identifications of the orbitals/states observed in STS• determination of the initial charge distribution within molecular is-lands as AFM only shows contrast, not quantitative charge assigned.• what will be seen in STS and EFS with photo-excitation• detailed determination of the charge states of individual molecules• additional effects due to the Coulomb repulsion6.1.2 Effects of the metal substrate, and stabilizingdifferent charge statesExperiments in this thesis have been performed predominantly on thin in-sulating NaCl bilayers, allowing for the simultaneous measurement by AFMand STM. These layers are used to decouple the molecule from the underly-ing substrate; however, there must be some degree of orbital overlap for themolecule to be in close enough proximity to the substrate to allow for tun-neling. An example of this is the case of PTCDA, which is influenced by thesubstrate to the extent that it alters the charge state of the molecule. Whilethis allows for investigation of transiently charged species, which do occurin functioning devices, the ability to stabilize charge allows for the investi-gation of energy alignment of different charge states. NC-AFM experimentson 10-30 ML NaCl have shown to stabilize charge states,[117] however thisdoes not allow for tunneling so simultaneous STS is not possible.We have found that on NaCl, trilayers are thick enough to allow forCuPc molecules to reversibly change persistent charge state in a hetero-molecular dimer, as indicated by differences in STM images (Figure 6.1).These changes are induced by a voltage pulse by the tip, either over theCuPc or the PTCDA.1126.1. Open questions and future workFigure 6.1: STM topography and corresponding frequency shift images ofreversible charging of a PTCDA/CuPc dimer on NaCl(3ML)/ Ag(111) (5nm x 3 nm, It = 2 pA, Vb = 0.5 VWe have also performed EFS spectroscopy on the heteromolecular dimerof PTCDA and CuPc, both on bilayer and trilayer NaCl. Although prelim-inary results show interesting charging events, higher resolution scanning(both spatial and energetic) is needed in order to extract meaningful infor-mation.6.1.3 Possibility to directly observe charge separationAll of the experiments performed here have relied on electron tunnelingor surface charge transfer to create charged species. To further investi-gate the optoelectronic properties of these OPV materials, we will examinethese systems under illumination, to observe the effect of excitation by aphoton (full experimental details can be found in Tanya Roussy’s MAScthesis[199].) This will more closely simulate a real device and allow inves-tigation of photo-excitation and photo-induced charge generation. Usingboth STS and EFS, we can observe how energy levels shift and local chargedistributions change on the submolecular level when an acceptor–donor pairis illuminated. We have performed this experiment with PTCDA/CuPc onNaCl(2ML)/Ag(111) but did not see any significant change upon illumina-tion. We believe this is due to the fact that PTCDA is negatively chargedand the level alignment is not well suited for photo-induced charge transfer.Preliminary results for PTCDA on NaCl(2ML)/Au(100) determined thatPTCDA molecules are neutral. The work function of gold is significantlylarger than that of copper or silver and no repulsion of surface electrons1136.1. Open questions and future workaround the molecule was observed. As well, there is not a close alignmentof the first two unoccupied molecular states as there is on Ag. Point spectraand STS maps are shown in Figure 6.2. From the point spectra of CuPc,both the occupied and unoccupied states shift towards the Fermi energy,indicating an effect due to the polarization energy, however overall shiftsupward or downward are not seen. This indicates there is no significantchange in molecular charge. No occupied states of PTCDA are seen, so acomparison of the occupied and unoccupied shifts could not be made.Another reason for the lack of any observable shift could be the quenchingof the excitation due to the nearby metal as often non-radiative energytransfer to the metal is faster than molecular fluorescence.[200, 201] Thiscould be prevented by using bulk or “bulk-like” thick films.[202, 203]1146.1. Open questions and future workFigure 6.2: PTCDA and CuPc on NaCl(2ML)/Au(100). a. STM topog-raphy (6 nm x 4 nm, It = 5 pA, Vb = 2 V). b. STS maps of PTCDAunoccupied states (6 nm x 4 nm). c. (dI/dV)/(I/V) point spectra of bothisolated CuPc and PTCDA molecules (dashed lines) as well as PTCDA andCuPc in the dimer (solid lines) (set-point: It = 5 pA, Vb = −2.5 V) d. STSmaps of CuPc states (6 nm x 4 nm).6.1.4 Novel materialsThe molecules examined, PTCDA and CuPc, were chosen to test ideasand measurement techniques for OPVs. These molecules are examples ofsmall units of the more complex molecules and polymers typically used inOPV devices. Investigating short oligomers of polymer materials, or di-rectly studying molecules used in vacuum deposited devices is importantto understand energy level alignment in materials optimized for true devicefunction. Collaborating with synthetic chemists will allow the investigation1156.1. Open questions and future workof new molecules and aid in design. The knowledge of how structure affectscharge transfer is key in order to optimize device efficiency.6.1.5 OutlookAs energy resources continue to diminish and pressures on our global en-vironment increase, optimizing the performance of new clean energy tech-nologies becomes even more essential. Organic photovoltaic (OPV) devicesare an emerging technology that has recently hit the market,[4–6, 10–14]however current device efficiencies are not amenable to mass deployment.Maximizing efficiency relies on a full understanding of how these devicesturn light into power.OPV materials rely on charge transfer at an interface between an electrondonating and an electron accepting material. There is a strong correlationbetween interfacial energy level alignment and local molecular geometry,which must be considered when creating new materials. Here, several fac-tors relating interfacial structure to the resulting electronic states and levelalignment at interfaces have been explored. If we can develop a completepicture of the properties that create ideal environments for charge transfer,we can use this information to synthesize new molecules that enable efficientcharge transfer and extraction, advancing future devices.Small molecules, like those studied here, offer significant promise fornew devices due to their high charge-carrier mobility. Creating new devicesfrom small molecules prepared by vacuum deposition utilizes technologyalready widely available.18 The combination of scanning probe techniqueswith submolecular spatial resolution combined with optical experiments hasthe potential to continue to advance our understanding and development ofnovel organic photovoltaic materials.18Similar vacuum deposition methods are used in the manufacturing of potato chipbags.116117Bibliography[1] Naomi Oreskes. The scientific consensus on climate change. Science,306(5702):1686–1686, December 2004.[2] Mikael Hok and Xu Tang. Depletion of fossil fuels and anthropogenicclimate change. A review. Energy Pol., 52:797–809, January 2013.[3] Susan Solomon, Gian-Kasper Plattner, Reto Knutti, and PierreFriedlingstein. Irreversible climate change due to carbon dioxide emis-sions. Proc. Natl. Acad. Sci. U.S.A., 106(6):1704–1709, February 2009.[4] Adolf Goetzberger, Christopher Hebling, and Hans-Werner Schock.Photovoltaic materials, history, status and outlook. Mat. Sci. andEngin: R, 40(1):1–46, January 2003.[5] K Kalyanasundaram and M Gra¨tzel. Themed issue: nanomaterials forenergy conversion and storage. J. Mat. Chem, 22(46):24190–24194,2012.[6] Nathan S Lewis. Toward cost-effective solar energy use. Science,315(5813):798–801, February 2007.[7] S E Gledhill, B Scott, and B A Gregg. Organic and nano-structuredcomposite photovoltaics: An overview. J. Mater. Res., 2005.[8] J R Sheats. Manufacturing and commercialization issues in organicelectronics. J. Mater. Res., 2004.[9] Caitlin McDowell and Guillermo C Bazan. Organic solar cells pro-cessed from green solvents. Curr. Opin. Green Sustainable Chem.,5:49–54, June 2017.[10] Noon Solar,[11] Konarka Announces Largest OPV Installa-tion, First Semi-Transparent BIPV Curtain Wall,118Bibliography[12] Yimu Zhao, Garrett A Meek, Benjamin G Levine, and Richard R Lunt.Near-infrared harvesting transparent luminescent solar concentrators.Adv. Opt. Mater., 2(7):606–611, July 2014.[13] The only device charging market umbrella,[14] Heliatek reaches efficiency record with 40% transparent organic solarcells[15] Miaomiao Li, Ke Gao, Xiangjian Wan, Qian Zhang, Bin Kan, RuoxiXia, Feng Liu, Xuan Yang, Huanran Feng, Wang Ni, YunchuangWang, Jiajun Peng, Hongtao Zhang, Ziqi Liang, Hin-Lap Yip, Xi-aobin Peng, Yong Cao, and Yongsheng Chen. Solution-processed or-ganic tandem solar cells with power conversion efficiencies greater than12 percent. Nat. Photon., 11(2):85–90, December 2016.[16] S R Forrest. The limits to organic photovoltaic cell efficiency. MRSBull., 2005.[17] Angele Reinders, Pierre Verlinden, and Alexandre Freundlich. Photo-voltaic solar energy: from fundamentals to applications. John Wiley& Sons, New York, NY, 2016.[18] Brian A Gregg and Mark C Hanna. Comparing organic to inorganicphotovoltaic cells: theory, experiment, and simulation. J. Appl. Phys.,93(6):3605–3614, March 2003.[19] Edgar A Silinsh. Organic molecular crystals: their electronic states,volume 16 of Springer Series in Solid-State Sciences. Springer, Berlin,1980.[20] R H Friend, R W Gymer, Holmes, AB, J H Burroughes, R N Marks,C Taliani, DDC Bradley, D A Dos Santos, J L Bredas, M Logdlund,and W R Salaneck. Electroluminescence in conjugated polymers. Na-ture, 397(6715):121–128, 1999.[21] C W Tang. Two-layer organic photovoltaic cell. Appl. Phys. Lett.,48(2):183–185, August 1998.119Bibliography[22] N Karl. Charge carrier transport in organic semiconductors. Synth.Met., 133-134:649–657, March 2003.[23] Christopher M Proctor, Martijn Kuik, and Thuc-Quyen Nguyen.Charge carrier recombination in organic solar cells. Prog. Polym. Sci.,38(12):1941–1960, December 2013.[24] Tracey M Clarke and James R Durrant. Charge photogeneration inorganic solar cells. Chem. Rev., 110(11):6736–6767, November 2010.[25] Jean-Luc Bre´das, Joseph E Norton, Je´roˆme Cornil, and VeaceslavCoropceanu. Molecular understanding of organic solar cells: The chal-lenges. Acc. Chem. Res., 42(11):1691–1699, November 2009.[26] S Matthew Menke, Wade A Luhman, and Russell J Holmes. Tailoredexciton diffusion in organic photovoltaic cells for enhanced power con-version efficiency. Nat. Mater., 12(2):152–157, February 2013.[27] Yasunari Tamai, Hideo Ohkita, Hiroaki Benten, and Shinzaburo Ito.Exciton diffusion in conjugated polymers: from fundamental under-standing to improvement in photovoltaic conversion efficiency. J. Phys.Chem. Lett., 6(17):3417–3428, September 2015.[28] Richard R Lunt, Jay B Benziger, and Stephen R Forrest. Relationshipbetween crystalline order and exciton diffusion length in molecularorganic semiconductors. Adv. Mater., 22(11):1233–1236, March 2010.[29] Dani M Stoltzfus, Jenny E Donaghey, Ardalan Armin, Paul E Shaw,Paul L Burn, and Paul Meredith. Charge generation pathways in or-ganic solar cells: assessing the contribution from the electron acceptor.Chem. Rev., 116(21):12920–12955, June 2016.[30] R Heitz, H Born, F Guffarth, O Stier, A Schliwa, A Hoffmann, andD Bimberg. Existence of a phonon bottleneck for excitons in quantumdots. Phys. Rev. B, 64(24):241305, November 2001.[31] M Knupfer. Exciton binding energies in organic semiconductors. Appl.Phys. A, 77(5):623–626, October 2003.[32] G Grancini, M Maiuri, D Fazzi, A Petrozza, H-J Egelhaaf, D Brida,G Cerullo, and G Lanzani. Hot exciton dissociation in polymer solarcells. Nat. Mater., 12(1):29–33, December 2012.120Bibliography[33] Roderick C I MacKenzie, Thomas Kirchartz, George F A Dibb, andJenny Nelson. Modelling nongeminate recombination in P3HT:PCBMsolar cells. J. Phys. Chem. C, 115(19):9806–9813, May 2011.[34] Franz Fuchs, Florent Caffy, Renaud Demadrille, Thierry Me´lin, andBenjamin Gre´vin. High-resolution Kelvin probe force microscopyimaging of interface dipoles and photogenerated charges in organicdonor–acceptor photovoltaic blends. ACS Nano, 10(1):739–746, Jan-uary 2016.[35] Hisao Ishii, Kiyoshi Sugiyama, Eisuke Ito, and Kazuhiko Seki. Energylevel alignment and interfacial electronic structures at organic/metaland organic/organic interfaces. Adv. Mater., 11(8):605–625, June1999.[36] Benjamin Stadtmu¨ller, Daniel Lu¨ftner, Martin Willenbockel, Eva MReinisch, Tomoki Sueyoshi, Georg Koller, Serguei Soubatch,Michael G Ramsey, Peter Puschnig, F Stefan Tautz, and ChristianKumpf. Unexpected interplay of bonding height and energy level align-ment at heteromolecular hybrid interfaces. Nat. Commun., 5:3685,2014.[37] G Yu, J Gao, J C Hummelen, F Wudl, and A J HEEGER. Poly-mer photovoltaic cells - enhanced efficiencies via a network of internaldonor-acceptor heterojunctions. Science, 270(5243):1789–1791, 1995.[38] Kla´ra Maturova´, Svetlana S van Bavel, Martijn M Wienk, Rene´ A JJanssen, and Martijn Kemerink. Description of the morphology de-pendent charge transport and performance of polymer:fullerene bulkheterojunction solar cells. Adv. Funct. Mater., 21(2):261–269, January2011.[39] H Sirringhaus, P J Brown, R H Friend, and M M Nielsen. Two-dimensional charge transport in self-organized, high-mobility conju-gated polymers. Nature, 401(6754):685–688, 1999.[40] Michael Graetzel, Rene´ A J Janssen, David B Mitzi, and Edward HSargent. Materials interface engineering for solution-processed photo-voltaics. Nature, 488(7411):304–312, August 2012.[41] Mijanur Rahaman Molla and Suhrit Ghosh. Structural variationson self-assembly and macroscopic properties of 1,4,5,8-naphthalene-diimide chromophores. Chem. Mater., 23(1):95–105, December 2010.121Bibliography[42] Kiarash Vakhshouri, Sameer Vajjala Kesava, Derek R Kozub, andEnrique D Gomez. Characterization of the mesoscopic structure inthe photoactive layer of organic solar cells: A focused review. Mater.Lett., 90:97–102, January 2013.[43] Shane R Yost, Lee-Ping Wang, and Troy Van Voorhis. Molecularinsight into the energy levels at the organic donor/acceptor interface:A quantum mechanics/molecular mechanics study. J. Phys. Chem. C,115(29):14431–14436, July 2011.[44] Chris Groves, Obadiah G Reid, and David S Ginger. Heterogeneityin polymer solar cells: local morphology and performance in organicphotovoltaics studied with scanning probe microscopy. Acc. Chem.Res., 43(5):612–620, February 2010.[45] Gang Li, Rui Zhu, and Yang Yang. Polymer solar cells. Nat. Photonics,6(3):153–161, February 2012.[46] M Willenbockel, B Stadtmu¨ller, K Scho¨nauer, F C Bocquet,D Lu¨ftner, E M Reinisch, T Ules, G Koller, C Kumpf, S Soubatch,P Puschnig, M G Ramsey, and F S Tautz. Energy offsets withina molecular monolayer: the influence of the molecular environment.New J. Phys., 15(3):033017, March 2013.[47] Kenneth R Graham, Clement Cabanetos, Justin P Jahnke, Matthew NIdso, Abdulrahman El Labban, Guy O Ngongang Ndjawa, ThomasHeumueller, Koen Vandewal, Alberto Salleo, Bradley F Chmelka,Aram Amassian, Pierre M Beaujuge, and Michael D McGehee. Im-portance of the donor:fullerene intermolecular arrangement for high-efficiency organic photovoltaics. J. Am. Chem. Soc., 136(27):9608–9618, July 2014.[48] A C Arias, J D MacKenzie, R Stevenson, J J M Halls, M Inbasekaran,E P Woo, D Richards, and R H Friend. Photovoltaic performanceand morphology of polyfluorene blends: A combined microscopic andphotovoltaic investigation. Macromolecules, 34(17):6005–6013, August2001.[49] Matthew T Lloyd, John E Anthony, and George G Malliaras. Pho-tovoltaics from soluble small molecules. Mater. Today, 10(11):34–41,November 2007.122Bibliography[50] Yongsheng Chen, Xiangjian Wan, and Guankui Long. High per-formance photovoltaic applications using solution-processed smallmolecules. Acc. Chem. Res., 46(11):2645–2655, July 2013.[51] G Yu, J Gao, J C Hummelen, F Wudl, and A J Heeger. Polymer pho-tovoltaic cells: enhanced efficiencies via a network of internal donor–acceptor heteroiunctions. Science, 1995.[52] Amaresh Mishra and Peter Ba¨uerle. Small molecule organic semi-conductors on the move: Promises for future solar energy technology.Angew. Chem. Int. Ed., 51(9):2020–2067, February 2012.[53] Frederik C Krebs. Degradation and stability of polymer and organicsolar cells. Sol. Energy Mater. Sol. Cells, 92(7):685–685, July 2008.[54] Mikkel Jørgensen, Kion Norrman, and Frederik C Krebs. Stabil-ity/degradation of polymer solar cells. Sol. Energy Mater. Sol. Cells,92(7):686–714, July 2008.[55] Matthew O Reese, Suren A Gevorgyan, Mikkel Jørgensen, EvaBundgaard, Sarah R Kurtz, David S Ginley, Dana C Olson,Matthew T Lloyd, Pasquale Morvillo, Eugene A Katz, AndreasElschner, Olivier Haillant, Travis R Currier, Vishal Shrotriya, Mar-tin Hermenau, Moritz Riede, Kiril R Kirov, Gregor Trimmel, ThomasRath, Olle Ingana¨s, Fengling Zhang, Mattias Andersson, Kristofer Tv-ingstedt, Monica Lira-Cantu, Darin Laird, Christine McGuiness, Srini-vas Jimmy Gowrisanker, Michael Pannone, Min Xiao, Jens Hauch,Roland Steim, Dean M DeLongchamp, Roland Ro¨sch, Harald Hoppe,Nieves Espinosa, Antonio Urbina, Gu¨lsah Yaman-Uzunoglu, Jo¨rg-Bernd Bonekamp, Albert J J M van Breemen, Claudio Girotto, EszterVoroshazi, and Frederik C Krebs. Consensus stability testing proto-cols for organic photovoltaic materials and devices. Sol. Energy Mater.Sol. Cells, 95(5):1253–1267, May 2011.[56] R Franke, B Maennig, A Petrich, and M Pfeiffer. Long-term stabilityof tandem solar cells containing small organic molecules. Sol. EnergyMater. Sol. Cells, 92(7):732–735, July 2008.[57] Z Ding, J Kettle, M Horie, S W Chang, G C Smith, A I Shames, andE A Katz. Efficient solar cells are more stable: the impact of polymermolecular weight on performance of organic photovoltaics. J. Mat.Chem. A, 4(19):7274–7280, 2016.123Bibliography[58] Bin Kan, Miaomiao Li, Qian Zhang, Feng Liu, Xiangjian Wan,Yunchuang Wang, Wang Ni, Guankui Long, Xuan Yang, HuanranFeng, Yi Zuo, Mingtao Zhang, Fei Huang, Yong Cao, Thomas PRussell, and Yongsheng Chen. A series of simple oligomer-like smallmolecules based on oligothiophenes for solution-processed solar cellswith high efficiency. J. Am. Chem. Soc., 137(11):3886–3893, March2015.[59] Yi-Hong Chen, Li-Yen Lin, Chih-Wei Lu, Francis Lin, Zheng-YuHuang, Hao-Wu Lin, Po-Han Wang, Yi-Hung Liu, Ken-Tsung Wong,Jianguo Wen, Dean J Miller, and Seth B Darling. Vacuum-depositedsmall-molecule organic solar cells with high power conversion efficien-cies by judicious molecular design and device optimization. J. Am.Chem. Soc., 134(33):13616–13623, August 2012.[60] Letian Dou, Jingbi You, Ziruo Hong, Zheng Xu, Gang Li, Robert AStreet, and Yang Yang. 25th anniversary article: A decade of or-ganic/polymeric photovoltaic research. Adv. Mater., 25(46):6642–6671, September 2013.[61] M C Scharber, D Mu¨hlbacher, M Koppe, P Denk, C Waldauf,A J HEEGER, and C J Brabec. Design rules for donors in bulk-heterojunction solar cells— towards 10% energy-conversion efficiency.Adv. Mater., 18(6):789–794, March 2006.[62] C J Brabec, A Cravino, and D Meissner. Origin of the open circuitvoltage of plastic solar cells. Adv. Funct. Mater., 2001.[63] S C Veenstra, A Heeres, G Hadziioannou, G A Sawatzky, and H TJonkman. On interface dipole layers between C 60 and Ag or Au.Appl. Phys. A, 75(6):661–666, December 2002.[64] E V Tsiper, Z G Soos, W Gao, and A Kahn. Electronic polarization atsurfaces and thin films of organic molecular crystals: PTCDA. Chem.Phys. Lett., 360(1-2):47–52, 2002.[65] Isabel Ferna´ndez Torrente, Katharina J Franke, and Jose Ignacio Pas-cual. Spectroscopy of C60 single molecules: the role of screening onenergy level alignment. J. Phys.: Condens. Matter, 20(18):184001–11,April 2008.[66] Steffen Duhm, Georg Heimel, Ingo Salzmann, Hendrik Glowatzki,Robert L Johnson, Antje Vollmer, Ju¨rgen P Rabe, and Norbert Koch.124BibliographyOrientation-dependent ionization energies and interface dipoles in or-dered molecular assemblies. Nature Materials, 7(4):326–332, April2008.[67] M A O Keefe, C J D Hetherington, Y C Wang, E C Nelson, J HTurner, C Kisielowski, J O Malm, R Mueller, J Ringnalda, M Pan,and A Thust. Sub-A˚ngstrom high-resolution transmission electron mi-croscopy at 300keV. Ultramicroscopy, 89(4):215–241, November 2001.[68] G Binnig, H Rohrer, Ch Gerber, and E Weibel. Surface studies byscanning tunneling microscopy. Phys. Rev. Lett., 49(1):57–61, July1982.[69] Russell Young, John Ward, and Fredric Scire. The topografiner: Aninstrument for measuring surface microtopography. Rev. Sci. Instrum.,43(7):999–1011, July 1972.[70] G Binnig, CF Quate, and C Gerber. Atomic force microscope. Phys.Rev. Lett., 56(9):930–933, March 1986.[71] Dimas G de Oteyza, Juan M Garca˜ Lastra, Martina Corso, Bryan PDoyle, Luca Floreano, Alberto Morgante, Yutaka Wakayama, An-gel Rubio, and J Enrique Ortega. Customized electronic couplingin self-assembled donor–acceptor nanostructures. Adv. Funct. Mater.,19(22):3567–3573, November 2009.[72] Liam S C Pingree, Obadiah G Reid, and David S Ginger. Electricalscanning probe microscopy on active organic electronic devices. Adv.Mater., 21(1):19–28, January 2009.[73] Gu¨nther Benstetter, Roland Biberger, and Dongping Liu. A review ofadvanced scanning probe microscope analysis of functional films andsemiconductor devices. Thin Solid Films, 517(17):5100–5105, July2009.[74] J Bardeen. Tunnelling from a Many-Particle Point of View. Phys.Rev. Lett., 6(2):57–59, 1961.[75] J Tersoff and D R Hamann. Theory and Application for the ScanningTunneling Microscope. Phys. Rev. Lett., 50(25):1998–2001, June 1983.[76] J Tersoff and D R Hamann. Theory of the Scanning Tunneling Micro-scope. Phys. Rev. B, 31(2):805–813, 1985.125Bibliography[77] Werner A Hofer, Adam S Foster, and Alexander L Shluger. Theoriesof scanning probe microscopes at the atomic scale. Rev. Mod. Phys.,75(4):1287–1331, October 2003.[78] Harold J W Zandvliet and Arie van Houselt. Scanning TunnelingSpectroscopy. Annu. Rev. Anal. Chem., 2(1):37–55, July 2009.[79] J B Pendry, A B Preˆtre, and B C H Krutzen. Theory of the scanningtunnelling microscope. J. Phys.: Condens. Matter, 3(24):4313–4321,June 1991.[80] Masaru Tsukada, Katsuyoshi Kobayashi, Nobuyuki Isshiki, and Hi-royuki Kageshima. First-principles theory of scanning tunneling mi-croscopy. Surf. Sci. Rep., 13(8):267–304, January 1991.[81] David Griffiths. Introduction to Quantum Mechanics. Cambridge Uni-versity Press, August 2016.[82] J A Stroscio, R M Feenstra, and A P Fein. Electronic-Structure of theSi(111) 2x1 Surface by Scanning-Tunneling Microscopy. Phys. Rev.Lett., 57(20):2579–2582, 1986.[83] A Greuling, M Rohlfing, R Temirov, F S Tautz, and F B Anders. Abinitio study of a mechanically gated molecule: From weak to strongcorrelation. Phys. Rev. B, 84(12):125413, September 2011.[84] B Koslowski, C Dietrich, A Tschetschetkin, and P Ziemann. Evalua-tion of scanning tunneling spectroscopy data: Approaching a quanti-tative determination of the electronic density of states. Phys. Rev. B,2007.[85] C Wagner, R Franke, and T Fritz. Evaluation of I/V curves in scan-ning tunneling spectroscopy of organic nanolayers. Phys. Rev. B,75(23):235432, June 2007.[86] A J Macdonald, Y S Tremblay-Johnston, S Grothe, S Chi, P Dosanjh,S Johnston, and S A Burke. Dispersing artifacts in FT-STS: a com-parison of set point effects across acquisition modes. Nanotechnology,27(41):414004, October 2016.[87] Jiutao Li, Wolf-Dieter Schneider, and Richard Berndt. Local den-sity of states from spectroscopic scanning-tunneling-microscope im-ages: Ag(111). Phys. Rev. B, 56(12):7656–7659, September 1997.126Bibliography[88] Karina Morgenstern and Jouko Nieminen. Imaging water on Ag(111):Field induced reorientation and contrast inversion. J. Chem. Phys.,120(22):10786, 2004.[89] K A Cochrane, A Schiffrin, T S Roussy, M Capsoni, and S A Burke.Pronounced polarization-induced energy level shifts at boundaries oforganic semiconductor nanostructures. Nat. Commun., 6:8312–8, Oc-tober 2015.[90] Jascha Repp, Gerhard Meyer, Fredrik E Olsson, and Mats Persson.Controlling the Charge State of Individual Gold Adatoms. Science,305(5683):493–495, July 2004.[91] Lucian Jdira, Peter Liljeroth, Eric Stoffels, Danie¨l Vanmaekelbergh,and Sylvia Speller. Size-dependent single-particle energy levels andinterparticle Coulomb interactions in CdSe quantum dots measuredby scanning tunneling spectroscopy. Phys. Rev. B, 73(11):115305,March 2006.[92] S Morita, F J Giessibl, E Meyer, and R Wiesendanger. NoncontactAtomic Force Microscopy, 2015.[93] U Du¨rig, H R Steinauer, and N Blanc. Dynamic force microscopyby means of the phase-controlled oscillator method. J. Appl. Phys.,82(8):3641–3651, June 1998.[94] Franz J Giessibl. Advances in atomic force microscopy. Rev. Mod.Phys., 75(3):949–983, July 2003.[95] Seizo Morita. Atomically resolved force microscopy. J. Vac. Sci. Tech-nol. A, 31(5):050802, 2013.[96] Franz J Giessibl. Atomic force microscopy: Si-(7x7) surface by atomicforce microscopy. Science, 267(5194):68–71, January 1995.[97] D J Griffiths. Introduction to Electrodynamics Prentice-Hall. UpperSaddle River, 1999.[98] Franz J Giessibl. Atomic resolution on Si(111)-(7x7) by noncontactatomic force microscopy with a force sensor based on a quartz tuningfork. Appl. Phys. Lett., 76(11):1470–1472, March 2000.[99] Boris J Albers, Marcus Liebmann, Todd C Schwendemann, Mehmet ZBaykara, Markus Heyde, Miquel Salmeron, Eric I Altman, and Udo D127BibliographySchwarz. Combined low-temperature scanning tunneling/atomic forcemicroscope for atomic resolution imaging and site-specific force spec-troscopy. Rev. Sci. Instrum., 79(3):033704, March 2008.[100] L Gross, F Mohn, N Moll, P Liljeroth, and G Meyer. The chemicalstructure of a molecule resolved by atomic force microscopy. Science,325(5944):1110–1114, August 2009.[101] Leo Gross, Fabian Mohn, Nikolaj Moll, Gerhard Meyer, Rainer Ebel,Wael M Abdel-Mageed, and Marcel Jaspars. Organic structure de-termination using atomic-resolution scanning probe microscopy. Nat.Chem., 2(10):821–825, August 2010.[102] G Kichin, C Wagner, F S Tautz, and R Temirov. Calibrating atomic-scale force sensors installed at the tip apex of a scanning tunnelingmicroscope. Phys. Rev. B, 87(8):081408, February 2013.[103] Prokop Hapala, Georgy Kichin, Christian Wagner, F Stefan Tautz,Ruslan Temirov, and Pavel Jel´ınek. Mechanism of high-resolutionSTM/AFM imaging with functionalized tips. Phys. Rev. B,90(8):085421, August 2014.[104] Christian Wagner, Matthew F B Green, Philipp Leinen, ThorstenDeilmann, Peter Kru¨ger, Michael Rohlfing, Ruslan Temirov, andF Stefan Tautz. Scanning quantum dot microscopy. Phys. Rev. Lett.,115(2):026101, July 2015.[105] Leo Gross, Fabian Mohn, Nikolaj Moll, Bruno Schuler, AlejandroCriado, Enrique Guitia´n, Diego Pen˜a, Andre´ Gourdon, and GerhardMeyer. Bond-order discrimination by atomic force microscopy. Sci-ence, 337(6100):1326–1329, September 2012.[106] Dimas G de Oteyza, Patrick Gorman, Yen-Chia Chen, SebastianWickenburg, Alexander Riss, Duncan J Mowbray, Grisha Etkin, ZahraPedramrazi, Hsin-Zon Tsai, Angel Rubio, Michael F Crommie, andFelix R Fischer. Direct imaging of covalent bond structure in single-molecule chemical reactions. Science, 340(6139):1434–1437, June 2013.[107] Bruno Schuler, Wei Liu, Alexandre Tkatchenko, Nikolaj Moll, GerhardMeyer, Anish Mistry, David Fox, and Leo Gross. Adsorption geometrydetermination of single molecules by atomic force microscopy. Phys.Rev. Lett., 111(10):106103, September 2013.128Bibliography[108] Alexander Riss, Alejandro Pe´rez Paz, Sebastian Wickenburg, Hsin-Zon Tsai, Dimas G de Oteyza, Aaron J Bradley, Miguel M Ugeda,Patrick Gorman, Han Sae Jung, Michael F Crommie, Angel Rubio,and Felix R Fischer. Imaging single-molecule reaction intermediatesstabilized by surface dissipation and entropy. Nat. Chem., 8(7):678–683, July 2016.[109] Florian Albrecht, Niko Pavlicˇek, Coral Herranz-Lancho, Mario Ruben,and Jascha Repp. Characterization of a surface reaction by means ofatomic force microscopy. J. Am. Chem. Soc., 137(23):7424–7428, June2015.[110] Sampsa K Ha¨ma¨la¨inen, Nadine van der Heijden, Joost van der Lit,Stephan den Hartog, Peter Liljeroth, and Ingmar Swart. Intermolecu-lar contrast in atomic force microscopy images without intermolecularbonds. Phys. Rev. Lett., 113(18):186102, October 2014.[111] M Neu, N Moll, L Gross, G Meyer, F J Giessibl, and J Repp. Imagecorrection for atomic force microscopy images with functionalized tips.Phys. Rev. B, 89(20):205407, May 2014.[112] Michael Ellner, Niko Pavlicˇek, Pablo Pou, Bruno Schuler, NikolajMoll, Gerhard Meyer, Leo Gross, and Rube´n Pere´z. The electric fieldof CO tips and its relevance for mtomic force microscopy. Nano Lett.,16(3):1974–1980, March 2016.[113] Leo Gross, Bruno Schuler, Fabian Mohn, Nikolaj Moll, Niko Pavlicˇek,Wolfram Steurer, Ivan Scivetti, Konstantinos Kotsis, Mats Persson,and Gerhard Meyer. Investigating atomic contrast in atomic forcemicroscopy and Kelvin probe force microscopy on ionic systems usingfunctionalized tips. Phys. Rev. B, 90(15):155455–11, October 2014.[114] J Zhang, P Chen, B Yuan, W Ji, Z Cheng, and X Qiu. Real-spaceidentification of intermolecular bonding with atomic force microscopy.Science, 342(6158):611–614, October 2013.[115] Nikolaj Moll, Bruno Schuler, Shigeki Kawai, Feng Xu, Lifen Peng,Akihiro Orita, Junzo Otera, Alessandro Curioni, Mathias Neu, JaschaRepp, Gerhard Meyer, and Leo Gross. Image distortions of a par-tially fluorinated hydrocarbon molecule in atomic force microscopywith carbon monoxide terminated tips. Nano Lett., 14(11):6127–6131,November 2014.129Bibliography[116] Hatem Labidi, Mohammad Koleini, Taleana Huff, Mark Salomons,Martin Cloutier, Jason Pitters, and Robert A Wolkow. Indicationsof chemical bond contrast in AFM images of a hydrogen-terminatedsilicon surface. Nat. Commun., 8:14222, 2017.[117] Wolfram Steurer, Shadi Fatayer, Leo Gross, and Gerhard Meyer.Probe-based measurement of lateral single-electron transfer betweenindividual molecules. Nat. Commun., 6:8353, September 2015.[118] S A Burke, J M LeDue, Y Miyahara, J M Topple, S Fostner, andP Gru¨tter. Determination of the local contact potential differenceof PTCDA on NaCl: a comparison of techniques. Nanotechnology,20(26):264012, July 2009.[119] Wilhelm Melitz, Jian Shen, Sangyeob Lee, Joon Sung Lee, Andrew CKummel, Ravi Droopad, and Edward T Yu. Scanning tunneling spec-troscopy and Kelvin probe force microscopy investigation of Fermienergy level pinning mechanism on InAs and InGaAs clean surfaces.J. Appl. Phys., 108(2):023711, July 2010.[120] Leo Gross, Fabian Mohn, Peter Liljeroth, Jascha Repp, Franz JGiessibl, and Gerhard Meyer. Measuring the charge state of an adatomwith noncontact atomic force microscopy. Science, 324(5933):1428–1431, June 2009.[121] Fabian Mohn, Leo Gross, Nikolaj Moll, and Gerhard Meyer. Imag-ing the charge distribution within a single molecule. Nat. Nanotech.,7(4):227–231, February 2012.[122] F Albrecht, J Repp, M Fleischmann, M Scheer, M Ondra´cˇek, andP Jel´ınek. Probing charges on the atomic scale by means of atomicforce microscopy. Phys. Rev. Lett., 115(7):076101–5, August 2015.[123] Bruno Schuler, Shi-Xia Liu, Yan Geng, Silvio Decurtins, GerhardMeyer, and Leo Gross. Contrast formation in Kelvin probe force mi-croscopy of single pi-conjugated molecules. Nano Lett., 14(6):3342–3346, June 2014.[124] L Gross, F Mohn, P Liljeroth, J Repp, F J Giessibl, and G Meyer.Measuring the charge state of an adatom with noncontact atomic forcemicroscopy. Science, 324(5933):1428–1431, June 2009.130Bibliography[125] B MacLeod. An ultra-low-vibration facility for housing a dilutiontemperature scanning tunneling microscope. University of BritishColumbia. Master’s Thesis. 2015.[126] J V Barth, H Brune, G Ertl, and R J Behm. Scanning tunneling mi-croscopy observations on the reconstructed Au(111) surface: Atomicstructure, long-range superstructure, rotational domains, and surfacedefects. Phys. Rev. B, 42(15):9307–9318, November 1990.[127] W Chen, V Madhavan, T Jamneala, and M F Crommie. ScanningTunneling Microscopy Observation of an Electronic Superlattice at theSurface of Clean Gold. Phys. Rev. Lett., 80(7):1469–1472, February1998.[128] Jascha Repp, Gerhard Meyer, Sladjana Stojkovic´, Andre´ Gourdon,and Christian Joachim. Molecules on Insulating Films: Scanning-Tunneling Microscopy Imaging of Individual Molecular Orbitals. Phys.Rev. Lett., 94(2):026803, January 2005.[129] C Chavy, C Joachim, and A Altibelli. Interpretation of STM images:C60 on the gold (110) surface. Chem. Phys. Lett., 214(6):569–575,November 1993.[130] Je´roˆme Lagoute, Kiyoshi Kanisawa, and Stefan Fo¨lsch. Manipulationand adsorption-site mapping of single pentacene molecules on Cu(111).Phys. Rev. B, 70(24):245415, December 2004.[131] Sarah-Charlotta Heidorn, Andre´ Sabellek, and Karina Morgenstern.Size Dependence of the Dispersion Relation for the Interface Statebetween NaCl(100) and Ag(111). Nano Lett., 14(1):13–17, January2014.[132] Ingmar Swart, Tobias Sonnleitner, Judith Niedenfu¨hr, and JaschaRepp. Controlled Lateral Manipulation of Molecules on InsulatingFilms by STM. Nano Lett., 12(2):1070–1074, February 2012.[133] K A Cochrane, A Schiffrin, T S Roussy, M Capsoni, and S A Burke.Pronounced polarization-induced energy level shifts at boundaries oforganic semiconductor nanostructures. Nat. Commun., 6:8312, Octo-ber 2015.[134] S Soubatch, C Weiss, R Temirov, and F Tautz. Site-SpecificPolarization Screening in Organic Thin Films. Phys. Rev. Lett.,102(17):177405, May 2009.131Bibliography[135] H M Zhang, J B Gustafsson, and L S O Johansson. STM study ofthe electronic structure of PTCDA on Ag/Si(111). Chem. Phys L˙ett.,485(1-3):69–76, January 2010.[136] Mathieu Linares, David Beljonne, Je´roˆme Cornil, Kelly Lancaster,Jean-Luc Bre´das, Stijn Verlaak, Alexander Mityashin, Paul Heremans,Andreas Fuchs, Christian Lennartz, Julien Ide´, Raphae¨l Me´reau,Philippe Aurel, Laurent Ducasse, and Fre´de´ric Castet. On the Inter-face Dipole at the Pentacene−Fullerene Heterojunction: A TheoreticalStudy. J. Phys. Chem. C, 114(7):3215–3224, February 2010.[137] I G Hill, A Kahn, Z G Soos, and R A Pascal Jr. Charge-separationenergy in films of pi-conjugated organic molecules. Chem. Phys. Lett.,327(3-4):181–188, 2000.[138] Naoki Sato, Kazuhiko Seki, and Hiroo Inokuchi. Polarization energiesof organic solids determined by ultraviolet photoelectron spectroscopy.J. Chem. Soc., Faraday Trans. 2, 77(9):1621, 1981.[139] Jian-Qiang Zhong, Xinming Qin, Jia-Lin Zhang, Satoshi Kera, NobuoUeno, Andrew Thye Shen Wee, Jinlong Yang, and Wei Chen. En-ergy Level Realignment in Weakly Interacting Donor–Acceptor BinaryMolecular Networks. ACS Nano, 8(2):1699–1707, February 2014.[140] T R Umbach, I Fernandez-Torrente, J N Ladenthin, J I Pascual, andK J Franke. Enhanced charge transfer in a monolayer of the organiccharge transfer complex TTF–TNAP on Au(111). J. Phys.: Condens.Matter, 24(35):354003, August 2012.[141] Roman Forker, Christian Golnik, Giovanni Pizzi, Thomas Dienel, andTorsten Fritz. Optical absorption spectra of ultrathin PTCDA filmson gold single crystals: Charge transfer beyond the first monolayer.Org. Electron., 10(8):1448–1453, December 2009.[142] Hiroyuki Yamane, Satoshi Kera, Koji Okudaira, Daisuke Yoshimura,Kazuhiko Seki, and Nobuo Ueno. Intermolecular energy-band disper-sion in PTCDA multilayers. Phys. Rev. B, 68(3):033102, July 2003.[143] F S Tautz. Structure and bonding of large aromatic molecules onnoble metal surfaces: The example of PTCDA. Prog. Surf. Sci., 82(9-12):479–520, September 2007.132Bibliography[144] M Mo¨bus, N Karl, and T Kobayashi. Structure of perylene-tetracarboxylic-dianhydride thin films on alkali halide crystal sub-strates. J. Cryst. Growth, 116(3-4):495–504, 1992.[145] S R Forrest. Ultrathin organic films grown by organic molecular beamdeposition and related techniques. Chem. Rev., 1997.[146] T Ogawa, K Kuwamoto, S Isoda, T Kobayashi, and N Karl. 3,4:9,10-Perylenetetracarboxylic dianhydride (PTCDA) by electron crystallog-raphy. Acta Crystallogr. B Struct. Sci. Cryst. Eng. Mater., 55(1):123–130, February 1999.[147] J Kro¨ger, H Jensen, R Berndt, R Rurali, and N Lorente. Molecularorbital shift of perylenetetracarboxylic-dianhydride on gold. Chem.Phys. Lett., 438(4-6):249–253, April 2007.[148] S Duhm, A Gerlach, I Salzmann, B Bro¨ker, R L Johnson, F Schreiber,and N Koch. PTCDA on Au(111), Ag(111) and Cu(111): Correla-tion of interface charge transfer to bonding distance. Org. Electron.,9(1):111–118, February 2008.[149] Michael Rohlfing, Ruslan Temirov, and Frank Tautz. Adsorptionstructure and scanning tunneling data of a prototype organic-inorganicinterface: PTCDA on Ag(111). Phys. Rev. B, 76(11):115421, Septem-ber 2007.[150] L Kilian, A Hauschild, R Temirov, S Soubatch, A Scho¨ll, A Ben-dounan, F Reinert, T L Lee, F Tautz, M Sokolowski, and E Umbach.Role of Intermolecular Interactions on the Electronic and GeometricStructure of a Large pi-Conjugated Molecule Adsorbed on a MetalSurface. Phys. Rev. Lett., 100(13):136103–4, April 2008.[151] S A Burke, W Ji, J M Mativetsky, J M Topple, S Fostner, H JGao, H Guo, and P Gru¨tter. Strain induced dewetting of a molec-ular system: bimodal growth of PTCDA on NaCl. Phys. Rev. Lett.,100(18):186104, May 2008.[152] E Le Moal, M Mu¨ller, O Bauer, and M Sokolowski. Stable andmetastable phases of PTCDA on epitaxial NaCl films on Ag (100).Phys. Rev. B, 2010.[153] S Heidorn, C Bertram, J Koch, K Boom, F Matthaei, A Safiei, J Henzl,and K Morgenstern. Influence of substrate surface-induced defects on133Bibliographythe interface state between NaCl(100) and Ag(111). J. Phys. Chem.C, 117(31):16095–16103, August 2013.[154] Fabian Mohn, Jascha Repp, Leo Gross, Gerhard Meyer, Matthew SDyer, and Mats Persson. Reversible Bond Formation in a gold-atom–organic-molecule complex as a molecular switch. Phys. Rev.Lett., 105(26):266102, December 2010.[155] H Aldahhak, W G Schmidt, and E Rauls. Adsorption of PTCDA onNaCl(100) and KCl(100). Surf. Sci., 617:242–248, November 2013.[156] B Such, G Goryl, S Godlewski, J J Kolodziej, and M Szymonski.PTCDA molecules on a KBr/InSb system: a low temperature STMstudy. Nanotechnology, 19(47):475705, October 2008.[157] Q Guo, A Paulheim, M Sokolowski, H Aldahhak, E Rauls, and W GSchmidt. Adsorption of PTCDA on Terraces and at Steps Sites of theKCl(100) Surface. J. Phys. Chem. C, 118(51):29911–29918, December2014.[158] M Grobis, A Wachowiak, R Yamachika, and M F Crommie. Tuningnegative differential resistance in a molecular film. Appl. Phys. Lett.,86(20):204102, 2005.[159] Yongqiang Xue, Supriyo Datta, Seunghun Hong, R Reifenberger, Ja-son I Henderson, and Clifford P Kubiak. Negative differential re-sistance in the scanning-tunneling spectroscopy of organic molecules.Phys. Rev. B, 59(12):R7852–R7855, March 1999.[160] Felix Marczinowski, Jens Wiebe, Focko Meier, Katsushi Hashimoto,and Roland Wiesendanger. Effect of charge manipulation on scan-ning tunneling spectra of single Mn acceptors in InAs. Phys. Rev. B,77(11):115318, March 2008.[161] R Bennewitz, M Bammerlin, and M Guggisberg. Aspects of dynamicforce microscopy on NaCl/Cu (111): resolution, tip-sample interac-tions and cantilever oscillation characteristics. Surf. Interface Anal.,27(5):462–466, 1999.[162] T Chasse´, C I Wu, I G Hill, and A Kahn. Band alignment at organic-inorganic semiconductor interfaces: α-NPD and CuPc on InP(110). J.Appl. Phys., 85(9):6589, 1999.134Bibliography[163] Haibo Wang, Patrick Amsalem, Georg Heimel, Ingo Salzmann, Nor-bert Koch, and Martin Oehzelt. Band-Bending in Organic Semicon-ductors: the Role of Alkali-Halide Interlayers. Adv. Mater., 26(6):925–930, December 2013.[164] Hans-Christoph Ploigt, Christophe Brun, Marina Pivetta, Franc¸oisPatthey, and Wolf-Dieter Schneider. Local work function changes de-termined by field emission resonances: NaCl/Ag(100). Phys. Rev. B,76(19):195404, November 2007.[165] A Kraft, R Temirov, S K M Henze, S Soubatch, M Rohlfing, andF S Tautz. Lateral adsorption geometry and site-specific electronicstructure of a large organic chemisorbate on a metal surface. Phys.Rev. B, 74(4):041402, July 2006.[166] R Hesper, L H Tjeng, and G A Sawatzky. Strongly reduced bandgap in a correlated insulator in close proximity to a metal. EPL,40(2):177–182, October 1997.[167] M Mura, X Sun, F Silly, H T Jonkman, G A D Briggs, M R Castell,and L N Kantorovich. Experimental and theoretical analysis of H-bonded supramolecular assemblies of PTCDA molecules. Phys. Rev.B, 81(19):195412, May 2010.[168] R Temirov, S Soubatch, A Luican, and F S Tautz. Free-electron-likedispersion in an organic monolayer film on a metal substrate. Nature,444(7117):350–353, November 2006.[169] I Horcas, R Ferna´ndez, J M Go´mez-Rodr´ıguez, J Colchero, J Go´mez-Herrero, and A M Baro. WSXM: A software for scanning probemicroscopy and a tool for nanotechnology. Rev. Sci. Instrum.,78(1):013705, 2007.[170] Jing Guo, Xiangzhi Meng, Ji Chen, Jinbo Peng, Jiming Sheng, Xin-Zheng Li, Limei Xu, Jun-Ren Shi, Enge Wang, and Ying Jiang. Real-space imaging of interfacial water with submolecular resolution. Nat.Mater., 13(2):1–6, January 2014.[171] R Hesper, L H Tjeng, and G A Sawatzky. Strongly reduced bandgap in a correlated insulator in close proximity to a metal. EPL,40(2):177–182, 1997.135Bibliography[172] Gabriele D’Avino, Davide Vanzo, and Zolta´n G Soos. Dielectric prop-erties of crystalline organic molecular films in the limit of zero overlap.J. Chem. Phys., 144(3):034702, January 2016.[173] Z G Soos and E V Tsiper. Charge redistribution and polarizationenergy of organic molecular crystals. Phys. Rev. B, 64(19): 195124,October 2001.[174] Benjamin J Topham, Manoranjan Kumar, and Zolta´n G Soos. Ioniza-tion potentials of crystalline organic thin films: Position dependencedue to molecular shape and charge redistribution. Chem. Phys. Lett.,493(4-6):251–254, June 2010.[175] Oliver T Hofmann, Patrick Rinke, Matthias Scheffler, andGeorg Heimel. Integer versusFractional Charge Transfer atMetal(/Insulator)/Organic Interfaces: Cu(/NaCl)/TCNE. ACSNano, 9(5):5391–5404, May 2015.[176] Gabriele D’Avino, Se´bastien Mothy, Luca Muccioli, Claudio Zannoni,Linjun Wang, Je´roˆme Cornil, David Beljonne, and Fre´de´ric Castet.Energetics of Electron–Hole Separation at P3HT/PCBM Heterojunc-tions. J. Phys. Chem. C, 117(25):12981–12990, June 2013.[177] D J Griffiths. Introduction to Electrodynamics. 2005.[178] W Steurer, J Repp, L Gross, I Scivetti, M Persson, and G Meyer.Manipulation of the charge state of single Au atoms on insulatingmultilayer films. Phys. Rev. Lett., 114(3):036801, January 2015.[179] N Kocic´, P Weiderer, S Keller, S Decurtins, S X Liu, and J Repp.Periodic Charging of Individual Molecules Coupled to the Motion ofan Atomic Force Microscopy Tip. Nano Lett., 15(7):4406–4411, June2015.[180] Martin Ondra´cˇek, Prokop Hapala, and Pavel Jel´ınek. Charge-state dy-namics in electrostatic force spectroscopy. Nanotechnology, 27(27):1–13, May 2016.[181] N Kocic´, S Decurtins, S X Liu, and J Repp. Forces from periodic charg-ing of adsorbed molecules. J. Chem. Phys., 146(9):092327, February2017.136Bibliography[182] J Peng, J Guo, P Hapala, D Cao, and R Ma. Submolecular-resolutionnon-invasive imaging of interfacial water with atomic force microscopy.arXiv, 2017.[183] Lynda Cockins, Yoichi Miyahara, Steven D Bennett, Aashish AClerk, Sergei Studenikin, Philip Poole, Andrew Sachrajda, and PeterGru¨tter. Energy levels of few-electron quantum dots imaged and char-acterized by atomic force microscopy. Proc. Natl. Acad. Sci. U.S.A.,107(21):9496–9501, May 2010.[184] Antoine Roy-Gobeil, Yoichi Miyahara, and Peter Gru¨tter. Revealingenergy level structure of individual quantum dots by tunneling ratemeasured by single-electron sensitive electrostatic force spectroscopy.Nano Lett., 15(4):2324–2328, March 2015.[185] Marina Pivetta, Franc¸ois Patthey, Massimiliano Stengel, AlfonsoBaldereschi, and Wolf-Dieter Schneider. Local work function Moire´pattern on ultrathin ionic films: NaCl on Ag(100). Phys. Rev. B,72(11):115404, September 2005.[186] S R Forrest. The limits to organic photovoltaic cell efficiency. MRSBulletin, 30(1):28–32, January 2005.[187] Ze-Lei Guan, Jong Bok Kim, Yueh-Lin Loo, and Antoine Kahn. Elec-tronic structure of the poly(3-hexylthiophene):indene-C60 bisadductbulk heterojunction. J. Appl. Phys., 110(4):043719, 2011.[188] F H Moser and A L Thomas. Phthalocyanine compounds. J. Chem.Educ., 1964.[189] K Y Law. Organic photoconductive materials: recent trends and de-velopments. Chem. Rev., 1993.[190] Benjamin Siegert, Andrea Donarini, and Milena Grifoni. Nonequi-librium spin crossover in copper phthalocyanine. Phys. Rev. B,93(12):121406, March 2016.[191] Fabian Schulz, Mari Ija¨s, Robert Drost, Sampsa K Ha¨ma¨la¨inen, AriHarju, Ari P Seitsonen, and Peter Liljeroth. Many-body transitions ina single molecule visualized by scanning tunnelling microscopy. Nat.Phys., 11(3):229–234, January 2015.137Bibliography[192] Noa Marom, Oded Hod, Gustavo E Scuseria, and Leeor Kronik. Elec-tronic structure of copper phthalocyanine: A comparative densityfunctional theory study. J. Chem. Phys., 128(16):164107, April 2008.[193] Pa¨ivi Ja¨rvinen, Sampsa K Ha¨ma¨la¨inen, Mari Ija¨s, Ari Harju, and PeterLiljeroth. Self-assembly and orbital imaging of metal phthalocyanineson a graphene model surface. J. Phys. Chem. C, 118(24):13320–13325,June 2014.[194] Andrew J Wallace, Bryce E Williamson, and Deborah L Crittenden.CASSCF-based explicit ligand field models clarify the ground stateelectronic structures of transition metal phthalocyanines (MPc; M =Mn, Fe, Co, Ni, Cu, Zn). Can. J. Chem., 94(12):1163–1168, October2016.[195] Noa Marom, Xinguo Ren, Jonathan E Moussa, James R Chelikowsky,and Leeor Kronik. Electronic structure of copper phthalocyanine fromG0W0calculations. Phys. Rev. B, 84(19):195143, November 2011.[196] C Uhlmann, I Swart, and J Repp. Controlling the Orbital Sequence inIndividual Cu-Phthalocyanine Molecules. Nano Lett., 13(2):777–780,February 2013.[197] R Ramprasad and N Shi. Polarizability of phthalocyanine based molec-ular systems: A first-principles electronic structure study. Appl. Phys.Lett., 88(22):222903, June 2006.[198] Caroline Henneke, Janina Felter, Daniel Schwarz, F Stefan Tautz, andChristian Kumpf. Controlling the growth of multiple ordered hetero-molecular phases by utilizing intermolecular repulsion. Nat. Mater.,109:7928, March 2017.[199] Tanya Roussy. Coupling spatially and spectrally resolved opticalmeasurements with a scanning probe system. University of BritishColumbia. Master’s Thesis. September 2016.[200] R Berndt, R Gaisch, J K Gimzewski, and B Reihl. Photon emissionat molecular resolution induced by a scanning tunneling microscope.Science, 1993.[201] Chao Zhang, Liuguo Chen, Rui Zhang, and Zhenchao Dong. Scanningtunneling microscope based nanoscale optical imaging of molecules onsurfaces. Jpn. J. Appl. Phys., 54(8S2):08LA01, August 2015.138Bibliography[202] X H Qiu. Vibrationally resolved fluorescence excited with submolecu-lar precision. Science, 299(5606):542–546, January 2003.[203] Elizabeta C´avar, Marie-Christine Blu¨m, Marina Pivetta, Franc¸oisPatthey, Majed Chergui, and Wolf-Dieter Schneider. Fluorescenceand phosphorescence from individual molecules excited by local elec-tron tunneling. Phys. Rev. Lett., 95(19):196102, November 2005.[204] M Prietsch, A Samsavar, and R Ludeke. Structural and electronicproperties of the Bi/GaP(110) interface. Phys. Rev. B, 43(14):11850–11856, May 1991.[205] Florian Albrecht, Martin Fleischmann, Manfred Scheer, Leo Gross,and Jascha Repp. Local tunneling decay length and Kelvin probeforce spectroscopy. Phys. Rev. B, 92(23):235443, December 2015.139Appendix ASTS Normalization andProcessingAll STS data is represented as (dI/dV)/(I/V) unless otherwise stated. Twocorrections were used: a correction for any current offset at zero bias inthe I(V) spectra and the addition of a normalization factor to remove thedivergence at zero bias in the (dI/dV)/(I/V) spectra.A.1 Bias offsetFirstly, to correct for any current discrepancies (likely due to leaky currentin instrumentation), an offset was added to the current values of the rawI(V) curves such that at zero bias, there was zero current. This offset wasfound from an average of the zero crossing of the forward and backwardspectra. This value was typically on the order of 1 x 10−13 A. Figure A.1demonstrates this adjustment.Figure A.1: I(V) curve of an isolated PTCDA on NaCl(2ML)/Ag(111)showing the raw forward and backward signals and the corrected signalafter subtraction of an offset. A scaled version (b.) is included to see thezero bias region.A.2 Normalization offsetThe normalization method used in this thesis is the division of the differ-entiated tunneling conductance by the total conductance, represented as140A.2. Normalization offset(dI/dV)/(I/V).[204] At zero bias, the normalized curve has an undefinedvalue (0/0), resulting in a divergence at the zero crossing. For semicon-ducting molecules on insulating surfaces, there is a large range of negligibletunneling current beginning and ending with the onset of the molecularstates. This results in the divergences occurring at the onset of the states.In order to resolve energy resonances closest to the Fermi energy (EF )an offset was added to the I(V) spectra (Equation A.1).I/V =√(I/V )2 + C2 (A.1)This value, C, was empirically chosen to be just large enough to removethe divergence, but not alter the bias value of any of the states. C<<I/V,typically on the order of 1 x 10−13. As long as this condition is met:I/V ≈ (I/V )dI/dVI/V≈ dIdVI/V(A.2)Figure A.2 shows the normalized spectra with and without the additionof the normalization factor, C. The divergence is clearly seen at the onsetof both the occupied and unoccupied states closest to EF .Figure A.2: Normalized STS of an isolated PTCDA on NaCl(2ML)/Ag(111)with and without the normalization correction offset C.141A.3. SmoothingA.3 SmoothingA moving boxcar average filter (n = 3) was used on all the STS data onthis thesis unless otherwise stated. We determined that this did not affectthe location of the states. With a typical voltage range of ∼4 V, and 512points, this resulted in a resolution of ∼12 mV.142A.3. SmoothingFigure A.3: Raw and n = 3 moving averaged smoothed spectroscopy. a.I(V), b. dI/dV and c., (dI/dV)/(I/V) spectra of raw and smoothed dataof a 12-molecule PTCDA nanoisland (A site) on NaCl(2ML)/Ag(111). Weapplied an n = 3 boxcar moving average filter before computing the numer-ical dI/dV, yielding a bias resolution of ∆V ∗n/2 = 12 meV. The smoothingprocedure does not alter the location of the peaks or obscure any features,given the widths of the tunneling resonances observed. The smoothing aidsmostly in the normalized spectra in regions where the tunneling current isnear zero.143A.4. Kappa maps of a dimer on NaCl(2ML)/Ag(111)A.4 Kappa maps of a dimer onNaCl(2ML)/Ag(111)An additional method of normalization (not used in this thesis) is compen-sating by determining the tunneling barrier decay length. The tunnelingcurrent has an exponential dependence on the tip-sample distance, z.I(z) = I0exp(−2κz) (A.3)The I(z) curves are fit and solved for κ, the tunneling barrier decay length;values are plotted as a function of (x,y) resulting in maps. A difficulty inthis method is the fact that these maps are bias dependent.Figure A.4 displays κmaps of a CuPc/PTCDA dimer on NaCl(2ML)/Ag(111)to of the dimer on the surface. It is known that this technique does not ac-curately reflect the VCPD.[205]Figure A.4: Kappa maps of a PTCDA/CuPc dimer on NaCl(2ML)/Ag(111).6 nm x 4 nm, It = 3 pA, a. Vb = −2.1 V, b. −0.75 V, and c. 1.0 V144Appendix BEFS Fitting and ModelingB.1 Background and theoryThe total force between an AFM tip and a sample is the sum of the longand short range forces:Ftotal = Felectrostatic + Fchemical + FV DW + Fadditional (B.1)where Felectrostatic can be represented in terms of capacitance gradient (∂C∂z )and total potential difference across the tip–sample junction (V) which corre-sponds to the difference between the applied bias and the “contact potentialdifference” (V = Vbias − VCPD).(V ).[92]Felectrostatic = −12∂C∂zV 2 (B.2)Due to the relation in the small amplitude limit∆ff0∝ ∂Ftotal∂z(B.3)we take the derivative of F with respect to z to obtain an equation for thefrequency shift:∆f ∝ −12∂2C∂z2V 2 +∂∂z(Fchemical + FV DW + Fadditional) (B.4)We assume that the chemical, van der Waals, and additional forces arebias independent, and thus constant over the experimental bias range. Thisreduces the equation for frequency shift to∆f(V ) = −12∂2C∂z2(Vbias − VCPD)2 + ∆fmax (B.5)where ∆fmax corresponds to the maximum frequency shift in the paraboliccurve.145B.2. Fitting of VCPD with charging eventsB.2 Fitting of VCPD with charging eventsThe additional effect of charging on the contact potential difference wasexamined. This results in jumps in the df(V ) spectra. To solve for multipleVCPD biases each segment of the curve must be fit separately. Finding theedge of each segment was performed by determining the location of deviationfrom parabolic behavior. For pixel-by-pixel data this must be done in anautomated fashion.To find an approximate location of the jumps in the parabola in order todetermine the range in which each segment should be fitted, a preliminaryfit was found from the center region of the parabola and this was subtractedfrom the data. The resulting difference plot clearly indicates where thejumps are. A straight line was fitted to the middle segment and the edgewas taken when the data deviates from this middle fit (Figure B.1). This isalso how the switching bias VSW was found for the Hubbard maps.Figure B.1: Representation of the analysis used to find the fit parametersand switching bias. a. the raw data, b. a preliminary fit, and c. thedifference between the data and the fit.Using these parameters, the measured df(V) was separated into threesegments with limits 50 meV from the jump bias to account for the gradualtransition. Each segment was fitted with a separate parabola and the max-146B.2. Fitting of VCPD with charging eventsimum for each curve was taken to be the VCPD of each of the charge statesof PTCDA.Figure B.2: Example spectra of a clover island showing three charge stateswith separate fits.147


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items