Employing X-ray Photoelectron Spectroscopy for Determining Layer Homogeneity in Mixed Polar Self-Assembled Monolayers

Self-assembled monolayers (SAMs) containing embedded dipolar groups offer the particular advantage of changing the electronic properties of a surface without affecting the SAM–ambient interface. Here we show that such systems can also be used for continuously tuning metal work functions by growing mixed monolayers consisting of molecules with different orientations of the embedded dipolar groups. To avoid injection hot-spots when using the SAM-modified electrodes in devices, a homogeneous mixing of the two components is crucial. We show that a combination of high-resolution X-ray photoelectron spectroscopy with state-of-the-art simulations is an ideal tool for probing the electrostatic homogeneity of the layers and thus for determining phase separation processes in polar adsorbate assemblies down to inhomogeneities at the molecular level.


S2. Details about the quantum-mechanical simulations
The simulations have been performed in the framework of density functional theory (DFT) using the VASP 2-5 package (version 5.3.2) and employing the repeated slab approach decoupling periodic replicas of the slab by a vacuum gap of ≈20Å in z-direction. The vacuum gap contained a self-consistently determined dipole layer. 6 For relaxing the geometries, we employed the advanced optimization tool GADGET 7 which enables performing geometry optimization in internal coordinates instead of Cartesian ones. The initial guess for the Hesse matrix was made according to Fischer's model. 8,7 We used the PBE functional 9 together with PAW 10,11 potentials (see Table S1) and a cutoff energy of 300 eV for the plane wave basis set.
For the geometry optimization we chose a convergence criterion of 10 -2 eV Å -1 for the maximum residual force per atom. The total energy in the self-consistency cycle of each DFT step was converged to at least 10 -6 eV enforcing at the same time a convergence of the zcomponent of the dipole moment per unit to 10 -4 eÅ . For the smallest considered surface unit cell (√3 x 3), a 8x5x1 Monkhorst-Pack 12 k-point grid was chosen, which was suitably rescaled for larger cells. No explicit van der Waals correction was included, as preliminary tests using the vdW-surf 13 approach for the pure films resulted in tilt angles clearly higher than the experimental ones. To what extent this is a consequence of an inappropriate description of the screening of the van der Waals interactions between the substrate and the adsorbate by the comparably thick adsorbate layer will require further investigations.
The unit cells included five layers of gold substrate with the topmost two layers allowed to relax in the geometry optimizations. The used gold lattice constant was optimized by applying the same methodology to avoid spurious surface relaxations. A value of 4.175 Å was found. As a starting condition for the geometry optimization the molecules were placed on fcc hollow sites of the Au(111) surface; in the course of the geometry optimization, they S4 moved to an intermediate position between hollow and bridge sites. For all investigated systems, except the striped phase, a geometry optimization was performed. The unit cell of the striped phase was too large for a geometry optimization of the whole system. Instead, this unit cell was created from the individual (geometry optimized) unit cells of the pure TP1up and TP1-down SAMs. Core-level energies were calculated employing the initial state approach, 14 which for systems like the present one is preferable over final state approaches for avoid charge-transfer dipole related artefacts. 15 Screening effects were included through an image charge model 16,17 in a post-processing step assuming a dielectric constant of the SAM of ε=3.0. XP spectra were calculated as described in ref 15 . In short, they were obtained from the individual core level energies employing a Gaussian broadening (variance = 0.1 eV), setting the incident photon energy to hν=350eV (in accordance to experiment) and including exponential damping 18 using a dimensionless damping coefficient β=0.65, 15 adjusted to reproduce the experimental spectra (for details see ref. 18 ). A variance of 0.1 eV was chosen for the Gaussian broadening, to fit the peak width of the calculated XP spectrum of the pure TP1-up SAM to the experimental spectrum. The same variance was then used to evaluate all other calculated spectra.   The core level energies of the substrate-side rings ring (in Figure S2) are not exactly the same for molecules of type TP1-up and TP1-down. This at the first glance appears somewhat counterintuitive, as the dipolar pyrimidine unit is located above this first ring and, therefore, should only energetically shift atoms above it. However, one can consider the whole SAM as a combination of two substructures, the TP1-up and TP1-down stripes. Therefore, the TP1up domains can be thought of as being embedded in the electrostatic potential of the surrounding TP1-down domains and vice versa. This influences the core level energies as described in detail for valence states in ref. 19 . This shift of the substrate side ring is not observed in the homogeneous 50:50 mixture (see Figure S3). In fact, in contrast to the striped SAM, the homogeneous 50:50 mixture (see Figure S3) shows no electrostatic shift of the upper phenylene ring between TP1-up and TP1down molecules. C1s core level energies in the top ring are the same for both types of molecules.

S5. Details on the spectral shape of pure and mixed SAMs
As was already discussed briefly in ref. 20 , the XP spectra of pure TP1-up and TP1-down SAMs contain several contributions. The main feature stems from the vacuum side phenyl ring due to the weakest associated attenuation of the photoemission signal. The deeper lying pyrimidine unit and the substrate side phenyl ring, which have different binding energies, give rise to the additional weak features in the vicinity of the main peak (also in part overlapping with the main peak). As the electrostatic shift of the top phenyl ring is of opposite sign for the TP1-down and TP1-up films, the additional features are situated at higher binding energies (more negative core level energies) than the main peak for TP1down and at lower binding energies (less negative core level energies) for TP1-up. This can also be clearly discerned in the calculated spectra for the pure films shown in Figure 3a of the main paper. The effect is still present for the mixed SAMs shown in Figure 3a, although less pronounced.
Upon closer inspection it is noticeable that the "shoulder" extends over a somewhat larger energy range for TP1-down than for TP1-up. This is caused by the orientation of the S9 pyrimidine dipole. In TP1-down the vacuum side phenyl ring is shifted to less negative C1s energies than the substrate side ring. The carbons in the pyrimidine unit however are chemically shifted to more negative core level energies compared to the substrate side ring (see previous chapter). In the case of TP1-up the chemical shift in the pyrimidine ring as well as the electrostatic shift of the top phenyl ring move the C1s energies to more negative values compared to the substrate side ring. As a consequence, in the TP1-up case the signals of the pyrimidine unit and the vacuum side phenyl ring partially overlap as they are closer in energy.
This observation also explains the slightly different peak heights found in the simulated spectrum of large domains (gray curve in Figure 3b of the main paper) produced as weighted superpositions of the pure TP1-up and TP1-down spectra.

S6. FWHM of the XP spectra of all investigated systems
The FWHM of the main peak of the XP spectra has be obtained in two ways: First by directly measuring the full width at half maximum of the highest peak in the spectrum; secondly, by fitting the original spectrum using 3 separate Gaussian peaks and then determining the FWHM of the dominant feature representing core-level excitations from the topmost ring. to a situation with all dipoles oriented parallel. The total cluster size in each case was chosen so that the electrostatic potential was converged to 0.0001 V. Figure S3 shows the calculated electrostatic potential for the various tested configurations relative to the situation with all dipoles aligned in the same direction.  Figure S4 shows that square and striped domain structures create similar electrostatic potentials above the 2D array. The difference in potential between the two configurations decreases with increasing domain size. These results show that using infinite stripes as domains is a reasonable approximation for what one can expect to encounter also for other domain shapes; i.e. striped domains do not give rise to any artifacts that qualitatively change the behavior of the monolayer.