A Single-Molecule Chemical Reaction Studied by High-Resolution Atomic Force Microscopy and Scanning Tunneling Microscopy Induced Light Emission

Abstract

Atomic force microscopy (AFM) as well as scanning tunneling microscopy induced light emission (STM-LE) are, each on their own, powerful tools used to investigate a large variety of properties of single molecules adsorbed on a surface. However, accessing both structural information by AFM as well as optical information by STM-LE on the same molecule so far remains elusive. We present a combined high-resolution AFM and STM-LE study on single metal-oxide phthalocyanines. Using atomic manipulation, the molecules can be deliberately reduced. We demonstrate structure elucidation and adsorption geometry determination of single molecules with atomic resolution combined with optical characterization by STM-LE and the possibility of investigating the change in a molecule’s exciton emission intensity by a chemical reaction.

Scanning tunneling microscopy induced light emission (STM-LE) on single molecules has recently emerged and advanced rapidly¹⁻⁴ with astounding resolution not only spatially but also energetically.⁵⁻⁷ For the intramolecular radiative transitions probed in STM-LE, different excitation and decay mechanisms, relying on the coupling of tip and sample, have been proposed.^2,3,7−11 However, whether an electronic and vibronic intramolecular transition is fostered or not can substantially depend on the exact environment, e.g., the adsorption position.¹²⁻¹⁴ Likewise, changes in the molecular geometry and chemical structure, like the conformation or the presence or absence or substitution of specific moieties, can intrinsically alter a molecule’s optical properties.^3,14 However, combining structural and optical information such as exciton energies in molecules remains challenging.

Because STM is sensitive to the electronic structure, STM-LE is ideally suited for probing optoelectronic properties of single adsorbed molecules, i.e., imaging the densities of molecular frontier orbitals and measuring fundamental gaps¹⁵⁻¹⁷ and exciton energies.^2,4,6,10,11 High-resolution atomic force microscopy (AFM), however, is highly sensitive to the forces between probe and substrate. This allows the measurement of the structure of molecules^18,19 and their adsorption geometry, including conformation,²⁰ adsorption orientation, site, and height.²¹ Accordingly, AFM is ideally suited to investigate the intermediates and products of on-surface reactions²²⁻²⁵ and reactions triggered by atom manipulation.²⁶⁻²⁹ The combination of AFM with STM-LE therefore enables the investigation of the change in a molecule’s optoelectronic properties throughout a chemical reaction. In addition, atomically precise assignment of symmetry axes of allowed radiative transitions to the molecular structure could be facilitated by overlaying STM-LE maps with high-resolution AFM images. The effect of adsorption height as well as adsorption position on the quenching of the radiative decay is accessible as well for suitable molecules, i.e., molecules with different stable adsorption conformations. Combining both techniques is therefore a promising route to link atomically resolved structural information with optical properties of individual molecules.

Here, we report on a combined AFM and STM-LE study, allowing for detailed structure determination with atomic resolution and controlled atom manipulation conjunct with the investigation of optoelectronic properties. We investigated vanadyl-phthalocyanine (VOPc) molecules decoupled from a Ag(111) substrate by an ultrathin NaCl layer.

Phthalocyanines (Pc) are well characterized because of their widespread use in organic light-emitting diodes and photovoltaics,³⁰⁻³² making them an excellent model system for STM-LE experiments.^3−6,10 Because they offer the possibility of induced on-surface chemical reactions, metal-oxide Pc are potentially interesting to investigate the influence of reduction on their electronic and optical properties. The reduced counterpart of VOPc, namely vanadium-phthalocyanine (VPc), could thus far not be synthesized by standard solution chemistry. It is therefore inaccessible to standard optical characterization techniques, while VOPc has already been characterized.^33,34

We were able to specifically dissociate the central oxygen using atom manipulation.^23,35−38 The successful reduction of VOPc to VPc could be confirmed by high-resolution AFM structural elucidation. In the STM-LE spectra, VOPc showed a characteristic light emission at 682 nm, while no light emission was detected from VPc.

Results and Discussion

Characterization of Vanadyl-Phthalocyanine

VOPc (Figure 1a) was deposited onto a cold (T ≈ 10 K) Ag(111) single crystal partially covered by bilayer NaCl [NaCl(2 ML)/Ag(111)], yielding well-dispersed single molecules. For the characterization and imaging, STM and STS as well as non-contact AFM with CO-functionalized tips were deployed^18,39 (see the Methods section for details on the sample preparation and experimental procedures).

Figure 1.

(a) Ball-and-stick model of VOPc. (b, c) Images of VOPc in O-up (left column) and O-down (central column) conformation and VPc (right column) acquired with (b) STM with CO-functionalized tip and (c) AFM at different tip heights. (d) AFM data resolving the adsorption position of VOPc O-up and O-down and VPc on bilayer NaCl. The grid indicates the Cl sites, which are resolved in the outer regions of the images by using a decreased tip height z. STM images were acquired at a set point of V = 0.2 V and I = 0.5 pA. z indicates the tip height relative to the STM set point. Positive z correspond to an increased tip–sample distance.

VOPc features two distinctively different adsorption conformations⁴⁰⁻⁴² on Ag(111) as well as on bilayer NaCl, which we observed with approximately equal occurrences. Because the aim was to perform STM-LE experiments, we will focus in the following only on VOPc adsorbed on bilayer NaCl, which is necessary for electronic decoupling. The two different VOPc species can be distinguished by their appearance in AFM and STM with a CO-functionalized tip, as shown in Figure 1b,c. From the AFM images, they can be assigned to a conformation with the central oxygen atom pointing away from the surface (O-up) and toward the surface (O-down), respectively.⁴⁰⁻⁴² For the O-up conformation, only the outermost phenylene groups can be resolved by AFM, while in the center, it exhibits an extended repulsive feature stemming from the central oxygen sticking out of the surface plane. In the O-down conformation, the full macrocycle can be resolved by AFM. For small tip–sample distances, the core of the macrocycle becomes distinctly more repulsive than the phenylene groups, indicating that the macrocycle is not fully planar but rather dome-shaped. This can be rationalized by the adsorption geometry; due to the oxygen pointing toward the surface in the O-down conformation, the macrocycle is pushed away from the surface in the center and bends slightly down at the outer part. In both conformations, the molecule adsorbs with the center on a Na site (Figure 1d). The reduced species VPc, shown in the right column in Figure 1, can be generated from VOPc O-up by atomic manipulation, as demonstrated in the following.

On-Surface Synthesis of VPc from VOPc by Atom Manipulation

Synthesizing VPc is known to be challenging as it immediately oxidizes to the more stable VOPc in oxygen containing environments. As of yet, its synthesis has only been achieved by Eguchi et al. by on-surface metalation of free-base Pc in UHV-environment.⁴³ An alternative route to synthesizing VPc is the controlled reduction of single VOPc molecules by SPM-based atom manipulation. In this approach, the highly localized nature of the electric field, tunneling current, or both between tip and molecule is exploited.

We found that the reduction reaction could be actuated reliably. To that end, we first positioned the tip above the center of a VOPc O-up at a set point of V = −2 V and I = 0.5 pA and then retracted the tip by about 2 Å. Next, the sample voltage was ramped from −2 to −4 V, resulting in currents on the order of 10 to 100 pA, and was held constant for up to 30 s while the tip position was kept fixed. To monitor whether and when a reaction occurred, tunneling current I and frequency shift Δf were recorded simultaneously. The dissociation of the oxygen usually occurred after a few to a few tens of seconds and was accompanied by a sharp drop in current and increase in Δf, as shown in Figure 2 for one exemplary reaction. Subsequent imaging with AFM confirmed the structural change (see Figure 1c); the molecular structure is clearly different from both observed VOPc conformations. Similar to VOPc O-down, the full macrocycle can be resolved by AFM. However, the outermost phenylene groups appear distinctly brighter compared to the central part. In the center, a bright, cross-shaped feature is visible in the AFM images, a similar contrast as has also been observed on other metal phthalocyanines such as FePc.^44,45 To quantify the non-planar adsorption geometries of the molecules, we performed force–distance spectroscopy [Δf(z)], yielding the tip height z*, i.e., the tip height at which the minimum Δf(z) is reached,²¹ with the tip placed above the center of the macrocycle (z_c^*) and a phenylene group (z_p), respectively. The difference in z* between both positions, in combination with the contrast in the AFM images, provides a measure for the shape of the molecule. For VOPc O-down, z_c^* is larger than z_p, corroborating the finding from the AFM image in Figure 1b that the molecule is dome-shaped. After the reaction, the bright appearance of the phenylene groups and the dark center in the AFM image indicate a clear change in the molecule’s shape. This is confirmed by force–distance spectroscopy, showing that indeed the phenylene groups are further away from the surface than the center of the molecule (z_c^* < z_p) (see the Supporting Information). This finding agrees with the expected change in the molecular geometry upon dissociation of the central oxygen. In addition, Figure 1d and the inset in Figure 2 show a newly arisen defect in the vicinity of the molecule after the reaction, presumably the dissociated oxygen atom, which is adsorbed on a Cl–Cl bridge site. For these reasons, we exclude a mere switching behavior between the O-up and O-down conformation.⁴²

Figure 2.

Tunnel current I and frequency shift Δf recorded over time during the reduction of a VOPc in O-up conformation at V = −4 V. The reaction is accompanied by a sharp drop in tunnel current and increase in Δf. The insets show constant-current STM images (I = 0.5 pA) before and after the spectrum was obtained.

Investigation of the adsorption position reveals that after the reaction, the molecule has rotated in-plane by about 45° and moved laterally such that now its center is located on a Cl site (see Figure 1d). The AFM images of VPc (Figure 1c, right column) show the phenylene groups with an apparently increased diameter and bond-like features appearing in the center of the rings. We assign this contrast to two different energetically degenerate, mirror symmetric adsorption orientations in which the symmetry axes of the molecule are rotated by a few degrees with respect to the symmetry axes of the surface. The adsorption orientation is switched under the influence of the tip, preferring the orientation in which an outer ring is located under the tip. The apparent bond-like features arising within the rings appear at the tip positions at which the molecular orientation switches.^29,46,47 This shuttling motion is already well-known for certain metal-phthalocyanines adsorbed on different surfaces.^3,48

STM-LE on VOPc and VPc

Even though there generally is no one-to-one correspondence between the electronic transitions probed by STS and the allowed optical transitions, they are often closely related (in STM-LE). Measuring the differential conductance thus can yield valuable information that complement the characterization of a molecule’s optical properties.

Figure 3 displays STS and STM data of the three different molecular species, i.e., VOPc in O-up and O-down conformation as well as VPc, on bilayer NaCl on Ag(111). The energy levels of VOPc O-up and O-down are shifted with respect to each other. We tentatively explain this shift by a concurrence of several effects. (i) The electrostatic dipole of the molecule, related to its partial negative charge accumulation at the oxygen, shifts both resonances to higher (lower) energies in the O-down (O-up) adsorption conformation. (ii) The increased adsorption height of the macrocycle in the O-down conformation will lead to an increased relative voltage drop across the NaCl film^15,49 and, thus, to an up-shifted negative ion resonance (NIR) and a down-shifted positive ion resonance (PIR) compared to the O-up conformation. Both effects work in the same direction in the case of the NIR, explaining its significant upshift for the O-down conformation with respect to the O-up conformation. In case of the PIR, which is found at similar energies for both conformations, the effects work in opposite directions. (iii) In addition, we cannot exclude shifts in the energies due to strain in the different adsorption geometries¹⁴ as well as a different response and reorganization of the NaCl due to the different adsorption geometries.⁵⁰⁻⁵² The spatial distribution of the molecular orbitals in VOPc in O-up and O-down conformation is very similar, however. Slight variations are only present in the center of the molecule, where the VO group is pointing in opposite directions.

Figure 3.

Differential conductance curves recorded on VOPc in O-up (orange) and O-down (black) conformation and VPc (blue). The insets show the respective orbital densities corresponding to the HOMO at negative and the LUMO at positive voltages, measured with a metallic tip.

After the reduction of VOPc to VPc, also the electronic structure of the molecule has significantly changed (Figure 3). The spatial distributions of both the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) look distinctly different compared to VOPc in either conformation. Additionally, the PIR for intact VOPc in both conformations is found at voltages below −2 V. In contrast, the reduced VPc exhibits a PIR already at around −0.7 V. Hence, imaging at bias voltages between −0.7 and −2 V serves as a fingerprint to differentiate between VPc and VOPc, as shown in the insets in Figure 2.

To investigate whether the observed changes in adsorption geometry and electronic structure also affect the molecule’s optical properties as VOPc is reduced to VPc, we recorded STM-LE spectra on both species. The STM-LE spectra on pristine VOPc in O-up conformation show a single peak at 682 nm (1.82 eV) for sample bias voltages V < −2.4 V with the tip positioned atop one of the extremities of the macrocycle (Figure 4a). The excitonic emission stems from a radiative transition from the degenerate Q-band to the ground state and is in good agreement with experimental results for the energy of the Q-band transition in (dissolved) VOPc.^33,34 Light emission could only be observed above VOPc in the O-up conformation because for O-down, it was not possible to establish stable conditions, i.e., the molecule would always jump to the tip or away for high bias voltages and currents, which would both be needed for inducing as well as detecting molecular luminescence on bilayer NaCl.

Figure 4.

(a) STM-LE spectrum recorded on a phenylene group of a VOPc (O-up) (V = −2.6 V, I = 350 pA). (b) Comparison of STM-LE spectra recorded on a phenylene group of a VOPc before and after tip-induced reduction using the exact same tip and parameters (V = −2.4 V, I = 150 pA). (c) Possible excitation mechanisms for molecular exciton formation. The energy level diagram on the left shows the initial state at V = 0 V applied bias. Upon applying a sufficient bias voltage, two possible excitation mechanisms can lead to exciton formation. Orange box: the tip’s electrochemical potential μ_t is pulled below the PIR, allowing the tunneling of an electron from the HOMO to the tip and thereby charging the molecule positively (1). This leads to a downward shift of the resonance for attaching electrons to the former LUMO (LUMO of the neutral molecule), pulling it below the sample’s electrochemical potential μ_s (2). The LUMO can subsequently be occupied with an electron from the sample, forming an exciton (3). Green box: electrons tunnel inelastically between tip and sample (only shown here for negative bias voltage), exciting a gap plasmon between tip and sample. This plasmon couples to the molecular exciton, leading to its formation.

After the dissociation of oxygen from VOPc O-up, no emission was observed on the product anymore, independent of the position of the tip above the molecule. Figure 4b shows a direct comparison of the molecular luminescence recorded with the tip positioned above one of the extremities of a VOPC O-up before and after its reduction using the exact same tip as well as the same STM set point and spectrum acquisition parameters. The peak at 682 nm, which is clearly visible in the spectrum recorded on VOPc (Figure 4b, left spectrum), disappeared in the VPc spectrum (Figure 4b, right spectrum). Also, no energetically shifted photon emission is detected on VPc for energies down to the near-infrared of 1.2 eV (see the Methods section for more details). Taking spectra at different applied bias voltages (−1.5 to −2.7 V and 2 to 2.4 V) also did not lead to a detectable signal in the emission. The absence of detectable luminescence could indicate either that no vibronic transition that can undergo a radiative decay is excited or that the radiative decay path is quenched compared to the situation in VOPc.

For the excitation of molecular excitons in STM-LE, two different pathways are usually discussed,^{2−4,7,8,10,11} shown in Figure 4c. They rely on either the emptying of the HOMO and subsequent occupation of the LUMO (charge-injection mediated) or a plasmon-mediated energy transfer to the molecule.

The charge-injection mediated process is illustrated in Figure 4c in the orange box. Upon applying a sufficiently high negative bias voltage, the tip’s electrochemical potential (μ_t) is shifted below the PIR, allowing the depopulation of the HOMO by tunneling of an electron into an unoccupied tip state (1). Charging the molecule positively leads to an energetic down-shift of the resonances for electron attachment due to reduced screening of the positive ion cores, which is partly compensated by reorganization within molecule and underlying substrate.⁵⁰ In addition, because the tip and sample form a double-barrier tunneling junction, the energetic positions of the molecular resonances with respect to the sample’s electrochemical potential (μ_s) are affected by the potential drop in the dielectric spacer layer separating molecule and metallic substrate. For bilayer NaCl, this can be estimated to be 5–10% of the applied bias voltage.^15,49 The sum of these effects can pull the resonance for attaching an electron to the former LUMO (LUMO of the neutral molecule) below the sample’s electrochemical potential, opening an additional tunneling channel (2). This allows electrons to tunnel into the former LUMO, creating an exciton (3).^2,3,11 Because the tunneling barrier for attaching electrons into the former LUMO is lower compared to the one for the former HOMO, the occupation of the LUMO is preferred.

The plasmon-mediated process (Figure 4c, green box) requires the occurrence of localized gap plasmon modes in the tip–sample junction. The molecule is excited via a plasmon-mediated energy transfer from inelastically tunneling electrons to the molecule. The coupling strength of exciton and gap plasmon is dictated by their energies and hence, in case of matching energies, exciton formation as well as its radiative decay is fostered.^7,8,10

In the charge-injection picture, the down-shift of the resonance for attaching electrons to the former LUMO upon charging the molecule positively at bias voltages of V = −2.6 V might be sufficient to allow its population in the case of VOPc. In VPc, however, the NIR is around 400 meV higher in energy than in VOPc. Hence, the level shifting might be insufficient to pull this resonance below the sample’s electrochemical potential in the positive charge state. This, in turn, would prevent the formation of an exciton in VPc in the charge-injection picture.

The decrease in fundamental gap upon reduction of VOPc observed in STS could indicate a concurring red shift of the excitonic emission that might also explain the absence of detectable emission. One possibility could be that excitonic emission in VPc occurs below our minimal measured energy of 1.2 eV. However, the exciton energy is challenging to quantitatively deduce from the change in fundamental gap because a direct general correlation is not given. Optical transitions are governed by symmetry and optical selection rules and, furthermore, many-body effects can have significant impact on both the involved energies and orbitals.^17,53,54 In metal Pc, the Q-band emission is driven by a transition between the ground state of A_1g symmetry and the first excited state of E_u symmetry of the orbitals delocalized over the macrocycle (π−π*).^55,56 The energy of this transition is usually only slightly affected by a change in the metallic center.⁵⁷⁻⁵⁹ This might indicate that the change in exciton energy in VOPc driven by the reduction of its metal oxide center is smaller than the change in fundamental gap of 1 eV might suggest.

For a change in exciton energy within our optical detection window, the absence of detectable luminescence could indicate a change in the cross-section of gap plasmon and molecular exciton, considering a plasmon-mediated excitation and decay. The plasmon resonance recorded at the bare double-layer NaCl shows a broad maximum around 630 nm, with still considerable intensity detected at around 680 nm (peak in luminescence spectrum of VOPc; see the Supporting Information). For larger wavelengths (smaller energies), the intensity of the plasmon resonance is significantly decreased, possibly resulting in a less efficient energy transfer between plasmon and exciton. For the decay of the molecular exciton, this means that less modes are available within the cavity between tip and sample. Hence, the radiative decay mechanism might be heavily quenched.

Quenching of the radiative decay channel could also be caused by changes in the environment or geometrical changes within the molecule.^3,12−14 A change in adsorption site for example, as observed for the reduction from VOPc to VPc, could significantly alter the relative likelihood of different decay pathways. This could facilitate a nonradiative decay of the molecular exciton, quenching the luminescence to below our detection limit.

Conclusions

In conclusion, we report on the realization of a combined AFM and STM-LE study, allowing for the investigation of the change in excitonic emission by a chemical reaction. As a model system, we used single VOPc molecules adsorbed on NaCl(2 ML)/Ag(111). Using atom manipulation, we were able to deliberately reduce VOPc to form VPc, which was confirmed by AFM atomic resolution imaging, adsorption site determination, and force–distance spectroscopy, and compare the emission characteristics before and after the chemical reaction using STM-LE. While for VOPc, a peak at 682 nm (1.82 eV) was clearly visible, no light emission was detected on VPc. Based on STS measurements on the different molecules, we relate this to a shift in energies of the molecular resonances with respect to the electrochemical potential. Possible explanations for the absence of excitonic emission after reduction to VPc are the hindrance of charge transfer into the LUMO (charge-injection mediated process), a quenching of radiative decay (plasmon-mediated process) or a shift of the exciton energy out of our optical detection window.

The combination of AFM with STM-LE allows the linking of a molecule’s structure with its optoelectronic properties. This approach can be extended using Kelvin probe force microscopy (KPFM) to incorporate the internal charge distribution, charge state, or dipole moment as well. Exploiting the AFM’s capabilities for precise atom manipulation, this can be used to investigate the change in a molecule’s properties induced by a chemical reaction. Due to the low-temperature and UHV environment, this provides access to the properties of molecules that cannot be synthesized by standard solution chemistry or where standard optical characterization is not possible, as shown for VPc. Moreover, STM-LE could assist AFM in the identification of molecules within complex molecular mixtures⁶⁰⁻⁶² by providing complementary information on the same individual molecule.

Methods

STM and AFM Measurements

The STM and AFM measurements were performed with a home-built combined STM/AFM setup operating at ultrahigh-vacuum (UHV) conditions (p ≈ 1 × 10^–10 mbar) and low temperatures (T ≈ 5 K). The microscope is equipped with a qPlus force sensor⁶³ operated in the frequency modulation mode⁶⁴ (resonance frequency f₀ ≈ 30 kHz, quality factor Q ≈ 100 000, spring constant k ≈ 1800 N m^–1, oscillation amplitude A ≈ 0.5 Å). The bias voltage is applied to the sample. All STM images were acquired in constant-current mode, AFM images were taken in constant-height mode at 0 V bias voltage. For optical detection, we used a spectrograph (Acton SP-300i, Princeton Instruments) coupled to a liquid nitrogen cooled CCD camera (PyLoN, Princeton Instruments) with a spectral resolution of about 0.2 nm and a solid angle for the detection of Ω ≈ 0.03. STM-LE spectra were recorded in an energy range of 1.28–3.07 eV (404–969 nm) and 1.18–2.53 eV (491–1054 nm) and by averaging over several frames, where each frame typically lasted 3 to 4 min, yielding total acquisition times per spectrum between 8 and 60 min. The shown spectra are background corrected.

Sample and Tip Preparation

As a substrate, we used a Ag(111) single crystal partially covered with (100)-oriented, 2 monolayer (ML) thick NaCl islands [NaCl(2 ML)/Ag(111)]. The Ag crystal was in situ cleaned by repeated Ne-ion sputtering and annealing (T ≈ 720 K) cycles. Subsequently, NaCl was evaporated, while the sample was kept at T ≈ 430 K such that (100)-oriented mostly double-layered NaCl islands were formed on the Ag substrate. The microscope tip consisted of a PtIr-wire (25 μm in diameter) that was sharpened using a focused ion beam followed by in situ indentations into the bare Ag surface to prepare a clean and atomically sharp tip. For tip apex passivation, CO was deposited on the cold (T ≈ 10 K) Ag surface by dosing gaseous CO into the UHV chamber. For functionalizing the tip apex, a CO molecule was picked up from the surface.^18,65 The VOPc molecules were deposited on the cold (T = 10 K) sample via sublimation from a Si-wafer that was flash-annealed to approximately 900 K within a few seconds.

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsnano.9b01852.

• Figures showing force–distance spectroscopy data on VOPc O-up, VOPc O-down, and VPc and STM-LE spectrum of the plasmon resonance on the bare bilayer NaCl (PDF)

The authors declare no competing financial interest.