Atomic-Scale Coupling of Photons to Single-Molecule Junctions

See allHide authors and affiliations

Science  02 Jun 2006:
Vol. 312, Issue 5778, pp. 1362-1365
DOI: 10.1126/science.1124881


Spatial resolution at the atomic scale has been achieved in the coupling of light to single molecules adsorbed on a surface. Electron transfer to a single molecule induced by green to near-infrared light in the junction of a scanning tunneling microscope (STM) exhibited spatially varying probability that is confined within the molecule. The mechanism involves photo-induced resonant tunneling in which a photoexcited electron in the STM tip is transferred to the molecule. The coupling of photons to the tunneling process provides a pathway to explore molecular dynamics with the combined capabilities of lasers and the STM.

The combination of optical excitation with ultrahigh spatial resolution would reveal new understanding of nanoscale structures and the elucidation of many light-sensitive, dynamic processes (15). To date, techniques available to achieve high spatial resolution with laser illumination are limited by diffraction to about half of the optical wavelength. The introduction of near-field scanning optical microscopes (NSOM), either aperture-based (2) or tip-enhanced (also called apertureless) (3), has extended the spatial resolution beyond the diffraction limit. The aperture-based NSOM has routinely reached a resolution of 50 to 100 nm and potentially down to 10 to 30 nm (2). Apertureless NSOM uses a strongly confined electric field by optically exciting surface plasmons localized at the apex of a sharp metallic tip; in principle, the highest spatial resolution possible by this technique is on the order of a few nanometers (3).

In contrast, the scanning tunneling microscope (STM) readily provides spatial resolution down to the atomic scale. Thus, if the photons couple to the tunneling electrons, it should be possible to achieve high spectral and temporal sensitivity simultaneously with the spatial resolution of STM. However, efforts in this direction have met with little success because of experimental challenges such as thermal effects from light illumination (6).

Here, we demonstrate the coupling of photons to a single molecule in a double-barrier junction of a STM through a two-step process of photo-induced resonant tunneling (7, 8), in which an electron is photoexcited to a higher level in the tip and then tunnels resonantly to a molecular state. This process is detected statistically by monitoring the result of such electron transfer—generation of a new singly occupied molecular orbital (SOMO). The electron transfer probability vanishes when the tip is positioned outside the electronic contour of the resonant molecular state and shows spatial variations within the contour.

The experiments were conducted with a home-built ultrahigh-vacuum (UHV) STM at the base temperature of 9.5 K (without laser illumination) (9) with silver (Ag) or tungsten (W) tips (10). To illuminate and align the laser to the junction, a pair of spherical lenses was mounted inside the UHV chamber. The lasers used were frequency-doubled Nd:YAG (532 nm), HeNe (633 nm), and Ti:Sapphire (800 nm). All were in the continuous wave (CW) mode and linearly polarized along the STM tip axis. Upon laser illumination of the junction, the temperature of STM rose linearly with the incident power (∼0.65 K/mW). To avoid the temperature drift and fluctuation, a laser stabilizer (Model BEOC-LPC, Brockton Electro-Optics Corp., Brockton, MA) was used just before the laser entered the UHV chamber to reduce the power instability to ∼0.1%.

The single-molecule double-barrier junction (schematically shown in Fig. 1A) was defined by positioning the STM tip over an individual molecule adsorbed on a thin (∼0.5 nm) insulating alumina film grown on the NiAl(110) surface (11). The two tunnel barriers in the junction are the vacuum gap between the STM tip and the molecule, and the oxide film between the molecule and NiAl. We studied magnesium porphine (MgP) (Fig. 1C, inset), a simple metalloporphyrin molecule that is involved in photosynthesis (12), which was thermally sublimed onto the oxidized surface at ∼10 K. Because of surface inhomogeneity, different MgP adsorption configurations are present (Fig. 1B).

Fig. 1.

(A) Schematic of a single-molecule double-barrier junction. (B) STM image showing individual magnesium porphine (MgP) molecules adsorbed on an oxidized NiAl(110) surface. The molecules marked M1 and M2 are the type of molecules studied here. The scan size is 10.9 nm by 10.9 nm, Vb = 1.5 V. This image and those in the following figures were taken at |I| = 30 pA. (C) dI/dV spectra measured over a molecule such as M1 and M2 in (B). The forward (red curve) and backward (green curve) scans are indicated by the two arrow heads. The two vertical arrows mark the positions of a sudden change in the dI/dV spectra. In contrast, the dI/dV spectrum (black curve), measured over the bare oxide surface adjacent to the molecule, shows no switching and hysteresis. The dI/dV signal was recorded with the lock-in technique. The tunnel gap was set at Vb = 1.5 V and I = 30 pA. The bias modulation was 10 mV (rms) at 400 Hz. The inset is the molecular structure of MgP. The asterisk (*) denotes a kink in the dI/dV spectrum because of two overlapping states (schematically shown in Fig. 5 as SUMO and LUMO-β).

Here, we focus on one specific type of MgP molecule, such as M1 and M2 in Fig. 1B. The direction of the two-lobe structure in the images taken at Vb > 1.2 V is always perpendicular to the [001] direction of the underlying NiAl(110), where Vb is the sample bias voltage with respect to the tip. In the absence of laser illumination, the differential tunneling conductance (dI/dV) spectra consistently exhibit stepwise changes and hysteresis. For the molecule shown in Fig. 1C, no states appeared until Vb = 0.55 V in the forward scan (ramping Vb up). At Vb ≈ 0.85 V, the dI/dV signal (as well as the tunneling current) suddenly dropped, indicating that the molecule was switched electronically from one state to the other. In the backward scan, the onset of unoccupied states is shifted up by 0.15 V, and a new occupied molecular state emerged at a threshold of Vb = –0.45 V. As the bias was ramped down, the molecule switched back to its original state. This hysteresis behavior could be cycled repeatedly.

The hysteresis, especially the appearance of a new occupied state, indicates that the molecule could be charged at positive bias and discharged at negative bias. The STM images taken at the onset of all of the states further confirm that the molecule with the new occupied state is negatively charged by one extra electron, whereas the one without it is neutral (to be discussed later in the text). The ability to control the molecular charge state is similar to the charging of individual Au atoms adsorbed on an ultrathin NaCl film on a copper surface (13). The ionic relaxation in the alumina, associated with the charging, stabilizes the negative ion, because alumina is even more polar than NaCl (13). In addition, the specific molecular adsorption state is also responsible for this charge transfer phenomenon.

Without laser illumination, the charge state of the molecule cannot be changed when the bias is set within the zero-conductance gap (∼ –0.4 V to ∼0.55 V). However, upon optical excitation in the junction, we observed charging of the molecule that could be controlled by photon energy and flux, sample bias, tip-molecule distance, and position of the tip within the molecule.

To detect and quantify this light-induced charge transfer process, we have adopted an experimental scheme, as shown in Fig. 2A (sequence 1 to 4). In this scheme, the STM feedback was kept on all of the time. We assume that the molecule was initially in its neutral state and set the tunneling parameters at the “parking” condition (Vp and Ip), where the molecule could not be charged with laser illumination except for photon energies larger than ∼2.1 eV (as discussed below). The STM tip was positioned over the molecule and tracked at the maximum of the topographic image. The bias was stepped to Vc and the tunneling current ramped to Ic. After holding there for a certain time thold, the tunneling parameters were returned to the parking condition. To probe whether the molecule is charged, the bias was ramped down to sufficiently large negative value (Vd) with I = Ip. If the molecule is charged, the STM tip will noticeably retract away from the molecule because of the presence of the newly occupied molecular state. At certain negative bias, the molecule will be discharged. Thus, this probing process also serves as a resetting (discharging) process. In addition, the charging probability away from the tracking position of the molecule could be measured by moving the tip to the desired location just before sequence 2. At the end of sequence 2, the tip was returned to the tracking position as shown in sequence 3. This complete cycle was repeated (three cycles are shown in Fig. 2B) and was under computer control.

Fig. 2.

(A) Experimental scheme (not to scale) used to detect the laser-induced charging probability in the single-molecule junctions. The STM feedback is kept on throughout the entire cycle. Time sequence 1 to 4 represents periods of (1) tracking and optional lateral offset, (2) charging, (3) lateral unoffset (optional), and (4) probing/discharging. The circled area illustrates the difference between charged (solid line) and uncharged (dashed line) molecules. We present results obtained with Vp = –0.3 V, Ip = –30 pA, and thold = 5 s. (B) Real-time trace of three successive cycles. The first and third cycles indicate that the molecule was charged by the photons during thold, whereas the molecule remained uncharged for the second cycle. Vc = 0.3 V, Ic = 50 pA, and lateral offset distance from the tracking point of the molecule is 8.8 Å.

The dependence of the charging probability on Vc shows a clear bias threshold (Fig. 3A), and the threshold decreases as the photon energy increases. The linearity of threshold versus photon energy (Fig. 3A, inset) indicates that light couples to the single-molecule junction quantum mechanically. In sharp contrast, without laser illumination, the charging probability was zero for Vc < 0.55 V, corresponding to the onset of unoccupied states for a neutral molecule. By adjusting the photon flux and the tunnel barrier width, the coupling of photons to the single-molecule junction could be controlled. At low photon flux, the coupling increased approximately linearly and saturated at higher fluxes (Fig. 3B). In addition, the coupling increased monotonically with closer tip-molecule distance, equivalent to the vacuum barrier width, as regulated by the tunneling current Ic (Fig. 3C).

Fig. 3.

(A) Charging probability as a function of Vc with laser illumination at different photon energies and compared to the probability without laser illumination (squares). Data were taken with an Ag tip on the same molecule as in Fig. 1C at 532 nm (2.34 eV, diamonds), 633 nm (1.96 eV, triangles), and 800 nm (1.55 eV, circles). The laser incident power was set at 0.2 mW. The dashed lines are drawn to guide the eyes. The inset shows the linearity of charging threshold with the photon energy. (B and C) Charging probability versus the incident power P and tunneling current Ic, respectively. The data were taken on another molecule using another Ag tip with photon energy of 1.96 eV (29). Parameters for (B) are Vc = 0.3 V, Ic = 50 pA and for (C) are P = 0.4 mW, Vc = 0.3 V. The probability for each data point in Fig. 3 and Fig. 4 is taken to be the ratio of the number of cycles, where the molecule was charged to a total of 150 cycles as described in Fig. 2 (30). Thus, the error bars are given by Embedded Image, reflected in the size of symbols.

The result in Fig. 3A shows that the sum of the photon energy and the bias threshold is ∼2.1 V. However, the threshold for charging without laser illumination could be as low as 0.6 V (Fig. 3A). This notable difference suggests that the photons couple to a higher molecular state, the LUMO+1, instead of the LUMO (lowest unoccupied molecular orbital). Thus, the energy difference between LUMO+1 (onset of 2.1 eV) and LUMO (onset of 0.55 eV) for the neutral molecule would be ∼1.55 eV. This difference is consistent with the 800-nm light observed in the tunneling electron–induced fluorescence spectra (10, 14). Density functional theory (DFT) calculation on this molecule shows that the LUMO is doubly degenerate and orthogonal, while the LUMO+1 is nondegenerate (15, 16, 17). The energy separation between them is ∼1.6 eV (15, 16).

Because the onset of the LUMO+1 for the neutral molecule is ∼2.1 eV, optical excitation with lower photon energy must be assisted by tuning the sample bias to vary the Fermi level of tip relative to LUMO+1 (18). The threshold bias for photo-induced charging is positive for photon energies 1.96 eV and 1.55 eV (Fig. 3A). Thus, the charging process occurred in the tip-molecule junction through the vacuum barrier. In the case of 2.34 eV photons, the photo-induced charging process could also occur in the molecule-substrate junction. Thus, the charging probability below its threshold is not zero (Fig. 3A). The electron transfer from the substrate to the molecule through the oxide barrier was observed by moving the tip outside the neutral molecule.

The photon coupling strongly depends on the spatial distribution of the molecular state as indicated by the spatial dependence of the charging probability for photon energy of 1.96 eV (Fig. 4). Images taken at Vc for the neutral molecules (before charging), corresponding to the geometry of MgP, are shown in Fig. 4, A and D, whereas Fig. 4, B and E, represent the contour of the molecular electronic states (19). A comparison of the variation of the charging probability and a line cut through the topographic image (Fig. 4, C and F) indicates that the spatial resolution of the photon coupling is as fine as the resolution of the STM images and approximately follows variations in the molecular electronic states. The spatial contrast in charging probability arises from variations in the LUMO+1, whereas the topographic line cut of the images is composed of contributions from all the states integrated below the imaging bias. This spatially dependent data further supports the mechanism of photo-induced electron transfer involving the tip-molecule junction. Combined with the bias-dependent data (Fig. 3A), the possibility of charging arising through local laser thermal heating is ruled out.

Fig. 4.

(A to C) The spatial dependence of the charging probability taken with an Ag tip, P = 0.4 mW, λ = 633 nm, Vc = 0.3 V, and Ic = 50 pA. (A and B) Images taken at 0.3 V (uncharged, 1.93 Å high), 2.0 V (charged, 4.17 Å high), respectively, without laser illumination. The scan size is 34 Å by 34 Å. (C) The curve connecting the squares represents the charging probability along different parts of the molecule as indicated by the white dots in images (A) and (B). The topographic line cut is taken along the dots from (B). (D to F) show another set of data taken with a W tip, P = 1.2 mW, λ = 633 nm, Vc = 0.25 V, and Ic = 50 pA. Images (D) and (E) were taken at 0.25 V (uncharged, 2.30 Å high) and 2.0 V (charged, 4.46 Å high), respectively, without laser illumination. The scan size is 23 Å by 23 Å. Given the difference in the laser power and that the charging threshold for the molecule in (A) to (C) is ∼0.1 V higher than that for the molecule in (D) to (F) (29), the Ag tip is determined to be more efficient than the W tip by a factor of ∼10.

The molecular states, in particular LUMO+1, for the neutral molecule could be derived as follows, although they were not directly observed in the dI/dV spectrum because of charging. The STM image (Fig. 5A) taken at the onset of the unoccupied state matches one of the calculated LUMOs, named LUMO-α (15, 16, 17, 19). The image taken at negative bias such as Fig. 5B, is relatively featureless and corresponds to the geometry of MgP (19). Once the molecule is charged, the STM images (Fig. 5, E and F) taken at the onsets of unoccupied and occupied states share the same pattern as the LUMO-α (Fig. 5A), indicating that only one electron is transferred into the LUMO-α of the neutral molecule. The newly occupied state is a SOMO, whereas the previously spin-degenerate state remains as a singly unoccupied molecular state (SUMO). They are separated by a coulomb energy U, as indicated in Fig. 5D. Because the peaks for SUMO and SOMO are not usually observed, a lower bound for U is given by the SOMO onset to the SUMO onset of ∼1.15 eV (Fig. 1C). Because the molecule is adsorbed on the polar oxide surface, the degeneracy of the LUMOs is most likely split, which is supported by the kink in the dI/dV spectrum (marked by * in Fig. 1C and Fig. 5D). The onset difference between LUMO+1 and SUMO is ∼1.6 eV (Fig. 1C), indicating that all of the unoccupied states of the charged molecule probably shift up rigidly by 0.15 V, with respect to those of the neutral molecule. These results allow us to roughly project the molecular states of a neutral molecule (dotted line in Fig. 5C). We first fit the dI/dV spectrum before charging for the LUMO-α to a Gaussian. The dI/dV spectrum (Vb > 0 portion) for the charged molecule is shifted down 0.15 V and added to half of the fitted LUMO-α Gaussian curve, because a LUMO-α spin state was split off to become a SOMO.

Fig. 5.

(A and B) Images of a neutral molecule taken at sample bias of 0.57 V and –0.4 V, respectively. (C) Energy diagram showing the molecular states of a neutral molecule (MgP). The arrowed solid and dashed lines represent a two-step process of photo-induced resonant tunneling through the vacuum barrier between the STM tip and the molecule. In comparison, the crossed-out process represents a one-step mechanism of photon-assisted resonant tunneling. (D) Energy diagram showing the molecular states of a negatively charged molecule (MgP). (E and F) Images of the charged molecule taken at 0.75 V and –0.4 V, respectively. The dashed lines for the molecular states in (C) and (D) show the extrapolated local density of states as discussed in the text. The size of the images is 21 Å by 21 Å.

The observed quantum coupling of photons to the single-molecule junction could be explained by photo-induced resonant tunneling, a two-step process involving excited states of the tip (shown in Fig. 5C) (7, 8). This mechanism is feasible because the lifetime of excited electrons in metals [for example, it is ∼10 fs in Ag (20)] is longer than the traveling time for electron tunneling (sub-fs) (21). Photon-assisted resonant tunneling from the tip directly to the molecule, as sketched in Fig. 5C and described in (22), is not likely, although this one-step mechanism has been observed with photons at microwave (23) and terahertz frequencies (24). In photon-assisted resonant tunneling, photon-assisted discharging directly from the SOMO to the unoccupied state in the tip or the substrate would be expected, but this was not observed. In contrast, to observe the discharging arising from photo-induced resonant tunneling, the photon has to first excite the electron from the SOMO to a higher unoccupied molecular state that depends on the photon energy (25). Furthermore, with the photon energy in the 1 to 2 eV range, the effective tunneling barrier height for photo-induced resonant tunneling is much lower than that for photon-assisted resonant tunneling (8). Thus, the two-step process dominates. However, we have not completely understood the reasons for photo-induced electron transfer to the LUMO+1, instead of the LUMO. One possible reason is that resonant tunneling to the LUMO+1 encounters a lower barrier height in comparison with the LUMO. The excess energy associated with the decay from the LUMO+1 to populate SOMO could be dissipated in the radiative and nonradiative transitions and in the reorganization energy of the alumina film in the vicinity of the negatively charged molecule (10, 13, 14).

The efficiency of photon coupling depends on the shape and elemental composition of the STM tip. For example, the Ag tip was more efficient than the W tip by a factor of about 10 (Fig. 4). The higher efficiency could be caused by the longer lifetime of excited electrons in the Ag tip than that in the W tip, which would lead to enhanced photo-induced resonant tunneling. In addition, the local electric field enhancement due to surface plasmons (3, 10) could also contribute to the photo-induced electron transfer.

The realization of the coupling of photons to the tunneling process could lead to experiments with chemical sensitivity at a submolecular level, such as the photo-induced current resulting from the intramolecular HOMO-LUMO transition (25) and laser-induced molecular fluorescence between LUMO+1 and LUMO. The use of photons is less destructive toward the molecule and involves different mechanisms compared with processes induced by tunneling electrons alone in the absence of light. Furthermore, the combination of lasers with the STM enables the exploration of molecular dynamics with simultaneous spatial, temporal, and energy resolutions.

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article