The direct catalyst-free transformation of graphite into diamond requires pressures far above the graphite-diamond equilibrium line and preferentially yields metastable hexagonal diamond at lower temperatures. Khaliullin, Eshet, Kühne, Behler and Parrinello (Nature Materials 10, 693 (2011)) explained this phenomenology with large-scale neural-network-potential simulations of homogeneous diamond nucleation in graphite, but their energetics were computed at zero temperature and provide only upper bounds on the nucleation barriers. They explicitly deferred the central unresolved question: a quantitative account, at finite temperature and experimentally relevant pressures, of heterogeneous nucleation on defects in the graphite lattice — how defect seeds such as stacking faults, grain boundaries, or pre-existing sp3sp^3-bonded regions, together with thermal fluctuations, lower the nucleation free-energy barriers and change the critical nucleus sizes. Later work has established mechanistically that grain boundaries and coherent graphite-diamond interfaces dominate real transformations, and the machine-learning-potential plus enhanced-sampling machinery needed for the quantitative study now exists, but no published work reports a systematic determination of heterogeneous nucleation barriers and critical sizes as functions of defect type, temperature, and pressure. The problem remains open.

Background

Graphite (stacked sp2sp^2-bonded carbon sheets) and diamond (three-dimensionally sp3sp^3-bonded carbon) are the two most familiar crystalline allotropes of carbon. Static compression of graphite without a catalyst converts it to diamond only at pressures of roughly 12 GPa or more — far above the graphite-diamond thermodynamic coexistence line (the Berman–Simon line, PP(GPa) ≈0.76+2.78×10−3T\approx 0.76 + 2.78\times10^{-3}T(K)) — and at comparatively low temperatures (about 1200–1700 K) the first product is often metastable hexagonal diamond (lonsdaleite, the wurtzite-structured polymorph) rather than the more stable cubic diamond. Explaining both the pressure gap and the polymorph selectivity is a classic problem in high-pressure carbon physics.

A nucleation barrier is the free-energy cost of forming a tiny embryo of the new phase; below a critical nucleus size the embryo shrinks, above it it grows. 'Homogeneous' nucleation assumes the embryo forms in a perfect crystal, whereas 'heterogeneous' nucleation assumes it is seeded at a defect — here, for example, a stacking fault (a local error in the ABAB... stacking of graphite's hexagonal planes), a grain boundary between misoriented crystallites, or a region already containing sp3sp^3 bonds. Defects typically reduce both the barrier and the critical size, so homogeneous barriers are upper bounds on the true (defect-assisted) barriers.

Khaliullin, Eshet, Kühne, Behler and Parrinello (2011) studied this transition with an artificial neural-network potential — an interatomic force field fitted to density-functional-theory energies so that very large systems (their simulation cell held 145,152 atoms) can be modeled with near-quantum accuracy. Seeding diamond nuclei of various sizes in a graphite matrix and relaxing at zero temperature, they showed that nucleation (not the previously proposed concerted, whole-crystal mechanism) accounts for the observed phenomenology: the lattice distortion accompanying a diamond embryo inhibits the transition at low pressure and directs it toward hexagonal diamond at higher pressure. Their computed barriers are zero-temperature upper bounds — near 20 GPa the critical nucleus reaches a radius of roughly 25–31 Å with a barrier of about 560–630 eV. They also noted that cubic diamond should nucleate around stacking faults common in hexagonal graphite, and that thermal fluctuations can slide the graphite basal planes, creating stacking sequences that favor cubic diamond. The paper closes by deferring exactly the question left open: 'A finite temperature MD study of the thermodynamics and kinetics for heterogeneous nucleation on defects in the graphite lattice will be carried out in the future.' (MD = molecular dynamics, atom-by-atom simulation of thermal motion.)

Subsequent work has clarified the mechanism landscape: finite-temperature MD showed diamond nuclei originate at graphite grain boundaries and grow fastest along specific crystal directions (Zhu et al., Matter 2020); recovered partially transformed samples reveal coherent graphite-diamond interfaces — structurally registered interfaces with mixed sp2sp^2/sp3sp^3 bonding — that govern the transformation (Luo et al., Nature 2022); atomic-resolution experiments plus simulation identified orthorhombic graphite as an intermediate phase (Luo et al., Carbon 2024). Meanwhile the machinery for quantitative finite-temperature nucleation studies has matured: machine-learning potentials combined with metadynamics (an enhanced-sampling technique that reconstructs free-energy landscapes by periodically biasing the system away from already-visited states) have resolved size-dependent nucleation in another solid-solid transition (Santos-Florez et al., PRL 2022), and machine-learning potentials trained specifically for nucleation in graphite have been reported (Erokhin et al., 2024). Finite-temperature nucleation rates have also been quantified for crystallization of diamond and graphite from liquid carbon (Donadio et al., Nature Communications 2025) — a different, liquid-side problem. What no later work supplies is the quantitative core of the 2011 open question: heterogeneous nucleation barriers and critical nucleus sizes for specific defect seeds in solid graphite, as functions of temperature and pressure, compared against the homogeneous upper bounds.

Problem Statement

For the direct graphite-to-diamond phase transition, determine quantitatively — by finite-temperature molecular dynamics or an equivalent statistical-mechanical treatment, at pressures and temperatures of experimental relevance — the thermodynamics and kinetics of heterogeneous diamond nucleation on defects in the graphite lattice. Specifically: (i) construct explicit defect seeds in graphite (such as stacking faults, grain boundaries, or pre-existing sp3sp^3-bonded regions) of specified atomic structure; (ii) compute the heterogeneous nucleation free-energy barriers and critical nucleus sizes for cubic and hexagonal diamond as functions of temperature and pressure; and (iii) establish how much these defect-mediated barriers and critical sizes are reduced relative to the zero-temperature homogeneous upper bounds (about 560–630 eV near 20 GPa) computed by Khaliullin et al. (2011), including the resulting temperature-dependent nucleation rates. A conclusive account must cover at least one representative defect seed of each of the three named classes — a stacking fault, a grain boundary, and a pre-existing sp3sp^3-bonded region — over the experimentally relevant range P≈10–30P \approx 10\text{–}30 GPa and T≈1000–2000T \approx 1000\text{–}2000 K, for both cubic and hexagonal diamond; a single defect type at a single (T,P)(T,P) point does not resolve the task. The common objective is a quantitative, defect-resolved account of how thermal fluctuations and lattice defects control the barrier, critical size, and kinetics of diamond nucleation from graphite under synthesis conditions.

References

  1. Rustam Z. Khaliullin, Hagai Eshet, Thomas D. Kühne, Jörg Behler, Michele Parrinello, "Nucleation mechanism for the direct graphite-to-diamond phase transition", Nature Materials 10, 693–697 (2011). DOI: 10.1038/nmat3078; arXiv:1101.1406. https://arxiv.org/abs/1101.1406
  2. Sheng-cai Zhu, Xiao-zhi Yan, Jin Liu, Artem R. Oganov, Qiang Zhu, "A Revisited Mechanism of the Graphite-to-Diamond Transition at High Temperature", Matter 3, 864–878 (2020). DOI: 10.1016/j.matt.2020.05.013. https://doi.org/10.1016/j.matt.2020.05.013
  3. Duan Luo, Liuxiang Yang, Hongxian Xie, Srilok Srinivasan, Jinshou Tian, Subramanian Sankaranarayanan, Ilke Arslan, Wenge Yang, Ho-kwang Mao, Jianguo Wen, "Atomistic evidence of nucleation mechanism for the direct graphite-to-diamond transformation", Carbon 229, 119538 (2024). DOI: 10.1016/j.carbon.2024.119538. https://doi.org/10.1016/j.carbon.2024.119538
  4. Kun Luo, Bing Liu, Wentao Hu, Xiao Dong, Yanbin Wang, Quan Huang, Yufei Gao, Lei Sun, Zhisheng Zhao, Yingju Wu, Yang Zhang, Mengdong Ma, Xiang-Feng Zhou, Julong He, Dongli Yu, Zhongyuan Liu, Bo Xu, Yongjun Tian, "Coherent interfaces govern direct transformation from graphite to diamond", Nature 607, 486–491 (2022). DOI: 10.1038/s41586-022-04863-2. https://doi.org/10.1038/s41586-022-04863-2
  5. Dezhou Guo, Kun Luo, Qi An, "Shear-promoted graphite-to-diamond phase transition at the grain boundary of nanocrystalline graphite", Functional Diamond 4, 2366807 (2024). DOI: 10.1080/26941112.2024.2366807. https://doi.org/10.1080/26941112.2024.2366807
  6. Jiajun Dong, Zhen Yao, Mingguang Yao, Rui Li, Kuo Hu, Luyao Zhu, Yan Wang, Huanhuan Sun, Bertil Sundqvist, Ke Yang, Bingbing Liu, "Decompression-Induced Diamond Formation from Graphite Sheared under Pressure", Physical Review Letters 124, 065701 (2020). DOI: 10.1103/PhysRevLett.124.065701. https://doi.org/10.1103/PhysRevLett.124.065701
  7. Davide Donadio, Margaret L. Berrens, Wanyu Zhao, Shunda Chen, Tianshu Li, "Metastability and Ostwald step rule in the crystallisation of diamond and graphite from molten carbon", Nature Communications 16, 6324 (2025). DOI: 10.1038/s41467-025-61674-5. https://doi.org/10.1038/s41467-025-61674-5
  8. Pedro A. Santos-Florez, Howard Yanxon, Byungkyun Kang, Yansun Yao, Qiang Zhu, "Size-Dependent Nucleation in Crystal Phase Transition from Machine Learning Metadynamics", Physical Review Letters 129, 185701 (2022). DOI: 10.1103/PhysRevLett.129.185701. https://doi.org/10.1103/PhysRevLett.129.185701
  9. S. V. Erokhin, M. A. Builova, P. B. Sorokin, "Training of Machine Learning Potentials for the Modeling of Nucleation in Graphite", Journal of Structural Chemistry 65, 831–839 (2024). DOI: 10.1134/S0022476624040188. https://doi.org/10.1134/S0022476624040188
  10. Tao Yu, Shengcai Zhu, "Hexagonal diamond formation mechanism: A review", Matter and Radiation at Extremes 11, 033801 (2026). DOI: 10.1063/5.0310909. https://doi.org/10.1063/5.0310909
  11. Peng Geng, Paulo S. Branicio, "Atomistic insights on the pressure-induced multi-layer graphene to diamond-like structure transformation", Carbon 175, 243–253 (2021). DOI: 10.1016/j.carbon.2021.01.007. https://doi.org/10.1016/j.carbon.2021.01.007