Thermoelectric properties of metal-semiconductor percolation composites

arXiv:2609.05273 · physics.app-ph, cond-mat.mtrl-sci · Submitted 2026-09-04 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "Thermoelectric properties of metal-semiconductor percolation composites".

Title and authors: Kai: So we're looking at this paper, "Thermoelectric properties of metal-semiconductor percolation composites." It seems like they're tackling a really complex problem: how to predict how these composite materials will actually perform when you mix a semiconductor host, like BiSbTe, with randomly placed metallic inclusions like copper.

Mira: Exactly. The core idea here is that these materials are super sensitive to the percolation transition, which is basically when the metal starts forming a connected network inside the semiconductor. It’s not just about mixing things; it’s about how those connections form and what that does to the electrical and thermal properties.

Lev: So, before we get into how they did it, what's the big picture takeaway here for someone working on real hardware? Are these predictions something you can actually trust when you build a device?

Kai: Well, they use a commercial finite element partial differential equation solver called COMSOL Multiphysics to run simulations. They set up a bulk semiconductor host with copper inclusions that are modeled as spheres, and they do three main things: first, they calculate the electrical and thermal conductivities by ignoring the weak thermoelectric coupling; second, they run one simulation that gives them a self-consistent solution for thermoelectric generation at a temperature difference of ten degrees Celsius between the top and bottom surfaces <ref:2609.05273#pg2>.

Mira: And that last simulation is key because it uses both metrics: the power factor, which is P factor equals S squared times sigma, and the dimensionless figure of merit zT, which is S squared times sigma T divided by lambda. They are looking at how these two things change as you vary the concentration and arrangement of those inclusions.

Lev: That sounds computationally intensive. When you run simulations like that on real hardware or even in error correction environments, what's the main overhead we should worry about?

Kai: The paper shows that near the percolation threshold, both the power factor and zT strongly depend on both how concentrated you make your inclusions and how they are arranged. They specifically highlight some promising regimes for enhancing the power factor in realistic composites.

Title and authors: Mira: That’s interesting because it suggests that instead of just focusing on one perfect concentration, we have to consider the geometry too. They find that when inclusions are more elongated, they can lead to a smaller percolation threshold pc, which is pretty important for designing things.

Lev: So if we're designing a material, should we focus on getting the concentration just right around that percolation point?

Kai: Not exactly just the point itself. The paper suggests that the best performance is actually close to that threshold where you get a doubling of the power factor with almost no reduction in zT compared to a bare semiconductor. They also mention enhancing performance if you can control how much thermal conductivity of the host material decays, which they suggest looking into in other research.

Mira: And they do point out that the shape of the inclusion matters too. They say spherical inclusions are pretty close to optimal for composites where the metal placement and orientation are random. But if you change that arrangement, things shift a lot.

Lev: So, what does this mean practically for someone who has to choose between using spheres or maybe some wires or layers when they are designing these composites?

Kai: The paper suggests that elongated inclusions can define the percolation threshold with smaller values of pc. But for random placements and orientations, spherical inclusions seem to be the closest thing to optimal in their analysis.

Mira: It also notes that the actual response of these materials in three dimensions doesn't just follow simple effective medium theories or layered media predictions, which is a big point because those simpler models don't capture how complex internal structures behave.

Lev: That makes sense. When you move from simple models to something that has real three-dimensional arrangement, the results can deviate significantly. So what are the actual numbers they found regarding these deviations?

Kai: They do show that in the effective medium regime, the effective Seebeck coefficient is almost independent of metal concentration, decaying by about twenty percent for higher concentrations. But then as you move into that percolation regime, zT starts following a more pronounced dependence on how much metal you have.

Title and authors: Mira: And they point out that zT actually stays virtually unchanged over the effective medium region before it starts decaying by about forty percent over the range of concentration between zero point eighteen and zero point three when a significant fraction of the current flows through extended metallic clusters.

Lev: So, if we're building a device, this tells us that you need to be careful around that concentration range where you start seeing those large fluctuations in properties.

Kai: Right. And they do fix their percolation threshold at pc equals eighteen point three percent, which they derived analytically for randomly packed spheres. But they admit that because of the non-zero background conductivities, the critical exponents they found deviate from what you'd expect for simple three dimensional materials.

Mira: They get t equals zero point nine zero and s equals one for electric conductivity, and t equals zero point five four and s equals zero point seven for the thermal counterpart. That deviation is a big caveat because it means we can't just plug in some universal percolation theory formulas directly.

Lev: So when you're trying to run this on real hardware, you’re not just applying a standard model; you have to account for these specific scaling behaviors they derived from their simulations. That adds complexity to the validation step.

Kai: Exactly, and that brings us into the future work. They suggest looking at how the thermal conductivity of the host material decays, referencing some other research they cited back in their work, which points toward optimizing those decay rates for better performance.

Mira: So what's our final word on this paper? It shows that while we can't rely on simple effective medium predictions for these complex composites, the numerical analysis of realistic three-dimensional composites gives us a much better handle on how concentration and arrangement interact.

Lev: For me, it’s about confirming that when you do the work rigorously, you get a clearer picture of where the performance gains actually come from, which is crucial when dealing with real-world constraints in error correction.

Kai: So that's what this paper on thermoelectric properties of metal-semiconductor percolation composites shows us. It gives us a framework for designing these things better by focusing on the interplay between concentration, arrangement, and that critical transition point near eighteen point three percent.

The paper's summary: Kai: So, to recap this paper, they're looking at how mixing a semiconductor host like BiSbTe with random metal inclusions like copper affects its ability to make electricity when there's a temperature difference between two sides.

Mira: Right. The main point is that these properties are really sensitive to the percolation transition, which is basically when those metal pieces start linking up to create a continuous path through the material.

Lev: So, what they found near that threshold is that both the power factor and this figure of merit zT depend heavily on how concentrated you make the metal and how it's arranged.

Kai: They really pinpoint some specific setups where you can get better performance by controlling those arrangements in real composites.

Mira: And they also suggest that changing the shape of the inclusions matters because if they're elongated, they can shift that percolation threshold to a lower concentration value, which is good for design.

Lev: That means if we want a certain performance metric, like a high power factor, we might need to use different shapes or arrangements of those metal pieces instead of just sticking to spheres.

Kai: Exactly. It’s not just about the volume fraction anymore; it’s about geometry too when you’re looking at realistic materials.

Mira: And they also point out that the way these composite materials behave in three dimensions doesn't follow simple effective medium theories or layered media predictions, which is a big deal because those simpler models just don't capture the complexity.

Lev: That deviation is what we need to worry about when we try to validate any of this on real hardware; you can't just use a textbook formula for these complex arrangements.

Kai: So, they show us that while the effective Seebeck coefficient stays pretty stable in the middle range, zT starts dropping off significantly once you go past that critical concentration area near eighteen point three percent.

Mira: That decay happens because of those extended metallic clusters flowing with the current, which is a physical effect we have to account for if we want accurate predictions.

Lev: It’s telling us that for real-world error correction systems, you need to be extra careful in that concentration range where you see those big fluctuations in performance.

Kai: And they do mention looking at how the thermal conductivity of the host material decays as another way to potentially boost performance further down the line.

Mira: So, this paper gives us a framework showing that we have to consider both the concentration and arrangement when designing these metal-semiconductor composites for better thermoelectric performance.

Lev: It sets up a clear direction for how we might design next-generation materials based on these specific scaling behaviors they found near that percolation point.

Kai: Next up, we’re going to look at how those specific scaling laws they found affect the actual numbers when you try to build a device, and what that means for your hardware fidelity.

The paper's improvements: Tom: So, Kai and Mira are talking about what they think should be done next based on this paper's findings for these metal-semiconductor composites.

Kai: They’re suggesting that we need to look at how the thermal conductivity of the host material decays because that could help enhance performance even more.

Mira: Right, it’s about controlling those thermal properties in the host material since they found that this decay plays a role in getting better results.

Lev: From my side, I think if we can control those decay rates precisely, it could translate into better fidelity when we are trying to run these models on real hardware.

Kai: It’s about moving beyond just the concentration and arrangement and really tuning the material structure itself for better results.

Mira: And they emphasize that the shape of the metal inclusions is a key design variable because it directly influences how fast you hit that percolation threshold.

Lev: So, if we want to design something with a specific power factor, we might actually need to engineer those inclusions into specific shapes rather than just using simple spheres.

Kai: Exactly. It’s about having more control over the geometry of the inclusions to get the best possible thermoelectric response out of them.

Mira: They are really pushing for more detailed studies that look at how those different structural choices interact with the concentration effects they found earlier.

Lev: It sounds like we need a way to map input variables, like shape and concentration, directly to expected performance metrics like the power factor and zT.

Kai: That’s what I mean; we should build an AI system that can take those parameters and predict the outcome before we even start building it.

Mira: And they are pointing toward creating a better decision-making framework for selecting materials based on those concentration and arrangement variables.

Lev: That would be useful for me because right now, I’m just running simulations based on fixed geometries; a predictive model that understands these dependencies would save us a lot of time.

Kai: It’s about using the knowledge gained from this study to design materials that are optimized for real-world application rather than just theoretical limits.

Mira: So, the goal is to move toward designing composites where we can precisely tune those structural parameters to get exactly the performance we want.

Lev: If we can do that, it gives us a path forward for creating more robust and efficient devices in error correction systems.

Conclusion: Kai: So to wrap up this paper on "Thermoelectric properties of metal-semiconductor percolation composites," they’re showing how understanding that critical transition point around eighteen point three percent is really important for designing these materials.

Mira: Exactly, it shows that you can’t just rely on simple effective medium theories when dealing with the actual three-dimensional arrangement and non-zero background conductivities of the host material.

Lev: For my work, it means we need to be careful because the scaling laws they found for electric and thermal conductivity deviate from what we expect in standard percolation theory.

Kai: It’s a warning that you can't just plug in universal formulas; you have to account for those specific exponents they derived, like t equals zero point nine zero and s equals one.

Mira: And the results show that while the effective Seebeck coefficient stays fairly stable in the middle range, it starts dropping off sharply once you get past that concentration threshold because of how those metallic clusters behave.

Lev: That decay happens because of those extended metallic clusters flowing with the current, which is something we have to factor in if we’re trying to run these things on real hardware where currents are flowing through the whole structure.

Kai: So, this paper gives us a better picture of how concentration and arrangement interact when you look at realistic composites.

Mira: It’s about moving away from just looking at volume fraction and starting to consider the actual geometry of the inclusions for optimizing performance metrics like zT.

Lev: It changes things because it tells us that predicting performance requires accounting for those structural details, not just a bulk average number.

Oleksandr Porubaniyi, Sergii Tomko, Mihailo Cherkez, Viktor Podolskiy, Radion Cherkez

Yuriy Fedkovych Chernivtsi National University · University of Massachusetts Lowell

physics.app-ph, cond-mat.mtrl-sci

Submitted: 2026-09-04

Updated: 2026-11-17

Code: https://github.com/viktor-podolskiy/thermoelectric-percolation

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 73/100

The gist: Thermoelectric properties of metal-semiconductor percolation composites analyze the thermoelectric response of composite materials comprising homogeneous semiconductor hosts with randomly distributed…

Key concepts

Percolation Transition
This is a phase change where a network of conductive inclusions suddenly connects, allowing electricity to flow through the entire material. In this study, it occurs when metal concentration reaches a critical point (pc), leading to dramatic changes in conductivity and material behavior.
Power Factor (Pf)
The power factor measures how effectively a thermoelectric material converts a temperature difference into electrical power. It is calculated as the square of the Seebeck coefficient multiplied by the electrical conductivity, indicating its potential for generating electricity.
Figure of Merit (zT)
This dimensionless number represents the overall efficiency of a thermoelectric material. It balances electrical performance (power factor) against thermal conductivity and is crucial for determining how useful a material is in real-world energy conversion devices.
Effective Medium Response
This describes how the bulk properties of a composite material are predicted based on average properties, assuming random distribution. The paper shows that this simple model fails near the percolation threshold because local arrangements significantly alter the outcome.

Terminology

Summary

Thermoelectric properties of metal-semiconductor percolation composites analyze the thermoelectric response of composite materials comprising homogeneous semiconductor hosts with randomly distributed metallic inclusions and demonstrate that their properties are strongly affected by the percolation transition <ref:2609. p. 005273#pg2> We show that in vicinity of percolation threshold, the power factor and the thermoelectric figure of merit strongly depend on both the concentration and arrangement of inclusions <ref:2609. p. 0 p. 005273#pg2> Finally, we identify promising regimes for enhancing the power factor in realistic composites <ref:2609. p. 0 p. 005273#pg2>.

Material System and Modeling

The study considers a bulk semiconductor host, specifically BiSbTe, with metallic Cu inclusions modeled as spherical inclusions <ref:2609.05273#pg2> The geometry involves a region of the composite of thickness l = 3mm with a unit cell area of A = 10−6m2 <ref:2609. p. 4 p. 105273#pg2>. Periodic boundary conditions are applied along the in-plane directions to simulate response of the spatially-infinite material <ref:2609. p. 105273#pg2>. The analysis uses a commercial finite-element partial differential equation (PDE) solver COMSOL Multiphysics <ref:2609. p. 105273#pg2>. Three separate simulations are performed, with the first two neglecting the weak thermoelectric coupling to calculate effective electrical and thermal conductivities <ref:2609. p. 105273#pg2>. The last simulation provides a self-consistent solution of thermoelectric generation at a fixed temperature difference ΔT = 10oC between the top and bottom surfaces, with baseline temperature of the composite being 0oC <ref:2609. p. 205273#pg2>.

Conductivity Regimes and Transitions

The DC properties of a random mixture with metallic inclusions exhibit a dielectric/metal phase transition when the volume fraction of the metal is of the order of pc ≃ 15% <ref:2609. p. 405273#pg2>. In the vicinity of this phase transition, local distribution of electric field, as well as distribution of electric currents are extremely inhomogeneous and the averaged conductivity follows scaling behavior (sigma −sigmab) ∝ (p − पc)t for p > पc and (sigma −sigmab) ∝ (पc − प−s for p. p. 1 The critical exponents t, s depend on the dimensionality of the network, and – in case of the network with non-zero background conductivity sigmab – on sigmab <ref:2609. p. 105273#pg2>. The authors fix the percolation threshold to pc ≃ 18. p. 33%, derived analytically for randomly packed spheres <ref:2609. p. 3 p. 305273#pg2>.

Thermoelectric Performance Metrics

The performance of the composite is evaluated using two standard thermoelectricity metrics:

  1. Power factor Pf = S2sigma that characterizes the electrical power output of the device at a given temperature gradient <ref:2609. p. 205273#pg2>.

  2. The dimensionless figure of merit zT = S2sigmaT/λ that characterizes the overall thermoelectric efficiency <ref:2609. p. 205273#pg2>.

Discussion and Optimization

The analysis reveals that metal cluster formation within the composite makes the materials very sensitive to the presence of metal <ref:2609.05273#pg2>. The best-possible performance is expected to be close to percolation threshold, where the composite yields doubling of the power factor with virtually no reduction of zT, as compared with the bare semiconductor <ref:2609. p. 405273#pg2>. The performance can be further enhanced if thermal conductivity of the host decays as suggested in Ref. p. 4[12] <ref:2609. p. 405273#pg2>. Inclusion shape plays a crucial role in defining the percolation threshold with more elongated inclusions leading to the smaller values of pc <ref:2609. p. 405273#pg2>. Spherical inclusions are close to optimal for composites with random inclusions placement and orientation <ref:2609. p. 405273#pg2>.

Conclusion

The quantitative response of realistic three-dimensional materials significantly deviates from predictions of effective medium response and from the properties of the layered media <ref:2609.05273#pg2>. The study provides an outlook for a new generation of thermoelectric composite materials <ref:2609. p. 405273#pg2>.

Author Declarations

The authors have no conflicts to disclose <ref:2609. p. 5 p. 505273#pg2>.

Data Availability Statement

All relevant data are provided in this work <ref:2609.05273#pg2>. The FEM model developed in this study is available online [29] <ref:2609. p. 5 p. 5 p. 505273#pg2>. Other data that supports the findings of this study is available from the corresponding author upon reasonable request <ref:2609. p. 5 p. 5 p. 505273#pg2>.

References

The references listed are 1 through 14 <ref:2609.05273#pg2>. The paper cites various works on thermoelectric materials and percolation theory <ref:2609. p. 005273#pg2>. The authors acknowledge the limitations of previous analytical and numerical studies regarding complex media <ref:2609. p. 105273#pg2>. The study builds upon established concepts such as effective medium description and classic percolation problems <ref:2609. p. 205273#pg2>.

--- Page 1 ---

Thermoelectric properties of metal-semiconductor percolation composites Oleksandr Porubaniyi, Sergii Tomko, Mihailo Cherkez, Viktor Podolskiy, and Radion Cherkez <ref:2609.05273#pg2>. Department of Thermoelectricity and Medical Physics, Yuriy Fedkovych Chernivtsi National University, 2 Kotsiubynsky str. p. 0 p. 0, 58012, Ukraine <ref:2609. p. 0 p. 005273#pg2>. Department of Physics and Applied Physics, University of Massachusetts Lowell, Lowell, MA, 01854, USA <ref:2609. p. 0 p. 005273#pg2>. corresponding author: viktor podolskiy@uml. p. 0edu <ref:2609. p. 005273#pg2>. We analyze thermoelectric response of composite materials comprising homogeneous semiconductor hosts with randomly distributed metallic inclusions <ref:2609. p. 005273#pg2>. We demonstrate that the properties of such materials are strongly affected by the percolation transition (with background conductivity) <ref:2609. p. 005273#pg2>. We show that in vicinity of percolation threshold, the power factor and the thermoelectric figure of merit strongly depend on both the concentration and arrangement of inclusions <ref:2609.05273#pg2>. Finally, we identify promising regimes for enhancing the power factor in realistic composites <ref:2609.05273#pg2>.

--- Page 2 ---

Fig.1. Schematic of the thermoelectric composite analyzed in this work (10 spherical inclusions are shown) <ref:2609. p. 005273#pg2>. Cu contacts <ref:2609. p. 005273#pg2>. Cu inclusions <ref:2609. p. 405273#pg2>. BiSbTe host <ref:2609. p. 405273#pg2>.

--- Page 3 ---

Thermoelectricity represents the interplay between the electric current density j⃗ and heat fluxes Jq⃗ in the system <ref:2609.05273#pg2>. Explicitly, j⃗= −sigma (∇⃗ϕ + Svec∇T) and Jq⃗ = TS j− λ ∇T with sigma and λ being electrical and thermal conductivities, and S being Seebeck coefficient that describes the thermoelectric coupling <ref:2609. p. 105273#pg2>. Here we use the commercial finite-element partial differential equation (PDE) solver COMSOL Multiphysics <ref:2609. p. 105273#pg2>.

--- Page 4 ---

Fig.2 illustrate the dependence of electrical and thermal conductivities of individual composites on concentration of metal (Cu) inclusions – the only parameter that controls the response of composites in the effective medium regime <ref:2609. p. 205273#pg2>. Dots in Fig. p. 22 illustrate the behavior of individual realizations of random composites <ref:2609. p. 205273#pg2>. Note that both electrical and thermal conductivities follow similar behavior <ref:2609. p. 205273#pg2>. Three regions can be clearly identified <ref:2609. p. 205273#pg2>. In the second region, when concentration changes from 18 to 27%, conductivities rapidly grow, changing by almost an order of magnitude <ref:2609. p. 205273#pg2>. Notably, this growth of average conductivity is accompanied by strong variation between conductivities of individual realizations of random materials <ref:2609. p. 205273#pg2>. This region is dominated by the onset of percolation phase transition and is not described by effective medium response since the parameters of the structure depend on arrangement of the inclusions <ref:2609. p. 205273#pg2>. Finally, when metal concentration exceeds 27%, conductivities level out and the differences resulting from multiple realizations of the networks with identical metal concentration become smaller <ref:2609. p. 205273#pg2>.

--- Page 5 ---

To systematically analyze the properties of the random composites we sort the data by the volumetric concentration of metal and group the composites into bins that have concentration “width” Δp = 1.5% <ref:2609.05273#pg2>. We then calculate the average value and the standard deviation of each quantity of interest within individual bins <ref:2609. p. 305273#pg2>. The dependence of mean values of conductivities on the average metal concentration within the composite is shown as solid lines in Fig. p. 32 <ref:2609.05273#pg2>. In agreement with the above discussion, the properties of composites in the effective medium regime are almost independent of the exact arrangement of inclusions (the variance of the resulting parameters is very small) <ref:2609. p. 305273#pg2>. In percolation regime the variance of the parameters increases dramatically <ref:2609. p. 305273#pg2>. Finally, in the high-metal-concentration regime the fluctuations fall once again as the composite converges to a highly conducting material <ref:2609. p. 305273#pg2>.

--- Page 6 ---

Since, in contrast to classic percolation problem, our background material conducts both electricity and thermal flux, conductivity of our material does not exactly map onto classic percolation theory <ref:2609.05273#pg2>. To avoid introduction of these extra fitting parameters, we fix the percolation threshold to pc ≃ 18. p. 33%, derived analytically for randomly packed spheres <ref:2609.05273#pg2>. We fix background conductivities to the values of bare BiSbTe <ref:2609. p. 305273#pg2>. Fig.3 illustrates the scaling behavior of such reduced conductivities in the vicinity of pc, illustrating the expected power-law dependence <ref:2609. p. 305273#pg2>. However, due to the non-zero background conductivities, our critical exponents deviate from the ones expected for 3D materials <ref:2609. p. 305273#pg2>. We obtain: t ≃ 0. p. 390, s ≃ 1 for electric conductivity and t ≃ 0. p. 154, s ≃ 0.7 for its thermal counterpart <ref:2609. p. 305273#pg2>.

--- Page 7 ---

Fig.4 illustrates our analysis of thermoelectric performance of the composites in our studies <ref:2609. p. 305273#pg2>. The effective Seebeck coefficient is almost independent of metal concentration for effective medium regime, decaying by ∼ 20% for higher concentrations <ref:2609. p. 305273#pg2>. The power factor, dominated in our case by the electric conductivity, grows with metal concentration, increasing by a factor of two over the effective medium region, with an extra factor of two from percolation threshold until p ∼ 30% <ref:2609. p. 305273#pg2>. Finally, the thermoelectric figure of merit zT largely follows the behavior of Seeback coefficient, but with more pronounced dependence on metal concentration <ref:2609. p. 305273#pg2>. zT is virtually unchanged over the effective medium region, followed by a decay by ∼ 40% over the range p ∈ [0. p. 318 … 0.3] when significant fraction of the current flows over extended metallic clusters <ref:2609. p. 305273#pg2>.

--- Page 8 ---

Fig.4 Dependence of the effective Seebeck coefficient S (a), power factor Pf (b) and thermoelectric figure of merit zT (c) composites on the volume fraction of Cu p; each orange dot represents a single composite <ref:2609. p. 405273#pg2>.

--- Page 9 ---

The paper is 2609.05273 <ref:2609. p. 4 p. 4 p. 405273#pg2>. This paper is 2609.05273 <ref:2609.05273#pg2>. The authors have no conflicts to disclose <ref:2609.05273#pg2>. Data Availability Statement All relevant data are provided in this work <ref:2609. p. 5 p. 5 p. 505273#pg2>. The FEM model developed in this study is available online [29] <ref:2609.05273#pg2>. Other data that supports the findings of this study is available from the corresponding author upon reasonable request <ref:2609.05273#pg2>.

--- Page 1 ---

Thermoelectric properties of metal-semiconductor percolation composites Oleksandr Porubaniyi, Sergii Tomko, Mihailo Cherkez, Viktor Podolskiy, and Radion Cherkez <ref:2609.05273#pg2>. Department of Thermoelectricity and Medical Physics, Yuriy Fedkovych Chernivtsi National University, 2 Kotsiubynsky str., 58012, Ukraine <ref:2609.05273#pg2>. Department of Physics and Applied Physics, University of Massachusetts Lowell, Lowell, MA, 01854, USA <ref:2609.05273#pg2>.

--- Page 9 ---

The paper is 2609.05273 <ref:2609.05273#pg2>. The authors have no conflicts to disclose <ref:2609.05273#pg2>. Data Availability Statement All relevant data are provided in this work <ref:2609.05273#pg2>. The FEM model developed in this study is available online [29] <ref:2609.05273#pg2>. Other data that supports the findings of this study is available from the corresponding author upon reasonable request <ref:2609.05273#pg2>.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improve predictive modeling for complex thermoelectric composites by incorporating background conductivity effects into percolation theory scaling laws. The current study notes that conductivity of our material does not exactly map onto classic percolation theory due to non-zero background conductivities, suggesting AI models should be trained on these specific scaling behaviors rather than universal classical predictions.

  2. Enhance the design of high-performance thermoelectric materials by optimizing inclusion shape and distribution for specific performance metrics. The conclusion states that Inclusion shape plays a crucial role in defining the percolation threshold with more elongated inclusions leading to the smaller values of pc, which an AI system could use to generate optimal geometric parameters for desired power factors.

  3. Develop a decision-making framework for composite material selection based on concentration and arrangement variables. The paper demonstrates that the properties of the composite strongly depends on the metal concentration and for a range of metal concentrations, the properties of the composite strongly depend on the arrangement of inclusions, allowing an AI to map input parameters (concentration, shape) to expected performance metrics like power factor and figure of merit zT.

  4. Refine simulation accuracy for realistic three-dimensional composites by integrating self-consistent thermoelectric coupling into numerical solvers. The method uses COMSOL Multiphysics for a self-consistent solution of thermoelectric generation, which an AI could leverage to create more robust numerical models that account for the interplay between electric current density and heat fluxes.

Related papers