Invited: Understanding the Photoinduced Reaction Dynamics of CO on Metal Surfaces with Neural Networks Potentials
ORAL · Invited
Abstract
Modelling the dynamics and reactivity of adsorbates on metals that are induced by near IR femtosecond laser pulses requires including the effect of the laser-excited electrons and, in many cases, also the effect of the highly excited surface lattice. Ab initio molecular dynamics with electronic friction and thermostats, (Te,Tl)-AIMDEF, do enable such a complex modelling [1,2]. However, these simulations come with a very large computational cost that severely limits statistics and propagation times. It is at this point that the use of multidimensional potential energy surfaces (PESs) based on the atomistic neural network (atNN) approach [3] is becoming the accurate alternative to ab initio molecular dynamics studies of diverse gas-surface reactions, including the mentioned photoinduced reactions that require a precise description of adsorbate-substrate and interadsorbate interactions under a highly excited and very changing environment.
Here, I will discuss different examples of accurate NN-PESs that we have constructed using the embedded atom neural network (EANN) method [4], which we find to be impressively accurate and flexible to account for all the necessities required to deal with photoinduced reactions [5]. Among other things, neural networks allowed us to understand the strong coverage dependence found experimentally in CO/Pd(111) [6], the large branching ratio between CO desorption and CO oxidation in Ru(0001) [7], and reveal the dynamical nature of the CO physisorption well that so far was only found in XPS experiments.
[1] M. Alducin, N. Camillone, S.-Y. Hong, J. I. Juaristi, Phys. Rev. Lett., 123, 246802 (2019).
[2] A. Tetenoire et al. J. Phys. Chem Lett. 13, 8516 (2022).
[3] J. Behler and M. Parrinello, Phys. Rev. Lett., 98, 146401 (2007).
[4] Y. Zhang, C. Hu, B. Jiang, J. Phys. Chem. Lett. Journal, 10, 4962 (2019).
[5] A. Serrano Jiménez et al., J. Chem. Theory Comput. 17, 4648 (2021).
[6] A. S. Muzas et al., J. Phys. Chem Lett. 15, 2587 (2024).
[7] I. Žugec et al., J. Amer. Chem. Soc. Au, 4, 1997 (2024).
Here, I will discuss different examples of accurate NN-PESs that we have constructed using the embedded atom neural network (EANN) method [4], which we find to be impressively accurate and flexible to account for all the necessities required to deal with photoinduced reactions [5]. Among other things, neural networks allowed us to understand the strong coverage dependence found experimentally in CO/Pd(111) [6], the large branching ratio between CO desorption and CO oxidation in Ru(0001) [7], and reveal the dynamical nature of the CO physisorption well that so far was only found in XPS experiments.
[1] M. Alducin, N. Camillone, S.-Y. Hong, J. I. Juaristi, Phys. Rev. Lett., 123, 246802 (2019).
[2] A. Tetenoire et al. J. Phys. Chem Lett. 13, 8516 (2022).
[3] J. Behler and M. Parrinello, Phys. Rev. Lett., 98, 146401 (2007).
[4] Y. Zhang, C. Hu, B. Jiang, J. Phys. Chem. Lett. Journal, 10, 4962 (2019).
[5] A. Serrano Jiménez et al., J. Chem. Theory Comput. 17, 4648 (2021).
[6] A. S. Muzas et al., J. Phys. Chem Lett. 15, 2587 (2024).
[7] I. Žugec et al., J. Amer. Chem. Soc. Au, 4, 1997 (2024).
–
Publication: - M. Alducin, N. Camillone, S.-Y. Hong, J. I. Juaristi, Phys. Rev. Lett., 123, 246802 (2019).<br>- A. Tetenoire et al. J. Phys. Chem Lett. 13, 8516 (2022).<br>- A. Serrano Jiménez et al., J. Chem. Theory Comput. 17, 4648 (2021).<br>- A. S. Muzas et al., J. Phys. Chem Lett. 15, 2587 (2024).<br>- I. Žugec et al., J. Amer. Chem. Soc. Au, 4, 1997 (2024).
Presenters
-
Maite Alducin
Centro de Física de Materiales CFM, Donostia-San Sebastian
Authors
-
Maite Alducin
Centro de Física de Materiales CFM, Donostia-San Sebastian