Optimizing Methane\nUptake on N/O Functionalized Graphene\nvia DFT, Machine Learning, and Uniform Manifold Approximation and\nProjection (UMAP) Techniques

Carbon materials possess active sites and functionalities\non the\nsurface that can attract prominent interest as solid adsorbents for\ndiverse gas adsorption. This study aimed to predict the optimized\nmethane uptake, adsorption energy (<i>E</i><sub>ad</sub>), and adsorbent rediscovery through multitechniques of neural, regression,\nclassifier ML-DFT, and Uniform Manifold Approximation and Projection\n(UMAP). Nitrogen and oxygen (N/O) functionalities and graphene, graphene\noxide (GO), and N-doped GO were applied to the methane storage medium.\nMulti-ML algorithms were employed for the adsorption energy of CH<sub>4</sub> uptake on (i) N/O functionalities such as pyridinic (N-<i>py</i>), carboxyl (O–II), oxidized (N-<i>x</i>), hydroxyl (O-<i>h</i>), Nitroso (N-<i>ni</i>), and Amine (primary, secondary<i>,</i> and tertiary).\n(ii) The graphene surfaces are decorated with N/O heteroatoms to construct\ngraphene oxide (GO) and N-doped GO. The DFT calculations were applied\nby PW91 and the Dmol<sup>3</sup> package. N/O-functionalities in the\ndistance of ∼2.0 to 3.1 Å groups obtained <i>E</i><sub>ad</sub> of approximately −2.0 to −4 eV. Further,\nML models accomplished the forthcoming rediscovery of CH<sub>4</sub> physisorption by using the multiadsorptive features of optimized\nadsorbents with an <i>R</i><sup>2</sup> of 0.99. ML-derived\nsensitivity analysis approach was applied to specifications such as\ndeformation adsorption energy, N/O functionality type, and optimized\nstructure. CH<sub>4</sub> adsorption specifications indicate sensitivity\nlevels of −0.03 to 0.02 eV. The synergetic DFT/ML approaches\ndistinguished the modeled and rediscovered phases of CH<sub>4</sub> adsorption on N/O functional groups and graphene structures. UMAP\nis employed as a new adsorbent screening approach to play a complementary\nrole in the ML modeling process.