Letter pubs.acs.org/NanoLett
Magnetism in Single Metalloorganic Complexes Formed by Atom Manipulation T. Choi,†,⊥ M. Badal,†,‡ S. Loth,§,¶ J.-W. Yoo,†,∥ C. P. Lutz,§ A. J. Heinrich,§ A. J. Epstein,† D. G. Stroud,† and J. A. Gupta*,† †
Department of Physics, The Ohio State University, Columbus, Ohio 43210, United States School of Natural Sciences, University of California, Merced, California 95343, United States § IBM Research Division, Almaden Research Center, San Jose, California 95120, United States ∥ School of Materials Science and Engineering, Ulsan National Institute of Science and Technology, Ulsan, Republic of Korea, 688-798 ‡
S Supporting Information *
ABSTRACT: The magnetic properties of molecular structures can be tailored by chemical synthesis or bottom-up assembly at the atomic scale. We used scanning tunneling microscopy to study charge and spin transfer in individual complexes of transition metals with the charge acceptor, tetracyanoethylene (TCNE). The complexes were formed on a thin insulator, Cu2N on Cu(100), by manipulation of individual atoms and molecules. The Cu2N layer decouples the complexes from Cu electron density, enabling direct imaging of the TCNE molecular orbitals as well as spin-flip inelastic electron tunneling spectroscopy. Results were obtained at low temperature down to 1 K and in magnetic fields up to 7 T in order to resolve splitting of spin states in the complexes. We also performed spin-polarized density functional theory calculations to compare with the experimental data. Our results indicate that charge transfer to TCNE leads to a change in spin magnitude, Kondo resonance, and magnetic anisotropy for the metal atoms. KEYWORDS: Metallorganic complexes, spin/charge transfer, atomic manipulation, spin flip spectroscopy, spin Hamiltonian, ultrathin insulating layer
A
Atomic resolution imaging and atom/molecule manipulation provide detailed information on the bonding between metal atoms and molecules, while tunneling spectroscopy can be used to probe how the electronic and magnetic properties evolve upon bond formation.13,16,17 For example, magnetism in single TCNE (tetracyanoethylene) molecules adsorbed on Cu(111)14 and in V[TCNE] complexes on Ag(100) formed by atom manipulation,13 produces a Kondo resonance in tunneling spectroscopy due to interaction with the substrate electrons. On Cu(111), multistable binding configurations with distinct charge and spin transfer were observed,14 echoing the variable oxidation states coexisting in the bulk materials. Here, we report STM studies of individual M[TCNE] complexes, where M is a magnetic (Co, Fe) or nonmagnetic (Cu) transition metal atom. Atomic manipulation is used to synthesize these complexes on an ultrathin insulating layer of Cu2N islands grown on Cu(100). Such layers have proven to be useful for decoupling adsorbates from the substrate electron density, thus enabling direct imaging of molecular orbitals similar to the gas-
n atomic-level understanding of spin and charge transfer across metal/organic interfaces is fundamental for a variety of pursuits, including information technologies (e.g., molecular electronics,1,2 molecular spintronics,3 quantum computing4), as well as surface chemistry (e.g., heterogeneous catalysis5). For example, in the family of organic-based semiconductors M[TCNE]x∼2, where M is a transition metal and TCNE is the strong charge acceptor tetracyanoethylene (C6N4), a variety of phenomena coupling electronic, optical and magnetic degrees of freedom are observed as one varies transition metal (V, Cr, Mn, Fe, Co, Ni) across the periodic table.6−8 Disorder is a central issue in such materials and arises from variability in the charge state of both metal and organic components, chemical stoichiometry, and purely structural contributions from materials growth in solvent, or solvent-free conditions. 9 It is perhaps counterintuitive that highly disordered V[TCNE]∼2 exhibits the highest ferromagnetic TC in the family (∼400 K), while crystalline Mn[TCNE]∼2, or Fe[TCNE]∼2 compounds with higher-spin transition metal atoms have much lower TC (107 and 121 K respectively). Such uncertainty motivates scanning tunneling microscopy (STM) studies of the smallest possible subunits of these materials, namely individual metallo-organic complexes.10−15 © 2014 American Chemical Society
Received: October 31, 2013 Revised: January 26, 2014 Published: February 3, 2014 1196
dx.doi.org/10.1021/nl404054v | Nano Lett. 2014, 14, 1196−1201
Nano Letters
Letter
Figure 1. Building metal−TCNE complexes on Cu2N/Cu(100). (a) STM image of the surface as prepared, showing a Co atom on Cu2N, and two TCNE molecules on the Cu(100) substrate. (b) One TCNE is picked up (dashed circle) onto the tip by varying voltage V and tip height Δz as shown in (f). Enhanced contrast is attributed to the TCNE-terminated tip. (c) A Co[TCNE] complex is formed after dropping the TCNE molecule off onto the Co atom. (d) A Cu atom is dropped off from the tip. (e) A nonmagnetic control, Cu[TCNE] is formed by picking up another TCNE molecule and putting it down onto the Cu atom. All images are taken at (0.15 V, 20 pA) and are Laplace filtered to emphasize local contrast. Scale bars = 1 nm. All images are taken at 5.3 K. (f) Plot of typical bias voltage, tip height, and measured tunneling current for picking up TCNE molecules from Cu substrate. The initial tip height is set at (0.1 V, 50 pA).
Figure 2. Comparison of spectra before/after bond formation taken with various tip locations. (a) STM image of Co[TCNE] on Cu2N (0.2 V, 40 pA) and TCNE adsorbed on a N vacancy in Cu2N island (0.15 V, 20 pA). All scale bars = 0.5 nm. (b) dI/dV spectroscopy on Co, TCNE, and Co[TCNE]. The tip height was set at 0.2 V, 50 pA. For clarity, red curves are offset by 0.1 pA/mV. Inset shows numerical derivative of the dI/dV signal (middle of (b)) enhancing the difference of vibrational IETS signals before and after bond formation (blue arrows). All data were taken at 5.3 K.
phase molecule16,18 and spin-flip spectroscopy of single atoms and clusters.19,20 STM imaging allows us to visualize bonding in the complexes, and high-resolution tunneling spectroscopy, combined with spin-polarized density functional theory calculations, allows us to study how charge transfer to TCNE influences the spin state and magnetic anisotropy of the magnetic atoms.
Figure 1a shows an STM image of a Cu2N island, surrounded by bare Cu(100) surface. Under our deposition conditions (c.f. Methods), we find metal atoms on both Cu2N and Cu regions but TCNE only on Cu regions or, rarely, adsorbed atop defects in the Cu2N islands (Figure 2a and Supporting Information Figure S2). This adsorption behavior reflects the fact that thin insulating films such as Cu2N typically have low sticking 1197
dx.doi.org/10.1021/nl404054v | Nano Lett. 2014, 14, 1196−1201
Nano Letters
Letter
Figure 3. Characterization of M[TCNE] complexes (M = Co, Fe, and Cu). (a−c) STM images of Co[TCNE], Fe[TCNE], and Cu[TCNE] complexes formed by atom manipulation (0.2 V, 20 pA, 0.25 V, 40 pA, and 0.2 V, 20 pA respectively). Scale bars = 1 nm. Inset of (a) shows the chemical structure of TCNE and calculated molecular orbital (LUMO) of gas-phase TCNE.14 (d−f) dI/dV spectroscopy on M[TCNE] complexes and isolated M atoms on Cu2N (M = Co, Fe, Cu respectively). The tip height was set at (60 mV, 30 pA), (40 mV, 50 pA), and (40 mV, 10 pA) in (d−f) respectively. For clarity, colored curves are offset by 0.15 pA/mV in (d,f) and 0.2 pA/mV in (e). All data were taken at 5.3 K.
appears at −0.15 V for Co in Co[TCNE] that is absent for the isolated atom. Spectra taken with the tip over the central CC bond in isolated TCNE shows steps at ∼32 meV (∼258 cm−1) due to inelastic scattering of tunneling electrons from, for example, vibrational modes.21 These steps are assigned to the =C-(CN)2 in-plane rocking mode,13,22 and are suppressed after complex formation.13 Similarly strong differences are observed with the tip positioned over the cyano groups. Unfortunately, we did not observe steps due to the CC bond (expected in the 170−200 mV range) or CN bond (∼280 mV), as prior Raman studies attributed changes in these mode energies to charge transfer in alkali[TCNE] salts.22 Selection rules for inelastic electron tunneling spectra (IETS) are still not well understood, and it is often the case that vibrational modes visible with other techniques do not appear in STM-IETS. Tunneling spectra were found to be independent of orientation within the complex (Supporting Information Figure S3), suggesting that interaction with Cu2N is secondary to the MTCNE bonding itself. Figure 3a−c shows STM images of Co, Fe, and Cu[TCNE] complexes, formed by atom/molecule manipulation as in Figure 1. We characterized the magnetic properties of the isolated metal atoms (Co, Fe, and Cu) and the metal atoms in complexes (Co[TCNE], Fe[TCNE], and Cu[TCNE]) by spinflip tunneling spectroscopy as seen in Figure 3d−f. We observe steplike features in conductance for the magnetic atoms due to the opening of an inelastic tunneling channel associated with spin-flip excitations. These spectra can be understood by an empirical spin-Hamiltonian20,23,24
coefficients for adsorbates. We therefore used atomic manipulation to form individual M[TCNE] complexes on the Cu2N island, as illustrated in Figure 1. First, the STM tip is located over a TCNE molecule adsorbed on the Cu substrate. After disabling the STM feedback loop, the tip is brought closer (∼0.3 nm) to the molecule while ramping the bias voltage (Figure 1f). This process results in the transfer of the TCNE molecule from the surface to the apex of the STM tip (nearly 100% success rate). Figure 1b shows an STM image taken with the TCNE terminated tip; the enhanced spatial resolution results from chemical interactions between the functionalized tip and the surface.20 Next, the TCNE molecule can be transferred back to the sample surface to form a complex with a metal atom on a Cu2N island. To achieve this, the TCNEterminated tip is located over the metal atom, and is brought closer by lowering the bias voltage (∼10 mV) and increasing the set current (∼7 nA) with the STM feedback enabled. As the tip is moved closer to the metal atom, a sharp reduction in tunneling current marks the drop-off of the TCNE molecule (∼80% success rate). Figure 1c shows an STM image of the area after forming a Co[TCNE] complex. We also formed a Cu[TCNE] complex on the same Cu2N island by first depositing a Cu atom from the STM tip (c.f. Figure 1d and Supporting Information Figures S1 and S8) and then repeating the TCNE transfer procedure (Figure 1e). Complexes were formed with one of four possible orientations with respect to the Cu2N lattice but always with the cyano group (CN) closest to the metal atom (Supporting Information Figure S3). Figure 2a shows a Co[TCNE] complex and TCNE molecule on Cu2N. Both closely resemble the lowest unoccupied molecular orbital (LUMO) of gas-phase TCNE (c.f. Supporting Information Figure S2), except that one cyano group (CN) appears larger in Co[TCNE] due to interaction with the metal atom. Perturbation by the STM tip during imaging or spectroscopy can rotate TCNE within the complexes without detachment, suggesting some degree of bonding (c.f. Supporting Information Figure S3). Tunneling spectra from individual Co, TCNE, and Co[TCNE] indicate differences in electronic structure also suggestive of bond formation (Figure 2b). For example, a prominent occupied orbital resonance
2 2 2 Ĥ = −gμB B⃗ ·S ̂ + DSẑ + E(Sx̂ − Sŷ )
(1)
Here g is the Lande g-value, μB is the Bohr magneton, B is the applied magnetic field, S is the atomic spin, D is an axial anisotropy, and E is a transverse anisotropy. The anisotropy terms result from the structure of the Cu2N lattice. The axial anisotropy breaks the degeneracy of ms levels, where ms is the magnitude of spin projection onto the z-axis, while the transverse mixes the ms levels. In the absence of transverse anisotropy, the states can be described as ms multiplets and the 1198
dx.doi.org/10.1021/nl404054v | Nano Lett. 2014, 14, 1196−1201
Nano Letters
Letter
uniaxial anisotropy introduces zero-field splitting of ms = ±1/2 and ms = ±3/2. The step at ±5.5 mV in Figure 3d corresponds to the transition between multiplets ms = ±1/2 and ms = ±3/2. This feature was also observed in a previous study of Co on Cu2N, giving values of g = 2.2, S = 3/2, D = 2.75 meV, and E = 0 meV.24 The positive value of D for the Co atom represents hard-axis and easy-plane anisotropy. The peak at 0 mV for Co is a Kondo resonance, which is not included in the spinHamiltonian model. Compared to Co on Cu(100) with a Kondo temperature of 88 K,25 the lower Kondo temperature (2.6 K) for Co on Cu2N indicates the decoupling effect of the intervening Cu2N.24 We now discuss tunneling spectroscopy of the M[TCNE] complexes after synthesis by atomic manipulation. Data taken on the Co atom of Co[TCNE] shown in Figure 3d show steps at a reduced voltage (∼1.9 meV), and no Kondo resonance. Similarly, spectra on the Fe atom of Fe[TCNE] show steps at a reduced voltage (Figure 3e). Coordination of these atoms to TCNE may introduce charge transfer and change the local electrostatic environment, both factors that can significantly change the magnetic anisotropy. Control measurements of nonmagnetic Cu atoms and Cu[TCNE] complexes reveal no steps in low voltage range (Figure 3f), suggesting that the steps are not associated with low energy vibrational modes.21 To confirm the spin-related origin of the low-energy steps and determine the spin magnitude and magnetic anisotropy, we performed measurements at lower temperature (1 K) and in magnetic fields up to 7 T. Figure 4a,b shows an STM image and
system was chosen following the convention for isolated Co on Cu2N.24 Our experimental data suggest that as a first approximation, we can neglect higher-order terms in the spin Hamiltonian due to the reduced crystal-field symmetry at the Co site, and the spin−spin interaction between Co and TCNE.26 Further discussed in the Supporting Information, the spin-flip spectra are independent of molecular orientation (Supporting Information Figure S3), and there are no spin-flip steps or Kondo features on TCNE within the complexes (c.f., Figure 2). A priori, the magnitude of the Co atomic spin is not known. Therefore, we simulated the field-dependent spectra with varying g, D, and E values for the two most likely Co spin values (S = 3/2 and S = 1) (c.f. Supporting Information Figure S7).24 The deviation between simulated and experimental curves is then calculated to evaluate the quality of the fit. Parameter-space plots of the deviation constrain g, D, and E to two local minima for each value of S. Comparably good fits are obtained for either S value (Figure 4c). Though we cannot more uniquely determine these parameters without further measurements with different field directions, our constraints clearly indicate significant differences in magnetic anisotropy between the isolated Co atom on Cu2N and the Co atom in Co[TCNE] complex (Figure 3 and Supporting Information Figure S6). To further support this conclusion and gain insight into the underlying mechanism, we performed spin-polarized DFT calculations (c.f. Methods). The spin value of the Co atom in the Co[TCNE] complex can differ from the bare Co atom due to charge transfer from the metal atom to TCNE, which is a strong charge acceptor. We find that the fully relaxed Co[TCNE] complex lies in a plane parallel to the Cu2N surface with Co (0.17 nm) and TCNE (0.32 nm) above as seen in Figure 5a. The Co atom of the Co[TCNE] complex attracts its neighboring two N atoms, and the Cu atom underneath the Co atom moves toward the bulk (Figure 5b).23,27 The calculated atomic positions of the Co atom and the neighboring atoms in the Cu2N island are very similar to those in the Co/ Cu2N system,27 suggesting that introduction of TCNE does not cause further surface relaxation. Thus, we attribute the changes in magnetic anisotropy to bond formation rather than structural deformation due to TCNE. To confirm this, we calculated charge densities of the Co[TCNE]/Cu2N system and find that there is a significant charge transfer from the Co atom (+0.72e) to the TCNE molecule (−0.58e) and Cu2N surface (−0.14e) as seen in Figure 5b. For clarity, the charge and magnetic moment distribution for only nearest neighboring atoms to Co atoms are shown. The net charge density in the Cu2N slab is −0.14e (c.f. Methods). This suggests that the Co atom forms chemical bonds with the TCNE molecule and Cu2N surface. The simulated STM topographic image agrees well with the experimental image (Figure 5d); both show a LUMO-like orbital associated with TCNE, and a bulge in one of the −C N lobes where the Co atom is bonded. We note that although the molecular orientation for the theoretical modeling differs from that in Figure 4, the comparison remains valid because the tunneling spectra were independent of TCNE orientation (c.f. Supporting Information Figure S3). We estimate the Co spin by calculating the magnetic moment distribution and the orbital-projected LDOS. Figure 5b shows a magnetic moment of 2 μB on the Co atom of the Co[TCNE] complex, which corresponds to a spin S = 1 on the Co atom. Additional moment density occurs on the TCNE molecule as well as nearest neighbor Cu and N atoms in the
Figure 4. Magnetic field dependence of spin-IETS steps (a) STM image of a Co atom and a Co[TCNE] complex (0.2 V, 20 pA). (b) Corresponding surface model showing Cu (red), N (blue), and Co (yellow) atoms, as well as a scale model of TCNE. The black arrow indicates the magnetic field orientation (Bext). Also shown is the coordinate system used in the spin-Hamiltonian model. (c) Spin-IETS steps as a function of magnetic field with the tip position indicated by the black dot in (a). The tip height was set at (30 mV, 1 nA). All spectra were taken at 1 K and offset by 0.15 along the y-axis for clarity. Solid and dotted lines indicate spin Hamiltonian calculations with (S = 3/2, D = −0.45 ± 0.025, E = 0 ± 0.025, g = 1.95 ± 0.025) and (S = 1, D = 0.3 ± 0.025, E = 0.8 ± 0.025, g = 2.3 ± 0.025) respectively.
schematic of the Co[TCNE] complex, as well as the magnetic field orientation with respect to the Cu2N lattice. As the magnetic field is increased, the steps in dI/dV systematically shift toward larger voltage (Figure 4c). These spectra can be analyzed within the spin-Hamiltonian model (eq 1) to understand how magnetic properties have changed as a result of complex formation. The coordinate 1199
dx.doi.org/10.1021/nl404054v | Nano Lett. 2014, 14, 1196−1201
Nano Letters
Letter
Figure 5. Spin-polarized DFT calculations of Co[TCNE]. (a) Side and top view of fully relaxed Co[TCNE] complex on Cu2N/Cu(100). (b) Bader analysis of charge and magnetic moment (parentheses) distribution. Here, we only display charge and magnetic moment distribution of nearest neighboring atoms of the Co atom. Total charge of Cu2N slab is −0.14e. (c) Top panel is the calculated 3d orbital local density of states (LDOS) for the Co atom in Co[TCNE]. The five 3d orbitals for spin up lie below EF while 3dyz and 3dx2 for spin down exhibit significant LDOS above EF (lower panel). Bottom panel shows comparison between the calculated 3d LDOS and dI/dV spectra. (d) A comparison of the DFT-simulated and experimental STM images (V = 0.2 V).
Cu2N surface. Using the same coordinate system as for the spin-Hamiltonian, the orbital-projected LDOS on the Co atom of the complex indicates that the 3dxz, 3dxy and 3dy2−z2 orbitals lie primarily below EF (0 eV) for both spin up and down, suggesting that these orbitals are fully occupied, while the 3dyz and 3dx2 orbitals have significant LDOS above EF for spin down and are only partially occupied (Figure 5c). We compare the calculated 3d LDOS with our tunneling spectra taken on isolated Co and Co in the Co[TCNE] complex (Figure 5c, bottom panel). The prominence of occupied states for Co[TCNE] compared to isolated Co suggests significant charge transfer. These occupied states are in good agreement with the calculated LDOS, allowing us to estimate a valence electron configuration of 3d84s0 for the Co atom in Co[TCNE], compared to a configuration of 3d74s2 for a free Co atom.27 This qualitative analysis is consistent with the charge distribution (transfer), the magnetic moment, and the orbital-projected LDOS of the Co atom. Therefore, our calculations suggest that the Co spin is reduced from S = 3/2 for the isolated Co on Cu2N to S = 1 in the Co[TCNE] complex. Given S = 1, our data (c.f. Figure 4c and Supporting Information Figure S7) indicate axial and transverse anisotropy parameters that are significantly different than for the isolated atom (D = 2.75, E = 0). This suggests that the bond formation within the complex redistributes charge and magnetic moment and breaks the symmetry of the local electrostatic environment. In summary, our studies illustrate a method for studying charge transfer in metallo-organic complexes at the atomic scale. A combination of atomic resolution imaging, tunneling spectroscopy, and density functional theory is used to develop a model for charge and spin redistribution in the complexes. We demonstrate atomic manipulation methods for constructing these complexes on a thin insulating surface, which facilitates detailed studies of the magnetic properties (spin magnitude,
Kondo resonance, and magnetic anisotropy). This approach provides valuable information toward a microscopic understanding of metal/organic interfaces, relevant for information technologies and heterogeneous catalysis. Methods. Data were taken in ultrahigh vacuum (
