Transport theory and simulations in hybrid structures: the need for a bottom-up approach in new semiconductor spintronic systems JAIRO SINOVA Texas A&M University Institute of Physics ASCR Institute of Physics ASCR Tomas Jungwirth, Vít Novák, Karel Vyborny, et al Hitachi Cambridge Joerg Wünderlich, A. Irvine, et al Hamburg University Jan Jacob, Bodo Krause-Kyora, et al International Symposium High Performance Computing in Nano-Spintronics Hamburg, November 30th, 2011 Research fueled by: Transport theory and simulations in hybrid structures: the need for a bottom-up approach in new spintronic systems I. Introduction: using the dual personality of the electron •Internal coupling of charge and spin: origin and present use II.Spintronic devices and the need to understand them microscopically III. Spin injection Hall effect FET: a new paradigm in exploiting SO coupling •Spin based FET: old and new paradigm in charge-spin transport II. Spin filters in InAs nano-wires III. A bottom to top approach: from mesoscopic to microscopic • Why is it important to approach the problem from the mesoscopic regime V. Theoretical tools of the mesoscopic world: • Landauer-Büttiker (coherent regime) • Non-Equilibrium Green’s Function approach: quantum Boltzmann Eq. • Beyond coherent transport: the basic master equation of the nonequilibrium density matrix VI. Computational Challenges of NEGF approach Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 2 The electron: the key character with dual personalities SPIN 1/2 Makes the electron antisocial: a fermion CHARGE Easy to manipulate: Coulomb interaction quantum mechanics E=p2/2m E→ iħ d/dt p→ -iħ d/dr + special relativity E2/c2=p2+m2c2 (E=mc2 for p=0) & spin = particles/antiparticles Dirac equation “Classical” external manipulation of charge & spin Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 3 Using charge and spin in information technology Using charge to create a field effect transistor: work horse of information processing Vg >0 S gate insulator semiconductor total wf antisymmetric = orbital wf antisymmetric × spin wf symmetric (aligned) D substrate HIGH tunablity of electronic transport properties the key to FET success in processing technology What about the internal communication between charge & spin? (spintronics) Sinova Using spin: Pauli exclusion principle and Coulomb repulsion →ferromagnetism work horse of information storage • Robust (can be as strong as bonding in solids) • Strong coupling to magnetic field (weak fields = anisotropy fields needed only to reorient macroscopic moment) e- International Symposium High Performance Computing in Nano-Spintronics, Hamburg 4 Internal communication between spin and charge:spinorbit coupling interaction (one of the few echoes of relativistic physics in the solid state) e- Classical explanation (in reality it arises from a second order expansion of Dirac equation around the non-relativistic limit) • “Impurity” potential • V(r) Produces an electric field In the rest frame of an electron the electric field generates an Motion of an electron effective magnetic field This gives an effective interaction with the electron’s magnetic moment s p V Beff Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 5 Internal communication between spin and charge:spinorbit coupling interaction (one of the few echoes of relativistic physics in the solid state) e- Classical explanation (in reality it arises from a second order expansion of Dirac equation around the non-relativistic limit) • “Impurity” potential • V(r) Produces an electric field In the rest frame of an electron the electric field generates an Motion of an electron effective magnetic field This gives an effective interaction with the electron’s magnetic moment V s p Beff Consequence #1 Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 6 Internal communication between spin and charge:spinorbit coupling interaction (one of the few echoes of relativistic physics in the solid state) e- Classical explanation (in reality it arises from a second order expansion of Dirac equation around the non-relativistic limit) • “Impurity” potential • V(r) Produces an electric field In the rest frame of an electron the electric field generates an Motion of an electron effective magnetic field This gives an effective interaction with the electron’s magnetic moment V s p Beff Consequence #2 Mott scattering Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 7 How spintronics has impacted your life: Metallic spintronics 1992 - dawn of (metallic) spintronics •Anisotropic magnetoresistance (AMR): In ferromagnets the current is sensitive to the relative direction of magnetization and current direction magnetization e- current Appreciable sensitivity, simple design, cheap BUT only a 2-8 % effect Giant magnetoresistance (GMR) read head - 1997 Fert, Grünberg et al. 1998 e- × Nobel Price 2007 Fert and Grünberg High sensitivity, very large effect 30-100% and are almost on and off states: “1” and “0” & magnetic memory bit Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 8 What next? The need for basic research Industry has been successful in doubling of transistor numbers on a chip approximately every 18 months (Moore’s law). Although expected to continue for several decades several major challenges will need to be faced. Circuit heat generation is one key limiting factor for scaling device speed Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 9 Transport theory and simulations in hybrid structures: the need for a bottom-up approach in new spintronic systems I. Introduction: using the dual personality of the electron •Internal coupling of charge and spin: origin and present use II.Spintronic devices and the need to understand them microscopically III. Spin injection Hall effect FET: a new paradigm in exploiting SO coupling •Spin based FET: old and new paradigm in charge-spin transport II. Spin filters in InAs nano-wires III. A bottom to top approach: from mesoscopic to microscopic • Why is it important to approach the problem from the mesoscopic regime V. Theoretical tools of the mesoscopic world: • Landauer-Büttiker (coherent regime) • Non-Equilibrium Green’s Function approach: quantum Boltzmann Eq. • Beyond coherent transport: the basic master equation of the nonequilibrium density matrix VI. Computational Challenges of NEGF approach Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 10 Towards a realistic spin-based non-magnetic FET device Can we achieve direct spin polarization injection, detection, and manipulation by electrical means in an all paramagnetic semiconductor system? Long standing paradigm: Datta-Das FET Unfortunately it has not worked : •no reliable detection of spin-polarization in a diagonal transport configuration •No long spin-coherence in a Rashba spinorbit coupled system (Dyakonov-Perel mechanism) Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 11 New paradigm using SO coupling: SO not so bad for dephasing Problem: Rashba SO coupling in the Datta-Das SFET is used for manipulation of spin (precession) BUT it dephases the spin too quickly (DP mechanism). 1) Can we use SO coupling to manipulate spin AND increase spin-coherence? Use the persistent spin-Helix state and control of SO coupling strength (Bernevig et al 06, Weber et al 07, Wunderlich et al 09, 10) • Can we detect the spin in a non-destructive way electrically? Use AHE to measure injected current polarization at the nano-scale electrically (Wunderlich, et al 09, 04) Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 12 Spin-dynamics in 2D electron gas with Rashba and Dresselhauss spin-orbit coupling 1) Can we use SO coupling to manipulate spin AND increase spin-coherence? a 2DEG is well described by the effective Hamiltonian: Rashba: from the asymmetry of the confinement in the z-direction ky [010] > 0, = 0 [110] kx [100] _ [110] Sinova Dresselhauss: from the broken inversion symmetry of the material, a bulk property ky [010] = 0, < 0 [110] kx [100] _ [110] International Symposium High Performance Computing in Nano-Spintronics, Hamburg 13 Effects of Rashba and Dresselhaus SO coupling ky [010] > 0, = 0 = - [110] ky [010] [110] kx [100] _ [110] = 0, < 0 kx [100] ky [010] [110] _ [110] kx [100] _ [110] Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 14 Spin-dynamics in 2D systems with Rashba and Dresselhauss SO coupling For the same distance traveled along [1-10], the spin precesses by exactly the same angle. [110] _ [110] _ [110] Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 15 Spin-helix state when α ≠ β In addition we performed MC Boltzmann transport calculations incorporating spin physics For Rashba or Dresselhaus by themselves NO oscillations are present; only and over damped solution exists; i.e. the spin-orbit coupling destroys the phase coherence. There must be TWO competing spin-orbit interactions for the spin to survive!!! Wunderlich, Irvine, Sinova, Jungwirth, et al, Nature Physics 09 Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 16 “Local” magnetization measurement through the Anomalous Hall Effect Spin dependent “force” deflects like-spin particles M⊥ _ __ majority FSO FSO I minority ρH=R0B ┴ +4π RsM┴ AHE is does NOT originate from any internal magnetic field created by M⊥; the field would have to be of the order of 100T!!! V Simple electrical measurement of out of plane magnetization (or spin polarization ~ n↑-n↓) InMnAs Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 17 Cartoon of the mechanisms contributing to AHE Skew scattering A ~σ~1/ni Vimp(r) (Δso>ħ/τ) λ*Vimp(r) (Δso<ħ/τ) Asymmetric scattering due to the spin-orbit coupling of the electron or the impurity. Known as Mott scattering. Intrinsic deflection B independent of impurity density Electrons deflect to the right or to the left as they are accelerated by an electric field ONLY because of the spin-orbit coupling in the periodic potential (electronics structure) E SO coupled quasiparticles Electrons have an “anomalous” velocity perpendicular to the electric field related to their Berry’s phase curvature which is nonzero when they have spin-orbit coupling. Side jump scattering B independent of impurity density Vimp(r) (Δso>ħ/τ) λ*Vimp(r) (Δso<ħ/τ) Electrons deflect first to one side due to the field created by the impurity and deflect back when they leave the impurity since the field is opposite resulting in a side step. They however come out in a different band so this gives rise to an anomalous velocity through scattering rates times side jump. Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 18 AHE contribution to Spin-injection Hall effect in a 2D gas Two types of contributions: i)S.O. from band structure interacting with the field (external and internal) •Bloch electrons interacting with S.O. part of the disorder Type (i) contribution much smaller in the weak SO coupled regime where the SOcoupled bands are not resolved, dominant contribution from type (ii) Crepieux et al PRB 01 Nozier et al J. Phys. 79 Lower bound estimate of skew scatt. contribution Wunderlich, Irvine, Sinova, Jungwirth, et al, Nature Physics 09 Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 19 Spin-injection Hall device measurements SIHE ↔ Anomalous Hall trans. signal VL Local Hall voltage changes sign and magnitude along a channel of 6 μm Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 20 Further experimental tests of the observed SIHE T = 250K Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 21 SiHE transistor Vb Vg VH I x VH Vg Vb x=1 m Spin Hall effect transitor: Wunderlich, Irvine, Sinova, Jungwirth, et al, Science 2010 Sinova + International Symposium High Performance Computing in Nano-Spintronics, Hamburg 22 SiHE transistor AND gate Vg2 [V] +0.1 0 -0.1 0 RH1 [] 12 6 H1 - Sinova - RH2 0 [] 3 6 - H2 International Symposium High Performance Computing in Nano-Spintronics, Hamburg 23 Next SiHE transistor: all electrical spin FET Key features: •Spin injection through thin Fe contacts into GaAs •Detection via SHE and Non-Local Hanle effect •Spin current control through charge current bias K. Olejnik, Wunderlich, et al Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 24 Semiconductor spin filters in InAs nanowires Jan Jacob’s group • Splits an unpolarized current into two oppositely spin-polarized currents A. A. Kiselev and K. W. Kim, Appl. Phys. Lett. 78, 775 (2001); J. Appl. Phys. 94, 4001 (2003). J. I. Ohe, M. Yamamoto, T. Ohtsuki, and J. Nitta, Phys. Rev. B 72, 041308 (2005). M. Yamamoto, T. Ohtsuki, and B. Kramer, Phys. Rev. B 72, 115321 (2005). A. W. Cummings, R. Akis, and D. K. Ferry, Appl. Phys. Lett. 89, 172115 (2006). M. Yamamoto, K. Dittmer, B. Kramer, and T. Ohtsuki, Physica E 32, 462 (2006). Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 25 12 Semiconductor spin filters in InAs nanowires • In-plane magnetic fields • Detect spin-Hall signal in filter outputs • Top-gated spin-filter cascades • Single quantum-point contacts • Top/Backgate-controlled spin-orbit coupling • Ferromagnet-Semiconductor Spin-Valves (together with OSU) Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 26 30 Transport theory and simulations in hybrid structures: the need for a bottom-up approach in new spintronic systems I. Introduction: using the dual personality of the electron •Internal coupling of charge and spin: origin and present use II.Spintronic devices and the need to understand them microscopically III. Spin injection Hall effect FET: a new paradigm in exploiting SO coupling •Spin based FET: old and new paradigm in charge-spin transport II. Spin filters in InAs nano-wires III. A bottom to top approach: from mesoscopic to microscopic • Why is it important to approach the problem from the mesoscopic regime V. Theoretical tools of the mesoscopic world: • Landauer-Büttiker (coherent regime) • Non-Equilibrium Green’s Function approach: quantum Boltzmann Eq. • Beyond coherent transport: the basic master equation of the nonequilibrium density matrix VI. Computational Challenges of NEGF approach Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 27 quantitative modeling of spin-current based devices from bottom up Semiclassical: Boltzmann-Monte-Carlo (2010) ✓ scale ~ 1μ Connection to effective Hamiltonians and boundary conditions of different phenomena Mesoscopic: Non-Equil. Green’s Function (2007) ✓ scale ~ 1nm Microscopic: Ab-initio (tight-binding) Interface effects of spin-injection and bulk spin coherence scale ~ 1Å Sinova Molecular level modeling COMPETING LENGTH SCALES OF THE SAME ORDER International Symposium High Performance Computing in Nano-Spintronics, Hamburg 28 Semiclassical bulk transport theory: Boltzmann “diagonal” transport Electrons are treated as Newtonian pinballs, but having an effective mass and g-factor determined by the band structure and undergo random quantum scattering Dynamics of the occupation number • Very successful for charge transport and micro-device modeling • Key feature: l.h.s. classical – r.h.s. quantum collision term • Main shortcoming: Does not treat systematically inter-band coherence or spin transport • Applicable in the diffusive metallic regime ASSUMES COLLISIONS ARE INSTANTANEOUS AND UNEVENTFUL SinovaFormalism International Symposium High Performance Computing in Nano-Spintronics, Hamburg NEGF 29 Semiclassical approach to bulk Hall transport: breakdown of collision assumption Modified Boltzmann Equation: Coordinate shift (Side jump): Golden Rule: Berry curvature: velocity: current: Sinitsyn et al PRB 06 Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 30 What is what in the modified Boltzmann treatment E a) Intrinsic deflection b) Side jump scattering c) Skew scattering SinovaFormalism International Symposium High Performance Computing in Nano-Spintronics, Hamburg NEGF 31 1. 2. 3. Quantum mesoscopic transport When the mean free path is of the same order as the system size then quantum interference effects appear In the presence of SO coupling bands are mixed: occupation number is not the only thing needed to describe transport Need to explore the non-equilibrium dynamics of the whole density matrix Nikolic et al 04 NEGF formalism is in a sense the quantum Boltzmann Equation: both the collisions AND dynamics between collisions are treated quantum mechanically SinovaFormalism International Symposium High Performance Computing in Nano-Spintronics, Hamburg NEGF 32 WHAT CAN WE ADDRESS FROM THIS MICROSCOPIC (1) What is the effect of the nature of the scattering on the induced spin-currents and spin coherence in general? (2) What is the nature of the spin-current? (3) Can these induced currents lead to strong enough spin-accumulation? (4) Does coherent transport matter? (5) What is the loss of spin-coherence in relation to scattering? (6) Is the effect more readily controlled at the mesoscopic scales? (7) What is the dissipative ratio in spintronic devices analogous to current devices? Advantages of NEGF approach: •No assumptions on spin-currents •Deals with intrinsic and extrinsic on an equal footing •Multi-band treatment comes out naturally •No assumed length scales •Inelastic processes can be incorporated Limitations: system sizes and extrapolation to the bulk regime SinovaFormalism International Symposium High Performance Computing in Nano-Spintronics, Hamburg NEGF 33 Mesoscopic transport via Landauer-Büttiker Formalism (Coherent limit of the NEGF formalism) Uncorrelated electrons injected from left lead. Single channel conductance: T transmission probability Sharvin conductance Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 34 1. Landauer-Büttiker Procedure in a Nutshell (real space) Calculate the leads self-energies (f-numberelt by the sample): Usually known or easily calculable 2. Compute the sample retarded GF: 3. Obtain transmission coefficients: 4. Calculate currents with Büttiker formula: See, e.g., Electronic Transport in Mesos. Systems by Datta Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 35 How to actually compute things: Tight-Binding Approximation Tight-binding lattice Continuous 2DEG Artificial length scale Effective mass Hamiltonian Tight-binding Hamiltonian Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 36 Example: Quantum Point Contact Conductance Experimental evidence of conductance quantization: Conductance of a ballistic 2DEG QPC: Van Wees et al, PRL 60, 848 (1988) Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 37 Limitations of Landauer-Büttiker •Limited to relatively small systems. •Must rely on finite size scaling to reach bulk values •Limited to coherent transport NEGF formalism advantages •Expresses physical quantities via correlation functions. •Studies non-coherent transport. •Can include the effect of interactions and dissipation gradually •Can make clearer connection to the bulk transport theories However, it still has same size limitation as LB Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 38 Nonequilibrium Green Function Formalism Classical transport: distribution function Quantum transport: density matrix Key correlation function Generalizes the density matrix. Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 39 Boltzmann vs. NEGF Functions (but not literally; r.h.s. knows about interband coherence) Electron distribution function Hole distribution function Scattering functions Do not take into account phase correlations. Sinova Electron correlation function Hole correlation function Scattering functions Include phase correlations. International Symposium High Performance Computing in Nano-Spintronics, Hamburg 40 Connection to semiclassical Boltzmann when looking in real space Using the relation shown and G in real space SinovaFormalism International Symposium High Performance Computing in Nano-Spintronics, Hamburg NEGF 41 NEGF Algorithm - Interactions 1. Compute self-energies only first iteration 2. Calculate the retarded GF 3. Find the correlation function 4. Calculate interaction self-energy 5. Check for convergence If not 6. Compute physical quantities Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 42 NEGF Algorithm – Molecular Transport The sample Hamiltonian depends on charge density: 4. Compute the charge density for biased sample: 1. Calculate self-energies of the electrodes: 5. Check for convergence: 2. Calculate the retarded GF: no 6. Compute physical quantities: 3. Compute the correlation function: Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 43 Nonequilibrium Spin Hall Accumulation in mesoscopic Rashba 2DEG: non-linear transport eV=0 +eV/2 -eV/2 Spin density (NEGF): PRL 95, 046601 (2005) SinovaFormalism International Symposium High Performance Computing in Nano-Spintronics, Hamburg NEGF 44 Transport theory and simulations in hybrid structures: the need for a bottom-up approach in new spintronic systems I. Introduction: using the dual personality of the electron •Internal coupling of charge and spin: origin and present use II.Spintronic devices and the need to understand them microscopically III. Spin injection Hall effect FET: a new paradigm in exploiting SO coupling •Spin based FET: old and new paradigm in charge-spin transport II. Spin filters in InAs nano-wires III. A bottom to top approach: from mesoscopic to microscopic • Why is it important to approach the problem from the mesoscopic regime V. Theoretical tools of the mesoscopic world: • Landauer-Büttiker (coherent regime) • Non-Equilibrium Green’s Function approach: quantum Boltzmann Eq. • Beyond coherent transport: the basic master equation of the nonequilibrium density matrix VI. Computational Challenges of NEGF approach Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 45 Usuki-transfer matrix method with GPUs (two terminal) Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 46 Key computational challenges The key issue with NEGF for realistic size microscopic calculations is how to best deal with large matrices (both multiplication and inversion) •In order to obtain bulk realistic values one needs to do finite size scaling of more than three small sizes •Can one dynamically deal with sparsity? •Not all eginevectors are created equal - how to select? •2-terminal vs. multi-terminal? •Conductance (two terminal) vs. G< (much more computationally expensive but contains everything) Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 47 Principal Outside Collaborators Jan Jacob Tomas Jungwirth Joerg Wunderlich Allan MacDonald U. Hamburg Texas A&M U. Cambridge-Hitachi U of Texas Inst. of Phys. ASCR U. of Nottingham Sinova Laurens Molenkamp Bryan Gallagher U. of Nottingham Würzburg Gerrit Bauer TU Delft and many others International Symposium High Performance Computing in Nano-Spintronics, Hamburg 48 Sinova’s group Xin Liu Texas A&M U. Vivek Amin Texas A&M U. Erin Vehstedt Texas A&M U. H. Gao Texas A&M U. Jacob Gyles Texas A&M U. Oleg Tretiakov (main PI Abanov) Texas A&M U. Previous members Liviu Zarbo Alexey Kovalev Ewelina Hankiewicz Nikolai Sinitsyn Texas A&M U. (PD-Texas A&M Univ.) PD-Texas A&M Univ. PD-Texas A&M Now: Now: Now: W2-Professor at Now: PD- ASCR PD-UC Riverside Würzburg University Staff LANL 49 International Symposium High Performance Computing in Nano-Spintronics, Hamburg Xiong-Jun Liu Texas A&M U. Now: PD- U. Maryland Sinova Mario Borunda Texas A&M Univ. Now: PD-Harvard Univ. What next? Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 50 Inversion of large sparse matrices Sinova International Symposium High Performance Computing in Nano-Spintronics, Hamburg 51