Session V2: Invited Session: Spin Caloritronics of Magnetic Structures and Devices

Thermal creation of stronger spin-transfer torque in oscillators and memories

Oscillators and magnetic random-access memories (MRAMs) investigated today rely on spin-transfer torque (STT) carried by an electric current flowing through a magnetic tunnel junction (MTJ) having barrier composition MgO.\footnote{See the STT review by D. Ralph and M. Stiles, J. Magn. Magn. Mater. \textbf{320, }1190 (2008).} Experiments confirm the theoretical upper bound \textit{$\tau $}$_{e}$=1/2 on the torque yield (defined as dimensionless torque per unit supplied electric current). This bound limits the performance potential of STT-MRAM in which current supplied by one transistor within each cell switches the information bit. Replacement of electric current with heat flow (supplied by a Joule heater) carried by magnons may provide a greater torque yield \textit{$\tau $}$_{h}$.\footnote{J. Slonczewski, Phys. Rev. B \textbf{82}, 054403 (2010).} The essential structure for this \textit{thermagnonic} spin transfer (TMST) comprises a stack of three nano layers: a spontaneously magnetized insulator (\textit{ferrite} for brevity), a non-magnetic metallic spacer, and the free metallic magnet responding to the transferred torque. Phonons carry most of the heat flowing through the ferrite. But spin-1 magnons also carry a portion of it and deposit pure spin polarization into the spacer whose free electrons transport it to the free magnet. Ferrite-metal interfaces also occur in a spin-Seebeck effect.\footnote{The talk by E. Saitoh in this Symposium} Principles of spin relaxation provide estimates of \textit{$\tau $}$_{h}$ based on existing data for sd-exchange, superexchange, and non-magnetic interfacial thermal resistance; \textit{$\tau $}$_{h}$ may exceed \textit{$\tau $}$_{e }$by one order of magnitude.\footnote{J. Slonczewski, Phys. Rev. B \textbf{82}, 054403 (2010).} Related results of an FMR spin-pumping experiment\footnote{B. Heinrich et al, Phys. Rev. Letts. \textbf{107}, 066604 (2011).} and DFT computations\footnote{The talk by K. Xia in this Symposium.} support the potential of TMST-MRAM. In the case of an oscillator, TMST could increase its efficiency and enable largely independent controls of frequency and output voltage.   

Spin pumping and spin Seebeck effect

Utilization of a spin current, a flow of electrons' spins in a solid, is the key technology in spintronics that will allow the achievement of efficient magnetic memories and computing devices. In this technology, generation and detection of spin currents are necessary. Here, we review inverse spin-Hall effect and spin-current-generation phenomena recently discovered both in metals and insulators: inverse spin-Hall effect, spin pumping, and spin Seebeck effect. (1)Spin pumping and spin torque in a Mott insulator system We found that spin pumping and spin torque effects appear also at an interface between Pt and an insulator YIG.. This means that we can connect a spin current carried by conduction electrons and a spin-wave spin current flowing in insulators. We demonstrate electric signal transmission by using these effects and interconversion of the spin currents [1]. (2) Spin Seebeck effect We have observed, by using the inverse spin-Hall effect [2], spin voltage generation from a heat current in a NiFe, named the spin-Seebeck effect [3]. Surprisingly, spin-Seebeck effect was found to appear even in insulators [4], a situation completely different from conventional charge Seebeck effect. The result implies an important role of elementary excitation in solids beside charge in the spin Seebeck effect. In the talk, we review the recent progress of the research on this effect. This research is collaboration with K. Ando, K. Uchida, Y. Kajiwara, S. Maekawa, G. E. W. Bauer, S. Takahashi, and J. Ieda. \\[4pt] [1] Y. Kajiwara and E. Saitoh et al. Nature 464 (2010) 262. \\[0pt] [2] E. Saitoh et al., Appl. Phys. Lett. 88 (2006) 182509. \\[0pt] [3] K. Uchida and E. Saitoh et al., Nature 455 (2008)778. \\[0pt] [4] K. Uchida and E. Saitoh et al.,Nature materials 9 (2010) 894 - 897.   

Computational spin caloritronics

Recent experimental and theoretical studies focused on spin-mediated heat currents in at interfaces between metals and insulators, where the latter can be either the barrier in a magnetic tunnel junction or a ferromagnetic insulator. A crucial parameter is the efficiency of spin injection and spin-transfer torque. In this talk, we will report realistic electronic structure calculations for two material systems. 1) The pertinent material parameter governing spin transfer and spin Seebeck effect is the spin mixing conductance that we calculate for the Silver-YIG (Yttrium-Iron-Garnett) interface. This turns out to be much larger than expected from the Stoner model. We find mixing conductance comparable to intermetallic interfaces, a surprising result that can be rationalized in terms of magnetic local moments at the interface. These results imply that the spin-mediated energy and information transmissivity of magnetic insulators is potentially much better than has been measured in early experiment, a result that has been experimentally corroborated very recently by several groups. 2). We demonstrate that the thermal spin-transfer torque (TST) in a junction Fe-MgO-Fe tunnel junctions with ultra thin barriers can amount to 10$^{-7}$J/m$^{2}$/K at room temperature, which is estimated to cause magnetization reversal for temperature differences over the barrier of the order of 10 K. The large TST for ultrathin barriers can be explained by multiple scattering due to interface states. Direct evidence for the existence of these states can be obtained by comparing shot noise calculations with recent experiments for high-quality junctions [Arakawa et al., Appl. Phys. Lett. 98, 202103(2011)]. *This work was carried out in collaboration with Xingtao Jia, Kai Liu and Gerrit E.W. Bauer.   

Magneto-Seebeck effect and thermal torques in magnetic tunnel junctions

Creating temperature gradients in magnetic nanostructures has resulted in a new research direction, i.e., the combination of magneto- and thermoelectric effects. Magnetic tunnel devices, known for application as magnetic sensor in hard disc drives or magnetic random access memories (MRAM) show large magnetoresistance. We show that in nanoscale magnetic tunnel junctions, the Seebeck voltage in a heat gradient can be controlled via the magnetization. The Seebeck coefficient changes during the transition from a parallel to an antiparallel magnetic configuration in a tunnel junction -- the magneto-Seebeck effect. In that respect, it is the analog to the tunneling magnetoresistance and thus is called tunneling magneto-Seebeck effect (or tunneling magnetothermopower). The change in Seebeck coefficients is in the order of the voltages known from the charge-Seebeck effect in semiconductors (up to 100 $\mu $V/K). Their size and sign can be delicately controlled by the composition of the electrodes' atomic layers adjacent to the barrier and the temperature and we observe a characteristic sign change from positive to negative magneto-Seebeck effects as theoretically predicted. It is known that generally strong electronic asymmetry at around the Fermi level results in a large Seebeck effect. Here the magnetization dependence of the charge-Seebeck coefficients varying up to $>$100{\%} for the parallel and the antiparallel originates from the half-metallic like transmission of the tunnel junction. Using heating with ultrafast laser pulses, these thermal gradients can be of up to 20 K across the tunnel barrier. We demonstrate that we can achieve the parameters predicted, where by thermal torques magnetization switching is expected. This allows to conceptually think of MRAM's driven by heat gradients only. \\[4pt] [1] M. Walter, et al. Nature Mater. 10, 742 (2011).   

Thermomagnonic spin transfer and Peltier effects in insulating magnets

The recent discovery of the spin Seebeck effect [1] in metals, insulators and semiconductors stimulated development of spincaloritronics [2]. The possibility of measuring the Onsager reciprocal spin Peltier effect has been investigated recently as well. In our theoretical work [3], we study the fictitious electromagnetic fields induced by magnetic textures which may offer an alternative route for observing the spin Peltier effect. Particularly, in an insulating ferromagnet a moving magnetic texture should effectively drive the spin (wave) current which in turn should lead to the heat current by the spin Peltier effect. We further study the coupled magnon energy transport and collective magnetization dynamics in ferromagnets with magnetic textures. We conclude that the analogy between the fictitious electromagnetic fields and real fields should lead to magnonic counterparts of such effects as the Hall effect, the Ettingshausen effect, the Nernst effect, and the Righi-Leduc effect. By constructing a phenomenological theory based on irreversible thermodynamics, we describe motion of domain walls by thermal gradients and generation of heat flows by magnetization dynamics. From microscopic description based on magnon kinetics, we estimate the transport coefficients and analyze the feasibility of energy-related applications (e.g. nanoscale heat pumps [4]) in insulating ferromagnets, such as yttrium iron garnet and europium oxide. Our estimates show that the viscous coupling effects between magnetization dynamics and magnon flows can be strong in materials with low spin densities (e.g. dilute magnetic systems) and narrow domain walls, which can allow the magnonic manipulation of magnetization dynamics and heat pumping.\\[4pt] [1] K. Uchida et al. Nature 455, 778 (2008).\\[0pt] [2] G. E. W. Bauer, A. H. MacDonald, S. Maekawa, Solid State Commun. 150, 459 (2010).\\[0pt] [3] A. A. Kovalev and Y. Tserkovnayk, arXiv:1106.3135.\\[0pt] [4] A. A. Kovalev and Y. Tserkovnyak, Solid State Commun. 150, 500 (2010).