Report

Single-Atom Spin-Flip Spectroscopy

See allHide authors and affiliations

Science  15 Oct 2004:
Vol. 306, Issue 5695, pp. 466-469
DOI: 10.1126/science.1101077

Abstract

We demonstrate the ability to measure the energy required to flip the spin of single adsorbed atoms. A low-temperature, high–magnetic field scanning tunneling microscope was used to measure the spin excitation spectra of individual manganese atoms adsorbed on Al2O3 islands on a NiAl surface. We find pronounced variations of the spin-flip spectra for manganese atoms in different local environments.

The magnetic properties of nanometer-scale structures are of fundamental interest and may play a role in future technologies, including classical and quantum computation. Such magnetic structures are composed of magnetic atoms in precise arrangements. The magnetic properties of each atom are profoundly influenced by its local environment. Magnetic properties of atoms in a solid can be probed by placing the atoms in tunnel junctions. Early experiments with planar metal-oxide-metal tunnel junctions doped with paramagnetic impurities exhibited surprisingly complex conductance spectra described as “zero-bias anomalies” (14). Such anomalies were shown to reflect both spin-flips driven by inelastic electron scattering and Kondo interactions of magnetic impurities with tunneling electrons (57). Single, albeit unknown, magnetic impurities were later studied in nanoscopic tunnel junctions (8, 9). Recently, magnetic properties of single-molecule transistors that incorporated either one or two magnetic atoms were probed by means of their elastic conductance spectra (10, 11). These measurements determined g values and showed field-split Kondo resonances due to the embedded magnetic atoms.

The scanning tunneling microscope (STM) offers the ability to study single magnetic moments in a precisely characterized local environment and to probe the variations in magnetic properties with atomic-scale spatial resolution. Previous STM studies of atomicscale magnetism include Kondo resonances of magnetic atoms on metal surfaces (12, 13), increased noise at the Larmor frequency (14, 15), and spin-polarized tunneling (16). We demonstrate a technique for measuring the spin excitation spectra of individual atoms adsorbed on a surface using inelastic electron tunneling spectroscopy (IETS) with a STM. Combined with the STM's capability to fabricate, image, and modify atomically precise structures, this technique provides a powerful new tool for studying and engineering the local magnetic properties of nanometer-scale systems.

IETS measures excitation energies, such as vibrational energies, of atoms or molecules within tunnel junctions (17, 18). Above a threshold voltage, electrons are able to transfer energy to these excitations during the tunneling process. This additional tunneling channel results in an upward step in conductance at the threshold voltage. For the measurements reported here, tunneling electrons lose energy to spin-flip excitations of single Mn atoms. The signature of Zeeman splitting in spin-flip IETS is a step up in conductance at an energy proportional to the applied magnetic field.

We used a home-built, ultrahigh-vacuum STM that reaches a base temperature of 0.6 K by means of a single-shot pumped 3He refrigerator. The STM is vibrationally isolated and at the same time thermally coupled to the 3He liquid by suspending the STM chamber directly above the liquid. We liquefied the 3He using the Joule-Thomson effect, obviating the need for a pumped 4He reservoir. Magnetic fields up to B = 7 T were applied in the plane of the sample. NiAl(110) samples were prepared in vacuum by repeated sputter/anneal cycles. Samples were then exposed to ∼10 Langmuir of O2 at ∼500 K and further annealed at 1200 K. This resulted in the growth of patches of Al2O3 two layers thick (0.5 nm) interspersed with regions of bare NiAl (19). Samples were then transferred into the STM, and Mn atoms were subsequently evaporated onto the cold surface. Mn has partially filled d-orbitals, and the free atom has a total spin of S = 5/2. The differential conductance, dI/dV, was measured using lock-in detection of the tunnel current I by adding a 50-μV root mean square modulation at 829 Hz to the sample bias voltage V.

A topograph of the partially oxidized NiAl surface (Fig. 1A) shows that the bare metal and the Al2O3 oxide regions are atomically flat. Contrast on the metal is caused by standing waves in surface-state electrons (20). The oxide has a nearly rectangular unit cell 1.06 nm by 1.79 nm, which yields a complex but nearly periodic pattern in the STM topograph (21). The cold sample was subsequently dosed with a small amount of Mn, and the same area was imaged again (Fig. 1B). Single Mn atoms are seen as protrusions with an apparent height of 0.13 nm on the bare metal surface and 0.16 nm on the oxide. The density of Mn atoms on the oxide is significantly smaller than on the metal, presumably due to a lower sticking probability and motion along the oxide surface during adsorption (22).

Fig. 1.

Comparison of Mn atoms on oxide and on metal. (A) STM constant-current topograph of a NiAl (110) surface partially covered with Al2O3 (upper right). Image: 20 nm by 10 nm; V = 100 mV, I = 50 pA. (B) Same area after dosing with Mn. (C) Conductance spectra at T = 0.6 K on the Mn atom on oxide (upper curves) measured at B = 7 T (black) and B = 0 T (red). The lower curves (shifted for clarity) were measured over the bare oxide surface. (D) Conductance spectra on a Mn atom on NiAl (upper curves) and on the bare NiAl surface (lower curves). All spectra in (C) and (D) were acquired with a nominal conductance of 10 nA/V (I = 50 pA at V = 5 mV) and normalized to unity for |V| > 2 mV to emphasize differences in low-bias features. (E) Topograph of the Mn atom on oxide. Image: 2.8 nm by 2.8 nm; B = 7 T, T = 0.6 K, V = 2 mV, I = 20 pA, VAC = 0.5 mVrms. (F) Spatial map of dI/dV acquired concurrently; an increased signal (light area) maps the spatial extent of the spin-flip conductance step.

The upper set of spectra in Fig. 1C shows the marked magnetic-field dependence of the conductance when the tip is positioned over a Mn atom on the oxide. At B = 7 T, the conductance is reduced near zero bias, with symmetric steps up to a ∼20% higher conductance at an energy of |Δ| ∼ 0.8 meV. These conductance steps are absent at B = 0. Furthermore, no conductance steps are observed when the tip is positioned over the bare oxide surface, over the bare metal surface, or over a Mn atom on the metal surface (Fig. 1, C and D). We verified that these conductance spectra are characteristic for single Mn atoms on oxide terraces and on bare NiAl(110) by measuring many Mn atoms with different atomic arrangements at the STM tip apex. The spatial extent of the conductance step can be visualized by measuring dI/dV while imaging the Mn atom (Fig. 1, E and F). We find that the dI/dV signal is localized to an area 1 nm in diameter, comparable to the atom's apparent lateral extent in the corresponding STM topograph.

The characteristic signature of spin-flip IETS is a step up in the differential conductance dI/dV at a bias voltage corresponding to the Zeeman energy Δ = gμBB, where μB = 57.9 μeV/T is the Bohr magneton and g = 2.0023 for a free electron. Figure 2A shows that the conductance step shifts to higher energy with increased field. Broadening of this step is due mainly to the effect of temperature, with contributions from the ac voltage modulation, spin lifetime, and instrumental noise. The thermal broadening of tip and sample densities of states can be calculated by twice convolving an intrinsically sharp step with the derivative of the Fermi-Dirac distribution (23). To fit our experimental data, we combine temperature with other sources of broadening by using an effective temperature (24). From the fits we extract the Zeeman splitting Δ for each value of magnetic field. We display the measured dI/dV curves in Fig. 2A by normalizing them to unity for voltages outside of the spin-flip region, by using the fit to establish the large-voltage conductance.

Fig. 2.

Shift of the spin-flip conductance step with magnetic field. (A) Conductance spectra (points) for an isolated Mn atom on oxide at different magnetic fields. Solid lines show fits to the temperature-broadened step model (see text). The data fit well to a step height of 20.5% for all fits except the highest field, where a slight tip modification changed the step height to 21.5%. The effective temperature in all curves was T = 0.85 K. All spectra were acquired with a nominal conductance of 10 nA/V (I = 50 pA at V = 5 mV) and normalized to unity for large |V| (see text). (B) Magnetic field dependence of the Zeeman energy Δ. Black points are extracted from the fits in (A), and red points were taken on a Mn atom near the edge of an oxide patch. Linear fits (black and red lines) constrained to Δ = 0 at B = 0 yield g values of 1.88 and 2.01, respectively.

The measured Zeeman splitting is proportional to the magnetic field (Fig. 2B, black points). The data fit well with a straight line through the origin and a slope that corresponds to g = 1.88 ± 0.02 (25). A different Mn atom, this one within 1 nm of the edge of an oxide patch, shows a significantly different g value (red points) of g = 2.01 ± 0.03. The only difference between these two Mn atoms is the local environment: They have different lateral distances to bare metal region; they may sit at different binding sites in the oxide unit cell; and perhaps more importantly, we expect the oxide patch to show reconstruction near the boundary to minimize its energy. We are not aware of any studies of the detailed atomic structure of the oxide patches near their boundaries. We have verified that the values we measure for Δ are insensitive to the height and lateral position of the tip, indicating that the tip serves as a nonperturbative probe of the spin properties of the adsorbed atom.

Mn atoms on the oxide that are laterally near metal-oxide interfaces (e.g., Fig. 3A) can exhibit spectra that are markedly different from those of the isolated Mn atoms on oxide terraces. Both of the interfacial Mn atoms in Fig. 3A have an apparent height similar to that of single Mn on oxide terraces. Whereas the left Mn atom shows an IETS spectrum similar to those in Figs. 1 and 2, the right Mn atom shows much larger (∼60%) steps in conductance (Fig. 3B). In addition, Fig. 3B shows a peak at zero field and zero bias. This peak splits and shifts to higher energy as the magnetic field is increased. Although the magnitude decreases sharply with field, the peaks remain clearly present at all magnetic fields. This behavior agrees well with the perturbation theory that was developed in the context of planar tunneling devices (5).

Fig. 3.

Conductance spectra of Mn showing Kondo resonances. (A) Topograph (6 nm by 10 nm) of Mn atoms bound near the interface between an oxide patch (upper half) and bare metal. (B) Conductance spectra of the Mn atom at top center in (A). The zero-field spectrum shows a Kondo resonance: a Lorentzian shaped rise in conductance near V = 0. At high fields a large spin-flip step dominates the Kondo signal. (C) Kondo resonance for a Mn atom (or cluster) near the boundary of an oxide patch that appears as a 0.17-nm-high protrusion (not pictured). The peak represents a factor-of-5 rise in conductance at zero bias. The Kondo peak broadens and splits symmetrically at higher magnetic fields.

The relative strength of the zero-bias conductance peak and spin-flip steps can vary dramatically. Figure 3C shows an example of a Mn atom (or cluster) where the conductance peak dominates the spin-flip steps. At zero field, the zero-bias conductance is enhanced by a factor of ∼5 relative to the background. The peak splits with magnetic field, and no clear spin-flip steps are observed.

The spectra in Fig. 3, B and C, show the hallmarks of a Kondo resonance: a narrow conductance peak with Lorentzian shape at zero bias that splits with magnetic field. The Kondo effect reflects the spin-flip interactions of conduction electrons with a localized magnetic impurity. The full width at half maximum of the resonance can be used to extract a Kondo temperature of TK ∼ 3 K in Fig. 3B, and TK ∼ 6 K in Fig. 3C (26, 27). The enhanced zero-bias conductance seen here is a simpler manifestation of Kondo physics than obtained in previous STM studies, where a more complicated Fano line shape for Kondo resonances reflected interference effects (12, 13, 26, 27). Unlike these earlier STM studies, where magnetic atoms were directly adsorbed on a metal surface, here the interaction between the Mn atom and the NiAl conduction electrons is mediated by an oxide film. We note that Mn adsorbed directly on NiAl does not show any Kondo signature in the 1- to 100-meV energy range studied here.

The zero-bias conductance peak for Mn on Al2O3 is comparable to Kondo effects observed in other nanostructures (911). However, the device characteristics in these nanostructures varied considerably, due presumably to uncontrolled variations in the molecular conformation, binding sites, electrode structures, and neighboring molecules. It is one of the strengths of STM to be able to characterize and control each of these variables.

The zero-bias anomaly in thin-film tunnel junctions showed a spin-flip channel with inhomogeneous broadening that was much larger than the sample temperature. This broad linewidth was attributed in part to the spatial average over impurities in the junction with differing g values (3, 4). Our observations indicate that such spatially averaged studies may reflect not only different g values but also site-dependent amplitudes for the spin-flip and Kondo channels.

The ability to directly measure the g value of individual atoms with the STM enables site-specific study of magnetic moments. When combined with the STM's capability to assemble atomically precise structures, spin excitations can now be studied in custom-engineered nanostructures. If atoms with spins can be coupled to each other in a controlled fashion, it might be possible to use the spin degree of freedom to transmit and process information on the atomic scale (28).

References and Notes

View Abstract

Navigate This Article