Controlled Atomic Doping of a Single C60 Molecule

See allHide authors and affiliations

Science  09 Apr 2004:
Vol. 304, Issue 5668, pp. 281-284
DOI: 10.1126/science.1095069


We report a method for controllably attaching an arbitrary number of charge dopant atoms directly to a single, isolated molecule. Charge-donating K atoms adsorbed on a silver surface were reversibly attached to a C60 molecule by moving it over K atoms with a scanning tunneling microscope tip. Spectroscopic measurements reveal that each attached K atom donates a constant amount of charge (∼0.6 electron charge) to the C60 host, thereby enabling its molecular electronic structure to be precisely and reversibly tuned.

The ability to modify the electronic properties of materials via charge donating or accepting dopant atoms plays a crucial role in semiconductor electronics (1). As devices shrink in size, the use of dopants (2, 3) will present new challenges, especially if variations in dopant distributions lead to large changes in interdevice performance. Progress in understanding and overcoming these limitations has been made through the synthesis of new molecular compounds (4) and also via electronic gating techniques (5), but the idea of placing charged dopant atoms directly onto individual molecular elements in situ has been less vigorously pursued. Here, we show how it is possible to tune molecular electronic properties by attaching an arbitrary number of charge-donating K atoms onto an isolated fullerene adsorbate. This method is reversible and should be applicable to more complex molecular structures.

The ability of bulk C60 to incorporate alkali metals and form fullerides has long been known (6, 7). The basis of our technique for creating molecular fullerides is the molecular manipulation capability of the scanning tunneling microscope (STM) (8, 9). By manipulating a single C60 molecule over successive K adsorbates on Ag(001), we formed KxC60 in a completely controllable manner for x ranging from 0 to 4. The electronic structure of a single KxC60 complex could be controllably tuned via the charge transferred from attached K atoms to the C60 host. The C60 molecular levels shift down in energy with respect to the Fermi level (EF), and new splitting in the states is induced near EF. Analysis of the molecular density of states (DOS) reveals that each K atom donates 0.6 electron to the C60 shell. The charge doping process can be reversed by dragging individual KxC60 complexes over surface impurities, thus removing the attached K ions one by one. This method yields a new level of control for tuning the electronic properties of molecular structures and opens previously unexplored possibilities for constructing molecular heterostructures with atomically precise spatially modulated electronic doping.

The experiments were performed with the use of a home-built ultrahigh-vacuum (UHV) STM cooled to 7 K. A polycrystalline PtIr tip was used for all measurements (10, 11). The Ag(001) sample was cooled to ∼80 K, dosed with C60, and transferred into the 7 K STM without ever leaving UHV. Potassium was deposited onto the 7 K sample with the use of a potassium getter manufactured by SAES Getters (Milan, Italy).

The process of constructing a K-doped C60 molecule (Fig. 1) involved moving the C60 over a K atom, which caused the K atom to spontaneously attach itself to the C60 molecule and create KC60 (Fig. 1B). Further manipulation of KC60 allowed the complex to be moved without detaching the K atom. An additional K atom could be added by moving the KC60 molecule over another K atom (Fig. 1C) to create K2C60. This “doping” process can be continued reliably, as shown by the fabrication of KxC60 structures up to x = 4 in Fig. 1. K doping levels up to x = 6 have also been achieved. The topography of a single C60 molecule did not undergo substantial changes after its conversion to KxC60, although occasional rotations are observed as well as a slight change in height and width. K4C60, for example, appears shorter than C60 by 3 ± 3% and wider by 9 ± 3% (Fig. 1E).

Fig. 1.

(A to D) Constant current topographs showing the formation of K4C60 (227 Å by 115 Å, V = 2.0 V, and I = 500 fA). The sliding resistances for manipulation at a sample bias of V = 10 mV for KxC60 were 100, 30, 40, 10, and 5 megohms for x values of 0, 1, 2, 3, and 4, respectively. (E) Radially averaged topographic cross sections of C60 and K4C60 acquired at V = 2.0 V and I = 500 fA. The apparent height of C60 (K4C60) was 4.8 ± 0.3 Å (4.7 ± 0.3 Å). The full width at a height of 0.25 Å (where the difference in width between C60 and K4 C60 is the largest) was 31.1 ± 2.1 Å (33.7 ± 2.1 Å) for C60 (K4C60). The error includes variation between different molecules, different averaging centers, and systematic errors from the data acquisition.

To measure the electronic changes that an individual C60 molecule undergoes during K doping, we performed differential conductance (dI/dV) measurements for each KxC60 molecule. dI/dV(r, V) gives a measure of the local DOS (LDOS) of a molecule at the tip position r and the energy E = EF + eV (12). For positive bias voltages, this allows the probing of the lowest unoccupied molecular orbital (LUMO) as well as higher lying orbitals (LUMO+1, etc.). Figure 2A shows spatially averaged dI/dV spectra for individual KxC60 complexes (x from 0 to 4). The undoped C60 spectrum (black curve, Fig. 2A) shows three main features located at 0.05, 0.5, and 1.7 V that correspond to the split LUMO states and LUMO+1 resonance (11). We refer to the split LUMO states of the undoped C60 as LUMOα (0.05 V) and LUMOβ (0.5 V).

Fig. 2.

(A) dI/dV spectra taken on KxC60 for x values of 0, 1, 2, 3, 4 (data shown was acquired from the molecule displayed in Fig. 1). The tip was stabilized at V0 = 2.0 V and I0 = 50 pA before taking spectra (Vmod = 20 mVrms). Each dI/dV curve is an average of ∼20 spectra taken across the molecule in a square array of points equally spaced by 3.8 Å. This averaging compensates for spatial LDOS variations due to the orbital structure of the molecule. (Inset) A 75 Å by 63 Å dI/dV map (V = 1.25 V and I = 50 pA) showing the dopant-induced energetic shift of the LUMO+1 state. At this voltage, the dI/dV signal is larger for K4C60 than for C60, which causes the doped molecule to “light up” and differentiate itself from the undoped molecule. (B to E) dI/dV maps for C60 (24 Å by 24 Å) and K4C60 (30 Å by 30 Å) show that energetically shifting LUMO and LUMO+1 states does not change their spatial distributions.

There are two main changes that occur to these spectral features as K atoms are added to a single C60 molecule. One is a monotonic shift toward EF of both the LUMO and LUMO+1 resonances with increasing doping. In the LUMO+1 resonance, the value at half-maximum of the left edge of the resonance shifts from ∼1.45 V for C60 to ∼1.15 V for K4C60. The other is that the LUMO profile was substantially altered by K doping. For example, the LUMOα resonance split for KC60, a gap opened up near EF for K2C60 and K3C60, and a peak rose near EF for K4C60. The overall width of the KxC60 LUMO manifold appears to narrow with increasing x without an appreciable LUMO shoulder appearing below EF for all doping levels. In general, dopant-induced spectral changes were only observed when K atoms were close enough to a C60 molecule so that they could not be distinguished from the C60 molecule in STM topographs. The dopant-induced spectral behavior shown here was reproduced for numerous C60 molecules with different STM tips.

Spatial maps of the LUMO+1 and LUMO resonances for C60 and K4C60 are shown in Fig. 2, B to E. Although some distortion was observed because of imperfections in the tip geometry, the spectral maps for the C60 and K4C60 molecular resonances show the same qualitative behavior (11). Although K doping shifts the energetic location of C60 molecular orbitals, it does not severely alter their spatial distribution.

Molecular manipulation with the STM tip was also used to reverse the doping of a KxC60 molecule by controllably removing K atoms one by one (Fig. 3, A and B). Moving KxC60 (in this case K4C60) over an impurity caused one K atom to unbind from the C60 molecule and stick to the impurity (13). Spectroscopic measurements verified that the atom removed was, in fact, a single K atom (Fig. 3C). This process worked reliably for all doping levels from KC60 to K4C60.

Fig. 3.

(A and B) Topographs showing K4C60 being converted to K3C60 by dragging it across a stationary impurity (285 Å by 75 Å, V = 2.0 V, and I = 500 fA). (C) Spatially averaged spectra taken for a K3C60 complex before (black line) and after (red line) adding an additional K ion to make K4C60. The dotted line shows the spectrum after this K4C60 complex is dragged over a surface impurity and reverted to K3C60. These spectra reveal how the molecular electronic structure can be toggled between K3C60 and K4C60 states by this method. Stabilizing parameters were V0 = 2.0 V and I0 = 70 pA before taking spectra.

We cannot determine the exact location of the K atoms in a KxC60 complex, but the lack of any substantial changes in C60 orbital structure caused by K-attachment implies that K atoms are not bound to the upper half of the C60 molecule and reside at the C60-Ag interface. Furthermore, when doped C60 molecules were (accidentally) picked up by the STM tip, all K-atom dopants remained on the substrate.

The K-induced spectroscopic shifts of C60 molecular orbitals can be quantified by fitting the observed resonances with standard line shapes. Here, we focused on the LUMO+1 resonance because it exhibits a simple energetic shift and no complex changes to its line shape. We found that each KxC60 LUMO+1 resonance can be accurately fit with the use of two Gaussian peaks (separated by 0.23 ± 0.05 V) and a linear background. The result of these fits is shown in Fig. 4. Both Gaussian peaks shift linearly toward EF at a rate of 85 ± 15 mV per attached K atom [Supporting Online Material (SOM) Text].

Fig. 4.

The LUMO+1 resonance for KxC60 (x ≤ 4) can be accurately described as the sum of two Gaussian peaks and a linear background. Here, we show how the center points of these two peaks evolve as x changes from 0 to 4. Error bars include variations in possible fit parameters as well as variations between different molecules, orientations, and measurements with different probe tips. The separation between the two LUMO+1 peaks remains essentially unchanged by the addition of K atoms, with a mean value of 0.23 ± 0.05 V. A linear fit to the LUMO+1 evolution gives an average shift toward EF of 85 ± 15 mV per added K atom.

The observed shift of the C60 spectroscopic features upon KxC60 formation is consistent with the notion that each K atom donates about 0.6 of an electron to the LUMO manifold of the C60 molecule. The charge transfer from K to C60 is estimated by assuming the charge rigidly fills the LUMO and does not change the energetic separation between C60 states. In this simple model, each electron added to the LUMO shifts the LUMO+1 toward the Fermi energy by the width of the LUMO resonance divided by its degeneracy (∼0.8 eV/6), which produces a shift of ∼130 mV per electron for C60 on Ag(001). Thus, the observed LUMO+1 shift of 85 mV per K atom indicates a charge transfer of ∼0.6 electron per added K. We note that previous work on K-doped bulk (6, 7) and monolayer (1416) C60 samples indicates that K is nearly fully ionized by C60, with one 4s electron transferred to C60 for all stoichiometries up to K6C60. The smaller charge transfer seen here could result from differences in how K atoms distribute themselves around a single, isolated C60 adsorbate versus the extended monolayers and bulk crystals studied previously. The remaining fractional 4s charge might be shared between the K and Ag(001) surface.

This simple picture, however, does not adequately explain the more complicated evolution of the LUMO resonance under K doping. In addition to a simple downshift in energy, the LUMO resonance line shape is altered substantially under K doping. These changes are not completely understood but likely result from a combination of three main effects: Jahn-Teller–like distortions (1719), asymmetric perturbation in the local electronic potential by the K ions (20), and electron-correlation effects (21). Electron transfer to the C60 cage can lead to distortions of the C60 lattice that remove some of the degeneracies of the C60 states [Jahn-Teller effect (22)]. The splittings of the C60 LUMO and LUMO+1 levels are predicted to be on the order of 0.1 to 0.3 eV for electron transfers of one to two electrons (17), consistent with some of the splittings observed in our study.

An additional perturbation to the C60 states arises from the distribution of charged K ions. The K ions, located on the C60-Ag interface, may create electric fields sufficient to split the C60 states (23) and cause state density to shift vertically with respect to the surface (24). This would diminish the ability of STM dI/dV spectroscopy to resolve states that are spatially shifted toward the bottom of the C60 molecule and away from the STM tip. The absence of substantial LUMO state density below EF might be a consequence of such a perturbation. On the other hand, electron-electron interactions (21) could also produce Coulomb gaps such as those observed near EF for K2C60 and K3C60. The exact nature of the LUMO behavior will require further study.

Supporting Online Material

SOM Text

References and Notes

View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article