Pith. sign in

REVIEW 3 major objections 5 minor 87 references

Machine learning interatomic potential for the low-modulus Ti-Nb-Zr alloys in the vicinity of dynamical instability

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read Simulation finds bone-like stiffness in Ti-Nb-Zr alloys.

desk verdict Solid MLIP study with an honest validation effort, but the headline low-modulus window is interpolated across an instability border and needs direct checking before it should be quoted. read the letter →

arxiv 2412.06270 v2 pith:L4LDZFUJ submitted 2024-12-09 cond-mat.mtrl-sci physics.comp-ph

classification cond-mat.mtrl-sciphysics.comp-ph
keywords machinelearninginteratomicpotentialmomenttensortitaniumalloyselasticmodulidynamicalinstabilityYoung'smodulusbiomedicalimplantselinvareffect
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper seeks to establish that a machine-learned interatomic potential can reliably simulate the finite-temperature elastic properties of multicomponent $\beta$-Ti$_{94-x}$Nb$_x$Zr$_6$ alloys even in the vicinity of dynamical instability. The authors train a moment tensor potential with an active-learning loop and show that it reproduces density-functional-theory elastic constants at 0 K across five Nb concentrations. Their central prediction is that near the boundary of dynamical and mechanical instability, the elastic moduli depend strongly nonlinearly on Nb content, producing a room-temperature Young's modulus of 30–40 GPa for alloys with 12–17 at.% Nb, matching human bone. They also predict elinvar behavior, meaning the moduli stay nearly constant over a wide temperature range, and strong anisotropy of directional Young's modulus that could enable texture-based design of implants. If correct, the work offers a practical computational route to designing low-stiffness biomedical titanium alloys without expensive ab initio molecular dynamics.

What carries the argument

The central machinery is a moment tensor potential, a machine-learned interatomic potential that expresses total energy, forces, and stresses as linear combinations of tensor-valued basis functions of local atomic environments. Training proceeds by active learning: configurations with high extrapolation grade are collected from molecular dynamics and re-evaluated with density functional theory, iteratively extending the dataset to include the distorted, low-Nb, low-temperature configurations that appear near instability. Elastic constants are then extracted from stress–strain relations in NVT molecular dynamics on supercells of 128,000 atoms, and polycrystalline moduli are obtained by Voigt–Reuss–Hill averaging. The key physical quantity is the tetragonal shear modulus $C' = (C_{11}-C_{12})/2$: as composition and temperature approach the boundary where $C'\to 0$, the material softens, the moduli become strongly nonlinear in composition, and directional anisotropy grows, producing the predicted low-modulus window.

What would settle it

Measure the room-temperature Young's modulus of single-phase β-Ti-Nb-Zr samples with 12–17 at.% Nb and well-characterized composition; if the modulus is clearly outside 30–40 GPa, the central prediction is wrong. A complementary check is to compute phonon dispersions for Ti$_{93}$Nb$_1$Zr$_6$ and Ti$_{88}$Nb$_6$Zr$_6$ near the predicted critical temperatures: the absence of imaginary modes below roughly 500 K and 390 K would contradict the claimed dynamical instability.

Watch

Extended reading notes

Core claim

The paper claims that a suitably trained moment tensor potential captures the temperature- and composition-dependent elastic response of $\beta$-Ti$_{94-x}$Nb$_x$Zr$_6$ solid solutions, including the weakly unstable low-Nb, low-temperature regime. On this basis the authors predict that Ti$_{93}$Nb$_1$Zr$_6$ and Ti$_{88}$Nb$_6$Zr$_6$ become dynamically and mechanically unstable below roughly 500 K and 390 K respectively, identified through anomalies in lattice parameters, mean-square displacements, and stress components. In the stable region, the computed Hill-averaged Young's modulus for 12–17 at.% Nb falls between 30 and 40 GPa at room temperature, comparable to human bone, and all mechanically stable compositions show elinvar-like weak temperature dependence of the elastic moduli. Near instability the alloys also exhibit strong elastic anisotropy, with the largest directional Young's modulus along [111] and the smallest along [100], which the authors connect to the same $C'$ softening that drives the low-modulus window.

Load-bearing premise

The predictions rest on the assumption that the machine-learned potential is accurate for the strongly distorted, low-symmetry atomic environments that appear in the unstable low-Nb alloys at 300–500 K, even though it was mostly trained on stable compositions and is validated by only 40 independent low-Nb configurations.

Editorial extensions

If this is right

  • If the prediction is correct, $\beta$-Ti$_{94-x}$Nb$_x$Zr$_6$ alloys with 12–17 at.% Nb give room-temperature Young's modulus of 30–40 GPa, matching the range needed for bone-implant compatibility.
  • The predicted elinvar behavior means that in mechanically stable compositions, elastic moduli remain nearly constant from 300 to 1300 K, so implants made from these alloys would resist stiffness changes with body temperature and processing heat.
  • The strong directional anisotropy, with maximum Young's modulus along [111] and minimum along [100], implies that controlling crystallographic texture during thermomechanical processing can tune the macroscopic stiffness without changing composition.
  • The identified instability boundaries, Ti$_{93}$Nb$_1$Zr$_6$ below about 500 K and Ti$_{88}$Nb$_6$Zr$_6$ below about 390 K, define a composition-temperature map that experimental synthesis could target to obtain metastable bcc alloys.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • This suggests the active-learning approach could be transferred to other metastable bcc alloys such as Ti-Mo or Ti-V, where the same $C'$-driven softening mechanism should produce analogous low-modulus windows.
  • If the 30–40 GPa window is confirmed experimentally, it would indicate that composition alone, without cold work or aging, can bring a $\beta$-titanium alloy to bone-like stiffness, which would simplify implant manufacturing.
  • The instability boundaries inferred from mean-square displacement and stress anomalies could be sharpened by direct anharmonic phonon calculations; a mismatch there would not overturn the modulus prediction but would refine the stability map.
  • The predicted anisotropy suggests that polycrystalline samples with random texture should exhibit a wider modulus spread than strongly textured ones; comparing textured and untextured samples of the same composition would test the texturing design idea.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper trains a moment tensor potential (MTP) for β-Ti94−xNbxZr6 alloys with x = 1, 6, 15, 22, and 30 at.% Nb using DFT data and an active-learning scheme, then uses MTP-MD (LAMMPS) to compute finite-temperature elastic constants and polycrystalline moduli over 300–1300 K. The central claims are: (i) an elinvar effect over a wide temperature range; (ii) a strongly nonlinear concentration dependence of elastic moduli near dynamical/mechanical instability that yields a predicted 30–40 GPa Young's modulus for compositions with 12–17 at.% Nb at room temperature; and (iii) strong anisotropy of directional Young's modulus in the vicinity of instability. The MTP is validated against PAW-DFT at 0 K on relaxed and unrelaxed SQS supercells, and on an independent 40-configuration test set drawn from the unstable low-Nb alloys.

Significance. If the central prediction is reliable, the paper would extend machine-learned interatomic potentials to a regime—dynamically unstable near-instability compositions—where classical potentials fail, and would identify a concrete composition window for low-modulus biomedical Ti alloys. The workflow is a strength: the MTP is fitted to DFT energies/forces/stresses, not to experimental elastic constants; the elastic properties emerge from MD stress–strain simulations; the training dataset and SQS structures are deposited; and an active-learning protocol is used to include low-temperature unstable configurations. The qualitative trends (decreasing modulus with decreasing Nb, increased anisotropy near instability, elinvar-like weak temperature dependence of the stable alloys) match available experiments and prior DFT studies. However, the central quantitative claim—the 30–40 GPa window at 12–17 at.% Nb—is sensitive to interpolation between sparse compositions and to the indirect inference of the instability border; this sensitivity is not yet quantified.

major comments (3)
  1. [3.4, Fig. 7] The statement that β-Ti94−xNbxZr6 with 12–17 at.% Nb exhibits EH between 30 and 40 GPa is an interpolation across only five simulated compositions (x = 1, 6, 15, 22, 30), with only Ti79Nb15Zr6 lying inside the claimed window; the boundaries at x = 12 and x = 17 are not directly simulated. Because the paper itself emphasizes a strongly nonlinear concentration dependence of the moduli near instability (Section 3.3, Fig. 5f and Fig. 6), the color-map interpolation in Fig. 7 cannot be assumed accurate across the entire window. Please perform MTP-MD for at least one or two additional compositions inside the window (e.g., x = 12 and x = 14) and, ideally, targeted DFT validation at those compositions at 0 K or at 300 K, to support the specific 30–40 GPa claim.
  2. [3.2–3.3, Figs. 3–5] The instability border shown as dashed/solid lines in Fig. 7 is inferred from the minimum of the MSD versus temperature (Fig. 3) and from the appearance of nonzero stress components (Fig. 4). These are indirect criteria, and the paper explicitly acknowledges in Section 3.3 that 'we cannot be certain that dynamical and mechanical instabilities coincide' and that pure β-Ti has non-Γ instabilities. Because the claimed 30–40 GPa modulus window is located immediately adjacent to this border, a shift of the border by only 2–3 at.% Nb or by 50–100 K would move or eliminate the window. The authors should substantiate the stability boundary with a more direct method, such as phonon dispersions for representative SQSs at 0 K, or at least a finite-temperature spectral-function analysis using the MTP, cross-checked against DFT for a small number of configurations.
  3. [Supplement S2, S5 (Fig. S3, Fig. S6, Table S1)] The validation of the MTP in the near-instability regime is limited: the independent test set (Supp. S2) comprises only 40 MD snapshots from the two low-Nb alloys and checks energies and stresses, not elastic constants; the 0 K unrelaxed-SQS comparison (Supp. S5) shows systematic errors ΔC12 ≈ 5–6 GPa and ΔC44 ≈ 3–4 GPa. Since the predicted softening is driven by C′ = (C11 − C12)/2 approaching zero, a 5–6 GPa error in C12 translates to a ~2.5–3 GPa error in C′, which could shift the composition at which C′ crosses zero by several at.% Nb and materially change the predicted EH. This error budget is not reflected in the uncertainty of the central 30–40 GPa window; please quantify the propagation of the MTP error (e.g., by re-fitting or bootstrap) and, if feasible, recompute the finite-temperature EH for at least one composition in the claimed window directly from DFT MD.
minor comments (5)
  1. [3.2, last paragraph] The critical temperatures are stated inconsistently: the text says 'Ti93Nb1Zr6 is dynamically unstable below 450–500 K and Ti88Nb6Zr6 is unstable bellow 350–400 K,' and then gives critical temperatures of '~400 and ~5 00 K, respectively.' The numbers appear to be swapped; please correct and ensure consistency with the instability borders in Fig. 7.
  2. [2.4] There are several typos: 'aroud' should be 'around', 'expalined' should be 'explained', 'disscus' should be 'discuss', and 'emphisize' should be 'emphasize'.
  3. [3.1] The phrase 'underesteemates C44' should read 'underestimates C44'.
  4. [3.2] The word 'anomality' should be 'anomaly' in 'This anomality can be attributed to the dynamical instability.'
  5. [3.4] The central claim of a 30–40 GPa EH window is stated without an explicit uncertainty estimate; please report the numerical uncertainty of the predicted EH values (e.g., from the averaging over orientations and from the MTP energy/stress errors) and add error bars or a confidence interval to Fig. 7.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: elastic moduli are emergent predictions from a DFT-trained MTP, not fitted targets.

full rationale

The central claims—the elinvar effect and the 30–40 GPa low-modulus window at 12–17 at.% Nb—are not circular. The MTP is fitted to DFT energies, forces, and stresses (Sections 2.3–2.4), not to elastic constants, polycrystalline moduli, or experimental bone-modulus values. The elastic constants are obtained from separate NVT stress-strain simulations (Section 2.4, Section 3.3), so the reported Cij, G, E, and EH are emergent predictions rather than fitted outputs. The instability borders are inferred from MSD minima and stress fluctuations in MTP-MD (Figures 3–4), which are distinct observables from the training labels; the paper explicitly acknowledges the uncertainty about whether dynamical and mechanical instabilities coincide (Section 3.3). The independent test set of 40 configurations from the low-Nb unstable alloys (Supplementary Section S2) provides external validation in the regime relevant to the low-modulus claim. Self-citations ([30], [31], [64]) are used to justify the MTP level choice and to connect to prior elinvar results, but they are not load-bearing: the present work validates the MTP against its own DFT calculations (Figure 1, Supplementary S5–S6) and against experimental Ti-Nb-Zr trends (Figure 8). The 12–17 at.% Nb / 30–40 GPa statement is an interpolation between computed compositions, which is a predictive use of the surrogate, not a reduction of the prediction to the inputs by construction. No specific circular step could be quoted, so the derivation chain is judged self-contained.

Assumptions & free parameters 3 free parameters · 7 assumptions · 0 invented entities

The predictions inherit all DFT and SQS modeling assumptions. The fitted MTP coefficients and hand-chosen hyperparameters are the main parameters paid for by the fitting procedure, and no new physical entities are introduced.

free parameters (3)
  • MTP coefficients (16g level, 380 basis functions) = Fit by linear regression to the 880-configuration DFT training set
    These coefficients define the potential energy surface from which all elastic constants and instability indicators are computed. They are fitted to DFT data, not derived from theory, so the finite-temperature predictions are surrogate outputs of that fit.
  • MTP cutoff radius = 5 Å
    Chosen after comparing RMSE for 5, 6, and 7 Å cutoffs in Supplementary Section S1. The cutoff affects the accuracy of energies and stresses in the unstable region.
  • Active-learning selection thresholds = Initial gamma = 150; final gamma < 1.2; gamma_select and gamma_breaking in MLIP-2
    These hand-chosen thresholds control which configurations enter the training set and therefore shape the potential in the low-temperature low-niobium region. The stopping criterion gamma < 1.2 is an assumption about MTP reliability.
assumptions (7)
  • domain assumption DFT-PBE with PAW accurately describes energies, forces, and stresses for Ti-Nb-Zr alloys.
    The MTP is trained and validated against VASP/DFT, so any DFT error transfers directly to the predicted elastic moduli. Invoked in Section 2.1.
  • domain assumption A single 128-atom SQS per composition represents the random solid solution sufficiently for elastic property prediction.
    One chemical realization per composition was used; no configurational averaging over multiple SQSs is performed. Invoked in Section 2.1 and Section 3.1.
  • ad hoc to paper The MTP extrapolation grade gamma < 1.2 is a sufficient criterion for reliability of MD predictions in the dynamically unstable region.
    The authors use gamma below 1.2 over 30,000 MD steps as the active-learning stopping criterion, with only 40 independent test configurations in the unstable region. Invoked in Section 2.4 and Supplementary Section S2.
  • domain assumption Linear stress-strain response with strains of ±2% and ±4% yields valid elastic constants even near mechanical instability.
    Used for all Cij extractions; near instability the stress-strain relation may be nonlinear or configuration-dependent. Invoked in Sections 3.1 and S7.
  • domain assumption The metastable bcc phase remains the relevant phase; transformations to omega, alpha, or martensite are neglected.
    The authors state that bcc Ti alloys are thermodynamically metastable and may transform under deformation; the simulations impose bcc-based SQS and MD cells. Invoked in the Introduction and Section 2.
  • domain assumption Molecular dynamics of 30,000 timesteps with 5,000 discarded yields converged thermal averages for lattice parameters, MSD, and stresses.
    No block-average convergence analysis is reported. Invoked in Sections 2.4 and 3.2.
  • standard math Voigt-Reuss-Hill averaging gives meaningful polycrystalline moduli for these anisotropic metastable alloys.
    Standard homogenization procedure; the spread between Voigt and Reuss is used as an anisotropy indicator. Invoked in Section S8.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine learning interatomic potential for the low-modulus Ti-Nb-Zr alloys in the vicinity of dynamical instability." pith.science (2026). https://pith.science/paper/L4LDZFUJ

@misc{pith2026241206270,
  author       = {Pith},
  title        = {Pith review of: Machine learning interatomic potential for the low-modulus Ti-Nb-Zr alloys in the vicinity of dynamical instability},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L4LDZFUJ}},
  note         = {Machine review of arXiv:2412.06270}
}
read the original abstract

Traditionally, alloying and thermal treatment are considered as the main tools for design of new materials. Application of first-principles simulations can significantly accelerate the process of materials design, however, to account for both, multicomponent chemical disorder and finite temperature effects in theoretical simulations is a challenging task. In this work we have trained machine learning interatomic potential to effectively simulate finite temperature elastic properties of multicomponent \beta-Ti94-xNbxZr6 alloys. Our simulations predict the presence of the elinvar effect for the wide range of temperatures. Importantly, we predict that in a vicinity of dynamical and mechanical instability, the \beta-Ti94-xNbxZr6 alloys demonstrate strongly non-linear concentration-dependence of elastic moduli, which leads to low values of moduli comparable to that of human bone. Moreover, these alloys demonstrate a strong anisotropy of directional Young's modulus which can be helpful for microstructure tailoring and design of materials with desired elastic properties.

Figures

Figures reproduced from arXiv: 2412.06270 by the authors.

Figure 1
Figure 1. Elastic constants 𝐶𝑖𝑗 of β-Ti94-xNbxZr6 alloys at T = 0 K. The PAW calculations are compared with MLIP predictions. The error bars in elastic constants 𝐶𝑖𝑗 are from averaging (see text above) [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Temperature dependence of (a) lattice parameters and (b) coefficient of linear thermal expansion of Ti94-xNbxZr6 alloys with different Nb concentrations. Lattice parameters are obtained from NpT calculations at P = 0. Each simulations box contains 128,000 atoms. Dashes lines indicate lattice parameters in regions of dynamical instability. Next, we used the information on lattice parameters and thermal expansion to f… view at source ↗
Figure 3
Figure 3. Temperature dependence of mean square displacements MSD of atoms from their ideal lattice positions in Ti94-xNbxZr6 alloys. The displacements were determined along three crystallographic directions: (a) MSDX, (b) MSDY and (c) MSDZ. The MSDs were determined within NVT simulations, accounting for the thermal expansion of the alloys. Each simulation box contained 128,000 atoms. Since the MSD shown in [PITH_FULL_IMAGE:… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Stress components 𝑆𝑖𝑗 in undistorted Ti94-xNbxZr6 alloys. The 𝑆𝑖𝑗 are shown for alloys with 1, 6 and 15 at.% Nb and at temperatures of 300 and 500 K. Each simulation box contains 128,000 atoms. 3.3 Elastic properties at finite-temperatures [PITH_FULL_IMAGE:figures/ful…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

87 extracted references · 41 canonical work pages

  1. [1]

    Grimvall, B

    G. Grimvall, B. Magyari-Köpe, V. Ozoliņš, and K.A. Persson. Lattice instabilities in metallic elements. Rev. Mod. Phys. 84 (2012). https://doi.org/10.1103/RevModPhys.84.945

  2. [2]

    Deringer, M.A

    V.L. Deringer, M.A. Caro and G. Csányi. Machine learning interatomic potentials as emerging tools for materials science . Adv. Mater. 31 (2019) p. 1902765. https://doi.org/10.1002/adma.201902765

  3. [3]

    J. Behler. Perspective: Machine learning potentials for atomistic simulations. J. Chem. Phys. 145 (2016) p. 170901. https://doi.org/10.1063/1.4966192

  4. [4]

    J. Behler . Neural network potential -energy surfaces in chemistry: a tool for large -scale simulations. Phys. Chem. Chem. Phys. 13 (2011) pp. 17930–17955. https://doi.org/10.1039/C1CP21668F

  5. [5]

    Grisafi, J

    A. Grisafi, J. Nigam, M. Ceriotti. Multi-scale approach for the prediction of atomic scale properties. Chem. Sci. 12 (2021) pp. 2078–2090. https://doi.org/10.1039/D0SC04934D

  6. [6]

    Zubatiuk and O

    T. Zubatiuk and O. Isayev. Development of multimodal machine learning potentials: Toward a physics-aware artificial intelligence . Acc. Chem. Res. 54 (2021) pp. 1575–1585. https://doi.org/10.1021/acs.accounts.0c00868

  7. [7]

    O. T. Unke and M. Meuwly, PhysNet: A neural network for predicting energies, forces, dipole moments, and partial charges . J. Chem. Theory Comput. 15 (2019) pp. 3678–3693. https://doi.org/10.1021/acs.jctc.9b00181

  8. [8]

    O.T. Unke, S. Chmiela, H.E. Sauceda, M. Gastegger, I. Poltavsky, K.T. Schütt, A. Tkatchenko and K. -R. Müller. Machine learning force fields . Chem. Rev. 121 (2021) pp. 10142–10186. https://doi.org/10.1021/acs.chemrev.0c01111

Show all 87 references
  1. [9]

    Friederich, F

    P. Friederich, F. Häse, J. Proppe, and A. Aspuru -Guzik. Machine-learned potentials for next - generation matter simulations. Nat. Mater 20 (2021) pp. 750–761. https://doi.org/10.1038/s41563- 020-0777-6

  2. [10]

    R. Drautz. Atomic cluster expansion for accurate and transferable interatomic potentials. Phys. Rev. B 99 (2019) 014104. https://doi.org/10.1103/PhysRevB.99.014104

  3. [11]

    Thompson, L.P

    A.P. Thompson, L.P. Swiler, C.R. Trott, S.M. Foiles, and G.J. Tucker. Spectral neighbor analysis method for automated generation of quantum -accurate interatomic potentials . J. Comp. Phys. 285 (2015) pp. 316–330. https://doi.org/10.1016/j.jcp.2014.12.018

  4. [12]

    Wood and A.P

    M.A. Wood and A.P. Thompson. Extending the accuracy of the SNAP interatomic potential form. J. Chem. Phys. 148 (2018) p. 241721. https://doi.org/10.1063/1.5017641

  5. [13]

    Lorenz, A

    S. Lorenz, A. Groß, and M. Scheffler. Representing high-dimensional potential-energy surfaces for reactions at surfaces by neural networks . Chem. Phys. Lett. 395 (2004) pp. 210–215. https://doi.org/10.1016/j.cplett.2004.07.076

  6. [14]

    Behler and M

    J. Behler and M. Parrinello. Generalized Neural-Network Representation of High-Dimensional Potential-Energy Surfaces . Phys. Rev. Lett. 98 (2007) p. 146401. https://doi.org/10.1103/PhysRevLett.98.146401

  7. [15]

    Liu and J

    M. Liu and J. R. Kitchin, SingleNN: A modified Behler-Parrinello neural network with shared weights for atomistic simulations with transferability . J. Phys. Chem. C 124 (2020) pp. 17811 – 17818. https://doi.org/10.1021/acs.jpcc.0c04225

  8. [16]

    Bartók, M.C

    A.P. Bartók, M.C. Payne, R. Kondor, and G. Csányi. Gaussian approximation potentials: The accuracy of quantum mechanics, without the electrons . Phys. Rev. Lett. 104 (2010) p. 136403. https://doi.org/10.1103/PhysRevLett.104.136403

  9. [17]

    Bartók and G

    A.P. Bartók and G. Csányi . Gaussian approximation potentials: A brief tutorial introduction . Int. J. Quant. Chem. 115 (2015) pp. 1051–1057. https://doi.org/10.1002/qua.24927

  10. [18]

    Klawohn, J .P

    S. Klawohn, J .P. Darby, J .R. Kermode, G . Csányi, M .A. Caro, A .P. Bartók Gaussian approximation potentials: Theory, software implementation and application examples . J. Chem. Phys. 159 (2023) p. 174108. https://doi.org/10.1063/5.0160898

  11. [19]

    Bartók, M.J

    A.P. Bartók, M.J. Gillan, F.R. Manby, and G. Csányi. Machine-learning approach for one- and two-body corrections to density functional theory: Applications to molecular and condensed water. Phys. Rev. B 88 (2013) p. 054104. https://doi.org/10.1103/physrevb.88.054104

  12. [20]

    Dragoni, T.D

    D. Dragoni, T.D. Daff, G. Csányi, and N. Marzari. Achieving DFT accuracy with a machine- learning interatomic potential: Thermomechanics and defects in bcc ferromagnetic iron. Phys. Rev. Mater. 2 (2018) p. 013808. https://doi.org/10.1103/physrevmaterials.2.013808

  13. [21]

    Raissi, P

    M. Raissi, P. Perdikaris, G.E. Karniadakis. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys. 378 (2019) pp. 686–707. https://doi.org/10.1016/j.jcp.2018.10.045

  14. [22]

    Purja Pun, R

    G.P. Purja Pun, R. Batra, R. Ramprasad, Y. Mishin. Physically informed artificial neural networks for atomistic modeling of materials. Nat. Commun. 10 (2019) pp. 1–10. https://doi.org/10.1038/s41467-019-10343-5

  15. [23]

    Karniadakis, I.G

    G.E. Karniadakis, I.G. Kevrekidis, L. Lu, P. Perdikaris, S . Wang, L. Yang. Physics-informed machine learning. Nat Rev Phys 3 (2021) pp. 422–440. https://doi.org/10.1038/s42254-021-00314- 5

  16. [24]

    Batzner, A

    S. Batzner, A. Musaelian, L. Sun, M . Geiger, J.P. Mailoa, M . Kornbluth, N. Molinari, T.E. Smidt and B. Kozinsky. E(3)-equivariant graph neural networks for data -efficient and accurate interatomic potentials. Nat Commun 13 (2022) p. 2453. https://doi.org/10.1038/s41467-022-29939- 5

  17. [25]

    Schütt, H.E

    K.T. Schütt, H.E. Sauceda, P. -J. Kindermans, A. Tkatchenko, and K. -R. Müller. SchNet - a deep learning architecture for molecules and materials . J. Chem. Phys. 148 (2018) p. 241722. https://doi.org/10.1063/1.5019779

  18. [26]

    Reiser, M

    P. Reiser, M. Neubert, A. Eberhard, L. Torresi, Ch. Zhou, Ch. Shao, H. Metni, C. van Hoesel, H. Schopmans, T. Sommer, P. Friederich. Graph neural networks for materials science and chemistry. Commun Mater 3, 93 (2022). https://doi.org/10.1038/s43246-022-00315-6

  19. [27]

    Batatia, D.P

    I. Batatia, D.P. Kovacs, G. Simm, C. Ortner, and G. Csányi. MACE: Higher order equivariant message passing neural networks for fast and accurate force fields . Adv. Neural Inf. Process. Syst. 35 (2022) pp. 11423–11436

  20. [28]

    A.V. Shapeev. Moment tensor potentials: A class of systematically improvable interatomic potentials. Multiscale Model. Simul. 14 (2016) pp. 1153–1173. https://doi.org/10.1137/15M1054183

  21. [29]

    Novikov, K

    I.S. Novikov, K. Gubaev, E.V. Podryabinkin1 and A.V. Shapeev. The MLIP package: moment tensor potentials with MPI and active learning . Mach. Learn.: Sci. Technol. 2 (2021) p. 025002. https://doi.org/10.1088/2632-2153/abc9fe

  22. [30]

    Shapeev, E.V

    A.V. Shapeev, E.V. Podryabinkin, K. Gubaev, F. Tasnádi and I.A. Abrikosov. Elinvar effect in β-Ti simulated by on-the-fly trained moment tensor potential. New J. Phys. 22 (2020) p. 113005. https://doi.org/10.1088/1367-2630/abc392

  23. [31]

    F. Bock, F. Tasnádi, I.A. Abrikosov. Active learning with moment tensor potentials to predict material properties: Ti0.5Al0.5N at elevated temperature J. Vac. Sci. Technol. A 42 (2024) p. 013412. https://doi.org/10.1116/6.0003260

  24. [32]

    Nelson, C.J

    K.R. Nelson, C.J. Burstone, A.J. Goldberg. Optimal welding of beta titanium orthodontic wires. Am J Orthod Dentofacial Orthop 92 (1987) p. 213. https://doi.org/10.1016/0889-5406(87)90414-8

  25. [33]

    M. Niinomi. Fatigue performance and cyto -toxicity of low rigidity titanium alloy, Ti –29Nb– 13Ta–4.6Zr. Biomaterials 24 (2003) pp. 2673-2683. https://doi.org/10.1016/S0142-9612(03)00069- 3

  26. [34]

    W.F. Ho, C.P. Ju, J.H. Chern Lin. Structure and properties of cast binary Ti –Mo alloys. Biomaterials 20 (1999) pp. 2115-2122. https://doi.org/10.1016/S0142-9612(99)00114-3

  27. [35]

    Goldberg, C.J

    J. Goldberg, C.J. Burstone. An evaluation of beta titanium alloys for use in orthodontic appliances. J Dent Res. 58 (1979) pp. 593-599. https://doi.org/10.1177/00220345790580020901

  28. [36]

    Lee, W.F

    C.M. Lee, W.F. Ho, C.P. Ju, J.H. Chern Lin. Structure and properties of Titanium–25 Niobium– x iron alloys. J Mater Sci Mater Med 13 (2002) pp. 695 –700. https://doi.org/10.1023/A:1015798011434

  29. [37]

    Banerjee, P.C

    R. Banerjee, P.C . Collins, D . Bhattacharyya, S . Banerjee, H.L . Fraser. Microstructural evolution in laser deposited compositionally graded α/β titanium -vanadium alloys. Acta Mater. 51 (2003) pp. 3277-3292. https://doi.org/10.1016/S1359-6454(03)00158-7

  30. [38]

    Banerjee, S

    R. Banerjee, S. Nag, J. Stechschulte, H.L. Fraser. Strengthening mechanisms in Ti–Nb–Zr–Ta and Ti –Mo–Zr–Fe orthopedic alloys . Biomaterials 25 (2004) pp. 3413-3419. https://doi.org/10.1016/j.biomaterials.2003.10.041

  31. [39]

    Kermanpur, H

    A. Kermanpur, H. Sepehri Amin, S. Ziaei-Rad, N. Nourbakhshnia, M. Mosaddeghfar. Failure analysis of Ti6Al4V gas turbine compressor blades . Engineering Failure Analysis 15 (2008) pp. 1052-1064. https://doi.org/10.1016/j.engfailanal.2007.11.018

  32. [40]

    Sh. Luo, J. Yao, J. Li, H. Du, H. Liu, F. Yu. Influence of forging velocity on temperature and phases of forged Ti-6Al-4V turbine blade. Journal of Materials Research and Technology 9 (2020) pp. 12043-12051. https://doi.org/10.1016/j.jmrt.2020.08.106

  33. [41]

    Zh.-Yu. He, L. Zhang, W.-R. Shan, Yu-Q. Zhang, R. Zhou, Y. -H. Jiang, J. Tan. Mechanical and corrosion properties of Ti -35Nb-7Zr-xHA composites fabricated by spark plasma sintering . Trans. Nonferrous Met. Soc. China 27 (2017) pp. 848–856. https://doi.org/10.1016/S1003- 6326(...

  34. [42]

    Hao, S.J

    Y.L. Hao, S.J. Li, S.Y. Sun, C.Y. Zheng, R. Yang. Elastic deformation behaviour of Ti-24Nb- 4Zr-7.9Sn for biomedical applications . Acta Biomater. 3 (2007) pp. 277 –286. https://doi.org/10.1016/j.actbio.2006.11.002

  35. [43]

    Okazaki, A New Ti–15Zr–4Nb–4Ta alloy for medical applications, Curr

    Y. Okazaki, A New Ti–15Zr–4Nb–4Ta alloy for medical applications, Curr. Opin. Solid State Mater. Sci. 5 (1) (2001) 45–53

  36. [44]

    C.E. Wen, Y. Yamada, K. Shimojima, Y. Chino, T. Asahina, M. Mabuchi, Processing and mechanical properties of autogenous titanium implant materials . J. Mater. Sci. - Mater. Med. 13 (2002) pp. 397–401. https://doi.org/10.1023/A:1014344819558

  37. [45]

    Moffat, U.R

    D.L. Moffat, U.R. Kattner. The stable and metastable Ti -Nb phase diagrams. Metall Trans A 19 (1988) pp. 2389–2397. https://doi.org/10.1007/BF02645466

  38. [46]

    Zhang, Y

    J. Zhang, Y. Li and W. Li. Metastable phase diagram on heating in quenched Ti -Nb high- temperature shape memory alloys. J Mater Sci 56 (2021) pp. 11456–11468. https://doi.org/10.1007/s10853-021-05814-4

  39. [47]

    Banerjee and J.C

    D. Banerjee and J.C. Williams. Perspectives on titanium science and technology. Acta Materialia, 61 (2013) pp. 844-879. https://doi.org/10.1016/j.actamat.2012.10.043

  40. [48]

    Devaraj, S

    A. Devaraj, S. Nag, R. Srinivasan, R.E.A. Williams, S. Banerjee, R. Banerjee, H.L. Fraser . Experimental evidence of concurrent compositional and structural instabilities leading to ω precipitation in titanium –molybdenum alloys . Acta Materialia 60 (2012) pp. 596-609. https:/...

  41. [49]

    Skripnyak, A.V

    N.V. Skripnyak, A.V. Ponomareva, M.P. Belov, E.A. Syutkin, A.V. Khvan, A.T. Dinsdale, I.A. Abrikosov. Mixing enthalpies of alloys with dynamical instability: bcc Ti -V system. Acta Mater. 188 (2020) 145. https://doi.org/10.1016/j.actamat.2020.01.056

  42. [50]

    Raabe, B

    D. Raabe, B. Sander, M. Friák, D. Ma, J. Neugebauer . Theory-guided bottom-up design of β- titanium alloys as biomaterials based on first principles calculations: Theory and experiments. Acta Mater. 55 (2007) pp. 4475-4487. https://doi.org/10.1016/j.actamat.2007.04.024

  43. [51]

    Karre, M.K

    R. Karre, M.K. Niranjan, S.R. Dey. First principles theoretical investigations of low Young's modulus beta Ti –Nb and Ti –Nb–Zr alloys compositions for biomedical applications . Materials Science and Engineering C 50 (2015) pp. 52-58. https://doi.org/10.1016/j.msec.2015.01.061

  44. [52]

    Q.-M. Hu, S. -J. Li, Yu -L. Hao, R. Yang, B. Johansson, L. Vitos. Phase stability and elastic modulus of Ti alloys containing Nb, Zr, and/or Sn from first -principles calculations. Appl. Phys. Lett. 93 (2008) p. 121902. https://doi.org/10.1063/1.2988270

  45. [53]

    J. H. Dai, X. Wu, Y. Song, R. Yang; Electronic structure mechanism of martensitic phase transformation in binary titanium alloys. J. Appl. Phys. 112 (2012) p. 123718. https://doi.org/10.1063/1.4770481

  46. [54]

    Skripnyak, F

    N.V. Skripnyak, F. Tasnádi, S.I. Simak, A.V. Ponomareva, J. Löfstrand, P. Berastegui, U. Jansson and I.A. Abrikosov. Achieving low elastic moduli of bcc Ti –V alloys in vicinity of mechanical instability. AIP Advances 10 (2020) p. 105322. https://doi.org/10.1063/5.0023347

  47. [55]

    Skripnyak, A.V

    N.V. Skripnyak, A.V. Ponomareva, M.P. Belov, I.A. Abrikosov. Ab initio calculations of elastic properties of alloys with mechanical instability: application to bcc Ti-V alloys. Mater. Des. 140 (2018) 357. https://doi.org/10.1016/j.matdes.2017.11.071

  48. [56]

    Huang, B

    L. Huang, B. Grabowski, J. Zhang, M. Lai, C.C. Tasan, S. Sandlobes, D. Raabe and J. Neugebauer. From electronic structure to phase diagrams: a bottom -up approach to understand the stability of titanium -transition metal alloys . Acta Mater. 113 (2016) p. 311. https://doi.org/...

  49. [57]

    P.E. Blöchl. Projector augmented-wave method. Phys. Rev. B 50 (1994) 17953. https://doi.org/10.1103/PhysRevB.50.17953

  50. [58]

    Kresse and J

    G. Kresse and J. Furthmüller. Efficiency of ab initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comput. Mater. Sci. 6 (1996) 15. https://doi.org/10.1016/0927-0256(96)00008-0

  51. [59]

    Kresse and J

    G. Kresse and J. Furthmüller. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 54 (1996) 11169. https://doi.org/10.1103/PhysRevB.54.11169

  52. [60]

    Ruban and I.A

    A.V . Ruban and I.A. Abrikosov. Configurational thermodynamics of alloys from first principles: Effective cluster interactions. Rep. Prog. Phys. 71 (2008) 046501. https://doi.org/10.1088/0034-4885/71/4/046501

  53. [61]

    Zunger, S.-H

    A. Zunger, S.-H. Wei, L. G. Ferreira, and J.E. Bernard. Special Quasirandom Structures. Phys. Rev. Lett. 65 (1990) 353. https://doi.org/10.1103/PhysRevLett.65.353

  54. [62]

    Perdew, K

    J.P . Perdew, K. Burke and M. Ernzerhof. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 77 (1996) 3865. https://doi.org/10.1103/PhysRevLett.77.3865

  55. [63]

    Hellmann

    H. Hellmann. Lebenslauf von hans hellmann in Hans Hellmann: Einführung in Die Quantenchemie: Mit biografischen Notizen von Hans Hellmann jr (2015) pp. 3–15. https://doi.org/10.1007/978-3-662-45967-6

  56. [64]

    Tasnádi, F

    F. Tasnádi, F. Bock, J. Tidholm, A.V. Shapeev, I.A. Abrikosov. Efficient prediction of elastic properties of Ti0.5Al0.5N at elevated temperature using machine learning interatomic potential, Thin Solid Films 737 (2021) p. 138927. https://doi.org/10.1016/j.tsf.2021.138927

  57. [65]

    Plimpton

    S. Plimpton. Fast Parallel Algorithms for Short-Range Molecular Dynamics. J Comp Phys 117 (1995) pp. 1-19. https://doi.org/10.1006/jcph.1995.1039

  58. [66]

    W. Voigt. Ueber die Beziehung zwischen den beiden Elasticitätsconstanten isotroper Körper. Annalen der Physik 274 (1889) pp. 573–587. https://doi.org/10.1002/andp.18892741206

  59. [67]

    A. Reuss. Berechnung der Fließgrenze von Mischkristallen auf Grund der Plastizitätsbedingung für Einkristalle. ZAMM – J. Appl, Z. angew. Math. Mech. 9 (1929) pp. 49–58. https://doi.org/10.1002/zamm.19290090104

  60. [68]

    R. Hill. The elastic behaviour of a crystalline aggregate. Proc. Phys. Soc. Section A 65 (1952) pp. 349–354. https://doi.org/10.1088/0370-1298/65/5/307

  61. [69]

    Wortman, R.A

    J.J. Wortman, R.A. Evans. Young’s modulus, Shear modulus, and Poisson’s ratio in silicon and germanium. J. Appl. Phys. 36 (1965) pp. 153–156. https://doi.org/10.1063/1.1713863

  62. [70]

    Zhang, R

    L. Zhang, R. Barrett, P. Cloetens, C. Detlefs, M. Sanchez Del Rio. Anisotropic elasticity of silicon and its application to the modelling of X-ray optics. J. Synchrotron Rad. 21 (2014) pp. 507–

  63. [71]

    M. Born. On the stability of crystal lattices. I. Math. Proc. Camb. Philos. Soc. 36 (1940) pp. 160–172. https://doi.org/10.1017/S0305004100017138

  64. [72]

    Tidholm, F

    J. Tidholm, F. Tasnádi, I.A. Abrikosov. Accurate prediction of high -temperature elastic constants of Ti0.5Al0.5N random alloy . Thin Solid Films 735 (2021) p. 138872. https://doi.org/10.1016/j.tsf.2021.138872

  65. [73]

    Saito, T

    T. Saito, T. Furuta, J.-H. Hwang, S. Kuramoto, K. Nishino, N. Suzuki, R. Chen, A. Yamada, K. Ito, Y. Seno, T. Nonaka, H. Ikehata, N. Nagasako, C. Iwamoto, Y. Ikuhara and T . Sakuma. Multifunctional Alloys Obtained via a Dislocation -Free Plastic Deformation Mechanism. Science ...

  66. [74]

    Asker, A.B

    C. Asker, A.B. Belonoshko, A.S. Mikhaylushkin, and I.A. Abrikosov. First-principles solution to the problem of Mo lattice stability . Phys. Rev. B 77 (2008) 220102(R). https://doi.org/10.1103/PhysRevB.77.220102

  67. [75]

    Korbmacher, A

    D. Korbmacher, A. Glensk, A.I. Duff, M.W. Finnis, B. Grabowski, J. Neugebauer. Ab initio based method to study structural phase transitions in dynamically unstable crystals, with new insights on the β to ω transformation in titanium. Phys Rev B 100 (2019) p. 104110. https://do...

  68. [76]

    Trinkle, M.D

    D.R. Trinkle, M.D. Jones, R.G. Hennig, S.P. Rudin, R.C. Albers, and J.W. Wilkins. Empirical tight-binding model for titanium phase transformations. Phys Rev B 73 (2006) p. 094123. https://doi.org/10.1103/PhysRevB.73.094123

  69. [77]

    Dubinskiy, G

    S. Dubinskiy, G. Markova, A. Baranova, V. Vvedenskiy, I. Minkova, S. Prokoshkin, V. Brailovski. A non-typical Elinvar effect on cooling of a beta Ti-Nb-Zr alloy. Materials Letters 314 (2022) p. 131870. https://doi.org/10.1016/j.matlet.2022.131870

  70. [78]

    Kim, H.Y

    K.M. Kim, H.Y. Kim, S. Miyazaki. Effect of Zr Content on Phase Stability, Deformation Behavior, and Young’s Modulus in Ti –Nb–Zr Alloys. Materials 13 (2020) 476. https://doi.org/10.3390/ma13020476

  71. [79]

    Inaekyan, V

    K. Inaekyan, V. Brailovski, S. Prokoshkin, V. Pushin, S. Dubinskiy, V. Sheremetyev . Comparative study of structure formation and mechanical behavior of age-hardened Ti–Nb–Zr and Ti–Nb–Ta shape memory alloys . Materials Characterization 103 (2015) pp. 65-74. https://doi.org/10...

  72. [80]

    Biesiekierski, J

    A. Biesiekierski, J. Lin, K. Munir, S. Ozan, Yu. Li, C. Wen. An investigation of the mechanical and microstructural evolution of a TiNbZr alloy with varied ageing time. Sci Rep 8 (2018) 5737. https://doi.org/10.1038/s41598-018-24155-y

  73. [81]

    Sheremetyev, S.D

    V.A. Sheremetyev, S.D. Prokoshkin, V. Brailovski, S.M. Dubinskiy, A.V. Korotitskiy, M.R. Filonov, M.I. Petrzhik. Investigation of the structure stability and superelastic behavior of thermomechanically treated Ti-Nb-Zr and Ti-Nb-Ta shape-memory alloys. Phys. Metals Metallogr. ...

  74. [82]

    Salvador, H.P

    C.A.F. Salvador, H.P. Van Landeghem, and R.A. Antunes. Selection of Ti Alloys for Bio - Implants: An Application of the Ashby Approach with Conflicting Objectives. Adv. Eng. Mater. 25 (2023) p. 2301169. https://doi.org/10.1002/adem.202301169

  75. [83]

    Qazi, H.J

    J.I. Qazi, H.J. Rack , B. Marquardt . High-strength metastable beta -titanium alloys for biomedical applications. JOM 56 (2004) pp. 49–51. https://doi.org/10.1007/s11837-004-0253-9

  76. [84]

    S. Pilz, T. Gustmann, F. Günther, M. Zimmermann, U. Kühn, A. Gebert. Controlling the Young’s modulus of a ß -type Ti-Nb alloy via strong texturing by LPBF. Materials & Design 216 (2022) p. 110516. https://doi.org/10.1016/j.matdes.2022.110516

  77. [85]

    M. Tane, S. Akita, T. Nakano K. Hagihara, Y. Umakoshi, M. Niinomi, H. Nakajima a. Peculiar elastic behavior of Ti –Nb–Ta–Zr single crystals. Acta Materialia 56 (2008) pp. 2856 -2863. https://doi.org/10.1016/j.actamat.2008.02.017

  78. [86]

    Data for: Machine learning interatomic potential for the low-modulus Ti-Nb-Zr alloys in the vicinity of dynamical instability

    Dataset: B.O. Mukhamedov, F. Tasnadi and I.A. Abrikosov (2025). “Data for: Machine learning interatomic potential for the low-modulus Ti-Nb-Zr alloys in the vicinity of dynamical instability”. Repository: https://public.openmaterialsdb.se/TiNbZr_MTP/TiNbZr_MTP.tar.gz Supplemen...

  79. [517]

    https://doi.org/10.1107/S1600577514004962

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.