Indirect Magnetic Coupling of Manganese Porphyrin to a

Dec 20, 2010 - Wolfgang Kuch , Matthias Bernien. Journal of .... H. C. Herper , M. Bernien , S. Bhandary , C. F. Hermanns , A. Krüger , J. Miguel , C...
0 downloads 0 Views 2MB Size
J. Phys. Chem. C 2011, 115, 1295–1301

1295

Indirect Magnetic Coupling of Manganese Porphyrin to a Ferromagnetic Cobalt Substrate† D. Chylarecka,‡ T. K. Kim,‡,∇ K. Tarafder,§ K. Mu¨ller,‡,O K. Go¨del,| I. Czekaj,⊥ C. Wa¨ckerlin,‡ M. Cinchetti,| Md. E. Ali,§,[ C. Piamonteze,# F. Schmitt,| J.-P. Wüstenberg,| C. Ziegler,| F. Nolting,# M. Aeschlimann,| P. M. Oppeneer,*,§ N. Ballav,*,‡ and T. A. Jung*,‡ Laboratory for Micro- and Nanotechnology, Paul Scherrer Institut, Villigen, Switzerland, Department of Physics and Astronomy, Uppsala UniVersity, Uppsala, Sweden, Department of Physics and Research Center OPTIMAS, UniVersity of Kaiserslautern, Kaiserslautern, Germany, General Energy Department, Paul Scherrer Institut, Villigen, Switzerland, and Swiss Light Source, Paul Scherrer Institut, Villigen, Switzerland ReceiVed: July 22, 2010; ReVised Manuscript ReceiVed: NoVember 12, 2010

The coupling mechanism of magnetic molecules to ferromagnetic surfaces is of scientific interest to design and tune molecular spintronic interfaces utilizing their molecular and surface architecture. Indirect magnetic coupling has been proposed earlier on the basis of density functional theory +U (DFT+U) calculations, for the magnetic coupling of manganese(II) porphyrin (MnP) molecules to thin Co films. Here we provide an experimental X-ray magnetic circular dichroism (XMCD) spectroscopy and scanning tunneling microscopy (STM) study of manganese(III) tetraphenylporphyrin chloride (MnTPPCl) on rough (exhibiting a high density of monatomic steps) and smooth (exhibiting a low density of monatomic steps) thin Co films grown on a Cu(001) single crystal toward the assessment of the magnetic coupling mechanism. After deposition onto the surface, MnTPPCl molecules were found to couple ferromagnetically to both rough and smooth Co substrates. For high molecular coverage, we observed higher XMCD signals at the Mn L-edges on the smooth Co substrate than on the rough Co substrate, as expected for the proposed indirect magnetic coupling mechanism on the basis of its predominance on the flat surface areas. In particular, DFT+U calculations predict a weak ferromagnetic molecule-substrate coupling only if the chloride ion of the MnTPPCl molecule orients away (Co-Mn-Cl) from the Co surface. Introduction Molecular spintronics, the emerging discipline to engineer spin-dependent electronic transport at interfaces with organic molecules, has rapidly attained the attention of the scientific community.1-5 A stimulating perspective is the possibility of designing new spintronic interfaces with tunable magnetic and electronic interactions enabled by the structure of the organic compound and potentially lower processing costs compared to inorganic materials. In organic magnetic molecules, the molecule’s spin is typically localized around unpaired electrons, e.g., at the open shell of a transition metal ion or at involved stable radical such as, nitric oxide. The spacing between individual spin systems is typically larger in organic than in inorganic spintronic materials thereby inducing low or no electronic conductivity and paramagnetism in the bulk form of the material. After adsorption on ferromagnetic surfaces, however, magnetic molecules have been demonstrated to couple magnetically to the substrate allowing for magnetic ordering at room temperature. The mechanism of the exchange coupling †

Part of the “Alfons Baiker Festschrift”. * To whom correspondence should be addressed. E-mail: thomas.jung@ psi.ch, [email protected], and [email protected]. ‡ Laboratory for Micro- and Nanotechnology, Paul Scherrer Institut. § Uppsala University. | University of Kaiserslautern. ⊥ General Energy Department, Paul Scherrer Institut. # Swiss Light Source, Paul Scherrer Institut. ∇ Present address: Leibniz Institute for Solid State and Materials Research, Dresden, Germany. O Present address: Center for Functional Nanomaterials, Brookhaven National Laboratory, Upton, New York 11973, United States. [ Present address: Theoretische Chemie, Ruhr-Universita¨t Bochum, Bochum, Germany.

to the surface atoms in this case crucially depends on the bonding. Therefore, detailed knowledge on the substrateadsorbate interaction is an important prerequisite for the detailed analysis and the design of the electronic and magnetic interface properties. Two important cases can be discriminated, i.e., the direct exchange coupling of orbitals containing unpaired electrons with the electronic system of the substrate and the indirect exchange coupling where the coupling derives from an intermediate atomic species linking the unpaired electron system in the molecule with the ferromagnetic substrate electrons. In recent studies, adsorbed metalloporphyrins and metallophthalocyanines on magnetized substrates were shown to possess a magnetic moment which either can be ferromagnetically coupled to the magnetic substrate or alternatively exhibits antiferromagnetic coupling through ad-atoms like oxygen.6-10 Only few studies provided evidence toward the detailed coupling mechanism by means of theoretical calculations.7,9 Notably, chemical reactions eventually change the chemical constitution of the surface and of the adsorbate and thereby the bonding situation as well as the coupling.11 Neither the exact species present in earlier experiments nor the adsorption geometry playing a crucial role in any physical effect has been unambiguously revealed in the earlier studies. In this paper, we present experimental X-ray magnetic circular dichroism (XMCD) evidence for the magnetic coupling mechanism of a manganese(III) tetraphenylporphyrin chloride (MnTPPCl) molecule to a premagnetized cobalt (Co) thin film. We complement the experiments by density functional theory (cluster DFT and DFT+U) calculations, investigating the coupling of manganese porphyrin chloride (MnPCl) to a Co surface. By combination of XMCD with complementary surface analysis, in particular

10.1021/jp106822s  2011 American Chemical Society Published on Web 12/20/2010

1296

J. Phys. Chem. C, Vol. 115, No. 4, 2011

scanning tunneling microscopy (STM) and X-ray photoelectron spectroscopy (XPS), detailed information about the chemical species adsorbed on the surface as well as an assessment of molecular orientation are provided. The MnTPPCl molecule consists of a porphyrin ring with four nitrogen (N) atoms coordinated to a manganese (Mn) ion in the center and an axially coordinated chloride (Cl) ion. Similarly to chloro(subphthalocyaninato) boron(III) (SubPc), the molecule has a dipole moment with a partial positive charge located on the central metal ion (Mn) and a partial negative charge located on the axial ligand (Cl).12,13 The Mn ion in bulk MnTPPCl is found in the 3+ oxidation state and in high-spin configuration with four unpaired electrons. However, as shown in earlier work, the Mn ion can be reduced to 2+ upon adsorption of MnTPPCl on a metallic Co substrate.6,7,10 A similar moleculesiron(III) octaethylporphyrin chloride (FeOEPCl)swas shown to couple ferromagnetically to premagnetized Co and nickel (Ni) thin films. The origin of this magnetic coupling was explained by DFT+U calculations to arise from an indirect exchange mechanism between iron and the substrate mediated by N atoms of the porphyrin ring.7 Notably the exact chemical species involved in the exchange coupling has not been revealed in this latter study. Due to the similarity of MnTPPCl and FeOEPCl also in their property to couple ferromagnetically to Co and Ni thin films and antiferromagnetically to oxygen-covered Co and Ni, we present further complementary studies toward the identification of the coupling mechanism which is at first sight expected to be similar for both these molecules, i.e., indirect.7,14,15 To experimentally verify the indirect exchange hypothesis for the MnTPPCl molecule, we evaporated it onto atomically smooth (step defect poor) and rough (step defect rich) Co films and compared the respective normalized XMCD signals at the Mn L-edges. The minimum spacing between Mn and Co in an assumed planar molecular orientation on the surface is calculated as ∼3.67 Å (discussed below). This calculated distance is too large for the overlap between the 3d orbitals of Mn and Co to occur, and therefore direct exchange coupling between Mn and Co is not probable. On the basis of this assignment, the direct exchange coupling between MnTPPCl and the Co film would become more dominant in the case of selective adsorption of molecules on steps or kinks of the substrate, where porphyrins tend to adsorb preferentially, due to the generally more favorable physical and/or chemical interaction of the π-system with the substrate atoms at closer proximity.16 In this case, only the proportion of molecules adsorbed on defects is expected to contribute to the observed circular dichroism. The present study aims to resolve this issue by investigating the magnetic coupling by XMCD spectroscopy and the adsorption geometry on rough and smooth ferromagnetic substrates by STM. Methods In the following subsections, the stages of the experiment as well as the details of theoretical calculations will be presented. Sample Preparation. All samples were prepared in a multichamber ultrahigh vacuum (UHV) system with a base pressure below 5 × 10-10 mbar. We have developed a reproducible method to prepare either smooth or rough Co films on Cu(001) single crystals. The Cu(001) single crystal (Mateck) was cleaned by repeated Ar+ sputtering/annealing cycles until no C1s, O1s, and Co2p peaks were identifiable by XPS measurement. About 20 monolayers (20 ML ∼ 35 Å) of Co film were deposited by electron-beam-induced evaporation onto a clean Cu(001) single

Chylarecka et al. crystal. The preparation of the Co films was performed in two steps: First, 10 ML of Co were evaporated onto the sample kept at room temperature. This provides a sufficient diffusion barrier to prevent segregation of Cu on the surface. After annealing of this 10 ML thick Co layer for 30 min at ∼240 °C, another 10 ML of Co were evaporated onto the sample kept at ∼190 °C to obtain atomically flat surfaces with a terrace size in the range of 25 nm. Low-energy electron diffraction (LEED) patterns show that the Co grows on the Cu(001) in the face centered cubic (fcc) structure, which is in good agreement with previous studies.10 The rough Co films were obtained by subsequent Ar+ ion bombardment (2 min 30 s, 0.7 eV, Isample ) 1.5 µA) at room temperature. No further annealing was performed before the MnTPPCl molecules (purchased from Porphyrin Systems, 98% purity) were evaporated in situ from a commercial molecular evaporator onto the smooth or rough Co thin films held at room temperature. The deposition rate was controlled by a quartz crystal microbalance (QCMB), and the coverage was calibrated by STM and XPS measurements. STM and XPS Measurements. STM and XPS measurements were carried out at room temperature. STM images were taken with electrochemically etched and Ar+ sputtered tungsten tips in constant current mode, and sample bias voltages as they are provided in the figure captions refer to a grounded tip. STM images were processed using the WsXM software.17 XPS was performed using a nonmonochromatized X-ray source (Al KR). XMCD Measurements. XMCD measurements were performed at the Surfaces/Interfaces: Microscopy (SIM) beamline of the Swiss Light Source (SLS) in a UHV chamber with base pressure of lower than ∼3 × 10-9 mbar. The beamline provides high brilliance X-ray light in the energy range of 130-2000 eV from two elliptical twin undulators, which permit switching of the photon helicity within a few seconds.18 Molecules were evaporated in situ onto the magnetized Co films in the XMCD chamber by using a commercial three-crucible evaporator. Co thin films were prepared as described in the Sample Preparation section, analyzed with XPS and STM, and then transferred to the SLS with a portable vacuum chamber with the base pressure of ∼3 × 10-10 mbar.19 The sample was magnetized by an electromagnet mounted outside of the chamber exerting a magnetic field of ∼125 mT at the sample position, which was sufficient to obtain single domain magnetization along the easy axis of the Co thin film substrates. The evaporation rate was controlled with a QCMB. XMCD spectra were recorded at room temperature in total electron yield (TEY) mode at Mn and Co L2,3 edges in remnant magnetization parallel to the sample plane. All spectra were recorded at the grazing incident angle of 70° and normalized to the incident photon flux. Dichroic spectra for two different molecular coverages were normalized to the respective background and average X-ray absorption spectroscopy (XAS) signals. Reverse magnetization was reproducibly achieved by inverting the externally applied magnetic field. DFT Calculations. DFT cluster calculations were performed for two porphyrin configurations: with and without the axial ligand (see insets in Figure 4). The Co(001) surface was represented by the Co14 cluster. The electronic structure of the clusters was calculated using the program code StoBe20 and the nonlocal generalized gradient approximation (GGA) functionals according to Perdew, Burke, and Ernzerhof (RPBE),21,22 to account for electron exchange and correlation. All Kohn-Sham orbitals are represented by linear combinations of atomic orbitals (LCAOs) using extended basis sets of contracted Gaussians from atom optimizations.23,24 During relaxation, the atoms in the supported clusters were allowed to move in 3D space.

Coupling of MnP to a Ferromagnetic Co Substrate We have furthermore performed DFT+U calculations to investigate the chemical and magnetic interaction of a MnPCl molecule with a ferromagnetic Co surface. Our first-principles DFT+U calculations have been performed with the VASP package, which uses pseudopotentials together with the projector augmented wave approach.25 In contrast to the cluster calculations, we represented the molecule and three atomic Co layers within a large supercell and applied periodic boundary conditions to capture the periodicity of the surface. Also, to reduce computational effort we have removed the four phenyl rings (and replaced these with hydrogens); i.e., the DFT+U simulation is performed for a MnPCl molecule. A kinetic energy cutoff of 400 eV was employed for the plane waves. For the DFT exchange-correlation functional, we used the PBE GGA.21 Recently, it has been reported26,14 that the commonly used DFTGGA approach fails in describing adequately the high-spin state of metalloporphyrins. A correct description of the molecular spin state can be achieved when strong Coulomb correlations within the open 3d shell are taken into account. The latter can appropriately be done through the DFT+U approach, in which an additional Coulomb interaction in the 3d shell is supplemented, defined through the on-site Coulomb and exchange parameters U and J. In our calculations, the U and J on Mn were taken to be 4 and 1 eV, respectively (see refs 26 and 14). In our simulations, we have modeled the metallic surface through three atomic layers of Co (adopting the fcc Co lattice parameter of 3.61 Å). We furthermore performed full 3D geometric optimizations of the porphyrin molecule and its position on the surface. To reduce the computational effort, we kept the bottom two layers of Co atoms fixed in our simulations. Through test calculations, we verified that this is indeed a good approximation; complete relaxation of all atomic distances revealed that the moleculesurface distances (and consequently, interactions) did not change, but a small relaxation of the top to second Co layer occurred. The whole simulation cell consisted of 188 atoms; for the reciprocal space sampling, we used 2 × 2 × 2 Monkhorst-Pack k-points. We performed the geometric relaxation of the MnPCl on the surface for two different configurations: in the first case, the Cl atom is on top of the molecule oriented away from the surface, whereas in the second case the Cl atom is between the porphyrin ring and surface. Results and Discussion Smooth and rough Co films can be clearly distinguished on the basis of their surface morphologies mapped by STM (Figure 1). While the smooth Co films are characterized by broad, rounded terraces with very few defects, the rough surface contains grooves, which form a characteristic “wrinkled” shape. The original morphology of the Co as deposited before sputtering can still be distinguished by the round terrace shape. Differences between the two substrates can be clearly recognized in the respective height profiles in Figure 1, which were taken along the lines depicted in the corresponding STM images over a distance of ∼50 nm. One can identify the two flat terraces ∼20 nm wide in the surface profile corresponding to the smooth sample, whereas in the case of the rough sample the height profile is more corrugated indicating far narrower flat areas (∼3 nm). MnTPPCl molecules were thermally sublimed onto the magnetized Co films with smooth or rough morphology. While there were no C 1s, N 1s, and Cl 2p XPS signals observed on clean Co substrates, the respective signals appeared after the evaporation of MnTPPCl molecules. Furthermore, the XPS data

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1297

Figure 1. 2D STM images of “smooth” and “rough” preparations of the Co substrates. (a) Smooth (I ) 0.12 nA, U ) -2.5 V) and (b) rough (I ) 0.25 nA, U ) -1.4 V) 20 ML thin Co films on Cu(001). Insets present respective height profiles along the ∼50 nm long lines depicted in the images. The crystallographic directions of the substrate are the same in both images and are indicated by arrows in (a).

obtained for a thick layer of MnTPPCl molecules revealed a ratio of 44.1:4.0:0.9 between C:N:Cl atoms which compares well to the chemical mass formula MnC44N4H28Cl and thus indicates that molecules stay intact during the evaporation process. The molecular integrity of the MnTPPCl molecules before and after the evaporation was further confirmed by a UV-vis spectroscopy (data shown in ref 10). The shape of the molecules in the STM images is cross-like, and the in-plane dimensions are calculated by DFT to be 1.24 × 1.24 nm2 (cf. Figure 2c). It is worthwhile to note that the phenyl substituents are oriented out of the porphyrin plane due to a steric repulsion mechanism.27 In a previous study of MnTPPCl and of FeOEPCl,6,7 it was shown by near-edge X-ray absorption fine structure spectroscopy (NEXAFS) that these molecules lie flat on the Co substrate. In the present work, STM images of MnTPPCl molecules on both smooth (Figure 2a) and

1298

J. Phys. Chem. C, Vol. 115, No. 4, 2011

Chylarecka et al.

Figure 2. Recognition of molecular features after adsorption of MnTPPCl on smooth and rough Co substrates. (a) 0.9 ML on smooth Co film, I ) 0.12 nA, U ) -2.3 V, and two different molecular shapes: with five and four lobes. (b) 0.3 ML on rough Co film, I ) 0.22 nA, U ) -1.2 V, and two molecular shapes. (c) Scheme of the MnTPPCl molecule. (d), (e) Two consecutive STM images displaying a change of the shape of the molecule from “five-lobe” to “four-lobe” during scanning, 10 × 10 nm2, I ) 0.12 nA, U ) -2.4 V (d), U ) -3.0 V (e). The crystallographic directions of the substrate are the same in each image and are indicated by arrows in (a).

rough (Figure 2b) magnetized Co films confirm the adsorbed flat-lying configuration of the MnTPPCl molecules on the surface. The size of the molecule measured in the STM is ∼1.17 × 1.33 nm2, which is in good agreement with the calculated value. The rectangular shape might be associated with the conformational flexure of the porphyrins along the dihedral phenyl-porphyrin bond.27 On the smooth Co substrate (Figure 2a), the molecules (coverage ∼0.9 ML) are adsorbed in a random fashion and do not apparently occupy preferred adsorption sites, indicating a significant substrate-molecular interaction which retards selfassembly of the layer, as reported earlier.10 Similarly, on the rough surface, where the displayed coverage is much smaller (∼0.3 ML, Figure 2b), the molecular adsorption also appears random. This observation can be explained by the considerable bonding strength and the correspondingly low diffusion of MnTPPCl on the Co substrate upon deposition and relaxation. Additionally to the well-defined cross-shaped molecules on the rough surface, we can also distinguish a significant number of isolated protrusions in Figure 2b. We tentatively assign these structures to the products of a molecular decomposition process on the substrate, possibly chlorine. This assignment is supported by our earlier reported evidence that MnTPPCl molecules adsorbed on metallic Co undergo dissociation leading to the removal of the Cl atom, which remains bound to the substrate.10 On the smooth Co surface shown in Figure 2a, such protrusions are also observed, but they are less apparent due to the almost full monolayer coverage.

On both types of Co substrates there are two kinds of MnTPPCl molecules noticeable in the STM images: characterized by four and by five lobes they resemble different faces of a dice (cf. zoom sections provided in Figures 2a and 2b). The fraction of five-lobe molecules counted from several STM images is ∼77% for both substrates. The difference between the two observed molecular geometries was recently attributed to a change in the oxidation state of Mn upon adsorption on the metallic substrate.10 However, no conclusive evidence was shown to unambiguously explain the observed effect, and therefore we will address this issue again in the following paragraphs. Figures 2d and 2e exhibit two consecutive STM images taken over the same area of ∼0.9 ML of MnTPPCl molecules on the smooth Co substrate, where one MnTPPCl molecule (marked by a black oval) changes its geometry from five-lobe to fourlobe. Any tip-induced effect that might explain this transition can be excluded, as the particular molecule was the only one affected in the image, and both types of molecules continue to be present in the image. Additionally, there was a dramatic difference neither in the bias voltage between the two images (-2.4 and -3.0 V, respectively) nor in the tunneling current (0.12 nA for both) that could induce such a change in the molecular appearance. We attribute this transition to a variation in the valence of the Mn ion, e.g., from the 2+ to 3+ oxidation state. Assuming the removal of Cl from MnTPPCl when adsorbed on Co, as evidenced in our earlier report,10 the resulting molecule

Coupling of MnP to a Ferromagnetic Co Substrate

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1299

Figure 3. Magnetization of Mn and Co in the MnTPPCl/Co system. (a) MnTPPCl molecules on smooth (blue) and rough (red) Co substrates for 0.3 (top) and 0.9 (bottom) ML coverage (for easy comparison, only one direction of the Co substrate magnetization (M+) is shown) and (b) XMCD spectra of the Co substrate for two opposite magnetization directions: M+ and M-.

would exhibit an empty position in the octahedral symmetry of Mn which provides a valence to be occupied by an externally supplied ligand. It is a well-known property of Mn(II)TPP to immediately oxidize in air,28 and therefore it is a reasonable assumption that it also reacts with the components of the residual gas in the UHV. Due to the low vacuum pressure of the order of 10-10 mbar in the experimental chamber, this reaction is very rare and difficult to observe. Consequent to this hypothesis, the five-lobe and the four-lobe molecular appearances are tentatively assigned to be Mn(II)TPP and Mn(III)TPPCl (or oxidized Mn(III)TPP), respectively.10 To address the question about the character of the magnetic interaction between MnTPPCl layers and the Co substrate, we have plotted Mn L2,3 edges XMCD spectra recorded for two coverages of MnTPPCl: ∼0.3 ML (Figure 3a, top) and ∼0.9 ML (Figure 3a, bottom) on smooth (black line) and rough (red line) Co substrates. In accordance with our previous results, ferromagnetic coupling of MnTPPCl molecules with both smooth and rough Co substrate was observed for two opposite magnetization directions.6,10 Figure 3a shows a clear effect of surface roughness on the XMCD signal for high coverage (0.9 ML) of MnTPPCl molecules which was neither observed nor reported earlier, while there is almost no difference observed between the XMCD signals recorded for smooth and for rough Co substrates for low molecular coverage (0.3 ML). The similar shape of the XMCD signal for 0.3 ML of MnTPPCl adsorbed on the smooth and on the rough substrate indicates that the adsorption statistics (on the step edge sites vs terrace sites) are similar on both. This means that at the same coverage (here 0.3 ML) the number of molecules adsorbed at step edges is about equal for smooth and for rough Co films. Therefore, no conclusion about the origin of the magnetic interaction can be drawn from the experiments at this coverage. At higher molecular coverage, however, the surface gets progressively occupied, and the adsorption site statistics may be altered by the tendency of the porphyrin to adhere to less favorable sites before nucleating a second layer. The significantly reduced (∼30%) XMCD signal for a close to full monolayer coverage of MnTPPCl on the rough Co substrate provides evidence that: (i) the statistics of adsorption is modified due to the high density of step edge sites present and (ii) the Mn coupling of MnTPPCl at the latter sites is reduced. Thus, our data support that the dichroism, i.e., the magnetic coupling, is favored by the flat adsorption geometry of the porphyrin and is reduced by molecules taking step or kink positions at higher molecular coverage on rough substrates. This also indicates that the

Figure 4. Calculated partial DOS of the Mn porphyrin with and without the axial Cl ligand. (a) With chlorine. (b) Without chlorine. Solid vertical bars indicate common Mn and N partial DOS peaks, and dashed vertical bars denote common N and Co partial DOS peaks. Black arrows represent spin-up and spin-down configuration, respectively. Insets represent the structure of the Mn porphyrin molecule above the Co14 cluster with the chlorine atom pointing up (a) and without chlorine (b).

chemical and/or electronic interaction of the molecules differs between the step edge position and the flat adsorption on the terrace. To complement the experimental results, we have performed two sets of theoretical calculations: (i) cluster DFT for Mn(II)TPP and Mn(III)TPPCl and (ii) DFT+U for Mn(III)PCl with the Cl atom pointing either up (toward vacuum) or down (toward the substrate). Insets of Figure 4 present the structure of the porphyrin placed above the Co(001) surface, which is represented by a Co14 cluster. The optimal Mn-Co distances in cluster-DFT calculations are equal to ∼4.15 and ∼3.67 Å for the Cl-ligated and -unligated Mn porphyrins, respectively. Whereas the N-Co distance is computed to be ∼3.88 Å, the stabilization energies for the Cl-ligated and -unligated Mn porphyrins have been determined as -0.29 and -0.33 eV, respectively. The fact that numerical models converge for both molecular configurations further supports our experimental evidence that both these forms can be present at the surface of the Co substrate. Figure 4 shows the spin-resolved partial density of states (DOS) for the porphyrin with and without chlorine supported at the Co cluster (black arrows represent spin-up and spin-down configuration, respectively). The hybridization of Mn and N electronic states can be realized from the coincidence of DOS peaks at the same energies for Mn 3d and N 2p states (marked by solid vertical lines in Figure 4). A small Co-N hybridization, however, is evidenced by the calculated partial

1300

J. Phys. Chem. C, Vol. 115, No. 4, 2011

Figure 5. Calculated spin-polarized partial DOS for a MnPCl molecule on a Co substrate with the Cl atom oriented away from the surface. Joint partial DOS peaks illustrate that both chlorine and nitrogen bond to Mn. The blue vertical bars illustrate a bonding of the molecule’s nitrogens to Co surface atoms and the green vertical bars a bonding between Mn and Cl within the molecule. The Fermi level (EF) is at 0 eV and the spin-up, respectively, spin-down partial DOS panels are indicated by the arrows.

DOS (dotted vertical lines), which suggests that there is a weak electronic interaction between the N centers of the porphyrin and the Co centers of the substrate. Such result was also shown by Wende et al. for the coupling of Fe(II)OEP to Co(001).7 Assuming that the coupling mechanism of Mn(II)TPP is somewhat similar to Fe(II)OEP, we have performed DFT+U calculations on an intact Mn(III)PCl molecule to complement the earlier work. In our calculations we focus on the chemical and magnetic interactions of the Mn(III)PCl with the Co surface. As expected, we find that the orientation of the Cl atom largely influences these interactions. The calculations were performed for the two possibilities of MnTPPCl adsorption: with Cl pointing (i) up (toward vacuum) and (ii) down (toward the substrate). Considering first the situation (i), we do obtain a stable molecular configuration, in which the central Mn atom has somewhat moved (by ∼0.4 Å) out of the porphyrin plane due to the bonding to the chlorine. The distance of the porphyrin ring to the Co substrate is 3.49 Å, typical for physisorbed porphyrins. The computed molecular spin is S ≈ 2, i.e., practically the same as the one expected for a free MnPCl molecule. The coupling of the molecule’s spin to that of the surface Co atoms is found to be weak, in the order of less than 5% of the MnP/Co interaction which calculates to 34 meV,15 yet a parallel (ferromagnetic) coupling is preferred. In Figure 5 we show the DFT+U computed partial DOS of this system. We use here the experience which has been gained in the past when using atom-resolved DOS plots for composite materials.14,15 The occurrence of narrow partial DOS peaks at the same energy exemplifies the occurrence of, in this case, bonding of Mn and Cl orbitals. In Figure 5 green vertical lines indicate coinciding Cl and Mn orbitals, and blue vertical lines indicate the positions of coinciding N and Co orbitals. Therefore, it is apparent that the Cl atom binds to Mn; furthermore, Mn is bound to the nitrogen through the chemical structure of the porphyrin, and N atoms exhibit bonding to the Co substrate. This observation illustrates the role of the nitrogen atoms in mediating an indirect magnetic coupling of the Mn and Co atoms. The situation (ii) in which the Cl atom is located between the Mn and the Co surface gives rise to a rather different chemical bonding and, consequently, magnetic interaction. This geometry is unfavorable for the bonding of a monovalent chlorine, which does not sustain two bonds, one to the Mn and one to Co atom. We find in the structural relaxations that the

Chylarecka et al. Cl atom moves away from the Mn and closer to a surface Co atom; i.e., the Mn-Cl bonding weakens in favor of a stronger Cl-Co bond. This shift of the chemical bonding is reflected in the molecule’s spin which is increased to S ≈ 5/2, which is in fact the high-spin value expected for a free MnP molecule.15 Also, the magnetic coupling of the molecule’s spin to the surface has now become weakly antiferromagnetic. This finding suggests that when the Cl binds to a Co atom a MnP molecule might be liberated and could interact ferromagnetically with the Co substrate at another location. Notably, this is supported by the experimental observation of the Cl dissociation as reported earlier.10 Overall, the presented theoretical calculations based on cluster DFT and DFT+U complemented our experimental XMCD and STM data. Specifically, we observe a pronounced Mn-XMCD signal (both on smooth and rough Co) at ∼639.9 eV and a shoulder at ∼641.2 eV. We assign the former (peak) structure to the Mn2+ species, i.e., to MnTPP formed by removal of Cl from MnTPPCl adsorbing onto the substrate surface with the Cl atom pointing toward Co as predicted by the DFT+U calculations. Consistently, we assign the second (shoulder) structure to the Mn3+ species, i.e., to MnTPPCl adsorbing with Cl pointing up. On the basis of the statistics of the molecular appearance in our STM data, we assign the former (peak) structure to the five-lobe molecular feature and the latter (shoulder) structure to the four-lobed feature. Both species are ferromagnetically coupled to Co as revealed by the XMCD data which is consistent with the presented DFT+U calculations. Conclusions Conclusive evidence has been provided from XMCD, STM, and first-principles cluster DFT and DFT+U calculations toward the identification of the molecular species involved in the magnetic coupling observed after deposition of MnTPPCl molecules to premagnetized smooth and rough Co thin film substrates. STM images evidence the flat-lying adsorption geometry of Mn porphyrin molecules on the surface. DFT calculations confirm the coexistence of ferromagnetically coupled Mn porphyrins, with and without Cl, and confirm the assignment of the species on the basis of the experimental results. Specifically, the geometry with the Cl atom located between Mn and the Co surface is not favorable in the calculations, while the geometry with the Cl oriented away from the surface is chemically more stable. A second conclusion results from the XMCD signal at the Mn L2,3-edge which was found to be higher on smooth than on rough Co films at high molecular coverage: this observation strongly supports the indirect coupling mechanism in full agreement with the presented calculations. Specifically, the overlap of the porphyrin N p-orbitals with the Co d-orbitals is observed to be strengthened as the Mn atom is moved out of the porphyrin plane. The approach taken in this study, to combine spectromicroscopy correlation experiments with theory, uniquely highlights the interplay of molecular geometry and chemical and magnetic interactions. Acknowledgment. We acknowledge financial support from the Swiss National Science Foundation (SNSF), National Center of Competence in Research (NCCR), and Research Equipment founding (R’Equip), Switzerland, and from the C. Tryggers Foundation, Sweden. We also acknowledge the Swedish National Infrastructure for Computing (SNIC). Part of this work has been performed at the Swiss Light Source (SLS), Paul Scherrer Institut, Villigen, Switzerland. The authors sincerely

Coupling of MnP to a Ferromagnetic Co Substrate thank Rolf Schelldorfer and Andrea Steger for technical support all throughout, Arantxa Fraile-Rodriguez for providing help during beamtimes at the SLS, Ernst Meyer for interesting and helpful discussions, and Jan Girovsky, Stefan Lach, Jens Sauther, Sabine Neuschwander, and Oleksiy Andreyev for assistance on lab-based and synchrotron experiments. N.B. thanks the Holcim Foundation for the Advancement of Scientific Research, Switzerland, for a research scholarship. References and Notes (1) Boyen, H.-G.; Ziemann, P.; Wiedwald, U.; Ivanova, V.; Kolb, D. M.; Sakong, S.; Gross, A.; Romanyuk, A.; Bu¨ttner, M.; Oelhafen, P. Nat. Mater. 2006, 5, 394. (2) Naber, W. J. M.; Faez, S.; van der Weil, W. G. J. Phys. D: Appl. Phys. 2007, 40, R205. (3) Bogani, L.; Wernsdorfer, W. Nat. Mater. 2008, 7, 179. (4) Cinchetti, M.; Heimer, K.; Wu¨stenberg, J.-P.; Andreyev, O.; Bauer, M.; Lach, S.; Ziegler, C.; Gao, Y.; Aeschlimann, M. Nat. Mater. 2008, 8, 115. (5) Dediu, V. A.; Hueso, L. E.; Bergenti, I.; Taliani, C. Nat. Mater. 2009, 8, 707. (6) Scheybal, A.; Ramsvik, C.; Bertschinger, R.; Putero, M.; Nolting, F.; Jung, T. A. Chem. Phys. Lett. 2005, 411, 214. (7) Wende, H.; Bernien, M.; Luo, J.; Sorg, C.; Ponpandian, N.; Kurde, J.; Miguel, J.; Piantek, M.; Xu, X.; Eckhold, Ph.; Kuch, W.; Baberschke, K.; Panchmatia, P. M.; Sanyal, B.; Oppeneer, P. M.; Eriksson, O. Nat. Mater. 2007, 26, 516. (8) Iacovita, C.; Ratei, M. V.; Heinrich, B. W.; Brumme, T.; Kortus, J.; Limot, L.; Bucher, J. P. Phys. ReV. Lett. 2008, 101, 116602. (9) Bernien, M.; Miguel, J.; Weis, C.; Ali, Md. E.; Kurde, J.; Krumme, B.; Panchmatia, P. M.; Sanyal, B.; Piantek, M.; Srivastava, P.; Baberschke, K.; Oppeneer, P. M.; Eriksson, O.; Kuch, W.; Wende, H. Phys. ReV. Lett. 2009, 102, 047202. (10) Chylarecka, D.; Wa¨ckerlin, C.; Kim, T. K.; Mu¨ller, K.; Nolting, F.; Kleibert, A.; Ballav, N.; Jung, T. A. J. Phys. Chem. Lett. 2010, 1, 1408. (11) Wa¨ckerlin, C.; Chylarecka, D.; Kleibert, A.; Mu¨ller, K.; Iacovita, C.; Nolting, F.; Jung, T. A.; Ballav, N. Nat. Commun. 2010, 1:61, doi: 10.1038/ncomms1057.

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1301 (12) de Wild, M.; Berner, S.; Suzuki, H.; Yanagi, H.; Schlettwein, Ivan, S. D.; Baratoff, A.; Gu¨ntherodt, H.-J.; Jung, T. A. Chem. Phys. Chem. 2002, 10, 881. (13) Berner, S.; de Wild, M.; Ramoino, L.; Ivan, S.; Baratoff, A.; Gu¨ntherodt, H.-J.; Suzuki, H.; Schlettwein, D.; Jung, T. A. Phys. ReV. B 2003, 68, 115410. (14) Oppeneer, P. M.; Panchmatia, P. M.; Sanyal, B.; Eriksson, O.; Ali, Md. E. Prog. Surf. Sci. 2009, 84, 18. (15) Ali, Md. E.; Sanyal, B.; Oppeneer, P. M. J. Phys. Chem. C 2009, 113, 14381. (16) Jung, T. A.; Schlittler, R. R.; Gimzewski, J. K. Nature 1997, 386, 696. (17) Horcas, I.; Fernandez, R.; Gomez-Rodriguez, J. M.; Colchero, J.; Gomez-Herrero, J.; Baro, A. M. ReV. Sci. Instrum. 2007, 78, 013705. (18) Flechsig, U.; Nolting, F.; Fraile Rodrı´guez, A.; Krempasky, J.; Quitmann, C.; Schmidt, T.; Spielmann, S.; Zimoch, D. AIP Conf. Proc. 2010, 1234, 319. (19) Scheybal, A.; Mu¨ller, K.; Bertschinger, R.; Wahl, M.; Bendounan, A.; Aebi, P.; Jung, T. A. Phys. ReV. B 2009, 79, 115406. (20) The program package StoBe is a modified version of the DFTLCGTO program package DeMon, originally developed by A. St.-Amant and D. Salahub (University of Montreal), with extensions by L. G. M. Pettersson and K. Hermann. (21) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77, 3865. (22) Hammer, B.; Hansen, L. B.; Norskov, J. K. Phys. ReV. B 1999, 59, 7413. (23) Labanowski, J. K.; Anzelm, J. W., Eds. Density Functional Methods in Chemistry; Springer -Verlag: New York, 1991. (24) Godbout, N.; Salahub, D. R.; Andzelm, J.; Wimmer, E. Can. J. Phys. 1992, 70, 560. (25) (a) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558–561. (b) Kresse, G.; Furthmu¨ller, J. Phys. ReV. B 1996, 54, 11169–11186. (c) Blo¨chl, P. E. Phys. ReV. B 1994, 50, 17953–17979. (26) Panchmatia, P. M.; Sanyal, B.; Oppeneer, P. M. Chem. Phys. 2008, 343, 47–60. (27) Jung, T. A.; Schlittler, R. R.; Gimzewski, J. K.; Tang, H.; Joachim, C. Science 1996, 271, 181. (28) Thomassen, P. J. PhD thesis; Radboud University Nijmegen, Netherlands, 2006.

JP106822S