148
T. Toyao et al.
most important descriptor is the periodic table group of the doped metal, followed by
surface energy, melting point, and atomic radius. This result indicates that E CH3 –E CH2
values, which govern the selectivity of catalytic reactions, derived from adsorption
energies can readily be predicted using ML.
3 Conclusion
In this chapter, we provided ample evidence that ML methods can be used to predict
the DFT-calculated adsorption energies of CH 3 , CH 2 , CH, C, and H on Cu-based
alloys. We demonstrated that ETR with 12 descriptors, which are readily available
from databases, had the best ML prediction accuracy in terms of RMSE. We also
demonstrated that the top three descriptors, namely the periodic table group, surface
energy, and melting point, for doped metals can be used to accurately predict adsorption energies. In addition, the difference between the adsorption energies of CH 3 and
CH 2 (E CH3 –E CH2 ) was predicted to optimize the utilization of methane.
The results obtained in this study suggest that ML methods have high potential for predicting the catalytic activity of materials and the selectivity of catalytic
reactions. An ML model can be applied in the screening of large libraries of alloys
and potentially other solid materials (e.g., oxides) widely used for CH 4 utilization.
The ultimate goal is to have enough knowledge of the factors determining catalytic
activity to tailor catalysts atom by atom. The catalytic properties of a material are in
principle determined by its electronic structure. The objective is thus the engineering
of electronic structure by changing the composition and physical structure. The aim
of controlling matter at the molecular scale by engineering the electronic structure
is not restricted to catalytic materials; it is a general challenge in chemistry, physics,
and materials science.
References
1. Klanner C, Farrusseng D, Baumes L, et al (2004) Angew Chemie - Int Ed 43:5347–5349.
2. Behler J, Article C (2011) Phys Chem Chem Phys 13:17930.
3. Nørskov JK, Bligaard T, Rossmeisl J, Christensen CH (2009) Nat Chem 1:37–46.
4. Hansgen DA, Vlachos DG, Chen JG (2010) Nat Chem 2:484–489.
5. Toyao T, Maeno Z, Takakusagi S, et al (2020) ACS Catal 10:2260–2297.
6. Suzuki K, Toyao T, Maeno Z, et al (2019) ChemCatChem 11:4537–4547.
7. Ras EJ, Rothenberg G (2014) RSC Adv 4:5963–5974.
8. Coperet C, Comas-Vives A, Larmier K, Copéret C (2017) Chem Commun 53:4296–4303.
9. Bligaard T, Bullock RM, Campbell CT, et al (2016) ACS Catal 6:2590–2602.
10. Takigawa I, Shimizu K, Tsuda K, Takakusagi S (2016) RSC Adv 6:52587–52595.
11. Ruban A, Hammer B, Norskov JK (1997) J Mol Catal A, Chem 115:421–429
12. Nørskov JK, Abild-Pedersen F, Studt F, Bligaard T (2011) Proc Natl Acad Sci U S A 108:937–
943.
13. Santen RAV, Neurock M, Shetty SG (2010) Chem Rev 110:2005–2048.
T. Toyao et al.
most important descriptor is the periodic table group of the doped metal, followed by
surface energy, melting point, and atomic radius. This result indicates that E CH3 –E CH2
values, which govern the selectivity of catalytic reactions, derived from adsorption
energies can readily be predicted using ML.
3 Conclusion
In this chapter, we provided ample evidence that ML methods can be used to predict
the DFT-calculated adsorption energies of CH 3 , CH 2 , CH, C, and H on Cu-based
alloys. We demonstrated that ETR with 12 descriptors, which are readily available
from databases, had the best ML prediction accuracy in terms of RMSE. We also
demonstrated that the top three descriptors, namely the periodic table group, surface
energy, and melting point, for doped metals can be used to accurately predict adsorption energies. In addition, the difference between the adsorption energies of CH 3 and
CH 2 (E CH3 –E CH2 ) was predicted to optimize the utilization of methane.
The results obtained in this study suggest that ML methods have high potential for predicting the catalytic activity of materials and the selectivity of catalytic
reactions. An ML model can be applied in the screening of large libraries of alloys
and potentially other solid materials (e.g., oxides) widely used for CH 4 utilization.
The ultimate goal is to have enough knowledge of the factors determining catalytic
activity to tailor catalysts atom by atom. The catalytic properties of a material are in
principle determined by its electronic structure. The objective is thus the engineering
of electronic structure by changing the composition and physical structure. The aim
of controlling matter at the molecular scale by engineering the electronic structure
is not restricted to catalytic materials; it is a general challenge in chemistry, physics,
and materials science.
References
1. Klanner C, Farrusseng D, Baumes L, et al (2004) Angew Chemie - Int Ed 43:5347–5349.
2. Behler J, Article C (2011) Phys Chem Chem Phys 13:17930.
3. Nørskov JK, Bligaard T, Rossmeisl J, Christensen CH (2009) Nat Chem 1:37–46.
4. Hansgen DA, Vlachos DG, Chen JG (2010) Nat Chem 2:484–489.
5. Toyao T, Maeno Z, Takakusagi S, et al (2020) ACS Catal 10:2260–2297.
6. Suzuki K, Toyao T, Maeno Z, et al (2019) ChemCatChem 11:4537–4547.
7. Ras EJ, Rothenberg G (2014) RSC Adv 4:5963–5974.
8. Coperet C, Comas-Vives A, Larmier K, Copéret C (2017) Chem Commun 53:4296–4303.
9. Bligaard T, Bullock RM, Campbell CT, et al (2016) ACS Catal 6:2590–2602.
10. Takigawa I, Shimizu K, Tsuda K, Takakusagi S (2016) RSC Adv 6:52587–52595.
11. Ruban A, Hammer B, Norskov JK (1997) J Mol Catal A, Chem 115:421–429
12. Nørskov JK, Abild-Pedersen F, Studt F, Bligaard T (2011) Proc Natl Acad Sci U S A 108:937–
943.
13. Santen RAV, Neurock M, Shetty SG (2010) Chem Rev 110:2005–2048.