High-Fidelity In Silico Dataset of Ordered Binary and Ternary MXenes with Predictive Synthesis Guidance

arXiv:2611.14687 · cond-mat.mtrl-sci · Submitted 2026-11-10 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.

Kai: We're looking at the paper: "High-Fidelity In Silico Dataset of Ordered Binary and Ternary MXenes with Predictive Synthesis Guidance".

Mira: Here's the gist of the paper. High-Fidelity In Silico Dataset of Ordered Binary and Ternary MXenes with Predictive Synthesis Guidance provides a robust foundation for developing machine-learned stability descriptors and for guiding experimental eWorts toward novel MXene discovery The dataset provides a robust foundation for developing machine-learned stability descriptors and for guiding experimental eWorts toward novel MXene discovery.

Kai: Dataset Scope and Methodology The study utilized density functional theory (DFT) with the r2SCAN functional to develop a high-fidelity dataset of 1,040 ordered binary and ternary bare delaminated MXenes, including mono-transition metal MXenes, in-plane ordered double-transition metal MXenes, in-plane ordered metal-divacancy MXenes, and out-of-plane ordered double-transitionmetal MXenes.

Mira: The methodology involved retrieving MAX phases from the Materials Project and then generating initial MXene structures by removing Al atoms and using topotactic structure templating.

Kai: Combinatorial permutations of these M and X elements yielded 78 m-MXenes, 312 i-MXenes, 624 o-MXenes, and 26 v-MXenes.

Title and authors: Kai: So, the authors are Ahmed H. Biby and Charles B. Musgrave from the University of Colorado Boulder and the University of Utah >

Mira: They’ve put this whole dataset into an open-access online repository, which is a big deal because it means anyone can use it to train their own models >

Lev: It's about making predictions for novel MXene discovery by giving the AI some real data to learn from, which is how we try to speed up the experimental process >

Kai: The title itself suggests they are providing guidance for synthesis, not just a list of structures, but telling us *how* to make them stable >

Mira: That's right. They are trying to move away from just guessing compositions and start using calculated stability data to guide those wet-lab experiments >

The paper's summary: Kai: The summary explains that they used density functional theory with the r-squared scan functional to generate this high fidelity dataset of one thousand forty ordered binary and ternary bare delaminated MXenes >

Mira: It goes into detail about the structural spaces they explored, including mono transition metal MXenes, in-plane ordered double transition metal MXenes, and out of plane ordered double transition metal MXenes >

Lev: They mention that they used combinatorial permutations of M and X elements to generate seven eighty m-MXenes, three hundred twelve i-MXenes, six twenty four o-MXenes, and two sixty four v-MXenes >

Kai: So it’s a very comprehensive structural exploration across all those different ordering types >

Mira: And the main finding they highlight is that all these MXenes are thermodynamically inaccessible under ambient equilibrium conditions because they lie above the convex hull under ambient equilibrium conditions >

The paper's improvements: Kai: The authors suggest a few ways to improve this dataset and its use. They focus on how stability can be assessed not just at room temperature and pressure, but by looking at non equilibrium conditions >

Mira: They show that eleven fifty predominantly carbide MXenes become thermodynamically stabilized when the temperature is elevated between five hundred Kelvin and two thousand Kelvin >

Lev: And they also found twenty three other MXenes are stabilized under high pressure of ten gigapascals, which gives us a concrete stability window to look for experimentally >

Kai: Plus, they identified three MXenes in the manganese iron carbon chemical space that are stabilized within an experimentally feasible range of carbon chemical potentials between negative six and one electron per carbon atom >

Conclusion: Mira: So, the paper concludes that this dataset lets us identify potentially undiscovered MXenes through a predictive synthesis approach and provides a foundation for training machine-learned stability descriptors >

Kai: It’s about using this data to map out structure composition stability trends by analyzing each studied MXene against its stability metrics >

Lev: And the limitations they point out is that they only studied bare delaminated MXenes, and they need experimental validation of these predicted stable MXenes to confirm their findings >

Mira: Overall, this work is about moving from just having structural information to having a predictive tool that can actually guide how we synthesize new materials >

Ahmed H. Biby, Charles B. Musgrave

University of Colorado Boulder · Department of Chemical and Biological Engineering, University of Colorado Boulder · Department of Chemical Engineering, University of Utah · Department of Materials Science and Engineering, University of Utah

cond-mat.mtrl-sci

Submitted: 2026-11-10

Updated: 2026-11-10

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

Importance score: 92/100

The gist: High-Fidelity In Silico Dataset of Ordered Binary and Ternary MXenes with Predictive Synthesis Guidance provides a robust foundation for developing machine-learned stability descriptors and for…

Key concepts

MXenes
MXenes are a class of 2D materials composed of transition metal carbides, nitrides, or carbonitrides. They are formed by removing aluminum atoms from MAX phases and exhibit unique electronic and structural properties that make them promising for various applications.
Thermodynamic Stability
This refers to whether a material can exist naturally at equilibrium under ambient conditions. The study found that most MXenes lie above the 'convex hull,' meaning they are thermodynamically inaccessible unless external factors like high heat or pressure are applied.
Stabilization Mechanisms
The paper explored three ways to make unstable MXenes stable: thermal stabilization (high temperature), pressure stabilization (high force), and chemical potential stabilization. These mechanisms help predict how to synthesize MXenes under non-equilibrium conditions.
Machine-Learned Descriptors
These are mathematical tools trained on the DFT data of 1,040 MXenes to predict their stability. By using these descriptors, AI can identify which new MXene structures are likely to be stable and suggest specific synthesis strategies.

Terminology

Summary

High-Fidelity In Silico Dataset of Ordered Binary and Ternary MXenes with Predictive Synthesis Guidance provides a robust foundation for developing machine-learned stability descriptors and for guiding experimental eWorts toward novel MXene discovery The dataset provides a robust foundation for developing machine-learned stability descriptors and for guiding experimental eWorts toward novel MXene discovery. p. 0

Dataset Scope and Methodology

The study utilized density functional theory (DFT) with the r2SCAN functional to develop a high-fidelity dataset of 1,040 ordered binary and ternary bare delaminated MXenes, including mono-transition metal MXenes, in-plane ordered double-transition metal MXenes, in-plane ordered metal-divacancy MXenes, and out-of-plane ordered double-transitionmetal MXenes. The methodology involved retrieving MAX phases from the Materials Project and then generating initial MXene structures by removing Al atoms and using topotactic structure templating. p. 2 Combinatorial permutations of these M and X elements yielded 78 m-MXenes, 312 i-MXenes, 624 o-MXenes, and 26 v-MXenes. p. 2 The computational details involved Meta-GGA spin-polarized DFT calculations using the Vienna Ab initio Simulation Package (VASP) with the r2SCAN exchange-correlation functional. p. 3

Thermodynamic Stability Findings

The analysis revealed that all MXenes are thermodynamically inaccessible under equilibrium conditions, as they lie above the convex hull under ambient equilibrium conditions. However, the study identified several synthesis strategies by assessing stability under non-equilibrium conditions. p. 9 Specifically, 115 predominantly carbide MXenes become thermodynamically stabilized at elevated temperatures (500–2000 K), and 23 are stabilized under high pressure (10 GPa). p. 0 Additionally, three MXenes in the Mn–Fe–C chemical space are stabilized within an experimentally feasible range of carbon chemical potentials. p. 0

Stabilization Mechanisms and Trends

The stabilization of MXenes was explored through three avenues: (a) elevated temperature (t-stabilization), (b) applied pressure (p-stabilization), and (c) modified chemical potential of X (µX-stabilization). The t-stabilized MXenes are primarily m- and oMXenes with only one i-MXene, i1-YScC. p. 5 Pressure stabilizes MXenes whose per-atom volume is sufficiently small relative to competing phases such that the VDp contributions lower their H(p) relative to competing phases. p. 6 The Mn–Fe–C chemical space was the only one in which stabilized MXenes (with n = 3) were predicted by adjusting the chemical potential of carbon within an experimentally feasible range of –6 to 1 eV/C atom. p. 9

Data Utility and Future Directions

This dataset enables the identification of potentially undiscovered MXenes via a predictive synthesis approach, supports the development of structure–composition–stability trends through analyzing each studied MXene with respect to its stability, and provides a foundation for training machine-learned MXene stability descriptors. The dataset is openly available in Zenodo under a CC BY 4. p. 90 license. As an outlook, this dataset can be expanded to include disordered and terminated MXenes, which significantly influence stability and properties. p. 9 The experimental validation of the predicted stable MXenes is also necessary to confirm these findings and guide synthesis. p. 9

Data Records

The dataset is openly available in Zenodo under a CC BY 4.0 license. It contains a table of all 1,040 MXenes in CSV and JSON formats. p. 9 The relaxed structure of each MXene is stored in a folder named by this identifier. p. 9 p. 9 p. 9 p. 9 p. 9 For each entry, the table lists ΔGf (eV/atom), ΔGd (eV/atom), Ehull (eV/atom), the decomposition products, the band gap (eV), and the layer group. p. 9 p. 9 p. 9 p. 9 p. 9 The minimum stabilizing temperatures and pressures of the t- and p-stabilized MXenes are listed in Table S3 of the Supporting Information. p. 6 p. 6 p. 6 p. 6 p. 6 The dataset is available in Zenodo at https://doi. p. 9 p. 9 p. 9 p. 9org/zenodo. p. 1 p. 1 p. 1 p. 124116692. The dataset contains a table of all 1,040 MXenes in CSV and JSON formats. p. 9 p. 9 p. 9 p. 9 The DFT-optimized structure of each MXene as a VASPCONTCAR file. p. 9 p. 9 p. 9 p. 9 A README describing each column. p. 9 p. 9 p. 9 p. 9 SHA-256 checksums for all files. p. 9 p. 9 p. 9 p. 9 Each of the 1,040 entries has a unique identifier consisting of two parts separated by a dash. p. 9 p. 9 p. 9 p. 9 The first part is a letter indicating the MXene category, as described in the Introduction. p. 2 p. 2 p. 2 p. 2 Followed by the value of n in the corresponding category formula. p. 9 p. 9 p. 9 p. 9 For binary (m- and v-) MXenes, the second part gives the M and X elements, while for ternary (o- and i-) MXenes it gives the M, M′, and X elements, respectively. p. 9 p. 9 p. 9 p. 9 For o-MXenes, M denotes the inner metal layer(s) and M′ the outer metal layers. p. 9 p. 9 p. 9 p. 9 For i-MXenes, M occupies two-thirds of the metal sites and M′ the remaining one-third. p. 2 p. 2 p. 2 p. 2 For example, the o-MXene with M = Sc, M′ = Y, X = C and n = 2 is “o2-ScYC”. p. 9 p. 9 p. 9 p. 9 The relaxed structure of each MXene is stored in a folder named by this identifier. For each entry, the table lists ΔGf (eV/atom), ΔGd (eV/atom), Ehull (eV/atom), the decomposition products, the band gap (eV), and the layer group. Ehull, ΔGf, and ΔGd are the energy above the hull and the formation and decomposition Gibbs energies of the DFT-optimized structures, computed relative to all competing phases, the elemental references, and all competing phases, respectively. p. 1 p. 1 p. 1 p. 1 The decomposition products are the competing phases that dominate under equilibrium conditions and inhibit formation of the corresponding MXene. p. 1 p. 1 p. 1 p. 1 The minimum stabilizing temperatures and pressures of the t- and p-stabilized MXenes are listed in Table S3 of the Supporting Information. The dataset is available in Zenodo at https://doi.org/zenodo.24116692. The dataset contains a table of all 1,040 MXenes in CSV and JSON formats. The DFT-optimized structure of each MXene as a VASPCONTCAR file. A README describing each column. SHA-256 checksums for all files. Each of the 1,040 entries has a unique identifier consisting of two parts separated by a dash. The first part is a letter indicating the MXene category, as described in the Introduction. Followed by the value of n in the corresponding category formula. For binary (m- and v-) MXenes, the second part gives the M and X elements, while for ternary (o- and i-) MXenes it gives the M, M′, and X elements, respectively. For o-MXenes, M denotes the inner metal layer(s) and M′ the outer metal layers. For i-MXenes, M occupies two-thirds of the metal sites and M′ the remaining one-third. For example, the o-MXene with M = Sc, M′ = Y, X = C and n = 2 is “o2-ScYC”. The relaxed structure of each MXene is stored in a folder named by this identifier. For each entry, the table lists ΔGf (eV/atom), ΔGd (eV/atom), Ehull (eV/atom), the decomposition products, the band gap (eV), and the layer group. Ehull, ΔGf, and ΔGd are the energy above the hull and the formation and decomposition Gibbs energies of the DFT-optimized structures, computed relative to all competing phases, the elemental references, and all competing phases, respectively. The decomposition products are the competing phases that dominate under equilibrium conditions and inhibit formation of the corresponding MXene. The minimum stabilizing temperatures and pressures of the t- and p-stabilized MXenes are listed in Table S3 of the Supporting Information. The dataset is available in Zenodo at https://doi.org/zenodo.24116692. The dataset contains a table of all 1,040 MXenes in CSV and JSON formats. The DFT-optimized structure of each MXene as a VASPCONTCAR file. A README describing each column. SHA-256 checksums for all files. Each of the 1,040 entries has a unique identifier consisting of two parts separated by a dash. The first part is a letter indicating the MXene category, as described in the Introduction. Followed by the value of n in the corresponding category formula. For binary (m- and v-) MXenes, the second part gives the M and X elements, while for ternary (o- and i-) MXenes it gives the M, M′, and X elements, respectively. For o-MXenes, M denotes the inner metal layer(s) and M′ the outer metal layers. For i-MXenes, M occupies two-thirds of the metal sites and M′ the remaining one-third. For example, the o-MXene with M = Sc, M′ = Y, X = C and n = 2 is “o2-ScYC”. The relaxed structure of each MXene is stored in a folder named by this identifier. For each entry, the table lists ΔGf (eV/atom), ΔGd (eV/atom), Ehull (eV/atom), the decomposition products, the band gap (eV), and the layer group. Ehull, ΔGf, and ΔGd are the energy above the hull and the formation and decomposition Gibbs energies of the DFT-optimized structures, computed relative to all competing phases, the elemental references, and all competing phases, respectively. The decomposition products are the competing phases that dominate under equilibrium conditions and inhibit formation of the corresponding MXene. The minimum stabilizing temperatures and pressures of the t- and p-stabilized MXenes are listed in Table S3 of the Supporting Information. The dataset is available in Zenodo at https://doi.org/zenodo.24116692. The dataset contains a table of all 1,040 MXenes in CSV and JSON formats. The DFT-optimized structure of each MXene as a VASPCONTCAR file. A README describing each column. SHA-256 checksums for all files. Each of the 1,040 entries has a unique identifier consisting of two parts separated by a dash. The first part is a letter indicating the MXene category, as described in the Introduction. Followed by the value of n in the corresponding category formula. For binary (m- and v-) MXenes, the second part gives the M and X elements, while for ternary (o- and i-) MXenes it gives the M, M′, and X elements, respectively. For o-MXenes, M denotes the inner metal layer(s) and M′ the outer metal layers. For i-MXenes, M occupies two-thirds of the metal sites and M′ the remaining one-third. For example, the o-MXene with M = Sc, M′ = Y, X = C and n = 2 is “o2-ScYC”. The relaxed structure of each MXene is stored in a folder named by this identifier. For each entry, the table lists ΔGf (eV/atom), ΔGd (eV/atom), Ehull (eV/atom), the decomposition products, the band gap (eV), and the layer group. Ehull, ΔGf, and ΔGd are the energy above the hull and the formation and decomposition Gibbs energies of the DFT-optimized structures, computed relative to all competing phases, the elemental references, and all competing phases, respectively. The decomposition products are the competing phases that dominate under equilibrium conditions and inhibit formation of the corresponding MXene. The minimum stabilizing temperatures and pressures of the t- and p-stabilized MXenes are listed in Table S3 of the Supporting Information.

Benchmarking and Comparison

For benchmarking, the mean absolute errors (MAEs) were 0.15 eV/atom for formation energy and 0. p. 508 eV/atom for Ehull, which are reasonable MAEs given the differences in exchange-correlation functionals. p. 5 Neither our calculations nor those from the C2DB database yielded negative/zero Ehull values, confirming that the studied MXenes are thermodynamically metastable under equilibrium conditions. p. 5

Key Structural Observations

The study of temperature-dependent stability revealed that 115 MXenes become thermodynamically stable at elevated temperatures, where all competing phases in the diagram were treated consistently within GibbsComputedStructureEntry. Most of the 115 t-stabilized MXenes are carbides, as tabulated in Table S3. p. 5 The trend suggests that a significant fraction of metastable MXenes may be pulled onto the convex hull when temperature-dependent terms are included. p. 5

Compositional Insights

Figure 2(C) shows that Sc and Y are the most frequently present elements among the stabilizable mMXenes. This may be because Sc and Y are early transition metals (group 3), which are relatively electropositive, and often form stable carbide/nitride frameworks. p. 6 In Figure 2(D), Zr and Hf are the most common inner-layer metals in stabilizable MXene carbides and nitrides. p. 7 Group-4 metals (Zr/Hf) are classic carbide/nitride framework formers and tend to stabilize metal–X networks. p. 7 Cr and Mn are the most frequent outer-layer metals associated with stabilizable MXene carbides and nitrides. p. 7

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. Improved AI can be used to predict synthesis conditions for novel MXenes by utilizing a machine-learned stability descriptor trained on this high-fidelity dataset of 1,040 ordered binary and ternary bare delaminated MXenes.

  2. The system can guide experimental eWorts toward novel MXene discovery by identifying synthesis strategies that can guide experimental eWorts based on the calculated formation and decomposition Gibbs energies, layer symmetry groups, and energies above the hull.

  3. The improved AI can assess the thermodynamic accessibility of any proposed MXene structure under ambient conditions by determining if it lies above the convex hull under ambient equilibrium conditions using r2SCAN functional calculations.

  4. The system can predict non-equilibrium synthesis routes by assessing stabilization under (a) elevated temperature, (b) applied pressure, and (c) modified chemical potential of X, specifically identifying that 115 predominantly carbide MXenes become thermodynamically stabilized at elevated temperatures (500–2000 K), and 23 are stabilized under high pressure (10 GPa).

  5. The system can identify specific chemical space windows for stabilization, such as finding that three MXenes in the Mn–Fe–C chemical space are stabilized within an experimentally feasible range of carbon chemical potentials.

Related papers