Orientation, electronic decoupling and band dispersion of heptacene on modified and nanopatterned copper surfaces

The adsorption of heptacene (7 A) on Cu(110) and Cu(110)-(2 × 1)-O was studied with scanning tunneling microscopy, photoemission orbital tomography and density functional calculations to reveal the influence of surface passivation on the molecular geometry and electronic states. We found that the charge transfer into the 7 A molecules on Cu(110) is completely suppressed for the oxygen-modified Cu surface. The molecules are aligned along the Cu-O rows and uncharged. They are tilted due to the geometry enforced by the substrate and the ability to maximize intermolecular π-π overlap, which leads to strong π-band dispersion. The HOMO-LUMO gap of these decoupled molecules is significantly larger than that reported on weakly interacting metal surfaces. Finally, the Cu-O stripe phase was used as a template for nanostructured molecular growth and to assess possible confinement effects.

These authors contributed equally to the work. * Authors to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
Future organic electronic and optoelectronic devices and sensors will be based on structures with length scales of larger conjugated molecules [1,2]. Such structures can be envisaged to be produced by top-down lithographic approaches [3], or they can be created via bottom-up molecular self organization [4,5]. An interesting nanoscopically patterned template for growing molecular nanostructures is the Cu(110)-p(2 × 1)-O reconstructed stripe phase, which consist of alternating stripes of clean Cu(110) and Cu -O [6,7] with a periodicity in the few nanometer range. Here, we investigate the growth and interaction of heptacene (7 A), a prototypical organic semiconductor whose length is on the order of the width of the stripes, on the surfaces of strongly interacting Cu(110) and passivated Cu(110)-(2 × 1)-O individually, as well as on the stripe phase. We resolve the geometric and electronic alignment of 7 A on the two surfaces in general, and compare it to its behavior in the nanoscopically confined system.
Research into the properties of higher acenes (n > 5, i.e. beyond pentacene) is presently attracting much interest from both the applied and basic science communities [8][9][10][11], despite the difficulties resulting from their sensitivity to air and light. The interest is driven by their expected superior properties [10,12,13] such as high lying highest occupied molecular orbital (HOMO) energy levels, an increasing carrier mobility and a decreasing band gap [14][15][16]. In a number of recent studies, on-surface synthesis strategies from stable precursors have been employed to access the properties of higher acenes, which revealed insights into the electronic properties of the molecules, for example the HOMO-lowest unoccupied molecular orbital (LUMO) gap reduction with increasing number of acene units, accessible on the single molecule level by scanning tunneling microscopy (STM) and atomic force microscopy [17][18][19][20][21]. However, often it is preferred to know the assembly, structure and electronic properties of monolayer films on surfaces, as those are responsible for energy level alignment or determine the growth and structure of multilayer thin films, which is desired knowledge for integration into optoelectronic devices. In the case of heptacene (7 A), evaporation from the diheptacene precursor, which decomposes into the monomer at high temperature, has proven successful to directly grow monolayer films on metal substrates [22][23][24], but has not yet been demonstrated for passivated surfaces.
For example, the adsorption and the location of the energy levels of 7 A on Cu(110) have been recently reported [22]. Due to the rectangular unit cell of the Cu(110) surface and the elongated shape of the 7 A molecule, a preferential orientation on the surface, as in the case of the shorter acenes (5 A, 6 A) [11,[25][26][27] can be expected. Indeed, in the work of Boné et al [22], STM and photoemission orbital tomography (POT) were used to find the preferential orientation of the 7 A molecule with its long molecular axis along the long axis of the Cu(110) unit cell and its aromatic plane parallel to the surface. However, especially at low temperatures (at room temperature and below), molecules may also appear in the adsorption structure rotated 90 • with respect to this. It is interesting to note that these two geometric adsorption states are also characterized by different charge transfer from the Cu surface to the molecules. While in the preferred orientation the LUMO is fully occupied and the LUMO + 1 is partially occupied, in the 90 • rotated molecule only the LUMO is occupied. The moleculesubstrate interaction is weaker on Ag(110) surfaces, where the 7 A LUMO gets only partially occupied [23], and for Au(111) [20], STM investigations suggest that electron transfer is completely disabled.
The preceding discussion shows that the electronic structure of the 7 A molecule is strongly altered by adsorption on clean metal surfaces. It is often useful to minimize this influence by decoupling the molecular layer from the metal surface by a thin dielectric layer. In the case of Cu(111) surfaces, this can be achieved, for example, by vapor deposition of a thin NaCl layer, and it should be noted that using this substrate system, it was possible for the first time to image electronic states (LUMO and HOMO) of pentacene using STM [28]. A simpler method is the passivation of the Cu by oxygen, and it has previously been shown for various molecular adsorbates that sufficient decoupling can be achieved [29][30][31][32]. In the present study, and for direct comparison with 7 A on Cu(110), we have chosen the oxygen reconstructed Cu(110)-(2 × 1)-O surface for decoupling [6]. Here, an interesting aspect is the formation of the so-called 'stripe' phase at oxygen coverages below the half-monolayer oxygen coverage. The periodicity of the stripes is around 7 nm, with respective widths of the stripes being tunable by the amount of oxygen exposure [6,7]. The stripe phase can be regarded as a nanostructuring of the Cu(110) surface [7,33,34].
The motivation for the present work arises from the above considerations. On the one hand, it will be verified that adsorption of 7 A on the fully covered Cu(110)-(2 × 1)-O surface achieves sufficient electronic decoupling and a changed orientation of the 7 A molecules due to the geometry of the unit cell. For this, STM and POT in combination with calculations using density function theory (DFT) are applied to resolve the geometric and electronic properties of 7 A. In addition, the possibility of preferential adsorption on the Cu(110)-(2 × 1)-O stripe phase, consisting of regions exposing the clean Cu(110) surface and the decoupling Cu(110)-(2 × 1)-O, and stabilization of extremely different charge states of 7 A molecules in neighboring areas, will be demonstrated.

Experimental details
All experiments were carried out under ultra-high vacuum conditions in two separate chambers, one for photoemission experiments and one for STM experiments. All preparation steps (sample cleaning, molecule deposition) were performed in dedicated preparation chambers directly attached to the STM and photoemission chambers. The Cu(110) single crystal surface was cleaned by repeated cycles of Ar + sputtering and annealing to 800 K until a sharp low energy electron diffraction (LEED) pattern was observed. The Cu(110)-(2 × 1)-O reconstruction was created by exposure of the clean Cu(110) surface to oxygen for 5 min at an O 2 background pressure of 3 × 10 −8 mbar and a substrate temperature of 573 K. This procedure yielded a completely O-covered surface. For the Cu(110)-(2 × 1)-O stripe phase the exposure time was reduced correspondingly. In the STM experiments, the successful creation of the stripe phase was observed by direct imaging, while in the photoemission experiments the change in work function was used as a measure of the oxygen coverage (supporting information S1) [34]. Heptacene (7 A) was deposited onto the surface by thermal evaporation of the diheptacene precursor from a home-made molecular evaporator. Prior to the deposition experiments, the evaporator was carefully degassed. Typical background pressure during 7 A deposition was in the range of 10 −9 mbar and the deposition rate was determined with a quartz crystal microbalance.
The STM experiments were performed with a CreaTec low-temperature STM using electrochemically etched tungsten tips. The bias voltage was applied to the sample and all images reported in this work were acquired at 77 K. Scanning tunneling spectroscopy (STS) was performed by measuring the dI/dV spectra using a lock-in amplifier (f mod = 877 Hz, A mod = 50 mV). The STS data presented in this work are normalized to the current signal.
POT measurements were performed at the NAWI Graz NanoPEEM core facility with the NanoESCA system by ScientaOmicron. The He I radiation from an unpolarized HIS 14 HD excitation source (Focus GmbH) was incident at an angle of 68 • to the surface normal. Due to the focusing mirror, the sample was illuminated by 30.6% s-polarized and 69.4% p-polarized light. This results in a decrease of intensity in the lower half of the momentum maps due to the polarization factor |A·k| 2 and a corresponding asymmetry of the momentum maps. The work function was determined from the secondary electron cut-off in a sample bias configuration.

Computational details
All calculations were performed within the framework of DFT employing the Vienna ab initio simulation package version 5.4.4 [35,36]. Exchange-correlation effects were approximated by the functional of Perdew-Burke-Ernzerhof [37] and van der Waals contributions were treated with Grimme's D3 dispersion correction [38]. We used the projector-augmented wave (PAW) method [39] assuming an energy cutoff of 400 eV. The full monolayer of 7 A on the (2 × 1)-O reconstructed Cu(110) surface was simulated in accordance with the experimental results with the 7 A molecules oriented along the O rows. We utilize the repeated slab approach using five substrate layers and a 30 Å vacuum layer. To prevent disturbing spurious electrical fields, a dipole layer was placed in the vacuum region [40]. The ionic positions of the structure were optimized on a Gamma-centered grid of (2 × 8 × 1) k-points until the remaining forces were below 0.01 eV Å −1 . During the optimization, the two bottom Cu layers of the slab were constrained. To evaluate the molecular orbital-specific dispersion, we recalculated the electronic properties of a freestanding 7 A monolayer on a denser k-mesh of 5 × 20 × 3 points. Based on this calculation, the angle-resolved photoemission intensity was simulated within the one-step model of photoemission [41] under the assumption that the wave function of the final state can be described as a plane wave [27].

Results and discussions
To demonstrate the impact of chemical modification and patterning of the Cu(110) surface on the orientation and electronic structure of 7 A molecules, we start by presenting STM results for the growth of the molecules on clean Cu(110) and a fully oxygen-covered Cu(110)-(2 × 1)-O surface, before discussing it for the stripe phase later. In figures 1(a)-(c), STM images of 7 A/Cu(110) at increasing 7 A coverage are shown, where the 7 A molecules are identified as rod-like protrusions. At low coverage (0.2 monolayers (ML), figure 1(a)), the molecules are aligned with their long molecular axis along the [110] substrate direction, that is, along the close-packed Cu atomic rows, in agreement with previous studies [22]. In addition to some isolated molecules, molecular pairs and larger aggregates arranged parallel in a row-like fashion are formed, which points to intermolecular interaction. The perpendicular distance between individual molecules in the pairs and larger row structures is 1.1 nm, which corresponds to three lattice distances in the [001] direction (figure 1(f) and supporting information S2(a)). Increasing the 7 A coverage to 0.7 ML leads to a denser packing ( figure 1(b)). However, the perpendicular distance between the molecules does not decrease uniformly, but has discrete values. While structures with a 1.1 nm separation are still present, the distance in the denser structures reduces to 0.74 nm, which is two lattice distances in [001] direction (figure 1(f) and supporting information S2(b)). At this coverage, most of the molecules show the preferred alignment along the [110] direction, but some are also found in the perpendicular arrangement, which is expected for deposition at room temperature according to previous photoemission studies [22]. To form a single orientation monolayer, a higher coverage deposition was performed at increased substrate temperature. The STM image of this preparation (figure 1(c)) shows the dense arrangement of molecules in the monolayer, with a perpendicular separation of 0.74 nm, and also some second layer molecules in a molecular arrangement similar to that found in the monolayer. Our results generally agree with former STM studies of the 7 A monolayer on Cu(110) [22]. In addition, we have performed STS to probe the electronic structure of the 7 A/Cu(110) system. As shown in figure 2 (red STS curve), no clear dI/dV peaks are seen around the Fermi level (E F ); however, shoulders are recognizable at bias voltages around +0.5 V and −1 V, respectively. As it is difficult to assign specific resonances without additional information, we refer to previous POT and DFT studies of 7 A/Cu(110) [22]. There it was found that significant charge transfer on hybridization leads to full occupation of the LUMO, which is shifted below E F , and significant partial occupation of the LUMO + 1, which is pinned to E F . The corresponding energies are indicated in figure 2. According to this comparison, we assign the shoulder at −1 V to a resonance of the occupied 7 A LUMO. There is strong substratemediated intermolecular dispersion for the LUMO + 1 (and to a lesser extent for the LUMO) observed in the band maps of 7 A/Cu(110) (supporting information S3). This presumably mediates the 7 A stacking and can explain the indistinct appearance of the LUMO + 1 in the STS.
To create a substrate that sufficiently decouples the 7 A molecules from the metal, the Cu(110) surface was oxidized at 773 K, which produces a complete monolayer of Cu(110)-  image in figure 1(e). The perpendicular intermolecular separation is 0.51 nm, which perfectly fits the added row periodicity (figure 1(f) and supporting information S2). We will show further below that such a tight packing is only possible because of an adsorption geometry with a significant tilt of the molecules. Furthermore, figure 1(e) reveals a regular lateral displacement of neighboring 7 A molecules in [001] direction, similar to what has been observed for 5 A on anatase TiO 2 (101) [44]. The corresponding angles with respect to [110] are ± 35 • , which corresponds to a shift of neighboring molecules by one lattice constant in [001] ( figure 1(f)). This information is relevant to understanding the electronic structure data, which will be presented below.
To prove that the 7 A molecules are electronically decoupled from the underlying metal substrate, STS was performed. As shown in figure 2 (blue curve), the dI/dV trace has two peaks, one below E F at V bias = − 0.85 V, which we tentatively assign to a resonance of the molecular HOMO, and one above E F at V bias = + 1.3 V corresponding to the LUMO. This assignment is corroborated by the POT analysis presented below and the calculated density of states (see supporting information S4).  The electronic properties of 7 A on the Cu(110)-(2 × 1)-O surface have additionally been investigated using UPS and POT. To this end, we compare in figure 3(a) the angleintegrated photoemission spectra and the orbital-resolved signal contributions obtained by k-space deconvolution [45] using the momentum maps of the orbitals of planar 7 A, for 7 A on Cu(110), and tilted 7 A for Cu(110)-(2 × 1)-O (Supporting Information S5 and S6). As shown previously, the signals at 0.15 eV, 1.10 eV and 1.4 eV binding energy (BE) in the spectrum of 7 A on Cu(110) can be assigned, by comparison with simulated maps of the 7 A molecular orbitals, to emission from the occupied former LUMO + 1 and LUMO, and the HOMO states, respectively (figure 3(b)) [22]. The distinction between LUMO + 1 and LUMO is based on the slightly different k of their emission maxima of 1.4 Å −1 and 1.3 Å −1 , respectively, due to the different number of lobes of the corresponding molecular orbitals. The emission of the HOMO has two maxima in the k map mirrored around k [001] = 0, reflecting the nodal plane on the long molecular axis.
Turning now to the spectrum of 7 A on Cu(110)-(2 × 1)-O, emission peaks appear at 0.7 eV and 1.7 eV BE, respectively. Identification of the orbitals, from which the emissions arise, is again performed by comparing experimental and simulated k maps ( figure 3(c)). Firstly, we note that compared to 7 A on Cu(110), on Cu(110)-(2 × 1)-O the molecular emissions appear 90 • rotated, from which we identify the long molecular axis of 7 A to be parallel to the [001] direction, in agreement with the STM data. At first glance these features appear to present momentum maps similar to the LUMO and LUMO + 1 on Cu with strongest intensity on the k [001] axis. However, their k values of 1.2 Å −1 and 1.0 Å −1 , respectively, are too small for the LUMOs and are close to the values expected for the HOMO and HOMO-1.
To resolve this discrepancy we take into consideration that due to the large corrugation of the reconstructed Cu(110)-(2 × 1)-O surface, the 7 A molecules may be adsorbed in a tilted geometry. Such molecular tilt strongly affects the photoemission distribution as the molecular orbital lobes in kspace are no longer perpendicular to the surface. This leads to emissions that lie either side of the axis for a flat lying molecule shifting onto the axis, as previously shown for pentacene [46,47]. Considering simulated k maps of the 7 A HOMO emission for different tilt angles (reported in the supporting information S7), we find best agreement with the experimental data for a tilt angle of 37 • , which perfectly reproduces the experimental HOMO and HOMO-1 k maps ( figure 3(c)). Together with the information from STM that the distance between neighboring 7 A molecules is 0.51 nm, corresponding to the periodicity of the reconstructed surface in the [110] direction, we derive the adsorption geometry of 7 A on Cu(110)-(2 × 1)-O shown in figure 3(d).
The close intermolecular spacing enforced by the Cu-O rows, along with the associated tilting of the aromatic planes, leads to direct π-π orbital overlap between neighboring molecules ( figure 3(d)). This will lead to strong πband dispersion in the 2D molecular monolayer. The degree of π-π overlap and concomitant dispersion is also influenced by the lateral shift between neighboring molecules. This is illustrated in figure 4(a) for the lateral shift suggested by the STM results of figure 1(e). Owing to the different number of nodes in the HOMO and HOMO-1, and the given displacement of two adjacent molecules, nodes and maxima of two adjacent HOMOs lead to a small hopping integral, while for the HOMO-1 this intermolecular overlap is maximized. This is born out in the calculations of the electronic structure of a free standing 7 A layer with this unit cell geometry, which shows a small dispersion (0.3 eV) for the HOMO derived band, while the HOMO-1 band has a dispersion of more than 0.5 eV. Figure 4(b) displays a 2D representation of the HOMO-1 derived band dispersion with energies expressed with a color code. As the band structure will only be observable in k-space regions where there is significant photoemission cross-section, these are effectively determined by the momentum maps of the isolated molecular orbitals [27]. The experimental band maps (E vs. k) obtained at slices through the momentum map intensity maxima of the HOMO and HOMO-1 at k = 1.2 Å −1 and k = 1.0 Å −1 , respectively, indeed display a weak dispersion in the HOMO and a strong one for the HOMO-1 and are in good agreement, both in terms of extent and intensity, with the theoretical band maps (figure 4(c)). A compilation of 2D HOMO and HOMO-1 band dispersions and theoretical band maps for free standing 7 A monolayers with a range of lateral shifts can be found in the supporting information S8. It should be noted that calculations of the electronic structure of the molecular monolayer on the Cu(110)-(2 × 1)-O substrate show no appreciable difference to that of the free standing monolayer proving that the oxygen reconstruction effectively decouples the molecules from the substrate.
After having understood the geometric and electronic properties on the Cu(110) and the fully oxygen-passivated substrate, now we turn to investigate how confinement effects on the Cu-O stripe phase may affect the interface properties. Figure 5(a) shows an STM image of the 0.5 ML stripe phase. The light areas correspond to the O-covered regions, and the dark areas correspond to the clean Cu(110) surface. The stripe period is 7 nm and the width of the oxygen stripes varies between 5 and 7 lattice constants, corresponding to 2.5-3.5 nm. After evaporation of 0.5 ML 7 A at room temperature, a surface structured with 7 A-covered Cu(110) regions and uncovered Cu(110)-(2 × 1)-O regions is obtained ( figure 5(b)) due to the larger adsorption energy of 7 A on the clean Cu(110), and, thus, preferential adsorption into these regions. The Cu(110) stripes are so narrow that there is room for only one molecule in the preferred adsorption direction. Due to geometric limitation, remaining free areas in the Cu(110) stripes, especially at the phase boundaries, are occupied by molecules rotated by 90 • .
Significantly, molecular adsorption has not disturbed the template and individual rows of 7 A could be produced. To achieve coverage also of the oxygen stripes, additional 7 A was evaporated onto the sample. In the STM image in figure 5(c), the different regions can be distinguished according to the orientation of the molecules. As described above, the Cu(110) regions are occupied by 7 A molecules aligned preferentially along the [110] direction. The heterogeneity in these regions in figure 4(c) results from the partial second layer formation. In the Cu-O regions, the molecules adopt the preferred 90 • rotated orientation, that is, along [001], with the same perpendicular distance (0.51 nm) and similar lateral shift of neighboring molecules as reported above for 7 A adsorption on the full Cu(110)-(2 × 1)-O substrate.
Finally, we check if the spatial confinement affects the electronic properties of the molecules. For this, we show in figure 6 the results of photoemission studies for a similar 7 A/stripe preparation as observed in STM. The black spectrum represents the angle-integrated photoemission spectrum in the valence region, where the frontier orbital molecular emissions should appear. In contrast to the results of figure 3 for the pure phases, the spectrum in figure 6 does not exhibit clear peaks. Despite this, already a visual inspection of the momentum maps recorded through this BE range (see supporting information S5) reveals distinct orbital signatures, akin to the ones observed for 7 A on the pure surfaces. This allows a k space deconvolution to be performed, the result of which is presented in figure 6 as colored lines for the contributions of 7 A on Cu and on Cu-O. Comparison with the k space deconvolution of the pure phases in figure 3 reveals that the indistinct electron energy distribution curve (black) is simply a sum of the contributions, with similar energy positions and spectral widths, of the differently oriented and charged 7 A molecules on the individual phases. Thus, it appears that there are no significant confinement effects.

Conclusions
We have investigated the geometric and electronic properties of 7 A on Cu(110), the passivated Cu(110)-(2 × 1)-O, and the Cu(110)-(2 × 1)-O stripe phase. While on Cu(110) the molecules interact strongly with the substrate, with almost three electrons transferred into the LUMO and LUMO + 1, the half monolayer of oxygen suffices to completely decouple the 7 A molecules, which remain uncharged. The uncharged molecules reorient along the Cu-O rows of the reconstruction and molecule-molecule interactions dominate, giving rise to strong π-π interaction between the now tilted molecules and concomitant π band dispersion.
In addition, it is shown that the Cu(110)-(2 × 1)-O stripe phase provides a stable nanoscopic template with lateral dimensions on the molecular scale. Despite their small dimensions, the stripes retain the properties of the original surfaces upon 7 A adsorption, leading to an array of alternating stripes occupied with neutral and more than doubly charged 7 A molecules, respectively, with no evidence for confinement effects despite the proximity of charged and uncharged molecules. These results suggest that the Cu-O stripe phase could provide a useful template for the investigation of onsurface reactions in confined areas or for templated threedimensional growth of heterogeneous systems.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.