Figure 1a and Figure 1b report the structural characterization of the ZnTPP/Fe(001)–p(1 × 1)O sample in the reciprocal and in direct space, respectively. The low-energy electron diffraction (LEED) pattern acquired on the ZnTPP/Fe(001)–p(1 × 1)O sample is characterized by a well-defined square lattice, where several diffraction orders are visible. Intense spots corresponding to the square lattice of the Fe(001)–p(1 ×1)O surface are marked with circles on the periphery of the screen. The coexisting LEED patterns of the Fe(001)–p(1 × 1)O surface and of the ZnTPP film allow for the quantitative evaluation of the overlayer lattice constant, which indicates that the molecules arrange themselves in a (5 × 5) commensurate array with respect to the Fe(001)–p(1 ×1)O surface, in agreement with previous results [46]. This order extends over large domains (hundreds of square nanometers wide) and tends to disappear as soon as additional molecules are deposited on top of the wetting layer. The formation of a well-ordered ZnTPP film with (5 × 5) periodicity is confirmed by the STM image displayed in Figure 1b, where individual ZnTPP molecules are resolved.

Figure 1: (a) LEED pattern of the system 1 ML ZnTPP /Fe(001)–p(1 × 1)O acquired with a beam energy equal to 55 eV. In the circles, the spots from the bare Fe(001)–p(1 × 1)O surface are highlighted. The arrows indicate the unit vectors in the reciprocal space of the substrate (long arrows) and after the deposition of the organic film (short arrows). (b) STM image of the ZnTPP overlayer. Tunneling parameters V = 1.5 V, I = 500 pA, image size 11 × 11 nm2. The red square indicates the (5 × 5) unit cell. In the lower right corner, the crystallographic directions are indicated.

Figure 2 focuses on the surface morphology for a sub-monolayer coverage of C60 on the ZnTPP/Fe(001)–p(1 × 1)O substrate. C60 forms a compact film, composed of hexagonal domains extending for hundreds of nanometers. By considering that the deposition has been performed with the substrate kept at room temperature, we can estimate that Ed for C60 diffusing on ZnTPP is significantly lower than 25 meV. It is worth to notice that the ZnTPP buffer layer remarkably decreases Ed with respect to the case of C60 deposited at room temperature directly on either the Fe(001) or Fe(001)–p(1 × 1)O surfaces. In the former case, the diffusion of C60 is completely hindered and fullerene forms a disordered film, while in the latter case a peculiar mode of growth, intermediate between diffusion-mediated and ballistic growth, is observed [23,50]. Figure 2b shows a blowup of one fullerene domain, where individual C60 molecules are visible inside a hexagonal lattice with a lattice parameter of about 1 nm, a value very similar to that measured in C60 films stabilized on either metallic [51] or oxide [25] substrates. Figure 2c shows the fast Fourier transform (FFT) calculated from the image reported in Figure 2a. Four hexagonal domains can be identified, differing by their angular orientation with respect to the substrate. Interestingly, the domains do not possess a well-defined epitaxial relation with respect to the (5 × 5) lattice of ZnTPP, indicating a weak interaction between the C60 film and the ZnTPP substrate, as confirmed by the spectroscopic measurements presented in the following.

Figure 2: (a) Large-scale STM image of a C60 wetting layer deposited on C60/Zn-TPP/Fe(001)-p(1 × 1)O. In the left top corner of the image, the ZnTPP layer is visible. (b) Zoomed image of the region marked by a dashed square in panel (a). (c) FFT of the image in panel (a), where four differently oriented hexagonal domains are marked. The rotation angles between the white and the pink, blue, and green hexagonal domains are 10°, 33°, and 44°, respectively. STM images have been acquired at V = 1.5 V and I = 500 pA.

The UPS spectra acquired on Fe(001)–p(1 × 1)O, ZnTPP Fe(001)–p(1 × 1)O, 1 ML C60/ZnTPP/Fe(001)–p(1 × 1)O, and 20 ML C60/Fe(001)–p(1 × 1)O samples are shown in Figure 3. The spectrum of Fe(001)–p(1 × 1)O is dominated by a large peak located at about 4.2 eV, which is attributed to O 2p states. This feature almost completely vanishes as soon as 1 ML of ZnTPP is deposited, indicating that oxygen remains buried at the ZnTPP/Fe(001)–p(1 × 1)O interface. In the 1 ML ZnTPP spectrum in Figure 3, the UPS peaks related to the main molecule ring and to the phenyl groups are labeled “R” and “Ph” [52,53], respectively, according to theoretical simulations performed on metal tetraphenyl porphyrins and metal porphyrins [54]. When an additional single layer of C60 is added to this system, new features appear. The photoemission signal from the underlying ZnTPP layer, albeit affected by the screening action of C60 (implying a rather large surface sensitivity of the technique, as also shown in [55] on a similar system), is still detected in those spectral regions not superimposed to the new C60 features. In particular, peaks “a” and “b” can be readily assigned to HOMO and HOMO−1 features and their energetic positions match with their equivalents when a very thick layer of C60 is grown directly on Fe(001)–p(1 × 1)O (top spectrum). The feature labeled “c” in Figure 3 is due to C 2p electrons [56]. Therefore, it is present with only slight modifications both in ZnTPP/Fe(001)–p(1 × 1)O and C60/ZnTPP/Fe(001)–p(1 × 1)O samples.

Figure 3: UPS spectra of the system Fe(001)–p(1 × 1)O at different coverages of ZnTPP and C60. The lowest spectrum is the one from the bare Fe(001)–p(1 × 1)O. The main features from Fe(001)–p(1 × 1)O (the peak due to oxygen, “O”), ZnTPP (both from the pyrrolic macroring, “R1” and “R2”, and from the phenyl subunits, “Ph1” and “Ph2”) and from C60 (“a”–“e”) are labeled and their evolution is indicated with dotted lines.

In order to determine the HOMO–LUMO gap of the C60 film, STS measurements have been acquired for both negative and positive bias to investigate the filled and empty electronic states, respectively. Figure 4 shows STS spectra acquired on the ZnTPP/Fe(001)–p(1 × 1)O surface (red) and on the C60/ZnTPP/Fe(001)–p(1 × 1)O system (black). The STS measurements acquired on ZnTPP/Fe(001)–p(1 × 1)O are in excellent agreement with those published in [43]. The STS curve referring to C60/ZnTPP/Fe(001)–p(1 × 1)O has been obtained by averaging several spectra acquired on equivalent C60 domains. We acquired also spectra in different locations of single C60 molecules, but not significant differences with a well-defined trend were observed. In the negative energy range (filled electronic states) a strong resonance centered at about −2.60 eV is present, which we attribute to HOMO states, in excellent agreement with UPS measurements (−2.56 eV). In the positive energy range (empty states) of the STS spectrum the LUMO peak is visible at 1.15 eV, resulting in an electronic gap equal to 3.75 eV.

Figure 4: Scanning tunneling spectrum acquired at constant tip–surface separation (open feedback loop) on the C60/ZnTPP/Fe(001)–p(1 × 1)O system (black) and on the ZnTPP/Fe(001)–p(1 × 1)O surface (red). The black curves have been obtained by averaging 30 single spectra taken on equivalent C60 domains. The set point before the acquisition of the spectra was set to V = 1.5 V and I = 1 nA.

Finally, work function measurements have been performed to evaluate the charge transfer between the different layers constituting the heterostructure. Generally, electron transfer from the substrate (overlayer) to the overlayer (substrate) induces an increase (decrease) of the work function with respect to the bare surface. For the work function measurements, the sample has been biased with a voltage of 10 V to detect the onset of the secondary electrons. The onset position is determined as the intersection of the zero-current line and the tangent to the rising edge of the data. Figure 5a displays a typical UPS spectrum in an energy range straddling the high-binding-energy cutoff of the secondary electrons, which we exploit for the evaluation of the work function for 1 ML ZnTPP/Fe(001)–p(1 × 1)O. The 10 eV offset due to the bias applied to the system has already been accounted for. In Figure 5b, the evolution of the work function for the different samples is presented. Starting from the bare substrate, the work function is reduced by about 0.3 eV after the deposition of 1 ML of ZnTPP, in agreement with previous measurements [42]. Such a decrease has been ascribed to charge transfer from ZnTPP to the Fe(001)–p(1 × 1)O substrate. When 1 ML of C60 is added, the variation of the work function is within the experimental error, indicating a negligible charge transfer on the surface region upon C60 adsorption.

Figure 5: (a) Work function acquired on the system 1 ML ZnTPP/Fe(001)–p(1 × 1)O. The two dashed lines indicate how the onset of the curve is determined. The 10 eV offset due to the bias applied to the system has been subtracted. (b) Summary of the values of the work functions acquired on the system Fe(001)–p(1 × 1)O at different molecular coverages.

The electronic properties of C60 adsorbed on ZnTPP deserve a deeper discussion. We recall that the difference between the energies of LUMO and HOMO orbitals of C60 at equilibrium is about γ = 1.6 eV, as determined experimentally [57] and theoretically [58]. However, the difference between the electron affinity and the ionization potential measured on isolated C60 (in the gas phase) is about Es = 4.95 eV [57], considerably higher than γ. This discrepancy is given by the fact that the ionization potential (electron affinity) is not simply the difference between the vacuum level and the HOMO (LUMO) energies of C60 at equilibrium, because an extra energy is required to remove (add) an electron from (to) the neutral molecule. Therefore, the gap measured with electron-based spectroscopic experiments is E = γ + U, where U is the on-site Coulomb energy [57]. The U term accounts for the fact that, when occupied states are probed, an electron is removed from the molecule, therefore the measured spectrum is not representative of the neutral but of the positively charged molecule. Similarly, when unoccupied states are probed, an electron is injected in the molecule and the system is negatively charged. For isolated C60 molecules, the charging energy is Us = Es − γ = 3.35 eV. In [57], Esper et al. measured γ by performing PES on C60 films highly doped with K. In this case, the LUMO orbitals were completely filled, therefore the charging energy was the same when HOMO and LUMO states were probed and the difference between the LUMO and HOMO energies was independent from U.

Generally, when C60 is adsorbed on a substrate, the U term is drastically reduced by the electrostatic screening provided by the metallic or molecular support. In the former case, when an electron is added or removed from C60, the charged molecule is screened by an opposite image charge underneath the metal surface, while, in the latter case, the screening is provided by electric dipoles induced on the organic substrate. In order to evaluate the coupling between C60 and the substrate, it is useful to quantify the reduction of the electronic gap E (or equivalently of the U term) with respect to that of the isolated molecule. In the case of the (111) surface of face-centered cubic bulk C60, the measured electronic gap is Eb = 3.50 eV [23]. Therefore, the charging energy is Ub = 1.90 eV. Defining ΔU as the variation of the Coulomb energy with respect to isolated C60, in the case of bulk C60, it is found ΔU = Ub − Us = −1.45 eV. Such a decrease of U can be ascribed to the polarization of the nine molecules surrounding each C60 located at the surface, six belonging to the topmost layer and three to the second layer. By considering an equal contribution for each molecule, every C60 provides a screening of about ΔU = −0.16 eV.

Starting from this observation, it is possible to evaluate the screening provided by the Fe(001)–p(1 × 1)O and ZnTPP/Fe(001)–p(1 × 1)O substrates on the C60 film (see Table 1). To this aim, we can assume that ΔU is the sum of two contributions, the first one due to the screening provided by six surrounding C60 molecules (ΔUsurf) and the second one provided by the substrate (ΔUsub). As for ΔUsurf, we consider for each sample the same value as found in bulk C60(111), because C60 forms a hexagonal lattice also on top of the other substrates. In the case of C60/Fe(001)–p(1 × 1)O, it is found ΔUsub = −0.59 eV. Therefore, the oxygen-passivated Fe(001) surface provides a higher screening with respect to a fullerene substrate. In contrast, for the C60/ZnTPP/Fe(001)–p(1 × 1)O sample, it is found ΔUsub = −0.24 eV, indicating a very low screening induced by the porphyrin buffer layer, even with respect to that provided by a substrate of bulk C60.

Table 1: Electronic coupling of C60 with the Fe(001)–p(1 × 1)O and ZnTPP/Fe(001)–p(1 × 1)O substrates. E is the energy gap measured by electron-based spectroscopic techniques. U = E − γ is the on-site Coulomb energy, where γ = 1.6 eV is the HOMO–LUMO energy difference at equilibrium. ΔUsurf and ΔUsub are variations of U with respect to the value of isolated C60 due to the topmost layer and the substrate, respectively.

System E (eV) U (eV) ΔUsurf (eV) ΔUsub (eV) isolated C60[42] 4.95 3.35 0 0 C60 bulk [22] 3.50 1.90 −0.96 −0.49 C60/Fe(001)–p(1 × 1)O [22] 3.40 1.80 −0.96 −0.59 C60/ZnTPP/Fe(001)–p(1 × 1)O 3.75 2.15 −0.96 −0.24