Session V36: Spin Transport in Metals including GMR

Spin-flipping with the antiferromagnets (AF) FeMn and IrMn

We investigate spin-flipping within the antiferromagnets AF = FeMn and IrMn and at their interfaces with Cu. To do so, we measure the Current-Perpendicular-to-Plane (CPP) magnetoresistance (MR) of sputtered Permalloy (Py = Ni(1-x)Fe(x) with x $\sim $ 0.2) based exchange-biased spin-valves (EBSVs) with an AF layer insert of variable thickness in the middle of the central 20 nm thick Cu layer. New measurements with thin inserts of IrMn are consistent with prior ones for thin FeMn inserts [1]---i.e., as little as 1 nm of IrMn placed in the middle of the EBSV causes the MR to drop by a factor of 50. Such a rapid drop corresponds to an effective spin-diffusion-length of only about 1 monolayer (ML), less than the $\sim $ 3 ML intermixing we expect at the AF/Cu interface. As in ref. [1], we attribute this rapid drop to strong spin-flipping at the AF/Cu interface. We also report new measurements extending both IrMn and FeMn thicknesses beyond 2 nm. In each case, the rate of decrease of MR becomes much slower, apparently corresponding to much longer spin-diffusion lengths within the bulk AFs. The work was supported in part by NSF grant DMR 08-04126. \\[4pt] [1] W. Park et al., Phys. Rev. \textbf{B62}, 1178 (2000)   

Dynamic Spin Valve

In this study we investigate a dynamic extension of well known classic spin valves. Ultra thin films were dc sputtered in a UHV chamber and their dynamic responses were measured by ferromagnetic resonance (FMR). Two Co layers, separated by a nonmagnetic Ag layer--thick enough to suppress exchange coupling--were deposited, with one of them coupled to a Gd underlayer, forming a Co(1)/Ag/Co(2)/Gd multilayer. At room temperature both Co(1) and Co(2) FMR's are observed for the external magnetic field in the plane of the film. The field for resonance of Co(2) is reduced significantly relative to Co(1), with the paramagnetic moment of the Gd apparently added to the Co magnetization, and the linewidth is broader. Spin pumping effects are minimal since Co(1) and Co(2) do not resonate at the same field.. The Co(2) FMR disappears at the T$_{C }$of Gd leaving the linewidth of the Co(1) FMR weakly temperature dependent down to the compensation temperature of the Co(2)/Gd bilayer. Below that point, the two Co layers in this dynamic spin valve device are in antiparallel alignment, leading to strong broadening of the Co(1) FMR with decreasing temperature.   

Current-induced dynamics in dual spin valves

Current perpendicular to plane passing through a standard ferromagnet/normal-metal/ferromagnet spin valve gives rise to a spin-transfer torque (STT) exerted on magnetic moments of the system, which is proportional to spin accumulation. In the regime of diffusive transport we investigate how the spin accumulation changes if a normal-metal/ferromagnet bilayer is added to the standard spin valve structure (forming a dual spin valve). It has been shown, that varying the angle between magnetizations of outermost fixed layers one can manipulate with the spin accumulation and tune the STT profile: large STT enhancement as well as wavy-like angular dependence, when STT disappears in certain noncollinear configuration, may be achieved. Employing macrospin simulations we predict the current-induced dynamics in a dual spin valves for both current directions. The possibility of ultra-fast switching as well as out-of-plane self-sustained precessions without a need of external magnetic field have been reported.   

Spin-Flipping at Co/Ni Interfaces

One of the few fundamental questions about electron-spin propagation across metal interfaces involving a ferromagnet (F) that is not yet answered is: `What is the probability (P) of spin-flipping across such interfaces?' We describe a technique for measuring the parameter $\delta $ [P = 1 - exp(-$\delta )$] for interfaces with one or two F-metals, in the current-perpendicular-to-plane (CPP) geometry, and apply it to Co/Ni interfaces. For Co/Ni, the technique involves combining measurements of sputtered simple [Co(3nm)/Ni(3nm)]$_{n}$ multilayers (n = number of repeats) with those on sputtered Permalloy (Py = Ni(1-x)Fe(x) with x $\sim $ 0.2) based double spin-valves with symmetric Py layers, containing [Co(3nm)/Ni(3nm)]$_{n}$ multilayer inserts in the middle of the 20 nm thick central Cu layer. In the spin-valves, the Py layer magnetizations are pinned by adjacent FeMn layers, so that an external magnetic field H can switch the ferromagnetically coupled [Co/Ni]$_{n}$ multilayer from parallel (P) to anti-parallel (AP) to the pinned Py layers. From CPP-MR measurements at 4.2K, we determine the Co/Ni interface resistance and spin-scattering asymmetry, and estimate $\delta $(Co/Ni) $\sim $ 0.3.   

Radiation Effects in GMR Devices

Current information technology relies heavily on magnetic materials via GMR read heads and magnetic random access memory (MRAM). The presumption is that these materials are radiation hard with respect to both photons and particles, potentially indicating utility for nuclear energy and space based applications. However, to date there are few detailed studies of magnetism in GMR devices in radioactive environments. This work explores the effects of gamma ray and neutron irradiation on GMR multilayers. The layer structure used in this experiment is Py/Cu/Py/FeMn/Ge. To study the effects of radiation three probes of magnetization, VSM, MR, and MOKE, are correlated pre and post radiation. We present characterization of the devices for multiple device geometries and doses up to 50Mrad for gamma rays and a minimum fast flux of (E$_{n}>$0.5MeV) of 6.3E12 nv for neutrons, both of which are well above the failure threshold for radiation-hard semiconducting devices.   

Correlation between bias fields and magnetoresistance in CoPt biased NiFe/Ta/NiFe heterosystems

Exchange coupled magnetic hard layer / soft layer (SL) thin films show SL biasing in close analogy to exchange bias systems with antiferromagnetic pinning.$^{1}$ Here we study CoPt(35nm)/NiFe(450nm)/Ta(d)/NiFe(450nm) heterostructures with d between 0.7 and 5nm. We use AGFM and SQUID to measure the overall magnetization reversal and minor loop behavior of the top NiFe layer followed by reversal of the CoPt-pinned NiFe layer. Magnetoresistance (MR) is measured by four-point methodology and modeled using magnetization data thus confirming the assumptions of uniform rotation of the top layer and exchange spring behavior of the pinned NiFe layer. In the absence of vector magnetometry, MR provides evidence for the distinct reversal mechanisms. $^{1}$Ch. Binek, S. Polisetty, Xi He and A. Berger, Phys. Rev. Lett. 96, 067201 (2006). Financial support by NSF through Career, MRSEC and the NRI.   

Magnetic coupling across various spacer layers probed by ferromagnetic resonance

Ferromagnetic resonance (FMR) is used to investigate the dynamical magnetic coupling between magnetic layers formed from an amorphous Co$_{56}$Fe$_{24}$B$_{20}$ alloy and from permalloy (Ni$_{81}$Fe$_{19})$ in sandwich structures. The ferromagnetic layers, each 10 nm thick, are separated by spacer layers formed from various non-magnetic metal and insulating materials including Cu, Pd, Pt, Ru, and MgO. The magnetization and magnetic anisotropy and the Gilbert damping are deduced from the ferromagnetic resonance signals by varying the applied magnetic field and the rf frequency. The magnetic layers are exchange coupled when the spacer layers are sufficiently thin. The critical thickness above which the layers are decoupled strongly depends on the spacer layer material. We show that dynamical exchange coupling due to spin pumping plays an important role.   

Quantum renormalization of the spin Hall effect.

Here we show, by a combined approach of density functional theory and quantum Monte Carlo simulation, that the Fe impurity in Au host metal, a historic Kondo system, shows an important new aspect of Kondo effect, i.e., a surprising role to enhance the relativistic spin-orbit interaction. The spin-orbit interaction, with the bare value of tens of meV, is tremendously enhanced, competing with the hybridization energy of the order of eV, leading to a gigantic spin Hall angle comparable to that observed in recent experiment. More generally we can, for the first time, quantify an essential difference between the anomalous Hall effect and the spin Hall effect.   

Temperature dependence of the non-local resistance and spin diffusion length in metallic lateral spin valves

We report measurements of the $T$ dependence of the non-local spin signal in lateral metallic spin valves with transparent ferromagnet (FM) / normal metal (N) interfaces. We have employed complementary spin valve and Hanle effect measurements to systematically study spin injection and relaxation. Devices were deposited in UHV from high purity sources of all four combinations of $Ni_{0.8}Fe_{0.2}$ or Co FM's and Cu or Al N channels. Devices of width 250 nm, thicknesses 200 - 400 nm, and FM contact separations ($d$ = 250 to 2000 nm) were fabricated on a single substrate using an \textit{in-situ} shadow masking technique. $\Delta R_{NL}$ measurements show a very different $T$ dependence for NiFe and Co devices while showing similar behavior for both Al and Cu. This is generally consistent with the measured resistivities and modeling based on the assumption of transparent interfaces. Fitting $\Delta R_{NL}$ vs. $d$ yields a spin diffusion length with weak $T$ dependence for both Al and Cu and is independent of FM. Hanle effect measurements confirm a consistent picture of spin relaxation. Work supported by the University of Minnesota NSF MRSEC.   

Reduced spin injection at high DC current in lateral spin valves

We report on non-local transport measurements in all metal lateral spin-valves with transparent contacts. We use direct current to investigate the symmetry in the injection process between a ferromagnet (FM) and a normal metal (NM). At high currents (around 1-2$\times $10$^{7}$A/cm$^{2})$ the non-local spin valve (NLSV) signal becomes larger when electrons are injected from the FM into the NM. A systematic study of the NLSV signal for different device lengths as a function of current magnitude and direction reveals the origin of this asymmetry. By fitting the near-exponential decay of the signal with distance the effects of FM polarization and NM spin diffusion length can be separated. These results show that the spin diffusion length of the NM is independent of current direction. However, the effective spin polarization of the FM appears to be larger when electrons are injected from FM into NM and smaller when electrons move in opposite direction. Possible explanations for this behavior will be discussed.   

Temporal evolution of spin diffusion lengths of Cu in nonlocal spin valves

Co/Cu nonlocal spin valves have been fabricated by shadow evaporation. The nonlocal spin signals have been measured every a few days over a 50-day-period after the fabrication of the devices. Unexpectedly, we found over 100{\%} increase of the spin signals. By analyzing the data for multiple devices fabricated on the same chip, we found the room temperature spin diffusion length of the Cu increases from 300 nm to 350 nm. The spin diffusion length at 4.2 K increases from 540 nm to $\sim $ 600 nm. The spin polarization shows a modest increase from 9{\%} to 11{\%} during the first 10 days, and then stays at the same value. We believe the increased spin diffusion length is related to the natural oxidation of the devices. Furthermore, we propose that Co impurities are implanted near the two side surfaces of the Cu wire during the shadow evaporation. These impurities limit the spin diffusion lengths of Cu in the as-fabricated devices. The gradual oxidation of the Cu surfaces will also oxidize the Co impurities and thus reduce the rate of spin-flip scatterings. Work supported by U.S. DOE Grant No. DE-FG02-07ER46374.   

Temperature dependence of current polarization in Ni$_{80}$Fe$_{20}$ by spin wave Doppler measurements

The temperature dependence of current polarization in ferromagnetic metals will be important for operation of spin-torque switched memories and domain wall devices in a wide temperature range. Here, we use the spin wave Doppler technique[1] to measure the temperature dependence of both the magnetization drift velocity $v(T)$ and the current polarization $P(T)$ in Ni$_{80}$Fe$_{20}$. We obtain these values from current-dependent shifts of the spin wave transmission resonance frequency for fixed-wavelength spin waves in current-carrying wires. For current densities of $10^{11}$ A/m$^2$, we obtain $v(T)$ decreasing from $4.8 \pm 0.3$ m/s to $4.1 \pm 0.1$ m/s and $P(T)$ dropping from 0.75$\pm$0.05 to 0.58$\pm$0.02 over a temperature range from 80 K to 340 K. [1] V. Vlaminck et al. Science 322, 410 (2008);   

A new approach to determine bulk spin polarization applied to Co$_{(1-x)}$Fe$_x$S$_2$

We report on a new method to determine the degree of bulk spin polarization in single crystal Co$_{(1-x)}$Fe$_x$S$_2$, by modeling magnetic Compton scattering with {\it ab initio} calculations. Spin-dependent Compton profiles were measured for CoS$_2$ and Co$_{0.9}$Fe$_{0.1}$S$_2$, along four and three different crystallographic directions respectively. The {\it ab initio} calculations were then refined by rigidly shifting the bands to provide the best fit between the calculated and experimental directional profiles for each sample. The bulk spin polarizations, $P$, corresponding to the spin-polarized density of states at the Fermi level, were then extracted from the {\it refined} calculations. The values were found to be $P=-72 \pm 6 \%$ and $P=18 \pm 7\%$ for CoS$_2$ and Co$_{0.9}$Fe$_{0.1}$S$_2$ respectively. Furthermore, determinations of $P$ weighted by the Fermi velocity ($v_F$ or $v_F^2$) were obtained, permitting a rigorous comparison with other experimental data and highlighting the experimental dependence of $P$ on $v_F$.   

Performing spin-polarized STM experiments theoretically

By introducing pathes on the hypersurface of the band energy contribution (free energy) to the magnetic anisotropy energy and of corresponding differences in the zz-like elements of the electric conductivity tensor, viewed as an implicit function of the free energy [1], experimental spin-polarized scanning tunnel microscope data, namely the recording of a differential current as a function of an externally applied field, can be analyzed and facilitate a direct comparison between theory and experiment. It is shown that along different pathes rather different flipping times of the direction of the magnetization can occur. In particular discussed are in terms of ``theoretical experiments'' horizontal and vertical movements of the tip, the influence of the magnetic properties of the tip as well as of typical samples. Furthermore, it is claimed that because of different time scales in a (presently still fictional) time- resolved mode for spin-polarized STM experiments even particular pathes on these hypersurfaces could be mapped out. \\[4pt] [1] P. Weinberger, Phys. Rev. B 80, 060403 R (2009)   

Andreev reflection spin polarization measurements of a weak ferromagnet PdNi

As the magnetic moment and the spin polarization of a ferromagnetic material have distinct origins, the existence of a relationship between these two different physical quantities has always been the topic of intense debate. We have studied a series of weakly ferromagnetic variable composition Pd$_{1-x}$Ni$_{x}$ thin film samples fabricated by sputter deposition. Point contact Andreev reflection (PCAR) spectroscopy~with electrochemically etched Nb tips was used to measure the spin polarization of these~samples. Spin polarization values was obtained by fitting the conductance data with the modified BTK model. To ensure the consistency of the results, the identical set of measurements has also been performed with Sn tips.~ Spin polarization of these samples will be compared to their magnetization (measured by a SQUID magnetometer) and the results will be discussed.