IBM Research Atomic-scale Engeered Spins at a Surface Chiung-Yuan Lin IBM Almaden Research Center © 2002 IBM Corporation © 2006 2005 IBM Corporation IBM Research Nanomagnetism and Information Technology Magnetism is at the heart of data storage. Courtesy of Hitachi Many novel computations schemes are based on manipulation of magnetic properties. J.R. Petta et al. Science 309, 2180 (2005) A. Imre et al. Science 311, 205 (2006) © 2006 IBM Corporation IBM Research Nanomagnets Fabricated nanomagnets can recreate model spin systems such as spin ice. R.F. Wang et al., Nature 439, 303 (2006) A small number of atomic spins can be coupled in metal clusters or molecular magnetic structures. Fe8, courtesy ESF. M.B. Knickelbein Phys. Rev. B 70, 14424 (2004) © 2006 IBM Corporation IBM Research Assembly and Measurement of Nanomagnets Top-down Bottom-up Atomic-scale control O P Manipulate structures P O © 2006 IBM Corporation IBM Research STM Studies of Atomic-Scale Spin-Coupling Manipulation on thin insulators: build individual nanomagnets with an STM 10Mn chain Mn atom Energy Energy |ST,m> |ST,m> |5/2,+5/2> |1,+1> |5/2,+3/2> Spin Excitation Spectroscopy: collective spin excitations of individual nanostructures |1,0> |5/2,+1/2> |5/2,-1/2> |1,-1> |5/2,-3/2> |0,0> |5/2,-5/2> Magnetic Field Science 312, 1021 (2006) © 2006 IBM Corporation IBM Research Keep it Simple: Free Mn Atom 4s 3d Mn: S = 5/2, L = 0, J = 5/2 Half filled d-shell Weak spin-orbit interactions © 2006 IBM Corporation IBM Research Scanning Tunneling Spectroscopy: LDOS Ef eV tip sample dI/dV 0 V Features in the local DOS are reflected in dI/dV. © 2006 IBM Corporation IBM Research Magnetic Atoms on Surfaces Magnetic atom Atom’s spin is screened by conduction electrons (Kondo effect) Metal surface Thin insulating layer A thin insulating layer may isolate the atomic spin © 2006 IBM Corporation IBM Research Inelastic Electron Tunneling Spectroscopy Ef D tip eV eV X sample D Non-magnetic tip Thin insulator Magnetic atom |eV| < >D Elastic Channel Open Inelastic InelasticChannel ChannelClosed Open dI/dV Non-magnetic sample -D kBT < D 0 © 2006 IBM Corporation D σe+σie σe eV IBM Research Methods of Electronic-structure Calculation Plane wave Atomic partial wave Atomic partial wave Atomic spheres Interstitial region Full-potential Linearized Augmented Plane Wave basis Periodic-slab geometry (5-layer Cu + 8-layer vacuum) Density Functional Theory Generalized Gradiant Approximation (GGA) PBE96: Perdew et al., PRL 77, 3865 (1996) Structure Optimization © 2006 IBM Corporation IBM Research FLAPW basis Periodic-slab geometry Cu vacuum Cu vacuum Cu vacuum Methods of Electronic-structure Calculation Cu (5-layer Cu + 8-layer vacuum) Density Functional Theory Generalized Gradiant Approximation (GGA) PBE96: Perdew et al., PRL 77, 3865 (1996) Structure Optimization © 2006 IBM Corporation IBM Research Methods of Electronic-structure Calculation 2 Veff i r i .i r 2m r Veff r V r r i r r r dr XC r 2 i FLAPW basis Periodic-slab geometry (5-layer Cu + 8-layer vacuum) Density Functional Theory Generalized Gradiant Approximation (GGA) PBE96: Perdew et al., PRL 77, 3865 (1996) Structure Optimization © 2006 IBM Corporation IBM Research Thin Insulator: CuN Islands on Cu(100) d0=2.55Å a0=3.60Å CuN a0=2d0 1nm N d0 Cu Mn Mn Mn Mn Cu(100) Mn Mn CuN monolayer Atomic resolution on CuN Mn atoms bind to Cu and N sites Cu(100) © 2006 IBM Corporation IBM Research DFT Calculation of Electron Density in CuN 0.25Å Cu+0.5 N-1 Cu+0.5 N-1 Cu+0.5 1.80Å Cu Cu N atoms are approximately coplanar with Cu atoms on CuN surface. © 2006 IBM Corporation IBM Research Manipulation of Mn on Cu(100) / CuN Pick up Atom Move tip in Apply 2.0V Pull tip back © 2006 IBM Corporation IBM Research Manipulation of Mn on Cu(100) / CuN Move tip in Apply -0.5V Pull tip back Pick up Atom Drop off © 2006 IBM Corporation IBM Research Spectroscopy of Mn Dimers N Cu Mn Mn Large step at ~6mV splits into three distinct steps at high fields 2.0 dI/dV (a.u.) B=7T 1.5 B=4T 1.0 B=0T 0.5 0.0 -10 -5 0 5 10 Voltage (mV) © 2006 IBM Corporation IBM Research Coupled Spins 5 4 … 1 0 S=5/2 S=5/2 ST = For ST=0 (singlet) the first excited state is ST=1 (triplet) E |ST,m> |1,+1> Three excitations around constant energy shift |1,0> |1,-1> |0,0> B © 2006 IBM Corporation IBM Research Chains of Mn Atoms Cu(100) 2 1nm 6 3 7 4 8 5 9 10Mn 1Mn CuN 1nm N Cu Mn Mn Mn IBM Almaden STM Lab has built chains of up to 10 Mn atoms on Cu binding sites © 2006 IBM Corporation IBM Research Spectroscopy of Mn Chains 10Mn 10 1nm 9Mn 9 6 8Mn 8 dI/dV [a.u.] 2 7 7Mn 6 6Mn 5 5Mn 4 4Mn 3 3Mn 2Mn 3 7 4 8 5 9 2 1Mn 1 0 -20 -10 0 10 10 20 Voltage [mV] Spectra change dramatically with each additional Mn atom. © 2006 IBM Corporation IBM Research Heisenberg Model of Spin Coupling J S Phenomenological Exchange Coupling J = Coupling strength Si = spin of ith atom H J i, j S i S j i, j Assumptions All spins are the same Nearest-neighbor coupling All J are the same J > 0 (antiferromagnetic coupling) N 1 H J S i S i 1 i 1 © 2006 IBM Corporation IBM Research Heisenberg Dimer Spectrum 6 J 20 S Energy [J] 15 5 5 4 4 4 3 3 3 3 2 2 2 2 2 1 0 1 0 1 0 1 0 1 0 10 5 1 0 0 1/2 1 3/2 2 5/2 SG=0 and SE=1 Atomic spin affects numbers of levels but not spacing First excited state at J 3 Atomic Spin © 2006 IBM Corporation IBM Research Determination of Spin Coupling Strength 2.5 From the dimer spectrum J=6.2meV dI/dV (a.u.) 2Mn 2.0 Variations in J of ±5% for different dimers at various locations 1.5 1.0 -25 -20 -15 -10 -5 0 5 J=6.2meV 10 15 20 25 Voltage (mV) © 2006 IBM Corporation IBM Research Determination of Atomic Spin 4.0 Using J = 6.2meV, we find S=5/2 3.5 dI/dV (a.u.) 3.0 3Mn S=3 2.5 S=5/2 STM determines both J and S! S=2 2Mn 2.0 1.5 1.0 -25 -20 -15 -10 -5 0 5 J=6.2meV 10 15 20 25 Voltage (mV) © 2006 IBM Corporation IBM Research Heisenberg Model for Longer Chains 7 6Mn 6 5 Use J = 6.2meV and S=5/2 5Mn Odd chains dI/dV (a.u.) 4 3 1 0 ground state spin = 5/2 excited state spin = 3/2 3Mn Even chains 2 4Mn 2Mn 1Mn -20 -10 0 10 ground state spin = 0 excited state spin = 1 20 Voltage (mV) © 2006 IBM Corporation IBM Research Unit Cells Used in Calculating Mn on CuN Single Mn, larger unit Mn Single dimer, Mn, smallest smallest unit unitcell cell cell N Cu Mn Mn 10.80Å 7.20Å 7.20Å © 2006 IBM Corporation IBM Research Electron Density with an Adsorbed Mn Atom Mn+ Cu+0.5 N -1.5 N -1.5 Cu+0.5 Cu Cu Cu • N atoms move farther out of surface Cu layer towards Mn atom. • Cu atom being pushed into the surface. • This “isolates” the free spin of Mn atom. © 2006 IBM Corporation IBM Research Mn Spin from DFT majority () minority () Free Mn atom 3d 5 © 2006 IBM Corporation S=5/2 IBM Research A new kind of atomic-scale magnet Mn N N Cu Cu Mn N Cu Cu Cu Surface N atoms isolate and bridge Mn atoms. This is a “surface” assembled magnet. © 2006 IBM Corporation IBM Research Control of Spin Coupling Strength 1.0 J=6.2meV dI/dV (a.u.) 0.5 0.0 J=2.7meV 1.0 0.5 0.0 -8 -6 -4 -2 0 2 4 6 8 Voltage (mV) STM can switch J by a factor of 2 by selecting the binding site © 2006 IBM Corporation IBM Research GGA+U GGA+U (strong Coulomb repulsion on Mn 3d) Calculating U by constraint GGA Calculating U • Lock d-orbital into the atomic sphere • Do GGA for Mn d3 d2.5 and d3 d1.5 • U =Δεd of the above two © 2006 IBM Corporation IBM Research Calculating Exchange Coupling H=J S1·S2 N Cu |±|S=5/2, Sz=±5/2 DFT total energies 2S2J= ++|H|++ +- |H| +- = E E © 2006 IBM Corporation IBM Research Calculating Exchange Coupling (in meV) GGA (U=0) GGA + U(calculated) GGA + U(calculated+1ev) STM Mn on Cu site Mn on N site 18.5 -1.8 (ferromagnetic!) 6.50 ±0.05 2.5 5.4 5.1 6.2±0.3 2.7 © 2006 IBM Corporation IBM Research Summary of theoretical work The nontrivial structure of the engineered spins requires DFT to determine. Calculated structure shows a new kind of molecular magnets. GGA+U produces correct S and very accurate J; very helpful for searching a system of desired S and J. © 2006 IBM Corporation IBM Research What’s Next Can we understand IETS processes? matrix elements, selection rules, transition strengths What is the origin of the exchange coupling? superexchange, delocalized electrons Are other interactions possible? vary distances, shapes, types of atoms Can we control anisotropy effects? Find a way to store and transfer spin information: bits and circuits based on atomic spins © 2006 IBM Corporation IBM Research Thanks to Barbara Jones Chris Cyrus Hirjibehedin Lutz Andreas Heinrich © 2006 IBM Corporation