Current research in current-driven magnetization dynamics S. Zhang, University of Missouri-Columbia Beijing, Feb. 14, 2006

Outlines Magentoelectronics started from discovery of giant magnetoresistive (GMR) effect Spin-dependent transport in magnetic metal based nanostructures Spin angular momemtum transfer: physics and potential technology Perspectives and conclusions

M.N. Baibich et al., Phys. Rev. Lett. 61, 2472 (1988). What is giant magnetoresistance? R

Origin of GMR—two current model e e e e EFEF A ferromagnet Different numbers of up and down electrons Up and down resistances Low resistance High resistance

GMR Reading head Bit width Bit length Conductor lead J MSpin valve Spin valve OR M NM M AF “0” “1”

Concert efforts: theorists, experiments and technologists on GMR Theorists: predict, explain, model and design GMR effects and devices Experimentalists: design, fabricate, characterize, and measure GMR devices Technologists: produce, evaluate, pattern, integrate, and deliver GMR devices It would be otherwise impossible to push the information storage so rapidly

History of magentic tapes and hard disks Now: 80Gbits/in2 5 years: 1 Terabits/in2 In 1988, giant Magnetoresistance (GMR) was discovered; in 1996, GMR reading heads were commercialized Since 2000: Virtually all writing heads are GMR heads

GND Magnetoelectronics: Magnetic Tunnel Junctions High tunneling probability Low resistance Low tunneling probability High resistance

Al-O barrier Cu (30) IrMn Co-Fe-B(4) Ta (5) IrMn (12) Al-O (0.8) Cu (20) Ta (5) Py (5) Ta (5) Co-Fe-B(4) V Source: Dr. Xiufeng Han

Brief History of TMJ 1974, M. Julliere (a graduate student) published an experiment article which claimed 14% TMR in Fe/Ge/Fe trilayers. A simple model was proposed (the paper became a sleeping giant). 1982, IBM reported 2% TMR on Ni/AlO/Ni. 1995, Moodera (MIT) and Miyazaki (Japan) reported 10% TMR for Co/AlO/Co. 1998, DARPA launched MRAM solicitation 1999, Motorola’s 128kB MRAM demo 2003, IBM, Motolora, 4Mb MRAM chip demo More than 10 startup MRAM companies formed. MRAM becomes internationally recognized future technology

Emerging non-volatile memory technologies Flow Spin Quantity FRAM PCRAM MRAM PFRAMSiC Bipolar PMC Molecular PolymerPerovskite NanoX’tal 3DROM

Current-driven spin torques GMR/TMR: magnetization states control spin transport (low-high resistance). Adverse effect: spin transport (spin current) affects magnetization states? Every action will have reaction—spin transfer

T spin angular momentum transfer? Momentum transfer—electromigration Angular momentum transfer—magnetization dynamics An impurity atom receives a force by absorbing a net momentum of electrons: electromigration is one of the major failure mechanisms in semiconductor devices. F The atom receives a torque by absorbing a net spin angular momentum of electrons: the spin torque can be used for spintronics

Interaction between spin polarized current and magnetization (J. Slonczewski, IBM) MpMp M Spin torque on the magnetic layer M

Current  torque on DW (Magnetic field  pressure on DW, ) Massless motion!! From Sadamichi Maekawa Current induced Domain wall motion

Magnetization dynamics: LLG equation (micromagnetics) LLG+spin torque Where Spin valve Domain wall

Novelty of spin transfer torques Manipulation of magnetization states by currents Self-sustained steady state magnetization dynamics Unusual thermal effects Interesting domain wall dynamics Dynamic phases: synchronization, modification and chaos

First observation of current induced magnetic switching by Spin torques Co1=2.5nm Co2=6.0nm Katine et. al., PRL (2000).

Self-sustained steady-states precession The first term is always negative (damping), the second term could be positive or negative (it even changes sign at different times). Energy damping and pumping: Limit cycle: the energy change is zero in an orbit

Calculated limit cycles

Kiselev et al., Nature (2003) Experimental identification of limit cycles

Unusual Thermal effects EbEb Neel-Brown relaxation : where is algebraic dependent on T, E Question: in the presence of the spin torque, how do we formulate the relaxation time? Thermal activation Difficulty: the spin torque is not conservative:

LLG equation at finite temperatures random field The magnetization receives following fields Precessional conservative field Non-conservative damping field Non-conservative spin torque field Diffusion field

Solution of Fokker-Planck equation is diffusion constant (dissipation-fluctuation relation) The probability energy density is: where

Experimental data interpretation Telegraph noise H J J H + J R Field alone Current alone H

H-I phase boundary of equal dwell times. Comparison with experiments Equal dwell times for P and AP states By simultaneously changing H and J, one can always have

Synchronization, modification and chaos Limit cycle + 1. Another oscillator 2. AC external field 3. AC external current Linear oscillator

Calculated limit cycles

Observation of synchronization by an AC current Rippard et al, PRL (2005)

Observation of mutual synchronization Kaka et al., Nature (2005); Mancoff et al, Nature (2005)

Observation of mutual synchronization

Narrower spectrum width at synchronization

Dynamic phases due to AC currents M M M M

Synchronization spectra x1

Modification spectra (beating) x2

Synchronization and modification agree well with experiments

Chaos spectra x3

Theories of spin torques in ferromagnets M e Berger, domain drag force, based an intuitive physics picture Bazaliy, et al, Waintal and Viret, a global pressure and a periodic torque Tatara and Kohno, spin transfer torque and momentum transfer torque. Zhang and Li, adiabatic torque and non-adiabatic torques Barnas and Maekawa, non-adiabatic torque relates to the damping of the adiabatic torque within a ballistic transport model for half- metallic materials

Spin torques in a domain wall Equation of motion for conduction electrons where Interaction between conduction electrons and magnetization:

If the wall is in steady motion, the current driven wall velocity is independent wall structure and pinning potentials Steady state wall motion Steady state wall velocity is thus

Wall velocity for different materials in a perfect wire M s (A/m)PWall velocity (m/s) Co14.46x Permalloy8x Fe 2 O x CrO x

Observed Domain wall motion in a nanowire Yamagushi et al., PRL (2004) Observed Wall velocity for

Vortex domain wall motion driven by current Wall transition: from vortex all to transverse wall

Magnetic tunnel Junction 1 0 Goal: using a reasonable current to switch magnetization, ideally less than 10 6 A/cm 2

Conductor lead J Oscillation of M (GHz) by a DC current Application 2: local AC magnetic field oscillators (generators) Local AC field (1000 Oe) with spatial resolution 10nm!

Application IV: concerns of CPP reading heads Bit width Bit length Conductor lead J MSpin valve “0” “1” The large current density in CPP reading heads may produce unwanted switching! Goal: eliminates current-induced switching for current density larger than 10 7 A/cm 2

Acknowledgement Students: Dr. Yu-nong Qi, Mr. Zhao-yang Yang, Mr. Jie-xuan He Postdoctoral: Dr. Z. Li (Postdoctoral) Collaborators: P. M. Levy (NYU) A. Fert (Orsay-Paris) Funded by: NSF-DMR, NSF-ECS, DARPA, NSFC