Spin-dependent thermoelectric effect and spin battery mechanism in triple quantum dots with Rashba spin–orbital interaction
Xu Wei-Ping1, 2, Zhang Yu-Ying1, Wang Qiang3, Nie Yi-Hang1, †,
Institute of Theoretical Physics and Department of Physics, Shanxi University, Taiyuan 030006, China
School of Applied Science, Taiyuan University of Science and Technology, Taiyuan 030024, China
Department of Physics, Taiyuan Normal University, Taiyuan 030001, China

 

† Corresponding author. E-mail: nieyh@sxu.edu.cn

Project supported by the National Natural Science Foundation of China (Grant Nos. 11274208 and 11447170).

Abstract
Abstract

We have studied spin-dependent thermoelectric transport through parallel triple quantum dots with Rashba spin–orbital interaction (RSOI) embedded in an Aharonov–Bohm interferometer connected symmetrically to leads using nonequilibrium Green’s function method in the linear response regime. Under the appropriate configuration of magnetic flux phase and RSOI phase, the spin figure of merit can be enhanced and is even larger than the charge figure of merit. In particular, the charge and spin thermopowers as functions of both the magnetic flux phase and the RSOI phase present quadruple-peak structures in the contour graphs. For some specific configuration of the two phases, the device can provide a mechanism that converts heat into a spin voltage when the charge thermopower vanishes while the spin thermopower is not zero, which is useful in realizing the thermal spin battery and inducing a pure spin current in the device.

1. Introduction

The efficiency of a material directly converting thermal energy into electrical energy or vice versa can be described by the dimensionless charge figure of merit (FOM) , where Sc is the charge Seebeck coefficient (thermopower), Gc is the electric conductance, κel(ph) is the electric (phonon) thermal conductance, and T is the operating temperature. In bulk materials, however, the value of ZcT remains at about one due to the restriction of the Wiedemann–Franz (W–F) law.[1] In recent years, the study of high efficient heat-electricity conversion in low-dimensional materials has become an active subject due to the effective improvement in conversion efficiency.[25] Thermoelectric effects in quantum dot (QD) systems have been widely studied both experimentally and theoretically.[614] In the QD systems, the thermoelectric effects can not only lead to considerable values of ZcT due to the strong violation of the W–F law[8] and small phonon contribution to the thermal conductance,[9] but also reveal many novel thermoelectric phenomena such as Coulomb blockade oscillations of the thermopower and the thermal conductance,[11] and the bipolar effect in the thermal conductance spectrum.[14] In the QD systems, the thermoelectric figure of merit can be largely enhanced by the Coulomb blockade effect.[14] Moreover, experiments have demonstrated the capability of measuring the Seebeck coefficients in atomic and molecular junctions,[1517] which inspires the theoretical study on the thermoelectric properties of nanojunctions.[1824]

Recently, the spin Seebeck effect has been observed in a metallic magnet using a spin-detection technique based on the spin Hall effect,[25] in which the temperature difference between the edges of a bulk ferromagnetic slab plays a role of the driving force for the spin voltage and spin current. This spin Seebeck effect is an analogy to the usual charge Seebeck effect, and can be applied directly to the thermal spin generators for driving spin current.[2628] It is more interesting that a heat flow can also been converted into a spin voltage in the magnetic insulator despite the absence of conduction electrons.[29] Using a similar experiment geometry, one also observed the spin Seebeck effect in ferromagnetic semiconductors.[30] The influence of the ferromagnetic leads on the thermoelectric transport properties has also been investigated in the QD systems with ferromagnetic electrodes.[3134] In addition, the quantum interference effects such as Fano effect, Dicke resonances, and Aharonov–Bohm oscillations have been proposed as methods of increasing the thermoelectric efficiency in the QD systems.[3539] Thus many efforts have been devoted to improving the thermoelectric efficiency in the QD systems. In particular, the Rashba spin–orbit interaction (RSOI) in the QD was used to produce the spin-dependent thermoelectric effect. In Refs. [34] (single QD system) and [28] (DQD system), the coexistence of the RSOI in the QD and the leads’ spin polarization enhanced the figure of merit of the spin thermoelectric effects. Reference [27] investigated the thermospin effects in parallel coupled DQD under the coaction of the Rashba spin–orbit interaction and the Zeeman splitting in the DQD, mainly involving the effect of the magnetic field applied to the DQD on the thermoelectric coefficient. However, in multi-QD systems, the research on the spin-dependent thermoelectric properties and the influence of the interference effects remains few. In this paper, we study the spin-dependent thermoelectric transport through parallel triple quantum dots (TQDs) with Rashba spin–orbit interaction (RSOI) embedded in an Aharonov–Bohm interferometer connected symmetrically to the leads, in which the coexistence of a bound state in the continuum (BIC)[40] and a Fano resonance results in wide regions in the space of parameters of large spin thermopower. Due to the presence of the RSOI, both the magnetic flux phase and the RSOI phase affect the thermoelectric transport properties. The spin-dependent electrical and thermal conductances, the thermopower, and the figure of merit as functions of the QD energy level are calculated in the linear-response regime. Under the appropriate configuration of the magnetic flux phase and RSOI phase, the spin figure of merit can be enhanced and even is larger than the charge figure of merit. In particular, the thermopower in both the magnetic flux phase and the RSOI phase is studied when other parameters are given. We find that the device can provide a mechanism of thermal spin battery that converts the heat into a spin voltage and induces a pure spin current.

2. Model

We consider a TQD molecule with RSOI coupled symmetrically to two leads (see Fig. 1). The RSOI is only in QD1 and QD3. The system can be modeled by a noninteracting Anderson Hamiltonian H = Hdot + Hlead + HT, with the QD Hamiltonian

where creates (annihilates) an electron with energy εi and spin σ in QDi (i = 1,2,3), and t denotes the interdot tunneling coupling. Hlead describes the noninteracting electrons in the left and right leads and reads

where is the creation (annihilation) operator of the electron with energy ε, momentum k, and spin σ in lead α. The last term HT in Hamiltonian H denotes tunneling between the QD and the electrodes and is expressed as

where describes tunneling coupling between QDi and lead α and is assumed to be independent of k. With the magnetic flux Φ threading through the AB ring and RSOI in QD1 and QD3, the tunneling matrix elements can be expressed in the form

where the extra spin-independent phase factor φ/4 and spin-dependent phase factor δσφRj/2 (j = 1,3) are added to the tunneling matrix elements with δ = 1, δ = −1, λL = 1, and λR = −1. φ = 2πΦ/ϕ0 is the Aharonov–Bohm phase with the flux quantum ϕ0 = h/e. φR1 and φR3 result from RSOI in QD1 and QD3, respectively. For simplicity, we have assumed that RSOI exists only inside QD and neglected the interlevel spin-flip since only one energy level exists inside each QD. The spin-dependent phase with βi the RSOI strength in QDi, where m* is the electron effective mass and Li is the length of QD. The magnitude of φRi can be tuned in experiment by a gate voltage or the QDs’ configuration.

Fig. 1. Schematic diagram of a coupled TQD system with RSOI in DQ1 and DQ3.

By using the nonequilibrium Green’s function technique, the electronic and heat currents flowing in the system can be written in the forms

where fα = [e(ωμα)/kBT + 1]−1 is the Fermi distribution function for the α electrode with chemical potential μα, temperature Tα, and Boltzmann constant kB. Tσ(ω) is the transmission coefficient corresponding to spin channel σ and is given by

Here and are, respectively, the retarded and the advanced Green’s functions in ω space with is defined by the matrix elements with θ(t) the step function. are matrices describing the coupling between the QDs and the leads, whose matrix elements are defined as .

In the linear response regime, the chemical potentials and the temperatures of the two leads are set to be μL = μR = μ and TL = TR = T. In this system, the chemical potential difference between the two leads is related to the spin channel. One can introduce a spin-dependent voltage ΔVσ = ΔV + δσΔVs with charge bias ΔV and spin bias ΔVs. The spin bias originates from spin accumulation. Corresponding, the spin-resolved electric and heat currents can be expressed as

where . The total charge current and spin current are defined as I = I + I and Is = II, respectively. In the linear-response approximation, in terms of , the spin-resolved electric conductance Gσ and the thermal conductance κelσ are given by and , respectively. The spin-resolved thermopower is defined as . Thus, the charge conductance Gc and thermopower Sc can be expressed as Gc = G + G and Sc = (S + S)/2, and the spin conductance Gs and thermopower Ss can be expressed as Gs = GG and Ss = (SS)/2, respectively. While the electron thermal conductance κ = ∑σκelσ. The corresponding charge and spin figures of merit are defined as and , respectively. For simplicity, we assume ε1 = ε2 = ε3 = ε. In order to diagonalize Hamiltonian Hdot, the QD operators are transformed by using the transformation[40,41]

which gives the diagonalized Hamiltonian of the isolated molecule

in the new subspace spanned by basis {|1,σ⟩,|2,σ⟩,|3,σ⟩}. Here Hdot|ı,σ⟩ = εi|ı,σ⟩ and correspond to three effective molecule states. By using the new QD operators, the Hamiltonian of the tunneling between the QDs and the leads can be expressed as

with the transformed tunneling coupling

For simplicity, we take V1 = V3 = V and V2 = βV with both V and β real, at the same time assume ϕR1 = −ϕR3 = ΔϕR/2. Thus, in the new subspace these coupling coefficients can be expressed as

with ϕσ = φδσΔϕR. Due to the abcence of RSOI in QD2, coupling coefficients and become real, thus choosing the appropriate matching of the magnetic flux phase and the RSOI phase; i.e., some especial values of ϕσ, may lead to some molecule states decoupling completely from the leads, forming a BIC. In the new subspace, the retarded Green’s function can be obtained from the equation of motion as

with

and . In terms of , the retarded Green function in the original subspace can be expressed as

with . In terms of Eqs. (8) and (21), one can calculate the transmission coefficient of the system, thus investigate the thermoelectric transport properties of the system.

3. Results and discussion

Now we investigate numerically the spin-depended thermoelectric transport through the system. For convenience, Γ = 2πρ(0)V2 is used as the energy unit with ρ(0) the density of states in the leads. The system temperature T is fixed to be 0.026Γ (the constant kB = 1). The other parameters are chosen as μL = μR = 0, t = 2Γ, and β = 1. In the presence of the RSOI and the magnetic flux, the electron transmission depends on the phase factors, which is presented in the transmission coefficient

The dependence of the transmission on the phase factor can be used to tune the thermoelectric coefficients via . According to the definition of the phase ϕσ, the RSOI phase is equivalent to the magnetic flux phase. However, actually, the influence of the two on the thermoelectric transport is not exactly the same. The influence of the RSOI phase is related to electron spin, which results in some interesting properties of thermoelectric transport. Figure 2 shows the thermoelectric coefficients as functions of the QD level for t = 2Γ and φ = 1.25π, ΔϕR = 0.25π. In general, the conductance should exhibit three resonance peaks corresponding to the three molecule-state levels. However, under the configuration of the magnetic flux phase and the RSOI phase, the coexistence of BIC and Fano resonance can occur for appropriate matching of the two phases. G (κel↓) presents an asymmetric resonance peak near the bonding state due to the Fano resonance in transmission, and the peak at ε = 0 merges with the peak near the antibonding state , forming a wider peak, see Figs. 2(a) and 2(b). While for G (κel↑), BIC appears near the antibonding state and the Fano resonance does not occur near the bonding state, which leads to large spin thermopower appearing in a wide range of energy. Figures 2(c) and 2(d) show the thermopower and the figure of merit as functions of the QD level. Due to the Fano resonance in the vicinity of the bonding state, the spin-dependent thermopower S presents a relatively large value, as a result, the spin thermopower Ss also presents a relatively large value, and the spin figure of merit ZsT is even larger than the charge figure of merit ZcT.

Fig. 2. Thermoelectric coefficients (a) G, (b) κ, (c) S, and (d) ZT as functions of the QD level ε for t = 2Γ. Here the phase factors are set to be φ = 1.25π and ΔϕR = 0.25π.

In order to observe comprehensively the modulation of the configuration of the magnetic flux phase and the RSOI phase on the thermoelectric coefficients, we present the thermopower as a function of both the magnetic flux phase φ and the RSOI phase ΔϕR in Fig. 3. It is convenient to consider the contour graphs as a function of two variables (φϕR) (the other parameters are ε = 0.01 and t = 2Γ). The spin-dependent thermopower Sσ presents a double-peak structure and these peaks locate at (φϕR) = (x, δσx + 4n ± 1/4), where n = 0,±1,±2,…, and x ∈ (−∞, ∞), see Figs. 3(a) and 3(b). The charge thermopower Sc and the spin thermopower Ss are made up of S and S. They present a quadruple-peak structure near (φϕR) = (4n,4m) (n,m = 0,±1,2,…) for Sc and near (φϕR) = (4n + 2,4m + 2) (n,m = 0,±1,2,…) for Ss. More precisely, these peaks, as shown in Figs. 3(c) and 3(d), locate at (φϕR) = (4n,4m±1/4) and (4n±1/4,4m) for the charge thermopower Sc, and at (φϕR) = (4n + 2,4m + 2±1/4) and (4n + 2±1/4,4m + 2) for the spin thermopower Ss, which can be used to optimize the thermoelectric efficiency. One can also find from Figs. 3(c) and 3(d) that for some specific configurations of the magnetic flux phase and the RSOI phase, when Ss is close to its peak value, Sc tends to zero, meaning that a pure spin current can be obtained near these specific configurations of the phases. In order to see this more clearly, in Fig. 4, we plot the thermopower S as a function of ε for (φϕR) = (−1.68,−1.96) (point M in Fig. 3). For this configuration of the phase, the amplitudes of S and S are equal at ε = 0.01, but present opposite signs. As a result, Sc vanishes while Ss ≈ 1 near ε = 0.01, the corresponding figure of merit is only 0.1. Further optimizing the operating parameters, one can obtain larger Ss and corresponding ZsT. For example, for(φϕR) = (1.25,0.25), Ss ≈ −1.5 and ZsT ≈ 0.6 near ε = −3, see Figs. 2(c) and 2(d). Hereby, under appropriate matching of the parameters, the system could be considered as an effective spin battery that can convert heat into a spin voltage and induce a pure spin current in the device. Moreover, the interdot tunneling coupling t also has an obvious effect on the thermoelectric properties of the system.

Fig. 3. Contour graphs of spin-dependent thermopower Sσ, charge thermopower Sc, and spin thermopower Ss as functions of the magnetic flux phase φ and the SROI phase ΔϕR for ε = 0.01Γ. The other parameters are the same as those in Fig. 2.
Fig. 4. Thermopower S as a function of the QD level for phase factor φ = −1.68π and ΔϕR = −1.96π. The other parameters are the same as those in Fig. 2.

For a given configuration of the magnetic flux phase and the RSOI phase, one can obtain maximal thermopower by choosing the appropriate interdot coupling. In Fig. 5, we plot the spin thermopower Ss as a function of the interdot tunneling coupling t for the configuration of the magnetic flux phase and the RSOI phase. In the vicinity of t = 0.6 Γ, the spin thermopower reaches its maximum .

Fig. 5. Spin thermopower Ss as a function of t for ε = 0.01Γ. The other parameters are the same as those in Fig. 4.
4. Conclusion

We have studied the spin-dependent thermoelectric transport through a parallel triple-dot system with RSOI embedded in an Aharonov–Bohm interferometer connected symmetrically to leads using the nonequilibrium Green’s function method in the linear response regime. Because the influences of the magnetic flux phase and the RSOI phase on the thermoelectric transport are not exactly the same, the thermopower can be considered as a function of both the magnetic flux phase φ and the RSOI phase ΔϕR when other parameters are given. The contour graphs of spin-dependent thermopower Sσ present double-peak structures in the specific (φϕR) region. While the charge thermopower Sc and the spin thermopower Ss present quadruple-peak structures. For some specific configurations of the magnetic flux phase and the RSOI phase, one can obtain the spin thermopower while the charge thermopower vanishes, which is useful in realizing the thermal spin battery that would allow to convert heat into a spin voltage and induce a pure spin current in the device. Moreover, under the appropriate configuration of the phase and parameters, one can obtain a relatively large spin thermopower and the spin figure of merit ZsT is even larger than the charge figure of merit ZcT.

Reference
1Majumdar A 2004 Science 303 777
2Hicks L DDrfesselhaus M S 1993 Phys. Rev. 47 016631
3Minnich A JDresselhaus M SRen Z FChen G 2009 Energy Environ. Sci. 2 466
4Bulka BKönig JPekola 2008 Phys. Rev. Lett. 100 066801
5Boukai A IBunimovich YTahir-Kheli JYu J KGoddard III W AHeath J R 2008 Nature 451 168
6Scheibner RBuhmann HReuter DKiselev M NMolenkamp L W 2005 Phys. Rev. Lett. 95 176602
7Scheibner RKOṅnig MReuter DWieck A DGould CBuhmann HMolenkamp L W 2008 New J. Phys. 10 083016
8Kubala BKOṅnig JPekola J 2008 Phys. Rev. Lett. 100 066801
9Kuo D M TChang Y C 2010 Phys. Rev. 81 205321
10Tsaousidou MTriberis G P 2010 J. Phys.: Condens. Matter 22 355304
11Zianni X 2007 Phys. Rev. 75 045344
12Wierzbicki MSwirkowicz R 2011 Phys. Rev. 84 075410
13Crepieux ASimkovic FCambon BMichelini F 2011 Phys. Rev. 83 153417
14Liu JSun Q FXie X C 2010 Phys. Rev. 81 245323
15Ludoph BVan Ruitenbeek J M 1999 Phys. Rev 59 12290
16Reddy PJang S YSegalman R AMajumdar A 2007 Science 315 1568
17Baheti KMalen J ADoak PReddy PJang S YTilley T DMajumdar ASegalman R A 2008 Nano Lett. 8 715
18Murphy P GMukerjee SMoore J E 2008 Phys. Rev. 78 161406
19Nozaki DSevincli HLi WGutierrez RCuniberti G 2010 Phys. Rev. 81 235406
20Finch C MGarcía-Suárez V MLambert C J 2009 Phys. Rev. 79 033405
21Markussen TJauho A PBrandbyge M 2009 Phys. Rev. 79 035415
22Liu Y SChen Y C 2009 Phys. Rev. 79 193101
23Xue H J T QZhang H CYin H TCui LHe Z L 2011 Chin. Phys. 20 027301
24Dubi YDi Ventra M 2011 Rev. Mod. Phys. 83 131
25Uchida KTakahashi SHarii KIeda JKoshibae WAndo KMae-kawa SSaitoh E 2008 Nature 455 778
26Uchida KOta THarii KAndo KSasage KNakayama HIkeda KSaitoh E 2009 IEEE Trans. Magn. 45 2386
27Xue H J T QZhang H CYin H TCui LHe Z L 2012 Chin. Phys. 21 037201
28Zheng JChi FGuo Y 2012 J. Phys.: Condens. Matter 24 265301
29Uchida KXiao JAdachi HOhe JTakahashi SIeda JOta TKajiwara YUmezawa HKawai HBauer G E WMaekawa SSaitoh E 2010 Nat. Mater. 9 894
30Jaworski C MYang JMack SAwschalom D DHeremans J PMyers R C 2010 Nat. Mater. 9 898
31Dubi YDi Ventra M 2009 Phys. Rev. 79 081302
32Swirkowicz RWierzbicki MBarnas J 2009 Phys. Rev. 80 195409
33Wang QXie H QJiao H JLi Z JNie Y H 2012 Chin. Phys. 21 117310
34Zheng JChi F 2012 J. Appl. Phys. 111 093702
35Wierzbicki MSwirkowicz R 2011 Phys. Rev. 84 075410
36Trocha PBarnaś J 2012 Phys. Rev. 85 085408
37Zheng JZhu M JChi F 2012 J. Low Temp. Phys. 166 208
38Gómez-Silva GÁ valos-Ovando Ode Guevara M L LOrellana P A 2012 J. Appl. Phys. 111 053704
39Wang QXie HNie Y HRen W 2013 Phys. Rev. 87 075102
40Guevara M L L DOrellana P A 2006 Phys. Rev. 73 205303
41Vallejo M Lde Guevara M L LOrellana P A 2010 Phys. Lett. 374 4928