Although an isolated individual molecule clearly has only one ionization potential, multiple values are found for molecules in ordered assemblies. Photoelectron spectroscopy of archetypical π-conjugated organic compounds on metal substrates combined with first-principles calculations and electrostatic modelling reveal the existence of a surface dipole built into molecular layers. Conceptually different from the surface dipole at metal surfaces, its origin lies in details of the molecular electronic structure and its magnitude depends on the orientation of molecules relative to the surface of an ordered assembly. Suitable pre-patterning of substrates to induce specific molecular orientations in subsequently grown films thus permits adjusting the ionization potential of one molecular species over up to 0.6 eV via control over monolayer morphology. In addition to providing in-depth understanding of this phenomenon, our study offers design guidelines for improved organic–organic heterojunctions, hole- or electron-blocking layers and reduced barriers for charge-carrier injection in organic electronic devices.
It is well established that the work function (Φ) of metals depends on the crystal face1,2,3. Φ is defined as the energy difference between the Fermi level (EF) and the electrostatic potential above the sample, the vacuum level (Vvac). For example, for copper, Φ values of the (100), (110) and (111) surfaces are spread over a range of 0.5 eV (refs 1,2). As EF is constant, this observation has been explained by the difference in the intrinsic ‘surface dipole’: variations of the geometric and, consequently, electronic structure cause a different amount of the electronic cloud to spill out of the bulk into the vacuum3,4. The resulting dipole raises Vvac to a larger or smaller extent and thus impacts Φ (refs 4,5). Note that this effect can only be observed for laterally extended surfaces, as the spatial region above the sample where Vvac is raised reaches farther away from the surface with increasing sample size (that is, area of the exposed surface)6,7. Small metal clusters with multiple facets of different crystal orientations have only one well-defined work function8,9.
For van der Waals crystals of non-dipolar molecules, surface dipoles and work-function anisotropy have not yet been explored6. Although variations of the ionization potential (the molecular equivalent of the work function) depending on the molecular orientation on a substrate have been reported before10,11,12,13,14,15,16, the prevalent interpretation in terms of variable photo-hole screening could never be satisfactorily quantified. Here, we propose a qualitatively different and novel explanation for the intriguing observation that one and the same molecule can have different—still well-defined—ionization potentials (ionization energies) if part of an ordered supramolecular structure.
We carried out X-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS) on α,ω-dihexyl-sexithiophene (DH6T) and α-sexithiophene (6T), on Ag(111). The ionization potentials of the molecules change by up to 0.6 eV depending on whether they are lying down flat on the substrate or standing upright. In contrast to previous attempts10,11,12,13,14,15,16, we rationalize these observations in terms of the collective electrostatic effect of the highly anisotropic intramolecular charge distribution on the basis of density-functional theory (DFT) calculations and electrostatic modelling. Supplementary studies on different substrates and molecules underline the universality of the observed effects and their explanation. We stress that the general concept is also valid for single crystals and ordered polymers.
As 6T and DH6T are used in organic field-effect transistors17,18,19,20,21,22,23 (OFETs), we discuss the immediate practical relevance of our findings in terms of the hole-injection barrier (HIB), a crucial parameter in organic electronic devices6,24,25. Pre-patterning an electrode with films of lying or standing DH6T enables subsequent growth of films of likewise lying or standing 6T molecules and thus permits lowering the HIB at the Ag–6T contact by 0.4 eV. Furthermore, we derive conceptual guidelines for molecular design to optimize the energy-level alignment at inorganic–organic and organic–organic heterojunctions. Our findings thus open new routes towards organic electronic devices with improved performance and functionality, not only OFETs but also organic light-emitting diodes and organic solar cells.
In general, the orientation of molecules in mono- and multilayers with respect to the substrate critically depends on the relative strengths of molecule–substrate interaction versus intermolecular interaction26,27. In the case of DH6T (and similar thiophene derivatives23), the molecules in the monolayer adsorb lying flat on metal surfaces28,29, whereas molecules in subsequent layers are ‘standing’ with their long axis close to the surface normal28. The experimental UPS spectra of DH6T on Ag(111) in the monolayer (L for ‘lying’) and multilayer (S for ‘standing’) regime are shown in Fig. 1a, and Fig. 1b shows the corresponding simulated spectra. In the former, three low-binding-energy peaks can be clearly distinguished with their respective maxima at 1.6 eV, 2.3 eV and 2.9 eV in the L-regime, and at 1.0 eV, 1.7 eV and 2.3 eV in the S-regime. These peaks are derived from the highest occupied molecular orbital (HOMO), the HOMO-1 and the HOMO-2 (ref. 30). All peaks are at 0.6 eV lower binding energy for the (standing) multilayer (S) compared with the (lying) monolayer (L). Note that the change in the intensity ratio between the HOMO and HOMO-1 peak from 1:1 (L) to almost 1:2 (S) is indicative of the different orientation of the long molecular axes with respect to the surface normal due to photoemission selection rules30. As commonly observed6,24,31, the adsorption of a molecular monolayer leads to a decrease of Φ of the metal substrate. In our case, ΔΦ≈−0.7 eV as determined from the secondary-electron cutoff (SECO; see the Methods section). This lowering of Vvac above the sample surface is often termed ‘interface dipole’6,24,31. No further reduction of Φ is observed on subsequent deposition of the multilayer (Fig. 1a). Consequently, the −0.6 eV binding-energy shift of the molecular levels directly translates into a reduction of the molecular ionization potential (that is, the energy difference between HOMO and Vvac) by this amount. To understand the physical origin of this shift, it is indispensable to investigate whether the core levels are affected in the same way as the valence levels. For DH6T, the XPS spectra of the sulphur 2p core levels are shown in Fig. 1c. Whereas only one doublet (2p3/2 at 164.40 eV and 2p1/2 at 165.65 eV) is observed in the L-regime, a second doublet appears in the S-regime which is shifted by 0.6 eV towards lower binding energy. The consistency with the UPS data confirms that all electronic states in the (lying) monolayer (L) are rigidly shifted to higher binding energy with respect to EF and Vvac.
In a second set of experiments, we investigated the orientation dependence of the ionization potential in ordered layers of 6T. In contrast to DH6T, 6T on Ag(111) does not show an abrupt transition in morphology from the first to the second layer. The first 6T layers adsorb lying flat on Ag(111) (refs 30,32) and only a slow, gradual transition to almost standing molecules was suggested for very thick (>200 nm) films30,32. An exemplary UPS spectrum of 150 Å 6T (multilayer) on Ag(111) is shown in Fig. 2a. Again, three peaks (HOMO, HOMO-1 and HOMO-2) can be identified with maxima at 1.8 eV, 2.5 eV and 3.1 eV. As the π-electronic structure of 6T is virtually identical to that of DH6T, the intensity ratio of the HOMO and HOMO-1 peaks of 1:1 is in accordance with the model of lying molecules30. Pre-patterning the Ag(111) substrate with a (lying) monolayer of DH6T and subsequent deposition of 6T does not change the valence spectrum of 6T (Fig. 2a). The interface dipole between DH6T and 6T is negligible (<0.1 eV) and is thus indicative of vacuum-level alignment6,7 at this organic–organic heterojunction. In the next experiment, the Ag(111) surface was pre-patterned with a bilayer of DH6T, that is, standing DH6T is now exposed on the surface. Deposition of 6T onto this modified substrate significantly alters the UPS spectrum of 6T (Fig. 2a). The valence levels are rigidly shifted by 0.4 eV towards lower binding energy and the intensity ratio HOMO/HOMO-1 changes to 1:2. We therefore propose the growth model shown in Fig. 3 for this organic heterostructure: owing to relatively strong π–π interactions, 6T grows lying down (L) on the lying DH6T monolayer; on the second (standing) DH6T layer, 6T also grows standing upright (S) as π–π interactions between 6T molecules dominate over the interaction with the now inert surface composed of the alkyl chains of the underlying (standing) DH6T layer. As vacuum-level alignment6,7 also prevails between the bilayer DH6T and 6T, the shift of the HOMO observed in Fig. 2a translates into a 0.4 eV lower ionization potential of 6T in the S-morphology compared with the L-morphology. Building the same DH6T–6T heterostructure on polycrystalline gold instead of Ag(111) yields essentially the same results (see the Supplementary Information), confirming that the observed effects are quite general for metal surfaces.
To understand our observations, it is important to consider that the kinetic energy of photoelectrons and thus the measured ionization potential is affected by the polarization of neighbouring matter by the photo-hole. For organic thin films on metals this includes (1) the metal substrate6,7,24,33 and (2) the surrounding molecules. With increasing thickness of the organic layer, the screening by the metal becomes less important for the topmost molecules (probed by XPS and UPS), resulting in an apparent shift of the molecular levels away from Vvac and thus an increase of the measured ionization potential. In our case, however, the ionization potential decreases as multilayers of DH6T are deposited onto the first (lying) layer. The same is true for 6T deposited on a bilayer of DH6T, where the ionization potential also decreases compared with direct deposition onto Ag(111) or onto a (lying) monolayer of DH6T. We conclude that screening of the photo-hole by the metal cannot account for the observed lowering of the molecular ionization potential.
It may be speculated, however, that the photo-hole is more efficiently screened by surrounding standing molecules than by surrounding flat-lying molecules and, for similar organic compounds, the impact of molecular orientation on the ionization potential has indeed been qualitatively rationalized in terms of the polarization energy depending on the packing density and/or morphology10,11,15,16. Here, we provide an upper limit for this proposed variation in polarization energy: a molecule in the topmost organic layer is surrounded by the metal substrate (at some distance), neighbouring molecules in the half-space below and by vacuum in the half-space above; a molecule deeper in the organic layer is also surrounded by molecules on top34. Clearly, the presence or absence of neighbouring molecules in the upper half-space must have a stronger effect on the polarization energy (and thus the measured ionization potential) than differences in the orientation of neighbouring molecules. Re-examining the XPS data in Fig. 1c, we find that the binding energy of the S(2p) peaks attributed to the first (lying) layer of DH6T does not change on deposition of subsequent layers of DH6T. To further confirm our reasoning, we carried out extra independent measurements (see the Supplementary Information), which yield an upper limit of 0.15 eV for the difference in polarization energy between the bulk and the surface35. We thus conclude that the 0.6 eV (0.4 eV) difference in the ionization potential between standing and lying DH6T (6T) cannot be explained in terms of photo-hole screening effects alone, and that another mechanism must be involved.
To understand the remarkable finding of one molecule having a different ionization potential depending on its orientation in an ordered supramolecular structure10,11,12,13,14,15,16,28, we carried out plane-wave-based DFT calculations using periodic boundary conditions and the repeated-slab approach on single layers of standing and lying DH6T and 6T molecules on the basis of available structural data (see the Methods section for details)13,29,36,37,38. The occupied density-of-states (DOS) for a lying (L) and standing (S) layer of DH6T is shown in Fig. 1b. In addition to good qualitative agreement with the experimental UPS spectra (Fig. 1a), we find that indeed, all molecular levels are closer to Vvac for the S-layer compared with the L-layer, that is, the ionization potential is lower for standing molecules. For 6T, DFT calculations yield similar results: the molecular levels are closer to Vvac for the S-layer, that is, the ionization potential is again lower compared with that for molecules in the L-layer. The corresponding DOS is shown in Fig. 2b.
To rationalize the fundamental mechanism that gives rise to this shift, we consider electrostatics on the molecular scale in analogy to the situation for extended metal surfaces versus metal nanoclusters (see above). Figure 4a shows the electrostatic potential (obtained from DFT calculations) around one isolated 6T molecule relative to its HOMO energy. The colour scheme (together with the energy scale) thus indicates the amount of work required to promote one electron out of the HOMO to any given point in space. Consistent with the observation that an isolated molecule has only one well-defined ionization potential39, the potential converges to a single value of Vvac (cyan) in any direction on a submolecular length scale. It becomes apparent though, that Vvac is higher directly above the (negatively charged) π-electron system (blue region marked ‘L’) than next to the hydrogen-terminated ends of the molecules (green region marked ‘S’). To a first approximation, the potential distributions of the individual molecules add up as molecules assemble into, for example, crystals or layers6. For molecules standing in a layer, the S-region dominates the electrostatic potential above the layer, whereas for lying molecules, the L-region determines the value of Vvac above the layer. This is shown in Fig. 4b,c, where the electrostatic potential of the molecular layers is plotted relative to their respective HOMO energies. Clearly, Vvac is higher (blue) above the layer with lying molecules than it is above the layer comprising only standing molecules (green), thus leading to the lower ionization potential of the latter. To confirm the validity of our model, we also carried out DFT calculations for lying and standing layers of pentacene, where the difference between the two respective ionization potentials was experimentally determined to be about 0.5 eV and no satisfactory explanation could be found11,13,14. In good agreement, our calculations yield a difference of 0.6 eV.
Whereas DFT calculations enable quantitative analysis, we offer an even simpler, purely electrostatic model to establish a more intuitive picture. We approximate the charge distribution corresponding to one 6T molecule (Fig. 5a) by a number of point charges (Fig. 5b): the π-electron system above and below each ring is clearly negatively charged; this is represented by negative point charges of −0.5e (elementary charge) placed 0.5 bohr above and below the molecular plane. These negative charges are compensated by a +1.0e point charge in the plane of the molecule. This pattern is repeated six times (one for each ring) along the long molecular axis with a spacing of 8 bohr (approximately equal to the distance between individual thiophene rings). From this model molecule, a two-dimensional (2D) (see the Methods section) molecular crystal is built (Fig. 5c), consisting of 21 standing model molecules in one layer with a distance of 6 bohr between their long molecular axes. Below that, one more such layer is placed with a gap of 10 bohr to the first layer.
This model crystal has two distinctly different crystal faces: one (top and bottom) is terminated by the point-charge pattern −0.5|+1.0|−0.5||−0.5|+1.0|−0.5||… that represents the hydrogen-terminated ends of the 6T molecules exposed in a standing layer; the other (left and right) is terminated by negatively charged sheets that represent the π-electron cloud exposed in a lying layer. The electrostatic potentials of all point charges are summed up to yield the potential within and around the model crystal (Fig. 5c). In analogy to the results from the DFT calculations, we find an extended region of lower electrostatic potential (green) over the hydrogen-terminated ends of the molecules, relevant for the standing layer (S). Above the π-system, there is an extended region of higher electrostatic potential (cyan), relevant for the lying layer (L). As a consequence, the work required to promote an electron from any one energy level within the model crystal (for example, the HOMO or the S(2p) core levels) into the spatial region above the hydrogen-terminated ends (S) is less than that for promoting an electron into the spatial region above the π-electron clouds (L). This difference can be measured as soon as the lateral extent of the supramolecular structure is large compared with a single molecule. The situation is equivalent to the presence of a considerable intrinsic surface dipole4,5 (negatively charged π-system, positively charged molecular plane below) at the surface of the lying molecular layers, whereas no such dipole occurs at the surface of standing layers. Note that, whereas the surface dipole of metals is ‘pushed back’ on adsorption of molecules (thus giving rise to an interface dipole)6,24,25,40,41, the intramolecular surface dipole remains unaffected by establishing contact to either the (metal) substrate or another organic layer. Our choice for the amount and separation of the charges in the model molecule and model crystal is justified a posteriori as the difference in Vvac above the π-system (L) and hydrogen-terminated ends (S) amounts to 0.4 eV (Fig. 5c), which compares favourably to the experimentally observed differences in the ionization potential of standing versus lying 6T layers (Fig. 2a). Our results are summarized in the energy-level diagram in Fig. 6.
We emphasize that the presence of an intrinsic surface dipole in molecular layers has important implications for organic electronics: Fig. 6b shows that an electronic heterojunction with a large energy-level offset can, in fact, be realized with only one molecular species (DH6T), a so far unexplored concept. As such heterojunctions play a crucial role in organic solar cells and in the context of hole- or electron-blocking layers in organic light-emitting diodes, we suggest that this energy-level offset may be tuned by chemically tailoring the end groups on the π-conjugated core (alkyl segments in DH6T). Inserting a dipole pointing towards the core by, for example, introducing heteroatoms or making them more electron withdrawing (for example, by fluorination23), increases the ionization potential of the standing layer (thus decreasing the offset); a dipole pointing away from the core or more electron-donating groups can be expected to further decrease the ionization potential of the S-layer (thus increasing the offset). As it determines the barrier for charge-carrier injection into the organic layer, the energy difference between the Fermi level of a (metallic) electrode and the conducting states in the active organic layer is also of uttermost importance for optimizing the performance of organic electronic devices6,24,25. For the occupied manifold of states discussed here, this is the HIB6,24. Although control over molecular orientation already allows considerably reducing the HIB into one and the same molecule by several tenths of an electronvolt, the strategies for chemical modification suggested above can be expected to contribute to a further lowering; similar considerations hold for the unoccupied manifold of states connected to the electron injection barrier in n-type OFETs (ref. 23).
As other important factors in organic electronic devices, for example, photoluminescence or charge-carrier mobility17,42,43, also depend on the orientation of the (intrinsically anisotropic) molecules, our approach of pre-patterning a metal surface with appropriate molecular species (shown here for DH6T) seems to be a promising tool for controlling the orientation of subsequently deposited molecules; for 6T, the gradual transition30,32 from lying to standing orientation can be reduced from hundreds of layers to only two.
UPS experiments were carried out at the FLIPPER II end-station at HASYLAB44 (Hamburg, Germany). The interconnected sample preparation chambers (base pressure 2×10−9 mbar) and analysis chamber (base pressure 2×10−10 mbar) enabled sample transfer without breaking ultrahigh-vacuum (UHV) conditions. The Ag(111) single crystal was cleaned by repeated Ar-ion sputtering and annealing cycles (up to 550 ∘C). DH6T (H. C. Starck GmbH) and 6T (Aldrich) were evaporated using resistively heated pinhole sources, at evaporation rates of about 1 Å min−1. The film mass thickness was monitored with a quartz crystal microbalance. Spectra were recorded with a double-pass cylindrical mirror analyser with an energy resolution of 150 meV and a photon energy of 22 eV. The SECOs were measured with the sample biased at −3.00 V. The work function was calculated by subtracting the sum of the total width (that is, SECO to Fermi level) of the photoelectron spectrum and the bias voltage from the photon energy. XPS experiments were carried out at the end-station SurICat (beamline PM4)45 at the BESSY synchrotron (Berlin, Germany). There, the UHV system consists of interconnected sample preparation (base pressure 1×10−8 mbar) and analysis (base pressure 1×10−10 mbar) chambers. The spectra were collected with a hemispherical electron energy analyser (Scienta SES 100) with 120 meV energy resolution at 20 eV pass energy. The photon energy was 400 eV. Further XPS spectra were measured with an Al Kα1/2 laboratory source in a custom UHV system. Sample preparation for XPS measurements was analogous to that for UPS measurements. All preparation steps and measurements were carried out at room temperature. The fitting of UPS and XPS spectra (Voigt peaks and Shirley background) was carried out with the program WINSPEC (Namur University).
In the DFT calculations, the repeated-slab approach was used with the vacuum region separating two consecutive molecular layers being ≥20 Å. The PW91 exchange-correlation functional was used. For the valence–core interactions, the projector augmented-wave method46 was used permitting the low kinetic energy cutoff of 20 Ryd for the plane-wave expansion of the valence Kohn–Sham orbitals. Monkhorst–Pack grids of 1×4 k-points (lying 6T and lying pentacene), 1×2 (lying DH6T) and 4×4 (all standing structures) were used for the integration of the 2D brillouin zone. The isolated 6T molecule was calculated in a 50×30×20 Å box at the Γ-point only. The atomic positions within the molecules were optimized until all remaining forces were ≤0.01 eV Å−1. All calculations were carried out with the VASP code47,48. The 3D graphics were produced with XCrysDen49.
In the absence of experimental structural data for (lying) monolayers on Ag(111), the surface unit cells of the respective molecules on Au(111) were used in the DFT calculations: a=25 Å, b=6 Å, γ=65.0∘ for 6T (ref. 29); a=38 Å, b=16 Å, γ=19.0∘ for DH6T (ref. 29); and a=5.76 Å, b=15.3 Å, γ=79.1∘ for pentacene13,36. As the actual thin-film structures for the (standing) multilayer systems are also not known from experiment, a single layer of standing molecules was cut out of the respective bulk structures for the DFT calculations; the lateral unit cells containing two molecules arranged in a typical herringbone fashion were taken to be a=7.851 Å, b=6.029 Å, γ=0∘ for 6T (ref. 37) and a=6.266 Å, b=7.742 Å, γ=84.68∘ for pentacene13,38; the long molecular axes in these structures are tilted by about 25∘ from the layer normal. For DH6T, a single layer of 6T was cut out of the 6T bulk structure and a hexyl chain was manually attached to either end of the molecules; in the course of the subsequent DFT calculations, the geometry of these hexyl chains was optimized.
Owing to intrinsic shortcomings of DFT, the calculated HOMO energies usually underestimate ionization potentials. As photo-hole screening is not included in standard DFT calculations, our calculated shifts in ionization potential have to be regarded as shifts of the initial electronic states before removal of the photo-electron. We attribute the overestimation of the shifts in ionization potential to the high degree of order and uniformity in the simulations (not necessarily present in experiment) and to possible discrepancies between the structures assumed for the calculations and the actual structures probed in experiment.
The 2D model molecular crystal in Fig. 5c is equivalent to a 3D crystal where the point-charge pattern shown in Fig. 5c is infinitely continued into and out of the plane of drawing. Note that in the case of 2D electrostatics, the potential, V, decreases with increasing distance, r, from a point charge as V ∝−ln(r2) instead of the familiar V ∝1/r for the 3D case.
The authors thank H. C. Starck GmbH for providing DH6T. N.K. acknowledges financial support by the Emmy Noether Program (DFG). G.H. is a Marie-Curie Fellow under the INSANE project (contract no. 021511). We thank L. Romaner for helpful discussions and the SFB 488 ‘Mesoscopically Organized Composites’ for financial support.