CP-AFM Molecular Tunnel Junctions with Alkyl Backbones Anchored Using Alkynyl and Thiol Groups: Microscopically Different Despite Phenomenological Similarity

In this paper, we report results on the electronic structure and transport properties of molecular junctions fabricated via conducting probe atomic force microscopy (CP-AFM) using self-assembled monolayers (SAMs) of n-alkyl chains anchored with acetylene groups (CnA; n = 8, 9, 10, and 12) on Ag, Au, and Pt electrodes. We found that the current–voltage (I–V) characteristics of CnA CP-AFM junctions can be very accurately reproduced by the same off-resonant single-level model (orSLM) successfully utilized previously for many other junctions. We demonstrate that important insight into the energy-level alignment can be gained from experimental data of transport (processed via the orSLM) and ultraviolet photoelectron spectroscopy combined with ab initio quantum chemical information based on the many-body outer valence Green’s function method. Measured conductance GAg < GAu < GPt is found to follow the same ordering as the metal work function ΦAu < ΦAu < ΦPt, a fact that points toward a transport mediated by an occupied molecular orbital (MO). Still, careful data analysis surprisingly revealed that transport is not dominated by the ubiquitous HOMO but rather by the HOMO–1. This is an important difference from other molecular tunnel junctions with p-type HOMO-mediated conduction investigated in the past, including the alkyl thiols (CnT) to which we refer in view of some similarities. Furthermore, unlike in CnT and other junctions anchored with thiol groups investigated in the past, the AFM tip causes in CnA an additional MO shift, whose independence of size (n) rules out significant image charge effects. Along with the prevalence of the HOMO–1 over the HOMO, the impact of the “second” (tip) electrode on the energy level alignment is another important finding that makes the CnA and CnT junctions different. What ultimately makes CnA unique at the microscopic level is a salient difference never reported previously, namely, that CnA’s alkyne functional group gives rise to two energetically close (HOMO and HOMO–1) orbitals. This distinguishes the present CnA from the CnT, whose HOMO stemming from its thiol group is well separated energetically from the other MOs.


Thickness of CnA Junction determined by XPS measurements
We utilized X-ray photoelectron spectroscopy (XPS) to confirm the presence of a CnA monolayer on metal surfaces and to estimate its thickness.By collecting photoelectrons at specific angles with respect to the surface, we increased the emission path lengths in the film by a factor of sinα, as illustrated in Figure S2A.The attenuation of the XPS signal was expressed as I = I 0 exp − d λ sin α where I 0 represents the photoelectron intensity from the bare substrate, I is the intensity from the monolayer-covered samples, d is the thickness of the monolayer, λ is the attenuation length, and α is the take-off angle between the surface and analyzer axis.In our experiment, we collected Ag 3d, Au 4f, and Pt 4f photoelectrons at a take-off angle of 45 degrees.
Previous studies have reported attenuation lengths for Ag 3d and Au 4f photoelectrons in organic SAMs of 3.6 nm and 4.2 nm, respectively.Considering that λ ∝ √ E, where E is the photoelectron kinetic energy, and that the Pt 4f core level binding energy is only slightly smaller than the Au 4f core level binding energy, we assumed that the attenuation length for Pt 4f is also close to 4.2 nm.
The SAMs of CnA on Ag, Au, and Pt substrates underwent characterization using X-ray Photoelectron Spectroscopy (XPS).Core level photoelectron signals for C 1s, Ag 3d, Au 4f, and Pt 4f were acquired with a take-off angle of 45 Based on this contrast, we calculated the thickness (d) of the SAM for different samples (Table S1, Figure S2A), with the listed values representing the averages of 2-3 samples.HOMO and/or HOMO-1 mediated conduction?
In view of the small energy separation between HOMO and HOMO-1 (E HOM O −E HOM O−1 0.2 eV, cf.Table 1), it might be tempting to consider the single level assumed by the (orSLM) approach that succeeded in accurately reproducing the measured currents (cf. Figure 3) to be an additive contribution of the HOMO and the HOMO-1 ("effective single level") rather than a genuine single effective level Speaking in general, given the fact that Γ HOM O = Γ HOM O (V ) depends on bias while Γ HOM O−1 = Γ HOM O (V ) does not (see discussion related to Figure 6 in the main text), depending on their interplay Γ ef f can or cannot significantly depend on bias.More precisely Data fitting using eq (1) could not be satisfactory if eq (S1) applied, i.e. if Γ ef f = Γ ef f (V ) significantly depended on bias, because Γ entering eq (1) is a number that does not depend on V .This implies that eq (1) is precisely like eq (S2), which applies for Γ HOM O (V ) In other words, although slightly closer to the Fermi level, in CnA the HOMO is much weaker coupled to electrodes that the HOMO-1, and this is why the p-type conduction of the CnA junction is dominated by the HOMO-1 and not by the HOMO To sum up, eq (S4) (which is just eq (1)) of the main text in specific notations) does apply and not eq (S1).

HOMO and HOMO-1 energies of isolated CnA molecules: comparison between OVGF and other methods
The small differences between the various MO energy offsets entering the analysis of the present paper makes it clear why utilizing a very accurate quantum chemical method is of paramount importance.The method based on the OVGF is a very elaborate manybody methods wherein the one-electron Green function includes expansions of the self-energy entering the electronic Dyson equation. 94This expansion is exact up to third order in the Coulomb electronic repulsion.Moreover, this third order expansion is supplemented by a geometrical approximation to higher (than third) orders in the electronic repulsion.Table S3 and Figure S8 give a flavor on how large are the differences between the OVGF estimates of the HOMO and HOMO-1 energies and other methods.Letting alone enormous differences from DFT values based on the Kohn-Sham (KS) orbital energies (amounting up to a few eV), we note that the ∆−DFT difference method 110 (a method often considered "acceptable") give an estimate for the HOMO (HOMO-1 cannot be computed using this method) that deviates from the OVGF value by about 0. Because what really matters in the analysis presented in the main text is the difference between the HOMO and HOMO-1 energies, with one exception (label "OVGF(zpm)", second numerical line in Table S3), the theoretical values presented in Table S3 do not contain corrections due to zero-point motion (zpm).The OVGF-based HOMO energies including zpm corrections shown in the second numerical line in Table S3) make comparison with the experimental ionization potentials (IP) 116 (first numerical line in Table S3).As visible in Table S3, the theoretical results perfectly reproduce the experimental data.

Figure
Figure S2: (A) XPS setup for SAM thickness measurements.(B) Dependence of SAM thickness on the molecular size for CnA adsorbed on the metal electrodes investigated in the present work.

Figure S5 :
Figure S5: UPS spectra of (A) bare Ag, Au and Pt, (B) CnA on Ag, Au and Pt.

Figure S6 :
Figure S6: UPS spectra of CnA on (A) Ag, (B) Au, and (C) Pt substrates (n= 8, 9, 10, 12).Binding energies are referenced to the Fermi level, E F = 0,eV.The spectral intensity of SAM-coated metal substrates was normalized to the intensity of bare metal substrates at 0 eV.The crossing points of the red lines indicate the onsets of the HOMOs; the values in eV represent the HOMO-Fermi energy offsets |ε CnA SAM 0

Figure S7 :
FigureS7: The HOMO-1 offset directly extracted from transport data clearly differs from the HOMO-1 offset deduced by combining UPS data and OVGF quantum chemical calculations.95,96This demonstrates that, contrary to alkane 44 and oligophenylene 43 monoand di-thiols, the ("second") tip electrode brings about an additional nonnegligible (measurable) HOMO shift towards the metal Fermi level.Our data by no means indicate an energy difference monotonically decreasing with the molecular size, a fact which rules out any image charge effects.

Figure
Figure S8: (A) Values of the HOMO energy E 0 HOM O for isolated CnA molecules of computed via OVGF depicted along with estimates using the most popular methods utilized in the literature indicated in the legends for the molecular sizes n considered in this paper.(B) Additional values of E 0 HOM O for 1-decyne (C10A).(C) Values of the HOMO-1 energy E 0 HOM O−1 for isolated CnA molecules of computed via OVGF and a few other methods enabling estimation of this quantity.
• .As illustrated in Figure S4, the intensity of the C 1s signal increased, while the intensities of Ag 3d, Au 4f, and Pt 4f signals decreased as the number of carbon atoms in CnA molecules increased.

Table S1
reveals that the thickness of the CnA SAM increases with the number of CH 2 units.

Table S1 :
Thickness in Å of the CnA SAMs on metals obtained via XPS

Table S2 :
HOMO energy offset deduced from transport (ε tran h , eV) and UPS (ε UPS