Regulating the femtosecond excited-state lifetime of a single molecule

See allHide authors and affiliations

Science  07 Sep 2018:
Vol. 361, Issue 6406, pp. 1012-1016
DOI: 10.1126/science.aat9688

Selectively exciting desorption

Reactions of molecules adsorbed on surfaces can be induced by injecting electrons from the tip of a scanning tunneling microscope. Rusimova et al. show that for the tip-induced desorption of toluene molecules from a silicon surface, two activation channels exist: One is invariant, but the other depends on the height of the tip above the surface. When the tip is very close to the molecule, it can quench the excitation. The decreased lifetime, in turn, decreases the desorption probability.

Science, this issue p. 1012


The key to controlling reactions of molecules induced with the current of a scanning tunneling microscope (STM) tip is the ultrashort intermediate excited ionic state. The initial condition of the excited state is set by the energy and position of the injected current; thereafter, its dynamics determines the reaction outcome. We show that a STM can directly and controllably influence the excited-state dynamics. For the STM-induced desorption of toluene molecules from the Si(111)-7x7 surface, as the tip approaches the molecule, the probability of manipulation drops by two orders of magnitude. A two-channel quenching of the excited state is proposed, consisting of an invariant surface channel and a tip height–dependent channel. We conclude that picometer tip proximity regulates the lifetime of the excited state from 10 femtoseconds to less than 0.1 femtoseconds.

Using the tip of a scanning tunneling microscope (STM) to initiate chemical reactions offers a route to controllable single-molecule chemistry (1, 2). Through the mechanical interaction between tip and target molecule, or by the electric field in the gap, the STM can induce molecular change across a ground-state potential energy landscape (3). The STM tunneling current, however, can generate excited states of a molecule and hence give enhanced specificity, and more varied outcomes, to the manipulation action [e.g., bond dissociation (4, 5), isomerization (6), or tautomerization (7)]. The specificity arises by controlling the energy (5) or position (7, 8) of the single electron (or hole) excitation within a single molecule. The ensuing molecular dynamics, and hence the final outcome, evolve naturally from that point. Having the ability to control and influence the dynamics of the excited state itself would open new paths to control matter, and its reactions, at the molecular level.

Here, we found that the lifetime of the positive ion of single toluene molecules on the Si(111)-7×7 surface can be directly controlled by the STM. By bringing the tip close to the molecule (600 to 800 pm), we regulated the excited state–mediated reaction outcome (molecular desorption) by more than two orders of magnitude. We correlate this to a reduction of the excited-state lifetime by approximately two orders of magnitude. We propose that a new electronic state is generated by the tip-molecule interaction that provides an additional decay channel for the excited state, thus quenching the excited state before its natural surface-limited lifetime elapses. We anticipate this work to be a starting point for other more complex molecular systems that yield multiple excited-state outcomes where this technique could be used to instigate, probe, and control them. The quenching process relies on fundamental quantum processes and should be applicable to a wide class of molecule/surface systems.

Multiple molecular adsorbates have been shown to react to the STM tunneling current (1), especially benzene and derivatives (5, 9). Broadly, the probability per electron of inducing a molecular reaction is higher on the Si(111)-7×7 surface than on the Si(100)-2×1 surface and is orders of magnitude higher than on metal surfaces (79). On metals, lifetimes of molecular ion states are as low as 0.1 fs (10), but the reduced density of states in semiconductors leads to longer excited-state lifetimes. The theory of dynamics induced by electronic transition (DIET) links greater lifetimes of excited states to higher probabilities of reaction (such as bond breaking or desorption) (11). Benzene, chlorobenzene, and toluene on Si(111)-7×7 have all been extensively studied (5, 9) and are highly sensitive to tunneling current.

Figure 1 shows a series of STM images charting the positive ion resonance (or negative-bias “hole”)–induced manipulation of a single chemisorbed toluene molecule on the Si(111)-7×7 surface. At the imaging conditions used (+1 V, 100 pA), chemisorbed toluene molecules were unperturbed by the STM (12) and appeared as dark features against the bright spots that were the adatoms of the silicon surface. Chemisorbed toluene molecules formed a 2,5-di-σ bonding configuration with the surface, forming one covalent bond to a silicon adatom (colored red) and one to a neighboring silicon rest atom (second-layer atoms with dangling bonds) (Fig. 1A). To manipulate the molecule, during a raster scan from bottom to top, we halted the tip atop the molecule and performed current injection (–1.3 V, 900 pA) for 8 s.

Fig. 1 STM imaging and time trace of single-molecule manipulation.

(A to C) High-resolution STM images and corresponding schematic diagrams of the manipulation procedure (imaging parameters: +1 V, 100 pA, 3 nm × 3 nm). (A) Before manipulation. A half unit cell of Si(111)-7×7 is outlined; the white circle atop the missing adatom–like dark spot location indicates the position of a single toluene molecule. (B) During manipulation. (C) After manipulation. (D) Time trace of the tip height during charge injection.

Figure 1D shows the tip height variation during this process. In step i, the tip was halted above the molecule. The feedback loop was disabled, and the tip retracted by 1 nm before the voltage was ramped to the desired manipulation value with the set-point current at 20 pA. In step ii, the feedback loop was then reengaged. In step iii, the set-point current changed to the required injection current, resulting in the tip approaching the surface closer than its initial value by an amount Δzi. Charge injection continued, and in this particular case, after 0.35 s of charge injection, the molecule-adatom bond was broken, leading to desorption. The underlying (bright) silicon adatom was exposed, causing the tip to withdraw by Δzm to restore the set-point current (step iv). Resuming the interrupted image scan of Fig. 1B resulted in a “half-moon” feature at the molecular adsorption site, typical of a manipulation event occurring mid-scan. Subsequent image scans (Fig. 1C) had the conventional Si(111)-7×7 surface appearance, including the silicon dangling bond at the original location of the toluene molecule.

From the fraction of ~120 toluene molecules that were manipulated after an injection time t, we deduced a time-dependent probability of manipulation P(t) for a single molecule, consistent with the first-order rate equation dP(t)/dt = k[1 – P(t)], where k is the rate of manipulation. Figure 2A illustrates this for injection parameters of +1.6 V, 450 pA. Figure 2B illustrates how the manipulation rate k varied with tunneling current for electron injection at +1.6 V. Figure 2C presents data for hole injections at –1.3 V and at –1.0 V. For electron injection, the rate increased linearly with injection current (see fit to Fig. 2B). For hole injection at low current (2 to 10 pA), we again found a linear dependence, but beyond 10 pA the rate of manipulation was approximately constant; the fitting function of Fig. 2C is discussed below.

Fig. 2 Rate of manipulation.

(A) Time dependence of the fractional manipulated molecule population (injection parameters: +1.6 V and 450 pA; 117 molecules). The dashed line shows the fit to P(t) = 1 – exp(–kt). (B) Rate of manipulation for electron injection at +1.6 V with a linear fit. (C) Rate of manipulation for hole injection at –1.3 V (solid circles) and at –1.0 V (open triangles). See text for fit details of tip-dependent model (dashed line). Error bars indicate SD.

The number n of electrons (or holes) that drive a single-molecule manipulation (4) leads to a power-law dependence of the rate k with current I, kIn. Hence, for electron injection, the near-linear dependence n = 0.8 ± 0.1 indicates a one-electron process (9). Similarly, at low currents, a one-hole process is responsible for desorption. However, for hole injection at currents above 10 pA, the near-constant rate implies a largely current-independent process. If the current is not driving the manipulation, what does?

The manipulation rates observed were a factor of 104 greater than those occurring in purely thermally driven desorption of toluene from Si(111)-7×7 (12). Hence, the presence of the STM tip is required for this manipulation to take place. Possible tip-molecule interactions that might drive manipulation are mechanical (i.e., a short-range chemical interaction between tip and molecule) or result from the electric field of the tunnel junction. However, we can rule out both. Figure 3A shows the tip height z during the electron and hole injections performed at different currents. The tip height z is the distance from the center of the bonding Si adatom to the center of the leading atom of the STM tip (see supplementary materials). In all cases, z exceeds 600 pm. To identify possible mechanical manipulation, we modified the manipulation experiments by setting the bias during step ii to 0 V, disabling the feedback loop, and setting z to a specific value (Fig. 3B, schematic). For each z value, we then measured the outcome of ~90 single-molecule manipulation experiments with an 8-s “exposure” of each target molecule. Little or no desorption was observed for z at or above 600 pm (Fig. 3B, shaded portion). Thus, in the height regime of the current-manipulation experiments, no mechanical manipulation occurred, and the desorption that did occur in Fig. 3B was consistent with that expected for a thermally driven process (see supplementary materials for z < 600 pm discussion).

Fig. 3 Mechanical and electric-field tip-induced interactions.

(A) Tip-surface separation as a function of tunneling current during charge injection. Solid circles, hole injection at –1.3 V; open diamonds, electron injection at +1.6 V. (B) Probability of manipulation after 8 s in the mechanical presence of the tip (0 V, 0 pA). (C) Estimated electric field (magnitude) in the junction between tip and surface as a function of the current I during the charge injection manipulation experiments. Solid circles, hole injection at –1.3 V; open diamonds, electron injection at +1.6 V. (D) Probability of manipulation after 8 s with only the electric field interaction (–10 V, 0 pA). Error bars indicate SD; some error bars are too small to see.

We eliminated the possibility of an electric field–induced manipulation mechanism by modifying step ii so that, with feedback disabled, the tip retracted an additional distance from the surface. We applied a –10 V bias to generate an electric field EV/z in the junction comparable to that in the current-injection experiments, and whose magnitudes are shown in Fig. 3C. In this case, however, there was no current. As shown by the data in Fig. 3D, without the current, there was little or no manipulation.

A similar linear to constant rate crossover appears in two previous studies (13, 14). There, tip-induced band bending (TIBB) was put forward as a possible explanation. Since then, detailed theoretical work and scanning tunneling spectroscopy show that TIBB only occurs if the semiconductor is in depletion (15, 16). For our work with n-type Si, this would be for electron injection. Therefore, TIBB cannot explain our hole injection results, nor the results of (13). The doping level here and in (14) also precludes any measurable TIBB even if it occurs in the depletion regime (17). Instead, the model proposed here is consistent with all three reports.

The final outcome of the molecular manipulation can be either that the molecule completely leaves the surface (desorption) or that it reattaches to the surface elsewhere (diffusion). We label an outcome as diffusion if, in an “after” STM image (e.g., Fig. 1C), the manipulated molecule appeared at an adjacent binding site. All other manipulation outcomes are classified as desorption. For all injection currents used, we found a branching ratio B of the probability of desorption to diffusion that was constant throughout the hole-injection experiments, Bh = 0.037 ± 0.004. It was also constant for electron injections with Be = 0.24 ± 0.03 over the reported range of currents. Furthermore, there was no evidence of other forms of manipulation, such as intramolecular bond dissociation (5), in either current regime.

Recasting the rate of manipulation in terms of the probability per injected charge of manipulation (electron or hole), Pe = ke/I (where e is the magnitude of the electron charge), yields Pe as a function of the tip height z during the current injections. For electron injection, as expected for a one-electron process (Fig. 4A), Pe was fairly constant over the range of z studied. For –1.3 V hole injection (Fig. 4B), Pe exponentially increased with z (i.e., decreasing current) until at ~800 pm, we found a near-constant region. Figure 4C shows data for –1.0 V hole injections (10 to 900 pA), where for all injections, we found a similar exponential increase in the manipulation probability as the tip withdrew from the surface.

Fig. 4 Manipulation suppression at close tip proximity.

(A) +1.6 V electron injection data from Fig. 2B recast as probability per electron as a function of the tip height. (B) –1.3 V hole injection data from Fig. 2C, recast as in (A). (C) –1.0 V hole injection. Blue lines in (A) and (B) show surface-limited excited-state dynamics; dashed red lines in (B) and (C) model tip-limited dynamics. Black curve in (B) is a fit to Eq. 1. Right-side y axes of (B) and (C) show the inferred excited-state lifetime τ of the positive ion with a value of 10 fs for purely surface-limited dynamics. (D) Schematic energy level diagrams depicting three regimes of tip manipulation suppression. VB is the bias voltage applied to the tip relative to the sample Fermi level EF.

The desorption of toluene, via a DIET process, occurs in three steps (18): (i) excitation by capture of the injected charge by the toluene molecule; (ii) dynamics, the evolution of the ionic molecule on its excited-state potential; and (iii) detachment, with decay of the state (neutralization) leaving a vibrationally excited neutral molecule and leading to molecule-surface bond breaking and the final outcome of desorption or diffusion. Given that for the same z change, Pe was constant for electron injection, we conclude that step i was also constant for hole injection. That is, the change in the “spot size” of the tunneling current caused by the change in the value of z did not change the fraction of the current captured by the molecule. Given the invariance of the branching ratio Bh, we further conclude that step iii was the same for all the experiments shown in Fig. 4B. Within a DIET model, what remains to influence the probability of manipulation is step ii, specifically the lifetime of the excited state (19).

For the similar system of benzene on Si(100), Alavi et al. (20) identified a hole excited-state lifetime of ~10 fs and a probability per hole injection similar to our findings. Further, in line with theoretical predictions for long-lived excited states (19), Alavi et al. reported a monotonic and near-linear dependence of the manipulation probability on the hole excited-state lifetime. Therefore, for our hole injections, we relate our maximum (i.e., constant region) manipulation probability per hole of 320 (±10) × 10–9, with an excited-state lifetime of 10 fs, and use the linear dependence Pe = βτ (where β = 32 × 10–9 fs–1) to map our measured probability of manipulation to an excited-state lifetime τ. The result is an excited-state lifetime that changes by two orders of magnitude, from 10 fs to <0.1 fs (Fig. 4, B and C, right-side y axes). A value on the order of 0.1 fs is more typical of that of adsorbates on metal surfaces (10), indicating that the proximity of the tip transforms the molecule-semiconductor system into a metal-molecule-semiconductor system.

Studies of cyclohexadiene on Si(100) (21, 22) have shown the creation of an interface electronic state at the location of the molecule as an STM tip approaches. The new state lies near the Fermi level, and in tandem with its creation, the highest occupied molecular orbital (HOMO) at –1.5 V broadens and decreases in intensity as the tip approaches closer. Given that our system also contains a π-bonding orbital on a six-member carbon ring that is di–σ-bonded to a Si substrate, we propose that at our higher currents (closest approach) a similar interfacial electronic state results in the reduced probability per hole of manipulation by providing a new decay channel for the excited state, which reduces its lifetime and concomitantly the probability of manipulation.

The lifetime of an excited state is the inverse of its relaxation rate R = 1/τ. We propose two components for the relaxation of the positive ion state: (i) a fixed rate arising from the presence of the surface, RS = 1/τS, with τS = 10 fs; and (ii) a z-dependent rate accounting for the effect of the tip, RT(z) = 1/τT(z), giving R = RS + RT(z). This tip-mediated relaxation channel will be related to the density of states of the interface state, ρi, through Fermi’s Golden Rule. An analogous scheme is used to describe the STM excitation, direct measurement, and z-dependent quenching of the millisecond spin excitation in single atoms (23).

Figure 4D presents schematic energy level diagrams for three regimes of tip height: (I) large tip-molecule separation with surface-dominated excited-state relaxation and thus reaction, (II) intermediate-z range with onset of (assumed near the Fermi level) tip-dependent interface state quenching, and (III) small-z separation with tip-dependent interface state quenching dominating the excited-state relaxation. These three regimes are indicated in the rate dependencies of Fig. 4, A to C.

For a tip-molecule system with localized electronic structure, the force between tip and molecule has been calculated as FIm with m between 1 and 2 (24). This calculation invoked a wave function overlap argument and should be broadly similar to the perturbative physics of the initial generation of an interface state by our STM tip. Thus, we make the connection RT(z) ∝ ρi(z) ∝ exp(−2κz)m, leading to a z-dependence of Pe ofEmbedded Image(1)where τT(z0) = τS and κ = 1.17 ± 0.06 A–1, as found from Fig. 3A. A surface-limited model, Pe = βτS, has a constant lifetime and therefore a constant manipulation probability. This surface-only model fits the +1.6 V electron injection in Fig. 4A. Below ~800 pm, there is a possible slight decrease in Pe, which suggests that the negative-ion state is also perturbed by the interface state. For –1.0 V hole injection shown in Fig. 4C, the fit is purely exponential, Pe(z) ∝ exp(2κz), corresponding to a tip-dominated dynamics. For –1.0 V, at all currents, the tip was near the molecule, hence the excited-state dynamics were always tip-limited. At –1.3 V, the tip was slightly farther removed from the surface. Thus, Fig. 4B shows a fit to Eq. 1 with m = 1.1 ± 0.1, and demonstrates a crossover at z0 = 830 ± 20 pm from a tip-limited to a surface-limited regime.

Our initial finding of a near-invariant rate of manipulation can therefore be reconciled with a one-hole process. For a one-hole process, the rate is defined as k = PeI/e. Combining this with the tip-dependent manipulation probability Pe of Eq. 1 gives the fit to the rate of manipulation (dashed line) in Fig. 2C. At large tip-molecule separation (low current), the traditional linear kI dependence is evident. For higher currents that led to a closer tip and quenching of the molecular excited state, the rate of manipulation became k = I/Im = I–0.1±0.1, giving the largely rate-invariant region of manipulation (Fig. 2C).

Molecules that require more than one electron (or hole) for manipulation, such as C-Cl dissociation of chlorobenzene (5) or diffusion of NH3 (25), should naturally be more sensitive to any tip modification of the excited state. Alongside other semiconducting molecule/surface systems, we would also expect any molecule/surface system that displays a long-lived excited state—for example, molecule/single–atomic layer insulator/metal systems (26)—to be sensitive to tip-induced modification of the excited state.

Supplementary Materials

Materials and Methods

Supplementary Text

References (2737)

References and Notes

Acknowledgments: We thank R. Palmer, D. Bird, and D. Wolverson for fruitful discussions. Funding: Supported by EPSRC grant EP/K00137X/1 (P.A.S.), a University of Bath studentship (K.R.R.), and EPSRC CDT CMP grant EP/L015544/1 (R.M.P.). Author contributions: K.R.R. was the primary experimentalist and performed the analysis. R.M.P., R.H., and F.L. performed subsets of the experiments; S.C. provided theoretic support and data interpretation; P.A.S. led the team, designed the experiment and the analysis; and K.R.R., S.C., and P.A.S. wrote the manuscript. Competing interests: Authors declare no competing interests. Data and materials availability: All data supporting this study are openly available from the University of Bath data archive at
View Abstract

Stay Connected to Science

Navigate This Article