Changes in adsorption heights upon self-assembly of bicomponent supramolecular networks

Codeposition of two molecular species [CuPc (donor) and PFP (acceptor)] on noble metal (111) surfaces leads to the self-assembly of an ordered mixed layer with maximized donor-acceptor contact area. The main driving force behind this arrangement is assumed to be the intermolecular C-F...H-C hydrogen-bond interactions. Such interactions would be maximized for a coplanar molecular arrangement. However, precise measurement of molecule-substrate distances in the molecular mixture reveals significantly larger adsorption heights for PFP than for CuPc. Most surprisingly, instead of leveling to increase hydrogen bond interactions, the height difference is enhanced in mixed layers as compared to the heights found in single component CuPc and PFP layers, resulting in an overall reduced interaction with the underlying substrate. The influence of the increased height of PFP on the interface dipole is investigated through work function measurements.


INTRODUCTION
The electronic properties of metal/organic interfaces define the charge carrier injection barriers in organic optoelectronic devices, which are key parameters for their efficiency. Such properties depend largely on the detailed interfacial structure, defined by the molecular orientation with respect to the substrate, the lateral distribution and the molecule-substrate distance. The latter in particular plays a central role in the interfacial energy level alignment. This can be directly concluded from the electronic interface model put forward by Vazquez et al., 1,2 which includes all of the commonly assumed processes determining potential changes at non-chemisorptive interfaces, such as permanent molecular dipoles, the "pillow effect" and interfacial charge transfer. The pillow effect arises from the compression, upon molecular adsorption, of the electron wavefunctions decaying into the vacuum from the metal surface.
This compression modifies the surface component of the work function of the metal, and, being driven by the Pauli repulsion exerted by the adsorbates, it depends on the molecule-substrate distance. Charge transfer is determined by the induced density of interfacial states (density of states in the molecular gap, related to the shift and broadening of the molecular levels interacting with the metal) and a screening parameter. Both depend on the molecule-substrate distance as well. The most precise way to determine that distance experimentally is by means of normal incidence X-ray Standing Waves (XSW). 3,4 This technique has been applied to a number of molecules on various surfaces. [5][6][7][8][9][10][11][12][13] Here we report the use of XSW to determine the molecule-substrate distances in donor-acceptor molecular blends, which are not only highly relevant for many organic devices, but also have interfacial properties often differing from those of the corresponding single component layers. 14 In this work we provide a complete structural characterization of the first monolayer of a stoichiometric 1:1 donor-acceptor mixture assembled on Ag(111) and Cu(111) substrates. The molecules used are copper phtalocyanine (CuPc, donor) and perfluoropentacene (PFP, acceptor), two piconjugated molecules known for their successful integration in optoelectronic devices. 15,16 The lateral order of the 2D blends and the molecule-substrate distances have been characterized using scanning tunneling microscopy (STM) and normal-incidence XSW, respectively. The former evidences highly crystalline molecular networks, in which each molecule is surrounded by the opposite species. The latter reveals how the molecules in the blends significantly change their molecule-substrate distances with respect to those in single component layers. Finally, we analyze the impact of those changes on the interface electronic properties by means of photoemission work function variation measurements.

EXPERIMENTAL DETAILS
All experiments were performed in ultrahigh vacuum chambers with a base pressure in the low 10 -10 mbar range. The Ag(111) and Cu(111) substrates were cleaned by standard Ar-ion sputtering and annealing cycles. Evaporation of molecules onto the clean substrates held at room temperature was monitored by a quartz crystal microbalance. STM measurements were performed at room temperature on Cu(111) and at 80 K on Ag(111), using a commercial Omicron VT-STM and electrochemically etched tungsten tips in constant current mode. STM images were analyzed using the WSxM software. 17 Errors in the determined unit cell parameters are the standard deviation obtained from several STM images.
Normal Incidence X-ray Standing Wave measurements were performed at the ID32 beamline of the European Synchrotron Radiation Facility in Grenoble, France. The beamline is equipped with a SPECS Phoibos 225 hemispherical electron analyzer, mounted at 90º with respect to the incoming beam, that can reach kinetic energies up to 15 keV and has energy resolution down to ∆E/E=10 -6 . The XSW measurements required photon energies around 2628.6 eV for Ag(111) and 2970.2 eV for Cu(111), corresponding to the (111) Bragg reflections of said substrates (111-layer spacings d 0 (Ag)=2.3586 Å and d 0 (Cu)=2.0871 Å). In order to minimize potential beam damage (see supporting information for more details), irradiation time was kept as low as possible during each measurement, and subsequent measurements were taken at different positions on the sample. The resulting reflectivity and photoelectron yield curves were fitted using the pyXSW software developed by Jerôme Roy. Due to the system's geometry (see above), multipole correction parameters are close to zero and may be disregarded. 11,18,19 A commercial SPECS 10/35 UV source was used for the UPS measurements. The angle of incidence was 45º and electrons were detected close to normal emission. The sample was biased at -24.22V in order to shift the spectrum to higher kinetic energies to avoid instrumental modification of the analyzer transmission due to stray electric and magnetic fields. The cut-off position was determined by a sigmoid fit. The energy position of the cut-off in different measurements of the clean substrates was found to be reliable, giving a standard deviation of 0.01 eV for Ag(111) and 0.03 eV for Cu(111).

RESULTS AND DISCUSSION
Constant current STM images show that depositing PFP, CuPc or molecular blends on Ag(111) and Cu(111) leads to the formation of ordered molecular layers in which molecules adopt a flat-lying configuration. The influence of the substrate is seen from the discrete azimuthal domains related to the substrate symmetry and observed in all cases. The structure of PFP or CuPc monolayers on Ag (111) and Cu(111) have already been reported on in the literature. 9,[20][21][22][23][24] We therefore place our focus on the 4 mixed layers. When co-deposited (or deposited sequentially one after the other, which renders the same results) in an approximately 1:1 ratio, the two molecules form an ordered structure similar to that found for closely related systems, 14,25,26 in which molecules of one type surround themselves by the other type, maximizing donor-acceptor contact and -C-F···H-C-intermolecular interactions (Fig. 1). The experimental unit cell parameters are summarized in Table 1. These, in combination with the measured domain orientations, allow us to put forth rectangular, commensurate structures as tentative epitaxial models. These are depicted in Fig. 1c, and the corresponding parameters are included in Table 1.  Normal incidence x-ray standing wave measurements of the donor-acceptor mixtures were performed probing the C, N and F atoms. Upon X-ray irradiation of a single crystal in proximity to Bragg scattering conditions, the incident and scattered waves interfere to produce a standing wave with precisely defined periodicity (that of the scattering planes) and changeable phase. Atoms immersed in the standing waves show a photoemission yield that depends on their position relative to the wave field's nodal planes. XSW measurements consist in recording core-level photoemission yields in dependence of the standing wave phase, whereby one obtains element-specific information on the atomic location with respect to the substrate crystal planes.
Because of the molecular compositions (see Fig. 1a), the values extracted for the N and F atoms can be unambiguously ascribed to CuPc and PFP, respectively. However, a more complex scenario is found for C, which is contained in both molecules. Based on previous high-resolution XPS studies on PFP and CuPc monolayers and the 1:1 mixture, 14 the signal from each molecule should be disentangled as follows. Three separate peaks are resolved in the mixture's C1s photoemission spectra as illustrated in in where n is a natural number and d 0 is the periodicity of the standing wave. As can be seen from the above expression, d H can only be determined within a multiple of d 0 . In the case of organic molecular (sub)monolayers, expected adsorption heights are in the range between 2.2 and 3.6 Å. 28 d 0 is therefore sufficiently large to make all n values but one unreasonable. Here we safely assume n = 1 in all cases.
C.F., C.P. and d H values for all analyzed species are summarized in Table 2 and a schematic representation of the molecular heights in the mix is shown in Fig. 3. Au, Cu and Ag, 9,10 N remains slightly lower than the C atoms. This is expected, as they form the central cage with the Cu atom that interacts most strongly with the metal substrate. In the mixture, this configuration, as well as the height of the CuPc molecules, remains practically unchanged, indicating a stronger interaction with the substrate than for PFP.
In single-component layers, CuPc lies closer to the substrate surface than PFP, both on Ag(111) and Cu(111). 7-10 Upon blend formation, the raising up of PFP further increases the height difference between donors and acceptors (Fig. 3). The intermolecular C-F···H-C interactions assumed to drive the selfassembly of the highly crystalline donor-acceptor networks would be strongest in a coplanar arrangement, which reduces the bond distance and enhances its linearity. 29 They are therefore expected to tend to level the molecular heights in the blends. However, contrary to expectations, we find that the height difference between molecules is increased in the mixed layer. The driving forces behind these surprising changes are unclear. Our hypotheses point either to substrate-mediated effects or to halogen- interactions between the PFP fluorines with the -orbitals above the CuPc (C-F···. The former seems most intuitive and may arise from changes in the interface electronics, 14,30 which in turn modify the 8 molecule-substrate interactions and the associated adsorption distances. The latter would profit from an intermolecular height offset in the order of that found experimentally. As opposed to the C-F···H-C interactions, which would tend to level the molecular heights and are known to be amongst the weakest hydrogen bonds, 29,31,32 C-F··· interactions have been shown to play an important role in organic crystal packing and their additional contribution to the intermolecular interactions therefore seems another plausible explanation to our findings. 31  On the whole, the increased molecule-substrate distance of PFP and unchanged distance for CuPc suggest an overall reduced interaction of the mixed molecular layer with the underlying substrate.
Reduced molecule-substrate interactions as a result of enhanced intermolecular interactions in molecular mixtures have been reported before and are consistent with general chemical concepts. 33,34 However, this is the first quantitative report on the associated changes in the molecule-substrate distances of donor-acceptor blends, which are in turn of utmost importance for the understanding of the interfacial electronic properties. 1,2,14 The changes in molecular height and conformation found in the mixed layer are expected to lead to   We hereby provide a direct measure of the effect of a molecule's adsorption height on vacuum level shifts and, in turn, interfacial energy level alignment.

ASSOCIATED CONTENT
Supporting information: XSW error analysis and beam damage (Figures S1), work function error analysis. This material is available free of charge via the Internet at http://pubs.acs.org.