Spintronic and electronic transport properties in graphene – The cornerstone for spin logic devices. 皮克宇* Department of Physics and Astronomy UC Riverside 4月26日, 2011 NTNU *Current location: Hitachi Global Storage Technologies Outline I. Introduction. II. Gate tunable spin transport in signal layer graphene at room temperature. III. Enhanced spin injection efficiency: Tunnel barrier study. IV. Spin relaxation mechanism in graphene: --- Charged impurities scattering. --- Chemical doping on graphene spin valves. Motivation for Spintronics Silicon electronics and the “end-of-the-roadmap”…. How to improve computers beyond the physics limits of existing technology? Spintronics: Utilize electron spin in addition to charge for information storage and processing. Spins for digital information OR Spin up “1” Spin down “0” Technological Approach Storage: Magnetic Hard Drives and Magnetic RAM use metal-based spintronics technologies. Ferromagnetic Materials: • Non-volatile • Radiation hard • Fast switching Logic: Silicon-based electronics are the dominant technology for microprocessors. Semiconducting Materials: • Tunable carrier concentration • Bipolar (electrons & holes) • Large on-off ratios for switches Spintronics may enable the integration of storage and logic for new, more powerful computing architectures. Hanan Dery et al., arXiv 1101.1497 (2011). Material Good electrical properties and potential good spintronic properties. Carbon Family (Z=6) ~ One of the candidates for the cornerstone of this bridge. 1D Carbon Nanotube K. Tsukagoshi, B. W. Alphenaar, and H. Ago, Nature 401, 572 (1999). 2D Graphene Discover in 2004 !! K. S. Novoselov et al., Science 306, 666 (2004). 3D Graphite M. Nishioka, and A. M. Goldman, Appl. Phys. Lett. 90, 252505 (2007). Properties of Graphene Electronic Band Structure Physical Structure Atomic sheet of carbon High mobility -- up to 200,000 cm2/Vs (typically 1,000 – 10,000 cm2/Vs). Zero gap semiconductor with linear dispersion: “massless Dirac fermions”. Tunable hole/electron carrier density by gate voltage. Possible for large scale device fabrication. Low intrinsic spin-orbit coupling C. Berger et al., Science 312, 1191 (2006). K. S. Kim et al., Nature 457, 706 (2009). Possibility for long spin lifetime at RT Graphene Spin transport 1. E. W. Hill et al., IEEE Trans. Magn. 42, 2694 (2006). (Prof. Geim’s group at Manchester ) 2. M. Ohishi et al., Jpn. J. Appl. Phys 46, L605 (2007). (Prof. Suzuki’s group at Osaka) 3. S. Cho et al., Appl. Phys. Lett. 91, 123105 (2007). (Prof. Fuhrer’s group at Maryland) 4. M. Nishioka, and A. M. Goldman, Appl. Phys. Lett. 90, 252505 (2007). (Prof. Goldman’s group at Minnesota) 5. N. Tombros et al., Nature, 571 (2007). (Prof. van Wees’ group at University of Groningen) 6. W. H. Wang et al., Phys. Rev. B (Rapid Comm.) 77, 020402 (2008). (Prof. Kawakami’s group at Riverside) Figure 2 in ref. 5. Figure 3 in ref. 5. Figure 4 in ref. 5. •Demonstrated the first gate tunable spin transport in graphene spin valve at room temperature. Observed Local and nonlocal magnetoresistance. Gate dependent non-local magnetoresistance. Hanle spin precession. Hybrid Spintronic Devices Spin Injector Spin Detector Lateral Spin Valve Ferromagnetic Electrodes _ 0 + Spin Transport Layer Desired Characteristics Graphene (beginning in 2007) Room temperature operation Yes High spin injection efficiency Yes (With tunnel barrier) Gate-tunable spin transport Spin transport over long distances Long spin lifetimes Allows spin manipulation Yes OK, 5 microns. Small graphene flakes. Theory: yes, Experiment: no Good potential Outline I. Introduction. II. Gate tunable spin transport in signal layer graphene at room temperature. III. Enhanced spin injection efficiency: Tunnel barrier study. IV. Spin relaxation mechanism in graphene: --- Charged impurities scattering. --- Chemical doping on graphene spin valves. Sample preparation Raman Identify single layer graphene with optical microscope and confirm with Raman spectrum. Sample preparation Co (7°) Co MgO (0°) MgO 2nm SLG SiO2 SLG Si SiO2 Optical Back Gate SEM SLG SLG Co Standard ebeam lithography 500 nm Device characterization I Contact resistance V 1.5 dV/dI (kΩ) Vg = 0 V E1 E2 E3 E4 R3pt 1.0 0 I R4pt V R3pt – R4pt 0.5 -200 0 I (μA) 200 E1 E2 E3 E4 Relectrode + Rcontact < 300 ohms Conductance (mG) V E1 E2 E3 E4 Co MgO SLG 1.5 Gate dependent resistance I Transparent contact of Co/SLG m ~ 2500 cm2/Vs 1.0 0.5 0.0 -60 -40 -20 0 20 Gate Voltage (V) 40 Spin Injection and Chemical Potential graphene FM e- Chemical Potential (Fermi level) m m m Density of states Density of states Spin-dependent Chemical potential Local and Nonlocal Magnetoresistance Local spin transport measurement: I charge current V Spin Injector Spin Detector spin current Non-local spin transport measurement: Spin Injector charge current Spin Detector + IINJ spin current - VNL Using lock-in detection M. Johnson, and R. H. Silsbee, PRL, 55, 1790 (1985) Nonlocal Magnetoresistance IINJ Parallel IINJ VNL Anti-Parallel VNL H H L L m Detector s Injector Spin up Vp>0 m Spin down Spin dependent chemical potential Spin dependent chemical potential Injector m Detector s Spin up VAP<0 m Nonlocal MR = (VP - VAP)/IINJ Spin down Nonlocal MR--- Temperature dependent Spin Signal Nonlocal MR = ΔRNL = ΔVNL/Iinj 80 200 0 ΔRNL -40 RNL (m) RNL (m) 40 100 RT -80 -100 0 H (mT) 100 0 0 100 200 Temperature(K) Room temperature spin transport 300 Nonlocal MR—Spacing dependence E7 ΔR (m) E6 E4 E2 SLG E3 E1 E5 1 um L (mm) Wei Han, K. Pi et al., APL. 94, 222109 (2009) 3μm L = 2 μm H (mT) L = 3 μm RNL (mΩ) RNl (mΩ) RNL (mΩ) 2μm L = 1 μm 1 λS ~1.6 μm H (mT) H (mT) Graphene spin valve Non-local signal (m) 40 spin injection efficiency is low. P~ 1%. 20 0 Non-local signal (m) 40 20 0 -20 -20 -40 -40 -600 -300 0 300 H (Oe) 600 -600 -300 0 300 H (Oe) 600 -600 -300 0 300 H (Oe) Non-local signal (m) 40 20 0 -20 -40 Gate tunable non-local spin signal 600 Hanle spin precession – spin lifetime measurement IINJ RNL (mΩ) H L = 3 μm 1.0 0.5 0 VNL L Diffusion coefficient -0.5 -1.0 Spin Lifetime -160 -80 0 H (mT) RNL 0 80 160 spin lifetime is “short”. L2 exp cos(Lt ) exp(t / s )dt 4 Dt 4 Dt 1 D = 0.025 m2/s s = 84 ps λs = 1.5 μm Challenges • Create spin polarized current in graphene. How to increase the spin injection efficiency? • Keep spin current polarized in graphene. What is the spin relaxation mechanism in graphene? Outline I. Introduction. II. Gate tunable spin transport in signal layer graphene at room temperature. III. Enhanced spin injection efficiency: Tunnel barrier study. IV. Spin relaxation mechanism in graphene: --- Charged impurities scattering. --- Chemical doping on graphene spin valves. Theoretical analysis How to achieve efficient spin injection? Ri Ri RF RF P 2 2 F 2 2 R RG RG RG L / G 2 L / G 1 G 4 RG e ( ) [ (1 ) e ] 2 2 2 2 1 PF 1 PJ 1 PF i 1 1 PJ i 1 PJ RNL Takahashi, et al, PRB 67, 052409 (2003) Co Tunneling contacts MgO SLG Insert a thin tunnel barrier to make R1, R2 >> RG How to fabricate pin-hole free tunnel barrier. RNL(Ω) 120 L=λG=W=2 μm PF=0.5, PJ=0.4 ρG=2 kΩ 60 0 Transparent contacts 0 20000 40000 Interface resistance (R1, R2 )(Ω) MgO Barrier with Ti adhesion layer 1 nm MgO on graphite (AFM) MgO Ti No Ti graphite RMS roughness: 0.766nm RMS roughness: 0.229nm W. H. Wang, W. Han et. al. ,Appl. Phys. Lett. 93, 183107 (2008). Tunneling spin injection into SLG Fabrication and Electrical characterization Co Ti/MgO(7°)Ti/MgO (9°) (0°) I Co TiO2 MgO I SLG SiO2 2-probe 3-probe 4 dV/dI (k) IDC (μA) SLG SiO2 200 8 0 -4 -8 V + - 300 K 150 100 50 300 K -0.6 -0.4 0 VDC(V) 0.3 0.6 0 -10 0 IDC (mA) 10 Tunneling spin injection into SLG Large Non-local MR with high spin injection efficiency PJ 2G L / N RNL e W G Johnson & Silsbee, PRL, 1985. Jedema, et al, Nature, 2002 . (0V ) 0.35mS , RNL (0V ) 130.4, W ~ 2.2 m m, L 2.1m m, G 2 m m, RNL=130 , PJ=31 % Wei Han, K. Pi et. al., PRL 105, 167202 (2010). Comparison of Co/SLG and Co/MgO/SLG Co Co MgO 2nm MgO 1nm SLG SLG SiO2 SiO2 L=1 mm 15 10 5 0 -5 -10 -15 -20 100 Non-local signal () Non-local signal (m) 20 Vg=0 V -600 -300 300 0 H (Oe) RNL= 0.02 3nm 600 P ~ 1% L=2.1 mm 50 0 -50 -100 Vg=0 V -800 -400 400 0 H (Oe) 800 RNL=130 P ~ 31% Tunnel barrier increases spin signal by factor of ~1,000 Theoretical analysis For Ohmic spin injection with Co/SLG Ri Ri RF RF P 2 2 F 2 2 RG RG RG RG L / N 2 L / G 1 4 RG e ( ) [ (1 ) e ] 2 2 2 2 1 PF 1 PJ 1 PF i 1 1 PJ i 1 PJ RNL RNL 4 pF 2 RF 2 e L / G 4 pF 2 RF 2 e L / G 1 R ( ) [ ] ~ G G 2 L / G 2 L / G 2 2 2 2 (1 pF ) RG 1 e (1 pF ) 1 e RG For Tunneling spin injection with Co/MgO/SLG Ri Ri RF RF P 2 2 F 2 2 R RG RG RG L / G 2 L / G 1 G 4 RG e ( ) [ (1 ) e ] 2 2 2 2 1 PF 1 PJ 1 PF i 1 1 PJ i 1 PJ RNL RNL RG G 2 L / G 1 PJ e ~ W G Gate Tuning of Spin Signal Drift-Diffusion Theory for Different Types of Contacts Proportional to graphene conductivity Inversely proportional to graphene conductivity Gate Tuning of Spin Signal Transparent contact Pin-hole contact Gate Tuning of Spin Signal Tunneling contact Characteristic gate dependence of tunneling spin injection is realized. Outline I. Introduction. II. Gate tunable spin transport in signal layer graphene at room temperature. III. Enhanced spin injection efficiency: Tunnel barrier study. IV. Spin relaxation mechanism in graphene: --- Charged impurities scattering. --- Chemical doping on graphene spin valves. Spin relaxation in graphene Experiment: Theory: Spin lifetime ~ 500 ps Spin lifetime ~ 100 ns – 1 ms (for single layer graphene) Two types of spin relaxation mechanisms: Elliot-Yafet mechanism D’yakonov-Perel mechanism defects Spin flip during momentum scattering events. Charged impurities (Coulomb) are the most important type of momentum scattering. spins precess in internal spin-orbit fields. Are charged impurities important for spin relaxation? C. Jozsa, et al., Phys. Rev. B, 80, 241403(R) (2009). N. Tombros, et al., Phys. Rev. Lett. 101, 046601 (2008). Experiment MBE cell I Co electrode + V Single-Layer Graphene (SLG) SiO2 Si (backgate) Charged impurities (we use Au in this study) Graphene spin valve device We add charged impurities onto a graphene spin valve to study its effect on spin lifetime. K. Pi, Wei Han et.al., Phys. Rev. Lett. 104, 187201 (2010). Challenges How to perform the experiment???? • With small amounts of adatom coverage, metal impurties will oxidize. • Clean environment and fine control of deposition rate. In-situ Measurement. Molecular beam epitaxy Growth. The UHV System Small MBE Chamber •Measure Transport Properties •Vary Temperature from 18K to 300K •Ports for 4 different materials •Apply a magnetic field SLG 500 nm Magnet SEM image In situ measurement Au is selected for this study because Au behaves as a point-like charged impurity on graphene. T=18 K m (cm2/Vs) Conductivity (mS) Gate dependent conductivity vs. Au deposition time Au 2 s Au 8 s Au 6 s Au 4 s No Au Au deposition (Sec) Gate Voltage (V) Deposition rate ~ 0.04 Å/min (5x1011 atom/cm2s) Coulomb scattering is the dominant charge scattering mechanism. K. M. McCreary, K. Pi et al., Phys. Rev. B 81, 115453 (2010). Without introducing extra spin scattering. Simulation Conductivity (mS) Effect of Au doping on non-local signal Introducing extra spin scattering. Au 2 s Au 8 s Au 6 s Au 4 s No Au Simulation Rnl () Rnl () Gate Voltage (V) Gate (V) Au doping does not introduce extra spin scattering. Gate (V) Hanle precession Directly compare spin lifetime between different amounts of Au doping. data fit ΔRNL (Ω) ΔRNL (Ω) data fit Au = 8 s Holes -0.01 0 0.01 -0.01 H(T) data fit Au = 0 s Holes -0.01 0 0 0.01 H(T) ΔRNL (Ω) ΔRNL (Ω) Au = 8 s Electrons Au = 0 s Electrons -0.01 0.01 data fit 0 H(T) H(T) data fit Au = 8 s Dirac Pt. -0.01 0 H(T) data fit ΔRNL (Ω) ΔRNL (Ω) 0.01 Au = 0 s Dirac Pt. -0.01 0 H(T) 0.01 0.01 Effect of charged impurities on spin lifetime Spin lifetime and the diffusion coefficient are determined from Hanle spin precession data Spin relaxation Momentum scattering Spin lifetime (ps) D (m2/s) 0.06 Dirac Pt. Electrons Holes 0.04 0.02 0.00 0 2 4 6 8 (2.9x1012 cm-2) Au deposition (sec) Au deposition (s) Charged impurities are not the dominant spin relaxation mechanism. Slight enhancement of spin lifetime • Spin relaxation mechanisms are correlated. 1/ s 1/ C 1/ j j c : Spin relaxation by Coulomb scattering. j : Spin relaxation by other defects (lattice defects, sp3 bound etc.). Recent study shows that Co contact plays an important role. Y. Gan et al., Small 4, 587 (2008). S. Molola et al., Appl. Phys. Lett. 94, 043106 (2009). Wei Han et al., arXiv 1012.3435 (2011). • Effect of D’yakonov-Perel mechanism. E-Y mechanism: s ~ m D-P mechanism: s ~ m-1 F. Guinea et al., Solid State comm. 149, 1140 (2009). Further study is needed. Enhancement of spin signal by chemical doping • At fixed gate voltage, Au doping can enhance conductivity. •No significant spin relaxation from charged impurities. Conductivity (mS) By Au doping we are able to enhance spin life time from 50 ps to 150 ps. 2.0 1.5 1.0 0.5 0.0 Possible to tune spin properties by chemical doping instead of applying high electric field (gate voltage). Conclusion Spin lifetime (ps) Achieved tunneling contact on graphene spin valves. Demonstrated charged impurities are not the dominant spin relaxation mechanism. Au deposition (s) Manipulation of spin transport in graphene by surface chemical doping. Acknowledgements Roland Kawakami Collaborators Wei Han Kathy McCreary Postdoc: Wei-Hua Wang (Academia Sinica in Taiwan) Yan Li Adrian Swartz Jared Wong Richard Chiang Wenzhong Bao Feng Miao Jeanie Lau (PI) Peng Wei Jing Shi (PI) Shan-Wen Tsai (PI) Francisco Guinea (PI) Mikhail Katsnelson (PI) Thank you. New physics in TM doped graphene system • Adatoms on Graphene; Wave function hybridization between TM and graphene may lead us to the new physics. --- Fe on graphene is predicted to result in 100% spin polarization. Y. Mao et al., Journal of Physics: Condensed Matter 20, 2008 (2008). --- Pt may induce localized magnetic states in Graphene. B. Uchoa et al., Phys. Rev. Lett. 101, 026805 (2008). • Hydrogen storage. --- AI doped graphene as hydrogen storage at room temperature. Z. M. Ao et al., J. Appl. Phys. 105, 074307 (2009). The UHV System We use same system to study the charge transfer and charge scattering mechanism of transition metals doped graphene. 5 mm Magnet SEM image Dirac point shift vs. Ti and Fe coverage Øgraphene = 4.5 eV No Ti (0 ML) Conductivity (mS) Conductivity (mS) ØTi = 4.3 eV 0.0038 ML 0.0077 ML 0.015 ML ØFe = 4.7 eV No Fe (0 ML) 0.041 ML 0.123 ML 0.205 ML Gate Voltage (V) Dirac Point (V) Dirac Point (V) Gate Voltage (V) 0 -40 -80 0.00 0.01 0.02 0 -30 -60 0.0 Ti coverage (ML) 0.1 0.2 Fe coverage (ML) Both Ti and Fe coverage show n-type doping Keyu Pi et al., PRB 80, 075406 (2009). Dirac point shift vs. Pt coverage No Pt (0 ML) 0.025 ML Dirac point shift (V) Conductivity (mS) ØPt = 5.9 eV 0.071 ML 0.127 ML Dirac Point (V) Gate Voltage (V) 0 -20 Pt-1 Pt-2 Fe-1 Fe-2 Fe-3 Ti-1 Ti-2 Ti-3 -40 TM coverage (ML) 0.00 0.05 0.10 0.15 Pt coverage (ML) Regardless of the metal work function, all TMs we have studied result in n-type doping when making contact with graphene. • The trend of Dirac point shift follows the work function. • All the Pt and Fe samples show the ntype doping behavior. Interfacial dipole Become n-type doping V(d) = tr(d) + c(d) V V V WM WG WG d d W W W WG E EFF tr(d) : The charge transfer between graphene and the metal (difference in work functions). EF Graphene Graphene Graphene EF Metal -q +q +q+q c(d) : the overlap of the metal and graphene wave functions c(d) = e−gd (a0 + a1d + a2d2) Highly depends on d. G. Giovannetti et al., Physical Review Letters 101, 026803 (2008). Possible reason for anomalous n-type doping p-type Graphene d n-type Transition metal --- An interfacial dipole having 0.9eV extra barrier for an equilibrium distance ~ 3.3 Å makes the required work function for p-type doping > 5.4eV. ( This explains why Fe with ØFe = 4.7 eV dopes n-type). --- Nano-clusters (smaller than ~ 3nm) have different work function values when compared with bulk material. G. Giovannetti et al., Physical Review Letters 101, 026803 (2008). M. A. Pushkin et al, Bulletin of the Russian Academy of Science: Physics 72, 878 (2008). Interfacial dipole Pt Coverage (Å) Dirac Point (V) 0 2 4 6 8 By Theoretical calculation, d increase as material coverage went from adatoms to continuous film. d d d AFM 2 Graphene AFM 1 0.62 ML 3.19 ML 10 nm 0 nm 0 0.87 1.75 2.62 3.50 AFM 1 AFM 2 Pt Coverage (ML) Experimental evidence of interfacial dipole. K. T. Chan, J. B. Neaton, and M. L. Cohen, Phys. Rev. B 77, 235430 2008. Scattering introduced by TM • Long range scattering. (Charge impurity) • Short-range scattering. (Point defect, wave function hybridization etc.) • Surface corrugations. (Ripple) F. Schedin, A. K. Geim, S. V. Morozov, E. W. Hill, P. Blake, M. I. Katsnelson, and K. S. Novoselov, Nature Mater. 6, 652 (2007). J.-H. Chen, C. Jang, S. Adam, M. S. Fuhrer, E. D. Williams, and M. Ishigami, Nature phys. 4, 377 (2008). Mobility change vs. TM coverage The electron and hole mobilities (μe, μh) are determined by taking a linear fit of the σ vs. n curve just away from the Dirac point (μe,h= |Δσ/Δne| ) 3 2.0 2 1.0 1 Ti-1 0 3 1.0 2 0.5 1 Conductivity (mS) 1.5 0.0 Fe-2 0 3 2.0 2 1.0 0.0 1 Pt-2 -4 -2 n 0 (1012 2 cm-2) 4 0.000 0.015 0 0.030 Coverage (ML) Mobility, m (103 cm2/Vs) 0.0 Fe data show strong electron hole asymmetry. Dirac point shift with TM coverage: Ti >Fe >Pt Mobility drop with TM coverage: Ti >Fe >Pt ? Dirac point shift vs. Mobility change Mobility change vs. Dirac point shift Normalized mobility, μ/μ0 Fitting equation: 0.1 ML μ/μ0 = (Γ0 + ΓTM)-1/Γ0-1 = (1 + ΓTM/Γ0)-1 ΓTM/Γ0 = (AVD,shift)β Pt-1 Pt-2 Ti-1 Ti-2 0.008 ML Ti and Pt fall on the universal curve. Coulomb scattering is the dominant effect. Dirac Point Shift (V) Fe-2 Electron Hole • Electron data follows the universal curve. μ/μ0 • Hole data is significantly different. • This implies some wave function hybridization in the Fe system. Dirac Point Shift (V) Keyu Pi, K. M. McCreary et al., PRB 80, 075406 (2009).