betaFeedback

WorksPhysics and Astronomy

Anisotropic Phonon Heat Flow and Thermoelectric Response in Tetragonal GeS2_2 and GeSe2_2

Abstract

The electronic structure, lattice dynamics, bonding, elastic response, and anisotropic thermoelectric transport properties of tetragonal GeS2_2 and GeSe2_2 were investigated using density functional theory, density functional perturbation theory, Wannier interpolation, and scattering-aware Boltzmann transport. The relaxed structures are mechanically and dynamically stable within the calculated harmonic description. The HSE03/Wannier band gaps are 2.48 eV for GeS2_2 and 1.23 eV for GeSe2_2, while substitution of S by Se lowers the upper phonon frequency from approximately 13.6 to 10.3 THz. The phonon Boltzmann transport calculations reveal pronounced lattice-transport anisotropy. Within the relaxation-time approximation, the 300 K in-plane and cross-plane lattice thermal conductivities are 26.86 and 1.19 W m−1^{-1} K−1^{-1} for GeS2_2, and 18.74 and 1.52 W m−1^{-1} K−1^{-1} for GeSe2_2, respectively. At 800 K, these values decrease to 10.22 and 0.46 W m−1^{-1} K−1^{-1} for GeS2_2, and 7.25 and 0.58 W m−1^{-1} K−1^{-1} for GeSe2_2. Frequency-resolved analysis shows that low-frequency phonons carry most of the heat, whereas the small cross-plane values reflect restricted out-of-plane phonon transport. Combining the ShengBTE RTA lattice tensors with AMSET electronic coefficients gives zT=0.257zT=0.257 for n-type cross-plane GeS2_2 at 800 K and 101910^{19} cm−3^{-3}. The corresponding PBE-AMSET estimate for GeSe2_2 is zT=0.066zT=0.066 for p-type cross-plane transport at 800 K and 3×10203\times10^{20} cm−3^{-3}. LOBSTER analysis identifies mixed covalent–ionic Ge–X bonding, with Ge–S bonds having a larger stabilizing ICOHP magnitude than Ge–Se bonds (−5.27-5.27 versus −4.74-4.74 eV per bond). These results identify tetragonal GeX2_2 compounds as strongly anisotropic thermoelectrics with moderate calculated zTzT values whose cross-plane response benefits from suppressed lattice heat transport.

Introduction

Growing energy demand and continued dependence on nonrenewable resources have increased the need for clean energy-conversion materials.[1] [2, 3] Thermoelectric materials are of interest because they directly convert heat into electricity and can also provide solid-state cooling, making them useful for waste-heat recovery in industrial, automotive, and other dissipative systems.[1] [4]

The performance of a thermoelectric material is commonly quantified by the dimensionless figure of merit, zTzT, defined as

zT=σS2Tκe+κl(1)zT = \frac{\sigma S^2 T}{\kappa_{\mathrm{e}}+\kappa_{\mathrm{l}}} \tag*{(1)}

where σ\sigma is the electrical conductivity, SS is the Seebeck coefficient, TT is the absolute temperature, and κe\kappa_{\mathrm{e}} and κl\kappa_{\mathrm{l}} are the electronic and lattice contributions to thermal conductivity, respectively.[4–6] High zTzT requires a large power factor, σS2\sigma S^2, and low total thermal conductivity. These quantities are mutually coupled: increasing carrier concentration can improve σ\sigma but often raises κe\kappa_{\mathrm{e}} through the Wiedemann–Franz relation, while lower carrier scattering may also increase heat transport.[2, 7–9] By contrast, κl\kappa_{\mathrm{l}} can often be reduced more independently by increasing phonon scattering through alloying, nanostructuring, dislocations, point defects, or hierarchical microstructures.[7, 9–11]

These design principles have been implemented across diverse thermoelectric material classes, including skutterudites, half-Heusler compounds, chalcogenides, and layered lowdimensional systems.[2, 12] Among them, IV–VI chalcogenides such as SnSe and SnTe have demonstrated record-high zTzT values through coordinated band and phonon engineering.[5–7, 10, 11] Despite these successes, several widely used materials still face practical challenges: Bi2Te3\mathrm{Bi}_2\mathrm{Te}_3-based alloys, for example, exhibit performance degradation at elevated temperatures, which lowers useful zTzT, whereas skutterudites and many half-Heusler compounds possess intrinsically high lattice thermal conductivity.[1] [2] These limitations motivate continued exploration of non-toxic, earth-abundant compounds that combine favorable electronic properties with intrinsically low lattice thermal conductivity.[12]

Layered and low-dimensional materials are promising thermoelectric candidates because structural anisotropy can influence, and partly decouple, electronic and phonon transport.[13–15] Group IV–VI chalcogenides containing Ge or Sn, such as GeS, GeSe, SnS, and SnSe, combine useful band structures with relatively abundant and environmentally compatible elements.[13, 16] Germanium dichalcogenides, GeX2\mathrm{GeX}_2 (X=S,Se\mathrm{X} = \mathrm{S}, \mathrm{Se}), have historically been studied mainly as glasses and for phase-change or resistive-switching applications. Recent work on metavalent bonding in main-group chalcogenides helps explain how these compounds can combine favorable electronic transport with strongly anharmonic phonon behavior, while their crystalline layered forms are now attracting interest for optoelectronic and nanoelectronic devices.[12] Experiments have shown strong in-plane structural, vibrational, and optical anisotropy in exfoliated GeS2\mathrm{GeS}_2 and GeSe2\mathrm{GeSe}_2 nanoflakes, and ultrathin β\beta-GeSe2\mathrm{GeSe}_2 flakes grown by van der Waals epitaxy show promising polarization-sensitive photodetection.[17–20]

Recent studies also highlight the optical and transport potential of GeS2\mathrm{GeS}_2. Atomic-scale characterization of monoclinic layered GeS2\mathrm{GeS}_2 nanostructures has linked strong in-plane anisotropy to anisotropic electronic and optical properties.[17] Layered GeS2\mathrm{GeS}_2 has also been reported to exhibit an exceptionally high refractive index and strong optical anisotropy in the blue and near-ultraviolet frequency ranges, making it relevant for short-wavelength nanophotonics and metasurfaces.[18] First-principles studies have further examined adsorption, electronic, and optical properties of GeS2\mathrm{GeS}_2 monolayers, as well as the dependence of calculated 2D band gaps on the chosen exchange–correlation functionals.[21–23]

Beyond the monoclinic layered phases, GeS2\mathrm{GeS}_2 also occurs in a tetragonal framework (tt-GeS2\mathrm{GeS}_2) with space group P42/nmcP4_2/nmc, where corner-sharing GeS4\mathrm{GeS}_4 tetrahedra form quasi-two-dimensional sheets oriented along the (0,0,1)(0,0,1) direction.[24, 25] This polymorph is a useful platform for connecting wide-gap semiconducting behavior, tetrahedral coordination, and anisotropic bonding with optical and transport properties. However, the electronic, optical, and phonon-transport properties of bulk tt-GeS2\mathrm{GeS}_2 remain comparatively underexplored from a thermoelectric perspective.[17, 18] The related GeSe2\mathrm{GeSe}_2 system is also less understood, especially in terms of anisotropic thermoelectric response. Prior work on anisotropic layered GeSe2\mathrm{GeSe}_2 reported a very small cleavage energy of about 0.05 J m−20.05\ \mathrm{J\ m}^{-2} and weak layer-dependent changes in the electronic and Raman spectra, showing that weak interlayer coupling is a realistic structural feature of this chemistry.[20]

In this work, the electronic structure and lattice thermal transport of bulk tt-GeSe2\mathrm{GeSe}_2 are investigated and compared with tt-GeSe2\mathrm{GeSe}_2 as a structural and chemical analogue. For tt-GeS2\mathrm{GeS}_2, we use the tetragonal reference structure with a=b=3.50 A˚a=b=3.50\ \text{\AA}, c=11.20 A˚c=11.20\ \text{\AA}, α=β=γ=90.00∘\alpha=\beta=\gamma=90.00^\circ, Ge atoms at (0.25,0.75,0.25)(0.25,0.75,0.25) and (0.75,0.25,0.75)(0.75,0.25,0.75), and S atoms at (0.25,0.25,0.37686)(0.25,0.25,0.37686), (0.75,0.75,0.62314)(0.75,0.75,0.62314), (0.25,0.25,0.87686)(0.25,0.25,0.87686), and (0.75,0.75,0.12314)(0.75,0.75,0.12314).[17] For tt-GeSe2\mathrm{GeSe}_2, the substituted and relaxed structure used in the calculations has a=b=3.703 A˚a=b=3.703\ \text{\AA} and c=11.284 A˚c=11.284\ \text{\AA}, with α=β=γ=90∘\alpha=\beta=\gamma=90^\circ; the Ge atoms occupy the same fractional positions, while Se atoms are located at (0.25,0.25,0.38604)(0.25,0.25,0.38604), (0.75,0.75,0.61396)(0.75,0.75,0.61396), (0.25,0.25,0.88604)(0.25,0.25,0.88604), and (0.75,0.75,0.11396)(0.75,0.75,0.11396). DFT and DFPT are used to obtain the electronic band structure, density of states, and elastic properties. Electronic transport is evaluated with state-dependent acoustic-deformation-potential, polar-optical-phonon, and ionized-impurity scattering,[26–28] while lattice thermal conductivity is obtained with ShengBTE.[29] Band-structure analysis, ICOOP, and ICOHP descriptors are then used to clarify the microscopic origin of the electronic and phonon transport behavior.

Computational details

Electronic structure and transport calculations

The electronic-structure calculations were carried out within Kohn–Sham density functional theory (DFT) using the plane-wave pseudopotential framework implemented in Quantum ESPRESSO.[30, 31] We focused on the tetragonal GeS2\mathrm{GeS}_2 phase (tt-GeS2\mathrm{GeS}_2) with space group P42/nmcP4_2/nmc. The reported reference structure has a=b=3.50 A˚a=b=3.50\ \text{\AA} and c=11.20 A˚c=11.20\ \text{\AA}, with α=β=γ=90∘\alpha=\beta=\gamma=90^\circ.[17] In the production Quantum ESPRESSO inputs used here, the tetragonal cell was defined by ibrav = 6, celldm(1) = 6.636200, and celldm(3) = 3.133016, corresponding to a=b=3.512 A˚a=b=3.512\ \text{\AA} and c=11.002 A˚c=11.002\ \text{\AA}, with two Ge and four S atoms per primitive cell.

Internal-coordinate optimization at the fixed tetragonal cell and ground-state electronic properties were obtained using the generalized gradient approximation (GGA) in the Perdew–Burke–Ernzerhof (PBE) formulation for the exchange–correlation functional,[32] combined with a semi-empirical Grimme D2 dispersion correction (vdw_corr = ‘grimme-d2’) to account for van der Waals interactions.[33, 34] Except for the separate LOBSTER projtion step discussed below, all QUANTUM ESPRESSO calculations used scalar-relativistic ultrasoft RRKJ pseudopotentials from the PSLibrary (Ge.pbe-n-rrkjus_psl.1.0.0.UPF, S.pbe-n-rrkjus_psl.1.0.0.UPF, and Se.pbe-n-rrkjus_psl.1.0.0.UPF).[35] This includes the PBE structural, elastic, band-structure, DFPT, deformation-potential, and AMSET-input calculations, as well as the HSE03 band-reference calculations. Keeping the same ultrasoft pseudopotential family across the main workflow avoids mixing structural, electronic, and response properties from different pseudopotential descriptions. The plane-wave kinetic-energy cutoff and charge-density cutoff were set to ecutwfc = 60 Ry and ecutrho = 480 Ry, respectively, and Brillouin-zone integrations were performed using an 8×8×48 \times8 \times4 Monkhorst–Pack kk-point mesh with occupations = ‘fixed’. Self-consistent PBE calculations employed a convergence threshold of conv_thr = 1.0×10−101.0 \times10^{-10} Ry on the total energy, with mixing_beta = 0.15 and a maximum of 200 electronic iterations.

To obtain more reliable band gaps for interpreting carrier excitation and transport trends, we performed additional self-consistent calculations using the screened HSE03 hybrid functional, as implemented in QUANTUM ESPRESSO.[36–38] In these calculations, the short-range Hartree–Fock exchange fraction and screening parameter were set to exx_fraction = 0.25 and screening_parameter = 0.106, respectively, with the qq-point mesh for exchange integrals defined by nqx1 = 2, nqx2 = 2, and nqx3 = 1. The same relaxed internal coordinates were used, and the Brillouin zone was sampled with a slightly coarser 6×6×26 \times6 \times2 kk-point grid (K_POINTS {automatic} 6 6 2 0 0 0). The electronic convergence threshold for HSE03 was set to conv_thr = 1.0×10−81.0 \times10^{-8} Ry with mixing_beta = 0.10 and up to 300 electronic iterations. This staged PBE-then-HSE03 workflow, in which cheaper GGA relaxations are followed by a single-point hybrid-functional correction, is a common strategy for obtaining improved semiconductor band gaps without the full cost of hybrid-functional geometry optimization, and has been used previously for related GeS2_2 systems.[22, 23]

For visualization, the PBE and HSE03 bands were interpolated with WANNIER90.[39–43] Electronic transport was calculated with AMSET v0.5.1, using its state-dependent Boltzmann transport formalism.[26] Because AMSET does not natively read Quantum ESPRESSO outputs, a Python interface based on the PWxml parser in PYMATGEN was used to extract the structure, kk points, eigenvalues, and electron count from data-file-schema.xml and write the AMSET band_structure_data.json input.[44] Orbital projections were represented by placeholder weights, so the wavefunction-overlap factors are approximate. ADP, POP, and IMP scattering were included using the calculated elastic, dielectric, deformation-potential, and optical-phonon inputs. Calculations covered 300–800 K and 101810^{18}–102210^{22} cm−3^{-3} on dense AMSET interpolation meshes of 81×81×2581 \times81 \times25 for GeS2_2 and 81×81×2781 \times81 \times27 for GeSe2_2. These meshes are post-processing grids generated from the converted dense NSCF band structures and are not the same as the Quantum ESPRESSO SCF/NSCF kk meshes. The converted PBE transport bands yield a 0.99 eV GeS2_2 gap, while GeSe2_2 lies very close to gap closure on the dense transport mesh, with a small ∼0.015\sim0.015 eV band overlap; this near-zero-gap PBE limit restricts quantitative high-temperature interpretation for GeSe2_2.[27, 28] As an independent constant-relaxation-time reference, the HSE03/Wannier bands were also post-processed with BOLTZTRAP2; these chemical-potential-dependent trends are included in the Supplementary Information and are not used for the quantitative zTzT values reported below.[45]

Harmonic lattice dynamics were investigated with DFPT as implemented in the PHonon component of QUANTUM ESPRESSO.[30, 31, 46] The PBE+DFT-D2 ground state used the same ultrasoft pseudopotentials, an 8×8×48 \times8 \times4 electronic kk mesh, and wave-function and charge-density cutoffs of 60 and 480 Ry, respectively. This choice is also efficient for response calculations involving dense qq-point sampling, Born effective charges, and dielectric tensors while retaining the same PBE+D2 structural description. Dynamical matrices were evaluated on a 20×20×620 \times20 \times6 qq mesh and transformed to real-space harmonic force constants with q2r.x. Born effective charges and the high-frequency dielectric tensor were obtained at Γ\Gamma using epsil=.true.; the corresponding non-analytic dipole–dipole correction was retained when interpolating the dispersion with matdyn.x.[47] The crystal acoustic sum rule was imposed, and the total phonon DOS was evaluated by tetrahedron integration on a 40×40×2040 \times40 \times20 uniform mesh with a 1 cm−1^{-1} energy spacing. An independent 4×4×24 \times4 \times2 finite-displacement calculation with PHONO3PY, supplied with the same dielectric tensor and Born charges, was used to verify the harmonic spectrum and the non-analytic splitting near Γ\Gamma.[48, 49]

The lattice thermal conductivity was subsequently calculated from second- and third-order force constants using the phonon Boltzmann transport equation implemented in SHENG­BTE.[29] Third-order force constants were generated using 4×4×24 \times4 \times2 finite-displacement supercells with thirdorder_espresso.py; after the force calculations, thirdorder_espresso.py scf.in reap 4 4 2 -7 was used for both compounds to retain anharmonic interactions through the seventh-nearest-neighbor shell. Kinetic-energy cutoffs of 65 and 520 Ry were used for the wave functions and charge density, and the Grimme D2 dispersion correction was included consistently to describe the interlayer interaction.[33, 34] The transport equation was sampled on a 20×20×620 \times20 \times6 phonon wave-vector mesh; both RTA and iterative tensors were checked, and the RTA tensor was used for the internally consistent zTzT estimates reported below.

Results and discussion

Crystal structures

Top and side views of tetragonal GeS₂

Figure 1. Crystal structure of tetragonal GeS₂ (space group P4₂/nmc): (a) top view and (b) side view. Blue and yellow spheres denote Ge and S atoms, respectively; translucent polyhedra in panel (a) show GeS₄ tetrahedra, and the projected unit cell is indicated in panel (b).

Figure 2. Crystal structure of tetragonal GeSe₂ (space group P4₂/nmc): (a) top view and (b) side view. Blue and green spheres denote Ge and Se atoms, respectively; translucent polyhedra in panel (a) show GeSe₄ tetrahedra, and the projected unit cell is indicated in panel (b).

GeX₂ (X = S, Se) crystallizes in a tetragonal structure with space group P4₂/nmc.[24, 25] The calculations were based on the reported tetragonal GeS₂ reference lattice, a=b=3.50a=b=3.50 Å and c=11.20c=11.20 Å, together with the corresponding internal coordinates.[17] The cell used in our GeS₂ SCF input is a=b=3.512a=b=3.512 Å and c=11.002c=11.002 Å, obtained from celldm(1) = 6.636200 and celldm(3) = 3.133016. These values remain close to the reported tetragonal reference structure.[24, 25]

The tetragonal GeSe₂ structure was generated by replacing S with Se in the same GeX₂ framework and using the corresponding SCF cell defined by celldm(1) = 6.998187741 and celldm(3) = 3.047055436. This gives a=b=3.703a=b=3.703 Å and c=11.284c=11.284 Å, with α=β=γ=90∘\alpha=\beta=\gamma=90^\circ. The larger lattice constants are consistent with the larger ionic size of Se compared with S. Because detailed experimental lattice parameters for GeSe₂ in the P4₂/nmc structure have not yet been reported, this geometry can serve as a first-principles reference for future studies. The final variable-cell relaxations gave residual pressures of only −0.03 kbar for GeS₂ and −0.01 kbar for GeSe₂; subsequent fixed-cell SCF calculations also remained close to zero pressure, with 0.13 and −0.02 kbar, respectively.

Figures 1 and 2 show that both compounds consist of corner-sharing GeX₄ tetrahedra arranged into layered GeX₂ sheets along the (0, 0, 1) direction. Each Ge⁴⁺ cation is coordinated by four chalcogen anions, forming a quasi-two-dimensional tetrahedral network analogous to a single slab of the high-pressure, layer-structured, tetragonal HgI₂-type phase.[25] The top views emphasize the in-plane periodic arrangement of Ge and chalcogen atoms, while the side views show the stacking of the GeX₂ layers along the cc axis.

Elastic properties

Elastic properties provide a direct measure of mechanical stability, stiffness, bonding anisotropy, and deformation response. The elastic constants were calculated with THERMO_PW, using the standard finite-strain approach interfaced with QUANTUM ESPRESSO.[30, 31, 50] For both compounds, the final static elastic calculations used frozen-ion = .false., allowing the internal atomic coordinates to relax under each imposed strain. GeS₂ was fitted using six strain geometries, a strain amplitude of 0.005, and a third-order polynomial. For the softer GeSe₂ lattice, a smaller strain amplitude of 0.002 with four strain geometries and a second-order polynomial was used to avoid unstable internal-coordinate relaxations. For tetragonal crystals, the stiffness tensor has six independent second-order elastic constants: C11C_{11}, C12C_{12}, C13C_{13}, C33C_{33}, C44C_{44}, and C66C_{66}. The calculated values are listed in Table 1. All values are reported in GPa after converting the THERMO_PW output from kbar using 1 kbar = 0.1 GPa.

Elastic constantGeS₂GeSe₂
C11C_{11}150.67124.05
C12C_{12}−2.264.60
C13C_{13}13.9320.92
C33C_{33}27.4837.32
C44C_{44}19.7027.04
C66C_{66}3.492.68

Table 1. Calculated relaxed-ion elastic constants of tetragonal GeS₂ and GeSe₂ in GPa.

The elastic constants satisfy the Born mechanical stability criteria for a tetragonal crystal: C11>∣C12∣C_{11} > \lvert C_{12}\rvert, 2C132<C33(C11+C12)2C_{13}^{2} < C_{33}(C_{11}+C_{12}), C44>0C_{44} > 0, and C66>0C_{66} > 0.[51] This confirms that the optimized tetragonal GeS₂ and GeSe₂ structures are mechanically stable against small homogeneous strains. Both compounds show strongly anisotropic stiffness: the in-plane constants are large, with C11=150.67C_{11}=150.67 GPa for GeS₂ and 124.05 GPa for GeSe₂, whereas the out-of-plane constants are much smaller, C33=27.48C_{33}=27.48 and 37.32 GPa, respectively. This contrast is consistent with the layered tetragonal framework and weaker resistance to deformation along the stacking direction.

The small C66C_{66} values indicate that both compounds are especially soft against in-plane shear deformation, even though their longitudinal and out-of-plane compression constants are much larger. This shear softness is important because soft deformation channels are often associated with anisotropic bonding, low phonon group velocities, and enhanced phonon scattering, all of which are favorable for reducing lattice thermal conductivity. The C12C_{12} values are small compared with C11C_{11}, indicating weak coupling between orthogonal in-plane normal strains.

ParameterGeS2_2GeSe2_2
BVB_V (GPa)42.2342.04
BRB_R (GPa)25.0032.82
BHB_H (GPa)33.6137.43
GVG_V (GPa)28.7927.28
GRG_R (GPa)10.649.46
GHG_H (GPa)19.7218.37
EVE_V (GPa)70.3867.29
ERE_R (GPa)27.9625.90
EHE_H (GPa)49.1746.60
νV\nu_V0.2220.233
νR\nu_R0.3140.368
νH\nu_H0.2470.268

Table 2. Voigt, Reuss, and Hill averaged relaxed-ion mechanical parameters of tetragonal GeS2_2 and GeSe2_2.

The Hill averages, which are the arithmetic means of the Voigt and Reuss bounds, give B=33.61B = 33.61 GPa, G=19.72G = 19.72 GPa, and E=49.17E = 49.17 GPa for GeS2_2.1 The corresponding values for GeSe2_2 are B=37.43B = 37.43 GPa, G=18.37G = 18.37 GPa, and E=46.60E = 46.60 GPa. These moderate values show that both tetragonal compounds are mechanically soft compared with densely bonded three-dimensional covalent solids. The Pugh ratio, B/GB/G, is 1.70 for GeS2_2 and 2.04 for GeSe2_2.2 Thus GeS2_2 lies slightly below the empirical ductile–brittle threshold of 1.75, while GeSe2_2 lies above it, indicating a somewhat more ductile average response.

The sound velocities and Debye temperatures obtained from the Voigt–Reuss–Hill elastic averages provide an additional indicator of heat transport. In the Slack model, the lattice thermal conductivity scales strongly with the Debye temperature, approximately through a θD3\theta_D^3 dependence.3 For GeS2_2, THERMO_PW gives bulk and shear sound velocities of 3.169 and 2.427 km s−1^{-1}, respectively, an average Debye sound velocity of 2.462 km s−1^{-1}, and a Debye temperature of 259.2 K. For GeSe2_2, the corresponding values are 2.750, 1.927, and 2.047 km s−1^{-1}, with a Debye temperature of 206.3 K. These low sound velocities and Debye temperatures are consistent with soft acoustic modes and support the low lattice thermal conductivity obtained from ShengBTE.

Overall, the elastic response confirms that both tetragonal compounds are mechanically stable and anisotropic, with pronounced shear softness that supports low lattice thermal conductivity and anisotropic thermoelectric behavior.

Phonon dispersion and vibrational density of states

Figure 3 contains 18 branches, as required for the six-atom primitive cell: three acoustic and 15 optical modes. No significant imaginary branch is observed along the sampled path; any tiny numerical deviation near Γ\Gamma, if present, is within the interpolation tolerance. The relaxed tetragonal GeS2_2 phase is therefore harmonically dynamically stable. This conclusion is complementary to the elastic stability conditions because the latter test only homogeneous strain, whereas the phonon calculation probes collective atomic displacements throughout the Brillouin zone. Similar use of phonon spectra to establish stability has been reported for other two-dimensional Ge–S phases, although their structures and dimensionality differ from the bulk tetragonal phase studied here.4

The spectrum separates into a lower manifold extending to approximately 7.2 THz and a higher optical manifold between about 9.7 and 13.6 THz, leaving a gap of roughly 2.5 THz. The comparatively dispersive low-frequency branches provide the main harmonic heat-carrying channels, whereas several optical branches are relatively flat and therefore have small group velocities. Their weak dispersion produces pronounced DOS maxima, particularly near 9.8–10.7 THz. Recent polarization-resolved Raman measurements and first-principles mode analysis of layered GeS2_2 assign the lower-frequency region mainly to Ge–S bending and tetrahedral breathing motions and the 340–450 cm−1^{-1} interval (approximately 10.2–13.5 THz) primarily to S-dominated Ge–S stretching.5 This assignment is consistent with the isolated high-frequency manifold and its upper limit near 13.6 THz in Figure 3.

Figure 3. DFPT phonon dispersion and total vibrational density of states of tetragonal GeS2_2. Frequencies are interpolated along Γ\Gamma–X–M–Γ\Gamma–Z–R–A–Z, while the DOS is integrated on a 40 × 40 × 20 uniform mesh. The non-analytical correction from the calculated dielectric tensor and Born effective charges is included.

Near Γ\Gamma, selected optical frequencies depend on whether the wave vector approaches in plane or along Γ\Gamma–Z. This is the expected non-analytic long-wavelength response produced by the anisotropic Born effective charges and dielectric screening, rather than an unstable or poorly converged mode.[46, 47] The two limiting approaches are therefore shown as separate path segments instead of being joined by an artificial vertical line. Finally, the harmonic dispersion and DOS alone do not determine phonon lifetimes: the dense, relatively flat optical manifold indicates low optical group velocities and provides many vibrational states, but the magnitude of anharmonic scattering must be established from third-order force constants and the phonon Boltzmann equation. Modern phonon-transport analyses likewise emphasize that group velocity, lifetime, and allowed scattering phase space must be considered together.[48, 55]

The corresponding GeSe2_2 spectrum in Figure 4 also contains three acoustic and 15 optical branches without significant imaginary frequencies, confirming harmonic stability of the substituted tetragonal structure. Replacing S by the heavier Se atom softens the complete spectrum: the upper frequency decreases from approximately 13.6 THz in GeS2_2 to 10.3 THz in GeSe2_2. This reduction is consistent with the mass dependence of lattice vibrations and with the weaker Ge–Se bonding inferred from the smaller magnitude of the Ge–Se ICOHP discussed below. The GeSe2_2 DOS is divided by reduced-DOS windows near 4.7–5.6 and 6.2–6.8 THz, and its strongest accumulation occurs around 7.8 THz, where several optical branches are nearly flat. In contrast, GeS2_2 exhibits a broader principal separation between the lower and upper manifolds, approximately 7.2–9.7 THz, and retains S-dominated stretching modes up to 13.6 THz.

Figure 4. DFPT phonon dispersion and total vibrational density of states of tetragonal GeSe2_2, calculated with the same path, non-analytical correction, and 40 × 40 × 20 DOS mesh used for GeS2_2.

The downward frequency shift and redistribution of optical states provide a harmonic basis for the different heat-transport responses of the two compounds, but they do not by themselves establish which material has the shorter phonon lifetime. The lower characteristic frequencies in GeSe2_2 increase thermal occupation at a given temperature, whereas the calculated conductivity additionally depends on branch-resolved velocities, anharmonic matrix elements, and scattering phase space. Accordingly, the dispersion/DOS comparison should be interpreted together with the calculated third-order force constants and κl\kappa_l, rather than as a standalone predictor of their conductivity ordering.[29, 49]

Spectral lattice thermal transport

Figure 5 summarizes the lattice-transport quantities most directly responsible for the thermal denominator of zTzT.[29] In the RTA treatment, GeS2_2 has κab=26.86\kappa_{ab}=26.86 W m−1^{-1} K−1^{-1} and κc=1.19\kappa_c=1.19 W m−1^{-1} K−1^{-1} at 300 K; these values decrease to 10.22 and 0.46 W m−1^{-1} K−1^{-1} at 800 K. GeSe2_2 shows smaller in-plane but slightly larger cross-plane thermal conductivity, with

κab=18.74\kappa_{ab}=18.74 W m−1^{-1} K−1^{-1} and κc=1.52\kappa_c=1.52 W m−1^{-1} K−1^{-1} at 300 K, decreasing to 7.25 and 0.58 W m−1^{-1} K−1^{-1} at 800 K. The iterative solution gives slightly larger values but preserves the same anisotropic contrast, as shown in the Supplementary Information. Because the thermoelectric analysis below uses the RTA tensor, the RTA values are used consistently in both the phonon and zTzT discussions. Following earlier lattice-transport analyses of low-κl\kappa_l thermoelectrics, we interpret these values using a combined descriptor set rather than a single metric: bonding strength, Grüneisen response, phonon velocities, lifetimes, and the frequency range over which heat-carrying modes accumulate all enter the final conductivity.[57, 58]

The temperature dependence provides an additional check on the physical origin of the low κc\kappa_c. Recent work on crystalline materials with intrinsically ultralow, glass-like heat transport shows that weak bonding, flat low-frequency rattling modes, strong multi-phonon scattering, and wave-like coherence can produce nearly temperature-independent κl\kappa_l.[59] The present GeX2_2 compounds behave differently: between 300 and 800 K, the calculated RTA conductivities follow approximately T−0.97T^{-0.97}–T−0.99T^{-0.99} for both in-plane and cross-plane components. Thus, tetragonal GeS2_2 and GeSe2_2 should be described as anisotropic low-cross-plane-conductivity crystals rather than glass-like thermal conductors within the present third-order ShengBTE treatment. Their low κc\kappa_c mainly reflects out-of-plane suppression of heat-carrying phonon velocities and finite anharmonic scattering in the layered network.

The RTA anisotropy ratio is larger in GeS2_2 (κab/κc≈22\kappa_{ab}/\kappa_c \approx22) than in GeSe2_2 (≈12.3\approx12.3–12.512.5), showing that Se substitution reduces the in-plane/cross-plane contrast even though both materials remain strongly anisotropic. The average mode Grüneisen parameter increases from 0.56 to 0.63 in GeS2_2 and from 0.65 to 0.68 in GeSe2_2 between 300 and 800 K, indicating moderate anharmonicity that strengthens with temperature. The weak-interlayer picture reported for layered GeSe2_2 is consistent with this anisotropic response, but our calculations also show that local Ge–Se bonds remain chemically stabilizing; therefore, the low κc\kappa_c should be assigned to anisotropic lattice dynamics rather than to a globally weak or mechanically unstable framework.[20] Together, the low cross-plane conductivity, persistent anisotropy, and finite anharmonic response explain why zTzT is larger along cc than in plane.

Figure 6. Mode-resolved ShengBTE scattering descriptors at 300 K for tetragonal GeS2_2 and GeSe2_2: (a,c) absolute mode Grüneisen parameters and (b,d) phonon–phonon lifetimes from final scattering rates. The cumulative conductivity, velocity, phase-space, spectral-conductivity, and iterative-tensor diagnostics are given in the Supplementary Information.

Figure 6 gives a mode-resolved view of the scattering channels behind the tensor averages in Figure 5. The largest ∣γλ∣|\gamma_\lambda| values occur mainly in low-frequency branches and selected optical groups, confirming that anharmonicity is distributed across several parts of the spectrum rather than confined to one isolated mode. This is important because recent first-principles thermoelectric studies show that low lattice conductivity is usually produced by the joint action of soft modes, reduced group velocities, short lifetimes, and enhanced scattering phase space, not by anharmonicity alone.[57, 58] At 300 K, many acoustic and low-lying optical modes retain picosecond-to-tens-of-picoseconds phonon–phonon lifetimes, while several higher-frequency optical modes have shorter lifetimes and contribute less to heat conduction because of their smaller velocities and weaker cumulative weight. Thus the low cross-plane conductivity is not caused by a single factor; it follows from anisotropic group velocities, finite anharmonic scattering, and the limited heat carried through the stacking direction.

Figure 5. Comparison of ShengBTE RTA lattice-transport descriptors for tetragonal GeS2_2 and GeSe2_2: (a) in-plane lattice thermal conductivity, (b) cross-plane lattice thermal conductivity, (c) anisotropy ratio κab/κc\kappa_{ab}/\kappa_c, and (d) average mode Grüneisen parameter.

Wannier-interpolated PBE and HSE03 band structures of tetragonal GeS2 and GeSe2

Figure 7. Wannier-interpolated PBE and HSE03 band structures of tetragonal GeS2_2 and GeSe2_2 along the Γ\Gamma–X–M–Γ\Gamma–Z–R–A–Z high-symmetry path. Panels (a) and (c) show the full dispersions, while panels (b) and (d) enlarge the corresponding band-edge regions. The y-axis in all panels is E−EVBME-E_{\mathrm{VBM}} in eV; panels (b) and (d) use a narrower y-axis window to resolve the band edges. For each functional, the energy zero is set to its own valence-band maximum; the colored horizontal guides and inset values identify the respective path band gaps.

The cumulative curves in the Supplementary Information show that most of the RTA lattice conductivity is accumulated below relatively low frequencies. For GeS2_2, 90% of the in-plane cumulative conductivity is reached by about 151 cm−1^{-1}, while the cross-plane component requires modes up to about 209 cm−1^{-1} because its total magnitude is small and more broadly distributed. For GeSe2_2, the corresponding 90% thresholds are lower, about 112 cm−1^{-1} for the in-plane component and 100 cm−1^{-1} for the cross-plane component. These lower characteristic frequencies are consistent with mass-induced softening after replacing S with Se, whereas the persistent separation between κab\kappa_{ab} and κc\kappa_c reflects the layered tetragonal network.

Insight from electronic structures

Figure 7 compares the PBE and HSE03 band dispersions after Wannier interpolation.[32, 36–42] Along the plotted high-symmetry path, the PBE gaps are 0.98 eV for GeS2_2 and 0.08 eV for GeSe2_2, whereas the corresponding HSE03/Wannier path gaps increase to 2.48 eV and 1.23 eV, respectively. The dense PBE mesh used for AMSET places GeSe2_2 even closer to gap closure, giving the small overlap noted in Sec. 2; therefore, both PBE descriptions consistently indicate that semilocal GeSe2_2 is near the semiconductor–semimetal boundary. Thus, HSE03 mainly shifts the conduction-band manifold upward while preserving the overall band topology near the valence-band edge. The much smaller gap of GeSe2_2 reflects the stronger energetic contribution of Se-derived states near the band edges and indicates easier carrier excitation than in GeSe2_2. Because semilocal PBE underestimates semiconductor band gaps, the HSE03 results are used as the electronic-structure reference for interpreting the thermoelectric transport trends, while PAW-PBE wave-functions are used only for the LOBSTER bonding analysis.

Carrier-dependent electronic transport

The AMSET transport tensors show a clear carrier asymmetry, as documented in the Supplementary Information. At 300 K and ∣N∣=1020 cm−3\lvert N\rvert= 10^{20}\ \mathrm{cm}^{-3}, trace-averaged σ\sigma is 4.96×1044.96 \times10^{4} (GeS2_2) and 2.76×104 S m−12.76 \times10^{4}\ \mathrm{S\ m}^{-1} (GeSe2_2) for n-type transport, compared with 3.24×1033.24 \times10^{3} and 1.39×103 S m−11.39 \times10^{3}\ \mathrm{S\ m}^{-1} for p-type transport. The corresponding n-type mobilities, 31.0 and 17.2 cm2 V−1 s−1\mathrm{cm}^{2}\ \mathrm{V}^{-1}\ \mathrm{s}^{-1}, exceed the p-type values of 2.02 and 0.49 cm2 V−1 s−1\mathrm{cm}^{2}\ \mathrm{V}^{-1}\ \mathrm{s}^{-1}. This difference combines band velocities with state-dependent scattering and cannot be attributed to effective mass or lifetime alone. Increasing concentration raises σ\sigma sublinearly because impurity scattering strongly reduces mobility; for n-type GeS2_2, increasing ∣N∣\lvert N\rvert from 102010^{20} to 1022 cm−310^{22}\ \mathrm{cm}^{-3} raises σ\sigma by only 2.4 times while μ\mu falls from 31.0 to 0.74 cm2 V−1 s−1\mathrm{cm}^{2}\ \mathrm{V}^{-1}\ \mathrm{s}^{-1}. [26, 27]

The κe\kappa_{\mathrm{e}} curves broadly follow σ\sigma, as both sample the same conducting states. At 1020 cm−310^{20}\ \mathrm{cm}^{-3} and 300 K, trace-averaged n-type κe\kappa_{\mathrm{e}} is 0.313 and 0.198 W m−1 K−1\mathrm{W\ m}^{-1}\ \mathrm{K}^{-1} for GeS2_2 and GeSe2_2, respectively. Heating lowers the n-type mobility in both systems, consistent with the shorter ADP+POP lifetimes at elevated temperature. [28] The p-type response of GeSe2_2 remains sensitive to the small PBE band overlap and should be regarded as qualitative until a gap-corrected AMSET calculation is performed. The carrier-concentration curves in the Supplementary Information connect these trends to the chemical-potential windows sampled by each carrier population. IMP scattering is included in the transport tensors but not in the plotted electron–phonon lifetime.

To clarify the bonding origin of the transport response, we performed LOBSTER analyses of GeS2_2 and GeSe2_2 using PAW-PBE single-point calculations on the same relaxed structures used in the main calculations. [60–63] PAW datasets were used only for this bonding step because LOBSTER reconstructs local-orbital populations and COHP/COOP/COBI descriptors from PAW-compatible wavefunctions. The PAW and ultrasoft calculations give nearly identical PBE band gaps for GeS2_2 (about 0.98–0.99 eV), indicating that the PAW calculation does not change the qualitative electronic structure. The low absolute charge spilling values of 0.90% for GeS2_2 and 0.84% for GeSe2_2 further confirm that the projection onto local orbitals is reliable for bonding analysis. [64]

The eight nearest-neighbor Ge–X bonds are nearly equivalent in both compounds, with bond lengths of 2.263 Å for Ge–S and 2.405 Å for Ge–Se. The ICOHP values are negative for both bonds, showing stabilizing Ge–chalcogen interactions; however, the magnitude decreases from about −5.27 eV per Ge–S bond to −4.74 eV per Ge–Se bond. The positive ICOOP values, 0.286 for GeS2_2 and 0.264 for GeSe2_2, indicate covalent orbital overlap in both compounds, while the Mulliken and Loewdin charges show charge transfer from Ge to the chalcogen atoms. Thus, the local Ge–S and Ge–Se bonds are best described as mixed covalent–ionic rather than purely ionic, with GeS2_2 showing slightly stronger local covalent bonding. Because the tetragonal P42/nmcP4_2/nmc structure is centrosymmetric, these local bond polarities cancel at the unit-cell level; therefore, bulk tetragonal GeS2_2 and GeSe2_2 are non-polar crystals despite having polar Ge–X bonds.

This bonding picture helps explain the thermoelectric behavior. Strong local Ge–X bonds preserve structural stability, but the layered tetrahedral framework remains elastically anisotropic and soft against shear, as reflected by the small C66C_{66} values and low Debye temperatures discussed above. Such a combination of mixed covalent–ionic bonding and soft anisotropic deformation channels can suppress phonon velocities and enhance phonon scattering without destroying the semiconducting electronic structure. The LOBSTER results therefore support the view that low lattice thermal conductivity in tetragonal GeX2_2 arises from structural anisotropy and shear softness rather than from weak local Ge–X bonding alone.

Anisotropic thermoelectric figure of merit

The anisotropic figure of merit was evaluated using Eq. (1), with SS, σ\sigma, and κe\kappa_{\mathrm{e}} taken from the AMSET carrier concentration and temperature and with the corresponding component of κl\kappa_{\mathrm{l}} taken from the ShengBTE RTA tensor. [4, 26, 29] The tensor components were combined consistently: xxxx electronic coefficients were paired with the in-plane lattice value κab=(κxx+κyy)/2\kappa_{ab} = (\kappa_{xx} + \kappa_{yy})/2, while the zzzz coefficients were paired with κc\kappa_{c}. This component-wise reconstruction follows the same self-consistency principle emphasized in recent thermoelectric data-curation work: reported zTzT values should be traceable to the underlying SS, σ\sigma, κ\kappa, and temperature values rather than treated as an independent number. [65] The heat maps in Figure 8 are plotted on the explicitly sampled carrier-concentration and temperature grid; the marked values are grid-point results rather than interpolation-derived optima. Recent first-principles analyses of thermoelectric materials emphasize the same point: zTzT is controlled by the simultaneous balance of Seebeck response, electrical conductivity, electronic thermal conductivity, and lattice thermal conductivity, so a low κl\kappa_l alone is not sufficient for high performance.[58, 66]

Figure 8. Cross-plane figure of merit zTczT_c of tetragonal GeS2_2 and GeSe2_2 obtained by combining AMSET electronic transport coefficients with the ShengBTE RTA lattice thermal conductivity component κc\kappa_c. The maps show the explicitly sampled temperature–carrier-concentration grid for n- and p-type transport; white circles mark the largest sampled value in each panel. Full component-resolved abab and cc carrier-concentration curves are provided in the Supplementary Information.

For n-type GeS2_2, the calculated cross-plane value is zTc=0.257zT_c = 0.257 at 800 K and n=1×1019 cm−3n = 1 \times10^{19}\ \mathrm{cm}^{-3}. At this point, the cross-plane power factor is only 0.159 mW m−1 K−20.159\ \mathrm{mW\ m^{-1}\ K^{-2}}, while κe+κl\kappa_e + \kappa_l is 0.496 W m−1 K−10.496\ \mathrm{W\ m^{-1}\ K^{-1}}. Thus, the low κc\kappa_c helps the denominator, but the numerator remains limited by modest cross-plane electrical conductivity (3.0×103 S m−13.0 \times10^3\ \mathrm{S\ m^{-1}}) and a finite Seebeck coefficient of −230 μV K−1-230\ \mathrm{\mu V\ K^{-1}} [Fig. 9]. At 300 K, the corresponding cross-plane value is 0.044 at 3×1018 cm−33 \times10^{18}\ \mathrm{cm}^{-3}. The favorable n-type concentration shifts upward with temperature because the conductivity gain from additional carriers initially outweighs the simultaneous reduction in ∣S∣|S| and increase in κe\kappa_e. In-plane n-type performance is much smaller, reaching only zTab=0.019zT_{ab} = 0.019 at 800 K and 3×1019 cm−33 \times10^{19}\ \mathrm{cm}^{-3}, chiefly because the in-plane lattice conductivity remains large.

The p-type response follows a different concentration dependence. At 800 K, the calculated values are 0.039 in plane at p=1×1021 cm−3p = 1 \times10^{21}\ \mathrm{cm}^{-3} and 0.104 along cc at p=1×1022 cm−3p = 1 \times10^{22}\ \mathrm{cm}^{-3}. The latter value occurs at the upper boundary of the concentration grid, so it should not be interpreted as a fully resolved optimum.

The Lorenz number in Figure 9(d) also varies with carrier type, transport component, and concentration, indicating that a fixed Wiedemann–Franz Lorenz number would not capture the electronic heat transport quantitatively in the dilute-to-moderately doped regime.[66, 67] Overall, cross-plane n-type transport gives the strongest calculated zTzT, despite the larger in-plane electrical conductivity, because the more than twofold reduction of κl\kappa_l along cc produces a substantially smaller thermal denominator. The predicted response remains moderate, indicating that pristine tetragonal GeS2_2 is not a high-zTzT material under the present scattering model, although its strong anisotropic contrast may be relevant when transport orientation can be controlled.

Figure 9. Transport factors controlling the 800 K thermoelectric response of tetragonal GeS2_2: (a) power factor, (b) total thermal conductivity entering the denominator of zTzT, (c) resulting zTzT, and (d) Lorenz number normalized by the Sommerfeld value L0=2.44×10−8 W Ω K−2L_0 = 2.44 \times10^{-8}\ \mathrm{W\ \Omega\ K^{-2}}. The dotted horizontal guides in panel (b) mark the RTA lattice components κab\kappa_{ab} and κc\kappa_c.

For GeSe2_2, the updated PBE-AMSET estimate gives smaller n-type values than for GeS2_2, with zTc=0.027zT_c = 0.027 at 800 K and n=1×1018 cm−3n = 1 \times10^{18}\ \mathrm{cm}^{-3}. For p-type cross-plane transport, the calculated value is zTc=0.066zT_c = 0.066 at 800 K and p=3×1020 cm−3p = 3 \times10^{20}\ \mathrm{cm}^{-3}. At this point, the cross-plane power factor is 0.050 mW m−1 K−20.050\ \mathrm{mW\ m^{-1}\ K^{-2}} and κe+κl\kappa_e + \kappa_l is 0.600 W m−1 K−10.600\ \mathrm{W\ m^{-1}\ K^{-1}}, so the low cross-plane lattice contribution is partly offset by modest electrical conductivity and a finite electronic heat current. Since the GeSe2_2 AMSET tensors are based on the semilocal PBE electronic structure, and the HSE03/Wannier calculation opens a substantially larger gap, the GeSe2_2 zTzT values should be regarded as qualitative until a gap-corrected scattering calculation is performed.

These zTzT values inherit the approximations of the two transport calculations. In particular, the AMSET inputs use PBE-derived bands and approximate orbital-overlap factors, while the ShengBTE RTA treatment neglects the iterative redistribution of nonequilibrium phonon populations.[26, 29] Grain boundaries, defects, and additional high-temperature scattering channels could change κl\kappa_l, whereas gap-corrected AMSET calculations could modify the electronic coefficients, especially for GeSe2_2.

Conclusion

We have presented a first-principles study of tetragonal GeS2_2 and GeSe2_2 that connects structure, bonding, lattice dynamics, and anisotropic thermoelectric transport within one consistent framework. Both compounds are mechanically and dynamically stable in the calculated tetragonal P42/nmcP4_2/nmc structure. HSE03/Wannier interpolation gives semiconducting gaps of 2.48 eV for GeS2_2 and 1.23 eV for GeSe2_2, while LOBSTER analysis shows mixed covalent–ionic Ge–X bonding. The stronger Ge–S interaction, reflected by the larger ICOHP magnitude (−5.27 eV per bond compared with −4.74 eV for Ge–Se), is consistent with the higher GeS2_2 phonon-frequency range.

The central result is the strong anisotropic separation between in-plane and cross-plane heat flow. At 300 K, the ShengBTE RTA lattice thermal conductivity changes from 26.86 to 1.19 W m−1^{-1} K−1^{-1} between the in-plane and cc components for GeS2_2, and from 18.74 to 1.52 W m−1^{-1} K−1^{-1} for GeSe2_2. At 800 K, the corresponding cross-plane values fall to 0.46 and 0.58 W m−1^{-1} K−1^{-1}. This suppressed out-of-plane thermal transport arises from the layered tetrahedral framework, shear softness, low-frequency heat-carrying phonons, and moderate anharmonicity.

When these lattice tensors are combined with scattering-aware AMSET electronic transport, the calculated thermoelectric response is strongest along the cross-plane direction. The calculated value is zTc=0.257zT_c = 0.257 for n-type GeS2_2 at 800 K and n=1×1019n = 1 \times10^{19} cm−3^{-3}. For GeSe2_2, the PBE-AMSET estimate gives zTc=0.066zT_c = 0.066 for p-type transport at 800 K and p=3×1020p = 3 \times10^{20} cm−3^{-3}, although this value should be treated as qualitative because the PBE transport bands are close to a band-overlap limit. Overall, the study shows that low κl\kappa_l alone is not enough to produce high zTzT: the power factor, mobility degradation at high carrier concentration, electronic heat transport, and Lorenz-number variation all control the final performance. Tetragonal GeX2_2 compounds therefore emerge as anisotropic thermoelectrics with moderate calculated zTzT values, with the most promising transport channel along the cc axis. Future gap-corrected scattering calculations, defect engineering, alloying, and strain control should be the most direct routes for testing whether the favorable cross-plane thermal response can be converted into higher thermoelectric efficiency.

Declaration of competing interest

The authors declare that they have no known conflicts of interest associated with this article.

Computing facilities

The computations were performed using the CINECA Leonardo supercomputing facility during the Master in High Performance Computing (MHPC) thesis work at the Abdus Salam International Centre for Theoretical Physics (ICTP).

CRediT authorship contribution statement

Neeraj Kulhari: Writing–original draft, Validation, Methodology, Investigation, Formal analysis, Data curation.

Krishna Swaroop Sharma: Writing–review and editing, Validation, Supervision, Project administration, Conceptualization.

K. C. Bharmu: Writing–review and editing, Validation, Supervision, Project administration, Conceptualization.

Computational Settings

The main text summarizes the methods used for the structural, elastic, electronic, phonon, and transport calculations. Table S1 collects the key numerical settings so that the reported figures can be reproduced more easily. The HSE03/Wannier band plots are used as band-structure references, while the quantitative thermoelectric figure of merit combines AMSET electronic coefficients with the ShengBTE relaxation-time-approximation (RTA) lattice thermal conductivity tensor.

CalculationMain settings
PBE structural/electronic calculationsQUANTUM ESPRESSO; PBE+DFT-D2; scalar-relativistic ultrasoft RRKJ PSLibrary pseudopotentials for the structural, elastic, band-structure, DFPT, deformation-potential, and AMSET-input calculations; 8×8×48 \times8 \times4 kk mesh; wave-function and charge-density cutoffs of 60 and 480 Ry.
HSE03 band referenceScreened hybrid functional using the same ultrasoft RRKJ PSLibrary pseudopotentials, with exx_fraction=0.25, screening_parameter=0.106, and 6×6×26 \times6 \times2 kk mesh.
Wannier interpolationPBE and HSE03 bands interpolated using WANNIER90; the plotted path is Γ\Gamma–X–M–Γ\Gamma–Z–R–A–Z.
Elastic constantsTHERMO_PW finite-strain workflow with relaxed internal coordinates, frozen_ions = .false.. GeS2\mathrm{GeS_2} used six strain geometries, strain amplitude 0.005, and third-order polynomial fitting. GeSe2\mathrm{GeSe_2} used four strain geometries, strain amplitude 0.002, and second-order polynomial fitting to stabilize the strained-structure relaxations.
AMSET transportADP, POP, and IMP scattering. AMSET post-processing interpolation meshes: 81×81×2581 \times81 \times25 for GeS2\mathrm{GeS_2} and 81×81×2781 \times81 \times27 for GeSe2\mathrm{GeSe_2}; these are separate from the Quantum ESPRESSO SCF/NSCF kk meshes. Carrier concentrations span 101810^{18} to 102210^{22} cm−3\mathrm{cm}^{-3}.
BoltzTraP2 diagnosticsHSE03/Wannier bands post-processed in the constant-relaxation-time approximation to inspect chemical-potential trends in SS, σ/τ\sigma/\tau, κe/τ\kappa_{\mathrm{e}}/\tau, power factor, Lorenz number, and DOS.
Phonons and ShengBTEHarmonic force constants from DFPT using the same PBE+DFT-D2 ultrasoft pseudopotential workflow; non-analytical correction from Born effective charges and dielectric tensor; third-order force constants from 4×4×24 \times4 \times2 finite-displacement supercells generated with thirdorder_expresso.py; thirdorder_expresso.py scf.in reap 4 4 2 -7 used for both compounds to retain interactions through the seventh-nearest-neighbor shell; ShengBTE transport sampled on a 20×20×620 \times20 \times6 phonon mesh.

Table S1. Main numerical settings used in the calculations.

Electronic-Structure and BoltzTraP2 Diagnostics

Figure S1 and Figure S2 show the HSE03/Wannier BoltzTraP2 transport curves plotted against chemical potential relative to the Fermi level. These plots are useful for identifying the carrier-type asymmetry and the sensitivity of SS, σ/τ\sigma/\tau, κe/τ\kappa_{\mathrm{e}}/\tau, and PF to the band-edge shape. They are not used directly for the final zTzT values because the main thermoelectric estimates use AMSET state-dependent scattering.

GeS₂ transport coefficients as functions of chemical potential, with panels (a) $S$ ($\mu$V K⁻¹), (b) $\sigma$ ($10^{6}$ S m⁻¹), (c) $\kappa_{\mathrm{e}}$ (W m⁻¹ K⁻¹), and (d) $S^{2}\sigma$ (mW m⁻¹ K⁻²), for 300 K, 500 K, and 800 K

Figure S1. HSE03/Wannier BoltzTraP2 transport coefficients of GeS₂ as functions of chemical potential for 300, 500, and 800 K.

Figure S2. HSE03/Wannier BoltzTraP2 transport coefficients of GeSe₂ as functions of chemical potential for 300, 500, and 800 K.

The Lorenz number, L=κe/(σT)L = \kappa_{\mathrm{e}}/(\sigma T), is shown in Figure S3. A constant Wiedemann–Franz value is a good approximation only for highly degenerate carriers. The present curves vary with chemical potential, temperature, and direction, showing that a fixed Lorenz number would be an oversimplification in the dilute and near-band-edge regions.

Figure S3. Lorenz number obtained from the BoltzTraP2 HSE03/Wannier transport tensors.

The electronic DOS from the trace files is plotted together with the Seebeck coefficient in Figure S4. Peaks in the DOS close to the band edges are relevant because a sharp transport distribution can enhance the Seebeck response, as expected from the Mahan–Sofo picture of thermoelectric optimization [1].

Figure S4. Comparison of electronic DOS and Seebeck coefficient trends from the HSE03/Wannier BoltzTraP2 calculations.

AMSET Carrier and Scattering Diagnostics

The AMSET carrier-dependent transport curves in Figure S5 and Figure S6 show how σ\sigma, κe\kappa_{\mathrm{e}}, and mobility vary with the sampled electron and hole concentrations. The separate 300 K concentration–chemical-potential plots below connect these carrier densities to the chemical-potential window used in the transport tensors. This is especially important for GeSe2_2, where the PBE transport bands are close to a band-overlap limit.

Figure S5. AMSET electrical conductivity, electronic thermal conductivity, and mobility of GeS₂ as functions of electron and hole concentration.

Figure S6. AMSET electrical conductivity, electronic thermal conductivity, and mobility of GeSe₂ as functions of electron and hole concentration.

Anisotropic zTzT Data

The main text reports selected zTzT values. Figure S10 and Figure S11 give the full component-resolved curves as functions of carrier concentration.

Figure S10. Carrier-concentration-dependent component-resolved zTzT values of GeS2_2 obtained by combining AMSET electronic coefficients with ShengBTE RTA lattice thermal conductivity.

Figure S11. Carrier-concentration-dependent component-resolved zTzT values of GeSe₂ obtained by combining AMSET electronic coefficients with ShengBTE RTA lattice thermal conductivity.

CompoundBranchConcentration (cm−3^{-3})zTzT
GeS2_2n-type, abab3×10193 \times10^{19}0.019
GeS2_2n-type, cc1×10191 \times10^{19}0.257
GeS2_2p-type, abab1×10211 \times10^{21}0.039
GeS2_2p-type, cc1×10221 \times10^{22}0.104
GeSe2_2n-type, abab1×10181 \times10^{18}0.003
GeSe2_2n-type, cc1×10181 \times10^{18}0.027
GeSe2_2p-type, abab1×10201 \times10^{20}0.006
GeSe2_2p-type, cc3×10203 \times10^{20}0.066

Table S2. Directional zTzT values at 800 K from the sampled carrier-concentration grid. The listed concentration is the grid point where the largest value for that branch occurs, not an interpolated optimum.

ShengBTE Lattice-Thermal-Transport Details

The main manuscript contains the compact mode-resolved comparison used for the discussion. The figures below retain the full per-material ShengBTE diagnostics: RTA and iterative conductivity tensors, anisotropy, cumulative conductivity, average Grüneisen parameters, frequency-resolved velocities, three-phonon phase space, lifetimes, and spectral conductivity. Table 3 lists the RTA values used in the main zTzT calculation.

DescriptorGeS2_2GeSe2_2
Charge spilling0.90%0.84%
Ge–X bond length (Å)2.2632.405
ICOHP (eV bond−1^{-1})−5.27−4.74
ICOOP0.2860.264
ICOBI0.900–
Mulliken charge Ge/X (ee)+0.54/−0.27+0.36/−0.18
Loewdin charge Ge/X (ee)+0.43/−0.21+0.27/−0.13

Table 3. LOBSTER bonding descriptors for tetragonal GeS2_2 and GeSe2_2 obtained from PAW-PBE single-point calculations on the relaxed structures. ICOHP and ICOOP values are reported per nearest-neighbor Ge–X bond at the Fermi level.

Figure S7. Electron and hole concentrations of GeS2_2 at 300 K as functions of chemical potential.

Figure S8. Electron and hole concentrations of GeSe2_2 at 300 K as functions of chemical potential.

Figure S9. AMSET relaxation-time comparison for GeS₂ and GeSe₂. The plotted lifetime contains the ADP and POP contributions; ionized-impurity scattering is included in the AMSET transport tensors but is not shown in this lifetime plot.

CompoundTT (K)κab\kappa_{ab} (W m−1^{-1} K−1^{-1})κc\kappa_c (W m−1^{-1} K−1^{-1})κab/κc\kappa_{ab}/\kappa_c
GeS₂30026.861.1922.58
GeS₂50016.250.7322.21
GeS₂80010.220.4622.03
GeSe₂30018.741.5212.29
GeSe₂50011.460.9212.45
GeSe₂8007.250.5812.55

Table S3. Selected RTA lattice thermal conductivities used in the thermoelectric analysis.

Figure S12. Transport factors controlling the 800 K thermoelectric response of tetragonal GeSe₂: (a) power factor, (b) total thermal conductivity entering the denominator of zTzT, (c) resulting zTzT, and (d) Lorenz number normalized by the Sommerfeld value L0=2.44×10−8 W Ω K−2L_0 = 2.44 \times10^{-8}\ \mathrm{W}\ \Omega\ \mathrm{K}^{-2}. This plot is the GeSe₂ counterpart of the GeS₂ factor decomposition shown in the main text.

Figure S13. Detailed ShengBTE lattice-transport summary for GeS2_2.

Figure S14. Detailed ShengBTE lattice-transport summary for GeSe2_2.

Six-panel plot titled “GeS₂ phonon transport (ShengBTE)” with in-plane (ab) and out-of-plane (c) curves, showing phonon group velocity, weighted phase space, phonon–phonon relaxation time, spectral thermal conductivity, cumulative thermal conductivity, and temperature-dependent thermal conductivity

Figure S15. Additional GeS₂ spectral phonon-transport descriptors from ShengBTE, including branch-resolved and cumulative quantities.

Figure S16. Additional GeSe2\mathrm{GeSe}_{2} spectral phonon-transport descriptors from ShengBTE, including branch-resolved and cumulative quantities.

LOBSTER Bonding Setup

The LOBSTER projections were performed using PAW-PBE single-point calculations because LOBSTER requires PAW-compatible wavefunctions. For GeS2, the local basis was Ge 3d 4s 4p and S 3s 3p. For GeSe2, the basis was Ge 3d 4s 4p and Se 4s 4p. COHP, COOP, and COBI pairs were generated over nearest-neighbor Ge–X distances; the resulting charge spilling values were below 1% for both compounds. The main manuscript reports the averaged nearest-neighbor ICOHP, ICOOP, and charge-transfer descriptors.

References

  1. [1]1 G. D. Mahan and J. O. Sofo, “The best thermoelectric,” Proceedings of the National Academy of Sciences of the United States of America 93(15), 7436–7439 (1996). doi:10.1073/pnas.93.15.7436.DOI

Paper details

vlnc:W_c39uqr3a
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents