Thermoelectric generation based on spin Seebeck effects

Ken-ichi Uchida,1, 2, ∗ Hiroto Adachi,3, 4 Takashi Kikkawa,1, 5 Akihiro Kirihara,4, 6 Masahiko Ishida,4, 6 Shinichi Yorozu,4, 6 Sadamichi Maekawa,3, 4 and Eiji Saitoh1, 3, 4, 5 1Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan 2PRESTO, Japan Science and Technology Agency, Saitama 332-0012, Japan 3Advanced Science Research Center, Japan Atomic Energy Agency, Tokai 319-1195, Japan 4Spin Quantum Rectification Project, ERATO, Japan Science and Technology Agency, Sendai 980-8577, Japan 5WPI Advanced Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan 6Smart Energy Research Laboratories, NEC Corporation, Tsukuba 305-8501, Japan (Dated: May 24, 2016) The spin Seebeck effect (SSE) refers to the generation of a spin current as a result of a temper- ature gradient in magnetic materials including insulators. The SSE is applicable to thermoelectric generation because the thermally generated spin current can be converted into a charge current via spin-orbit interaction in conductive materials adjacent to the magnets. The insulator-based SSE de- vice exhibits unconventional characteristics potentially useful for thermoelectric applications, such as simple structure, device-design flexibility, and convenient scaling capability. In this article, we review recent studies on the SSE from the viewpoint of thermoelectric applications. Firstly, we in- troduce the thermoelectric generation process and measurement configuration of the SSE, followed by showing fundamental characteristics of the SSE device. Secondly, a theory of the thermoelec- tric conversion efficiency of the SSE device is presented, which clarifies the difference between the SSE and conventional thermoelectric effects and the efficiency limit of the SSE device. Finally, we show preliminary demonstrations of the SSE in various device structures for future thermoelectric applications and discuss prospects of the SSE-based thermoelectric technologies. KEYWORDS — spintronics; spin current; spin Seebeck effect; inverse spin Hall effect; anomalous Nernst effect; thermoelectric generation; magnetic material; thin film

I. INTRODUCTION ture difference owing to the collinear orientation of ESE and ∇T . A thermoelectric module based on the Seebeck effect usually consists of a number of Π-structured ther- Heat is an omnipresent and abundant source of energy, mocouples, i.e., junctions of two materials with different and thus the efficient utilization of thermal energy, such Seebeck coefficients; the thermocouples are serially con- as waste heat recovery and solar heat power generation, is nected to enhance the thermoelectric output [Fig. 1(b)]. indispensable for the realization of future sustainable so- The output voltage of the thermoelectric module can be ciety. Thermoelectric generation is one of the promising enhanced in proportion to the number of the thermocou- technologies for making effective use of heat since it en- ple elements, although such cascaded structure requires ables direct conversion from thermal energy to electrical costly fabrication processes. power [1–3]. As thermoelectric devices consist of solid- The efficiency of the thermoelectric generation based state materials and have no moving parts, they are silent, on the Seebeck effect is characterized by the dimension- reliable, and scalable. Most of the prevalent thermoelec- less figure of merit Z T , defined as tric generation technologies are based on the Seebeck ef- SE fect, discovered by T. J. Seebeck in 1821 [4, 5]. Although α2 Z T = T, (1) a variety of thermoelectric effects were found to appear SE κρ in electric conductors subjected to temperature gradients after the discovery of the Seebeck effect [6], the earliest where κ, ρ, and T are the thermal conductivity, electrical resistivity, and absolute temperature of thermoelectric thermoelectric phenomenon is still the key player in the 2 thermoelectric technology because of its relatively-high materials [1]. The factor α /ρ is known as a thermoelec- output power and versatility. tric power factor. The efficiency of the Seebeck device for electricity generation is defined as η = (energy pro- The Seebeck effect refers to the generation of an elec- vided to a load)/(heat energy absorbed at the hot end arXiv:1604.00477v2 [cond-mat.mtrl-sci] 21 May 2016 tric field ESE as a result of a temperature gradient ∇T of the device), which is a function of internal and load in a conductor, where the direction of ESE is parallel ∗ resistances. The optimized efficiency ηSE of the Seebeck to that of ∇T [Fig. 1(a)]. The thermopower induced device is then given as a function of the figure of merit by the Seebeck effect is represented by the Seebeck co- as follows: efficient: α ≡ ESE/∇T , which is equal to the ratio of p the generated electric voltage to the applied tempera- ∗ ∆T 1 + ZSET − 1 ηSE = , (2) p Tc Th 1 + ZSET + Th

where ∆T = Th − Tc is the temperature difference be- ∗ Electronic address: [email protected] tween the hot and cold ends of the device, Th(c) is the 2

Seebeck effect Longitudinal spin Seebeck effect (LSSE) (e) Developing LSSE devices

(a) (c) metallic film conductor lx coating method flexible device

ESE ly EISHE ∇T ∇T

lz jS z y M x magnetic insulator Fundamental element serial connection enlarging area flexible substrate (b) (d) metallic film all-ferromagnetic device multilayer stack Module structure thermocouple ∇T ∇T magnetic insulator ferromagnetic metal

FIG. 1: (a),(b) Schematic illustrations of the fundamental element and module structure of the thermoelectric device based on the Seebeck effect. (c),(d) Schematic illustrations of the fundamental element and module structure of the thermoelectric device based on the longitudinal spin Seebeck effect (LSSE). ∇T , ESE(ISHE), M, and jS denote the temperature gradient, electric field generated by the Seebeck effect [inverse spin Hall effect (ISHE)], magnetization vector, and spatial direction of the thermally generated spin current, respectively. lx(y) is the length of the metallic film of the LSSE device along the x (y) direction and lz is the thickness of the metallic film. (e) Schematic illustrations of the LSSE-device structures being developed for future thermoelectric applications.

temperature at the hot (cold) end, and T is the average recently [7, 8, 18–22]; since the Wiedemann-Franz law is temperature between the hot and cold ends [4, 5]. Im- violated in anisotropic materials, they are also useful for ∗ portantly, ηSE monotonically increases with increasing optimizing heat and charge conductions separately. ∗ ZSET ; ηSE goes to the Carnot efficiency ηC = ∆T/Th in Nearly 200 years after the discovery of the Seebeck ef- the limit of ZSET → ∞. Therefore, many efforts in ther- fect, a novel thermoelectric generation principle was dis- moelectric research are dedicated to improve the figure covered in the field of spintronics [23–25]. This novel of merit of thermoelectric materials. principle can be said as a spin counterpart of the See- beck effect: “spin Seebeck effect” (SSE) [26]. The SSE Equations (1) and (2) mean that materials having a refers to the generation of a spin current, a flow of spin large Seebeck coefficient, low thermal conductivity, and angular momentum [27, 28], as a result of a temperature low electrical resistivity are necessary for improving the gradient in a magnetic material. The SSE is applicable thermoelectric conversion efficiency. Importantly, low κ to the construction of thermoelectric generators because and ρ values make it possible to suppress the energy loss the spin current generated by the SSE can be converted due to heat conduction and Joule dissipation, respec- into a charge current via the spin-orbit interaction, or the tively. However, in isotropic metals, the Wiedemann- inverse spin Hall effect (ISHE) [29–34], in a conductive Franz law (κeρ = LeT with the electronic Lorenz num- thin film (mostly, a paramagnetic metal film) adjacent to ber Le) limits this improvement when κ is dominated by the magnetic material (see Sec. II for details). Here, the the electronic thermal conductivity κe. A conventional direction of the electric field induced by the ISHE EISHE way to overcome this limitation is to use thermoelectric in the conductive film is perpendicular to that of ∇T in semiconductors, where the thermal conductance is usu- the magnetic material [Fig. 1(c)], a configuration differ- ally dominated by phonons while the electrical conduc- ent from the conventional Seebeck effect. Since the SSE tance is determined by charge carriers, and thus κ and appears not only in ferromagnetic metals [26, 35–38] and ρ are separated according to the kind of the carriers [1]. semiconductors [39, 40] but also in magnetic (mostly, fer- Recently, to improve the figure of merit of thermoelectric rimagnetic) insulators [41, 42], it enables the conversion materials by reducing phonon thermal conductivity with- of heat energy in insulators into electrical energy in ad- out affecting electrical conductivity, not only exploration jacent conductors [43, 44], which was impossible if only of new materials [7, 8] but also nanotechnology-based conventional thermoelectric technologies were used. The phonon engineering [9–17] have been conducted. Fur- thermoelectric technology based on the SSE is still in an thermore, research on the thermoelectric properties of early phase of its development, and the efficiency is very anisotropic materials has attracted increasing attention small at the present stage. However, as reviewed in this 3 article, the SSE exhibits various unconventional features ventional Seebeck device due to the different device con- suitable for thermoelectric applications. figurations and driving principles [130–132]. One of the In this review, we focus on the so-called longitudinal purposes of this review article is to formulate the ther- SSE (LSSE) [42, 44] in metallic film/magnetic insula- moelectric figure of merit and conversion efficiency of the tor junction systems, depicted in Fig. 1(c). After the LSSE device, and to clarify the upper limit of the effi- first report on the SSE [26], this phenomenon has been ciency. measured mainly in two different device structures, i.e., The advantages of the LSSE device include not only the longitudinal and transverse configurations [45], where the device-design flexibility, discussed above, but also the spin currents parallel and perpendicular to ∇T are mea- following characteristics. The LSSE device has conve- sured, respectively. The longitudinal configuration has nient scaling capability, where the thermoelectric output been used in most of the recent SSE studies because the increases simply by extending the device area because insulator-based LSSE systems are useful for the exclusive the total amount of the thermally generated spin cur- detection of spin-current-induced signals and suitable for rents increases as the device becomes larger [43]. Here, thermoelectric applications owing to its simple and ver- the electric field induced by the ISHE at each point of the satile structure. The first observation of the LSSE was LSSE device is integrated into the output LSSE voltage: reported in 2010 by using a junction system compris- VSSE = EISHEly, where EISHE is the magnitude of EISHE ing a paramagnetic metal Pt and a ferrimagnetic insula- and ly is the length of the metallic layer along the y direc- tion [Fig. 1(c)]. Since the internal resistance of the metal- tor Y3Fe5O12 (YIG) [42]. The Pt/YIG junction system is now recognized as a model system for studying SSE lic layer of the LSSE device is Rmetal = ρly/lxlz with lx physics, since Pt and YIG enable efficient spin-charge and lz respectively being the length along the x direc- conversion and pure detection of spin-current effects, re- tion and thickness of the metallic film, the maximum ex- spectively. Since this demonstration, the LSSE has been tractable electrical power Pmax is proportional to the area 2 observed in various combinations of magnetic insulators of the LSSE device: Pmax ∝ VSSE/Rmetal ∝ lxly [see Sec. and conductive films [42–110], as shown in Fig. 2. In con- III C]. Such a straightforward scaling law makes it possi- trast, in the transverse configuration, the SSE measure- ble to enhance the thermoelectric output simply by en- ments may be disturbed by thermal conductivity mis- larging the device area [Fig. 1(d)]. To exploit the above match [111, 112] between a film and a substrate that characteristics of the LSSE device, versatile and low-cost induces parasitic LSSE [113] or anomalous Nernst effect fabrication methods have been developed, such as coat- (ANE) [6, 114–121], requiring careful thermal design of ing [43] and plating [106] technologies [Fig. 1(e)]. These the sample and measurement systems. In the transverse methods enable the implementation of simple-structure, SSE measurements using conductive ferromagnets, a pla- large-area, and even flexible LSSE devices onto various nar Nernst effect (PNE) [6, 122–125] may also arise. In heat sources, as shown in Secs. V A and V B. the past several years, various thermoelectric measure- The thermoelectric generation based on the LSSE in- ments have been performed in the transverse configura- volves the following three factors: tion [111–113, 122–125], and the quantitative separation (1) heat-current/spin-current conversion efficiency in of the transverse SSE from the parasitic ANE and PNE magnetic materials, has been reported in a ferromagnetic metal [38]. In ad- dition to the longitudinal and transverse configurations, (2) spin-angular-momentum transfer efficiency across a non-local geometry has also been used recently to in- metal/insulator interfaces, characterized by the vestigate the length scale of the SSE [126–129]. spin mixing conductance [133–136], The LSSE in insulators allows a new approach to im- (3) spin-current/charge-current conversion efficiency in prove the thermoelectric figure of merit. In the insulator- metallic films, characterized by the spin Hall angle based LSSE device, the heat and charge currents flow in [34, 137]. different parts of the device: κ is the thermal conduc- tivity of the magnetic insulator and ρ is the electrical A direct approach to enhance the performance of the resistivity of the adjacent metal, such that κ and ρ in the spin-current-driven thermoelectric generation is to im- LSSE device are segregated according to the part of the prove the LSSE itself [the factor (1)], which realizes effi- device elements [44]. Therefore, the denominator of the cient thermal spin-current generation [see Sec. V D]. In figure of merit, κρ, of the LSSE device is free from the addition to this approach, the improvement of the spin Wiedemann-Franz law, and allowed to be optimized if mixing conductance [the factor (2)] and the spin Hall one selects the combination of a magnetic insulator with angle [the factor (3)] are also essential. For example, low thermal conductivity and a metallic film with low to realize the efficient spin-angular-momentum transfer electrical resistivity. This is one of the unique character- across the metal/insulator interface, the improvement of istics of the LSSE device, although, in addition to the κρ crystal quality and interface condition of the LSSE de- factor, the LSSE thermopower itself must be improved. vice by annealing has been conducted [69, 138, 139]. The Importantly, in fact, the relation between the figure of spin mixing conductance can also be enhanced by in- merit and the thermoelectric conversion efficiency of the serting an ultra-thin ferromagnetic interlayer between a LSSE device should be different from that of the con- paramagnetic metal and a ferrimagnetic insulator owing 4

Conductive film Simple metal Alloy Bilayer metal Oxide

Pt Au Ir Pd Ni* W Ta Mo Nb Cr Ti NiFe* FePt* IrMn AuCu Pt Pt Pt Pt CoFeB* Co* IrO2 SrRuO3* /Cu /Au /FeCu* /Ti /Ti /Cu Magnetic insulator positive θSH negative θSH

Ref. No. Garnet ferrite 42 50,51 76,103 64 64,75 49 91 56,58,61 96 63 108 51,95 57,59 87 92

Y3Fe5O12 (YIG)

Y3-xRxFe5-yMyO12 55,84 44 44 44 44 44 44 44 80 R = Ca, Nd, Gd, Bi M = Al, Mn, Ga, V, In, Zr Fig. 12 of this article 102 Gd3Fe5O12 86 86 86 Gd3Ga5O12 (paramagnetic)

Spinel ferrite 46 (Mn,Zn)Fe2O4 82 70,82 70 54,100 Fe3O4 53 NiFe2O4 85,107 CoxFe3-xO4 (x = 0.25, 0.75, 1) 99 CoCr2O4 (multiferroic) 106 106 Ni0.2Zn0.3Fe2.5O4

Hexagonal ferrite 77 BaFe12O19 110 Ba0.5Sr1.5Zn2Fe12O22 (multiferroic)

Perovskite 78 89 La2NiMnO6 89 La0.67Sr0.33MnO3 86

DyScO3 (paramagnetic)

Corundum 104

Cr2O3 (antiferromagnetic)

Rutile 105 105 105

MnF2 (antiferromagnetic)

FIG. 2: Combination of magnetic insulators and conductive films used for the observation of the LSSE. The solid circles show the material combinations in which the LSSE have been observed, where the red (blue) circles represent the fact that the sign of the observed LSSE voltage is the same as (opposite to) that in the Pt/YIG systems. The color of the circle showing the Pt/Gd3Fe5O12 system is black, since the LSSE in this system exhibits sign reversals as a function of temperature. The numbers near the circles correspond to the reference numbers. The asterisks in the list of the conductive films represent ferromagnetic metals, which enable the hybrid thermoelectric generation based on the LSSE and anomalous Nernst effect (ANE). Typical values for the spin diffusion lengths and spin Hall angles of the conductive materials are summarized in the review on the spin Hall effects [34].

to the increase of magnetic moment density at the inter- but also these techniques are necessary for optimizing face [80, 134]. To realize the efficient spin-charge con- the LSSE devices. Although the conventional LSSE ex- version, the ISHE has been measured in various metals periments have mainly been performed using paramag- [44, 140–142], alloys [56, 58, 61, 63, 70, 96, 108, 143– netic metal/ferrimagnetic insulator junction systems, all- 145], semiconductors [146–149], oxides [89, 92, 150, 151], ferromagnetic devices [56, 58, 61, 96], i.e., ferromagnetic and organic materials [152, 153]. The spin mixing metal/ferrimagnetic (or ferromagnetic) insulator junc- conductance and spin Hall angle can be estimated by tion systems, and alternately-stacked metal/insulator means of various techniques, such as micro-wave-induced multilayer films [100] have recently been recognized as spin pumping [29, 30, 61, 63, 141, 142, 145–150, 152– useful tools for improving the thermoelectric performance 157], spin Hall magnetoresistance [158–162], and non- of the LSSE devices, as shown in Secs. V C and V D. local methods [31, 33, 140, 143–145, 151, 163–165], In this article, we review recent studies on the LSSE where these phenomena can be measured by apply- from the viewpoint of thermoelectric applications. This ing microwaves or charge currents to similar paramag- article is organized as follows. We start with an expla- net/ferromagnet junction systems instead of temperature nation of the thermoelectric generation process and mea- gradients. Therefore, not only the LSSE experiments surement configuration of the LSSE in Sec. II, followed 5 by experimental demonstrations of the basic characteris- (a) V tics of the LSSE devices in Sec. III. In Sec. IV, we formu- heat bath late the thermoelectric figure of merit and conversion effi- metallic film ciency of the LSSE devices and compare the results with those for conventional thermoelectric devices. In Sec. V, magnetic insulator slab without substrate we show preliminary demonstrations for future thermo- ∇T or electric applications of the LSSE devices, where we focus film on substrate on the device structures depicted in Fig. 1(e). The last Sec. VI is devoted to the conclusions and prospects.

H EISHE II. THERMOELECTRIC GENERATION (b) s PROCESS AND MEASUREMENT z CONFIGURATION y x jS

Figure 3(a) shows a schematic illustration of the stan- M dard configuration used for the LSSE measurements. The basic structure of the LSSE device consists of a metallic FIG. 3: (a) A schematic illustration of the typical experimen- film formed on a magnetic insulator. Here, the magnetic tal configuration for measuring the LSSE in a junction system insulator is a slab without a substrate or a film formed comprising a metallic film and a magnetic insulator. ∇T is on a substrate, while the metallic layer must be a thin applied across the metal/insulator interface by sandwiching film as explained later. In the metal/insulator junction the LSSE device between two heat baths with different tem- in the longitudinal configuration, the temperature gradi- peratures. When a soft magnetic insulator is used, an external magnetic field H (with the magnitude H) is applied to align ent, ∇T , is applied perpendicular to the metal/insulator the magnetization of the insulator along the x direction. (b) interface (along the z direction) and the spatial direction A schematic illustration of the ISHE induced by the LSSE. of the spin current induced by the LSSE is parallel to the ˆs denotes the unit vector along the electron-spin polarization ∇T direction [42]. The LSSE is usually measured with (|| M) in the metallic film. applying an external magnetic field to align the magne- tization of the magnetic insulator along the x direction [Fig. 3(a)]. However, the external magnetic field is not indispensable for the LSSE-based thermoelectric gener- excitations of localized magnetic moments, in the mag- ation if the device comprises a hard-magnetic material netic insulator driven by the temperature gradient [166– [77, 85]. Under this condition, a DC electric voltage dif- 182], since the LSSE appears even when a conduction ference V between the ends of the metallic film of the electrons’ contribution is completely frozen out. This LSSE device is measured. nonequilibrium magnon dynamics in the magnetic insu- To apply the temperature gradient in the direction per- lator can interact with conduction-electron spins in the pendicular to the metal/insulator interface, the LSSE attached metal and transfer the spin angular momentum device is usually sandwiched between two heat baths across at the metal/insulator interface via the interface s- which are stabilized at different temperatures [Fig. 3(a)] d exchange interaction, which is described in terms of the [42, 44]. By heating or cooling the heat baths, one can spin mixing conductance [133–136]. This spin-angular- apply a reversible temperature gradient to the LSSE de- momentum transfer induces a conduction-electron spin vice since its top and bottom surfaces are directly con- current in the metallic film. This spin current is con- nected to the heat bath [45]. Owing to this high temper- verted into the electric field, EISHE, by the aforemen- ature controllability, the setup depicted in Fig. 3(a) is tioned ISHE in the metallic layer if the spin-orbit inter- mostly used in the LSSE measurements; all the experi- action of the metal is strong [Fig. 3(b)]. When the mag- ments shown in this article were performed in this setup. netization M of the magnetic insulator is aligned along In contrast, the temperature gradient can be applied to the x direction, EISHE is generated in the metallic film the LSSE device also by using Joule heating in an on-chip along the y direction according to the relation [27] heater [70, 75, 76, 82], laser heating [47, 48, 57, 66, 67], E = (θ ρ)ˆs × j , (3) and microwave heating [52, 68]. Although these methods ISHE SH S enable the measurements of the LSSE without using the where θSH, ˆs, and jS are the spin Hall angle, unit vector specialized temperature-gradient generators, it is difficult along the electron-spin polarization (|| M) in the metal- to determine the temperature distributions; the meth- lic layer, and spatial direction of the spin current (with ods are sometimes combined with numerical simulations the magnitude of spin current density jS) induced by the [48, 57, 67, 68, 82]. LSSE, respectively. Thus, the thermoelectric voltage can The thermoelectric generation mechanism of the LSSE be generated by the combination of the LSSE and ISHE device is summarized as follows. The driving force of the and observed by measuring the voltage V between the LSSE is nonequilibrium dynamics of magnons, collective ends of the metallic layer, where the sign of V is deter- 6 mined by that of θSH (Fig. 2) [44]. This thermoelectric The above experiments indicate that the observed generation mechanism works below the Curie tempera- thermoelectric voltage is attributed to the LSSE. How- ture of the magnetic insulator [74]. Here, to detect the ever, to exclusively establish the LSSE, the contribution LSSE voltage, the thickness of the metallic layer has to be of the LSSE has to be separated from that of the ANE comparable to its spin diffusion length [34, 137], typically [6]. Since Pt is a paramagnetic metal and YIG is a very a few to several hundreds of nanometers, since the spin good insulator, the ANE does not exist in the Pt/YIG current injected into the metallic layer and the resultant system in the ordinary sense. In this system, however, EISHE exist only in the vicinity of the metal/insulator weak ferromagnetism may be induced in the Pt layer interface due to the spin relaxation. We also note that, in the vicinity of the Pt/YIG interface due to a static when a highly resistive insulator is used, the contribu- magnetic proximity effect [183–190] because Pt is near tion from thermoelectric phenomena in itinerant mag- the Stoner ferromagnetic instability [191, 192]. In fact, nets, such as the conventional Seebeck and Nernst effects, when the thickness of Pt is very thin (< 3 nm), weak is eliminated in the LSSE device [42, 58]. ferromagnetic signals were observed by means of X-ray magnetic circular dichroism [184–186]. If the proximity- induced ferromagnetism induces the ANE in the Pt layer, III. BASIC CHARACTERISTICS OF SPIN the ISHE voltage induced by the LSSE in the Pt/YIG SEEBECK DEVICES system may be contaminated by the proximity-induced ANE in the Pt layer. This possibility was pointed out by In this section, by using the Pt/YIG model systems, Huang et al. in 2012 on the basis of magnetoresistance we demonstrate the basic characteristics of the LSSE de- and Hall measurements [183], although the anisotropic vices, such as the dependence of the thermoelectric volt- magnetoresistance in Pt/YIG systems was subsequently age on the temperature, temperature gradient and its shown to be attributed to the spin Hall magnetoresis- direction, magnetic field and its direction, and device ge- tance [158–162]. Here, the electric field induced by the ometry. ANE is generated according to the relation [44]

EANE = N M × ∇T, (4) A. Fundamental experiments where N is the anomalous Nernst coefficient. This con- figuration is similar to that of the LSSE since EANE is Here, we show the fundamental properties of the LSSE generated along the y direction when ∇T || z and M || x [44, 51, 58]. The Pt/YIG junction system used here con- [compare Figs. 5(a)-5(c)]. After the problem presenta- sists of a single-crystalline YIG slab and a Pt film sput- tion by Huang et al., the pure detection of the LSSE tered on the well-polished surface of the YIG. The lengths was realized by using Au/YIG systems [50, 51], where of the Pt film (YIG slab) along the x, y, and z directions Au is free from the magnetic proximity effect because are lx = 2 mm (Lx = 2 mm), ly = 6 mm (Ly = 6 mm), its electronic structure is far from the Stoner instabil- and lz = 10 nm (Lz = 1 mm), respectively. To generate ity [191]. To investigate the magnetic proximity effect in ∇T along the z direction, the temperatures of the heat Pt/YIG systems, microwave spectroscopy measurements baths attached to the top and bottom of the Pt/YIG- were also carried out [187, 190]. Guo et al. theoreti- slab sample were stabilized to 300 K and 300 K + ∆T , cally investigated the proximity-induced ANE in Pt and respectively. The external magnetic field H (with the Pd within Berry-phase formalism based on relativistic magnitude H) was applied to the Pt/YIG-slab sample in band-structure calculations [189]. the x-y plane at an angle θ to the y direction [see the The clear separation of the LSSE from the proximity- inset to Fig. 4(a)]. induced ANE was reported in [51, 58] by comparing Figure 4(a) shows V between the ends of the Pt film transverse thermoelectric voltage in the Pt/YIG system in the Pt/YIG-slab sample as a function of ∆T at H = in in-plane magnetized (IM) and perpendicularly magne- 1.2 kOe. When H was applied along the x direction (θ = tized (PM) configurations. In the IM (PM) configuration, 90◦), the magnitude of V was found to be proportional to H is applied parallel (perpendicular) to the Pt/YIG in- ∆T . As shown in Fig. 4(b), the sign of V for finite values terface and ∇T is applied perpendicular (parallel) to the of ∆T is reversed in response to the sign reversal of H, interface, as shown in Fig. 5(d) [5(h)]. The IM configura- indicating that the V signal in the Pt film is affected by tion is the same as the LSSE setup (Fig. 3), where both the magnetization direction of the YIG slab (note that the LSSE and ANE can appear if they exist. In the PM the magnetization of the YIG slab is aligned along the configuration, the ANE signal can appear since the tem- H direction when H > 0.5 kOe). To confirm the origin perature gradient, magnetization, and inter-electrode di- of this signal, we also measured the θ dependence of V rection are at right angles to one another [Eq. (4)], while in the same Pt/YIG-slab sample at ∆T = 10 K and the LSSE voltage should disappear due to the symmetry H = 1.2 kOe [Fig. 4(c)]. The V signal was observed of the ISHE [Eq. (3)], where ˆs || jS in the PM config- to vary with θ in a sinusoidal pattern and vanish when uration. Therefore, the quantitative comparison of the θ = 0 and 180◦, a situation consistent with the symmetry voltage between these configurations enables the estima- of the ISHE induced by the LSSE described in Eq. (3). tion of the ANE contamination in the Pt/YIG system. 7

(a) 50 (b) 50 (c) 20 Pt/YIG-slab ΔT = 30 K

∇T 25 K 10 V) V) 20 K V) μ μ μ

( ( 15 K (

V θ V V 25 z H 0 0 y 10 K 5 K x Voltage ○ Voltage Voltage θ = 90 0 K -10 ΔT = 10 K H = 1.2 kOe θ = 90○ H = 1.2 kOe -50 -20 0 10 20 30 -2 -1 0 1 2 0 90 180 270 360 Temperature difference ΔT (K) Magnetic field H (kOe) Magnetic field angle θ (deg)

FIG. 4: (a) The temperature-difference ∆T dependence of the voltage V in the Pt/YIG-slab sample at H = 1.2 kOe and θ = 90◦, measured when ∇T || − z. θ denotes the angle between H and the y direction. (b) H dependence of V in the Pt/YIG-slab sample for various values of ∆T at θ = 90◦. (c) θ dependence of V in the Pt/YIG-slab sample at ∆T = 10 K and H = 1.2 kOe.

Figures 5(e) and 5(i) show the H dependence of the tion of the proximity-induced ANE voltage. Recently, to voltage normalized by the device length along the y di- validate the availability of the proximity-induced ANE, rection and the temperature gradient in the YIG slab, we investigated the ANE in alternately-stacked Pt/Fe V/(Ly∇TYIG), in the Pt/YIG-slab sample in the IM and multilayer films, which were designed to enhance the PM configurations, respectively [44, 58]. The magnitude proximity-induced ANE intentionally; the proximity- of V/(Ly∇TYIG) in the IM configuration was found to be ANE contribution in the Pt/Fe multilayer films is ex- much greater than that in the PM configuration. Here, pected to be up to two orders of magnitude greater than the magnitude of the normal Nernst voltage, which is the that in the conventional Pt/YIG system [120]. However, H-linear component of V/(Ly∇TYIG), in the Pt/YIG- even in these multilayer systems, no clear evidence for the slab sample in the PM configuration is comparable to existence of the proximity-induced ANE was observed. that in the Pt/paramagnetic Gd3Ga5O12 (GGG)-slab Therefore, the potential of the magnetic proximity ef- sample and in a Pt plate without a substrate [58], con- fect as a thermoelectric generation mechanism is still un- firming that the in-plane temperature gradient is gener- known. In contrast, the standard ANE in ferromagnetic ated in the Pt/YIG-slab sample in the PM configuration materials has already been investigated as a novel ther- (note that the Pt/GGG-slab sample exhibits only a nor- moelectric conversion technology [118, 121]. Especially, mal Nernst effect and no LSSE voltage except at very low we focus on ferromagnetic metals as conductive layer ma- temperatures [82]). The voltage behavior in the Pt/YIG- terials of the LSSE device since they enable the hybrid slab sample is completely different from that in a ferro- thermoelectric generation by the LSSE and ANE [44] and magnetic Ni81Fe19 film on a GGG substrate, where only replacement of noble metals, such as Pt and Au, in the the ANE and small normal Nernst effect appear; in the LSSE device, as discussed in Sec. V C. Ni81Fe19 film, the isotropic ANE voltage was observed in both the IM and PM configurations, when the voltage is normalized by the temperature gradient in the Ni81Fe19 B. High magnetic field dependence film [Figs. 5(g) and 5(k)] [44, 58]. The above results clearly show that the transverse thermoelectric voltage Recently, several research groups reported the investi- in the Pt/YIG system is dominated by the ISHE voltage gation of the high-magnetic-field response of the LSSE in induced by the LSSE and that the proximity-ANE con- Pt/YIG systems with different YIG thicknesses at vari- tamination is negligibly small. In [58], the contribution of ous temperatures [93, 94, 98, 103]. These studies provide the proximity-induced ANE voltage in the Pt/YIG sys- a significant guideline for optimizing the thickness of the tem was estimated to be less than 0.1 % of the LSSE magnetic insulator layer of the LSSE device since the voltage. As shown in Figs. 5(f) and 5(j), similar results high-magnetic-field dependence of the LSSE was found were obtained in Au/YIG systems. to be associated with the characteristic lengths of the The experimental results shown in Fig. 5 clearly LSSE. In this subsection, with showing our experimental demonstrate that the thermoelectric voltage in the results for the Pt/YIG systems [93], we review the be- Pt/YIG system is due entirely to the ISHE induced by havior of the LSSE in a high magnetic field range and its the LSSE. Although the separation of the LSSE from interpretation. the ANE is very important from the viewpoint of funda- The observation of the SSE in insulators revealed that mental physics, even the proximity-induced ANE should the magnon excitation plays a key role in this phe- be utilized effectively from the viewpoint of thermoelec- nomenon. After the pioneering theoretical work by Xiao tric applications because the thermoelectric output of et al. [166], the SSE is mainly described in terms of the the LSSE device might be enhanced by the superposi- effective magnon temperature Tm in a ferrimagnetic insu- 8

(a) LSSE paramagnetic metal In-plane magnetized configuration Perpendicularly magnetized configuration z y metal film (d) H (h) V z x V V EISHE y Ly x ∇T jS ∇T Lz insulator ∇T L slab M x H ferrimagnetic insulator 1.0 (e) 1.0 (i) Pt/YIG-slab Pt/YIG-slab V/K) V/K)

μ μ Pt/GGG-slab

(b) Anomalous Nernst effect (ANE) ) ( ) ( 0 0 YIG/GGG YIG/GGG

V T T ∇ ∇ y y L L

EANE /( Pt/GGG-slab /( V -1.0 V -1.0 0.2 (f) 0.2 (j) Au/YIG-slab Au/YIG-slab V/K) V/K)

μ 0.1 μ 0.1 Au/GGG-slab

∇T M ) ( ) ( ferromagnetic 0 0

conductor YIG/GGG YIG/GGG T T ∇ ∇ y -0.1 y -0.1 L L

/( Au/GGG-slab /( (c) Proximity-induced ANE V V -0.2 -0.2 1.0 (g) 1.0 (k) Ni Fe /GGG-slab Ni Fe /GGG-slab induced M V 81 19 81 19 V/K) V/K) μ μ EANE ) ( ) (

NiFe 0 NiFe 0 T T ∇ ∇ y y L L /( /( V V ∇T -1.0 -1.0 M -5 0 5 -50 0 50 Magnetic field H (kOe) Magnetic field H (kOe)

FIG. 5: (a)-(c) Schematic illustrations of the LSSE (a), ANE (b), and proximity-induced ANE (c). EANE denotes the electric field generated by the ANE. (d) A schematic illustration of the metal-film/insulator-slab junction system in the in-plane magnetized (IM) configuration. (e),(f) H dependence of V/(Ly∇TYIG(GGG)) in the Pt/YIG (GGG)-slab and Au/YIG (GGG)- slab samples in the IM configuration. (g) H dependence of V/(Ly∇TNiFe) in the Ni81Fe19/GGG-slab sample in the IM configuration. (h) A schematic illustration of the metal-film/insulator-slab junction system in the perpendicularly magnetized (PM) configuration. (i),(j) H dependence of V/(Ly∇TYIG(GGG)) in the Pt/YIG (GGG)-slab and Au/YIG (GGG)-slab samples in the PM configuration. (k) H dependence of V/(Ly∇TNiFe) in the Ni81Fe19/GGG-slab sample in the PM configuration. Lx, Ly, and Lz are the lengths of the insulator slab along the x, y, and z directions, respectively. ∇TYIG(GGG) and ∇TNiFe denote the temperature gradients in the YIG (GGG) slab and Ni81Fe19 film, respectively. The sample consists of a 10-nm-thick Pt, Au, or Ni81Fe19 film formed on a single-crystalline YIG or GGG slab. All the measurements were performed at T = 300 K.

lator and effective electron temperature Te in an attached SSE were also developed [169–171, 174–182]. However, paramagnetic metal; when the effective magnon-electron microscopic understanding of the relation between the temperature difference is induced by an external temper- magnon excitation and thermally generated spin current ature gradient, a spin current is generated across the fer- is yet to be fully established, and more detailed studies rimagnet/paramagnet interface due to the thermal spin are necessary. Since the magnon excitation is modulated pumping. Subsequently, Adachi et al. developed linear- by a magnetic field due to the Zeeman gap gµBH, the response theories of the magnon- and phonon-mediated ISHE voltage induced by the SSE can also be affected by SSEs [167, 168, 172]. Hoffman et al. formulated a the magnetic field. Therefore, systematic measurements Landau-Lifshitz-Gilbert theory of the SSE to investigate of the magnetic-field-induced response of the SSE become the thickness dependence and length scale of the SSE powerful tools for unraveling the thermo-spin conversion [173]. In 2014, Rezende et al. discussed the SSE in terms mechanism based on the magnon excitation. of a bulk magnon spin current created by a tempera- In Fig. 6(a), we show the transverse thermopower ture gradient in a ferrimagnetic insulator [60]. Further- S (≡ (V/∆T )(Lz/Ly)) in the Pt/YIG-slab sample as more, various theoretical models of the magnon-driven a function of H for several values of the temperature 9

(a) (b) 10-5 5 1.0 T = 300 K Pt/YIG-slab S (μV/K) 0 0 10-6 -1.0 Pt/YIG-slab

-5 ) H = 1.8 kOe

1.0 /f.u. H = 80 kOe T = 10 K B S (μV/K) 5

LSSE thermopower -7 (μ 10

M 10 100 0 0 Temperature T (K)

(c) 0 -5 -1.0 (%) 20 Magnetization LSSE δ

LSSE thermopower 0.5 T = 5 K 5 40

0 0 60 Experiment 80 Calculation -0.5 -5 LSSE suppression 100 -50 0 50 10 100 Magnetic field H (kOe) Temperature T (K)

FIG. 6: (a) H dependence of the LSSE thermopower S in the Pt/YIG-slab sample and the magnetization M of the YIG slab at T = 300 K, 10 K, and 5 K, measured when H was swept between ±90 kOe. The sample consists of a 5-nm-thick Pt film sputtered on the top surface of a single-crystalline YIG slab. (b) Double logarithmic plot of the T dependence of S in the Pt/YIG-slab sample at H = 1.8 kOe (closed circles) and 80 kOe (open triangles). (c) T dependence of the suppression of the LSSE voltage by magnetic fields δSSE in the Pt/YIG-slab sample (circles). A blue line shows the T dependence of δSSE calculated based on the conventional SSE model (see [93] for details).

T , measured when H was swept between ±90 kOe. The suppression of the LSSE voltage and the conventional for- clear LSSE voltage was observed in the Pt/YIG-slab sam- mulation at relatively-high temperatures is attributed to ple at all the temperatures and its magnitude at each the fact that the small Zeeman energy is defeated by ther- temperature gradually decreases with increasing H after mal fluctuations when gµBH  kBT in the conventional taking the maximum value, while the magnitude of M models, where g, µB, and kB are the g-factor, Bohr mag- is almost constant after the saturation. This suppression neton, and Boltzmann constant, respectively (note that of the LSSE voltage becomes apparent by applying high the magnon gap energy at H = 80 kOe corresponds to magnetic fields, while it is very small in the conventional gµBH/kB = 10.7 K); to affect the magnon excitation by LSSE measurements in a low field range. The suppres- magnetic fields, the magnon energy has to be compara- sion of the LSSE voltage was shown to be irrelevant to ble to or less than the Zeeman energy. In contrast, the the normal Nernst effect in the Pt film [93]. observed large suppression of the LSSE voltage in the The magnetic-field-induced suppression of the LSSE Pt/YIG-slab sample indicates that the magnon excita- voltage in the Pt/YIG-slab sample increases with de- tion relevant to the LSSE is affected by magnetic fields creasing the temperature. Figure 6(c) shows the T de- even at around room temperature. This result suggests pendence of the suppression of the LSSE thermopower that low-frequency magnons of which the energy is com- δSSE in the same Pt/YIG-slab sample, where δSSE is de- parable to the Zeeman energy provide a dominant contri- fined as (Smax − S80kOe)/Smax with Smax and S80kOe re- bution to the LSSE; the thermo-spin conversion efficiency spectively being the S values at the maximum point and of the LSSE has magnon-frequency dependence, which is at H = 80 kOe [the T dependence of S at the posi- not included in the conventional SSE theories. In [93] tive H values is shown in Fig. 6(b)]. The field-induced and [193], the origin of this spectral non-uniform thermo- suppression in the Pt/YIG-slab sample was observed to spin conversion is discussed in terms of the frequency de- be almost constant above 30 K and strongly enhanced pendence of a magnon thermalization (energy relaxation) below 30 K; the δSSE value in the Pt/YIG-slab sample length and a magnon diffusion length, respectively. It is reaches ∼70 % at T = 5 K. The field-induced suppres- notable that lower frequency magnons exhibit the longer sion of the LSSE for T > 30 K cannot be explained by characteristic lengths in general [126–129, 166, 193–199]. the conventional SSE models, while that for T < 30 K The above results and discussions indicate that, to seemingly agrees with numerical calculations based on maximize the thermoelectric output of the LSSE de- the thermal spin pumping mechanism [Fig. 6(c)] (see vice, the thickness of the magnetic insulator has to be [93] for details). The inconsistency between the observed greater than the characteristic lengths of low-frequency 10

(a) Pt/YIG-slab Pt/YIG-film C. Current-voltage-power characteristics and scaling law YIG thickness 10.42 μm 1.09 μm 0.31 μm 1.0 1 mm Now we demonstrate the current-voltage-power char- 0 acteristics of the LSSE device by using a Pt/BiY2Fe5O12

S (μV/K) (Bi:YIG) bilayer film [44]. To determine the characteris- tics, we attached a load resistance Rload to the Pt layer

LSSE thermopower -1.0 and measured the voltage across the load resistance Vload 2 -50 0 50 -50 0 50 -50 0 50 -50 0 50 to estimate the output power Vload/Rload with chang- H (kOe) H (kOe) H (kOe) H (kOe) ing the Rload value, while most of the LSSE experiments (b) (c) have been performed under the open-circuit condition. 1.0 0 2 In Fig. 8(a), we show Vload/Rload generated from the Pt/Bi:YIG-film sample at ∆T = 11.2 K as a function of

value 10 Rload. We found that the maximum output power can S 0.5 (%) (μV/K) be extracted when Rload is equal to the internal resis- LSSE max δ

S 20 tance of the Pt layer: RPt = 33 Ω. This behavior is Maximum T = 300 K LSSE suppression T = 300 K also confirmed by the linear current-voltage relation and parabolic current-power relation for the Pt/Bi:YIG-film 0 30 10-7 10-6 10-5 10-4 10-3 10-7 10-6 10-5 10-4 10-3 sample in Fig. 8(b). The current-voltage-power charac- YIG thickness (m) YIG thickness (m) teristics and optimizing condition for the load resistance for the LSSE devices are the same as those for any linear FIG. 7: (a) H dependence of S in the Pt/YIG-slab sam- generators, such as batteries and conventional thermo- ple with the YIG thickness of 1 mm and in the Pt/YIG- electric modules. film samples with the YIG thickness of 10.42 µm, 1.09 µm, and 0.31 µm at T = 300 K, measured when H was swept Next, we experimentally verify the scaling law of the between ±90 kOe. The YIG films were grown on single- LSSE device: the output power is proportional to the crystalline GGG substrates by a liquid phase epitaxy method. device area, introduced in Sec. I. To do this, we inves- (b) YIG-thickness dependence of Smax at T = 300 K. (c) tigated the dependence of the LSSE voltage on the size, YIG-thickness dependence of δSSE at T = 300 K. i.e., length and width, of the metallic layer of the LSSE device as follows [101]. For measuring the metallic-layer-length dependence of the LSSE, we used the structure called “spin Hall ther- mopile” [49]. The spin Hall thermopile consists of an alternating array of two different metallic wires with dif- ferent spin Hall angles, connected in series in a zigzag configuration. The sample used here is the spin Hall ther- magnons providing a strong contribution to the LSSE, mopile comprising 10-nm-thick Pt wires and 30-nm-thick since the contribution from the long-range magnons can Au wires formed on a Bi:YIG film [see the inset to Fig. be limited by boundary conditions in thin magnetic in- 8(c)]. In this sample, the LSSE voltage in each wire is sulators [93]. In fact, several research groups demon- integrated into the total voltage between the ends of the strated that, by using the Pt/YIG-slab and Pt/YIG-film Pt-Au thermopile and the total voltage predominantly systems, the magnitude of the LSSE thermopower mono- comes from the Pt wires, because the spin Hall angle tonically decreases with decreasing the thickness of YIG and resistivity of Pt are greater than those of Au and the [our experimental results are shown in Figs. 7(a) and thickness of the Pt wires is less than that of the Au wires 7(b)] [93, 95, 103]. Significantly, the suppression of the (note that the magnitude of the LSSE voltage increases LSSE by high magnetic fields, δSSE, also monotonically with decreasing the thickness of the metallic layer except decreases with decreasing the YIG thickness [Figs. 7(a) for ultrathin regions [58, 64, 69]). Figure 8(c) shows the and 7(c)]. This behavior indicates that the contribution dependence of the LSSE voltage per unit temperature tot of low-frequency magnons, which govern the LSSE sup- difference, V/∆T , on the total length of the Pt wires ly tot pression in the Pt/YIG-slab sample, fades away in the in the Pt-Au thermopile, where ly is changed by chang- Pt/YIG-film samples when the YIG thickness is less than ing the number of the Pt-Au pairs. As expected above, tot their characteristic lengths and that only remaining con- the magnitude of V/∆T increases in proportion to ly . tribution from high-frequency magnons, which have en- In addition to the Pt-length dependence, we also inves- ergy much greater than the Zeeman energy and provide tigated the Pt-width dependence. We fabricated Pt films a weak contribution to the LSSE, appears in the thin with different widths on fixed-size Bi:YIG films, where YIG-film samples. As reviewed in this subsection, the the length and thickness of the Pt films are fixed [see the measurements of the high-magnetic-field response of the inset to Fig. 8(d)]. In Fig. 8(d), V/∆T and the Pt-film LSSE are useful for characterizing the properties of the resistance RPt are plotted as a function of the Pt width LSSE devices associated with magnon excitation. lx. We found that the magnitude of V/∆T is almost in- 11

(a) (b) 10 30 5

) Pmax )

ΔT = 11.2 K ΔT = 11.2 K W

)

V 25

(p

(pW

μ 4

(

load

load load

/R 1 /R 2 20

2 Vload

load

load 3 R load load 15 resistance 2 0.1 RPt = 33 Ω 10 Pt 1

Bi:YIG On-load voltage V 5

On-load power V

On-load power V 0.01 0 101 102 103 104 0 0.2 0.4 0.6

Load resistance Rload (Ω) On-load current Iload (pA)

(c) 15 (d) 1.5 600

V Pt 500 Au

(Ω)

10 Bi:YIG 1.0 400 Pt

) /K ) /K

V z V V μ y μ 300 x lx Pt V/ΔT ( 5 V/ΔT ( 0.5 Bi:YIG 200

Pt resistance R 100

Voltage / Temperature difference Voltage / Temperature difference 0 0 10 20 30 40 50 0 1 2 tot Total Pt length ly (mm) Pt width lx (mm)

2 FIG. 8: (a) On-load power Vload/Rload induced by the LSSE in the Pt/Bi:YIG-film sample as a function of the load resistance 2 Rload at ∆T = 11.2 K. Vload denotes the on-load voltage. (b) Vload and Vload/Rload induced by the LSSE in the Pt/Bi:YIG-film sample as a function of the on-load current Iload = Vload/Rload. The Pt/Bi:YIG-film sample used for the experiments (a) and (b) consists of a 10-nm-thick Pt film with the size of 16 × 16 mm2 and a 120-nm-thick Bi:YIG film formed on a 0.5-mm-thick tot SGGG wafer of 1-inch diameter. (c) V/∆T as a function of the total Pt length ly in the [Pt-Au thermopile]/Bi:YIG-film samples. (d) V/∆T and the Pt-film resistance RPt as a function of the Pt width lx in the Pt/Bi:YIG-film samples. All the measurements were performed at room temperature.

dependent of lx, while RPt is inversely proportional to called a transverse device [5]. Next, based on the discus- lx. sion of the ANE generators, we formulate the efficiency tot The obtained scaling results, i.e., V/∆T ∝ ly and of the LSSE devices. V/(RPt∆T ) ∝ lx, clearly indicate that a larger film area leads to larger thermoelectric output power that can be extracted to an external load. Here, we would like to em- A. Review of efficiency of thermoelectric phasize again that the scaling law of the LSSE devices is generators based on anomalous Nernst effects quite different from that of conventional thermoelectric devices based on the Seebeck effect, where the thermo- As pointed out above, it is quite instructive to review electric output scales with the number of thermocouples the efficiency calculation of the ANE device [200–202]. connected in series (see Sec. I and Fig. 1). We begin with the following transport equations [203]: E = ρj + α∇T + N M × ∇T, (5) q = Πj − κ∇T + N T M × j, (6) IV. THEORY OF EFFICIENCY OF SPIN SEEBECK THERMOELECTRIC DEVICES where E, j, and q are an applied electric field, charge current density (with the magnitude j), and heat current In this section, we present a theory of the efficiency of density, respectively. Π = αT is the Peltier coefficient. thermoelectric devices based on the LSSE. First, we give The (anomalous) Hall coefficient is disregarded as it is a brief review of the efficiency of thermoelectric genera- irrelevant to the present discussion, and the Leduc-Righi tors based on the ANE. This is because the ANE has a coefficient [204, 205] is assumed to be negligibly small. similarity to the LSSE in that the induced electric field Here we note that the following discussion is applicable is orthogonal to the applied temperature gradient, and not only to the ANE devices but also to normal Nernst a thermoelectric device possessing this characteristic is devices [6] if the spontaneous magnetization is replaced 12

2 with an external magnetic field, while this article focuses where Pout = RloadIy is the electrical power output and on the comparison between the LSSE and ANE devices Qh is the thermal power input from the hot reservoir. because of their similar characteristics. It should also The latter quantity is calculated as follows [208, 209]. be added that the thermoelectric conversion efficiency of First, we assume a uniform charge current distribution normal Nernst generators, or Ettingshausen coolers, in along the z direction, i.e., jy(z) = j = constant. Then, semimetals can be much greater than that of the ANE from Domenicali’s equation [Eq. (8)] under the condition and LSSE devices, although an external magnetic field T (z = 0) = Th and T (z = lz) = Tc, the temperature must be applied [206]. distribution is calculated to be We first calculate the amount of heat evolved per unit 2 time and volumes, which governs the temperature profile T (z) = Th − (z/lz)∆T + (ρj /2κ)z(lz − z), (12) of the system in the steady state. This quantity is given where the distribution of T (z) is schematically shown by −∇·qE [203], where qE = q+φj is the energy current density with φ being the electrochemical potential. Using in Fig. 9(a). Next, we evaluate the magnitude of the Eqs. (5) and (6), we obtain heat current at the hot reservoir, i.e., qh = N MsT j − κ(∇zT )z=0, using Eq. (6). Then, the rate of heat re- 2 2 − ∇ · qE = ρj + κ∇ T moval from the hot reservoir, defined by Qh = (lxly)qh, − T j · ∇α + 2N (j × M) · ∇T, (7) is calculated to be where the first and second terms on the right hand N MsTh R 2 Qh = Iy + K∆T − Iy , (13) side represent the Joule heating and thermal conduc- (lz/ly) 2 tion. The third term, which comes from the temper- ature dependence of the Seebeck coefficient via the re- where K = κ(lxly/lz) is the thermal conductance in the lation ∇α = (dα/dT )∇T , is the Thompson effect and z direction and R = ρ(ly/lxlz) is the electrical resistance will be neglected hereafter to simplify the argument. Be- in the y direction. sides, since the last term can be interpreted as a change From Eqs. (11)-(13), the efficiency is expressed as a in the Thompson effect due to the presence of magne- function of r = Rload/R: tization [203], this term is also discarded. Therefore, by considering the conservation law of energy flux, i.e., ηANE = ηCf(r), (14) ∇ · qE = 0, we obtain the following equation: where ηC = ∆T/Th is the Carnot efficiency. Here, the ρj2 + κ∇2T = 0, (8) characteristic function, r which determines the temperature distribution in ther- f(r) = , (15) (1+r)2 ∆T moelectric devices, known as Domenicali’s equation [207]. −(1 + r) + 0 − ZANETh 2Th Let us focus on the ANE device shown in Fig. 9(a). Here, a temperature gradient ∇zT = −∆T/lz (∆T = is defined by the isothermal figure of merit Th −Tc) is applied along the z axis, and an external mag- 2 netic field is applied along the x axis to align the mag- 0 (N Ms) ZANE = 0 , (16) netization, M = Msˆx. Under the isothermal condition κ ρ (∇yT = 0) as well as the open circuit condition (j = 0), 0 the ANE induces an electric field Ey = N Ms∆T/lz along where κ is the thermal conductivity at zero electric field. the y axis, and a charge current density is driven in the Note the minus sign in the first term of the denominator y direction, i.e., j = jy ˆy. Since the electromotive force is in Eq. (15), which is peculiar to transverse devices [200]. given by By maximizing ηANE with respect to r, the optimized ∗ efficiency ηANE in the limit of ∆T  Th is calculated to Vemf = −N Ms∆T (ly/lz), (9) be  ∗  and the total resistance in the circuit is given by Rtot = ∗ 1 − r η = ηC , (17) R+Rload with R and Rload respectively being the internal ANE 1 + r∗ and load resistances, the resultant charge current I = Iy ˆy q ∗ 0 is calculated to be r = 1 − ZANETh, (18)

Vemf −N Ms∆T (ly/lz) Iy = = , (10) where the functional form is quite different from the con- Rtot R(1 + r) ventional Seebeck device because of the fact that the where we have introduced a dimensionless variable r ≡ ANE device has a symmetry of transverse devices. Note that, by contrast, the maximum power output is given by Rload/R. Now we discuss the efficiency of the ANE device. The the impedance matching condition R = Rload as is well efficiency is defined by known [208]. Two comments are in order. First, according to the ηANE = Pout/Qh, (11) argument of [210] and [211], the isothermal figure of merit 13

(a) ANE-based thermoelectric device (b) LSSE-based thermoelectric device z z x x y y

lx lx M M

paramagnetic ferrimagnetic ferromagnetic conductor metal insulator Rload Rload ly ly

lz lz Lz hot cold hot cold

Temperature T (z) Temperature T (z)

Th Th T0

Tc Tc z z 0 lz -lz 0 Lz

FIG. 9: (a) A schematic illustration of the ANE-based thermoelectric device and its temperature profile along the z direction. (b) A schematic illustration of the LSSE-based thermoelectric device and its temperature profile along the z direction. The load resistance Rload is attached to the ferromagnetic conductor (paramagnetic metal) of the ANE (LSSE) device. lx, ly, and lz are the lengths of the ferromagnetic conductor (paramagnetic metal) of the ANE (LSSE) device along the x, y, and z directions, respectively. Th(c) is the temperature at the hot (cold) end of the devices and T0 is the temperature at the metal/insulator interface of the LSSE device.

0 ZANET in Eq. (15) is related to the adiabatic figure of pression; otherwise such a simple result would never be 2 merit ZANET = (N Ms) /κρ as obtained and a more involved numerical approach would be required, which is beyond our scope. Also, we as- 0 ZANET sume that there is no discontinuous jump in the temper- ZANET = , (19) 1 + ZANET ature distribution at the P/F interface. In this sense, 0 the present approach is complementary to the calcula- such that 0 ≤ ZANET ≤ 1, where κ is the thermal con- ductivity at zero charge current. Therefore, the maxi- tion in [130], where only the temperature difference at 0 the P/F interface was considered as a driving force. In- mum allowed efficiency is obtained for ZANET = 1 (i.e., ∗ max stead, we assume that the temperature is a smooth func- r = 0), achieving the Carnot efficiency: ηANE = ηC. 0 tion across the P/F interface, and the temperature at the Second, the figure of merit ZANE in Eq. (16) is deter- mined by the thermal conductivity κ0 and the electrical interface T0 is determined by the heat current conserva- resistivity ρ that are defined in the same material. tion. Here, the relationship between the present contin- uous temperature model and a three temperature model used in [57, 130, 166] is given in the following way. The B. Efficiency of thermoelectric generators based on present approach is constructed based on the formalism spin Seebeck effects developed in [196], and the magnon accumulation [see Eq. (11) of [196]] corresponds to the effective magnon-phonon In this subsection, based on the discussion in the pre- temperature difference. It has been demonstrated that vious subsection, we discuss the efficiency of the LSSE the present single temperature gradient model provides device. We consider a device shown in Fig. 9(b), where us with an excellent description of the SSE in magnetic a bilayer of a paramagnetic metal P and a ferrimagnetic multilayers [100]. insulator F is sandwiched between hot and cold reser- We start our discussion from the following transport equations for P: voirs. We assume that the thermal conductivity of P κP is much greater than that of F κ , such that the temper- F j = σ E + θ ˆs × j , (20) ature gradient develops inside F and thus the tempera- P SH S ture gradient inside P is negligibly small in comparison jS = σPES + θSHˆs × j, (21) to F. Looking at Table I given in [212], this assumption q = −κP∇T, (22) seems moderately reasonable. Note that we made this assumption in order to keep the mathematical manipula- where ES is the spin voltage gradient and σP is the elec- tions manageable and to extract a simple analytical ex- trical conductivity of P. For F, we assume the following 14 transport equations: where the coefficient c2 (0 < c2 < 1) represents the ratio of the spin current injected back into F to the spin-Hall jS = σmES − ζm∇T, (23) drift current δjS in P. Note that the heat current ac- q = ΠmjS − κF∇T, (24) companied by the spin current given in Eq. (25), i.e., q = Πmc1ζm(∆T/Lz)ˆz, is absorbed into the definition of where ES in F is determined by the magnon density κF as κF − Πmc1ζm → κF. gradient ∇nm, σm is the magnon conductivity, and ζm Now we can calculate the thermal power input Qh from (Πm) is the magnon Seebeck (Peltier) coefficient. Here, the hot reservoir. We assume that Domenicali’s equation mag ph [Eq. (8)] approximately holds for P, neglecting spin Joule the thermal conductivity κF = κF + κF of F includes both magnon contribution κmag as well as phonon con- heating [213]. Using heat current conservation at the P/F F 2 ph interface, we have qh = κF(∆T/Lz) − |δq| − ρPj , where tribution κF , and Onsager’s reciprocity [212] requires ρP = 1/σP. From this, we have Πm = ζmT/σm. Let us first calculate the open circuit voltage under 2 Qh = KF∆T + c2ΠmT (ly/lz)Iy − RIy (29) a temperature bias ∆T = Th − Tc between the hot and cold reservoirs. As stated before, because of the condition for Qh = (lxly)qh. At the same time, the output power κP  κF, the temperature gradient develops inside F as is given by ∇T = −[(Th − T0)/Lz]ˆz ≈ −(∆T/Lz)ˆz. According to Eq. (23), this temperature gradient drives a spin current 2 Pout = RloadIy , (30) jS = ζm(∆T/Lz)ˆz inside F, which then injects an amount of so that after a lengthy but straightforward calculation, the efficiency ηSSE = Pout/Qh is obtained: jS = c1ζm(∆T/Lz)ˆz (25) ηSSE(r) = ηC g(r), (31) into P, where the coefficient c1 (0 < c1 < 1) represents the ratio of the spin current injected into P to the ther- where the characteristic function g(r) is defined by mal drift magnon current in F. Note that the parameter r c1 is proportional to the spin mixing conductance. If g(r) = 2 , (32) we adopt an approach of [60] and [196] where the conti- ( Lz ) (1+r) − ( Lz )ξ(1 + r) − ∆T lz ZSSETh lz Th nuity of jS at the P/F interface is postulated, the con- ξ = (c2/c1)(σP/σm), (33) stant c1 is given by c1 ∝ [cosh(Lz/Λ) − 1]/ sinh(Lz/Λ), where Λ is the magnon diffusion length of F. Inside P, and the figure of merit of the LSSE device ZSSE is given thanks to the ISHE [Eq. (20)], the injected spin current by jS = [c1(ζm∆T )/Lz]ˆz is converted into a charge current 2 j = −[θSHc1ζm∆T/Lz]ˆy when ˆs = ˆx. Then, in the open (θSHαS) circuit condition, this current is compensated by an elec- ZSSE = . (34) κFρP tromotive force By maximizing ηSSE with respect to r, the optimized Vemf = −θSHαS∆T (Ly/Lz), (26) efficiency in the limit of ∆T  Th is calculated to be

   ∗  where αS = (1/2)c1ζm/σP. Here, the factor 1/2 in αS ∗ lz 1 1 − r ηSSE = ηC ∗ , (35) comes from averaging over the thickness of P, where we Lz ξ 1 + r approximated the hyperbolic variation of jS(z) by a lin- ∗ p ear curve by assuming that the thickness of P is of the r = 1 − ξZSSETh, (36) order of its spin diffusion length λ (∼ l ). For a circuit z where the functional form again has a characteristic of containing the total resistance R = R + R , the cur- tot load transverse devices. Note that, because the result is quite rent I = I ˆy driven by the open circuit voltage [Eq. (26)] y different from that of [130] and several important steps is given by to arrive at the final result therein are missing, we must conclude that it is almost impossible for us to draw par- θSHαS∆T (Ly/Lz) Iy = − (27) allel between the two approaches. In Table I, we summa- R(1 + r) rize parameters that influence the efficiency of the LSSE thermoelectric generators. where r = Rload/R as before. Owing to Eq. (21), the resultant charge current density j = [Iy/(lxlz)]ˆy drives an additional spin current δjS = [(θSHIy)/(lxlz)]ˆz, a part of which is injected back into F. Then, due to Eq. (24), C. Discussion this spin current is accompanied by a heat current First of all, it is important to note that, in comparison ΠmθSHIy with the ANE device [Eq. (16)], the figure of merit of the δq = c2 ˆz, (28) (lxlz) LSSE device is determined by the thermal conductivity 15

TABLE I: Parameters that influence the efficiency of the LSSE thermoelectric generators.

θSH Spin Hall angle of P αS Spin Seebeck coefficient [Eq. (26)] κF Total thermal conductivity of F ρP Electrical resistivity of P lz Thickness of P Lz Thickness of F σP Electrical conductivity of P, σP = 1/ρP [Eq. (20)] σm Magnon conductivity of F [Eq. (23)] c1 Ratio of spin current injected into P to thermal drift magnon current in F [Eq. (25)] c2 Ratio of spin current injected back into F to spin-Hall drift current in P [Eq. (28)] λ Spin diffusion length of P Λ Magnon diffusion length of F

∗ κF and the electrical resistivity ρP that are defined in two the situation r = 0 cannot be realized. Note also that, different materials, reflecting the fact that the charge cur- although at first glance it appears that a use of a thin rent flows only through P whereas the thermal resistance magnetic layer (Lz  lz) would result in a very large 0 is mainly given by F. Therefore, in contrast to ZANE of efficiency, this naive expectation does not work since 1/ξ the ANE device, ZSSE of the LSSE device is free from is proportional to Lz/Λ for Lz/Λ  1. Anyway, as is the limitation of the Wiedemann-Franz law. obvious from the structure of the present calculation, ∗ The optimized efficiency ηSSE is determined by three the efficiency of the LSSE device is bound by the con- factors: the figure of merit ZSSE, the spin converting pa- straint of the second law of thermodynamics. Therefore, rameter ξ, and the geometrical factor lz/Lz. Thus, for there must be a condition (lz/Lz)(1/ξ) ≤ 1 to ensure ∗ the calculation of ηSSE, we need to estimate the mag- that the upper limit of the maximum allowed efficiency nitude of ξ [Eq. (33)], in addition to the ZSSE value. is the Carnot efficiency, a microscopic derivation of which The parameter ξ may be evaluated in the following way. is left to future studies. However, the efficiency of present First, since both the parameters c1 and c2 measure the LSSE devices is still very low; making use of typical ratio of a spin injection current to a bulk spin current, we values for Pt/YIG systems: lz = 10 nm, Lz = 1 µm, could set c2/c1 ≈ 1 for a rough estimate (a microscopic ρP = 0.1 µΩm, κF = 7 W/mK, θSHαS = 1 µV/K, and determination of the factor c2/c1 is beyond our scope ξ = 1, the figure of merit and the ratio of the optimized since the present discussion is devoted to a phenomeno- efficiency to the Carnot efficiency are respectively esti- −4 ∗ −4 logical description of the efficiency). Next, the conduc- mated to be ZSSET = 4×10 and ηSSE/ηC = 1×10 % ∗ tivity ratio may be estimated using Einstein relations, at T = 300 K, where ηSSE has a very little dependence 2 2 σP = e DPgP(F) and σm = e Dmnm/(kBT ), where DP on ξ when ξZSSET  1. Therefore, the dramatic en- and Dm are respectively the electron diffusion coefficient hancement of the LSSE thermopower and the reduction of P and magnon diffusion coefficient of F, gP(F) is the of thermal conductivity of F are necessary for realistic electron density of states per unit volume at the Fermi thermoelectric applications of the LSSE. The determina- energy, and nm is the magnon number density. With tion of the optimum geometrical factor, lz/Lz, is also im- these estimates, we obtain portant for improving the efficiency, while the increase of lz/Lz may conflict with the thickness dependence of the  D  g (ε ) ξ ≈ P P F , (37) LSSE because the LSSE thermopower usually decreases Dm nm/(kBT ) with increasing lz [50, 58, 64, 69, 91] and decreasing Lz [43, 93, 95, 98, 103]. R where nm can be calculated by nm = dgm()fBE() with the magnon density of states per unit volume gm() and the Bose-Einstein distribution function f (). BE V. DEMONSTRATIONS FOR Now we discuss the upper limit of the efficiency η∗ . SSE THERMOELECTRIC APPLICATIONS Similarly to the isothermal figure of merit of the ANE device [Eq. (19)], the isothermal LSSE figure of merit is In this section, we introduce several approaches for expected to satisfy 0 ≤ ZSSET ≤ 1. Then, the maxi- mum allowed efficiency of the LSSE device [Eq. (35)] is future thermoelectric applications of the LSSE devices, ∗ which include the demonstration of the LSSE-based ther- achieved for r = 0 (i.e., ξZSSETh = 1), which yields moelectric generation in coating films, flexible devices,   max lz 1 all-ferromagnetic devices, and multilayer films. The for- ηSSE = ηC . (38) mer two structures are under development to exploit the Lz ξ versatility and scalability of the LSSE devices, while the Note that the existence of such a maximum allowed effi- latter two are to improve their thermoelectric perfor- ciency requires a condition that 0 ≤ 1/ξ ≤ 1; otherwise mance. 16

A. Spin thermoelectric coating (a) MOD method (b)

Spin coating Pt/Bi:YIG-film 10

The concept of a coating-based LSSE device called V) μ

“spin thermoelectric (STE) coating” was first proposed MOD ( solution V in 2012 [43]. The STE-coating device consists of a 0 single Pt film metal/magnetic insulator bilayer film directly coated on a SGGG heat source. Thanks to the simple structure and straight- Voltage -10 forward scaling law of the LSSE, the STE coating is po- tentially applicable to the implementation of large-area -10 -5 0 5 10 thermoelectric devices onto various-shaped heat sources, Drying Temperature difference ΔT (K) (c) such as the surfaces of electronic instruments and auto- Annealing Pt/Bi:YIG-film mobile bodies. The STE coating was demonstrated by using a spin- single Pt film ΔT = 8 K coating method, which may realize large-area thermo- Bi:YIG

electric devices in a highly productive way. Here, we V)

μ 4 K used Bi:YIG and Pt as the magnetic insulator and metal- ( V lic film layers, respectively. The STE-coating films were 0 K formed by using the simple fabrication steps illustrated in Fig. 10(a). First, a Bi:YIG film was formed on Voltage -4 K Sputtering a (GdCa)3(GaMgZr)5O12 (substituted gadolinium gal- lium garnet: SGGG) substrate, which has good lattice- -8 K Pt 0 matching properties with Bi:YIG, by means of a metal- 10.0 μV organic decomposition (MOD) method [43]. The MOD -200 0 200 method consists of simple three step processes: spin coat- SGGG Bi:YIG Magnetic field H (Oe) ing, drying, and annealing of the MOD solution contain- ing the constituent elements, which enables the fabrica- FIG. 10: (a) Preparation process for the Pt/Bi:YIG STE- tion of thin magnetic insulator films without using spe- coating device. (b) ∆T dependence of V in the Pt/Bi:YIG- cialized and costly equipment, such as vacuum deposi- film sample and single Pt film. (c) H dependence of V in tion apparatus. Subsequently, the Pt film was formed by the Pt/Bi:YIG-film sample and single Pt film for various val- sputtering over the whole surface of the Bi:YIG film. The ues of ∆T . All the measurements were performed at room thicknesses of the Bi:YIG and Pt films are 120 nm and temperature. 10 nm, respectively. Importantly, although the surface area of the samples used here is small (8 × 2 mm2), the above process can be easily applied to large-area manu- facturing, since it does not require patterning steps such ples connected electrically in series, it is vulnerable to as photolithography and electron-beam lithography. bending stresses, making it difficult to construct flexible In Figs. 10(b) and 10(c), we respectively show the ∆T devices. In contrast, the LSSE device has a high affinity and H dependences of V in the Pt/Bi:YIG STE-coating for flexible thermoelectric generation owing to its simple film. The clear LSSE voltage was found to appear even structure and scaling law; it will be applicable on various in this thin film structure prepared by the MOD coat- heat sources, paving the way to versatile thermoelectric ing method. We also checked that no signal appears generators or sensors. in a plain Pt film sputtered directly on a SGGG sub- Recently, we demonstrated the construction of a LSSE- strate, indicating that the thin Bi:YIG film works as a based flexible thermoelectric sheet [106]. To construct thermoelectric generator. We found that a crystalline the flexible LSSE device, we used spray-coating tech- Bi:YIG film can be grown even onto a glass substrate nique called “ferrite plating” [214, 215], which enables by using the MOD method [43], and the STE coating the fabrication of ferrimagnetic ferrite thin films. Since is applicable even onto amorphous surfaces. Such versa- conventional ferrite-film preparation techniques, includ- tile implementation of thermoelectric functions may open ing sputtering [85], liquid phase epitaxy [139], and pulsed opportunities for various applications making full use of laser deposition [135], require high temperature processes omnipresent heat. (ranging from 400 ◦C to 800 ◦C) for crystallizing ferrites, they cannot be used for the formation of ferrite films on heat-labile soft materials, such as plastics. By contrast, B. Flexible spin Seebeck devices the ferrite plating method is based on chemical reaction processes, and thus does not need any high temperature Most of the conventional thermoelectric devices are processes, enabling the coating of ferrite films on a vari- rigid, and not easily applicable onto curved or uneven ety of substrates including flexible plastic films. heat sources. Because the conventional thermoelectric The fabrication process of the ferrite plating method device consists of a number of Π-structured thermocou- is very simple, which also requires no specialized and 17

(a) Ferrite plating method (b) ture can be suitable for the flexible LSSE devices because of the following two reasons. First, in the LSSE configu- reaction oxidizer solution ration, the spin current is generated along the columnar structure, and thus less affected by grain scattering. Sec- ond, the columnar grain boundary can work as a stress- relieving cushion when the film is bent, leading to the high flexibility and bending tolerance. substrate Pt/Ni0.2Zn0.3Fe2.5O4/polyimide rotating heating stage The ferrite plating method potentially enables the di- rect coating of thermoelectric functions onto various sur- (c) faces over a large area, expanding the versatility and utility of the LSSE devices. Significantly, the flexible Pt thermoelectric sheet based on the LSSE, demonstrated here, has remarkably low thermal resistance because of the thick-substrate-free structure. Therefore, it may be Ni0.2Zn0.3Fe2.5O4 suitable for heat-flow sensing applications [106].

substrate 500 nm

FIG. 11: (a) A schematic illustration of the ferrite-plating C. Hybrid thermoelectric generation based on spin method. (b) A photograph of a LSSE-based flexible thermo- Seebeck and anomalous Nernst effects electric sheet, where a Pt/Ni0.2Zn0.3Fe2.5O4 film was formed on a 25-µm-thick polyimide flexible substrate. (c) A scanning As shown in Fig. 2, most of the LSSE experiments electron microscope image of a Ni0.2Zn0.3Fe2.5O4 film grown to date have been performed by using Pt as a metal- on a thermally oxidized silicon substrate by the ferrite-plating lic layer. Pt is useful for investigating the physics of the method. LSSE, since it enables sensitive detection of the thermally generated spin currents and the spin-current-related pa- rameters, such as the spin diffusion length and spin Hall angle, of Pt have been well studied [34]. However, Pt costly equipment. As schematically illustrated in Fig. is unsuitable for thermoelectric applications of the LSSE 11(a), the ferrite film can be grown simply by spraying because of its high material cost. Furthermore, even the an aqueous reaction solution and an oxidizer simultane- spin-charge conversion efficiency of Pt is insufficient for ously onto a substrate, mounted on a rotating heating obtaining adequate thermoelectric performance. stage [214, 215]. The thickness of the ferrite film can be Recently, the ISHE has been investigated not only in controlled by the spray volume and time and can be de- paramagnetic materials but also in ferromagnetic mate- pendent on surface conditions of substrates. Importantly, rials by means of the LSSE experiments [56, 58, 61, 70, all the processes of the ferrite plating method can be ◦ 87, 96] (Fig. 2). The main purpose of these studies performed below 100 C, applicable to heat-labile plastic is microscopic understanding of the ISHE in ferromag- substrates. nets. In contrast, we focus on ferromagnetic materials By using the ferrite plating method, we successfully as low-cost replacements for Pt in the LSSE devices. fabricated the flexible thermoelectric sheet based on the Here, as a simple model case, we consider a ferromag- LSSE [see a photograph in Fig. 11(b)]. The flexible netic metal/ferrimagnetic insulator junction system un- LSSE device consists of a ferrimagnetic Ni0.2Zn0.3Fe2.5O4 der a temperature gradient. In this system, the ANE thin film, grown on a flexible polyimide substrate by is induced in the ferromagnetic metal layer in a conven- the ferrite plating, and a Pt film, sputtered on the tional manner. In addition to the ANE contribution, if Ni0.2Zn0.3Fe2.5O4 film. As shown in Fig. 11(b), the the spin current is injected into the ferromagnetic metal Pt/Ni0.2Zn0.3Fe2.5O4/polyimide sheet is highly flexible from the ferrimagnetic insulator via the LSSE and the and bendable. We observed clear LSSE signals in the ISHE appears in the ferromagnetic metal, the hybrid flexible thermoelectric sheet, where the magnitude of the thermoelectric generation based on the combination of LSSE signals is comparable to that in the conventional the LSSE and ANE is realized [Fig. 12(a)]. The pre- rigid LSSE devices [106]. We also demonstrated that the vious studies showed that the direction of EISHE is the ferrite plating method is applicable to not only plastic same as that of EANE in various ferromagnetic materials substrates but also glass substrates; it is useful for con- [56, 58, 61, 70, 87, 96]. Even when the spin Hall angle of structing low-cost LSSE devices. the ferromagnetic metal is smaller than that of Pt, the A noticeable feature of the ferrite film grown by the fer- shortfall of the ISHE voltage can be compensated by the rite plating method is its columnar grain structure [see superposition of the ANE voltage. In fact, the thermo- the cross-sectional scanning electron microscope image of electric output of all-ferromagnetic devices can even be the Ni0.2Zn0.3Fe2.5O4 film in Fig. 11(c)]. Here, the crys- better than that of the conventional Pt-based devices, as tal orientation of Ni0.2Zn0.3Fe2.5O4 is coherently aligned demonstrated below. within each columnar grain [106]. Such columnar struc- In Figs. 12(b) and 12(c), we compare the trans- 18

(a) LSSE + ANE (b) (c) LSSE + ANE ferromagnetic 20 Ni/Bi:YIG-film Pt/Bi:YIG-film metal V LSSE ANE V) 10 μ

EISHE + EANE ( V 0 jS single Ni film M Voltage -10 single Pt film

∇T M -20 ΔT = 7.9 K

ferrimagnetic insulator -400 -200 0 200 400 -10 -5 0 5 10 Magnetic field H (Oe) Temperature difference ΔT (K)

FIG. 12: (a) A schematic illustration of the hybrid thermoelectric generation based on the LSSE and ANE in a ferromagnetic metal/ferrimagnetic insulator junction system. (b) H dependence of V in the Ni/Bi:YIG-film sample, Pt/Bi:YIG-film sample, single Ni film, and single Pt film at ∆T = 7.9 K. (c) ∆T dependence of V in the Ni/Bi:YIG-film sample, Pt/Bi:YIG-film sample, single Ni film, and single Pt film. The thicknesses of the Ni, Pt, and Bi:YIG films are 10 nm, 10 nm, and 120 nm, respectively. All the measurements were performed at room temperature.

verse thermoelectric voltage between ferromagnetic enhancement is unaccompanied by the increase of the in- Ni/Bi:YIG-film and conventional Pt/Bi:YIG-film sys- ternal resistance, the output power also increases with tems. The Bi:YIG layer of the Ni/Bi:YIG-film and increasing n, a situation different from the case of the Pt/Bi:YIG-film samples were formed on a SGGG sub- LSSE-voltage enhancement by the spin Hall thermopile strate by means of the MOD method. We found that [44, 49]. The observed n dependence of the LSSE volt- the Ni/Bi:YIG-film sample exhibits the clear V signal of age in the [P/F] ×n multilayer systems is beyond con- which the magnitude is more than twice greater than ventional expectations based on the situation that the that of the ANE voltage in a plain Ni film directly systems are merely regarded as several independent P/F placed on the substrate, indicating that the signal in the bilayers electrically connected in parallel, where the out- Ni/Bi:YIG-film sample is attributed to the superposition put voltage is not enhanced while the output power is of the ANE in the Ni layer and the ISHE induced by enhanced owing to the reduction of the internal resis- the spin current injected from the Bi:YIG layer. Impor- tance [Fig. 14(a)] [100]. Importantly, this LSSE-voltage tantly, the magnitude of V in the Ni/Bi:YIG-film sample enhancement cannot be explained even when the spin- is even greater than that in the conventional Pt/Bi:YIG- current injection into P from both the top and bottom film sample [Figs. 12(b) and 12(c)]. Since the electri- F layers is taken into account, where the upper limit of cal resistivity of Ni is comparable to or smaller than the LSSE enhancement is twice of the voltage in the sin- that of Pt, the Ni/Bi:YIG-film sample shows the bet- gle P/F bilayer; as shown in Fig. 13(b), the observed ter thermoelectric performance than the Pt/Bi:YIG-film enhancement is much greater than this upper limit. sample in terms of not only the output voltage but also Our current interpretation of the mechanism of the the output current or power. This result suggests that LSSE enhancement in the P/F multilayer systems is sum- the hybrid thermoelectric generation using ferromagnetic marized as follows [100]. The essence of the LSSE en- metal/ferrimagnetic insulator junction systems will be hancement is the boundary conditions for spin currents useful for constructing noble-metal-free, low-cost, and ef- flowing normal to the P/F interfaces, which affect the ficient LSSE devices. magnitude and spatial profile of the spin currents gen- erated by the LSSE. Here we assume the following two boundary conditions: (i) spin currents must disappear D. Spin Seebeck effect in multilayer devices at the top and bottom surfaces of the multilayer systems and (ii) spin currents in the P and F layers are continuous To realize efficient thermal spin-current generation, at the interfaces. Although spin currents in paramagnetic the LSSE has recently been investigated in multi- metals and ferrimagnetic insulators are respectively car- layer systems comprising alternately-stacked paramagnet ried by conduction electrons and spin waves, the bound- (P)/ferromagnet (F) films [88, 89, 100]. The recent stud- ary condition (ii) allows us to treat these spin currents ies have revealed that the LSSE voltage in [P/F] × n in the same manner in the following phenomenological systems significantly and monotonically increases with discussions. Let us now compare a spin-current profile increasing the number of the P/F bilayers n. For ex- in a P1/F1/P2/F2 system with that in a F1/P2/F2 sys- ample, in [100], the magnitude of the LSSE voltage in tem without the top P layer (P1) [Fig. 14(b)], where P1 [Pt/Fe3O4] × 6 systems was observed to be enhanced and P2 are good spin sinks. According to the boundary by a factor of 4–6 compared with that in [Pt/Fe3O4] × condition (i), the spin current is eliminated at the top of 1 bilayer systems (Fig. 13). Since this LSSE-voltage the F1 layer in the F1/P2/F2 system. However, this is 19

(a) [Pt/Fe3O4] × n (n = 6) and boundary value. This phenomenological interpreta- Pt tion is consistent with the experimental results; similar n dependence of the LSSE voltage was observed in the [Pt/Fe3O4] × n systems [compare the inset to Fig. 13(b)

Fe3O4 with that to Fig. 14(c)] [100]. The [Pt/Fe3O4] × 6 sys- tem holds the current record of the LSSE thermopower at room temperature (Fig. 15). Since the enhancement of the LSSE based on this mechanism strongly depends on 50 nm MgO the spin diffusion length (magnon diffusion length) of the P (F) layer, the determination of optimum thicknesses of each layer and optimum P/F material combination (b) 10 n = 6 is crucial for further improvement of the thermoelectric T = 300 K performance of the LSSE devices. n = 3 5 n = 2

) /K VI. CONCLUSIONS AND PROSPECTS

V n = 1 μ 0 0.8 In this article, we reviewed the experimental results V / Δ T ( ) 0.6 K / of the LSSE in insulator-based systems from the view- -5 0.4

S (μ V 0.2 point of thermoelectric applications. The basic structure

Voltage / Temperature difference of the LSSE device is a simple bilayer film comprising a 0 1 2 3 4 5 6 7 magnetic insulator and a metallic or other conductive -10 Number of bilayers n material; an out-of-plane temperature gradient in the -8 -6 -4 -2 0 2 4 6 8 magnetic insulator layer induces an in-plane electric field Magnetic field H (kOe) in the metallic layer via spin-current generation across FIG. 13: (a) Scanning transmission electron microscope im- the interface. Owing to this configuration, the LSSE de- age of the cross section of the [Pt/Fe3O4] × 6 system, where vice exhibits the following unique characteristics. Firstly, the thickness of the Pt (Fe3O4) layers is 17 nm (34 nm). The the figure of merit of the LSSE device is free from the magnified views of the Pt layer and Fe3O4/MgO interface Wiedemann-Franz law, and can be optimized by selecting show the high crystalline quality of the multilayers. (b) H de- the combination of a magnetic insulator with low thermal pendence of V/∆T in the [Pt/Fe3O4] ×n systems for various conductivity and a metallic film with low electrical resis- values of the Pt/Fe3O4-bilayer number n at T = 300 K. The tivity. Secondly, the LSSE device follows the convenient inset to (b) shows the n dependence of S in the [Pt/Fe3O4] ×n scaling law; the maximum extractable power of the LSSE systems. The red dotted line represents the upper limit of the device is proportional to the device area, and can be en- LSSE voltage for the parallel circuit of connected Pt/Fe O 3 4 hanced simply enlarging the device without constructing bilayers. complicated modules. The simple structure and scaling law of the LSSE device make it possible to implement thermoelectric functions onto various heat sources by us- ing versatile and low-cost fabrication processes. not the case for the P1/F1/P2/F2 system; the spin cur- The thermoelectric performance of the LSSE device rent remains a large value at the P1/F1 interface, while cannot be simply compared with that of conventional it must disappear at the top of the P1 layer. As shown thermoelectric devices based on the Seebeck effect be- in Fig. 14(b), this difference results in the enhancement cause of the different device configurations and driving of the spin currents near the P1/F1 interface, a situation principles. The theory presented in Sec. IV shows that consistent with the prediction in [173]. By applying the the thermoelectric conversion efficiency of the LSSE de- above discussion, we calculated out-of-plane spin-current vice is characterized by the figure of merit ZSSET , de- profiles in the [P/F] × n systems for various values of scribed by Eq. (34). Here, ZSSET is defined as pro- n [Fig. 14(c)] and the n dependence of the spin-current portional to the square of the LSSE thermopower and magnitude averaged over all the P layers hjSi [the inset inversely proportional to the thermal conductivity of a to Fig. 14(c)] (note that hjSi can be regarded as an ob- magnetic insulator and electrical resistivity of a metallic servable quantity in the measurements of the LSSE in the film, which is similar to the case of the conventional ther- P/F multilayer systems, since the P layers are electrically moelectric devices. However, the figure of merit of the connected with each other due to the small thickness and LSSE device is defined in the range of 0 ≤ ZSSET ≤ 1, resultant small out-of-plane electrical resistance of the F which is a characteristic of transverse thermoelectric de- layers [100]). Importantly, the magnitude of hjSi mono- vices. tonically increases with increasing n. The physics be- Despite the potential advantages of the LSSE device, hind is that, thanks to the multilayer structure, the spin the current stage of the SSE research is still far from real- current in the P interlayers acquires a new length scale istic thermoelectric applications. This is mainly because 20

(a) (c) Vpar Vpar << Vmul Vmul ∇T

n = 8 paramagnet (P) ferromagnet (F) [P/F] × n

n = 4

∇T ∇T n = 2

(b) ∇T

F2 P2 F1 P1 n = 1

0.4 0.4

n = 8 with P1 0.3 0.3

n = 4 (arb. unit) (arb. unit) S S j j 0.2 without P 0.2 1 n = 2 0.3 0.2

0.1 (arb. unit)

0.1 0.1 > Spin current Spin current S j

n = 1 < 0 2 4 6 8 10 Number of bilayers n

0 20 40 60 80 100 0 100 200 Position along z direction (nm) Position along z direction (nm)

FIG. 14: (a) Comparison of the LSSE voltage between paramagnet (P)/ferromagnet (F) bilayers electrically connected in parallel and an alternately-stacked P/F multilayer film. The LSSE voltage in the latter Vmul was found to be much greater than that in the former Vpar. (b) Comparison of the spin-current jS profiles between P1/F1/P2/F2 and F1/P2/F2 systems. (c) jS profiles calculated for the [P/F] × n multilayer systems for various values of the P/F-bilayer number n. The inset to (c) shows the n dependence of the spin-current magnitude averaged over all the P layers hjSi. The discussion on these calculation results is detailed in [100].

the LSSE thermopower is very small at present, although since even the current record of the thermoelectric perfor- there is plenty of scope for the performance improvement. mance of the LSSE is still inadequate, continuous devel- The LSSE thermopower can be enhanced by improving opment of the LSSE-based thermoelectric technology is the thermal spin-current generation efficiency in the mag- necessary. For realistic thermoelectric applications, the netic insulator, spin Hall angle in the metallic layer, and reduction of the thermal conductivity (electrical resis- spin mixing conductance at the metal/insulator interface. tivity) of the insulator layer (metallic layer) for the im- In fact, owing to the efforts after the discovery of the provement of ZSSET and optimum thermal design for the LSSE, the thermoelectric performance of the LSSE de- stable and continuous operation of the LSSE device are vice is being improved, especially, in recent years with also indispensable. the advent of multilayer systems as shown in the time- In the field of spintronics, in addition to the SSE, series trends of the LSSE thermopower S [Fig. 15(a)] a variety of novel phenomena in which the interplay and the LSSE power factor [Fig. 15(b)] in typical LSSE of spin and heat plays a crucial role were discovered devices. Here, we define the LSSE power factor as the [6, 25]. Some of the thermo-spin phenomena, such as output power normalized by the applied temperature gra- spin-dependent Seebeck effects [216, 217], Seebeck effects dient and by the device area: in magnetic tunnel junctions [218], and magnon-drag ef- S2  V L 2 l V 2 1 1 fects [219], are also potentially applicable to thermoelec- = z y = , (39) R ∆T L Rl R ∇T 2 l l tric generation in nano-structured spintronics devices. s y x x y Furthermore, at low temperatures, giant thermopower where Rs (= Rlx/ly) is the sheet resistance of the metal- was observed in transverse SSE devices comprising non- lic layer of the LSSE device and we assume lx(y) = Lx(y). magnetic semiconductors [220]. Although the origin of Figure 15(b) emphasizes again that multilayer systems this giant thermoelectric effect is different from the SSE are useful for improving the output power of the LSSE discussed in this article, it represents a great potential of device due to the combination of the voltage enhance- spin-based thermoelectric technologies. However, these ment and internal resistance reduction. Nevertheless, thermo-spin phenomena appear only in conductors; the 21

(a) LSSE thermopower voltage per unit temperature difference (b) LSSE power factor square of LSSE thermopower at T = 300 K normalized by device dimensions at T = 300 K normalized by sheet resistance

2 2 S (μV/K) S /Rs (pW/K ) 5 Garnet ferrite 0.2 Spinel ferrite Pt(1.5 nm)/YIG

4

3 Raw data: V/ΔT = 25 μV/ K 0.1 Pt/Fe3O4 Pt(3 nm)/YIG multilayers 2

conductive layer thickness > 5 nm Pt/Fe O 1 3 4 Pt(3 nm)/YIG multilayers Pt(1.5 nm)/YIG

0 year 0 year 2010 2012 2014 2016 2010 2012 2014 2016

2 FIG. 15: Time-series trends of the LSSE thermopower S (a) and the LSSE power factor S /Rs (b) at T = 300 K in the typical LSSE devices [42–44, 46, 49, 51, 53–55, 58, 71, 74, 80, 81, 85, 90, 92, 93, 100, 120]. Although the output voltage can be increased simply by reducing the conductive layer thickness of the LSSE devices, such thin films cannot exhibit high output power due to their high internal resistance (see gray circle data points). For fair comparison, the experimental data for the LSSE devices with the spin Hall thermopiles are not shown in these trends. For example, in [44], the large LSSE voltage of V/∆T ∼ 0.4 mV/K was realized by using the Pt-Au thermopile structure, although its internal resistance is very high.

utilization of insulators is an exclusive feature of the SSE- A. Miura, L. Morell´on,T. Murakami, R. C. Myers, Y. based thermoelectric generation. We anticipate that this Nakamura, T. Niizeki, T. Nonaka, J. Ohe, Y. Ohnuma, unique feature of the SSE will lead to various thermo- Y. Oikawa, T. Ota, T. Oyake, Z. Qiu, R. Ramos, S. electric applications, including energy harvesters, ther- M. Rezende, K. Sato, T. Seki, J. Shiomi, Y. Shiomi, H. mometers, infrared sensors, position detectors [47], and Someya, S. Takahashi, K. Takanashi, Y. Tserkovnyak, B. user-interface devices. J. van Wees, J. Xiao, N. Yamamoto, and A. Yagmur for valuable discussions.

ACKNOWLEDGMENTS This work was supported by PRESTO “Phase Inter- faces for Highly Efficient Energy Utilization” from JST, The authors thank M. H. Aguirre, M. Akimoto, P. A. Japan, Grant-in-Aid for Scientific Research (A) (No. Algarabel, A. Anad´on,J. Barker, G. E. W. Bauer, S. 15H02012), Grant-in-Aid for Challenging Exploratory R. Boona, C. L. Chien, S. Daimon, S. T. B. Goennen- Research (No. 26600067), Grant-in-Aid for Scientific Re- wein, J. P. Heremans, B. Hillebrands, T. Hioki, D. Hou, search on Innovative Area, “Nano Spin Conversion Sci- S. Y. Huang, M. R. Ibarra, J. Ieda, R. Iguchi, K. Ihara, ence” (No. 26103005) from MEXT, Japan, and NEC Y. Iwasaki, H. Jin, X. F. Jin, D. Kikuchi, M. Kl¨aui,S. Corporation. T.K. is supported by JSPS through a re- Kohmoto, K. Kondo, I. Lucas, T. Manako, A. Matsuba, search fellowship for young scientists (No. 15J08026).

[1] F. J. DiSalvo, “Thermoelectric cooling and power gen- vol. 4, no. 87, pp. 46860-46874, Sep. 2014. eration,” Science, vol. 285, no. 5428, pp. 703-706, Jul. [4] D. M. Rowe ed., CRC Handbook of Thermoelectrics. 1999. CRC Press, 1995. [2] L. E. Bell, “Cooling, heating, generating power, and re- [5] H. J. Goldsmid, Introduction to Thermoelectricity. covering waste heat with thermoelectric systems,” Sci- Springer, 2009. ence, vol. 321, no. 5895, pp. 1457-1461, Sep. 2008. [6] S. R. Boona, R. C. Myers, and J. P. Heremans, “Spin [3] P. Sundarraj, D. Maity, S. S. Roy, and R. A. Taylor, caloritronics,” Energy Environ. Sci., vol. 7, no. 3, pp. “Recent advances in thermoelectric materials and solar 885-910, Mar. 2014. thermoelectric generators - a critical review,” RSC Adv., [7] L.-D. Zhao, S.-H. Lo, Y. Zhang, H. Sun, G. Tan, 22

C. Uher, C. Wolverton, V. P. Dravid, and M. G. and opportunities in two-dimensional materials beyond Kanatzidis, “Ultralow thermal conductivity and high graphene,” ACS NANO, vol. 7, no. 4, pp. 2898-2926, thermoelectric figure of merit in SnSe crystals,” Nature, Apr. 2013. vol. 508, no. 7496, pp. 373-377, Apr. 2014. [22] R. Fei, A. Faghaninia, R. Soklaski, J.-A. Yan, C. Lo, [8] J. P. Heremans, “Thermoelectricity: The ugly duck- and L. Yang, “Enhanced thermoelectric efficiency via ling,” Nature, vol. 508, no. 7496, pp. 327-328, Apr. 2014. orthogonal electrical and thermal conductances in phos- [9] L. D. Hicks and M. S. Dresselhaus, “Effect of quantum- phorene,” Nano Lett., vol. 14, no. 11, pp. 6393-6399, well structures on the thermoelectric figure of merit,” Nov. 2014. Phys. Rev. B, vol. 47, no. 19, pp. 12727-12731, May [23] S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. 1993. M. Daughton, S. von Moln´ar, M. L. Roukes, A. Y. [10] A. I. Hochbaum, R. Chen, R. D. Delgado, W. Liang, E. Chtchelkanova, and D. M. Treger, “Spintronics: A spin- C. Garnett, M. Najarian, A. Majumdar, and P. Yang, based electronics vision for the future,” Science, vol. “Enhanced thermoelectric performance of rough silicon 294, no. 5546, pp. 1488-1495, Nov. 2001. nanowires,” Nature, vol. 451, no. 7175, pp. 163-167, Jan. [24]I. Zuti´c,ˇ J. Fabian, and S. Das Sarma, “Spintronics: 2008. Fundamentals and applications,” Rev. Mod. Phys., vol. [11] B. Poudel, Q. Hao, Y. Ma, Y. Lan, A. Minnich, B. Yu, 76, no. 2, pp. 323-410, Apr. 2004. X. Yan, D. Wang, A. Muto, D. Vashaee, X. Chen, J. [25] G. E. W. Bauer, E. Saitoh, and B. J. van Wees, “Spin Liu, M. S. Dresselhaus, G. Chen, and Z. Ren, “High- caloritronics,” Nature Mater., vol. 11, no. 5, pp. 391- thermoelectric performance of nanostructured bismuth 399, May 2012. antimony telluride bulk alloys,” Science, vol. 320, no. [26] K. Uchida, S. Takahashi, K. Harii, J. Ieda, W. Koshibae, 5876, pp. 634-638, May 2008. K. Ando, S. Maekawa, and E. Saitoh, “Observation of [12] A. J. Minnich, M. S. Dresselhaus, Z. F. Ren, and G. the spin Seebeck effect,” Nature, vol. 455, no. 7214, pp. Chen, “Bulk nanostructured thermoelectric materials: 778-781, Oct. 2008. current research and future prospects,” Energy Environ. [27] S. Maekawa, E. Saitoh, S. O. Valenzuela, and T. Kimura Sci., vol. 2, no. 5, pp. 466-479, May 2009. eds., Spin Current. Oxford University Press, Oxford, [13] J.-F. Li, W.-S. Liu, L.-D. Zhao, and M. Zhou, “High- 2012. performance nanostructured thermoelectric materials,” [28] S. Maekawa, H. Adachi, K. Uchida, J. Ieda, and E. NPG Asia Mater., vol. 2, no. 4, pp. 152-158, Oct. 2010. Saitoh, “Spin current: Experimental and theoretical [14] P. E. Hopkins, C. M. Reinke, M. F. Su, R. H. Olsson, III, aspects,” J. Phys. Soc. Japan, vol. 82, no. 10, pp. E. A. Shaner, Z. C. Leseman, J. R. Serrano, L. M. Phin- 102002/1-102002/23, Oct. 2013. ney, and I. El-Kady, “Reduction in the thermal conduc- [29] A. Azevedo, L. H. Vilela-Le˜ao,R. L. Rodr´ıguez-Su´arez, tivity of single crystalline silicon by phononic crystal A. B. Oliveira, and S. M. Rezende, “dc effect in fer- patterning,” Nano Lett., vol. 11, no. 1, pp. 107-112, Jan romagnetic resonance: Evidence of the spin-pumping 2011. effect?,” J. Appl. Phys., vol. 97, no. 10, pp. 10C715/1- [15] M. Kashiwagi, S. Hirata, K. Harada, Y. Zheng, K. 10C715/3, May 2005. Miyazaki, M. Yahiro, and C. Adachi, “Enhanced figure [30] E. Saitoh, M. Ueda, H. Miyajima, and G. Tatara, “Con- of merit of a porous thin film of bismuth antimony tel- version of spin current into charge current at room tem- luride,” Appl. Phys. Lett., vol. 98, no. 2, pp. 023114/1- perature: Inverse spin-Hall effect,” Appl. Phys. Lett., 023114/3, Jan. 2011. vol. 88, no. 18, pp. 182509/1-182509/3, May 2006. [16] J. Maire and M. Nomura, “Reduced thermal conduc- [31] S. O. Valenzuela and M. Tinkham, “Direct electronic tivities of Si one-dimensional periodic structure and measurement of the spin Hall effect,” Nature, vol. 442, nanowire,” Jpn. J. Appl. Phys., vol. 53, no. 6S, pp. no. 7099, pp. 176-179, Jul. 2006. 06JE09/1-06JE09/4, Jun. 2014. [32] M. V. Costache, M. Sladkov, S. M. Watts, C. H. van [17] A. Miura, S. Zhou, T. Nozaki, and J. Shiomi, der Wal, and B. J. van Wees, “Electrical detection of “Crystalline-amorphous silicon nanocomposites with re- spin pumping due to the precessing magnetization of a duced thermal conductivity for bulk thermoelectrics,” single ferromagnet,” Phys. Rev. Lett., vol. 97, no. 21, ACS Appl. Mater. Interfaces, vol. 7, no. 24, pp. 13484- pp. 216603/1-216603/4, Nov. 2006. 13489, Jun. 2015. [33] T. Kimura, Y. Otani, T. Sato, S. Takahashi, and S. [18] D. T. Morelli, J. Heremans, M. Sakamoto, and C. Uher, Maekawa, “Room-temperature reversible spin Hall ef- “Anisotropic heat conduction in diacetylenes,” Phys. fect,” Phys. Rev. Lett., vol. 98, no. 15, pp. 156601/1- Rev. Lett., vol. 57, no. 7, pp. 869-872, Aug. 1986. 156601/4, Apr. 2007. [19] A. A. Balandin, “Thermal properties of graphene and [34] J. Sinova, S. O. Valenzuela, J. Wunderlich, C. H. Back, nanostructured carbon materials,” Nature Mater., vol. and T. Jungwirth, “Spin Hall effects,” Rev. Mod. Phys., 10, no. 8, pp. 569-581, Aug. 2011. vol. 87, no. 4, pp. 1213-1259, Oct. 2015. [20] J. L. Cohn, B. D. White, C. A. M. dos Santos, and J. [35] K. Uchida, T. Ota, K. Harii, K. Ando, H. Nakayama, J. Neumeier, “Giant Nernst effect and bipolarity in the and E. Saitoh, “Electric detection of the spin-Seebeck quasi-one-dimensional metal Li0.9Mo6O17,” Phys. Rev. effect in ferromagnetic metals,” J. Appl. Phys., vol. 107, Lett., vol. 108, no. 5, pp. 056604/1-056604/5, Feb. 2012. no. 9, pp. 09A951/1-09A951/5, May 2010. [21] S. Z. Butler, S. M. Hollen, L. Cao, Y. Cui, J. A. Gupta, [36] S. Bosu, Y. Sakuraba, K. Uchida, K. Saito, T. Ota, E. H. R. Guti´errez,T. F. Heinz, S. S. Hong, J. Huang, Saitoh, and K. Takanashi, “Spin Seebeck effect in thin A. F. Ismach, E. Johnston-Halperin, M. Kuno, V. V. films of the Heusler compound Co2MnSi,” Phys. Rev. Plashnitsa, R. D. Robinson, R. S. Ruoff, S. Salahud- B, vol. 83, no. 22, pp. 224401/1-22440/6, Jun. 2011. din, J. Shan, L. Shi, M. G. Spencer, M. Terrones, W. [37] K. Uchida, H. Adachi, T. An, T. Ota, M. Toda, B. Windl, and J. E. Goldberger, “Progress, challenges, Hillebrands, S. Maekawa, and E. Saitoh, “Long-range 23

spin Seebeck effect and acoustic spin pumping,” Nature Uchida, V. I. Vasyuchka, A. V. Chumak, A. A. Serga, Mater., vol. 10, no. 10, pp. 737-741, Oct. 2011. E. Saitoh, and B. Hillebrands, “Heat-induced damping [38] S. H. Wang, L. K. Zou, J. W. Cai, B. G. Shen, and J. modification in yttrium iron garnet/platinum hetero- R. Sun, “Transverse thermoelectric effects in platinum structures,” Appl. Phys. Lett., vol. 102, no. 6, pp. strips on permalloy films,” Phys. Rev. B, vol. 88, no. 21, 062417/1-062417/4, Feb. 2013. pp. 214304/1-214304/6, Dec. 2013. [53] D. Meier, T. Kuschel, L. Shen, A. Gupta, T. Kikkawa, [39] C. M. Jaworski, J. Yang, S. Mack, D. D. Awschalom, K. Uchida, E. Saitoh, J.-M. Schmalhorst, and G. Reiss, J. P. Heremans, and R. C. Myers, “Observation of the “Thermally driven spin and charge currents in thin spin-Seebeck effect in a ferromagnetic semiconductor,” NiFe2O4/Pt films,” Phys. Rev. B, vol. 87, no. 5, pp. Nature Mater., vol. 9, no. 11, pp. 898-903, Nov. 2010. 054421/1-054421/5, Feb. 2013. [40] C. M. Jaworski, J. Yang, S. Mack, D. D. Awschalom, R. [54] R. Ramos , T. Kikkawa, K. Uchida, H. Adachi, I. Lucas, C. Myers, and J. P. Heremans, “Spin-Seebeck effect: A M. H. Aguirre, P. Algarabel, L. Morell´on,S. Maekawa, phonon driven spin distribution,” Phys. Rev. Lett., vol. E. Saitoh, and M. R. Ibarra, “Observation of the spin 106, no. 18, pp. 186601/1-186601/4, May 2011. Seebeck effect in epitaxial Fe3O4 thin films,” Appl. [41] K. Uchida, J. Xiao, H. Adachi, J. Ohe, S. Takahashi, J. Phys. Lett., vol. 102, no. 7, pp. 072413/1-072413/5, Feb. Ieda, T. Ota, Y. Kajiwara, H. Umezawa, H. Kawai, G. 2013. E. W. Bauer, S. Maekawa, and E. Saitoh, “Spin Seebeck [55] K. Uchida, T. Nonaka, T. Kikkawa, Y. Kajiwara, and E. insulator,” Nature Mater., vol. 9, no. 11, pp. 894-897, Saitoh, “Longitudinal spin Seebeck effect in various gar- Nov. 2010. net ferrites,” Phys. Rev. B, vol. 87, no. 10, pp. 104412/1- [42] K. Uchida, H. Adachi, T. Ota, H. Nakayama, S. 104412/6, Mar. 2013. Maekawa, and E. Saitoh, “Observation of longitudinal [56] B. F. Miao, S. Y. Huang, D. Qu, and C. L. Chien, “In- spin-Seebeck effect in magnetic insulators,” Appl. Phys. verse spin Hall effect in a ferromagnetic metal,” Phys. Lett., vol. 97, no. 17, pp. 172505/1-172505/3, Oct. 2010. Rev. Lett., vol. 111, no. 6, pp. 066602/1-066602/5, Aug. [43] A. Kirihara, K. Uchida, Y. Kajiwara, M. Ishida, Y. 2013. Nakamura, T. Manako, E. Saitoh, and S. Yorozu, “Spin- [57] M. Schreier, A. Kamra, M. Weiler, J. Xiao, G. E. W. current-driven thermoelectric coating,” Nature Mater., Bauer, R. Gross, and S. T. B. Goennenwein, “Magnon, vol. 11, no. 8, pp. 686-689, Aug. 2012. phonon, and electron temperature profiles and the spin [44] K. Uchida, M. Ishida, T. Kikkawa, A. Kirihara, T. Seebeck effect in magnetic insulator/normal metal hy- Murakami, and E. Saitoh, “Longitudinal spin See- brid structures,” Phys. Rev. B, vol. 88, no. 9, pp. beck effect: from fundamentals to applications,” J. 094410/1-094410/13, Sep. 2013. Phys.: Condens. Matter, vol. 26, no. 34, pp. 343202/1- [58] T. Kikkawa, K. Uchida, S. Daimon, Y. Shiomi, H. 343202/15, Aug. 2014. Adachi, Z. Qiu, D. Hou, X.-F. Jin, S. Maekawa, and E. [45] K. Uchida, T. Ota, H. Adachi, J. Xiao, T. Nonaka, Y. Saitoh, “Separation of longitudinal spin Seebeck effect Kajiwara, G. E. W. Bauer, S. Maekawa, and E. Saitoh, from anomalous Nernst effect: Determination of origin “Thermal spin pumping and magnon-phonon-mediated of transverse thermoelectric voltage in metal/insulator spin-Seebeck effect,” J. Appl. Phys., vol. 111, no. 10, junctions,” Phys. Rev. B, vol. 88, no. 21, pp. 214403/1- pp. 103903/1-103903/12, May 2012. 214403/11, Dec. 2013. [46] K. Uchida, T. Nonaka, T. Ota, and E. Saitoh, “Lon- [59] M. Schreier, N. Roschewsky, E. Dobler, S. Meyer, H. gitudinal spin-Seebeck effect in sintered polycrystalline Huebl, R. Gross, and S. T. B. Goennenwein, “Current (Mn, Zn)Fe2O4,” Appl. Phys. Lett., vol. 97, no. 26, pp. heating induced spin Seebeck effect,” Appl. Phys. Lett., 262504/1-262504/3, Dec. 2010. vol. 103, no. 24, pp. 242404/1-242404/4, Dec. 2013. [47] K. Uchida, A. Kirihara, M. Ishida, R. Takahashi, and [60] S. M. Rezende, R. L. Rodr´ıguez-Su´arez,R. O. Cunha, E. Saitoh, “Local spin-Seebeck effect enabling two- A. R. Rodrigues, F. L. A. Machado, G. A. Fonseca dimensional position sensing,” Jpn. J. Appl. Phys., vol. Guerra, J. C. Lopez Ortiz, and A. Azevedo, “Magnon 50, no. 12R, pp. 120211/1-120211/3, Dec. 2011. spin-current theory for the longitudinal spin-Seebeck [48] M. Weiler, M. Althammer, F. D. Czeschka, H. Huebl, M. effect,” Phys. Rev. B, vol. 89, no. 1, pp. 014416/1- S. Wagner, M. Opel, I.-M. Imort, G. Reiss, A. Thomas, 014416/10, Jan. 2014. R. Gross, and S. T. B. Goennenwein, “Local charge [61] A. Azevedo, O. Alves Santos, G. A. Fonseca Guerra, and spin currents in magnetothermal landscapes,” Phys. R. O. Cunha, R. Rodr´ıguez-Su´arez,and S. M. Rezende, Rev. Lett., vol. 108, no. 10, pp. 106602/1-106602/5, “Competing spin pumping effects in magnetic hybrid Mar. 2012. structures,” Appl. Phys. Lett., vol. 104, no. 5, pp. [49] K. Uchida, T. Nonaka, T. Yoshino, T. Kikkawa, D. 052402/1-052402/5, Feb. 2014. Kikuchi, and E. Saitoh, “Enhancement of spin-Seebeck [62] G. Siegel, M. C. Prestgard, S. Teng, and A. Tiwari, voltage by spin-Hall thermopile,” Appl. Phys. Express, “Robust longitudinal spin-Seebeck effect in Bi-YIG thin vol. 5, no. 9, pp. 093001/1-093001/3, Sep. 2012. films,” Sci. Rep., vol. 4, pp. 4429/1-4429/6, Mar. 2014. [50] D. Qu, S. Y. Huang, J. Hu, R. Wu, and C. L. Chien, [63] J. B. S. Mendes, R. O. Cunha, O. Alves Santos, P. R. T. “Intrinsic spin Seebeck effect in Au/YIG,” Phys. Rev. Ribeiro, F. L. A. Machado, R. L. Rodr´ıguez-Su´arez,A. Lett., vol. 110, no. 6, pp. 067206/1-067206/5, Feb. 2013. Azevedo, and S. M. Rezende, “Large inverse spin Hall [51] T. Kikkawa, K. Uchida, Y. Shiomi, Z. Qiu, D. Hou, D. effect in the antiferromagnetic metal Ir20Mn80,” Phys. Tian, H. Nakayama, X.-F. Jin, and E. Saitoh, “Longitu- Rev. B, vol. 89, no. 14, pp. 140406(R)/1-140406(R)/5, dinal spin Seebeck effect free from the proximity Nernst Apr. 2014. effect,” Phys. Rev. Lett., vol. 110, no. 6, pp. 067207/1- [64] D. Qu, S. Y. Huang, B. F. Miao, S. X. Huang, and C. L. 067207/5, Feb. 2013. Chien, “Self-consistent determination of spin Hall angles [52] M. B. Jungfleisch, T. An, K. Ando, Y. Kajiwara, K. in selected 5d metals by thermal spin injection,” Phys. 24

Rev. B, vol. 89, no. 14, pp. 140407(R)/1-140407(R)/5, Phys. Lett., vol. 105, no. 24, pp. 242412/1-242412/4, Apr. 2014. Dec. 2014. [65] A. Kehlberger, G. Jakob, M. C. Onbasli, D. H. Kim, C. [78] Y. Shiomi and E. Saitoh, “Paramagnetic spin pump- A. Ross, and M. Kl¨aui,“Investigation of the magnetic ing,” Phys. Rev. Lett., vol. 113, no. 26, pp. 266602/1- properties of insulating thin films using the longitudinal 266602/5, Dec. 2014. spin Seebeck effect,” J. Appl. Phys., vol. 115, no. 17, pp. [79] M. Schreier, G. E. W. Bauer, V. I. Vasyuchka, J. Flipse, 17C731/1-17C731/3, May 2014. K. Uchida, J. Lotze, V. Lauer, A. V. Chumak, A. A. [66] N. Roschewsky, M. Schreier, A. Kamra, F. Schade, K. Serga, S. Daimon, T. Kikkawa, E. Saitoh, B. J. van Ganzhorn, S. Meyer, H. Huebl, S. Gepr¨ags,R. Gross, Wees, B. Hillebrands, R. Gross, and S. T. B. Goen- and S. T. B. Goennenwein, “Time resolved spin Seebeck nenwein, “Sign of inverse spin Hall voltages generated effect experiments,” Appl. Phys. Lett., vol. 104, no. 20, by ferromagnetic resonance and temperature gradients pp. 202410/1-202410/4, May 2014. in yttrium iron garnet platinum bilayers,” J. Phys. D: [67] M. Agrawal, V. I. Vasyuchka, A. A. Serga, A. Kirihara, Appl. Phys., vol. 48, no. 2, 025001/1-025001/5, Jan. P. Pirro, T. Langner, M. B. Jungfleisch, A. V. Chu- 2015. mak, E. Th. Papaioannou, and B. Hillebrands, “Role of [80] D. Kikuchi, M. Ishida, K. Uchida, Z. Qiu, T. Murakami, bulk-magnon transport in the temporal evolution of the and E. Saitoh, “Enhancement of spin-Seebeck effect by longitudinal spin-Seebeck effect,” Phys. Rev. B, vol. 89, inserting ultra-thin Fe70Cu30 interlayer,” Appl. Phys. no. 22, pp. 224414/1-224414/5, Jun. 2014. Lett., vol. 106, no. 8, pp. 082401/1-082401/4, Feb. 2015. [68] M. Agrawal, A. A. Serga, V. Lauer, E. Th. Papaioan- [81] Z. Qiu, D. Hou, K. Uchida, and E. Saitoh, “Influence nou, B. Hillebrands, and V. I. Vasyuchka, “Microwave- of interface condition on spin-Seebeck effects,” J. Phys. induced spin currents in ferromagnetic-insulator|normal D: Appl. Phys., vol. 48, no. 16, pp. 164013/1-164013/5, metal bilayer system,” Appl. Phys. Lett., vol. 105, no. Apr. 2015. 9, pp. 092404/1-092404/4, Sep. 2014. [82] S. M. Wu, F. Y. Fradin, J. Hoffman, A. Hoffmann, and [69] Y. Saiga, K. Mizunuma, Y. Kono, J. C. Ryu, H. Ono, A. Bhattacharya, “Spin Seebeck devices using local on- M. Kohda, and E. Okuno, “Platinum thickness depen- chip heating,” J. Appl. Phys., vol. 117, pp. 17C509/1- dence and annealing effect of the spin-Seebeck voltage in 17C509/4, May 2015. platinum/yttrium iron garnet structures,” Appl. Phys. [83] A. Sola, M. Kuepferling, V. Basso, M. Pasquale, T. Express, vol. 7, no. 9, pp. 093001/1-093001/4, Sep. 2014. Kikkawa, K. Uchida, and E. Saitoh, “Evaluation of [70] S. M. Wu, J. Hoffman, J. E. Pearson, and A. Bhat- thermal gradients in longitudinal spin Seebeck effect tacharya, “Unambiguous separation of the inverse spin measurements,” J. Appl. Phys., vol. 117, no. 17, pp. Hall and anomalous Nernst effects within a ferromag- 17C510/1-17C510/4, May 2015. netic metal using the spin Seebeck effect,” Appl. Phys. [84] H. Asada, A. Kuwahara, N. Sakata, T. Ono, T. Lett., vol. 105, no. 9, pp. 092409/1-092409/4, Sep. 2014. Ishibashi, A. Meguro, T. Hashinaka, K. Kishimoto, [71] J. Lustikova, Y. Shiomi, Z. Qiu, T. Kikkawa, R. Iguchi, and T. Koyanagi, “Longitudinal spin Seebeck effect K. Uchida, and E. Saitoh, “Spin current generation from in Nd2BiFe5−xGaxO12 prepared on gadolinium gal- sputtered Y3Fe5O12 films,” J. Appl. Phys., vol. 116, no. lium garnet (001) by metal organic decomposition 15, pp. 153902/1-153902/6, Oct. 2014. method,” J. Appl. Phys., vol. 117, no. 17, pp. 17C724/1- [72] A. Aqeel, I. J. Vera-Marun, B. J. van Wees, and T. T. M. 17C724/3, May 2015. Palstra, “Surface sensitivity of the spin Seebeck effect,” [85] T. Niizeki, T. Kikkawa, K. Uchida, M. Oka, K. Z. J. Appl. Phys., vol. 116, no. 15, pp. 153705/1-153705/5, Suzuki, H. Yanagihara, E. Kita, and E. Saitoh, “Ob- Oct. 2014. servation of longitudinal spin-Seebeck effect in cobalt- [73] W. X. Wang, S. H. Wang, L. K. Zou, J. W. Cai, Z. G. ferrite epitaxial thin films,” AIP Adv., vol. 5, no. 5, pp. Sun, and J. R. Sun, “Joule heating-induced coexisted 053603/1-053603/6, May 2015. spin Seebeck effect and spin Hall magnetoresistance in [86] S. M. Wu, J. E. Pearson, and A. Bhattacharya, “Param- the platinum/Y3Fe5O12 structure,” Appl. Phys. Lett., agnetic spin Seebeck effect,” Phys. Rev. Lett., vol. 114, vol. 105, no. 18, pp. 182403/1-182403/5, Nov. 2014. no. 18, pp. 186602/1-186602/5, May 2015. [74] K. Uchida, T. Kikkawa, A. Miura, J. Shiomi, and E. [87] D. Tian, Y. Li, D. Qu, X.-F. Jin, and C. L. Chien, “Sep- Saitoh, “Quantitative temperature dependence of longi- aration of spin Seebeck effect and anomalous Nernst ef- tudinal spin Seebeck effect at high temperatures,” Phys. fect in Co/Cu/YIG,” Appl. Phys. Lett, vol. 106, no. 21, Rev. X, vol. 4, no. 4, pp. 041023/1-041023/9, Nov. 2014. pp. 212407/1-212407/4, May 2015. [75] N. Vlietstra, J. Shan, B. J. van Wees, M. Isasa, F. [88] K.-D. Lee, D.-J. Kim, H. Y. Lee, S.-H. Kim, J.-H. Lee, Casanova, and J. Ben Youssef, “Simultaneous detec- K.-M. Lee, J.-R. Jeong, K.-S. Lee, H.-S. Song, J.-W. tion of the spin-Hall magnetoresistance and the spin- Sohn, S.-C. Shin, and B.-G. Park, “Thermoelectric sig- Seebeck effect in platinum and tantalum on yttrium iron nal enhancement by reconciling the spin Seebeck and garnet,” Phys. Rev. B, vol. 90, no. 17, pp. 174436/1- anomalous Nernst effects in ferromagnet/non-magnet 174436/8, Nov. 2014. multilayers,” Sci. Rep., vol. 5, pp. 10249/1-10249/10, [76] Y. Xu, B. Yang, C. Tang, Z. Jiang, M. Schneider, R. May 2015. Whig, and J. Shi, “Heat-driven spin transport in a fer- [89] Y. Shiomi, Y. Handa, T. Kikkawa, and E. romagnetic metal,” Appl. Phys. Lett., vol. 105, no. 24, Saitoh, “Transverse thermoelectric effect in pp. 242404/1-242404/4, Dec. 2014. La0.67Sr0.33MnO3|SrRuO3 superlattices,” Appl. Phys. [77] P. Li, D. Ellsworth, H. Chang, P. Janantha, D. Richard- Lett., vol. 106, no. 23, pp. 232403/1-232403/4, Jun. son, F. Shah, P. Phillips, T. Vijayasarathy, and M. Wu, 2015. “Generation of pure spin currents via spin Seebeck ef- [90] K. Uchida, J. Ohe, T. Kikkawa, S. Daimon, D. Hou, fect in self-biased hexagonal ferrite thin films,” Appl. Z. Qiu, and E. Saitoh, “Intrinsic surface magnetic 25

anisotropy in Y3Fe5O12 as the origin of low-magnetic- Feb. 2016. field behavior of the spin Seebeck effect,” Phys. Rev. B, [103] E.-J. Guo, A. Kehlberger, J. Cramer, G. Jakob, vol. 92, no. 1, pp. 014415/1-014415/8, Jul. 2015. and M. Kl¨aui, “Origin of the thickness-dependent [91] D. Qu, S. Y. Huang, and C. L. Chien, “Inverse spin low-temperature enhancement of spin Seebeck Hall effect in Cr: Independence of antiferromagnetic or- effect in YIG films,” 2015. [Online]. Available: dering,” Phys. Rev. B, vol. 92, no. 2, pp. 020418(R)/1- http://arxiv.org/abs/1506.06037 020418(R)/4, Jul. 2015. [104] S. Seki, T. Ideue, M. Kubota, Y. Kozuka, R. Takagi, M. [92] Z. Qiu, D. Hou, T. Kikkawa, K. Uchida, and E. Saitoh, Nakamura, Y. Kaneko, M. Kawasaki, and Y. Tokura, “All-oxide spin Seebeck effects,” Appl. Phys. Express, “Thermal generation of spin current in an antiferromag- vol. 8, no. 8, pp. 083001/1-083001/3, Aug. 2015. net,” Phys. Rev. Lett., vol. 115, no. 26, pp. 266601/1- [93] T. Kikkawa, K. Uchida, S. Daimon, Z. Qiu, Y. Shiomi, 266601/5, Dec. 2015. and E. Saitoh, “Critical suppression of spin Seebeck ef- [105] S. M. Wu, W. Zhang, A. KC, P. Borisov, J. E. Pear- fect by magnetic fields,” Phys. Rev. B, vol. 92, no. 6, son, J. S. Jiang, D. Lederman, A. Hoffmann, and A. pp. 064413/1-064413/9, Aug. 2015. Bhattacharya, “Antiferromagnetic spin Seebeck effect,” [94] H. Jin, S. R. Boona, Z. Yang, R. C. Myers, and J. P. Phys. Rev. Lett., vol. 116, no. 9, pp. 097204/1-097204/5, Heremans, “Effect of the magnon dispersion on the lon- Mar. 2016. gitudinal spin Seebeck effect in yttrium iron garnets,” [106] A. Kirihara, K. Kondo, M. Ishida, K. Ihara, Y. Iwasaki, Phys. Rev. B, vol. 92, no. 5, pp. 054436/1-054436/8, H. Someya, A. Matsuba, K. Uchida, E. Saitoh, N. Aug. 2015. Yamamoto, S. Kohmoto, and T. Murakami, “Flexible [95] A. Kehlberger, U. Ritzmann, D. Hinzke, E.-J. Guo, J. heat-flow sensing sheets based on the longitudinal spin Cramer, G. Jakob, M. C. Onbasli, D. H. Kim, C. A. Seebeck effect using one-dimensional spin-current con- Ross, M. B. Jungfleisch, B. Hillebrands, U. Nowak, and ducting films,” Sci. Rep., vol. 6, pp. 23114/1-23114/7, M. Kl¨aui, “Length scale of the spin Seebeck effect,” Mar. 2016. Phys. Rev. Lett., vol. 115, no. 9, pp. 096602/1-096602/5, [107] E.-J. Guo, A. Herklotz, A. Kehlberger, J. Cramer, G. Aug. 2015. Jakob, and M. Kl¨aui,“Thermal generation of spin cur- [96] T. Seki, K. Uchida, T. Kikkawa, Z. Qiu, E. Saitoh, rent in epitaxial CoFe2O4 thin films,” Appl. Phys. Lett., and K. Takanashi, “Observation of inverse spin Hall vol. 108, no. 2, pp. 022403/1-022403/5, Jan. 2016. effect in ferromagnetic FePt alloys using spin Seebeck [108] L. K. Zou, S. H. Wang, Y. Zhang, J. R. Sun, J. W. effect,” Appl. Phys. Lett., vol. 107, no. 9, pp. 092401/1- Cai, and S. S. Kang, “Large extrinsic spin Hall effect in 092401/4, Aug. 2015. Au-Cu alloys by extensive atomic disorder scattering,” [97] S. Wang, L. Zou, X. Zhang, J. Cai, S. Wang, B. Shen, Phys. Rev. B, vol. 93, no. 1, pp. 014422/1-014422/6, and J. Sun, “Spin Seebeck effect and spin Hall magne- Jan. 2016. toresistance at high temperatures for a Pt/yttrium iron [109] B. F. Miao, S. Y. Huang, D. Qu, and C. L. Chien, garnet hybrid structure,” Nanoscale, vol. 7, no. 42, pp. “Absence of anomalous Nernst effect in spin Seebeck 17812-17819, Nov. 2015. effect of Pt/YIG,” AIP Adv., vol. 6, no. 1, pp. 015018/1- [98] U. Ritzmann, D. Hinzke, A. Kehlberger, E.-J. Guo, M. 015018/6, Jan. 2016. Kl¨aui,and U. Nowak, “Magnetic field control of the [110] R. Takagi, Y. Tokunaga, T. Ideue, Y. Taguchi, Y. spin Seebeck effect,” Phys. Rev. B, vol. 92, no. 17, pp. Tokura, and S. Seki, “Thermal generation of spin cur- 174411/1-174411/5, Nov. 2015. rent in a multiferroic helimagnet,” APL Mater., vol. 4, [99] A. Aqeel, N. Vlietstra, J. A. Heuver, G. E. W. Bauer, B. no. 3, pp. 032502/1-032502/7, Mar. 2016. Noheda, B. J. van Wees, and T. T. M. Palstra, “Spin- [111] S. Y. Huang, W. G. Wang, S. F. Lee, J. Kwo, and Hall magnetoresistance and spin Seebeck effect in spin- C. L. Chien, “Intrinsic spin-dependent thermal trans- spiral and paramagnetic phases of multiferroic CoCr2O4 port,” Phys. Rev. Lett., vol. 107, no. 21, pp. 216604/1- films,” Phys. Rev. B, vol. 92, no. 22, pp. 224410/1- 216604/4, Nov. 2011. 224410/8, Dec. 2015. [112] S. Bosu, Y. Sakuraba, K. Uchida, K. Saito, W. [100] R. Ramos, T. Kikkawa, M. H. Aguirre, I. Lucas, A. Kobayashi, E. Saitoh, and K. Takanashi, “Thermal ar- Anad´on,T. Oyake, K. Uchida, H. Adachi, J. Shiomi, tifact on the spin Seebeck effect in metallic thin films P. A. Algarabel, L. Morell´on,S. Maekawa, E. Saitoh, deposited on MgO substrates,” J. Appl. Phys., vol. 111, and M. R. Ibarra, “Unconventional scaling and signif- no. 7, pp. 07B106/1-07B106/3, Apr. 2012. icant enhancement of the spin Seebeck effect in multi- [113] D. Meier, D. Reinhardt, M. van Straaten, C. Klewe, layers,” Phys. Rev. B, vol. 92, no. 22, pp. 220407(R)/1- M. Althammer, M. Schreier, S. T. B. Goennenwein, A. 220407(R)/5, Dec. 2015. Gupta, M. Schmid, C. H. Back, J.-M. Schmalhorst, T. [101] A. Kirihara, M. Ishida, K. Uchida, H. Someya, Y. Kuschel, and G. Reiss, “Longitudinal spin Seebeck ef- Iwasaki, K. Ihara, S. Kohmoto, E. Saitoh, and T. Mu- fect contribution in transverse spin Seebeck effect ex- rakami, “Spin-Seebeck thermoelectric converter,” IEEE periments in Pt/YIG and Pt/NFO,” Nature Commun., International Nanoelectronics Conference, no. 7460433, vol. 6, pp. 8211/1-8211/7, Sep. 2015. pp. 1-3, July 2014. [114] W.-L. Lee, S. Watauchi, V. L. Miller, R. J. Cava, and [102] S. Gepr¨ags,A. Kehlberger, F. D. Coletta, Z. Qiu, E.-J. N. P. Ong, “Anomalous Hall heat current and Nernst Guo, T. Schulz, C. Mix, S. Meyer, A. Kamra, M. Al- effect in the CuCr2Se4−xBrx ferromagnet,” Phys. Rev. thammer, H. Huebl, G. Jakob, Y. Ohnuma, H. Adachi, Lett., vol. 93, no. 22, pp. 226601/1-226601/4, Nov. 2004. J. Barker, S. Maekawa, G. E. W. Bauer, E. Saitoh, [115] T. Miyasato, N. Abe, T. Fujii, A. Asamitsu, S. On- R. Gross, S. T. B. Goennenwein, and M. Kl¨aui,“Ori- oda, Y. Onose, N. Nagaosa, and Y. Tokura, “Crossover gin of the spin Seebeck effect in compensated ferrimag- behavior of the anomalous Hall effect and anomalous nets,” Nature Commun., vol. 7, pp. 10452/1-10452/6, Nernst effect in itinerant ferromagnets,” Phys. Rev. 26

Lett., vol. 99, no. 8, pp. 086602/1-086602/4, Aug. 2007. vol. 93, no. 2, pp. 020403(R)/1-020403(R)/5, Jan. 2016. [116] Y. Pu, D. Chiba, F. Matsukura, H. Ohno, and J. Shi, [130] A. B. Cahaya, O. A. Tretiakov, and G. E. W. Bauer, “Mott relation for anomalous Hall and Nernst effects “Spin Seebeck power generators,” Appl. Phys. Lett., vol. in Ga1−xMnxAs ferromagnetic semiconductors,” Phys. 104, no. 4, pp. 042402/1-042402/4, Jan. 2014. Rev. Lett., vol. 101, no. 11, pp. 117208/1-117208/4, Sep. [131] T. Liao, J. Lin, G. Su, B. Lin, and J. Chen, “Opti- 2008. mum design of a nanoscale spin-Seebeck power device,” [117] M. Mizuguchi, S. Ohata, K. Uchida, E. Saitoh, and K. Nanoscale, vol. 7, no. 17, pp. 7920-7926, May 2015. Takanashi, “Anomalous Nernst effect in an L10-ordered [132] A. B. Cahaya, O. A. Tretiakov, and G. E. W. Bauer, epitaxial FePt thin film,” Appl. Phys. Express, vol. 5, “Spin Seebeck power conversion,” IEEE Trans. Magn., no. 9, pp. 093002/1-093002/3, Sep. 2012. vol. 51, no. 9, pp. 0800414/1-0800414/14, Sep. 2015. [118] Y. Sakuraba, K. Hasegawa, M. Mizuguchi, T. Kubota, [133] Y. Tserkovnyak, A. Brataas, G. E. W. Bauer, and B. S. Mizukami, T. Miyazaki, and K. Takanashi, “Anoma- I. Halperin, “Nonlocal magnetization dynamics in fer- lous Nernst effect in L10-FePt/MnGa thermopiles for romagnetic heterostructures,” Rev. Mod. Phys., vol. 77, new thermoelectric applications,” Appl. Phys. Express, no. 4, pp. 1375-1421, Dec. 2005. vol. 6, no. 3, pp. 033003/1-033003/4, Mar. 2013. [134] X. Jia, K. Liu, K. Xia, and G. E. W. Bauer, “Spin trans- [119] R. Ramos, M. H. Aguirre, A. Anad´on,J. Blasco, I. fer torque on magnetic insulators,” Europhys. Lett., vol. Lucas, K. Uchida, P. A. Algarabel L. Morell´on, E. 96, no. 1, pp. 17005/1-17005/6, Oct. 2011. Saitoh, and M. R. Ibarra, “Anomalous Nernst effect of [135] M. Weiler, M. Althammer, M. Schreier, J. Lotze, M. Fe3O4 single crystal,” Phys. Rev. B, vol. 90, no. 5, pp. Pernpeintner, S. Meyer, H. Huebl, R. Gross, A. Kamra, 054422/1-054422/6, Aug. 2014. J. Xiao, Y.-T. Chen, H. Jiao, G. E. W. Bauer, and S. T. [120] K. Uchida, T. Kikkawa, T. Seki, T. Oyake, J. Shiomi, B. Goennenwein, “Experimental test of the spin mixing Z. Qiu, K. Takanashi, and E. Saitoh, “Enhancement interface conductivity concept,” Phys. Rev. Lett., vol. of anomalous Nernst effects in metallic multilayers free 111, no. 17, pp. 176601/1-176601/5, Oct. 2013. from proximity-induced magnetism,” Phys. Rev. B, vol. [136] Y. Ohnuma, H. Adachi, E. Saitoh, and S. Maekawa, 92, no. 9, pp. 094414/1-094414/6, Sep. 2015. “Enhanced dc spin pumping into a fluctuating ferro- [121] Y. Sakuraba, “Potential of thermoelectric power gener- magnet near TC,” Phys. Rev. B, vol. 89, no. 17, pp. ation using anomalous Nernst effect in magnetic mate- 174417/1-174417/10, May 2014. rials,” Scripta Mater., vol. 111, pp. 29-32, Jan. 2016. [137] H. Jiao and G. E. W. Bauer, “Spin backflow and ac [122] A. D. Avery, M. R. Pufall, and B. L. Zink, “Observation voltage generation by spin pumping and the inverse of the planar Nernst effect in permalloy and nickel thin spin Hall effect,” Phys. Rev. Lett., vol. 110, no. 21, pp. films with in-plane thermal gradients,” Phys. Rev. Lett., 217602/1-217602/5, May 2013. vol. 109, no. 19, pp. 196602/1-196602/5, Nov. 2012. [138] M. B. Jungfleisch, V. Lauer, R. Neb, A. V. Chumak, and [123] M. Schmid, S. Srichandan, D. Meier, T. Kuschel, J.-M. B. Hillebrands, “Improvement of the yttrium iron gar- Schmalhorst, M. Vogel, G. Reiss, C. Strunk, and C. H. net/platinum interface for spin pumping-based applica- Back,“Transverse spin Seebeck effect versus anomalous tions,” Appl. Phys. Lett., vol. 103, no. 2, pp. 022411/1- and planar Nernst effects in permalloy thin films,” Phys. 022411/4, Jul. 2013. Rev. Lett., vol. 111, no. 18, pp. 187201/1-187201/5, Oct. [139] Z. Qiu, K. Ando, K. Uchida, Y. Kajiwara, R. Taka- 2013. hashi, H. Nakayama, T. An, Y. Fujikawa, and E. Saitoh, [124] D. Meier, D. Reinhardt, M. Schmid, C. H. Back, J.-M. “Spin mixing conductance at a well-controlled plat- Schmalhorst, T. Kuschel, and G. Reiss, “Influence of inum/yttrium iron garnet interface,” Appl. Phys. Lett., heat flow directions on Nernst effects in Py/Pt bilayers,” vol. 103, no. 9, pp. 092404/1-092404/4, Aug. 2013. Phys. Rev. B, vol. 88, no. 18, pp. 184425/1-184425/6, [140] M. Morota, Y. Niimi, K. Ohnishi, D. H. Wei, T. Tanaka, Nov. 2013. H. Kontani, T. Kimura, and Y. Otani, “Indication of [125] I. V. Soldatov, N. Panarina, C. Hess, L. Schultz, intrinsic spin Hall effect in 4d and 5d transition metals,” and R. Sch¨afer,“Thermoelectric effects and magnetic Phys. Rev. B, vol. 83, no. 17, pp. 174405/1-174405/5, anisotropy of Ga1−xMnxAs thin films,” Phys. Rev. B, May 2011. vol. 90, no. 10, pp. 104423/1-104423/9, Sep. 2014. [141] H. L. Wang, C. H. Du, Y. Pu, R. Adur, P. C. Ham- [126] L. J. Cornelissen, J. Liu, R. A. Duine, J. Ben Youssef, mel, and F. Y. Yang, “Scaling of spin Hall angle in 3d, and B. J. van Wees, “Long-distance transport of 4d, and 5d metals from Y3Fe5O12/metal spin pump- magnon spin information in a magnetic insulator at ing,” Phys. Rev. Lett., vol. 112, no. 19, pp. 197201/1- room temperature,” Nature Phys., vol. 11, no. 12, pp. 197201/5, May 2014. 1022-1026, Dec. 2015. [142] S. Singh, M. Anguera, E. del Barco, R. Springell, and [127] S. T. B. Goennenwein, R. Schlitz, M. Pernpeintner, K. C. W. Miller, “Moderate positive spin Hall angle in ura- Ganzhorn, M. Althammer, R. Gross, and H. Huebl, nium,” Appl. Phys. Lett., vol. 107, no. 23, pp. 232403/1- “Non-local magnetoresistance in YIG/Pt nanostruc- 232403/5, Dec. 2015. tures,” Appl. Phys. Lett., vol. 107, no. 17, pp. 172405/1- [143] Y. Niimi, M. Morota, D. H. Wei, C. Deranlot, M. 172405/4, Oct. 2015. Basletic, A. Hamzic, A. Fert, and Y. Otani, “Extrinsic [128] B. L. Giles, Z. Yang, J. Jamison, and R. C. Myers, spin Hall effect induced by iridium impurities in cop- “Long-range pure magnon spin diffusion observed in a per,” Phys. Rev. Lett., vol. 106, no. 12, pp. 126601/1- nonlocal spin-Seebeck geometry,” Phys. Rev. B, vol. 92, 126601/4, Mar. 2011. no. 22, pp. 224415/1-224415/13, Dec. 2015. [144] Y. Niimi, Y. Kawanishi, D. H. Wei, C. Deranlot, H. [129] L. J. Cornelissen and B. J. van Wees, “Magnetic field X. Yang, M. Chshiev, T. Valet, A. Fert, and Y. Otani, dependence of the magnon spin diffusion length in the “Giant spin Hall effect induced by skew scattering from magnetic insulator yttrium iron garnet,” Phys. Rev. B, bismuth impurities inside thin film CuBi alloys,” Phys. 27

Rev. Lett., vol. 109, no. 15, pp. 156602/1-156602/5, Oct. mer, S. T. B. Goennenwein, E. Saitoh, and G. E. W. 2012. Bauer, “Theory of spin Hall magnetoresistance,” Phys. [145] P. Laczkowski, J.-C. Rojas-S´anchez, W. Savero-Torres, Rev. B, vol. 87, no. 14, pp. 144411/1-144411/9, Apr. H. Jaffr`es,N. Reyren, C. Deranlot, L. Notin, C. Beign´e, 2013. A. Marty, J.-P. Attan´e,L. Vila, J.-M. George, and A. [159] H. Nakayama, M. Althammer, Y.-T. Chen, K. Uchida, Fert, “Experimental evidences of a large extrinsic spin Y. Kajiwara, D. Kikuchi, T. Ohtani, S. Gepr¨ags,M. Hall effect in AuW alloy,” Appl. Phys. Lett., vol. 104, Opel, S. Takahashi, R. Gross, G. E. W. Bauer, S. T. no. 14, pp. 142403/1-142403/5, Apr. 2014. B. Goennenwein, and E. Saitoh, “Spin Hall magne- [146] K. Ando, S. Takahashi, J. Ieda, H. Kurebayashi, T. toresistance induced by a nonequilibrium proximity ef- Trypiniotis, C. H. W. Barnes, S. Maekawa, and E. fect,” Phys. Rev. Lett., vol. 110, no. 20, pp. 206601/1- Saitoh, “Electrically tunable spin injector free from the 206601/5, May 2013. impedance mismatch problem,” Nature Mater., vol. 10, [160] C. Hahn, G. de Loubens, O. Klein, M. Viret, V. V. no. 9, pp. 655-659, Sep. 2011. Naletov, and J. Ben Youssef, “Comparative measure- [147] L. Chen, F. Matsukura, and H. Ohno, “Direct-current ments of inverse spin Hall effects and magnetoresistance voltages in (Ga,Mn)As structures induced by ferromag- in YIG/Pt and YIG/Ta,” Phys. Rev. B, vol. 87, no. 17, netic resonance,” Nature Commun., vol. 4, pp. 2055/1- pp. 174417/1-174417/8, May 2013. 2055/6, Jun. 2013. [161] N. Vlietstra, J. Shan, V. Castel, J. Ben Youssef, and B. [148] J.-C. Rojas-S´anchez, M. Cubukcu, A. Jain, C. J. van Wees, “Spin-Hall magnetoresistance in platinum Vergnaud, C. Portemont, C. Ducruet, A. Barski, A. on yttrium iron garnet: Dependence on platinum thick- Marty, L. Vila, J.-P. Attan´e, E. Augendre, G. Des- ness and in-plane/out-of-plane magnetization,” Phys. fonds, S. Gambarelli, H. Jaffr`es,J.-M. George, and M. Rev. B, vol. 87, no. 18, pp. 184421/1-184421/5, May Jamet, “Spin pumping and inverse spin Hall effect in 2013. germanium,” Phys. Rev. B, vol. 88, no. 6, pp. 064403/1- [162] M. Althammer, S. Meyer, H. Nakayama, M. Schreier, 064403/15, Aug. 2013. S. Altmannshofer, M. Weiler, H. Huebl, S. Gepr¨ags, [149] J.-C. Lee, L.-W. Huang, D.-S. Hung, T.-H. Chiang, M. Opel, R. Gross, D. Meier, C. Klewe, T. Kuschel, J. C. A. Huang, J.-Z. Liang, and S.-F. Lee, “Inverse J.-M. Schmalhorst, G. Reiss, L. Shen, A. Gupta, Y.- spin Hall effect induced by spin pumping into semicon- T. Chen, G. E. W. Bauer, E. Saitoh, and S. T. B. ducting ZnO,” Appl. Phys. Lett., vol. 104, no. 5, pp. Goennenwein, “Quantitative study of the spin Hall mag- 052401/1-052401/4, Feb. 2014. netoresistance in ferromagnetic insulator/normal metal [150] Z. Qiu, Y. Kajiwara, K. Ando, Y. Fujikawa, K. Uchida, hybrids,” Phys. Rev. B, vol. 87, no. 22, pp. 224401/1- T. Tashiro, K. Harii, T. Yoshino, and E. Saitoh, “All- 224401/15, Jun. 2013. oxide system for spin pumping,” Appl. Phys. Lett., vol. [163] F. J. Jedema, A. T. Filip, and B. J. van Wees, “Electri- 100, no. 2, pp. 022402/1-022402/3, Jan. 2012. cal spin injection and accumulation at room tempera- [151] K. Fujiwara, Y. Fukuma, J. Matsuno, H. Idzuchi, Y. ture in an all-metal mesoscopic spin valve,” Nature, vol. Niimi, Y. Otani, and H. Takagi, “5d iridium oxide as a 410, no. 6826, pp. 345-348, Mar. 2001. material for spin-current detection,” Nature Commun., [164] F. J. Jedema, H. B. Heersche, A. T. Filip, J. J. A. Basel- vol. 4, pp. 2893/1-2893/6, Dec. 2013. mans, and B. J. van Wees, “Electrical detection of spin [152] K. Ando, S. Watanabe, S. Mooser, E. Saitoh, and H. Sir- precession in a metallic mesoscopic spin valve,” Nature, ringhaus, “Solution-processed organic spin-charge con- vol. 416, no. 6882, pp. 713-716, Apr. 2002. verter,” Nature Mater., vol. 12, no. 7, pp. 622-627, Jul. [165] Y. Niimi and Y. Otani, “Reciprocal spin Hall effects 2013. in conductors with strong spin-orbit coupling: a re- [153] Z. Qiu, M. Uruichi, D. Hou, K. Uchida, H. M. Ya- view,” Rep. Prog. Phys., vol. 78, no. 12, pp. 124501/1- mamoto, and E. Saitoh, “Spin-current injection and de- 124501/19, Dec. 2015. tection in κ-(BEDT-TTF)2Cu[N(CN)2]Br,” AIP Adv., [166] J. Xiao, G. E. W. Bauer, K. Uchida, E. Saitoh, and vol. 5, no. 5, pp. 057167/1-057167/7, May 2015. S. Maekawa, “Theory of magnon-driven spin Seebeck [154] R. H. Silsbee, A. Janossy, and P. Monod, “Coupling effect,” Phys. Rev. B, vol. 81, no. 21, pp. 214418/1- between ferromagnetic and conduction-spin-resonance 214418/8, Jun. 2010. modes at a ferromagnetic—normal-metal interface,” [167] H. Adachi, K. Uchida, E. Saitoh, J. Ohe, S. Takahashi, Phys. Rev. B, vol. 19, no. 9, pp. 4382-4399, May 1979. and S. Maekawa, “Gigantic enhancement of spin See- [155] Y. Tserkovnyak, A. Brataas, and G. E. W. Bauer, “En- beck effect by phonon drag,” Appl. Phys. Lett., vol. 97, hanced Gilbert damping in thin ferromagnetic films,” no. 25, pp. 252506/1-252506/3, Dec. 2010. Phys. Rev. Lett., vol. 88, no. 11, pp. 117601/1-117601/4, [168] H. Adachi, J. Ohe, S. Takahashi, and S. Maekawa, Feb. 2002. “Linear-response theory of spin Seebeck effect in fer- [156] S. Mizukami, Y. Ando, and T. Miyazaki, “Effect of spin romagnetic insulators,” Phys. Rev. B, vol. 83, no. 9, diffusion on Gilbert damping for a very thin permalloy 094410/1-094410/6, Mar. 2011. layer in Cu/permalloy/Cu/Pt films,” Phys. Rev. B, vol. [169] J. Ohe, H. Adachi, S. Takahashi, and S. Maekawa, “Nu- 66, no. 10, pp. 104413/1-104413/9, Sep. 2002. merical study on the spin Seebeck effect,” Phys. Rev. B, [157] Y. Kajiwara, K. Harii, S. Takahashi, J. Ohe, K. Uchida, vol. 83, no. 11, 115118/1-115118/5, Mar. 2011. M. Mizuguchi, H. Umezawa, H. Kawai, K. Ando, K. [170] S. A. Bender, R. A. Duine, and Y. Tserkovnyak, Takanashi, S. Maekawa, and E. Saitoh, “Transmission “Electronic pumping of quasiequilibrium Bose-Einstein- of electrical signals by spin-wave interconversion in a condensed magnons,” Phys. Rev. Lett., vol. 108, no. 24, magnetic insulator,” Nature, vol. 464, no. 7286, pp. 262- pp. 246601/1-246601/5, Jun. 2012. 266, Mar. 2010. [171] Y. Ohnuma, H. Adachi, E. Saitoh, and S. Maekawa, [158] Y.-T. Chen, S. Takahashi, H. Nakayama, M. Altham- “Spin Seebeck effect in antiferromagnets and compen- 28

sated ferrimagnets,” Phys. Rev. B, vol. 87, no. 1, pp. Comment on “Pt magnetic polarization on Y3Fe5O12 014423/1-014423/7, Jan. 2013. and magnetotransport characteristics”, 2013. [Online]. [172] H. Adachi, K. Uchida, E. Saitoh, and S. Maekawa, Available: http://arxiv.org/abs/1307.4869 “Theory of the spin Seebeck effect,” Rep. Prog. Phys., [187] Y. Sun, H. Chang, M. Kabatek, Y.-Y. Song, Z. Wang, vol. 76, no. 3, pp. 036501/1-036501/20, Mar. 2013. M. Jantz, W. Schneider, M. Wu, E. Montoya, B. Kar- [173] S. Hoffman, K. Sato, and Y. Tserkovnyak, “Landau- dasz, B. Heinrich, S. G. E. te Velthuis, H. Schultheiss, Lifshitz theory of the longitudinal spin Seebeck effect,” and A. Hoffmann, “Damping in yttrium iron garnet Phys. Rev. B, vol. 88, no. 6, pp. 064408/1-064408/8, nanoscale films capped by platinum,” Phys. Rev. Lett., Aug. 2013. vol. 111, no. 10, pp. 106601/1-106601/5, Sep. 2013. [174] L. Chotorlishvili, Z. Toklikishvili, V. K. Dugaev, J. [188] B. F. Miao, S. Y. Huang, D. Qu, and C. L. Chien, Barna´s,S. Trimper, and J. Berakdar, “Fokker-Planck “Physical origins of the new magnetoresistance in approach to the theory of the magnon-driven spin See- Pt/YIG,” Phys. Rev. Lett., vol. 112, no. 23, pp. beck effect,” Phys. Rev. B, vol. 88, no. 14, pp. 144429/1- 236601/1-236601/5, Jun. 2014. 144429/6, Oct. 2013. [189] G. Y. Guo, Q. Niu, and N. Nagaosa, “Anomalous Nernst [175] J. Ren, “Predicted rectification and negative differential and Hall effects in magnetized platinum and palladium,” spin Seebeck effect at magnetic interfaces,” Phys. Rev. Phys. Rev. B, vol. 89, no. 21, pp. 214406/1-214406/6, B, vol. 88, no. 22, pp. 220406(R)/1-220406(R)/5, Dec. Jun. 2014. 2013. [190] M. Haertinger, C. H. Back, J. Lotze, M. Weiler, S. [176] S. A. Bender and Y. Tserkovnyak, “Interfacial spin Gepr¨ags, H. Huebl, S. T. B. Goennenwein, and G. and heat transfer between metals and magnetic insu- Woltersdorf, “Spin pumping in YIG/Pt bilayers as a lators,” Phys. Rev. B, vol. 91, no. 14, pp. 140402(R)/1- function of layer thickness,” Phys. Rev. B, vol. 92, no. 140402(R)/5, Apr. 2015. 5, pp. 054437/1-054437/6, Aug. 2015. [177] I. I. Lyapilin, M. S. Okorokov, and V. V. Ustinov, “Spin [191] D. A. Papaconstantopoulos, Handbook of the Band effects induced by thermal perturbation in a normal Structure of Elemental Solids. Plenum Press, 1986. metal/magnetic insulator system,” Phys. Rev. B, vol. [192] H. Ibach and H. L¨uth, Solid-State Physics: An Introduc- 91, no. 19, pp.195309/1-195309/7, May 2015. tion to Principles of Materials Science. Springer, 2009. [178] S. R. Etesami, L. Chotorlishvili, and J. Berakdar, [193] S. M. Rezende, R. L. Rodr´ıguez-Su´arez,R. O. Cunha, “Spectral characteristics of time resolved magnonic spin J. C. L´opez Ortiz, and A. Azevedo, “Bulk magnon spin Seebeck effect,” Appl. Phys. Lett., vol. 107, no. 13, pp. current theory for the longitudinal spin Seebeck effect,” 132402/1-132402/4, Sep. 2015. J. Magn. Magn. Mater., vol. 400, pp. 171-177, Feb. [179] A. Brataas, H. Skarsv˚ag,E. G. Tveten, and E. L. 2016. Fjærbu, “Heat transport between antiferromagnetic in- [194] D. J. Sanders and D. Walton, “Effect of magnon-phonon sulators and normal metals,” Phys. Rev. B, vol. 92, no. thermal relaxation on heat transport by magnons,” 18, pp. 180414(R)/1-180414(R)/5, Nov. 2015. Phys. Rev. B, vol. 15, no. 3, pp. 1489-1494, Feb. 1977. [180] L. Chotorlishvili, Z.Toklikishvili, S. R. Etesami, V. [195] S. S.-L. Zhang and S. Zhang, “Magnon mediated elec- K. Dugaev, J. Barna´s, and J. Berakdar, “Magnon- tric current drag across a ferromagnetic insulator layer,” driven longitudinal spin Seebeck effect in F |N and Phys. Rev. Lett., vol. 109, no. 9, pp. 096603/1-096603/5, N|F |N structures: Role of asymmetric in-plane mag- Aug. 2012. netic anisotropy,” J. Magn. Magn. Mater., vol. 396, pp. [196] S. S.-L. Zhang and S. Zhang, “Spin convertance at mag- 254-262, Dec. 2015. netic interfaces,” Phys. Rev. B, vol. 86, no. 21, pp. [181] S. M. Rezende, R. L. Rodr´ıguez-Su´arez, and A. 214424/1-214424/7, Dec. 2012. Azevedo, “Theory of the spin Seebeck effect in antifer- [197] M. Agrawal, V. I. Vasyuchka, A. A. Serga, A. D. romagnets,” Phys. Rev. B, vol. 93, no. 1, pp. 014425/1- Karenowska, G. A. Melkov, and B. Hillebrands, “Direct 014425/9, Jan. 2016. measurement of magnon temperature: New insight into [182] J. Xiao and G. E. W. Bauer, “Transport between met- magnon-phonon coupling in magnetic insulators,” Phys. als and magnetic insulators,” 2015. [Online]. Available: Rev. Lett., vol. 111, no. 10, pp. 107204/1-107204/5, Sep. http://arxiv.org/abs/1508.02486 2013. [183] S. Y. Huang, X. Fan, D. Qu, Y. P. Chen, W. G. Wang, [198] A. R¨uckriegel, P. Kopietz, D. A. Bozhko, A. A. Serga, J. Wu, T. Y. Chen, J. Q. Xiao, and C. L. Chien, “Trans- and B. Hillebrands, “Magnetoelastic modes and lifetime port magnetic proximity effects in platinum,” Phys. of magnons in thin yttrium iron garnet films,” Phys. Rev. Lett., vol. 109, no. 10, pp. 107204/1-107204/5, Sep. Rev. B, vol. 89, no. 18, pp. 184413/1-184413/14, May 2012. 2014. [184] S. Gepr¨ags,S. Meyer, S. Altmannshofer, M. Opel, F. [199] S. R. Boona and J. P. Heremans, “Magnon thermal Wilhelm, A. Rogalev, R. Gross, and S. T. B. Goennen- mean free path in yttrium iron garnet,” Phys. Rev. B, wein, “Investigation of induced Pt magnetic polariza- vol. 90, no. 6, pp. 064421/1-064421/8, Aug. 2014. tion in Pt/Y3Fe5O12 bilayers,” Appl. Phys. Lett., vol. [200] T. C. Harman and J. M. Honig, “Theory of galvano- 101, no. 26, pp. 262407/1-262407/4, Dec. 2012. thermomagnetic energy conversion devices. I. Genera- [185] Y. M. Lu, Y. Choi, C. M. Ortega, X. M. Cheng, J. W. tors,” J. Appl. Phys., vol. 33, no. 11, pp. 3178-3188, Cai, S. Y. Huang, L. Sun, and C. L. Chien, “Pt mag- Nov. 1962. netic polarization on Y3Fe5O12 and magnetotransport [201] T. C. Harman and J. M. Honig, “Theory of galvano- characteristics,” Phys. Rev. Lett., vol. 110, no. 14, pp. thermomagnetic energy conversion devices. II. Refriger- 147207/1-147207/5, Apr. 2013. ators and heat pumps,” J. Appl. Phys., vol. 33, no. 11, [186] S. Gepr¨ags,S. T. B. Goennenwein, M. Schneider, F. pp. 3188-3194, Nov. 1962. Wilhelm, K. Ollefs, A. Rogalev, M. Opel, and R. Gross, [202] T. C. Harman and J. M. Honig, “Theory of galvano- 29

thermomagnetic energy conversion devices. III. Genera- vol. 11, no. 3, pp. 199-202, Mar. 2012. tors constructed from anisotropic materials,” J. Appl. [220] C. M. Jaworski, R. C. Myers, E. Johnston-Halperin, and Phys., vol. 34, no. 1, pp. 189-194, Nov. 1963. J. P. Heremans, “Giant spin Seebeck effect in a non- [203] L. D. Landau, E. M. Lifshitz, and L. P. Pitaevskii, magnetic material,” Nature, vol. 487, no. 7406, pp. 210- Electrodynamics of Continuous Media. Pergamon Press, 213, Jul. 2012. 1984. [204] Y. Zhang, N. P. Ong, Z. A. Xu, K. Krishana, R. Gagnon, and L. Taillefer, “Determining the Wiedemann-Franz Profiles: ratio from the thermal Hall conductivity: Application to Cu and YBa2Cu3O6.95,” Phys. Rev. Lett., vol. 84 no. 10, pp. 2219-2222, Mar. 2000. Ken-ichi Uchida was born in Kanagawa Prefecture, [205] Y. Shiomi, Y. Onose, and Y. Tokura, “Effect of scat- Japan, in 1986. He received the B.Eng. and M.Sc. Eng. tering on intrinsic anomalous Hall effect investigated by degrees from Keio University, Yokohama, Japan, in 2008 Lorenz ratio,” Phys. Rev. B, vol. 81 no. 5, pp. 054414/1- and 2009. respectively, and the Ph.D. degree from To- 054414/6, Feb. 2010. hoku University, Sendai, Japan, in 2012. [206] W. M. Yim and A. Amith, “Bi-Sb alloys for magneto- He was an Assistant Professor at the Institute for Ma- thermoelectric and thermomagnetic cooling,” Solid- terials Research, Tohoku University from 2012 to 2014. State Electron., vol. 15, no. 10, pp. 1141-1165, Oct. 1972. He has been an Associate Professor at the Institute for [207] C. A. Domenicali, “Stationary temperature distribution in an electrically heated conductor,” J. Appl. Phys., vol. Materials Research, Tohoku University, since 2014, and 25, no. 10, pp. 1310-1311, Oct. 1954. a PRESTO Researcher at Japan Science and Technology [208] R. R. Heikes and R. W. Ure, Jr., Thermoelectricity: Sci- Agency, since 2012. He has worked on spintronics and ence and Engineering. New York: Interscience Publish- spin caloritronics. He has developed thermal, acoustic, ers Inc., 1961. and plasmonic methods of generating spin currents and [209] G. D. Mahan, “Good Thermoelectrics,” in Solid State realized thermoelectric generation using insulators. Physics, vol. 51, pp. 81-157, Sep. 1997. [210] R. B. Horst, “Thermomagnetic figure of merit: Bis- Hiroto Adachi was born in Aichi Prefecture, Japan, in muth,” J. Appl. Phys., vol. 34, no. 11, pp. 3246-3254, 1975. He received the B.Sc., M.Sc., and D.Sc. degrees Nov. 1963. from Kyoto University, Kyoto, Japan, in 1998, 2000, and [211] R. T. Delves, “Figure of merit for Ettingshausen cool- ing,” Brit. J. Appl. Phys., vol. 15, no. 1, pp. 105-106, 2003, respectively. Jan. 1964. He was a Postdoctoral Research Assistant at Okayama [212] J. Flipse, F. K. Dejene, D. Wagenaar, G. E. W. Bauer, University, Okayama, Japan, from 2003 to 2007, at the J. Ben Youssef, and B. J. van Wees, “Observation of the Swiss Federal Institute of Technology (ETH) Zurich, spin Peltier effect for magnetic insulators,” Phys. Rev. Zurich, Switzerland, from 2007 to 2009, at the Insti- Lett., vol. 113, no. 2, pp. 027601/1-027601/5, Jul. 2014. tute for Materials Research, Tohoku University, Sendai, [213] A. A. Tulapurkar and Y. Suzuki, “Boltzmann approach Japan, from 2009 to 2010, and at the Advanced Science to dissipation produced by a spin-polarized current,” Research Center, Japan Atomic Energy Agency, Tokai, Phys. Rev. B, vol. 83, no. 1, pp. 012401/1-012401/4, Japan, from 2010 to 2011. Since 2011, he has been a Se- Jan. 2011. [214] M. Abe, T. Itoh, and Y. Tamaura, “Magnetic and bio- nior PD Fellow at the Advanced Science Research Cen- magnetic films obtained by ferrite plating in aqueous ter, Japan Atomic Energy Agency. His research interests solution,” Thin Solid Films, vol. 216, no. 1, pp. 155- include spintronics and superconductivity. 161, Aug. 1992. [215] K. Kondo, S. Yoshida, H. Ono, and M. Abe, “Spin Takashi Kikkawa was born in Kanagawa Prefecture, sprayed Ni(-Zn)-Co ferrite films with natural resonance Japan, in 1991. He received the B.Sc. and M.Sc. de- frequency exceeding 3 GHz,” J. Appl. Phys., vol. 101, grees from Tohoku University, Sendai, Japan, in 2013 no. 9, pp. 09M502/1-09M502/3, May 2007. and 2015, respectively. He is currently working toward [216] A. Slachter, F. L. Bakker, J.-P. Adam, and B. J. van the Ph.D. degree in science at Tohoku University. Wees, “Thermally driven spin injection from a ferro- magnet into a non-magnetic metal,” Nature Phys., vol. Since 2015, he has been supported by the Japan So- 6, no. 11, pp. 879-882, Nov. 2010. ciety for the Promotion of Science through a research [217] J.-C. Le Breton, S. Sharma, H. Saito, S. Yuasa, and fellowship for young scientists. His research interest in- R. Jansen, “Thermal spin current from a ferromagnet cludes the investigation of thermo-spin effects in mag- to silicon by Seebeck spin tunnelling,” Nature, vol. 475, netic materials. no. 7354, pp. 82-85, Jul. 2011. [218] M. Walter, J. Walowski, V. Zbarsky, M. M¨unzenberg, Akihiro Kirihara was born in Saga Prefecture, Japan, M. Sch¨afers,D. Ebke, G. Reiss, A. Thomas, P. Peret- in 1979. He received the B.Eng. and M.Eng. degrees zki, M. Seibt, J. S. Moodera, M. Czerner, M. Bachmann, and C. Heiliger, “Seebeck effect in magnetic tunnel junc- from the University of Tokyo, Tokyo, Japan, in 2002 and tions,” Nature Mater., vol. 10, no. 10, pp. 742-746, Oct. 2004, respectively. 2011. He has belonged to NEC Corporation since 2004. [219] M. V. Costache, G. Bridoux, I. Neumann, and S. O. He was a Visiting Researcher at Technische Universit¨at Valenzuela, “Magnon-drag thermopile,” Nature Mater., Kaiserslautern, Kaiserslautern, Germany, from 2013 to 30

2014. Since 2014, he has also been a Research Member tute for Materials Research, Tohoku University, Sendai, of the ERATO “Spin Quantum Rectification” project at Japan, a Professor in the Faculty of Engineering, Nagoya Japan Science and Technology Agency. His research in- University, Nagoya, Japan, from 1988 to 1997, and a Pro- terests include spintronics, spin caloritronics, and ther- fessor at the Institute for Materials Research, Tohoku moelectric conversion devices. University from 1997 to 2010. Since 2010, he has been a Director of the Advanced Science Research Center, Masahiko Ishida was born in Fukushima Prefecture, Japan Atomic Energy Agency, Tokai, Japan. His main Japan, in 1972. He received the B.Eng., M.Eng., and research interests include theories of electronic proper- D.Eng. degrees from University of Tsukuba, Tsukuba, ties in strongly correlated electron systems, in particular, Japan, in 1995, 1997, and 1999, respectively. high-temperature superconductors and orbital physics in After spending one year of postdoctoral research at transition metal oxides, and theories of transport phe- University of Tsukuba, he joined Fundamental Research nomena in magnetic nanostructures. Laboratories, NEC Corporation in 2000. He was also a Dr. Maekawa received Fellow of Institute of Physics Visiting Researcher at the California Nanosystems Insti- (U.K.) in 1999, the Humboldt Award (Germany) in 2001, tute, University of California, Los Angeles, California, APS Fellow in 2007, the IUPAP Magnetism Award and USA, from 2007 to 2008. Since 2014, he has been lead- Neel Medal in 2012, the Honoris Cause Doctorate of Uni- ing the Spin Thermoelectrics Team at Smart Energy Re- versity of Zaragoza (Spain) in 2013, and several Japanese search Laboratories, NEC Corporation, and the spin See- awards. beck effect utilization group of the ERATO “Spin Quan- tum Rectification” project at Japan Science and Technol- Eiji Saitoh was born in Tokyo, Japan, in 1971. He ogy Agency. He has worked on various fields of electronic received the B.Eng., M.Eng., and Ph.D. degrees from and sensing device fabrication. the University of Tokyo, Tokyo, Japan, in 1996, 1998, and 2001, respectively. Shinichi Yorozu was born in Ibaraki Prefecture, Japan, He was a Research Associate in the Department of in 1965. He received the B.Eng. degree from Hokkaido Physics, Keio University, Yokohama, Japan, from 2001 University, Hokkaido, Japan, in 1988, and the M.Eng. to 2006, and an Assistant Professor in the Department and Ph.D. degrees from the University of Tokyo, Tokyo, of Applied Physics and Physico-Informatics, Keio Uni- Japan, in 1990 and 1993, respectively. versity, from 2006 to 2009. He has been a Professor He joined Fundamental Research Laboratories, NEC at the Institute for Materials Research, Tohoku Univer- Corporation in 1993. He was a Visiting Researcher at the sity, Sendai, Japan, since 2009, a Professor at the WPI State University of New York at Stony Brook, New York, Advanced Institute for Materials Research, Tohoku Uni- USA, from 1997 to 1998. He was a Senior Manager of versity, since 2012, and a Visiting Group Leader at the the Fundamental and Environmental Research Laborato- Advance Science Research Center, Japan Atomic Energy ries, NEC Corporation in 2005. Since 2015, he has been a Agency, Tokai, Japan, since 2010. Since 2014, he has Deputy General Manager of Smart Energy Research Lab- been a Research Director of the ERATO “Spin Quantum oratories, NEC Corporation. He has responsibility for Rectification” project at Japan Science and Technology the research area of nanotechnology and quantum tech- Agency. He was also a Visiting Scholar at the Cavendish nology, which include spintronics, nano-photonic devices, Laboratory, University of Cambridge, Cambridge, U.K., quantum information, and printed electronics. in 2004 and a PRESTO Researcher at the Japan Sci- ence and Technology Agency, from 2007 to 2011. He has Sadamichi Maekawa was born in Nara Prefecture, worked on various fields of condensed matter physics such Japan, in 1946. He received the B.Sc. and M.Sc. de- as strongly correlated electron systems, nanomagnetics, grees from Osaka University, Osaka, Japan, in 1969 and and spintronics. His research is now focused on spin- 1971, respectively, and the D.Sc. degree from Tohoku related phenomena including the generation, detection, University, Sendai, Japan, in 1975. and manipulation of spin currents. He was a Research Associate from 1971 to 1982, and Dr. Saitoh received the IUPAP Young Scientist Award an Associate Professor from 1982 to 1988 at the Insti- in 2009 and many Japanese awards.