This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Transcription

1 This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier s archiving and manuscript policies are encouraged to visit:

2 Progress in Surface Science 85 (2010) Contents lists available at ScienceDirect Progress in Surface Science journal homepage: Review Scanning tunneling microscopy of functional nanostructures on solid surfaces: Manipulation, self-assembly, and applications H.-J. Gao *, Li Gao Institute of Physics, Chinese Academy of Sciences, P.O. Box 603, Beijing , China article info abstract Commissioning Editor: J.G. Hou Keywords: Scanning tunneling microscopy Manipulation Perylene Pentacene Iron phthalocyanine Ag(1 1 0) Au(1 1 1) Rotaxane Conductance transition Ultra-high density data storage The manipulation, self-assembly, and application of functional nanostructures on solid surfaces are fundamental issues for the development of electronics and optoelectronics. For a future molecular electronics the fabrication of high-quality organic thin films on metal surfaces is crucial, which can be achieved by thermal evaporation for various organic/metal systems. The switching property of single molecules can be manipulated and measured, revealing a possibility to realize single molecular devices. Manipulation of a local conductance transition in organic thin films, used for ultra-high density data storage, has also been achieved based on several different mechanisms. The stability, reversibility, and repeatability of the local conductance transition have been improved by molecular design. In this article, we will summarize our recent scanning tunneling microscopy studies on these issues and discuss their perspectives. Ó 2009 Elsevier Ltd. All rights reserved. Contents 1. Introduction Growth of organic molecules on noble metal surfaces Abbreviations: AFM, atomic force microscopy; DFT, density functional theory; EDP, electron diffraction pattern; HOMO, highest occupied molecular orbital; HOPG, highly oriented pyrolytic graphite; ITO, indium tin oxide; LB, Langmuir Blodgett; LEED, low energy electron diffraction; LUMO, lowest unoccupied molecular orbital; MBE, molecular beam epitaxy; STM/STS, scanning tunneling microscopy/spectroscopy; TED, transmission electron diffraction; TEM, transmission electron microscopy; UHV, ultra-high vacuum. * Corresponding author. Tel.: ; fax: address: hjgao@iphy.ac.cn (H.-J. Gao) /$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi: /j.progsurf

3 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Monitoring organic thin film growth by MBE LEED Polycyclic aromatic hydrocarbons Phthalocyanine molecules Long alkyl chain substituted molecules Structural and conductance transitions of single rotaxane molecules CH 2 Cl 2 insulating layer STM measurements Molecular manipulation by STM Understanding the geometry of molecules by DFT and MM calculations Conductance transition and ultra-high density data storage on molecular thin films Conductance transition induced by phase transition Conductance transition induced by rupture of hydrogen bonds Conductance transition induced by inter-molecular charge transfer N,N 0 -dimethyl-n 0 (3-nitrobenzylidene)-p-phenylene-diamine (DMNBPDA) [88] ,1,2-Tricyano-2-[(4-dimethylaminophenyl)ethynyl]ethene (TDMEE) [228] Cyano-2,6-dimethyl-4-hydroxy azobenzene (CDHAB) [232] Conductance transition induced by transition of molecular structure Conclusions and outlook Acknowledgements References Introduction Functional nanostructures on solid surfaces are an important concern not only for the academic but also for the industrial community in view of their importance in the development of electronics and optoelectronics. Exploring new materials and developing new techniques for materials preparation [1,2] are of the heart in most of the developments in science and technology. The surface science community is well placed to explore functional materials due to versatile surface analysis techniques, including the scanning tunneling microscopy (STM). STM allows us to see the surface with atomic-resolution by quantum mechanical effect of tunneling. STM can resolve the local electronic structure on the atomic scale on all kinds of conducting surfaces, under various environments with little damage to the sample. Since its invention [3 7], STM has become more and more powerful due to the rapid development of STM-related techniques, and has been helping to expand considerably the scope of atomic scale research [8 18]. The primary advantage of STM is its capability of high-resolution imaging. For example, both the rest atoms and the adatoms of Si(1 1 1)-(7 7) surface simultaneously were observed by STM [19]. STM has been widely used to study the adsorption behavior of atoms or molecules on surfaces [20 25]. Another important application of STM is manipulation at nanometer scale, where single atoms or molecules can be assembled into various desired nanostructures on surfaces. The atomic-scale apex of STM tip can also be used to perform local modifications on surfaces, which is promising for ultra-high density data storage from the perspective of technological applications. Organic semiconductors have great potential applications in electronics and optoelectronics [26]. For example, organic thin-film transistors (OTFTs), which offer the possibility of applications for flexible displays, all-plastic smart cards, as well as organic light-emitting devices (OLEDs), have received widespread attention in recent years [27 34]. In organic devices, carrier transport and luminescent behavior are governed by the orientation and packing of molecules [35 37]. Therefore, enormous efforts have been made to understand and precisely control the formation of organic thin films with various structures [38 47]. Practical applications of organic devices require the use of inexpensive high-quality organic films deposited on various substrates, such as metal, oxidized silicon [48] and silicon surfaces. In this article, we focus on our recent studies on organic/metal systems. Organic semiconductors on single-crystalline metal surfaces are typical systems for injection contacts in OTFTs and OLEDs. While organic multilayers are already applied in current organic devices, the interfaces

4 30 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) between organic molecules and metal surfaces play a crucial role in the performance of the organic devices [49 72]. Organic semiconductors on single-crystalline metal surfaces are typical systems for injection contacts in organic thin-film transistors (OTFTs) and organic light-emitting diodes (OLEDs). Polycyclic aromatic, metal phthalocyanine and quinacridone molecules are among the most important candidates for molecular electronics and optoelectronics. Fabricating high-quality organic/metal interfaces is of primary importance for device performance and presents many challenges. Molecular devices are based on the functional properties of individual molecules [73]. The conductance switching property of organic molecules has potential applications in both, single molecular switches and ultra-high density information storage. Ultra-high density data storage is one important key to the development of information technology [74 81]. Nanometer-scale recording on organic thin films by STM through a conductance transition is a promising strategy [82 100]. However, proper recording mechanisms should be explored in order to obtain reliable performance. In this article, we will review our recent STM studies on these issues and discuss their perspectives. We will start from monitoring the organic thin film growth at the initial stage at single molecular scale, and then focus on the structural and physical properties and potential applications, in particular, the conductance transition and switching at single molecular scale. More specifically, our topics in this article will include: (1) STM, low energy electron diffraction (LEED) and density functional theory (DFT) studies of organic/metal interfaces and growth; (2) manipulation of the structural and the conductance transition of single rotaxane molecules. In this section, rotaxanes, one kind of inter-locked supermolecules, the most promising materials for future nanometer-scale recording is presented; and (3) ultra-high density data storage by means of the local conductance transition in molecular thin films. 2. Growth of organic molecules on noble metal surfaces In this section, we will review monitoring the growth process a few kinds of typical molecules. An organic molecular beam and low energy electron diffraction (OMBE LEED) technique is introduced at first, which was demonstrated to be an efficient way to control the growth process and the resulting structure and physical properties, together with STM imaging and DFT calculations. Pentacene (C 22 H 14 ), an aromatic molecule composed of five benzene rings, is a promising material in organic electronics [31 34,37, ]. Pentacene thin films can be used as the channel layer of OTFTs [31 34,101,102] because the charge mobility of pentacene in OTFTs has reached or even surpassed that of amorphous Si in Si-TFTs [101,102]. Previous studies on pentacene thin films include X-ray [29], atomic force microscopy [30], scanning tunneling microscopy [107], as well as theoretical studies of the adsorption of a single pentacene molecule on different surfaces [108]. Perylene (C 20 H 12 ) is a planar aromatic molecule with rather large intrinsic charge-carrier mobility at low temperatures [109,110]. Perylene has exhibited excellent performance in electronic and lightemitting devices [111]. The formation of perylene adlayers has been studied on metal [109, ] and semiconductor [124,125] surfaces. So far, multilayer growth has been conducted on Si(1 1 1) [124], Au(1 1 1) [112] and Cu(1 1 0) [112,117,120] surfaces. Some previous studies show that perylene molecules assemble with the p-plane oriented almost or completely parallel to the substrate in the multilayer regime [117,124]. Other studies show that a planar or near-planar orientation of perylene molecules is limited to the first monolayer, and the transition to the bulk structure occurs with increasing film thickness [112,120]. Metal phthalocyanine (MPc) materials have wide applications in gas-sensing devices, photovoltaic applications, light-emitting diodes, solar and fuel cells, organic field effect transistors (OFETs), pigments, and dyes [ ]. Hipps and co-workers carried out a series of studies on the self-assembly of MPcs (FePc, NiPc, CoPc, CuPc) on a Au(1 1 1) surface in UHV [ ], and found that different central metal atoms present a different brightness contrast in STM images due to different d orbital occupations. Itaya and co-workers succeeded in preparing well-defined adlayers of MPcs (CoPc, CuPc, ZnPc) on a Au(1 1 1) surface by immersing an Au(1 1 1) substrate into a benzene solution saturated with the molecules by means of an electrochemical STM (ECSTM) [ ]. However, their studies

5 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) mainly focused on monolayers. To get a better understanding of the growth process, it is necessary to study the growth behavior at different coverages [136,137]. Quinacridone and its derivatives (QA) are well-known organic pigments and dopant emitters that display excellent chemical stability [ ]. They show very pronounced photovoltaic and photoconductive activities. Because of their good electrochemical stability in solids and light photoluminescent in dilute solutions, quinacridone and its derivatives are promising materials for the fabrication of high-performance organic light-emitting devices [ ]. Moreover, many investigations on quinacridone derivatives have been reported to explore the assembly and structural properties in solid liquid interface. For example, Feyter group applied STM to study the aggregation behavior and two-dimensional (2D) order of 2,3,9,10-tetra(dodecyloxy)quinacridone, 2,9-di(2-undecyltridecyl- 1-oxy)quinacridone, and N,N 0 -dimethyl-substituted analogues on highly oriented pyrolytic graphite (HOPG) [145,146]. Zhang and co-workers reported 2D self-assemblies of a series of N,N 0 - dialkyl-substituted quinacridone derivatives on a HOPG-solution interface. They found that this system can be fine-tuned by co-adsorbing with monofunctional acid or bifunctional dicarboxylic acids [ ] and can also form chiral racemoids in the co-adsorption structures [148]. Gold and silver are commonly used as the contact materials for source and drain electrodes in OFETs. Metallic surfaces are thermodynamically unstable in a cleaved-bulk configuration, and some of them reconstruct into an atomic arrangement different from the bulk one. For example, the bare Au(1 1 1) surface shows a herringbone reconstruction pattern exhibiting an ordered array of domain boundaries between surface regions with different atomic stacking, face-centered-cubic (fcc) and hexagonal-close-packed (hcp) stacking [ ]. By contrast, the bare Ag(1 1 0) surface has the same atomic arrangement with the bulk crystal Monitoring organic thin film growth by MBE LEED Deposition of organic molecules from the gas phase in UHV, sometimes referred to as organic molecular beam deposition (OMBD) or organic molecular beam epitaxy (OMBE), is an ideal means to achieve monolayer control over the growth of organic thin films with extremely high chemical purity and structural precision. In the experiments described in detail in this review, we used a MBE LEED apparatus, which integrates MBE with LEED and allows in situ recording of diffraction patterns in real time during molecular deposition [109,155,156]. Fig. 1 shows a schematic of the MBE LEED apparatus. It is perfectly suitable to investigate the growth of organic thin films on metal surfaces [115,116, ]. In the following, the evolution of the LEED pattern will be described for pentacene on Ag(1 1 0). Similar evolution processes have been observed for other adsorption systems included in this section. STM experiments were performed using an Omicron UHV STM. The single crystal Ag(1 1 0) and Au(1 1 1) surfaces were cleaned by repeated cycles of Ar + sputtering and subsequent annealing at a pressure of mbar. Subsequently, the quality of the cleaned surfaces was checked by LEED and STM before thermal evaporation experiments. All the as-received organic materials were effectively purified using the temperature gradient sublimation method [162] and immediately loaded into the sublimation cells. The cells were kept at elevated temperatures for outgassing before the thermal evaporation experiments Polycyclic aromatic hydrocarbons Polycyclic aromatic hydrocarbons (PAHs) consist of fused aromatic rings and do not contain heteroatoms or carry substitutes. PAHs exhibit characteristic UV absorbance spectra. Most of them are also fluorescent, emitting characteristic wavelength of light when they are excited. The extended p electron electronic structures of PAHs lead to these spectra and semi-conducting behaviors for some large PAHs. Therefore, PAHs have been important candidates for the electronic and optoelectronic devices. The unique structures of PAHs results into a characteristic type of molecular adsorption on noble metal surfaces. In the following, we discuss their adsorption behavior by three specific cases: pentacene on Ag(1 1 0), perylene on Ag(1 1 0), and perylene on Au(1 1 1). By these three cases, we can have some idea about the difference between different PAHs as well as the effect of metal surfaces on molecular adsorption.

6 32 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fluorescence screen Electron gun Sample Three crucibles Fig. 1. Schematic of the MBE LEED apparatus. Fig. 2 shows the recorded LEED patterns during the deposition of a pentacene monolayer on Ag(1 1 0). The silver substrate was kept at room temperature during deposition. As the crucible temperature increases from room temperature to 140 C, there is no change in the LEED pattern, indicating no additional molecules are deposited on the Ag(1 1 0) surface. When the crucible temperature rises gradually to 145 C, very few molecules adsorb onto the substrate. At this stage the LEED pattern looks similar to a halo, as shown in Fig. 2a. Fig. 2b shows that the diffraction intensity is enhanced and the halo assumes an elliptical pattern with further deposition. The ellipse finally decays and some spots appear in Fig. 2d, indicating a periodic structure of the adsorbate on the substrate. Fig. 2e and f shows sharp diffraction spots, corresponding to a well ordered structure of pentacene on the Ag(1 1 0) surface. This diffraction pattern evolution indicates that before nucleation and growth the pentacene molecules are highly mobile on the Ag(1 1 0) surface at room temperature. With increasing molecular coverage, the diffraction pattern does not change any more, as shown in Fig. 2f, which indicates that there is no structural transition with further deposition. While structural transitions have been observed for many other kinds of molecules with increasing molecular coverage, no dose-induced structural transition occurs in this molecular system, suggesting that domains grow laterally without change of their internal structure. Fig. 2g is obtained at higher electron beam energy of 34 ev, and includes the diffraction spots of both silver substrate and pentacene film. Two different structures can be deduced from Fig. 2e g, which are caused by two domain orientations mirrored at a crystal plane of the Ag(1 1 0) substrate [163,164]. The lattice relationship between the pentacene film and the Ag(1 1 0) substrate is: b 1 b 2 ¼ a1 a 2 Here, (b 1, b 2 ) are the unit cell vectors of the pentacene film, and (a 1, a 2 ) are the unit cell vectors of the Ag(1 1 0) substrate. Annealing after growth, in many cases, leads to a structural transition. For our experiments on pentacene/ag(1 1 0), the position and the intensities of the diffraction spots do not change significantly when the sample temperature is increased gradually from room temperature to 140 C, indicating that no structural transition occurs. However, when the sample temperature is increased to 145 C, which is close to the temperature used for normal film preparation, the LEED diffraction spots slowly become dark (Fig. 2i) and finally disappear, indicating the desorption of the molecules from the substrate. Generally speaking, the desorption temperature of a molecule on silver is higher than the molecular sublimation temperature if there is covalent bonding between the molecule and the substrate [164,165]. Here the desorption temperature of 145 C is a little bit lower than the sublimation temperature. Thus, the interaction between pentacene molecules and the Ag(1 1 0) substrate is a weak van der Waals interaction, different from the chemical adsorption of O atoms and polar groups on Ag(1 1 0).

7 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 2. In situ LEED patterns during molecular deposition showing the structural evolution of the pentacene molecules on the Ag(1 1 0) surface. (a) Beam energy E = 13 ev, substrate temperature T s =20 C, evaporation temperature T v = 145 C, deposition time t = 20 s; (b) E = 13 ev, T s =20 C, T v = 150 C, t = 30 s; (c) E = 13 ev, T s =20 C, T v = 151 C, t = 40 s; (d) E = 13 ev, T s =20 C, T v = 152 C, t = 50 s; (e) E = 13 ev, T s =20 C, T v = 153 C, t = 60 s; (f) E = 13 ev, T s =20 C, T v = 154 C, t = 65 s; (g) E = 34 ev, T s =20 C, T v = 154 C, t = 65 s; (h) E = 13 ev, T s = 140 C; (i) E = 13 ev, T s = 145 C. Molecular mechanics calculations have been conducted to find the favorable adsorption site and molecular orientation of pentacene on the Ag(1 1 0) surface [166]. The MM+ force field computational scheme, which is improved from the MM2 force field, was used. In this method, the configuration of the model system in equilibrium is obtained by minimizing the energy. Our calculations reveal that: (1) pentacene molecules prefer to adopt a flat-lying geometry on the Ag(1 1 0) surface; (2) the center of the molecular plane prefers to be on the bridge site between two adjacent topmost Ag atoms; (3) there are two favorable molecular orientations. Fig. 3 shows the two optimized configurations for a single pentacene molecule on the Ag(1 1 0) surface. Besides Ag(1 1 0), perylene molecules adopt a flat-lying geometry also on Cu(1 1 0) [107] and Si(0 0 1) surfaces [108]. In order to determine the favorable molecular orientations of the pentacene in the flat-lying mode, we calculated 5 5 pentacene unit cells adsorbed on the Ag(1 1 0) substrate. Based on the calculated results for a single molecule, we propose four possible molecular orientations A, B, C, and D on the Ag(1 1 0) substrate, as shown in Fig. 4. The lattice relationship between the pentacene film and the Ag(1 1 0) substrate is deduced from the LEED pattern in Fig. 2. The distance between the molecular layer and the top layer of the substrate surface is set to be about 5 Å initially. After the optimization,

8 34 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) molecules in either A or B orientation are shifted from their original orientations. Orientation A finally changes to orientation C (see Fig. 4c), and orientation B changes to orientation D (see Fig. 4d). The calculated lattice parameters are consistent with the parameters deduced from the LEED pattern with an error of ±5%. For orientations A and B, the distance between the pentacene layer and the substrate surface is finally reduced to 3.3 Å. We have also calculated the single point energy (E sp ) and the optimized energy (E opt ) for orientations C and D. The results are the same as for orientations A and B, and the distance between the pentacene layer and the substrate surface is also reduced to 3.3 Å, indicating that orientations C and D are favorable for pentacene molecules on the Ag(1 1 0) substrate. Fig. 3. Two optimized molecular orientations for pentacene on Ag(1 1 0). Fig. 4. Schematic diagram of pentacene molecules on Ag(1 1 0) in four specific orientations.

9 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) For perylene molecules on the same metal surface Ag(1 1 0), preferential adsorption at step edges is observed in the initial growth stage. With increasing coverage but still within submonolayer coverage, molecules distribute homogeneously on the silver surface showing no molecular order (see Fig. 5a). Quasi-ordered molecular arrangements were observed when the coverage was increased close to one monolayer (see Fig. 5b). Highly ordered arrangements were obtained at monolayer coverage (see Fig. 5c and d). LEED pattern reveal that the perylene superstructure has two domain orientations mirrored at a crystal plane of the Ag(1 1 0) substrate, induced by the twofold symmetry of the substrate. The commensurate superstructure determined by the LEED pattern is approximately, b 1 ¼ 4 1 a1 1 3 b 2 a 2 Here, (b 1, b 2 ) and (a 1, a 2 ) are the unit cell vectors of the perylene monolayer and of the Ag(1 1 0) substrate, respectively. STM results at 77 K show the same superstructure. In many cases, the difference in the thermal expansion coefficient of the metal substrate and the organic adlayers leads to structural Fig. 5. (a) STM image (35 nm 35 nm, U = 1.8 V, I = 0.04 na) showing the scenario after one submonolayer, with a molecular coverage of molecules/cm 2, was cooled down to 7 K. (b) STM image (35 nm 35 nm, U = 1.7 V, I = 0.11 na) showing the scenario after another submonolayer, with a molecular coverage of molecules/cm 2, was cooled down to 77 K. (c) High-resolution STM image (30 nm 30 nm, U = 0.9 V, I = 0.21 na) showing the domain boundaries. (d) STM image (115 nm 115 nm, U = 1.7 V, I = 0.08 na) of the highly ordered perylene monolayer without any domain boundaries.

10 36 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) transitions with increasing temperature. However, such a transition was not observed in our experiments. Several other superstructures have been observed in previous studies [109]. In our experiments, when the deposition rate changes from ML/min, corresponding to the deposition rate for the uniform monolayer, to ML/min, many metastable structures appear. Fig. 6 shows a STM image of one model structure including two domains due to the twofold symmetry of the Ag(1 1 0) substrate. By analyzing the STM images of the model structure and those of the uncovered Ag(1 1 0) surface at atomic-resolution, we determine the model structure as: c 1 ¼ 4 0 a1 1 2 c 2 a 2 Here, (c 1, c 2 ) and (a 1, a 2 ) are the unit cell vectors of the model metastable structure and of the substrate lattice, respectively. The perylene superstructure on Ag(1 1 0) is very sensitive to growth conditions, which reveals a poor self-assembly ability in this case. Molecular self-assembly is driven by the surface structure of the substrate and the balance between inter-molecular interaction and molecule substrate interaction. In some cases, the surface structure can serve as a template for the formation of one-dimensional or two-dimensional molecular nanostructures [50,51,167]. In other cases, the molecular self-assembly on flat and unstructured surfaces can be realized by strong molecule substrate interaction or intermolecular interaction [107,168]. For some organic/metal systems, such as perylene/ag(1 1 0), the molecule substrate interaction, the inter-molecular interaction and the template-influence of the substrate are all quite weak, in which case the fabrication of highly ordered molecular nanostructures is quite sensitive to growth conditions due to the lack of an internal dominating driving force for selfassembly. These systems can therefore be classified as weakly interacting. The growth process is inherently a non-equilibrium phenomenon governed by the competition between kinetics and thermodynamics [169]. The diffusion of a molecule on a flat terrace is the most important kinetic process in the monolayer growth. The surface diffusion coefficient D is related to the site to site hopping rate of a molecule k s by D = a 2 k s, where a is the effective hopping distance between sites and k s is proportional to exp( V s /k B T), where V s is the potential-energy barrier from site to site, T is the substrate temperature, and k B is the Boltzmann constant. In our experiment, if perylene molecules are deposited on the Ag(1 1 0) surface at a constant deposition rate F, then the ratio D/F determines the average distance that a perylene molecule has to travel to arrive at its final adsorption Fig. 6. STM image (6 nm 6 nm, U = 0.8 V, I = 0.17 na) of the modal metastable structure.

11 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) site before the coverage is increased up to one monolayer, because perylene molecules do not aggregate into molecular islands in the submonolayer coverage region. Therefore, the ratio of deposition to diffusion rate D/F is the key parameter characterizing growth kinetics. In the case of large values of D/ F, growth occurs close to equilibrium conditions; that is, the molecules have enough time to explore the potential energy surface so that the system reaches a minimum energy configuration. On the contrary, in the case of small values of D/F, the growth is essentially determined by kinetics; individual processes, especially those leading to metastable structures, are increasingly important. The favorable adsorption configuration for perylene on Ag(1 1 0) surface is shown in Fig. 7a [115,116]. The simulated STM image (see Fig. 7b) of the fully relaxed stable structure in Fig. 7a, obtained by calculating the density of states (DOS) using the Tersoff Hamann approach [170], is in good agreement with the experimental STM image (see Fig. 7c) using identical tunneling conditions of U = 1.2 V and I = 0.13 na in experiment and simulation. Substrate is an important factor for molecular adsorption. Normally molecules exhibit different adsorption behavior on different substrates. Fig. 8a is a typical STM image showing the scenario at 78 K after 0.1 ML perylene molecules were deposited onto the Au(1 1 1) surface at 255 K. Individual molecules cannot be clearly imaged at this coverage at 78 K due to rapid diffusion over the gold surface. On increasing the molecular coverage up to 0.3 ML, some bright molecular aggregates were observed in our STM measurements at 78 K (see Fig. 8b). The molecular aggregates can hardly diffuse over the surface but vibrate locally, thus they can be imaged as bright aggregates. The adsorption locations of all aggregates are very selectively influenced by the surface reconstruction of the Au(1 1 1) substrate. Ordered molecular arrangements form when the coverage increases up to one monolayer, as shown in Fig. 8c and d. Combining LEED results and STM measurements, we determined the monolayer lattice as a (4 4) structure: b 1 b 2 ¼ a1 a 2 Here, a 1 and a 2 are the unit cell vectors of Au(1 1 1), b 1 and b 2 are the unit cell vectors of the perylene monolayer superstructure. In the STM images, individual perylene molecules extend 1 nm in the ½11 2Š direction and 0.6 nm in the ½1 10Š direction (see Fig. 8d). This indicates that the molecular long axis is along the ½11 2Š direction, in agreement with the studies by Yoshimoto et al. [123]. Fig. 9 depicts the supposed model for a perylene monolayer on a Au(1 1 1) surface. According to our theoretical calculations, the bridge site is the most stable adsorption site [158]. The growth of the second layer is completely different from the one of the first layer. When additional molecules are evaporated onto the first layer, two-dimensional molecular islands form with an ordered arrangement, as shown in Fig. 10a and b. The sticking coefficient for the molecules impinging on the first monolayer is much lower than the one for the molecules impinging on the gold surface. Only a small number of molecular islands were observed after further depositing molecules onto Fig. 7. (a) The model of the favorable superstructure obtained with energy optimization. (b) Simulated STM image ( 1.2 V) of the favorable superstructure. (c) Experimental STM image (4 nm 4 nm, U = 1.2 V, I = 0.13 na) of perylene superstructure.

12 38 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 8. (a) STM image (50 nm 50 nm, U = 2.3 V, I = 0.08 na) of molecular diffusion. Molecular coverage is 0.1 ML. (b) STM image (50 nm 50 nm, U = 1.3 V, I = 0.05 na) of 0.3 ML perylene molecules on the reconstructed Au(1 1 1) surface. (c) STM image (20 nm 20 nm, U = 0.7 V, I = 0.03 na) of perylene monolayer. (d) STM image (8 nm 8 nm, U = 1.0 V, I = 0.05 na) of perylene monolayer. The sizes of individual molecules are indicated. Molecular long axis is along the ½11 2Š direction. the first monolayer using the same incident flux for 100 min, the typical growth time for the first monolayer. The apparent height of the molecular islands is 4 5 Å (see the line profile in Fig. 10a), varying with the tunneling conditions in STM scanning. Fig. 10b represents a typical STM image showing the molecular arrangement of the molecular island. The line profile from C to D in Fig. 10b shows that molecules assemble in a dimer-like arrangement in the ½112Š direction. Fig. 11a depicts a typical high-resolution STM image showing both the molecular arrangement of the first layer and the one of the second layer. The substrate orientation can be determined from the first layer. Consequently, the structure of the second layer can be determined as: c 1 1:5 3:6 a1 ¼ 4 8 c 2 a 2 In the above equation, c 1 and c 2 are the unit cell vectors of the second layer. Molecules form an oblique two-dimensional Bravais lattice with two molecules at each Bravais lattice point. We determine that the molecules in the second layer adopt a tilted configuration, instead of a flat-lying configuration, according to the molecular density derived from the STM measurements. Fig. 11b illustrates the supposed structural model of the second layer.

13 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 9. Supposed structural model for a perylene monolayer on a Au(1 1 1) surface. The bridge site is selected according to our DFT calculations. The difference in growth mode between the first two layers is ascribed to their different environment with respect to the competition between molecule molecule and molecule substrate interactions. The dominant force for the growth of the first layer is the molecule substrate interaction. The growth of the second layer is dominated by molecule molecule interaction, which results in the growth of ordered molecular islands. The existence of the first layer leads to a remarkable decrease of the interaction between the substrate and the molecules of the second layer. The first layer prevents the direct bonding between the second layer and the gold substrate on the one hand, and weakens the vertical extension of the electronic states of the gold substrate on the other hand. In the di/dv spectra (Fig. 12) measured on the first layer, there is a gap of 2.3 ev between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO), located at 0.9 ev below E F and at 1.4 ev above E F, respectively. The surface state of the gold substrate was not detected in the di/dv measurement. The growth mode for perylene on Au(1 1 1) was also observed for many PAHs [112]. Generally the planar PAHs molecules form a monolayer on metal surfaces with the molecular plane parallel to the substrate [62]. In most cases, including perylene/au(1 1 1), however, the bulk structure of PAHs does not include a low-index and closely packed plane whose molecular arrangements are the same as the molecular monolayer on the metal surface. From the second layer onward, perylene films adopt a different packing arrangement than in the first monolayer, close to the bulk structure. For perylene/ Au(1 1 1), the molecular island is an intermediate structure, with molecules tilting from the surface and forming a dimeric arrangement. The surface structure of an epitaxial perylene multilayer crystal is the herringbone packing of dimeric perylene molecules in the ab-plane of the monoclinic a-perylene single crystal [111,112,171]. Although the structure of the molecular island in the second layer is different from the final multilayer structure, the former has two features of the latter, i.e. dimeric arrangement and a tilting molecular orientation Phthalocyanine molecules Phthalocyanine molecules are macrocyclic compounds including an alternating nitrogen atom carbon atom ring structure. Hydrogen and metal cations can be coordinated into the molecular center by coordinate bonds with the four isoindole nitrogen atoms. Most of the elements are able to coordinate to the phthalocyanine macrocycle. So it will be interesting to explore how the central atoms affect the

14 40 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) a A B b [112] D 10 nm 2 nm C c 0.5 A B d 0.05 C D Height (nm) Height (nm) Distance (nm) Distance (nm) Fig. 10. (a) STM image (90 nm 90 nm, U = 0.8 V, I = 0.08 na) of a perylene island on the first layer. The line profile from A to B in the image is shown in (c). (b) STM image (15 nm 15 nm, U = 0.4 V, I = 0.04 na) on the perylene island. The line profile from C to D in the image is shown in (d). Fig. 11. (a) STM image (40 nm 40 nm, U = 0.4 V, I = 0.04 na) showing the molecular arrangements of both the first layer (I) and the second layer (II). (b) The proposed structural model of the second layer. (c 1, c 2 ) are the unit cell vectors of the second layer. Side view is along the c 1 direction.

15 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Tunneling Current (na) 0.02 perylene monolayer on Au(111): 78 K Sample Bias (V) di/dv (a.u.) Fig. 12. I V and di/dv spectra measured on the perylene monolayer on the Au(1 1 1) surface. adsorption the phthalocyanine molecules on metal surfaces. In the following, we discuss one specific molecule FePc on Au(1 1 1) surface. Preferential adsorption at step edges was observed in the initial growth stage. Most of the molecules stand against the step edges. When the step edges are fully decorated, with increasing coverage FePc molecules begin to adsorb on the terraces. The STM image of a single FePc molecule is a cross with a bright spot at the center. Fig. 13a shows a typical STM image of 0.1 ML FePc molecules on Au(1 1 1). At this coverage, almost all FePc molecules are adsorbed on the fcc regions and at the elbow positions of the reconstructed Au(1 1 1) surface. Fig. 13b shows a typical STM image of 0.3 ML FePc molecules on Au(1 1 1). At this coverage, FePc molecules can also be observed on hcp regions of the reconstructed Au(1 1 1) surface. At submonolayer coverage, there are two adsorption configurations for FePc molecules on the Au(1 1 1) substrate. For one configuration, called Configuration I, the cross is directed in the ½1 10Š and ½11 2Š directions of the Au(1 1 1) substrate; for the other configuration, called Configuration II, the cross rotates with respect to the molecular center by 15 compared to Configuration I. Our theoretical simulations reveal that the central Fe atom is at the bridge site in Configuration I, and at the Fig. 13. STM images of a FePc submonolayer on Au(1 1 1). (a) 0.1 ML (50 nm 50 nm, U= 1.5 V, I = 0.05 na); (b) 0.3 ML (50 nm 50 nm, U= 2.0 V, I=0.05 na). The set of three arrows indicates the close-packed directions of the Au(1 1 1) substrate. The orientations of all images are identical. Broken white lines indicate domain walls of the Au(1 1 1) surface reconstruction.

16 42 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) top site in Configuration II, as shown in Fig. 14. Initial simulations using a PBE exchange correlation functionals show that the adsorption energy of Configuration I is 13 mev higher than the one of Configuration II [160]. But further calculations using PW91, LDA and BLYP functionals reveal that Configuration II is more stable than Configuration I [172]. A highly-ordered FePc monolayer was fabricated by elevating the substrate temperature up to 390 K during thermal evaporation [159]. Fig. 15a shows a typical STM image of a highly-ordered FePc monolayer on Au(1 1 1). The high-resolution STM image of the FePc monolayer (see Fig. 15b) presents directly the real-space configuration of individual molecules and the periodic structure of the organic monolayer. Based on STM and LEED measurements, we determined the molecular superstructure as: b 1 ¼ 5 0 a1 3 6 b 2 a 2 Here, a 1 and a 2 are the unit cell vectors of the Au(1 1 1) substrate, b 1 and b 2 are the unit cell vectors of FePc monolayer. Due to the threefold symmetry of the Au(1 1 1) substrate, three equivalent domains of the monolayer are observed in our STM measurements [ ]. The inset of Fig. 15a presents a sketch of the three symmetry-equivalent domains ([A], [B] and [C]) with the underlying Au(1 1 1) substrate. A line defect in the ordered monolayer is marked by double dashed lines in Fig. 15b. Line defects make the lattice slightly shifted. Fig. 15c shows a large scale STM image of the [A] domain, which consist of two equivalent domains, [A1] and [A2]. Fig. 15d shows the highly ordered molecular arrangements around the domain boundary between [A1] and [A2]. It is obvious that both the lattice and the molecular orientation are identical between [A1] and [A2]. Additional FePc molecules were deposited onto the highly ordered monolayer under identical experimental conditions, and FePc multilayers of different coverages were fabricated. The sticking coefficient of the molecules impinging on the first monolayer is much lower than the one on the bare Au(1 1 1) substrate, and consequently, more time is needed for the deposition of the same amount of molecules. Fig. 16a shows a large scale STM image of the FePc multilayer at a coverage of 1.3 ML. Most of the molecules form well-ordered islands and only few molecules randomly adsorb on the first layer as clusters or individual molecules. All the step edges of the molecular islands are oriented along the ½ 12 1Š and ½10 1Š directions. The size of the islands gets larger with increasing coverage up to 1.7 ML (see Fig. 16b). Fig. 17a shows a STM image of the second layer. LEED and STM measurements reveal that the superstructure of the second layer is the same as the one of the first layer. FePc molecules in the Fig. 14. Optimized adsorption configurations for experimentally observed Configuration I and Configuration II.

17 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 15. STM images of the first monolayer of FePc on Au(1 1 1). (a) 300 nm 300 nm, U= 1.2 V, I=0.05 na; (b) 20 nm 20 nm, U= 0.4 V, I=0.05 na; (c) 60 nm 60 nm, U= 0.4 V, I=0.05 na; (d) 10 nm 10 nm, U= 0.4 V, I=0.05 na. The inset of (a) shows a sketch of three symmetry-equivalent domains ([A], [B] and [C]) with the underlying Au(1 1 1) substrate. One line defect and one unit cell are marked with dashed lines and a square in (b), respectively. Two orientation-domains in the first FePc monolayer on Au(1 1 1) are shown in (c). The dashed-dotted lines indicate the domain boundaries. High-resolution STM image (d) shows the orientations of molecules in both of the orientation-domains. second layer are imaged as triangular structure with only three lobes and a central protrusion, marked by a dotted circle in Fig. 17b, indicating a non-planar adsorption configuration. The non-polar adsorption configuration was also observed for a single FePc molecule adsorbed on the first layer, as shown in Fig. 17c. The molecule is tilted along the ½10 1Š direction, with a lobe downward to the substrate and trapped at the central hole of one unit cell of the first layer. As a consequence, the downward lobe appears dark, while the other three lobes appear bright in Fig. 17c. As to the molecules of the second layer, the downward lobe is not observable, so that the molecule is imaged as a triangle structure with only three lobes and a central protrusion in Fig. 17b. Fig. 17d shows the molecular arrangements of both, the first and the second layer. The boundary between the [B1] and [B2] domains in the first layer is marked by a dashed line. The island of the second layer is on top of the [B2] domain with straight step edges along the ½10 1Š and ½1 21Š directions of the Au(1 1 1) substrate. The molecule at the island corner presents a similar configuration to the one of the single molecule in Fig. 17c. However, there is a slight difference between them, which is possibly due to inter-molecular interactions within the island. The intersections of the dashed square lattice in the image represent the iron atoms of the FePc molecules in the [B2] domain, and those of the dash dotted square lattice for the iron atoms of the FePc molecules in the second layer, which clearly indicates that the superstructure of the second layer is

18 44 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 16. Large-scale STM images of a FePc multilayer on Au(1 1 1). (a) 1.3 ML (120 nm 120 nm, U= 2.2 V, I=0.05 na); (b) 1.7 ML (120 nm 120 nm, U= 2.2 V, I=0.04 na). The growth mode of the second layer differs totally from the one of the first layer. identical to the one of the first layer and that the central metal atoms of FePc in the second layer are not exactly on top of the Fe atoms in the first layer. As shown in Fig. 17c, each FePc molecule has four tilted directions on the first layer. In Fig. 18a, there are two different domains (d 1 and d 2 ) of the second layer on the [C1] domain of the first layer. It is clear that the two domains have an identical quadratic superstructure, but different tilted directions. The domain boundary between the two domains is marked with dashed lines, in which molecules prefer another adsorption configuration. Understanding and controlling the defect formation during the epitaxial growth process are of great significance to improve the performance of organic devices. As shown in Fig. 18b, the line defect A in the first layer extends to the second layer by inducing a line defect B. Therefore, the quality of the first monolayer is of primary importance for the growth of high-quality organic films Long alkyl chain substituted molecules Molecular synthesis offers a variety of molecules, which provides one well-controlled method for tuning device properties in molecular electronics. Molecular arrangement is crucial for the molecular devices. And decorating molecules without changing their primary physical properties is an effective method for modulating the molecular arrangement. Here, we studied the adsorption of long alkyl chain substituted molecules on the metal surface [25]. We can clearly see that the alkyl chain plays an important role in the molecular superstructure. Fig. 19a shows the molecular structure of a quinacridone derivative named QA16C. At dilute molecular coverage, it is hard to obtain stable STM images at room temperature due to the high mobility of the molecules on the silver surface. Fig. 19b is an STM image obtained at 78 K, showing that the molecules absorb both on terraces and at step edges. Isolated molecules on terraces have two different molecular orientations. The alkyl chains of QA16C are observable in the STM image with sub-molecular resolution, as shown in Fig. 19c. The two molecular orientations on silver terraces reveal that the molecule substrate interaction is strong enough to determine the adsorption sites of isolated QA16C molecules on Ag(1 1 0) [59]. The adsorption configuration at the step edges is different from that on terraces, as shown in Fig. 19d. At step edges, molecules are ordered up to several tens of nanometers along Ag ½20209Š step edges. Fig. 19e shows the details of the molecular arrangement at step edges. The inter-molecular distance, marked as l, is 2.9 ± 0.1 nm. Each molecule is imaged as a small protrusion of about 0.05 nm in height with respect to the upper silver terrace. These molecular protrusions are much smaller than on terraces (0.11 ± 0.01 nm). We propose that the molecules adsorbed at step edges adopt a different orientation of lying at the corner because adsorbates usually interact more

19 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 17. STM images of the second FePc layer (a) and (b) on Au(1 1 1). (a) 50 nm 50 nm, U= 2.2 V, I=0.05 na; (b) 10 nm 10 nm, U= 2.2 V, I=0.05 na. One unit cell and one FePc molecule are marked with a square and a circle, respectively, in (b). (c) STM image (10 nm 10 nm, U= 0.4 V, I=0.05 na) of a single FePc molecule adsorbed on the first layer. (d) STM image (10 nm 10 nm, U= 0.4 V, I=0.05 na) of the first and second FePc layer. The adsorption configuration of FePc molecules on the first layer and the superstructure relation between the first and second layer on Au(1 1 1) are shown in (c) and (d), respectively. strongly with the substrate atoms on a transition metal surface at the kinks and step edges than on flat terraces. It should be pointed out that molecules prefer to adsorb at double step edges instead of single step ones, as shown in the dashed rectangle in Fig. 19d. In general, the step edges on a clean Ag(1 1 0) substrate are along ½110Š and [0 0 1] directions, and extend like an irregular sawtooth to tens of nanometers scale [115]. The reason for the molecular arrangement along the Ag ½20209Š direction at dilute molecular coverage might be that the adsorbed QA16C molecules drive the silver atoms at step edges to align in this specific direction to achieve the most stable configuration. With increasing coverage, molecules tend to aggregate into molecular islands developing from the step edges onto the lower terraces. These islands have highly ordered molecular arrangements with a row-like structure. Fig. 20a is an STM image of one monolayer of QA16C molecules on Ag(1 1 0), showing large-area ordered row-like nanopatterns. The uniform QA16C row-like structures, extending along ½553Š directions toward the step edge, can extend several hundred nanometers depending on the width of terraces. The distance between neighboring rows is 2.91 ± 0.05 nm. Normally, there exists only one symmetrical domain even on a large silver terrace, and the domain covers the terrace compactly with

20 46 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 18. (a) STM image (20 nm 20 nm, U= 2.2 V, I=0.05 na) of two domains in the second layer. (b) STM image (40 nm 40 nm, U= 2.2 V, I=0.05 na) showing the relation between the defects of the first layer and that of the second layer on Au(1 1 1). few defects as shown in Fig. 20a. Fig. 20b shows an infrequent boundary between two symmetrical QA16C domains on the same terrace. This boundary is irregular with some holes due to the mismatch between these two structures. Fig. 20c presents the details of the monolayer row-like structure. The QA16C molecules are displayed in this high-resolution STM image using a ball-and-stick model. It is obvious that the QA16C molecules are organized into rows along the ½553Š direction, and separated by the alkyl chains, which are interdigitated over their full length and aligned with their long axis parallel to the substrate surface. The angle between the QA16C backbone and the Ag ½110Š direction is 21 ± 2. The alkyl chains are almost parallel to each other at an angle of 108 ± 2 with respect to the molecular backbone. High-resolution STM image of alkyl chains, as shown in Fig. 20d, shows further structural details along the interdigitated alkyl chains. The zigzag-shape row of spots for every alkyl chain is visible, and the distance between alternate spots in the row, about 0.26 ± 0.01 nm, is consistent with the calculated value of nm in alkyl chains [176]. This result suggests that the alkyl chain plane lies parallel to the substrate. The Fourier transforms of large number of images of the molecular adlayer in this configuration provide mean values of the unit cell parameters which are a = 1.07 ± 0.03 nm, b = 2.95 ± 0.03 nm, with a rotation angle of 103 ± 2, as shown in Fig. 20c. We propose the following matrix to describe the commensurate superstructure of QA16C molecules on Ag(1 1 0): b 1 b 2 ¼ a1 a 2 Here, a 1 and a 2 are the unit cell vectors of Ag(1 1 0), b 1 and b 2 are the unit cell vectors of the QA16C monolayer superstructure. Fig. 20e is the close-packing model of QA16C on Ag(1 1 0). In this model, the alkyl chains operate as spacers to adjust the inter-molecular distance in the saturated QA16C monolayer. The strong intermolecular interaction is helpful to the stability of the final row-like QA16C structure. Fig. 21a shows a typical STM image of 1.3 ML QA16C molecules on Ag(1 1 0). The lengths of the rows vary from 10 to 30 nm. The observed distances between two neighboring rows are 1.82 ± 0.02 nm, 2.83 ± 0.02 nm, and their integer multiples. The angle between the second layer rows and the Ag ½1 10Š direction is 23 ± 2. For the second layer, the details of the interdigitated alkyl chains can hardly be clearly imaged due to the existence of the first monolayer. Through the dispersed single molecules in Fig. 21a, we can identify the adsorption sites of the molecules in the second layer with

21 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 19. (a) Molecular structure of quinacridone derivative (QA16C). At dilute coverage, the molecular orientations adsorbed on large open silver terraces (b) and at step edges (d) are quite different. (b) On large open silver terraces, two distinct molecular orientations, indicated by black and white circles, respectively, exist exclusively, implying a strong substrate-molecule interaction. Ag ½110Š direction is also indicated by the white arrow. 70 nm 70 nm, U = 1.2 V, I = 0.05 na. (c) High-resolution STM image of single QA16C on silver terrace. 15 nm 15 nm, U = 1.2 V, I = 0.05 na. (d) QA16C molecules adsorb at step edge of Ag(1 1 0). Most of these molecules are observed to align regularly with a length up to several tens of nanometer along the Ag ½20209Š direction. This molecular alignment can be observed only at double step edges while few molecules are adsorbed at single step edges as shown in the dashed rectangle. 120 nm 120 nm, U = 0.8 V, I = 0.2 na. (e) High-resolution STM image of QA16C molecules at silver step edge. 30 nm 30 nm, U = 0.8 V, I = 0.2 na. respect to the first one. The molecular backbones in the second layer are on top of the molecular backbones in the first layer. The interaction between the layers before, the p p stacking interaction enhances the adsorption stability because the additional molecules stack on the conjugated backbones

22 48 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 20. (a) Large scale STM image of QA16C monolayer on Ag(1 1 0). 150 nm 150 nm, U = 2.0 V, I = 0.05 na. (b) Two symmetrical QA16C domains are observed on the same terrace. This boundary is irregular and some holes can be observed due to the mismatch between these two structures. 37 nm 37 nm, U = 1.0 V, I = 0.1 na. (c) High-resolution STM image of QA16C monolayer on Ag(1 1 0) surface. The QA16C molecule is displayed in this image by a ball-and-stick model. QA16C molecules are organized into rows along the ½5 53Š direction, and separated by the alkyl chains, which are interdigitated over their full length and aligned with their long axis parallel to the substrate surface. 10 nm 10 nm, U = 1.0 V, I = 0.21 na. (d) Details of alkyl chains between QA16C molecules suggest that the alkyl chain plane lies parallel to the substrate nm 3.36 nm, U = 1.5 V, I = 0.6 na. (e) The proposed model of QA16C on Ag(1 1 0) surface.

23 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) of nether QA16C. This arrangement is quite different from that in the first layer adsorbed on Ag(1 1 0) surface directly. For monolayer adsorption, the chemical bond between the oxygen and the silver atoms determines the QA16C adsorption sites on the substrate. Taking into account of the relatively large distance between substrate and the second molecular layer, the additional QA16C layer should be less affected by the silver substrate. Fig. 21c presents the details of the QA16C row-like structures. The close-up reveals that the flat-lying molecules interact predominantly via their functional groups. The backbones of QA16C are parallel to each other, with an angle of 74 ± 2 to the Ag ½1 10Š direction, and modulated to keep their distances. Fig. 21d shows the line profile across a QA16C row from P to Q, from which we derive an average distance of 1.15 ± 0.02 nm between neighboring molecules in the rows. Fig. 21b shows the proposed model for the QA16C row-like structure. The adsorption sites of the additional molecules adsorbed on the first molecular layer are determined by the inter-molecular interactions of the p p stacking stabilization between two neighboring layers. In order to avoid the overlapping between QA16C in the rows, the backbones of QA16C in the upper and lower layers are slightly staggered with an angle of about 30, as shown in the lower left inset of Fig. 21b. The interdigitated alkyl chains also appear as spacers to separate the neighboring rows in the same way as they do in the first monolayer. In this model, the distance between the oxygen and the hydrogen of neighboring molecules in the rows is 2.35 Å, as shown in the upper right inset of Fig. 21b. Previous studies revealed that the C HO bond interaction occurs when the distance between the oxygen and the hydrogen is within the range of Å [56]. Therefore, in the second layer a hydrogen bond is Fig. 21. (a) A typical STM image of 1.3 ML QA16C on Ag(1 1 0). QA16C molecules in the second layer form row-like structure. 80 nm 80 nm, U = 1.1 V, I = 0.03 na. (b) Proposed structure. The green and white molecules are in the first and second layer, respectively. Ag atoms under QA16C are not shown in this sketch. (c) Details of QA16C row-like structure in the second layer. 20 nm 20 nm, U = 1.1 V, I = 0.03 na. (d) Line profile across the QA16C row marked by the dashed arrow line from P to Q shown in (c).

24 50 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) formed between two neighboring molecules in the QA16C row-like structures. These hydrogen bonds, as attractive inter-molecular interactions, play an important role in forming the stable structure of rows in the second layer. Besides hydrogen bonds, other interactions in the second layer, such as alkyl alkyl interaction, may also contribute to the stability of the second layer structure since the interdigitated alkyl chains appear. The formation of the second layer is more complex than that of the first layer. In the second layer structure, only a part of molecules have the opportunity for p p stacking to the first layer. At the beginning of the adsorption of additional QA16C molecules onto the first monolayer, the p p stacking affects the initial adsorption site. However, the structure stability in the second layer is mostly improved by intrinsic inter-molecular interactions, such as hydrogen bonds and alkyl alkyl interactions. The second organic molecule layer on noble metal substrate, for larger distance between molecules and substrate atoms, should be less affected by molecule substrate interaction. Consequently, in the second QA16C layer, the hydrogen bonds, alkyl alkyl interaction and p p interaction dominate the self-assembly to form a stable structure. 3. Structural and conductance transitions of single rotaxane molecules After presenting the monitoring, understanding, and controlling the growth of molecules on metal surfaces, let us discuss the physical properties and the potential applications. Rotaxane-based molecules offer significant potential for applications in functional molecular electronic devices [126, ]. Rotaxanes themselves have shown considerable potential for applications both in functional molecular electronic devices [126,177,178, , ] and in nanoscale data recording [81]. Structural bistability of some rotaxane molecules has been observed in the solution phase. These molecules contain a p-electron-deficient ring component that encircles a dumbbell-shaped component incorporating two p-electron-rich recognition sites in its rod section and terminated by bulky stoppers. The ring component, cyclobis(paraquat-p-phenylene) (CBPQT 4+ ), can move back and forth between the two p-electron-rich recognition sites on the dumbbell-shaped backbone in response to external stimuli, resulting in the switching between two stable structures. Structural switching can lead to conductance switching, which makes rotaxane molecules potential candidates for ultra-high density information storage. However, in all practical applications, the rotaxane molecules would have to be in the solid state. The conductance switching of rotaxanes is well defined in the solution phase, but it is disputable in the solid-state phase [179, ]. Our recent studies [191,192] show that the structural bistability does exist for single H2 rotaxane molecules on a buffer-layered solid surface and that the structure transition is accompanied by a conductance switching at 77 K. In addition, STM demonstrated its capability of manipulating intra-molecular structure, measuring local electronic states, and inducing local conductance transition at a single molecule scale [172,193] CH 2 Cl 2 insulating layer A sample of single rotaxane molecules was prepared on a buffer-layered Au(1 1 1) surface for a subsequent STM study. Our experimental objective was to determine whether the interface effect or the inherent property of rotaxane molecules are responsible for the conductance switching in rotaxanebased solid-state devices. In order to investigate the inherent property of rotaxane molecules, the interaction between the rotaxane molecule and the underlying gold substrate should be effectively screened [1,194]. In our experiments, an insulating buffer layer is self-assembled between the molecules and the substrate, which effectively weakens the interaction between the rotaxane molecules and the gold. A small molecule, CH 2 Cl 2, was used as the insulating buffer-layer material because CH 2 Cl 2 molecules self-assemble easily on the Au(1 1 1) surface, and are also a good solvent for rotaxane molecules. In our experiments high self-assembly ability was observed for CH 2 Cl 2 molecules on the Au(1 1 1) surface. The self-assembled CH 2 Cl 2 monolayer was prepared by dropping a CH 2 Cl 2 solvent onto the clean Au(1 1 1) surface in a nitrogen environment at room temperature. Subsequently, the sample was annealed at 330 K in UHV. STM observations revealed high-quality self-assembled monolayer (SAMs), as shown in Fig. 22a. Auger electron spectroscopy measurements on the SAM

25 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) showed the signature of chlorine (Cl) [195,196], but no signals from oxygen (O) and nitrogen (N), see Fig. 22c and d, which confirms that the observed self-assembled structures are composed of CH 2 Cl 2 molecules. Fig. 22a shows that the CH 2 Cl 2 molecules are arranged into stripes. To prepare the sample of isolated H2 rotaxane molecules on a surface, we first dissolved the rotaxane molecules in CH 2 Cl 2 solution, with a molarity of mol/l, then placed several drops of the rotaxane/ch 2 Cl 2 solution onto a clean Au(1 1 1) surface which was prepared by using standard UHV methods [197]. To check whether the rotaxane molecules were isolated from the gold substrate by the SAM using this preparation method, voltage pulses were used to remove the rotaxane molecules from their adsorption sites. Subsequent STM images of the same area showed that there was no defect in the SAM at the original molecular adsorption sites, which confirms that the rotaxane molecules were actually adsorbed on the SAM STM measurements STM observations show that isolated rotaxane molecules are dispersed on the self-assembled CH 2 Cl 2 monolayer (see Fig. 22b). Various molecular conformations have been observed in our STM experiments. In Fig. 22b, molecules 1 and 2 have the same conformation, while molecule 3 has a different one. We observed in total 25 different conformations for the rotaxane molecules (see Fig. 23a). According to our STM experiments, if the distance between two adjacent molecules is less than 5 nm, Fig. 22. (a) STM image (20 nm 20 nm) of a CH 2 Cl 2 self-assembled monolayer after annealing at 85 C. Bias voltage U = 1.3 V, current I = 0.08 na. (b) STM image (30 nm 30 nm) of rotaxane molecules on the CH 2 Cl 2 self-assembled monolayer. Scanning parameters: U = 1.2 V, I = 0.03 na. (c) Auger electron spectroscopy ( ev) of a CH 2 Cl 2 /Au(1 1 1) sample. Carbon (C) was detected. Oxygen (O) and nitrogen (N) were not detected. (d) Auger electron spectroscopy ( ev) of a CH 2 Cl 2 /Au(1 1 1) sample. Cl was detected.

26 52 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) the two molecules tend to adopt the same conformation. One may therefore conclude that the conformation of adjacent molecules is influenced only at small inter-molecular spacing Molecular manipulation by STM We also conducted scanning tunneling spectroscopy measurements on the adsorbed rotaxane molecules. As STM is capable of high-spatial-resolution measurements, we collected I V data at several points on a single rotaxane molecule (see Fig. 23b). Our measurements show that there is little difference between the I V curves across the molecule, indicating that electron tunneling is occurring via the whole molecule. As the molecular conformation may have an influence on the I V characteristics [198], we also performed I V measurements on H2 molecules with different conformations. Numerous measurements showed that the molecular conformation has no obvious effect on the I V curves. For the as-deposited rotaxane molecules a zero-conductance gap of 1.8 ev is measured. A structure/conductance transition in a single rotaxane molecule was observed by varying the polarity of the voltage bias applied between tip and sample in the STM experiments. Fig. 24a d (together with the corresponding measured I V curves (see Fig. 24e)) shows sequential STM images where the polarity of the bias voltage has been reversed between subsequent scans. The change of the bias voltage from 1.3 V to +1.3 V led to changes in both the molecular STM image and the I V curve. When the bias voltage changed from +1.3 V back to 1.3 V, both, the molecular STM image and the I V curve revert. This process was reversible and repeatable in our experiments. The above results, however, are not sufficient to put forward the argument that any structural transition induces a conductance switching for single rotaxane molecules. Firstly, the observed change in the molecular STM images may have been induced by the difference between the spatial distribution of the LUMO and HOMO orbitals. Secondly, the difference in I V curves may have been induced by the difference in the tip-sample distance under different set points for open feedback loop measurements [ ]. Fig. 23. (a) STM images showing 25 types of molecular conformations for individual rotaxane molecules in our STM observation. Scanning parameters: U = 1.3 V, I = 0.03 na. Image size: 10 nm 10 nm for 1 17; 8 nm 8 nm for 18 and 19; 7 nm 7 nm for 20; 6 nm 6 nm for 21 23; 5 nm 5 nm for 24; 4 nm 4 nm for 25. The molecular apparent size varies considerably with the molecular conformation. (b) I V curves taken at different locations of a single rotaxane molecule are roughly the same. Scale bar is 2 nm. Both, the STM image and I V curves are recorded at 77 K.

27 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 24. (a d) Reversible conductance transitions induced sequentially in an individual rotaxane molecule by the STM field. STM images show different but consistent forms on sequentially reversing the bias polarity. The scale bar is 1 nm. (e) I V characteristics. (f) STM images of twelve different molecules. (g) The time intervals between the polarity change and structure/ conductance switching for 1 8 molecules in (f). Magenta represents the time interval between polarity change from 1.3 V to +1.3 V and the structure/conductance switching. Olive represents the time interval between polarity change from +1.3 V to 1.3 V and the structure/conductance switching. Line represents switching occurred at a moment during that time range. Dot represents switching occurred at that moment. (h) The sketch map for the statistics in (g). Four cases (#1, #2, #3, and #4) are listed. When t=0, the polarity of the scanning voltage is reversed. For the case of #1, the transition moment lies between 0 and T 1 ; For the case of #2, the transition moment is at t=t 2 ; For the case of #3, the transition moment lies between T 3 and T 4 ; For the case of #4, the transition moment is at t=t 5. The sequential STM images shown in Fig. 25a e, however, exclude these possibilities. Fig. 25a and b show that when the bias voltage changes from 1.3 V to +1.3 V, the molecular image and the I V curve are not immediately changed. The subsequent STM image (Fig. 25c) captures the instant of change for the molecular image. Fig. 25d shows that both, the structure and the I V curves have changed. The analysis of the process depicted in Fig. 25a d leads to two conclusions. First, the change of the molecular STM image is induced by the structural change and not by the difference between LUMO and HOMO imaging. Second, the change in I V curves represents the conductance switching, and is not an experimental artifact. Finally, we changed the bias voltage back to the initial negative polarity, and the molecular image returned precisely to its initial state (Fig. 25e), with the I V curves corresponding to the initial low-conductivity state. The observed correlation between structural change and conductance change is reversible and repeatable. This suggests that the conductance transition is induced by the intra-molecular structural transition. In our STM experiments, not all the molecules on the surface showed a structure/conductance transition. At present, we are not certain whether the conformation is responsible for this behavior, because of the following two facts. First, it is quite difficult to identify the precise conformation of the molecules on the surface based on the STM images alone. Second, as the rotaxane molecules were dispersed on the substrate, interaction between the molecules and the substrate might have influenced the switching properties of the molecules. In our STM observations, we have observed an ordered layer of the solvent molecule (CH 2 Cl 2 ) co-existing with the rotaxane molecules on the Au substrate. This ordered layer isolates the molecules from the Au substrate and decreases the interaction between the rotaxanes and the Au. However, experimentally, different molecular conformations generally correspond to different molecular appearances in the STM images, and we show some experimental data in Fig. 24f. We performed transition experiments on twelve molecules, of which eight (1 8 in Fig. 24f) showed a structure/conductance transition, while four (9 12 in Fig. 24f) did not. Molecules 1 and 3 have the same molecular conformation, as do molecules 6 and 11. According to these experimental results, it is not safe to say that conductance switching can always be observed for one molecular conformation and never for another. We can only say that the molecular conformation influences the probability for conductance switching.

28 54 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 25. A sequence of STM images of a single rotaxane molecule showing a reversible conductance transition correlated with a change in molecular structure. (a, b) STM images obtained with changed bias polarity show a similar image, indicating a similar HOMO and LUMO distribution. (c) As the tip approaches the molecule, a sudden structural transition induces the molecule to jump across the substrate. (d) The molecule stays in this new position for the next scan. (e) It recovers its original structure perfectly when the voltage bias is set to its original polarity. The green lines and circles mark features on the substrate for reference; the scale bar size is 1 nm. I V characteristics show that the conformation change is associated with a transition to a high-conductance state. Numbers on the images represent positions of the I V measurements. In our STM experiments, the structure/conductance switching was realized by switching the polarity of the tip-sample bias voltage between 1.3 V and +1.3 V. The switching is a sudden incident, as shown in Fig. 24c. In some cases, for example, #2 and #4 in Fig. 24h, we captured the switching moment because the switching occurred while the STM tip was scanning over the molecule, also seen in Fig. 25c. In this case, we know exactly the time interval between the polarity change and the structure/ conductance switching. However, in the other cases, #1 and #3 in Fig. 24h, we did not capture the switching moment because the switching occurred while the STM tip was scanning over the SAM surface rather than over the molecules. In this case, we can only give a range for the time interval between the polarity change and the structure/conductance switching. Fig. 24g shows statistics for the above time interval for molecules 1 8 in Fig. 24f Understanding the geometry of molecules by DFT and MM calculations Stoddart and co-workers pointed out that the conductivity of rotaxane molecules is mainly determined by the CBPQT +, tetrathiafulvalene (TTF) and 1,5-dioxynaphthalene (DNP) groups, and that the electron tunneling occurs via the whole molecule if these groups are crowded together [ ]. I V measurements taken at different points of the molecule, as shown in Fig. 23b, indicate that the electrons tunnel through the whole molecule. Therefore, the CBPQT 4+, TTF and DNP groups should be crowded together. It was also predicted that the conductance switching is induced by the movement of the ring component CBPQT 4+ between DNP and TTF recognition sites, when activated by an external stimulus. In STM experiments, the external stimulus is the inelastic electron tunneling process transferring the energy of inelastic tunneling electrons to the ring component. There are two factors that influence the structure/conductance transition of single rotaxane molecules: the underlying CH 2 Cl 2 insulating layer, and the molecular conformation. On one hand, the CH 2 Cl 2 buffer layer effectively weakens the coupling between the Au substrate and the rotaxane molecules. Experiments on rotaxane

29 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) molecules directly adsorbed on the Au substrate all failed to realize a structure/conductance transition, which is ascribed to the strong bonding between the Au substrate and the molecules (e.g., the S atoms in the rotaxane molecule). Hence, the existence of the CH 2 Cl 2 buffer layer is a prerequisite for conductance switching. On the other hand, not all rotaxane molecules isolated by the buffer layer show structure/conductance transitions, which means that molecular conformation is another crucial factor. Although 25 different molecular conformations have been observed in our STM experiments, none of them is a fully extended structure, as shown in Fig. 26a. To identify the molecular conformation of a free rotaxane molecule, we carried out fully optimized DFT calculations and classical forcefield calculations. Owing to numerical limitations and the size of the molecule, we initially built two simplified models by removing the two large stoppers and the four PF 6 anions from the rotaxane molecule see 1 and 2 in Fig. 26b. Structure optimization was performed using the Gaussian 03 program package at the DFT-B3LYP(3-21G) level for the two models [ ]. Whether the CBPQT 4+ is at the a CH 2 O O CH 2 O CH 2 O O CH 2 O stopper 4PF 6 - N N + + S S CH 2 S SCH 2 S S + TTF + N N CBPQT 4+ CH2 O DNP OCH 2 stopper OCH 2 O OCH 2 OCH O 2 OCH 2 b 1 2 H (backbone) C (backbone) O S H (CBPQT) C (CBPQT) F P N 3 Fig. 26. (a) Structure model of the rotaxane molecule. The end parts enclosed by dashed line are the stoppers preventing the CBPQT 4+ disengaging from the molecular backbone. The CBPQT 4+ is colored in blue. The tetrathiafulvalene (TTF) and 1,5- dioxynaphthalene (DNP) are colored in red. (b) Fully optimized structure of a simplified model for molecule with the CBPQT 4+ ring positioned over the TTF group (1) and over the DNP group (2). 3 is the result of molecular mechanics calculation of the entire molecule. All calculations confirm a curved conformation for the rotaxane molecule. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

30 56 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) TTF or DNP sites, the simplified backbone is not linear but curved; see 1 and 2 in Fig. 26b. In addition, structure optimization for the entire rotaxane molecule was carried out using molecular mechanics with an MM+ force field [166]. This result also showed a curved backbone (see 3 in Fig. 26b). These curved structures indicate that, for rotaxane molecules, the backbone cannot form a linear molecule because of the two stoppers and the CBPQT 4+ component. The curved structure provides more space for CBPQT 4+ to move between TTF and DNP sites. In addition, the calculation results explain the ellipsoidal appearance of the rotaxane molecules on CH 2 Cl 2 /Au(1 1 1), and also the reason why the rotaxane molecule adopts so many different conformations. Rotaxane is a large, flexible molecule that can adopt many possible molecular conformations, some of which facilitate the movement of the CBPQT 4+ in the case of individual rotaxane molecules on a solid surface. The molecular conformation influences the probability for conductance switching. The following part will review the conductance and nanorecording using various systems with different mechanisms, which provides a wider range of possibility for potential applications of these kinds of nanostructures. 4. Conductance transition and ultra-high density data storage on molecular thin films Ultra-high density data storage is critical for the future development of information technology [74 81]. To this end, nanometer-scale recording on organic thin films by STM through a conductance transition is a promising strategy [82 92]. In order to improve the performance of data storage, it is extremely important to explore a suitable mechanism for a conductance transition towards practical applications. In this section, we review several typical mechanisms for conductance transitions in the nanometer-scale recording on organic thin films by STM Conductance transition induced by phase transition We have achieved a reversible conductance transition in a crystalline organic complex on a scale close to the dimensions of the unit cell [87]. The two conjugated organic compounds shown in Fig. 27a 3-nitrobenzal malonitrile (NBMN) and 1,4-phenylenediamine (pda), were used to form the complex. NBMN was prepared following recopies in the literature [207], and the two materials were mixed together with a 1:1 molar ratio and were subsequently vacuum evaporated. High-resistivity Fig. 27. Molecular structures of (a) NBMN and pda, (b) ( )-TADDOL and coumarin, (c) DMNBPDA, (d) TDMEE and (e) CDHAB.

31 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) polycrystalline NBMN-pDA films were prepared by thermal evaporation onto substrates at room temperature at a rate of 65 nm/min. Films for conductance transition experiments were deposited to a thickness of 20 nm on a freshly cleaved HOPG substrate. Samples for macroscopic four-probe I V measurements were deposited to a thickness of 200 nm on substrates of glass coated with conductive indium tin oxide (ITO) films. For TEM, films were simultaneously deposited onto carbon films and subsequently examined in a JEOL 200CX. The coverage was found to be very uniform. X-ray diffraction shows that the polycrystalline films have a triclinic structure. In contrast, for deposition rates >6 nm/ min, the resulting films were conductive and structural characterization shows that these more rapidly deposited films are amorphous. Standard four-probe measurements of I V characteristics on the 200 nm thick films, shown in Fig. 28a, demonstrate electrical bistability. Increasing the voltage stepwise from zero produced the open points in Fig. 28a, but when the voltage was increased above 3.2 V, the film abruptly switched to a conductive state. In the conductive state, stepwise sampling of the current as a function of the applied voltage produced the solid points in Fig. 28a. The resistivity of the high-impedance state is 10 8 X cm, and the one of the conductive state is 10 4 X cm. The transition time from the highimpedance state to the conductive state was measured to be 80 ns, as shown in Fig. 28b. Local conductance transitions were induced in the high-resistivity films by STM. A typical image of the film is shown in Fig. 29a. As shown in Fig. 29b and c, a 3 3 array and an A pattern could be formed by applying voltage pulses to the STM tip in constant height mode. Each bright spot in the image represents a high-conductance region, i.e., regions that have been switched from insulator to conductor. The shadow effect on each mark is due to the fact that the feedback circuits were not completely set to zero during the scan. All marks were erased on heating the sample in situ above 423 K, when the conductance of the local regions recovered their original insulating states. It was possible to induce the reverse transition in individual marks by applying a reverse-polarity voltage pulse of 4.5 V for 50 ms, as shown in Fig. 29d and e. If the same pulse was applied without reversing the polarity, the mark was incompletely erased. Therefore the process is due to the combined effect of applied field and local heating induced by the current passing through the conducting region. In principle, the ultimate resolution should be the size of the molecular complex, 1 nm. Fig. 29f demonstrates the resolution on this length scale. The marks are separated by approximately 1.7 nm. Local I V characteristics of the film measured before and after mark formation confirmed the conductance transition. As shown in curve a of Fig. 30, with the NBMN-pDA film in its initial high-impedance state, the applied voltage can be varied from 0 to 2.1 V but the tunneling current remains small (about 0.2 na). Beyond a threshold of 2.1 V, the film changes to a conductive state with the I V relation of curve b. Curve b clearly demonstrates conductivity and eliminates deposited charge as a possible mechanism of the mark formation. Curve c is the I V curve for the graphite substrate, which is inconsistent with the one of the film in its conductive state. This observation excludes the possibility that the conducting state is reached by simply burning a hole through the film. Fig. 28. (a) I V relation of a 200 nm film showing the low and high-conductance states. The voltage threshold is 3.2 V. (b) Transient conductance measurements on the 200 nm film showing a transition time of 80 ns.

32 58 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 29. STM images of the NBMN-pDA film on HOPG. (a) An image of the surface of the film showing crystalline order, image size: 6 nm 6 nm; (b) a 3 nm 3 nm array formed by voltage pulses of 4 V and 1 ms width; (c) an A pattern formed by voltage pulses of 3.5 V and 2 ms; (d) and (e) STM images after erasing marks one at a time with reverse-polarity voltage pulses of 4.5 V and 50 ms; (f) resolution test using voltage pulses of 4.2 V and 10 ms. The distance between neighboring marks is 1.7 nm. Scan conditions are U=0.1 V, I=0.4 na for (a); U=0.19 V, I=0.19 na for (b) (f); constant height mode. Several possible mechanisms for the conductance transition were considered resulting in the identification of a candidate mechanism that is in agreement with all available experimental data for this system. The first candidate considered was the charge transfer mechanism [208]. In the charge transfer mechanism, hypothetically, a voltage pulse above some threshold brings about a shift in electron

33 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 30. Typical STM current voltage relations. Curve a: Before the voltage pulse, with the film in a highly insulating state, which becomes conductive above a voltage threshold of 2.1 V; curve b: I V relation after the voltage pulse, indicating a transition to the conductive state; curve c: linear I V relation from the HOPG substrate. density into a charge-separated state. While such an excited electronic state could presumably impart conductance properties to the molecule different from those of the ground state, in the present case, the charge transfer mechanism is excluded by the transition time measurements. The transition time from the initial high-impedance state to the conductive state is 80 ns, many orders of magnitude longer than even the most sluggish molecular electronic transitions. Furthermore, it is highly unlikely that an appreciable population could be sustained in the excited state (charge transfer) for the hours to days that the film remains in the conductive state [208]. The second possible mechanism is that the individual molecular complexes of NBMN-pDA might undergo an internal rotation into an excited molecular conformation whose geometry completes the conjugated p structure across the entire complex. This is essentially the conformational switching mechanism of Joachim and Launay [209]. In a molecule containing two or more benzene rings in sequence, the electrical resistance increases by about 10 4 when the angle between the planes of adjacent rings is changed from 0 to 90 [210]. In the present case, the individual components contain extended p-conjugated structures. If these components were brought into alignment in an excited conformation, completing the p structure across the entire complex, it would result in greatly enhanced conduction in the excited conformation. The present system proved to be too large to test the conformational isomerization hypothesis computationally, but we were able to carry out electronic structure calculations on a closely related system that exhibits the same electrical bistability behavior. Local conductance transitions have been demonstrated on the nanometer scale in NBPDA [208]. The structure of NBPDA is shown in Fig. 31(inset). We searched for stable conformations of this molecule at the Hartree Fock/SCF level of theory. As shown in Fig. 31, NBPDA exhibits two stable conformers. The predicted harmonic vibrational frequencies for the more stable of these two conformers reproduce the dominant peaks and features of the FTIR spectrum of crystalline NBPDA very well. In both conformers, however, the two benzene rings are coplanar. Since the p p conjugation extends over the entire NBPDA system in both conformers, we conclude that local conductance transformations observed in NBPDA do not arise from conformational isomerization, and by analogy it seems highly unlikely that such a mechanism would be at the root of the same phenomena in the very closely related NBMN-pDA system. The most likely source of the local conductance transition in NBMN-pDA thin films is a mechanism whereby the applied electric pulse leads to a reorientation of one or more entire molecules, in effect introducing local disorder into the crystalline thin film. The experiments demonstrate that the crystalline films are insulating and the amorphous films are relatively conducting. This result is less surprising when one notes that this is an organic system, and carbon, while an excellent insulator when in the form of diamond, is much more conducting in various less-ordered forms. It follows naturally that if a local area of the film is structurally disordered, that region will become conducting. Since the NBMN-

34 60 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 31. The molecular structure of NBPDA (inset) and potential energy curves for internal rotations of NBPDA. Solid circles: rotation about the (C (3), N (1), C (4), C (5) ) dihedral angle. Open circles: rotation about the (C (1), C (2), C (3), C (5) ) dihedral angle. pda system has a permanent dipole, an electric field pulse could lead to molecular reorientation. Such a reorientation would effectively introduce local disorder into the film. This induced local disorder hypothesis is also consistent with the mechanism for erasure. Only reversed polarity pulses completely reverse the conductance transition, and they must be applied for a longer duration than the write pulse. Obviously, if the write pulse acting on the permanent dipole twists the molecule in one direction, reversing the twist will require the application of force of the opposite sign. Furthermore, the duration must be sufficient to allow time for the order to redevelop. Finally, the 80 ns characteristic transition time for the conductance transition, although much too slow for an electronic or conformational transition, is perfectly reasonable for reorientation of a large molecule. To test the hypothesis that the conductive region of the film is disordered, TEM and electron-diffraction studies were carried out on a film before and after switching. The results are shown in Fig. 32. As anticipated, the high-impedance film is crystalline, but after switching to the conductive state, the TEM and diffraction studies indicate an amorphous film Conductance transition induced by rupture of hydrogen bonds The rupture of hydrogen bonds within supramolecules is responsible for the conductance transition in the ultra-high density data recording using a new supramolecular structure of ( )-TADDOL (2,2-dimethyl-a,a,a 0,a 0,-tetraphenyldioxolane-4,5-dimethanol) and coumarin (see Fig. 27b) [89]. A highly-ordered crystalline TADDOL-coumarin (TC) thin film was self-assembled via hydrogen bonding and p-stacking among ( )-TADDOL and coumarin molecules. We successfully realized ultra-high density data storage in the crystalline thin film using STM, achieving a data-storage density of about bits/cm 2. In order to obtain a high-quality film, it is crucial to select appropriate components and control the growth rate. Bulk TC crystals are available according to reported procedures [211]. The supramolecular crystals as a whole were 1:1 inclusion complexes of ( )-TADDOL with coumarin, which contained their respective hydrogen-bonding recognition subunits, i.e., hydroxyl and carbonyl group. To prepare films for data storage, TC crystalline complexes were redissolved in a methylbenzene-methylene dichloride mixture (5:1) to form a solution. The drop-casting method was used for preparing films on freshly cleaved HOPG substrates. The thickness of the film could be modulated by adjusting the

35 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 32. Structural characterization of NBMN-pDA films. Before switching (nonconducting): (a) TEM and (b) diffraction. After switching (conducting): (c) TEM and (d) diffraction. concentration of the solution. The TEM image shows a uniform and rather flat surface, and the electron-diffraction pattern reveals a highly crystalline nature of the thin film (see Fig. 36a), when a thin film was prepared with a solution concentration of M. Such a uniform smooth thin film with a suitably large single-crystallized region is the basis for practical technological applications. Fig. 33a shows a typical STM image of the thin film. Using the same tip, an atomic image of the HOPG substrate surface can be observed (Fig. 33a, inset). From Fig. 33a, it is clear that TC molecules are arranged periodically. The molecular plane of coumarin is almost parallel to the HOPG substrate surface due to the strong van der Waals force resulting from the enforced flat adsorption geometry between them; the carbonyl groups of coumarin interact with the ( )-TADDOL molecule by hydrogen bonding, at the same time the face-to-face p p interactions of aromatic rings occur between coumarin-coumarin and TADDOL TADDOL molecules, as shown in Fig. 33b. The cooperative effects between them lead to the highly ordered structure of the thin film on the HOPG substrate surface. Recording experiments were performed with an STM (Solver P-47 STM, NT-MDT) under ambient conditions using electro-chemically etched tungsten tips. In our experiments, the STM tip was very stable, and the images of HOPG with atomic-resolution were obtained before and after the recording experiments. To induce the recording dots, pulsed voltages were applied between the STM tip and the thin film. Fig. 34a shows a typical STM image of the recorded pattern on the TC thin film. The recording marks illustrate a 3 4 matrix pattern in a scanning area of 70 nm 70 nm. The size of the recording dots is about 2.2 nm in diameter. The recording dots were formed by applying a program-controlled

36 62 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 33. (a) A typical STM image of TC crystalline thin film. Scanning parameters: V bias = 0.36 V, I ref = 0.32 na; the inset shows the image of HOPG. (b) The stacking arrangement of ( )-TADDOL (the green structures) and coumarin molecules (the blue structures) on HOPG substrate in the crystalline film. Dotted lines represent the stabilizing hydrogen bonds between the molecules. From the model, the hydrogen bonding interactions between ( )-TADDOL and coumarin molecules, p p interactions of aromatic rings between coumarin-coumarin and TADDOL-TADDOL molecules can be clearly observed. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) pulsed voltage of 3.4 V and 2.3 ms duration. During the scanning period, the recorded marks were very stable and could last more than 24 h. In principle, the ultimate resolution should be the size of the molecular complex, i.e. about 1.0 nm. Fig. 34b demonstrates a resolution on this length scale. It is shown that the marks could be written very close to each other (the marks are separated by approximately 1.0 nm) without merging, implying a potential recording density of about bits/cm 2. From the experiments with different pulse conditions, it was confirmed that the dot size strongly depended on the amplitude and width of the applied pulsed voltage. For voltage pulses above the threshold voltage, the probability for the dot formation was almost 100%. Moreover, the dot sizes became larger with increasing amplitude of the applied pulsed voltage from 3.2 V to 6.2 V. When the value of the voltage was above 6.2 V, a hole could be clearly seen on the thin film surface. Fourier-transform IR (FTIR) spectra of a bulk TC crystal were compared with the ones from the crystal after the action of an electric field using a Keithley 4200 Semiconductor Characterization System. The FTIR results under different voltage conditions are shown in Fig. 35. The hydrogen bonds would break off when the value of the voltage reaches a certain threshold value U 1 (9.6 V), leading to a blue-shift of the characteristic IR absorption t C@O from 1700 cm 1 to 1724 cm 1, as shown in Fig. 35b [212]. By raising the voltage continuously to a higher value U 2 (12.8 V), the t C@O absorption at 1700 cm 1 vanished (see Fig. 35c), which could be attributed to the evaporation of the coumarin component from the TC crystalline complexes. On the basis of the above FTIR results, we propose the following mechanism for the information dot formation: when a pulsed voltage (>3.2 V) is applied to a TC thin crystalline film, H-bonds between ( )-TADDOL and coumarin molecules are broken, leading to a local cracking of the crystal. The latter was confirmed by electron diffraction measurement. The EDP varied from a clear spot pattern to a diffraction ring, as shown in Fig. 36. When the pulsed voltage reached a higher value (>6.2 V), the coumarin component was vaporized away, leading to the formation of the holes Conductance transition induced by inter-molecular charge transfer Charge transfer is an important physicochemical process and the interactions between electron donors and acceptors play a fundamental role in crystal design [213,214], synthetic processes [215], and even potential applications in molecular devices [216]. Metal/tetracyano-p-quinodimethane (TCNQ)

37 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 34. (a) STM image of a recorded pattern obtained by applying voltage pulses of 3.4 V and 2.3 ms. The scanning area is 70 nm 70 nm. The tunneling conditions are: V bias = 0.42 V, I ref = 0.62 na. Scan mode: constant current. The average size of marks is about 2.2 nm in diameter. (b) Resolution test using voltage pulses of 3.2 V and 5.0 ms. The space between two information dots is about 1.0 nm. complexes [217,218] are one of the typical materials for data recording via charge transfer, in which metals are used as electron donors and organic molecules are used as electron acceptors. Experimental and theoretical studies suggested that the efficiency of inter-molecular charge transfer varies substantially in molecular assemblies with different packing modes [215,216, ]. The problems with these organic metal complexes are that their properties are locally inhomogeneous and that it is difficult to produce well-dispersed films from them. In most cases, the film adopts polycrystalline or even amorphous states and defects exist. In contrast to organic metal complexes, however, all-organic films possess uniform physical chemical properties. So far, we have successfully designed and synthesized a series of organic complexes, which are stable at ambient temperatures and pressures and have produced organic thin films. We have tested and confirmed with STM the validation of these organic thin films in nanometer scale data storage [85,87,222,223] N,N 0 -dimethyl-n 0 (3-nitrobenzylidene)-p-phenylene-diamine (DMNBPDA) [88] The DMNBPDA molecule (see Fig. 27c) has a strong electron donor, N(CH 3 ) 2, and an electron acceptor, NO 2. It is more stable than NBPDA. The thin film was deposited on freshly cleaved HOPG substrates by a vacuum deposition method. In the STM measurements, the film showed good bistability. By applying properly pulsed voltages between STM tip and substrate, we succeeded in writing a letter y on the DMNBPDA film (Fig. 37a). The average diameter of the marks is 1.1 nm, and the distance between two dots can reach 1.5 nm. The corresponding data-storage density is more than bits/cm 2. I V curves show that the conductance transition occurs after a voltage pulse is applied to the film (Fig. 37b). I V experimental results are consistent with our previous interpretation that the recording mechanism was due to charge transfer in the recorded regions [208,224]. The coexistence of a strong electron acceptor, NO 2 and a donor, N(CH 3 ) 2 in the molecule suggests that electron delocalization may be induced by charge transfer in the recorded region in the thin film. Generally, charge transfer includes both intra-molecular and inter-molecular transfer. In our case, however, the former approach could hardly be responsible because intra-molecular charge transfer is unstable

38 64 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 35. The FTIR spectra of a bulk of TC crystal before and after adding voltage: (a) FTIR of the TC crystal (b) after adding voltage U 1 (9.6 V) (c) after adding voltage U 2 (12.8 V). Fig. 36. The process of electron bombardment on a TC crystalline film. (a) Initial electron-diffraction pattern, indicating highly crystalline film. (b) After bombarding with electrons, the pattern was replaced by a diffraction ring as a result of electronbombardment induced rupturing of hydrogen bonds between ( )-TADDOL and coumarin molecules. while the recording marks are observed to be extremely stable, with no apparent change during 10 h of scanning. We therefore feel justified to attribute the recording mechanism to inter-molecular charge transfer. To confirm the mechanism, the UV vis spectrum of the DMNBPDA thin film was compared to the one of the film after the action of an electric field using a standard four-probe method with a threshold voltage of 8.2 V, see Fig. 38. The spectra shown in Fig. 38a and b exhibit a band with k max at 397 nm and 399 nm, corresponding to the absorption of DMNBPDA molecules without interaction among molecules, while the spectrum after application of the electrical field exhibits two additional bands at 495 nm and 629 nm besides the peak at 399 nm (Fig. 38b). The new bands are attributed to the intensive interaction among molecules, especially inter-molecular charge transfer [225,226]. Considering the special structure of DMNBPDA, i.e. the strong electron acceptor and electron donor in the molecule, the interaction is most likely due to the inter-molecular charge transfer.

39 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) a b 10 Current(nA) II I Voltage (V) Fig. 37. (a) A y pattern formed by nine dots on the DMNBPDA thin film. STM worked in constant height mode. U=0.36 V, I=0.21 na. The average size of marks in diameter is 1.1 nm; (b) Typical I V curves of the unrecorded and recorded regions in the DMNBPDA thin film. Curve I is related to the recorded region and curve II is related to the unrecorded region. a 397 abs b abs Wavelength (nm) o Experiment data Gauss fit of data Gauss fit peak 1 for data Gauss fit peak 2 for data Gauss fit peak 3 for data Gauss fit peak 3 for data Wavelength (nm) Fig. 38. The UV vis spectra of the DMNBPDA thin film (a) before and (b) after application of an electrical field.

40 66 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) The charge-density distributions of DMNBPDA were calculated using the hybrid Hartree Fock DFT molecular method of B3LYP [205,206] with a 6-31G * basis set [227]. The result is shown in Fig. 39a, which indicates that electron transfer may take place in two nearest-neighboring molecular systems from N(CH 3 ) 2 to NO 2 when induced by an electric field. The result also shows that the HOMO and LUMO are symmetric, as shown in Fig. 39b and c. This result demonstrates the possibility that the HOMO of one molecule may interact with the LUMO of its neighboring molecule for charge transfer. Thus the recording mechanism can be explained as follows: when a crucial voltage pulse is applied to the film, charge is transferred from N(CH 3 ) 2 of one molecule to NO 2 of another molecule, i.e. the p electron is delocalized, leading to a relative resistance decrease of the system. The analysis of the thin film is helpful to further understand the recording mechanism. Atomic force microscopy (AFM) observations, shown in Fig. 40a and b, reveal that the top surfaces of the films are very smooth exhibiting only a few steps. From the analysis of the topography and the phase image, it is clear that the topographically higher areas (bright areas in phase image) reflect the DMNBPDA Fig. 39. (a) Charge-density distributions of DMNBPDA (the solid circle is positive charge, and the hollow circle is representative of negative charge); (b) HOMO and (c) LUMO of DMNBPDA.

41 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) thin film and the relatively lower topographic locations (dark areas in phase image) correspond to HOPG. The line profile analysis reveals the step height to be about 1.4 nm (see Fig. 40a and b), almost coincident with the full molecular length of DMNBPDA (1.36 nm, shown in Fig. 40c). This observation suggests that the organic film has a thickness of a monolayer and that the DMNPBDA molecules may stand perpendicular to the HOPG surface plane. Fig. 40d shows a typical STM image at molecular resolution. Using the same tip, an image of an HOPG sample at atomic-resolution can be observed easily. Based on Fig. 40d, we find that the molecules of DMNBPDA are arranged periodically, and that the dimension of the periodic unit is 0.43 nm 0.54 nm. Considering the inter-molecular charge transfer, we think that the molecules are arranged inversely. Fig. 40e shows the model for the molecular arrangement in the film. The orderly arrangements of the organic molecules may arise from various factors such as the nature of the organic molecules, the electrostatic interaction between the neighboring molecules, and the influence of the substrate. In fact, the orderly arrangement of a DMNBPDA thin film is very important for data storage. In a few experiments, it was troublesome to write data on the thin films under the same conditions, and it was also difficult to obtain a molecular image using STM. This is because, as we speculate, DMNBPDA molecules are arranged irregularly in those films. It is therefore more difficult for the inter-molecular charge transfer to occur when the pulse voltage is applied ,1,2-Tricyano-2-[(4-dimethylaminophenyl)ethynyl]ethene (TDMEE) [228] TDMEE is a highly conjugated molecule, which features a strong electron-donating N(CH 3 ) 2 group and a strong electron-accepting tricyanomonoethynylethene core in its rigid planar structure, as shown in Fig. 27d. It is very stable and can be sublimed without decomposition (100 C, 0.1 mbar) Fig. 40. (a) AFM image of a DMNBPDA thin film deposited on the surface of HOPG; (b) Profile of the white line in (a) showing a step height of about 1.4 nm; (c) The length of a single molecule of DMNBPDA. (d) A typical STM image of DMNBPDA thin film, U=0.42 V, I=0.67 na; (e) Model for the arrangement of DMNBPDA on HOPG.

42 68 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) [229]. Most importantly, TDMEE molecules undergo p-stacking in the crystal, thereby exhibiting a favorable antiparallel dipolar alignment [230]. By using vacuum deposition, highly ordered, crystalline TDMEE thin films for nanoscale data storage were successfully assembled by taking advantage of inter-molecular p p and donor acceptor interactions [228]. TDMEE was synthesized according to a protocol in the literature [229]. The thin films for nanoscale data storage were deposited on a freshly cleaved HOPG substrate using vacuum deposition. Samples for TEM observation were deposited on a Cu grid coated with an amorphous carbon supporting film. Fig. 41a shows a TEM image of the TDMEE thin film and its representative electron-diffraction pattern. It indicates that the as-prepared thin film has a large-scale uniform and rather flat surface with an extended single-crystal region, which would be favorable for nanoscale data recording. Fig. 41b illustrates the packing mode of the TDMEE molecules in the crystal structure [230]. As shown in Fig. 41b, the assembly adopts a regular alternate donor acceptor stack, offered by cofacially orientated TDMEE molecules, with N,N-dimethylanilino (DMA) donor and cyanoethynylethene (CEE) acceptor moieties placed at a distance of 3.44 Å. Such a close antiparallel dipolar alignment in the stacks represents an ideal arrangement for field-induced inter-molecular charge transfer between molecules in the neighboring layers, due to strong inter-molecular interactions and high electronic coupling [215,220]. This is further verified by subsequent recording experiments and IR spectra. Recording experiments were performed with a NT-MDT Solver P-47 STM in constant height mode using tips made of tungsten wires by electrochemical etching. Fig. 42a presents a typical STM image of recorded marks on a TDMEE thin film. As shown in Fig. 42a, a v pattern consisting of five recording dots is formed after applying a voltage pulse of 2.64 V at 10 ms duration. The average diameter of the recorded dots is about 2.1 nm, corresponding to a potential storage density of about bit/cm 2.In this case, the recorded pattern could continuously be scanned by STM without discernable changes for several hours, indicating the high stability of the pattern under ambient conditions. Fig. 42b shows typical I V curves of the film before and after recording. The electrical resistance of the recorded region is much lower than the one of the unrecorded region, which indicates that a conductance transition occurs after application of the voltage pulse to the film. This transition leads to the bright dots observed by STM. The I V experimental results indicate that the applied electric field could induce charge transfer and produce delocalized electrons in the recorded region [88]. In general, charge transfer includes both, the intra- and inter-molecular modes. In the present case, however, the possibility of the former could be excluded, because intra-molecular charge transfer is unstable while the as-recorded pattern exhibits particular stability. Combining this observation with the packing characteristics of the TDMEE molecules in the crystal structure mentioned above, it is suggested that inter-molecular charge transfer is responsible for the dot formation. Fig. 41. (a) TEM image and corresponding electron-diffraction pattern (insert) of a TDMEE thin film, showing that the film is well crystallized. (b) The packing arrangement of TDMEE molecules in the crystalline thin film, showing the regular alternate donor acceptor stacks.

43 bcurrent (na) Author's personal copy H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) a 10 nm I II Voltage (V) Fig. 42. (a) STM image of a nanoscale v pattern recorded on the TDMEE thin film by applying a voltage pulse of 2.64 V for 10 ms. U=0.34 V, I=0.06 na. Constant height mode. The average diameter of each mark is about 2.1 nm. (b) Typical STM I V curves for (I) unrecorded and (II) recorded regions of the TDMEE thin film. According to the microreflection absorption infrared spectra, Fig. 43, the TDMEE thin film has a coupled stretch band of the three CN groups and the acetylenic fragments in the tricyanomonoethynylethene core at 2170 cm 1 (Fig. 43b), which is nearly identical to the one of the powder materials (Fig. 43a). After the action of an electric field with a threshold voltage of about 7.5 V using a Keithley 4200 Semiconductor system, the IR spectrum of the film (Fig. 43c) exhibits a 36 cm 1 blue-shift of the absorption peak as compared to the one observed in Fig. 43b. The hypsochromic shift of this peak, which is assigned to the tricyanomonoethynylethene moiety, reveals that charge transfer may have taken place in the film during the recording process [231]. Considering the recording stability and the packing characteristics of the TDMEE molecules mentioned above, charge transfer is thought to be inter-molecular. Fig. 43. The microreflection absorption FTIR spectra of (a) a TDMEE powder, (b) a TDMEE thin film before and (c) after applying a voltage.

44 70 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 44a c show the charge-density distributions and the HOMO and LUMO of the TDMEE molecule, respectively. The HOMO was found to be almost entirely located on the DMA moiety, while the LUMO was found to be almost entirely located on the CEE moiety. The calculated HOMO LUMO gap is 2.56 ev. The results suggest that the HOMO of one molecule may interact with the LUMO of its neighboring molecule for charge transfer. Here it is worth noting that the solid-state packing of TDMEE molecules, i.e., cofacially alternate donor acceptor stacks, contributes to the facilitation of electron transfer from the N(CH 3 ) 2 group to the tricyanomonoethynylethene moiety in a neighboring molecule at a critical electric field [215]. The recording mechanism can be explained as follows. The TDMEE Fig. 44. (a) Charge-density distributions of TDMEE (hollow circles: positive charge, solid circles: negative charge). (b) HOMO and (c) LUMO of the TDMEE molecule.

45 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) thin film is originally in a high-resistance state, in which electron-conducting pathways do not exist. When a crucial voltage pulse is applied to the film, charge is transferred from the N(CH 3 ) 2 residue of one molecule to the tricyanomonoethynylethene moiety of a neighboring molecule. The charge transfer brings about an inter-molecular electron-conducting pathway, which ensures the conductance transition of the thin film from a high-resistance to a low-resistance state forming a bright information dot Cyano-2,6-dimethyl-4-hydroxy azobenzene (CDHAB) [232] For a single CDHAB molecule, as shown in Fig. 27e, a hydroxyl group and a cyano group are included, which act not only as electron donor and acceptor groups, respectively, but also as hydrogen-bonding recognition subunits. Based on molecular self-organization, CDHAB molecules were assembled into a highly-ordered crystalline thin film on the freshly cleaved HOPG substrate. Fig. 45a displays the AFM topographic image of the thin film on an HOPG substrate. The observed area is about lm 2. The topography clearly illustrates the flat form of the thin film in micrometer sized areas. The film thickness is about 5 nm as determined by AFM measurements. In order to investigate the internal structure of the thin film, the transmission EDP has been measured. Fig. 45b displays the TEM image of the thin film and its representative selected-area EDP (Fig. 45b, inset). The TEM image shows a uniform and flat surface, and the EDP shows a fairly clear spot pattern, indicating the highly crystalline nature of the thin film. Fig. 46a shows a typical STM image of a CDHAB thin film, where the CDHAB molecules are arranged periodically on the HOPG substrate; the periodic distances along the x- and y-directions marked in the image are 1.5 and 0.8 nm, respectively, while the angle between them is 53. These values are perfectly consistent with the periods along the long axis of CDHAB molecules and between adjacent rows of molecules, as well as the angle between these, as shown in the model of the stacking arrangement of CDHAB molecules in a bulk crystal (Fig. 46b). Considering the structure of CDHAB, we can reasonably assume that the cooperative interaction effects of hydrogen bonding, p p interactions and van der Waals forces lead to the orderly arrangement of CDHAB molecules on the HOPG substrate [ ]. Such a uniform and smooth thin film with suitably large, highly-ordered regions is the basis for technological applications. A flat region was selected for the information recording experiments. To induce the recording dots, pulsed voltages of 0 to +6.0 V in amplitude and ms in width were applied between the conductive probe and the crystalline thin film. To avoid a tip crash, the STM tip was lifted up several angstroms from the tunneling position and the feedback loop was disabled before applying a pulse voltage. By applying voltage pulses of +2.6 V and 3.0 ms between STM tip and substrate, we succeeded in writing a letter y pattern on the crystalline CDHAB thin film. Fig. 47a shows the STM images of the recorded dot patterns in a scanning area of nm 2. One bright dot corresponds to a recorded Fig. 45. (a) AFM image of a CDHAB thin film on HOPG prepared by self-assembly. 10 lm 10 lm. (b) TEM image of a CDHAB thin film and its electron-diffraction pattern in the inset.

46 72 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 46. (a) A typical STM image of a crystalline CDHAB thin film showing its crystalline order. U=0.36 V, I=0.25 na. The periods along the x- and y-directions marked in the image are 1.5 and 0.8 nm, respectively, and the angle between them is 53. (b) A structural model to show the stacking arrangement of CDHAB molecule in the crystalline thin film. The periods along the long axis of a CDHAB molecule and between adjacent rows as well as the angle between them are Å, Å, and 54.30, respectively. mark. The average dot size is about 1.8 nm in diameter. In our experiment, the recorded marks were very stable and no obvious changes could be observed during the continuous scanning process of 6 h. Further studies have shown that when the pulsed voltage was lower than a threshold voltage of +2.6 V, no dots were recorded, but for voltage pulses above this threshold, the probability of dot formation was almost 100%. Furthermore, when a reverse-polarity voltage pulse higher than an absolute value of 2.2 V was applied to the recorded region, the marks could be erased. Fig. 47b and c indicate the situation after one and two information dots have been erased, respectively. By applying a pulsed voltage on the same region of the thin film again, a mark can be rewritten on it (Fig. 47d). With alternate exposure of the film to positive and negative voltages beyond the threshold value, more than ten write-read-erase cycles have been demonstrated in the same area. The local electrical properties of CDHAB storage thin films reported here have been characterized by examining the current voltage (I V) characteristics using STM (Fig. 47e). From the curves it is clear that the CDHAB film shows a different conductance before and after the formation of information marks. As shown by curve I in Fig. 47e, the tunneling current remains low in the CDHAB film in its initial state. After a voltage pulse higher than the threshold (+2.6 V), the film shows a low-impedance during the same voltage conditions (curve II) at the recorded region. The I V characteristics of the film before and after recording suggest that the formation of the information dots is due to the electrical switching of the CDHAB film from a high-impedance state to a lowimpedance state. It is reasonable to assume that such a conductance change might be caused by a reversible inter-molecular charge transfer, as has been frequently observed in other donor acceptor compounds or complexes [88,228, ]. To further validate this assumption, we have studied the conductive behavior of macroscopic CDHAB films and the corresponding characteristic spectra. For this study, a thin film of about 200 nm thickness was used. I V curves measured on an ITO-coated glass substrate are shown in Fig. 48a. Curve I implies that at the low applied voltage the film shows a high-impedance state (OFF state). Beyond a threshold of about +2.8 V, an electrical transition from the high-impedance state to a low-impedance state (ON state) takes place with an abrupt increase in current. The film shows good stability in this high-conductivity state during the subsequent voltage scan (curve II). The high-conductivity state could be returned to the low-conductivity state (OFF state) by applying a negative bias as indicated in curve III, where the current suddenly drops at 2.5 V. After the film has returned to the low-conductivity state, it can be maintained in this state during a negative bias scan and can be switched back to the high-conductivity state by applying a positive bias higher than the threshold, resulting in an OFF ON OFF ON reversible trait. The CDHAB film exhibits consistent electrical bistability under macroscopic conditions (Fig. 47e). Transient conductance measurements performed on the same sample have shown a short transition time of about 45 ns. Furthermore, UV vis spectra of

47 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) a b c d e 2 1 E II Current (na) 0-1 I Voltage (V) Fig. 47. STM images of the CDHAB film during the information recording experiments. U=0.38 V, I=0.28 na. (a) A recorded pattern (letter y ) by a voltage pulse of V for 3.0 ms; Image size: 60 nm 60 nm. (b c) Patterns after erasing one and two marks, respectively, by a voltage pulse of 2.2 V for 3.0 ms. (d) Pattern after rewriting one mark by a voltage pulse of +2.6 V and 3.0 ms. (e) Typical STM I V curves in the unrecorded (curve I) and recorded region (curve II). the CDHAB thin film in two different electrically stable states were compared. Fig. 48b and c exhibit the UV vis spectra before and after the action of an electrical field higher than the threshold voltage of +2.8 V, respectively. In addition to the intrinsic bands of CDHAB film at 346 and 456 nm before the application of an electrical field (Fig. 48b), the spectrum of the CDHAB film after the application of an electrical field (Fig. 48c) exhibits two additional bands at 427 and 498 nm, indicating the intensive interaction among molecules, especially inter-molecular charge transfer [88,225,226,240,241]. When the film is switched back to the high-impedance state, the UV vis spectrum shows the initial state again.

48 74 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) a Current (ma) IV III II I b Voltage (V) 346 Abs c Wavelength (nm) Abs Wavelength (nm) Fig. 48. (a) I V characteristics of CDHAB films on ITO-coated glass, exhibiting the conductance transition from the low- to the high-conductivity state in curve I, the memory effect of the high-conductivity state in curve II, and the recovery of the lowconductivity state with the application of a reverse voltage scan in curves III and IV. UV vis spectra of the CDHAB film before (b) and after (c) application of an electrical field higher than the threshold voltage of +2.8 V. The experimental spectrum (solid line) is fitted with four components at 348 nm (red), 427 nm (blue), 457 nm (cyan), and 498 nm (green) to give the combined signal (magenta). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

49 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Conductance transition induced by transition of molecular structure For practical applications of organic molecules with electronic bistability, a stable switching in ambient conditions is required [ ]. Rotaxanes represent a new class of supermolecules and show stability at various chemical states in solution [177,245,246]. Rotaxane molecules have shown significant potential to be used as building blocks for molecular electronics [247,248]. The key structure of the molecules contains a p-electron-deficient macrocycle that is locked around a p-electronrich component along the molecular thread with two bulky stoppers. The macrocycle can move between different p-electron-rich components upon external stimuli, resulting in a switching of the electronic configuration [177]. This makes rotaxane an interesting candidate for nanorecording. We reported on stable conductance transitions in solid state rotaxane-based Langmuir Blodgett (LB) thin films [249]. We have achieved reproducible nanometer-scale recording on these films by STM. Fig. 49a illustrates the molecular structure of rotaxane 14PF 6 (H1). This molecule has shown conclusive evidence of switching between two distinct configurations. Such a switching originates from the reversible mechanical movement of the CBPQT 4+ (macrocycle) between TTF and DNP recognition sites triggered by chemical stimuli. Theoretical calculations [188,189] have also revealed that a gain of in conductivity can be expected when the CBPQT 4+ moves from TTF to DNP. We first prepared the H1 films on an ITO-coated glass substrate using common LB technique. The substrate was dipped into the solution and lifted 50 times for multiple H1 coating. The average thickness of the films was 70 nm. Using a two terminal junction, we observed a fast conductance-switching phenomenon with an on/ off ratio of 100 in the H1 LB film. The I V characteristic of the film (Fig. 49b) shows a reversible electrical bistability and a nonvolatile memory effect in the film. The first voltage scanning from 0 to 2 V, described by curve I, shows a sharp current increase starting at 1.4 V, which indicates the transition from a low-conductivity state (OFF state) to a high-conductivity state (ON state). In curve II, the I V data measured after 12 h from the first measurement clearly show that the ON state in the sample is maintained for a long time. The OFF state can be gradually recovered by applying reverse voltage Fig. 49. (a) Molecular structure of rotaxane H1. (b) I V characteristics of the thin H1 films on ITO glass, exhibiting a conductance transition from the low- to the high-conductivity state in curve I, the memory effect of the high-conductivity state in curve II, and the recovery of the low-conductivity state with reverse voltage scans in curves III V. (c) Transition time measurement of the conductance switching, showing a very short transition time of about 60 ns.

50 76 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) scanning (curves III, IV, and V). A transient conductance measurement (Fig. 49c) performed on the same H1 sample shows a short transition time of 60 ns. To explore the origin of the reversible ON OFF transition in the H1/ITO sample, we performed a series of control experiments. I V measurements of the H1 films supported on HOPG substrates gave the same OFF, ON, and Recovery behaviors under external voltage scans. Similar I V measurements on a layer of poly(methyl methacrylate) (PMMA) with a thickness of 100 nm revealed a sudden conductance transition from low- to high-conductance states, but no resumption of the OFF state could be obtained by applying reverse voltage scans. These experiments demonstrate that the reversible ON OFF transition and nonvolatile memory effects of the H1 films originate from the rotaxane molecules. This stable conductance switching of the H1 films indicates that the film can be a good candidate for stable and reproducible nanorecording. We utilized the conductance switching of the H1 thin films for STM-based nanorecording. A film of about 20-nm thickness was deposited on a HOPG substrate (H1/HOPG). Fig. 50a c show the formation of recording dots, which were written one by one by applying a positive voltage pulse of 2 V and 5 ms duration via the STM tip. No recording dot was found if a negative voltage pulse was applied to the films, indicating that the recording is dependent on the voltage polarity. Additionally, no recording dots could be written on the pure HOPG substrate when a similar voltage pulse was applied. Fig. 50d shows the I V measurements on the original film and on the dots formed. It clearly shows a conductivity increase at the dot location. This I V characteristic is highly reproducible, suggesting that the voltage pulse causes a conductance transition of the H1 molecules and produces the recording dots. This voltage pulse-induced conductance transition manifested by the dots on the films can be observed directly in conductive contact AFM experiments. The surface morphology and the corresponding local conductance information can be acquired simultaneously and independently in the conductive contact AFM mode since the tip is in contact with the film in a constant force mode during scanning. We first applied a rectangular voltage pulse (+3 V, 5 ms) on the H1/HOPG sample through the AFM probe, and then acquired the AFM height and current images as shown in Fig. 51. The AFM current images and height images before and after applying the voltage pulse are recorded as Fig. 51a, a 0, b and b 0, respectively. A bright spot of about 8 nm in diameter can be observed clearly in Fig. 51b (current image), but no specific feature was recorded in the height image (Fig. 51b 0 ) at the corresponding position. It can be seen clearly from the height profile across the mark on the height image a b e c d Current (na) Voltage (V) Fig. 50. STM images of the recording dots written on the H1 thin films. (a c) Recording dots written one by one on application of voltage pulses to the STM tip. (d) Typical I V characteristics measured on the original films (I) and on the induced recording dots (II). (e) STM images of a 5 4 recording dots array on the H1/HOPG sample. This film has been left exposed to air for about 2 months since its preparation. STM was performed in constant current mode with set points of U = 0.65 V and I = 0.05 na.

51 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) (Fig. 51a 00 and Fig. 51b 00 ) that the shape of the height profile does not change and that the height difference on the site of the mark before and after the recording is less than 0.3 nm. This result rules out the possibility of destruction of the films by the voltage pulse as well as evaporation of material from the tip onto the surface. Our AFM measurements give direct proof that the recording dots originate from the conductance transition in the H1 film. In addition, we tried to erase the bright marks by applying a negative voltage pulse to them. But this was not successful even though we changed the amplitude and the duration time of the pluses. With a large negative pulse, the films were destroyed. This nonerasability might be due to a potential barrier of the molecule between the off and on states. Further modification of the molecule is probably needed to perform the erasability of the recording. Our further experimental results show that the H1/HOPG samples are very stable in ambient conditions. The conductance switching phenomena can be repeatedly observed on the same sample even after the sample has been left exposed in air a b 1 2 a b a b Heihgt (nm) h=4.42nm Height (nm) h=4.64nm nm nm Fig. 51. AFM current image, height image and the sketch map of a height profile across the height image before (a, a 0,a 00, respectively) and after (b, b 0,b 00, respectively) a voltage pulse was applied on the H1 films through the AFM probe. After applying the voltage pulse, a clear bright mark appears in the current image (b) at the region where the voltage pulse was applied. No morphology changes can be found in the height image (b 0 ). The numbers of 1 and 2 in b indicate the positions of the darkest region in the height image and the region on the formed bright mark, respectively. These two regions are also indicated in the height images of a 0 and b 0. The profiles across the two regions in a 0 and b 0 are shown in a 00 and b 00, respectively. It can be seen from a 00 and b 00 that the shape of the height profile does not change and that the height difference on the site of the mark before and after applying the voltage pulse is less than 0.3 nm, indicating that the voltage pulse caused a conductance change in the H1 thin film without a distinct surface morphology damage.

52 78 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) for about 2 months. Fig. 50e shows a 5 4 array of the recording dots written one by one by a STM tip (about 2.2 V and 5 ms) on the two-month-old H1/HOPG sample. The writing is generally reproducible. If the writing occasionally fails during the first attempt, only little changes are needed to adjust the voltage pulse for a subsequent successful writing. The shape and size of the dots are not uniform, which may be due to different configurations of local regions in the H1 films or to some unexpected changes in the external environment when the voltage pulses were applied. The combination of stable switching behavior and ease of control makes H1 quite applicable for nanorecording. Similar conductance transitions in rotaxane multilayers connected to two terminal junctions have been reported in the literature [179,187]. It is proposed that the switching originates from the transitions of the macrocycle. This hypothesis is supported by theoretical calculations [188,189]. Recently, it was argued that this type of conductance transition is due to the formation of metal filaments, rather than to the movements of the macrocycle in the rotaxanes [190]. Our experiments show no obvious evidence of the formation of strong conducting filaments [ ]. First, if the metallic filaments are formed, a very small resistance in the high conductive state and a very high ON/OFF ratio will be obtained. Our ON/OFF ratio of 10 2 is not consistent with the case where conducting filaments are formed ( ). Second, the threshold voltage should increase linearly with the film thickness under the condition of metallic filament formation. This is not in agreement with our results where the threshold voltage has no obvious relationship to the film thickness. Third, our comparative experiments performed on PMMA thin films suggest that the conductance transition depends on inherent properties of the H1 molecules. Another possibility for this switching may be a crystallinity change of the thin film. However, the AFM measurements show that the height difference on the site of the marks before and after recording is less than 0.3 nm. This observation suggests that a crystallinity change is not the main reason for the recording. Therefore, the conductance switching behavior of H1 films in our experiment is believed to originate from the machine-like motion of the macrocycle between two recognizing sites. This resembles a transient net oxidized or net reduced state of the molecules when current flows through the asymmetric tunnel barriers in the two terminal junctions [179]. It is of interest to determine the orientation and configuration of the H1 molecule on HOPG. We calculate several possible configurations of H1 on HOPG by MM. The MM+ force field computational method was used. To this end, totally 24 configurations were chosen by considering the position of the macrocycle, the molecular orientation, the adsorption modes (such as the tilted molecule, the lyingaside molecule and the flat molecule) and the effect of convergence limitation. Six fully-optimized stable configurations in three couples have been obtained (see Fig. 52). The configurations shown in Fig. 52a, a 0 correspond to a possible stable geometry for the ideal LB film, in which the macrocycle moves up and down along the molecular thread with a relatively low barrier. This kind of geometry is mentioned in many other similar systems. Fig. 52b and b 0 show the arch structures of the H1 molecule. The molecular plane of the obstacles at the two sides is parallel to the HOPG surface due to the p p interaction. It appears as if the molecular backbones are suspended between two obstacles. This geometry has a lower potential barrier for the movement of the macrocycle. The last couple shown in Fig. 52c and c 0 gives a reverse energy relationship compared to the configurations of Fig. 52a, a 0 and b, b 0. In terms of Fig. 52a, a 0 and b, b 0, the calculated total energy of the former one (the macrocycle is on the TTF group) is lower than the latter one (the macrocycle is on the DNP group) in each couple. But when the molecule is in the configuration of the last couple, the total energy of the molecule is about 7 kcal/mol lower in Fig. 52c 0 (the macrocycle is on the DNP group) than in Fig. 52c (the macrocycle is on the TTF group). This result can give a possible explanation for the difficulty in erasability. Reversible, erasable, and rewritable nanoscale recording on organic thin films is of practical importance in ultra-high density information storage. Although writing high-conductance nanoscale marks on rotaxane H1 thin films was possible, erasing of the marks showed some difficulties. We achieved the reversible, erasable, and rewritable nanorecording on a different film, the rotaxane H2 LB [253]. The written dots, with a feature size of 3 nm, show significant stability in air at room temperature. Fig. 53a shows the structure of the H2 molecule. It is a variation of the H1 molecule [254]. An H2 molecule consists of a p-electron-deficient ring cyclobis(paraquat-p-phenylene) (CBPQT 4+ ) and a dumbbell-shaped component. The dumbbell component has two p-electron-rich recognition sites (TTF

53 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) Fig. 52. Possible molecular configurations of H1 molecule on HOPG obtained by molecular mechanics calculations. a and a 0, two possible tilted configurations with the macrocycle at TTF (a) and DNP (a 0 ). b and b 0, two possible arch configurations with the macrocycle at TTF (b) and DNP (b 0 ), in which the macrocycle can move between TTF and DNP parallel to the HOPG surface with relative low barrier. c and c 0, two possible flat configurations with the macrocycle at TTF (c) and DNP (c 0 ). and DNP; see Fig. 53a) and is terminated by bulky stoppers. The ring encircles part of the dumbbell, making them mechanically inter-locked with each other. The ring can move back and forth between the two different p-electron-rich recognition sites in response to the external stimuli, resulting in the switching between the two stable structures. Recent theoretical calculations [188,189,255] have revealed that the movement of the CBPQT 4+ ring is accompanied by a change in the molecular electronic structure. The difference between H1 and H2 molecules lies in the spacer between TTF and DNP. In H2, a rigid cyclohexyl spacer replaces the soft alkyl spacer ( (CH 2 ) 5 ) in H1 [254]. For STM recording, the H2 rotaxane thin films with 20-nm thickness were prepared on HOPG substrates using the LB technique; for macroscopic I V characterization, the films with a thickness of 70 nm were prepared on the ITO-coated glass substrates; and for the micro-raman studies, the thin films were also prepared on ITO-coated glass substrates. The thickness of the film is about 100 nm, with some protuberant islands of 400 nm in height. The micro-raman spectra were acquired on these plateaus with an excitation wavelength of nm. By applying voltage pulses to the H2 thin films through the STM tip, we realized repeatable and rewritable nanorecording (Fig. 53b). Nanoscale dots can be written repeatedly with voltage pulses of typically 2.0 V and ms duration. This case is similar to the one of the H1 thin films [249]. What is more interesting is that the marks written on the H2 thin films can be erased, re-recorded, and re-erased on the same site. In the whole recording process, the dots maintain a size of 3 nm and are stable in air at room temperature for more than 12 h. The marks are directly visible in the current image in conductive contact AFM characterizations, but invisible in the topographic image (see Fig. 53c), suggesting that the appearance of dots is indeed due to conductance transitions of the H2 molecules. Our further studies show that the H2 films are suitable

54 80 H.-J. Gao, L. Gao / Progress in Surface Science 85 (2010) a b CH 2 O O CH 2 O CH 2 O O CH 2 O stopper N CH 2 S N 4PF 6 - S S (1) (2) (3) S S N TTF + SCH 2 CH2 O N DNP CBPQT 4+ OCH 2 stopper OCH 2 O OCH 2 OCH O 2 OCH 2 (4) (5) (6) c Fig. 53. (a) Molecular structure of the H2 molecule. (b) Frames 1 3 show bright marks written one by one using STM; frames 4 6 show the erasing, rewriting, and re-erasing on the same recording site, with U = 0.8 V and I =0.05 na. The voltage pulse for recording was 2 V for 3 ms, the pulse for erasing was 2 V for 3 ms. Scale bar is 6 nm. (c) Topographic (left) and current (right) AFM images of two marks written by conductive contact AFM on a 8 nm thick H2 films. Scale bar is 10 nm. for the reversible nanorecording even after being stored in air at room temperature for one month. To verify the origin of the conductance transitions of the H2 molecules, we performed macroscopic I V measurements on the H2 films using a standard I V characterization system (Keithley, model 4200SCS). The ITO substrate served as one electrode. We used a freshly cleaved HOPG plate as another electrode, which was pressed against the H2 thin film to ensure a good contact. Fig. 54 shows the I V