Phonon-Glass Electron-Crystal Design

• PGEC is a design concept that minimizes lattice thermal conductivity while maintaining high electron mobility to maximize thermoelectric efficiency.
• Realization approaches include nanostructured semiconductors, clathrates with rattling guest ions, and molecular crystals with quasi-1D bands to control heat and charge transport.
• Data-driven screening using the lattice-to-total thermal conductivity ratio accelerates the discovery and optimization of high-performance thermoelectric materials.

The phonon-glass electron-crystal (PGEC) design concept represents a fundamental paradigm in the search for high-performance thermoelectric materials. PGEC envisions materials that, at the microscopic scale, suppress lattice (phononic) heat conduction as in structural glasses, while retaining the band-like (delocalized) electrical transport typical of high-mobility crystals. The concept operationalizes the tradeoff central to maximizing the thermoelectric figure of merit, $$zT = S^2 \sigma T / (\kappa_\mathrm{el} + \kappa_\mathrm{lat})$$, where a large Seebeck coefficient $S$, high electrical conductivity $\sigma$, and minimal lattice thermal conductivity $\kappa_\mathrm{lat}$ yield optimal conversion efficiency. Approaches to PGEC span nanostructured crystalline semiconductors such as Si or Ge with mesoscale disorder, complex bulk crystals hosting dynamic guest atoms (clathrates), and molecular crystals exhibiting quasi-one-dimensional electronic bands embedded in low-frequency phonon glass frameworks. The PGEC design principle now extends into machine-learning-accelerated materials discovery, with quantitative descriptors such as the lattice-to-total thermal conductivity ratio providing actionable metrics for large-scale screening.

1. PGEC Principle: Quantitative Framework

The PGEC paradigm is formally defined by the simultaneous minimization of $\kappa_\mathrm{lat}$ (phonon-glass behavior) and maximization of $\sigma$ (electron-crystal behavior). The efficiency metric is

$$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$

where $\kappa_\mathrm{e}$ and $\kappa_\mathrm{lat}$ denote electronic and lattice contributions to the total thermal conductivity $\kappa$. Slack's original proposal established that ideal thermoelectric performance is achieved by combining glass-like phonon transport—manifested as low Debye-level $S$0 or plateaued, amorphous-like $S$1—with metallic or crystalline electronic properties—large $S$2 (via high-mobility, high carrier concentration) and finite $S$3 (Mott relation) (Takabatake et al., 2014).

A rigorous PGEC descriptor has emerged: the lattice-to-total thermal conductivity ratio $S$4. High-$S$5 materials consistently exhibit $S$6, indicating a near-equal sharing of heat conduction between phonons and electrons, thus defining a quantitative target for both discovery and device optimization (Sun et al., 26 Nov 2025).

2. Mechanisms for Realizing PGEC Behavior

The suppression of $S$7 without compromising $S$8 is achieved by selectively scattering phonons while permitting carrier percolation through crystalline domains. Key mechanisms include:

• Nanostructuring: Introduction of mesoscale features (nanopores, inclusions, defective regions) with characteristic length $S$9 lying between the electron mean free path $\sigma$0 and the phonon mean free path $\sigma$1: $\sigma$2. Phonons experience strong boundary or defect scattering (Matthiessen’s rule), while electrons retain long-range coherence (Bah et al., 2022, Zhu et al., 2017, Tainoff et al., 2013).
• Phonon Resonance and Rattlers: In host-guest systems such as clathrates, loosely bound guest ions ("rattlers") within a crystalline framework introduce low-frequency local Einstein oscillators that resonantly scatter propagating acoustic phonons, destroying the correlation length for heat transport (resonant and TLS tunneling models) without disrupting electronic delocalization (Takabatake et al., 2014).
• Quasi-Low-Dimensional Electronic Bands: In molecular or organic frameworks, $\sigma$3–$\sigma$4 stacking yields quasi-1D dispersive bands embedded within a van der Waals (vdW) lattice of weakly coupled stacks. The result is a sharp electronic density of states at band edges (boosting $\sigma$5) with glassy, ultra-low $\sigma$6 due to weak bonding (Mi et al., 2015).

3. Experimental and Computational Realizations

Nanopatterned Crystalline Si and Ge

• Nanoporous Silicon Devices: Arrays of 40 nm pores at 100 nm pitch in 62 nm Si yield a reduction in $\sigma$7 by a factor of $\sigma$8 (to 83 W/mK) with only a minor decrease in $\sigma$9 (down to half the bulk value under defect-limited conditions), and a $\kappa_\mathrm{lat}$0 enhancement in $\kappa_\mathrm{lat}$1 via modified $\kappa_\mathrm{lat}$2 around $\kappa_\mathrm{lat}$3 (Mott relation) (Bah et al., 2022). Power densities up to 6 mW/cm$\kappa_\mathrm{lat}$4 for $\kappa_\mathrm{lat}$5 K and 400 nW/cm$\kappa_\mathrm{lat}$6 at $\kappa_\mathrm{lat}$7 K have been demonstrated in CMOS-compatible architectures.
• Ion-Beam Patterned Silicon: Focused Xe$\kappa_\mathrm{lat}$8 irradiation yields cylindrical amorphous inclusions (diameter $\kappa_\mathrm{lat}$9 nm, spacing $\kappa_\mathrm{lat}$0 nm), reducing $\kappa_\mathrm{lat}$1 by $\kappa_\mathrm{lat}$2 (to $\kappa_\mathrm{lat}$3 W/mK) with $\kappa_\mathrm{lat}$4 retaining $\kappa_\mathrm{lat}$5 of bulk and S unperturbed, resulting in $\kappa_\mathrm{lat}$6 up to 0.5 at room temperature (Zhu et al., 2017).
• Mn-Doped Germanium Films: Ge$\kappa_\mathrm{lat}$7Mn$\kappa_\mathrm{lat}$8 thin films with $\kappa_\mathrm{lat}$910-25 nm Ge$\sigma$0Mn$\sigma$1 inclusions achieve $\sigma$2 suppression by $\sigma$3 (to 2 W/mK), maintain $\sigma$4 S/m, and exhibit $\sigma$5 at 300 K. Interface quality and nominally bulk electronic mean-free-path ($\sigma$6) ensure electron-crystal behavior (Tainoff et al., 2013).

Cage-Structured Clathrates

Type-I clathrate compounds (e.g., Ba$\sigma$7Ga$\sigma$8Ge$\sigma$9, Sr$$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$0Ga$$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$1Ge$$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$2) combine a rigid crystalline cage with encaged guest atoms. The guest ion “rattler” modes (Einstein modes with $$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$3 K) sharply reduce $$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$4 to values as low as 0.4 W/mK (e.g., in $$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$5-Ba$$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$6Ga$$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$7Sn$$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$8) while the dispersive conduction bands ensure metallic electrical behavior. These systems routinely achieve $$zT = \frac{S^2 \sigma T}{\kappa_\mathrm{e} + \kappa_\mathrm{lat}}$$9 at high temperatures (Takabatake et al., 2014).

Quasi-1D Electronic Structures in Phonon-Glass Frameworks

Bulk Bis-Dithienothiophene molecular crystals (BDTMC) present a π–π stacked structure yielding quasi-one-dimensional valence bands and “glassy” phonon transport (κ$\kappa_\mathrm{e}$0 W/mK at 300 K). DFT+Boltzmann and Green-Kubo MD calculations yield $\kappa_\mathrm{e}$1 under optimum hole-doping (p$\kappa_\mathrm{e}$2 cm$\kappa_\mathrm{e}$3, $\kappa_\mathrm{e}$4 μV/K, $\kappa_\mathrm{e}$5 S/m) (Mi et al., 2015).

4. Data-Driven PGEC Descriptor and Large-Scale Screening

Predictive PGEC optimization now integrates experimental informatics and machine learning. The lattice-to-total thermal conductivity ratio $\kappa_\mathrm{e}$6 emerges as a robust descriptor: high-$\kappa_\mathrm{e}$7 compounds universally cluster near $\kappa_\mathrm{e}$8, confirming the PGEC ideal as a quantitative metric (Sun et al., 26 Nov 2025). Tree-based regressors (Random Forest) trained on $\kappa_\mathrm{e}$970,000 entries enable rapid prediction of $\kappa_\mathrm{lat}$0, $\kappa_\mathrm{lat}$1, and $\kappa_\mathrm{lat}$2, allowing high-throughput screening of $\kappa_\mathrm{lat}$3100,000 systems. Application to Materials Project databases yielded 2,522 ultralow-$\kappa_\mathrm{lat}$4 candidates, with R-guided chemical modification (doping, alloying) tuning $\kappa_\mathrm{lat}$5 and $\kappa_\mathrm{lat}$6 to target the $\kappa_\mathrm{lat}$7 regime and optimize $\kappa_\mathrm{lat}$8.

5. Materials Design Guidelines and Optimization Strategies

Table: Representative PGEC Realizations and Design Parameters

System	PGEC Suppression Mechanism	Reported $\kappa_\mathrm{lat}$9
Nanoporous Si	$\kappa$0 nm, $\kappa$1 nm pores	$\kappa$20.02
Si irradiated	$\kappa$3 nm, $\kappa$4 nm amorphous	$\kappa$50.5
Ge:Mn Films	$\kappa$6–25 nm, $\kappa$7–100 nm	0.15
Clathrates	Guest "rattlers" in cages	0.5–0.7 (@800K)
BDTMC π–π Org.	Q1D HOMO bands, vdW glass	1.48

Optimizing PGEC materials requires:

• Ensuring geometric separation between electron and phonon mean free paths: $\kappa$8
• Maximizing interface roughness for diffuse phonon scattering (minimal specularity, $\kappa$9)
• Engineering resonant scattering (rattler scale/frequency) or mass/dynamic disorder
• Modulating carrier concentration and band offset (fine-tune S, decouple S/$S$00)
• Using hierarchical or multi-scale defect structures for broad-band phonon blocking
• Integrating machine-learned doping/alloying schemes to shift $S$01 as indicated by computational screening (Sun et al., 26 Nov 2025)

6. Limitations and Future Directions

While substantial $S$02 suppression is achieved, electrical resistance and interface transport still often limit absolute $S$03 values (e.g., in nanoporous Si devices $S$04 due to residual high $S$05 W/mK and contact resistance (Bah et al., 2022)). The amorphous limit ($S$06 W/mK for Si) requires more aggressive nanostructuring, hierarchical feature introduction, or band engineering. For clathrates, tuning guest-cage coupling, incorporating mixed guests, and optimizing stoichiometry and carrier type offer further paths. For molecular crystals, chemical/strain modulation of inter-stack coupling and impurity scattering minimization are active levers (Mi et al., 2015).

Data-driven frameworks now enable rapid identification of low-$S$07, high-$S$08 regimes, with the $S$09 metric supporting efficient screening and compositional tuning (Sun et al., 26 Nov 2025). Integrative approaches combining nanoscale phonon engineering, electronic band design, and high-throughput computational discovery define the contemporary landscape of PGEC research.

7. Perspectives and Emerging Research Directions

PGEC is now an established yet evolving branch in thermoelectrics, with multidimensional approaches converging: bottom-up nanostructuring in monolithic semiconductors, top-down guest-host frameworks, and emergent low-dimensionality in organic or hybrid systems. Further development will likely exploit:

• Phononic crystals and coherent phonon bandgap engineering
• Carrier-selective scattering/energy filtering at engineered boundaries
• Automated materials optimization via explainable AI models targeting the ideal $S$10
• Environmentally benign, earth-abundant compositions (clathrates, silicides, chalcogenides)
• On-chip integration for energy harvesting/cooling in microelectronics and IoT (Bah et al., 2022, Sun et al., 26 Nov 2025)

High-throughput computational and experimental advances continue to expand the PGEC design space, providing actionable guidance for synthesizing next-generation thermoelectric materials with optimized performance.