| The increasingly serious energy crisis and dramatically deteriorated environment make it urgent to develop new clean energy technology.Thermoelectric(TE)material is regarded as one of the most promising alternatives,since it can convert waste heat into electrical energy directly.Here,the TE conversion efficiency can be generally evaluated by a dimensionless figure-of-merit ZT=S2σT/(κe+κL),where S,σ,T,κe and κL are the Seebeck coefficient,the electrical conductivity,the absolute temperature,the electronic and lattice thermal conductivity,respectively.However,it is usually difficult to significantly improve the ZT value of a given material since those transport coefficients are inherently coupled with each other.In this thesis,we give a complete study of the structural,phonon,electronic,and TE transport properties of several superlattices and van der Waals heterostructures(vdWHs)constructed by stacking some two-dimensional metarials in different styles(periodic or nonperiodic along vertical or horizontal directions).On the other hand,to accelerate the screening of high performance TE materials,we adopt machine learning(ML)approach to propose an effective descriptor by which the energy gap of vdWHs can be predicted in a high-throughput way.The main content of the thesis includes:To improve the TE performance of transition metal dichalcogenides,we construct a van der Waals superlattice(vdWSL)formed by periodic stacking of the WS2 and WTe2 monolayers along the vertical direction.Both the ab-initio molecular dynamics simulation(AIMD)and the phonon spectrum calculations indicate that the superlattice is rather stable.Due to the coexistence of strong covalent(W-S,W-Te)and weak van der Waals bonding(S-Te),the superlattice exhibits large anharmonicity and ultralow lattice thermal conductivity(0.12 W/mK@800 K).Besides,the system displays a good electronic transport performance caused by the weak electron dispersion and high band degeneracy along the vertical direction.As a result,the ZT value of n-type vdWSL can reach 2.4 at 800 K,which is rather rare for the transition metal dichalcogenides materials.In addition to perpendicular stacking,we design a BAs/BSb two-dimensional(2D)superlattice by periodic arrangement of BAs and BSb nanoribbons along the horizontal direction.As there is no imaginary frequency in the phonon spectrum,such a 2D structure is kinetically stable.Due to the coexistence of six different covalent bonds in the system,the lattice thermal conductivity of the BAs/BSb superlattice is much lower than those of the BAs and BSb monolayers.In addition,the superlattice exhibits a moderate band gap(0.79 eV)and high carrier mobility,which suggests its good electronic transport.As a consequence,the ZT value of n-type superlattice can be optimized to～3.1 at 800 K.Meanwhile,the ZT of the p-type system can reach～2.9,which is beneficial for the TE modules that need both p-and n-type legs with comparable conversion efficiency.Unlike periodic superlattices,we can also construct nonperiodic vdWHs by stacking two-dimensional materials.On the other hand,most good TE materials are composed of rare elements,which suggests a relatively higher cost.Considering these two points,we investigate a vdWH formed by stacking the black phosphorus and blue phosphorus monolayers along the vertical direction.It is found that the black phosphorus/blue phosphorus vdWH is rather stable even at higher temperature.Compared with those of the two constituent monolayers,the κL of the vdWH is significantly reduced caused by the weak layer-layer interactions.Besides,the vdWH has good electronic transport performance due to type-Ⅱ band alignment,reduced band gap,and multi-valley structure.As a result,the ZT value of n-type system can reach 1.6 at 300 K,which is much higher than those of the two components.In addition,the ZT value of n-type system exceeds 2.0 in the temperature range from 400 to 800 K,and the maximum ZT value can reach 3.2 at 700 K.Due to the presence of earth-abundant and light phosphorus,the black phosphorus/blue phosphorus vdWH could find promising thermoelectric applications in a wide temperature range.In the above works,we focus on the TE properties of few specific superlattices or vdWH.To effectively search high performance TE materials,we adopt a ML method(compressed sensing)to propose a three-dimensional(3D)descriptor by which the band gap of vdWH can be obtained in a high-throughput style.The vdWH is formed by stacking the graphene-like monolayers of group-IIIA,IVA,and VA elements,with a total of 325 systems as the training data.The strong predictive power of our descriptor is demonstrated by high Pearson correlation coefficient for both the training(94%)and testing(92%)sets.By utilizing such a 3D descriptor,we have rapidly predicted the band gaps of 7140 possible vdWHs,including 6331 systems exhibiting small gaps(0～1 eV),which offers a large space to screen high performance TE heterostructures.Taking the BSb/AsP vdWH as an example,the descriptor predicts a gap of 0.89 eV.Further first-principles calculations reveal that,in a wide range of carrier concentration,the vdWH can have high Seebeck coefficients with comparable values for the p-and n-type systems.At a temperature of 700 K,the ZT values of both p-and n-type systems can be optimized to～3.3,suggesting its very promising TE performance operated in the intermediate temperature region. |
