2026 Volume 35 Issue 3
Article Contents

Yayu Wang, Jue Hou, Ming Yang, Xingli Zhang. Reducing lattice thermal conductivity via phonon engineering: Strategies for high-performance thermoelectrics[J]. Chinese Physics B, 2026, 35(3): 037201. doi: 10.1088/1674-1056/ae1c22
Citation: Yayu Wang, Jue Hou, Ming Yang, Xingli Zhang. Reducing lattice thermal conductivity via phonon engineering: Strategies for high-performance thermoelectrics[J]. Chinese Physics B, 2026, 35(3): 037201. doi: 10.1088/1674-1056/ae1c22

Reducing lattice thermal conductivity via phonon engineering: Strategies for high-performance thermoelectrics

  • Received Date: 03/09/2025
    Accepted Date: 21/10/2025
    Available Online: 15/03/2026
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(10)  /  Tables(1)

Article Metrics

Article views(449) PDF downloads(2) Cited by(0)

Access History

Reducing lattice thermal conductivity via phonon engineering: Strategies for high-performance thermoelectrics

Abstract: 

Thermoelectric materials convert heat directly into electricity and are therefore promising for energy harvesting and environmental applications. Ideal high-performance thermoelectrics combine ultralow lattice thermal conductivity, κL, with high carrier mobility, a paradigm commonly termed phonon-glass electron-crystal. However, strong coupling between electronic and phononic transport complicates simultaneous optimization of these properties. Because κL is largely independent of electronic transport, targeted suppression of κL is an effective route to partially decouple heat and charge transport. This review summarizes recent advances in reducing κL via two complementary approaches: phonon engineering of bulk nanostructured systems and phonon engineering of low-dimensional materials. In bulk systems, κL may be minimized while retaining high electrical conductivity and maximizing the thermoelectric figure of merit ZT by controlling three fundamental phonon parameters: the volumetric specific heat cv, the phonon group velocity vg, and the phonon relaxation time τ. Low-dimensional architectures, including superlattices, nanowires, and nanocomposites, supply additional levers to suppress lattice heat transport and to tailor the electronic structure. Integrating multiscale and multimodal phonon-control strategies enables significant reductions in κL without sacrificing electronic performance, thereby advancing the phonon-glass electron-crystal paradigm.

1.   Introduction
  • The accelerating global shortage of resources and mounting environmental concerns are creating an urgent need for sustainable clean-energy technologies. Thermoelectric materials and devices — which contain no moving parts — can directly convert heat into electricity via the Seebeck effect (Fig. 1),[1] or convert electrical energy into heating or cooling via the Peltier effect (Fig. 1).[1] Consequently, thermoelectric devices can harvest thermal energy from sources such as solar heat and industrial waste heat, thereby improving overall resource-use efficiency. Owing to these attributes, thermoelectrics have attracted considerable interest and found applications in industrial waste-heat recovery,[2] power generation from natural heat sources,[3] wearable thermal energy systems,[4] automotive waste-heat recovery,[5] thermoelectric refrigeration,[6] microprocessor cooling, and cooling of medical devices.[7,8] They are therefore considered promising solutions to energy supply and environmental challenges.

    The performance of a thermoelectric material is quantified by the dimensionless figure of merit ZT,[9] defined as ZT = TS2σ/k, where T is the temperature, S is the Seebeck coefficient, σ is the electrical conductivity, and κ is mainly composed of the electronic and lattice thermal conductivities (κe and κL).[10] High ZT requires a large power factor (S2σ) combined with low κ; however, S, σ and κ are mutually interdependent, so improving one quantity typically perturbs the others. In particular, S and σ vary inversely with carrier concentration n, producing a fundamental trade-off in device optimization.[11] Thus, strategies that partially decouple electronic and phononic transport are essential to increase ZT.

    Among the transport coefficients, the lattice thermal conductivity κL is relatively independent and critically important for thermoelectric performance;[12,13] reducing κL often yields the largest practical gains in ZT.[14–16] From kinetic theory, κL depends on the phonon volumetric specific heat cv, the phonon group velocity vg, and the phonon relaxation time τ.[14] Accordingly, three principal routes exist to reduce κL: (i) shorten τ tuned by defect engineering (e.g., zero-dimensional point defects,[17–19] one-dimensional dislocations,[20–22] and two-dimensional interfaces such as grain boundaries, nanotwins, and stacking faults[23–26]), (ii) reduce vg via lattice softening or heavy-atom substitution,[27–29] and (iii) lower cv (for example, via superionic behavior[30]). Although these strategies have been demonstrated effectively in three-dimensional materials, a key limitation remains: the inverse relationship between S and σ with carrier concentration constrains further increases in ZT.[31] In the late 1990s Hicks and Dresselhaus[32] proposed that reduced dimensionality can increase the electronic density of states near the Fermi level, thereby enhancing S, while surface and interface scattering can preferentially suppress phonon transport with limited degradation of σ.[33–35] This insight motivates two complementary nanoscale paradigms for improving thermoelectric performance: the phonon engineering of bulk nanostructured materials and the phonon engineering of low-dimensional nanomaterials.

    This review summarizes recent progress in suppressing the κL of thermoelectric materials, with emphasis on two complementary approaches: phonon engineering of bulk nanostructured systems and phonon engineering of low-dimensional nanomaterials. In bulk systems, κL may be minimized while preserving high electrical conductivity and maximizing the thermoelectric figure of merit ZT by tuning three primary phonon parameters: the volumetric specific heat cv, the phonon group velocity vg and the phonon relaxation time τ. We then examine how low-dimensional architectures, including superlattices, nanowires and nanocomposites, further reduce lattice heat transport and thereby enhance thermoelectric performance. Finally, we highlight remaining challenges and opportunities for translating these strategies into practical, scalable thermoelectric devices.

2.   Phonon engineering for bulk nanostructured materials
  • Phonon relaxation time τ can be tuned by introducing scattering centers of different dimensionalities, including zero-dimensional (0D) point defects, one-dimensional (1D) line defects, and two-dimensional (2D) planar defects.

  • Point defects are atomic-scale lattice perturbations, typically classified as vacancies, substitutional impurities, or interstitials. They influence electronic transport by altering the carrier concentration (e.g., the Hall carrier concentration nH) and efficiently scatter short-wavelength phonons via local mass and force-constant disorder, thereby reducing the lattice thermal conductivity.[36] The effect of point defects on phonon relaxation can be described within the Debye–Callaway framework,[37] in which the defect-induced phonon scattering rate often follows a Rayleigh-type frequency dependence (approximately ∝ ω4) for small defects.[38] For substituted elements (point defects), the associated phonon scattering rate is often written as τp−1=Vω44πν3Γ,[17] where Γ is the scattering parameter. According to the models of Slack and Abeles,[39,40]Γ can be decomposed into mass-fluctuation and strain-field contributions, Γ = ΓM+ ΓS, where ΓM accounts for mass disorder (mass contrast between host and substituent atoms) and ΓS describes scattering due to local strain/force-constant perturbations. These terms can be evaluated from the defect concentrations and the corresponding mass and size/elastic mismatches. Therefore, the strong mass fluctuation and strain-field fluctuation induced by point defects can significantly reduce κL.[41]

    In the Cu2SnSe3 system (Fig. 2(a)),[17] partial substitution of Cu by different dopants produces markedly different effects on the lattice thermal conductivity κL. Replacing 10% of Cu with Ni or Zn yields only a negligible change in κL, whereas 10% substitution by heavier or more size-mismatched elements such as Cd, Ag, or Al causes a pronounced reduction in κL. This behavior is primarily attributable to stronger mass-fluctuation and strain-field scattering induced by the latter dopants. Figure 2(b) shows that introducing extrinsic defects (Cd, Ga, In, Mn, Ge, Zn, and Al) systematically suppresses κL across the entire measured temperature range; notably, substituting 10% of Sn with Al reduces κL by ∼ 71.5% at 300 K and by ∼ 50.9% at 850 K.[17] Wang et al.[18] experimentally demonstrated that Pb incorporation into cubic SnSe synergistically improves thermoelectric transport by concurrently increasing the hole concentration and reducing the lattice thermal conductivity κL. At a Pb content of x = 0.08, the peak power factor at 300 K increases to 20 μW⋅cm−1⋅K−2. The atomic-size mismatch introduced by Pb generates enhanced internal microstrain, which elevates phonon scattering and lowers κL; with increasing Pb concentration, lattice-distortion (strain) fluctuations intensify, further promoting phonon attenuation. These local bonding inhomogeneities and the associated increase in lattice anharmonicity therefore produce additional reductions in κL. A complementary and increasingly adopted strategy to suppress κL is targeted doping with atoms that possess stereochemically active lone-pair s electrons, which induce local bonding disorder and amplify lattice anharmonicity.[43] Although compounds that intrinsically contain active lone pairs are scarce, intentional incorporation of lone-pair dopants can create localized low-frequency vibrational modes and severe local distortions, yielding phonon scattering that far exceeds predictions from simple mass-fluctuation or point-defect models. The effectiveness of this “anharmonic-atom” doping mechanism has been demonstrated in CuFeS2 chalcopyrite. Substitution of Cu by In (which carries stereochemically active lone pairs) distorts the local coordination away from ideal tetrahedral symmetry (Fig. 2(c)) and induces localized low-frequency modes that couple strongly to the host phonon spectrum, thereby substantially increasing lattice anharmonicity. As shown in Fig. 2(d), only 3 at.% In on the Cu site reduces κL by > 40%.[42] A comparable phenomenon occurs in Bi-doped GeTe, where 10 at.% Bi yields ∼ 37% reduction in κL at 740 K, an effect attributed to the strong anharmonicity introduced by Bi lone-pair cations.[42]

    In addition to elemental doping, an appropriate concentration of vacancies can optimize carrier scattering, thereby balancing electrical conductivity σ and the Seebeck coefficient S to improve the power factor S2σ.[44,45] Vacancies also introduce additional phonon-scattering centers, shortening phonon mean free paths and slowing heat transport, which effectively reduces the lattice thermal conductivity κL. For example, in polycrystalline SnSe, α-radiation–induced intrinsic Sn vacancies enhance phonon scattering and reduce κL to 0.22 W⋅m−1⋅K−1, yielding a high ZT of 2.4 at 800 K.[19] Similarly, Chen et al.[46] obtained an ultralow κL of 0.4 W⋅m−1⋅K−1 in a SnTe–AgSbTe2 alloy by increasing the vacancy concentration, underscoring the important role of vacancy defects in enhancing thermoelectric performance.

    Various point-defect engineering strategies have been developed to suppress the κL of thermoelectric materials. In addition to conventional elemental doping and vacancy engineering, high-entropy design, in which multiple distinct elements share a single crystallographic sublattice, has emerged as an effective and widely adopted strategy.[47,48] Entropy engineering acts via two synergistic mechanisms: (1) entropy-driven structural stabilization, which can preserve or even enhance electronic transport by stabilizing higher-symmetry single-phase structures and promoting band convergence; and (2) severe lattice distortion and chemical/force-constant disorder, which induce strong mass- and strain-field fluctuations that generate dense, multiscale phonon scattering and dramatically reduce κL.[49] These combined effects enable high-entropy compounds to achieve unusually low lattice thermal conductivities while retaining favorable electronic properties (e.g., long-range carrier transport in an ordered lattice coexisting with atom-scale disorder). Thermodynamically, the high-entropy effect is rooted in the Gibbs free-energy relation ΔG = ΔH – TΔS < 0, where ΔH, T and ΔS denote the enthalpy change, absolute temperature, and configurational entropy change, respectively.[50]

    Increasing configurational entropy lowers ΔG and can therefore stabilize single-phase (often high-symmetry) structures over multiphase mixtures, as illustrated schematically in Fig. 3(a).[50] The absence of secondary-phase interfaces reduces carrier scattering at phase boundaries and — when combined with enhanced band degeneracy in the stabilized phase — supports excellent electronic transport. Concurrently, the severe lattice-distortion effect produced by atomic-radius and bonding mismatches increases lattice anharmonicity, preferentially scattering mid- to high-frequency phonons and further suppressing κL. Moreover, lattice distortion and chemical disorder can impede long-range defect motion (e.g., dislocation glide), which modifies mechanical behavior and can help sustain the engineered microstructure. Within the complex high-entropy matrix, synergistic interactions among multiple elements therefore enhance both thermoelectric and mechanical properties (Fig. 3(b)).[50]

    A representative example is GeTe-based high-entropy alloys and compounds. The band structure of cubic GeTe is analogous to those of PbTe and SnTe: the valence-band maximum (VBM) and conduction-band minimum (CBM) occur at the L point of the Brillouin zone (fourfold valley degeneracy, Nv = 4), yielding a direct band gap at L. However, a secondary valence band along the Σ direction lies close in energy to the L band; this Σ band has a high nominal degeneracy (Nv = 12) and in cubic GeTe the L–Σ energy offset ΔEL–Σ is small (on the order of 60–70 meV).[51,52] Such a small offset facilitates thermal population and band convergence at elevated temperature, providing a route to increase the effective density-of-states mass and thereby to enhance the Seebeck coefficient and power factor. By contrast, the L–Σ offsets reported for PbTe and SnTe are larger (∼ 100 meV and ∼ 300 meV, respectively), which renders band convergence less accessible thermally.[53,54] Directional lattice distortion (for example, the rhombohedral distortion along [111]) perturbs the L-point symmetry and shifts the relative energies of the L and Σ bands; in the distorted (rhombohedral) phase the Σ-derived band can become the valence-band maximum, while the L manifold is split by symmetry lowering. Such splitting of valley degeneracies in the low-symmetry phase hinders optimal band convergence and explains why the high-symmetry cubic phase is generally the superior thermoelectric platform at elevated temperatures.[55]

    In a representative study (Fig. 4), alloying LiSbTe2 into GeTe was found to stabilize the cubic phase across 300–773 K.[56] During alloying, Li and Sb partially substitute on Ge sites in both the cubic and rhombohedral phases, which suppresses the stereochemical activity of the cation lone pair and lowers the rhombohedral–cubic transition temperature. This stabilization of the cubic phase promotes multiband convergence and enhances phonon scattering, leading to concurrently improved electronic transport and reduced lattice thermal conductivity. Consequently, the alloy attains a high average thermoelectric figure of merit, ZTavg ≈ 1.38, over the 300–773 K range.[56] Wang et al.[56] synthesized single-phase high-entropy alloys in the Ge0.64Ag0.12Sb0.12X0.12Te system (X = Sn, Mn, Pb). Analysis of these GeTe-based samples revealed that local chemical fluctuations are governed by the intrinsic properties of the alloying elements — primarily differences in atomic mass, ionic radius and electronegativity. Beyond increasing configurational entropy via multicomponent alloying, the degree of local structural disorder can be further tuned by varying the elemental composition. Larger elemental mismatch produces enhanced lattice distortions and strain fields, which intensify phonon scattering and substantially suppress the lattice thermal conductivity. Replacing Sn or Mn with Pb enhanced elemental fluctuations and reduced the room-temperature lattice thermal conductivity from ∼ 0.88 W⋅m−1⋅K−1 and ∼ 0.73 W⋅m−1⋅K−1 to ∼ 0.61 W⋅m−1⋅K−1, respectively — substantially lower than that of pristine GeTe.[57] By tuning the alloy composition to optimize the carrier concentration, the GeTe–AgSbPb composition achieved a peak figure of merit, ZTmax, that increased from ∼ 1.9 to ∼ 2.4, demonstrating the potential of the high-entropy strategy for achieving high-performance thermoelectric materials.[57]

  • Point (0D) defects produce Rayleigh-like scattering with a scattering rate proportional to ω4, and therefore predominantly scatter high-frequency phonons. One-dimensional defects (dislocations) give rise to a scattering rate that scales approximately linearly with frequency, i.e., τD ∼ ω−1, and are thus particularly effective at scattering mid-frequency phonons.[58] The role of dislocations in thermoelectric materials — especially their impact on thermal transport — has therefore attracted considerable interest.

    The contributions of dislocation cores and of their surrounding strain fields to phonon scattering (Fig. 5) can be expressed phenomenologically as[20,59]

    where ND is the dislocation density, r is the effective core radius, BD the Burgers vector, γ the Grüneisen parameter, ω the phonon angular frequency and v the phonon group velocity. Thus τDC ∝ ω−3 and τDS ∝ ω−1, consistent with preferential scattering of mid-frequency phonons by dislocations and of higher-frequency modes by point defects. Such dislocation-mediated scattering efficiently reduces the contribution of mid-frequency phonons to heat transport, thereby lowering the κL (Figs. 5(a) and 5(b)).[20] Dislocation-induced reductions in κL become substantial when the dislocation density exceeds approximately 1012 cm−2.[59,60] Achieving such high densities without degrading electrical transport remains a central challenge. Therefore, dislocation engineering that introduces dense dislocation networks and the associated strain fields provides an effective route to suppress κL and improve thermoelectric performance.

    In conventional engineering materials, dislocations are typically introduced by plastic deformation to improve mechanical properties.[62] However, most thermoelectric materials are brittle, so conventional deformation rarely generates dislocations.[60] Consequently, dislocations in thermoelectric semiconductors are usually introduced via specialized techniques, for example, interstitial-atom incorporation,[21] vacancy engineering,[22] liquid-phase extrusion,[63] hot deformation,[64] alloying[18] and ultrahigh-pressure sintering.[65]

    Interstitial-atom incorporation inserts atoms into lattice interstitial sites, producing local strain that promotes dislocation nucleation. Zhang et al.[21] showed that Cu doping in Pb0.9985Gd0.0015Se leads to aggregation of Cu interstitials, nucleation of high-density dislocation networks, and a substantial reduction of κL. However, dislocation densities attainable by interstitial or vacancy incorporation alone are constrained by dopant solubility in the host matrix, which makes it difficult to reach ultrahigh densities (e.g., ND > 1015 m−2) in thermoelectric materials. To overcome this limitation, Xu et al.[61] developed a two-step strategy. First, heavy Te–S coalloying in a PbSe matrix produces large lattice distortions that increase the driving force for dislocation formation. Second, Cu interstitial doping in the optimized Pb1.02Se0.72Te0.20S0.08 composition further accelerates dislocation nucleation. By these successive steps, the Pb1.02Se0.72Te0.20S0.08 sample doped with 0.3 at.% Cu attained an ultrahigh dislocation density of 5.4 × 1016 m−2, substantially exceeding previously reported values for n-type PbSe. The dense dislocation network yields ultralow lattice thermal conductivity over the measured temperature range, with κL reduced to 0.42 W⋅m−1⋅K−1 at room temperature and a minimum κL≈ 0.29 W⋅m−1⋅K−1 at 573 K.[61] Importantly, these dislocations have only a modest effect on carrier transport: the material retains a reasonable weighted mobility and exhibits markedly improved thermoelectric performance, reaching a room-temperature ZT of 0.62 and an average ZT ≈ 0.9 over 300–573 K. High-angle annular dark-field STEM imaging and strain mapping reveal detailed dislocation structures and their strain fields (HAADF-STEM, GPA and EDS results are presented in Fig. 5).[61] Individual edge dislocations were directly observed with Burgers vector (b = 1/2[0 1 1]), and geometric phase analysis (GPA) shows pronounced linear strain fields in the εxx and εyy components (edge-dislocation image and Burgers circuit: Figs. 5(c) and 5(d); GPA: Figs. 5(e) and 5(f)). Energy-dispersive x-ray spectroscopy (EDS) mapping indicates Cu enrichment near edge dislocations, supporting the role of Cu interstitials in promoting nucleation (Fig. 5(g)). A screw dislocation with Burgers vector (b = 1/3[1 1 1]) was also identified and exhibits strong localized strain fields (HRTEM/HAADF image: Fig. 5(h)). The larger spatial extent of screw-dislocation strain fields (GPA: Figs. 5(i) and 5(j)) likely produces stronger phonon scattering. The contrast between edge- and screw-type strain fields highlights how tailored dislocation architectures can be engineered to suppress lattice heat transport in thermoelectric materials.

    In vacancy engineering, intentionally introduced cation vacancies accumulate in the crystal lattice and produce local strain that promotes dislocation nucleation, often during subsequent thermal annealing. Mg2Sn single crystals grown by melting intrinsically contain Mg vacancies and dislocation cores; both undoped and B-doped specimens therefore show reduced lattice thermal conductivity. For example, undoped crystals prepared by melting display κL ≈ 4.3 W⋅m−1⋅K−1 at 300 K, substantially lower than ∼ 7.3 W⋅m−1⋅K−1 reported for other undoped samples.[66,67] Huang et al.[22] further decoupled carrier and phonon transport by B doping, thereby improving thermoelectric performance. Boron increases both the Mg-vacancy concentration and the density of dislocation cores: vacancy clustering and the strain fields induced by doping nucleate dislocations whose cores and surrounding strain fields effectively scatter mid-frequency phonons. Consequently, the lattice thermal conductivity of Mg2Sn single crystals can be driven toward theoretical minima (κL min ≈ 0.65 W⋅m−1⋅K−1).[22]

    Liquid-phase extrusion expels excess elements during sintering, inducing lattice stress that promotes dislocation formation. Jung et al.[63] synthesized Bi1.7Sb0.3Te3 by transient liquid-phase sintering (LPS) to simultaneously optimize σ and κL. Te-rich Bi1.7Sb0.3Te3 ingots were melt-spun into powders with excess Te; these powders undergo LPS at ∼ 723 K. Applying external pressure during LPS facilitates the removal of excess Te, causing rapid pellet shrinkage and the formation of dense, interwoven dislocation networks. The resulting interwoven dislocations at grain boundaries strongly scatter phonons, reducing κL to 0.4 W⋅m−1⋅K−1 at 300 K. Concurrently, the Te-rich environment promotes Bi-on-Te antisite formation, which enhances electronic transport and increases σ to 2.2 × 105 S⋅m−1 at 300 K, thereby markedly improving thermoelectric performance.

    Hot deformation refers to mechanical processing at elevated temperature (e.g., hot pressing) that enhances ductility and thereby promotes dislocation formation. Chen et al.[64] prepared Bi0.5Sb1.5Te3 (BST) polycrystals by zone melting followed by hot deformation. Zone-melted specimens exhibited a well-ordered, defect-free layered microstructure, whereas hot-deformed samples developed dense dislocation networks. These networks strongly scatter mid-frequency phonons, producing a marked reduction in the κL. Geometric phase analysis (GPA) of the εxx strain component revealed substantial strain fields introduced by the deformation. Such multiscale defect architectures, comprising dislocations and grain boundaries, serve as highly effective phonon-scattering centers and substantially suppress κL.

    Alloying and doping similarly modify the local coordination environment: atomic-size mismatch between host and dopant induces strain that nucleates dislocations. For example, Wang et al.[18] synthesized Pb-doped SnSe by vacuum melting, mechanical activation (MA) and hot-press sintering (HPS). The atomic-size mismatch introduced by Pb generates strain fluctuations that nucleate additional dislocation arrays, thereby enhancing phonon scattering and improving thermal transport.

    Ultrahigh-pressure (UHP) sintering produces shear stresses that exceed the threshold for dislocation nucleation in ceramics, thereby enabling extensive dislocation glide and multiplication. Unlike metals and metavalently bonded chalcogenides, ceramic oxides typically resist formation of high-density dislocations under conventional processing owing to strong covalent/ionic bonding and intrinsic brittleness. UHP sintering at gigapascal pressures overcomes this limitation: experiments indicate that GPa-level shear stresses are sufficient to break atomic bonds and nucleate dislocations. Yin et al.[65] reported a dislocation density of approximately 9.1 × 1016 m−2 in Bi0.96Pb0.04CuSeO ceramics sintered at 2 GPa using UHP processing. Inverse fast-Fourier-transform (IFFT) analysis reveals geometrically necessary dislocations at grain boundaries that produce pronounced εxx and εxy strain fields. High-resolution analysis locates edge dislocations with Burgers vector b = a/2[1 1 1]; these dislocations interconnect and coalesce into mesoscale strain domains. The resulting dense dislocation networks strongly suppress phonon transport: the κL of BiCuSeO falls to 0.13 W⋅m−1⋅K−1 at 767 K, approaching the material’s diffusive limit (κdiff ≈ 0.07), while the compound reaches a record thermoelectric figure of merit, ZT ≈ 1.69 at 767 K, surpassing previously reported BiCuSeO-based materials.[65]

    To summarize, effectively introducing dislocations enhances mid-frequency phonon scattering, thereby substantially reducing lattice thermal conductivity (κL). By strategically selecting dislocation engineering methods, both thermoelectric and mechanical properties can be optimized, enabling the development of high-performance thermoelectric materials.

  • Interfacial defects are also concerned because of their strong scattering effect on low-frequency phonons (τInter ∼ ω0).[58] The most common two-dimensional (2D) defects (interfacial defects) in thermoelectric materials include grain boundaries, twin boundaries, and stacking faults.[1,68–70]

    Grain boundaries, which are the interfaces between grains in polycrystalline materials, play a critical role in phonon scattering.[71] Figure 6(a) shows how thermal conductivity varies with temperature for single-crystal and polycrystalline PbSe.[13,72] While single crystals effectively have infinite grain size, reducing the grain size to the micrometer scale lowers PbSe’s lattice thermal conductivity from 4.5 W⋅m−1⋅K−1 to approximately 30% of its original value. This reduction arises primarily from the enhanced phonon scattering at grain boundaries.

    Grain growth is a diffusion-driven process in which grain boundaries migrate, causing small grains to disappear and large grains to coarsen. Precipitates impede this evolution because lattice mismatch with the matrix raises interfacial energy and thus slows grain-boundary motion. In addition, precipitates trap or scatter mobile species, particularly cations, which reduces the atomic flux into grain interiors and promotes cation-vacancy formation. The consequent reduction in atomic mobility refines the grain-size distribution and stabilizes precipitates at grain boundaries, thereby pinning boundary migration.[73,74]

    In n-type Cu0.01PbSe, heavy Ge doping drives defect evolution and modifies the electronic structure, markedly improving thermoelectric performance near room temperature. High Ge concentrations induce grain refinement and produce uniformly distributed pseudo-nanostructured vacancy clusters (Fig. 6(b)).[23] These clusters enhance phonon scattering and reduce the lattice thermal conductivity to an ultralow value of 0.59 W⋅m−1⋅K−1 at 300 K. Concurrently, Ge increases the electron effective mass, which compensates for PbSe’s intrinsically low Seebeck coefficient and, together with optimized carrier concentration, helps to decouple electronic and phononic transport. Elemental mapping of the n-type Cu0.01PbSe sample shows Ge and Cu enrichment accompanied by Pb depletion. The resulting precipitates further refine the grain structure, increase grain-boundary density, and enhance phonon scattering.[23]

    Fu et al.[24] similarly introduced sub-microscale grain boundaries into Fe1.05Nb0.75Ti0.25Sb and incorporated abundant atomic-scale point defects to form hierarchical phonon-scattering centers. This architecture reduced the lattice thermal conductivity by ∼ 80% and yielded a peak ZT ≈ 1.34 at 1150 K. Although mosaic crystals appear single-crystalline at the macroscopic scale, they consist of numerous nanoscale crystallites with varying sizes and orientations and contain abundant low-angle grain boundaries (LAGBs).[75,76] Such microstructures can strengthen thermoelectric performance.[76] To mitigate the trade-off that point defects and other lattice imperfections — while effective at reducing κL — often impede electronic transport and reduce carrier mobility, Yang et al. introduced dual-site (Ag–Cl) doping in Bi2S3 to induce dense LAGBs.[77] This approach simultaneously increased carrier concentration and mobility while suppressing κL. Consequently, κL was reduced to 0.30 W⋅m−1⋅K−1, yielding a peak ZT ≈ 0.9 at 676 K and an average ZT ≈ 0.52 across 323–676 K.[77]

    A twin boundary is a planar defect across which the crystal lattice is mirrored. Twins typically form during processes such as plastic deformation, annealing, or rapid solidification, where stress or thermal driving forces promote coordinated lattice reorientation. Coherent nanotwins, which retain an ordered atomic registry across the boundary, preferentially scatter mid- to low-frequency phonons while leaving carrier mobility largely intact; this selectivity can enhance thermoelectric performance by reducing lattice thermal conductivity without severely degrading electrical transport. Li et al.[25] introduced dense twin networks into Cu2SnS3 by mechanical alloying followed by spark-plasma sintering; the twin engineering reduced the lattice thermal conductivity to ∼ 0.46 W⋅m−1⋅K−1 at 773 K and yielded a peak ZT ≈ 0.98. By contrast, CTS samples synthesized without twin engineering (e.g., by conventional solid-state or wet-chemical routes) display much higher κ1 (≈ 2.4 W⋅m−1⋅K−1 at ∼ 320 K), underscoring the effectiveness of twin engineering in suppressing lattice heat transport.[78,79] Zheng et al.[80] fabricated core–shell Cu5FeS4 nano-icosahedra via a colloidal route, producing high densities of coherent twin planes. By reducing particle size (to ∼ 130 nm) through Cu/Fe precursor tuning, they increased twin density (∼ 1.6 × 10−1 nm−1) and obtained bulk samples with κ1 ≈ 0.45 W⋅m−1⋅K−1 at 773 K and a peak ZT ≈ 0.95, illustrating the effectiveness of “twin + composition” engineering for simultaneous phonon suppression and electrical optimization. High twin-plane densities can decouple phonon and carrier transport by matching the twin spacing to dominant phonon mean free paths while remaining much larger than electronic mean free paths, thus preferentially scattering phonons over charge carriers.

    Stacking faults, which arise from deviations in the atomic stacking sequence, scatter phonons and thereby reduce the lattice thermal conductivity κL, improving thermoelectric performance. Such faults emerge during plastic deformation, rapid annealing, or alloying, for example when an FCC sequence shifts from ABCABC⋯ to ABCACB⋯. In Cd-doped high-entropy PbSe0.5S0.25Te0.25, lattice mismatch at the rock-salt PbSe/CdS interfaces generates interfacial strain that is relaxed via the formation of dense stacking faults.[26] These stacking faults act as efficient scattering centers for mid-frequency phonons, thereby reducing the κL. Analogous nanoscale stacking faults have been reported to scatter phonons in TlGaSe2[81] and in (GeTe)1−x(Bi2Te3)x[82] composites, contributing to suppressed κL in those systems. Ming et al.[83] deliberately engineered dense stacking-fault networks in Cu2SnSe3 by lowering the intrinsic stacking-fault energy; the resulting planar defects strongly impede heat-carrying phonons and reduce κL.

    Overall, in bulk nanostructured materials, the primary strategy for reducing thermal conductivity is to modify the lattice structure. Introducing defects increases phonon scattering centers and shortens phonon relaxation times. Among various types of defects, linear and planar defects enhance phonon scattering without significantly degrading carrier mobility, unlike point defects. Furthermore, planar defects can selectively scatter low-energy carriers and mid-frequency phonons, making them a highly promising approach for concurrently lowering thermal conductivity and enhancing the Seebeck coefficient.

  • Lattice thermal conductivity depends critically on the phonon group velocity vg, which, for long-wavelength acoustic modes, is commonly approximated by the sound velocity vs. Low sound velocities generally correlate with low lattice thermal conductivities. Phonon dynamics are governed by the nature and strength of chemical bonding: the sound velocity scales as vs∼k/M, where k denotes the bond stiffness (force constant) and M the average atomic mass.[84] Consequently, weak bonds and heavy constituent elements yield low vs and hence reduced lattice thermal conductivity.[84]

    Liu et al.[27] investigated Cd substitution in p-type zintl AB2X2 compounds (A = Mg, Ca, Ba, Sr, Yb, Eu; B = Zn, Cd, Mg; X = Sb, Bi) using density-functional theory, combining chemical-bond analysis with electron-localization (ELF) mapping. Figure 7(a) presents the ELF on the (100) plane of Ca8Mg14Cd2Bi16; by definition, ELF = 1 corresponds to perfect localization, ELF = 0.5 to a homogeneous electron gas and ELF = 0 to complete delocalization. The ELF around Ca sites remains essentially unchanged, indicating that Cd substitution at Mg sites does not alter the interlayer ionic character. Conversely, the reduced ELF between Cd and Bi (relative to Mg–Bi) indicates a weakening of covalent bonding within the B–X framework. Phonon-dispersion calculations (Figs. 7(b) and 7(c)) reveal pronounced softening of acoustic modes in Ca8Mg14Cd2Bi16: the cutoff frequencies at the K and L points decrease from 1.66 THz and 0.89 THz to 0.67 THz and 0.61 THz, respectively.[27] This lattice softening, together with enhanced phonon scattering, accounts for the observed reduction in κL. Similarly, Tian et al.[28] reported that substituting Ag for Mn in MnSb2Te4 weakens chemical bonding and induces lattice softening, producing an ∼ 12% reduction in sound velocity. Consequently, the room-temperature thermal conductivity decreased from 0.90 W⋅m−1⋅K−1 to 0.55 W⋅m−1⋅K−1 (∼ 39% reduction). Zhu et al.[85] experimentally demonstrated that Pb incorporation into GeTe weakens chemical bonding and breaks lattice symmetry, inducing lattice softening. This conclusion is supported by the flattening of low-frequency acoustic phonon branches and a concomitant decrease in phonon group velocity v, both of which lead to a pronounced reduction in lattice thermal conductivity.

    Unlike conventional defect-induced phonon scattering, lattice softening remains effective across a broad temperature range and consistently reduces the κL. Hanus et al.[29] systematically demonstrated this distinction. In their schematic comparison (Fig. 8),[29] solid circles denote experimental κL for single-crystal Si (sc-Si) and nanocrystalline Si (nc-Si; average grain size ≈ 42 nm), while the curves show calculated κL that include lattice-softening effects and/or scattering mechanisms (phonon–phonon scattering τpp−1 and phonon–grain-boundary scattering τgbs−1). The only parameters changed between sc-Si and nc-Si were the sound velocity vs and the grain size, both measured experimentally. The reduction from the black to the yellow curve reflects lattice softening alone, whereas the drop to the blue dashed curve reflects grain–boundary scattering; when both mechanisms are included (blue solid curve), the calculated κL closely matches the nc-Si data over the entire temperature range. Notably, phonon scattering dominates at low temperatures but becomes progressively less effective at high temperatures, whereas lattice softening yields a nearly uniform reduction in κL across all temperatures, as shown in Fig. 8.[29]

    Doping or alloying is a widely used strategy to induce lattice softening. Han et al.[86] reported that substituting Hf or Ta for Nb in NbFeSb, owing to the large atomic masses and, in the case of Hf, its aliovalent character, produces avoided crossings between optical and acoustic phonon branches and reduces the acoustic phonon group velocity. With 10% Ta or Hf substitution, the phonon group velocity Vg decreases from 3428 m⋅s−1 to 3388 m⋅s−1 (Ta) and 3349 m⋅s−1 (Hf); 10% Hf doping in particular yields a ∼ 65% reduction in lattice thermal conductivity. Xiang et al.[87] synthesized Zr3Ni2.5Co0.5Sb4 by Co-doping Zr3Ni3Sb4; phonon dispersions reveal strong flattening of optical branches and a reduction of the acoustic cutoff frequency from 95 cm−1 to ≤ 65 cm−1, consistent with pronounced lattice softening. Ni–Co disorder further introduces localized rattler-like modes that undergo avoided crossings with low-frequency acoustic phonons, flattening the branches and extending phonon lifetimes in that frequency window; consequently, the lattice thermal conductivity is reduced by ∼ 42%.

    A heavy alloying approach, such as incorporating AgSbTe2 into MnGeTe2, can simultaneously modulate both electronic and lattice properties. The heavier Ag and Sb atoms enhance spin–orbit coupling and increase the density-of-states effective mass (m*) by approximately 50%, while maintaining a high Seebeck coefficient at elevated hole concentrations.[88] Alloying also softens the lattice: both transverse and longitudinal sound velocities decrease with increasing AgSbTe2 content, which suppresses κL toward the amorphous limit. Shen et al.[89] showed that substituting Cd for Zn in ZnSb produces strong point-defect scattering and concurrently softens acoustic phonons, leading to an approximately 44% reduction in κL at room temperature for Zn0.7Cd0.3Sb.

    Moreover, internal strain fields introduced by structural defects can induce lattice softening. For example, Shen et al.[90] reported that excess Ag in AgIn5Se8 generates abundant point defects, twin boundaries and stacking faults; these defects produce internal strain fields that soften the lattice and enhance phonon scattering, thereby reducing the κL. As the Ag content x increases from 0 to 0.03 and 0.07, the average sound velocity vs decreases from 2028 m⋅s−1 to 1905 m⋅s−1 and 1674 m⋅s−1, respectively. Similarly, materials containing high dislocation densities (for example, Na0.03Eu0.03Sn0.02Pb0.92Te)[91] exhibit strain-induced phonon softening and corresponding κL suppression. Distinct from direct phonon–dislocation scattering, the strain fields associated with dense dislocation networks reduce the local sound velocity through localized anharmonicity, which in turn suppresses κL (κL scales approximately with vs3),[92] thus providing an efficient route to lower thermal conductivity without necessarily introducing strong carrier scattering.

  • Reducing specific heat alone offers limited leverage for lowering thermal conductivity because, under the Debye approximation, the high-temperature heat capacity of a crystalline solid approaches the Dulong–Petit limit (cv ≈ 3NkB). Superionic conductors circumvent this constraint: above the superionic transition temperature, mobile cations become highly disordered and exhibit liquid-like dynamics. Local hopping and atomic rearrangements in the cation sublattice suppress shear-wave propagation and disrupt phonon heat transport, so that cv can decrease from ∼ 3 NkB to roughly 2–2.5NkB and the lattice thermal conductivity is substantially reduced. For example, Cu2−xSe — where a rigid Se sublattice coexists with a liquid-like Cu sublattice — has been reported to attain ZT ≈ 1.5 near 1000 K, an effect attributed to the suppression of transverse phonon modes by the mobile Cu ions.[93] Likewise, in Ag(Cu)CrSe2, the onset of superionic disorder of Ag/Cu on tetrahedral sites leads to pronounced anharmonicity and strong phonon scattering, which reduces cv and yields ultralow lattice thermal conductivity.[30] These examples illustrate that superionicity provides an alternative route to lower κL via changes in both specific heat and phonon transport, complementary to strategies based on increased phonon scattering alone.

    Modifying the lattice architecture of bulk nanostructured materials is a principal strategy to suppress κL. Defect engineering introduces scattering centers that shorten phonon relaxation times: point defects predominantly scatter high-frequency phonons, dislocations target mid-frequency modes, and interfaces or planar faults (e.g., nanotwins, low-angle grain boundaries) impede low-frequency, long-wavelength phonons. Unlike point defects, which often degrade carrier mobility, well-designed one- and two-dimensional defects can scatter phonons effectively while preserving or even improving electronic transport via energy-filtering effects, thereby providing simultaneous reduction of κL and enhancement of the Seebeck coefficient. Complementary approaches include lattice softening — via mass substitution or alloying — that reduces phonon group velocities (vg), and superionic behavior, in which liquid-like cation dynamics suppress transverse phonons and lower the effective specific heat (cv); both mechanisms reduce κL across broad temperature ranges. As shown in Table 1, a comparison of phonon-engineering strategies is summarized, including the primary phonon frequency ranges they target, their typical advantages and disadvantages, and the practical limitations of their applications. Crucially, these mechanisms act synergistically: by combining phonon scatterers that operate at different frequency bands with lattice-softening or superionic effects, one can suppress thermal transport across the whole phonon spectrum while retaining favorable electronic properties. This multi-scale, multi-mechanism paradigm therefore offers a robust route to approaching the phonon-glass, electron-crystal ideal and to realizing high-performance thermoelectric materials.

3.   Phonon engineering of low-dimensional structures
  • Despite their success in three-dimensional systems, conventional phonon-engineering strategies still face a fundamental limitation: the trade-off between the Seebeck coefficient and electrical conductivity as carrier concentration varies, which constrains further ZT improvements. In 1993, Hicks and Dresselhaus[32] proposed that low dimensionality and quantum-confinement effects can sharpen the electronic density of states near the Fermi level and thereby increase the Seebeck coefficient, while surface and interface scatterings preferentially suppress phonon transport.[33–35] Building on this idea, micro- and nanostructuring, including superlattices, nanocomposites, and nanowires, has been developed to decouple electronic and phononic transport and to achieve substantially reduced lattice thermal conductivity and enhanced thermoelectric performance.

  • Superlattice thin films are periodic structures formed by stacking alternating layers of two or more materials, with individual layer thicknesses typically ranging from a few to several tens of atomic layers. Phonon transport in such superlattices can be either incoherent (particle-like) or coherent (wave-like), depending on the relation between the phonon mean free path (or coherence length) and the superlattice period. When the phonon mean free path is shorter than the period, transport is dominated by incoherent particle-like phonons and can be described by particle-based models in which phonons scatter independently at interfaces; conversely, when the period is comparable to or smaller than relevant phonon wavelengths or coherence lengths, wave effects and coherent transport become significant. In the incoherent regime, thermal conductivity is effectively reduced by enhancing interface scattering — for example, by increasing interface density or roughness, introducing nano-inclusions, or maximizing the mismatch in phonon density of states between constituent layers. Superlattice films are commonly fabricated by sputtering,[94] thermal evaporation,[95] metal–organic chemical vapor deposition (MOCVD)[96] and molecular beam epitaxy (MBE),[34] each providing different trade-offs between interface sharpness, compositional control and throughput.

    Recent advances in epitaxial-film synthesis enable atomic-scale engineering of phononic states. Superlattices (SLs), i.e., periodic stacks of alternating layers, redefine lattice periodicity via a supercell of thickness Na (where N is the number of unit cells and a is the unit-cell thickness). This produces zone folding of the phonon dispersion and reduces the Brillouin zone to 2π/Na.[99] At the reduced zone, zone-folded acoustic (ZA) phonons appear (Fig. 9(a)) and their energies are tunable through N.[97] When Na is shorter than the phonon coherence length λ (coherent regime, λ > Na), wave interference opens gaps at the zone boundary; ZA modes form standing waves at these gaps and the thermal conductivity κ is reduced. Hence κ tends to decrease as Na increases (stronger zone folding). Conversely, when Na is much larger than λ (incoherent regime, λ < Na), scattering destroys coherence, the zone-folding gaps close, and k increases with Na despite the larger number of zone folds.[97] As a result κ is non-monotonic and reaches a minimum near the coherence–incoherence crossover (Na ≈ λ). Because phonon coherence depends on the relative lengths Na and λ, κ can be engineered either geometrically (by changing Na) or by tuning λ through phonon renormalization associated with electronic or magnetic phase transitions.[100] Recent studies of oxide SLs stress that designing both the electronic contribution to thermal conductivity and the phonon spectrum is essential to realize high thermoelectric performance.[100] Although artificial, zone-folded phonons in SLs are now well established, modulation of these modes by correlated electronic or magnetic phase transitions remains uncommon. Recent experiments on oxide superlattices illustrate how control of phonon coherence and zone folding tunes thermal transport. For example, SrRuO3/SrTiO3 superlattices show a clear coherence–incoherence crossover: by varying the interface density n = 2/Na (with Na the supercell thickness), the cross-plane thermal conductivity κ displays a minimum at n ≈ 0.7 nm−1 (Na ≈ 2.8 nm), corresponding to λ ≈ 2.8 nm The minimum κ is ∼ 2.1 W⋅m−1⋅K−1, slightly above the alloy limit, and the measured n dependence is quantitatively captured by Simkin–Mahan calculations that include zone-folding effects.[97] TDTR, Raman and SHG measurements show that structural phase transitions (polar/AFD ordering and ferromagnetism) in the constituent layers strongly modify phonon coherence and mean free paths; temperature- or strain-driven symmetry changes therefore can further suppress κ by disrupting coherence and opening or closing zone-folding gaps. These results demonstrate that superlattice geometry combined with correlated structural and electronic phase transitions provides powerful, complementary levers to suppress heat transport by phonon engineering.[97] Fu et al.[101] fabricated Bi2Te3/Sb2Te3 superlattices and reported a reduced thermal conductivity of 0.33 W⋅m−1⋅K−1 at 275 K, compared with 2.38 and 1.56 for bulk Sb2Te3 and Bi2Te3, respectively; the reduction was attributed to enhanced phonon scattering from the SL’s high interface density. Similarly, Yang et al.[98] prepared Sb2Te3–Sb2Se3 superlattice films whose room-temperature κ is lower than that of single-component Sb2Te3; in particular, a sample with 2 nm:2 nm layer thicknesses exhibited κ = 0.62 W⋅m−1⋅K−1, nearly half that of Sb2Te3 (1.23 W⋅m−1⋅K−1), demonstrating the effectiveness of SL architectures in suppressing heat transport (Fig. 9(b)).

    When the phonon mean free path (MFP) is comparable to or exceeds the superlattice period, phonons display wave-like behavior: they can coherently propagate across several periods, preserve phase information and undergo multiple reflections at interfaces. Under these conditions, coherent transmission and multiple-interface reflections produce interference effects that modify phonon dispersion, opening or widening phononic band gaps and reducing phonon group velocities. Introducing aperiodicity can disrupt coherent transport and promote phonon localization, thereby yielding ultralow thermal conductance in suitably designed superlattices.[102–104] For example, Hu et al.[105] reported that aperiodic Si/Ge superlattices can exhibit thermal conductivities an order of magnitude lower than periodic counterparts, which they attributed to Anderson localization of coherent phonons. Jue Wang and co-workers[106] theoretically examined periodic and two classes of aperiodic MoS2–MoSe2 lateral superlattices, finding that increasing the length of the aperiodic region reduces phonon conductance and can induce localization; the most strongly aperiodic (‘wild’) structures show the lowest conductance for a given length. Phonon transmission analysis indicates that suppressed transmission in these lateral superlattices stems from enhanced anharmonic scattering and interface disorder, which together reduce κL and improve the thermoelectric figure of merit. Finally, tuning the interlayer twist angle in van der Waals heterostructures provides an additional route to modulate phonon transport: studies report a strong, material-dependent twist-angle dependence of thermal conductivity, although the absolute magnitudes vary with the system and measurement or computational details. Moreover, tuning the interlayer twist angle of the superlattice provides another pathway to further reduce the lattice thermal conductivity. For example, Singh et al.[107] twisted the interlayer angle of CrI3/1T-MoS2, achieving ultralow lattice thermal conductivities of 0.0341 W⋅m−1⋅K−1 and 0.0034 W⋅m−1⋅K−1 at twist angles of 23.41° and 30°, respectively.

  • Nanowires, as prototypical one-dimensional materials, have diameters that are much smaller than those of bulk counterparts, which enhances phonon boundary scattering. Silicon nanowires (Si-NWs) are particularly promising for thermoelectric applications because they can substantially reduce thermal conductivity while largely preserving the Seebeck coefficient and electrical conductivity of bulk silicon. Lee et al.[108] systematically tuned the Si-NW diameter, doping level and surface roughness to optimize phonon scattering. They found that reducing the nanowire cross-sectional area intensifies phonon–boundary scattering and markedly lowers thermal conductivity, and increasing surface roughness introduces additional scattering sites that further suppress the heat transport.

    Figures 10(a) and 10(b) compare measured thermal conductivities for vertical Si-NWs as a function of diameter and roughness; Fig. 10(a) additionally includes data for vapor–liquid–solid (VLS)-grown Si-NWs. For example, increasing the root-mean-square surface roughness from 2.28 nm to 6.88 nm reduces thermal conductivity by ∼ 25.5%, and decreasing the diameter from 350 nm to 200 nm yields an additional ∼ 27.5% reduction. Phonon mean-free-path calculations indicate that rough surfaces predominantly suppress long-wavelength phonons, whereas diameter reduction shortens the MFP across a broad portion of the phonon spectrum. Furthermore, B- and P-doped Si-NWs (200-nm diameter) exhibit thermal conductivities 71.5% and 58.4% lower than those of undoped rough wires, corresponding to values ∼ 14.8× and ∼ 10.1× smaller than bulk silicon (150 W⋅m−1⋅K−1), respectively. Kang et al.[109] synthesized tellurium nanowires with a thermal conductivity of ∼ 1.0 W⋅m−1⋅K−1 — substantially lower than bulk Te pellets — and Shah et al.[110] showed that surface passivation of Te nanowires can further reduce thermal conductivity to ∼ 0.16 W⋅m−1⋅K−1 by enhancing phonon scattering at nanowire–matrix interfaces. Heris et al.[111] examined how transverse cross-sectional geometry affects the thermal conductivity of Si and Ge nanowires by comparing circular, triangular, diamond, hexagonal, and square profiles. They found that nanowires with more edges or corners exhibit lower thermal conductivity and that conductivity increases with the cross section’s surface-to-volume ratio (perimeter-to-area ratio). Extending their work, they modeled hollow nanowires by removing core atoms and observed significant conductivity reductions across all geometries. This decline stems from the hollow design’s further increase in surface-to-volume ratio, which enhances phonon confinement and scattering, yielding much lower thermal conductivity than in solid nanowires.

  • Nanocomposites are nano-heterostructures in which nanoscale inclusions, such as fine grains or nanoparticles, are embedded within a host matrix and act as additional phonon-scattering centers.[112] In this review, the term “nanocomposites” specifically refers to bulk matrices containing nanoscale inclusions whose thermal transport is governed by interfacial scattering and related nanoscale effects. When the inclusions are metallic or semimetallic, they can donate electrons to increase electrical conductivity and generate an energy-filtering effect that enhances the Seebeck coefficient. For example, Zheng et al.[113] fabricated a hybrid 0D–2D heterostructure by in situ growth of Au nanoparticles on hexagonal Sb2Te3 nanoplates; this architecture simultaneously optimized electrical conductivity, thermal conductivity and the Seebeck coefficient, yielding an ∼ 110% increase in ZT relative to pristine Sb2Te3. Kim et al.[114] reported that incorporation of an Sb2O3 nanophase into Bi0.5Sb1.5Te3 enhanced the Seebeck coefficient via energy filtering and reduced thermal conductivity by ∼ 77% owing to strong phonon scattering at interfaces. Building on this concept, Minsu Kim and co-workers[115]constructed 1D/2D nanocomposites comprising Sb2O3 nanowires and layered double hydroxides (LDHs); the resulting increase in 1D/2D heterointerface density produced intense interfacial phonon scattering, effectively suppressing lattice thermal conductivity and improving ZT.

    Low-dimensional architectures such as superlattices, nanowires and nanocomposites provide tunable parameters at the length and interface scales that facilitate partial decoupling of electronic and phononic transport, thereby substantially suppressing κL and improving thermoelectric performance. Superlattices operate in two transport regimes: an incoherent regime dominated by interface scattering and a coherent regime dominated by wave interference. In the incoherent regime, increasing interface density, enhancing interface roughness or introducing aperiodicity strengthens interface scattering. In the coherent or semi coherent regime, multiple-interface reflections and interference can open phononic band gaps or induce phonon localization, which reduce phonon group velocities and κL. Nanowires intensify boundary and geometric scattering through diameter reduction, increased surface roughness, modified cross sectional shapes or hollow architectures; doping and surface treatments can act synergistically to lower κL while preserving or improving electronic transport. Nanocomposites introduce additional interfacial scattering centers by embedding nanoscale inclusions within a host matrix. When inclusions are metallic or semimetallic, they can also provide charge donation or energy filtering to optimize electronic properties. Experimental demonstrations confirm the efficacy of low-dimensional designs for dramatically reducing κL and enhancing ZT. Overall, by tuning interface density, size, surface roughness, aperiodicity and interlayer twist, low-dimensional phonon engineering supplies a multiscale, multimodal toolbox to minimize heat conduction without sacrificing electronic performance; however, controllable fabrication, interface precision and scalable manufacturability remain major challenges for practical deployment

4.   Conclusion
  • The pursuit of high thermoelectric performance has motivated extensive efforts towards designing materials with intrinsically low lattice thermal conductivity. This review examines the role of phonon engineering in reducing lattice thermal conductivity and summarizes recent advances in two complementary approaches, namely, phonon engineering of bulk nanostructured systems and phonon engineering of low-dimensional materials.

    In bulk nanostructured systems, defect engineering introduces scattering centers that selectively target phonons across the frequency spectrum. Point defects effectively scatter high frequency phonons and can concurrently tune carrier concentration. Dense dislocation networks mainly suppress mid-frequency phonons and may be generated by doping, vacancy clustering, liquid phase extrusion, hot deformation or ultrahigh pressure sintering. Interfacial defects such as grain boundaries, twin planes and stacking faults preferentially impede low frequency long wavelength phonons while exerting limited influence on carrier mobility. Lattice softening and modulation of specific heat provide additional suppression of lattice thermal conductivity: heavy atom substitution, alloying and bond weakening reduce phonon group velocities, whereas superionic behavior characterized by liquid like cation dynamics lowers the effective heat capacity and suppresses transverse phonon transport. Importantly, these mechanisms can act synergistically so that combining frequency targeted scattering with lattice softening or superionic effects suppresses heat conduction across the full phonon spectrum while preserving favorable electronic properties.

    Phonon engineering in low dimensional materials offers further opportunities to reduce lattice thermal conductivity. Architectures such as superlattices, nanowires and nanocomposites provide tunable design levers through length scale and interface engineering, enabling more effective decoupling of electronic and phononic transport. Superlattices may operate in an incoherent regime dominated by interface scattering or in a coherent regime dominated by wave interference; increasing interface density, enhancing interface roughness or introducing aperiodicity strengthens phonon scattering, while in the coherent regime multiple reflections and interference can open phononic band gaps or induce localization, both of which reduce phonon group velocities and thermal conductivity. Nanowires enhance boundary and geometric scattering through diameter reduction, increased surface roughness, cross sectional design or hollow architectures; appropriate doping and surface treatments can further suppress lattice thermal conductivity while maintaining or improving electrical transport. Nanocomposites formed by embedding metallic, semimetallic or insulating nanoprecipitates in a host matrix generate abundant interfacial scattering sites and, when metallic inclusions are used, can enable energy filtering or charge transfer to optimize electronic properties. Numerous experimental demonstrations confirm that low dimensional designs can substantially reduce lattice thermal conductivity and enhance the thermoelectric figure of merit.

    The practical relevance of these strategies, however, hinges on manufacturability, cost and environmental considerations — topics discussed next. From a translational perspective, practical deployment requires careful assessment of process, economic and environmental factors. Many phonon engineering approaches, notably point defect doping and nanocomposite formation, are compatible with conventional metallurgical and powder metallurgy routes such as alloy melting and ball milling followed by hot pressing, and thus have scale up potential. Powder metallurgy methods including gas atomization and low temperature pulverization, together with emerging ink based printing techniques, show promise for higher throughput and lower cost device fabrication.[116–118] By contrast, laboratory scale, atomically precise strategies such as molecular beam epitaxy grown superlattices and twist engineering of van der Waals heterostructures remain slow and costly and therefore face substantial barriers to industrialization.[119] Material supply and environmental constraints further complicate route selection, because many high performance systems rely on scarce or toxic elements, notably Te and Pb, which pose supply risk and health concerns and motivate research on Pb and Te free alternatives.[120,121] Finally, a persistent trade off exists between κL suppression and electronic transport: many effective phonon scattering motifs reduce κL but simultaneously increase carrier scattering, degrading mobility and electrical conductivity. Strategies such as energy filtering, band engineering to increase density of states effective mass, and careful matching of phonon and electron scattering length scales can mitigate these penalties, but they add design complexity and require extensive optimization.

    In summary, integrating multiscale, multimodal phonon control strategies can substantially reduce lattice thermal conductivity while preserving electronic performance, thereby advancing the phonon glass electron crystal paradigm. To translate high performance concepts into manufacturable devices, future research should prioritize scalable processing routes and environmentally benign chemistries.

Figure (10)  Table (1) Reference (121)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return