hide
Free keywords:
-
Abstract:
In the context of the impending oil shortage in the near future, it is vital to find an economical and environmental-friendly route to produce alternative building blocks for chemical industry, such as propene and acrylic acid. Many attempts have been done in the past decades in order to implement an efficient and suitable catalyst for alkane oxidation. However, no real breakthrough has beenachieved in the catalytic community since the discovery of MoVTeNb M1 phase-pure catalyst patented by Mitsubishi Corp. in the early 1990s. The lack of homogeneity of data presented in the literature, regarding the many proposed catalytic systems, addresses the new oxidation catalysis research back to fundamental roots, since only a deep understanding and a systematic study of the intrinsic nature of the catalyst may constitute the key for the future. To this latter extent, the implementation of a catalyst requires the study of relative simple and flexible system, such as perovskites, since these bulk oxide catalysts present a high level of flexibility without destroying their structure. Mathematical tools, such as high-throughput DFT calculations, machine learning (ML) and artificial intelligence, have already been widely employed for accelerating the discovery of possible new perovskite-like structures [1-5] and for searching for key-descriptors for the catalytic performance. The present project deals with the synthesis, characterization and catalytic testing of twenty-three phase-pure perovskite-like catalysts (two of them are 97% phase-pure perovskite-like catalysts) in order to serve as basis for artificial intelligence analysis. For this purpose, Mn has been chosen as B-site cation combined with an other metal belonging to the 3d or 4d row of the periodic table (the synergistic effect is claimed to be beneficial to the catalytic performance). The selected element at the B-site have similar ionic radius among each other (ionic radii in six coordination: Co2+ = 0.65 Å, Zn2+= 0.74 Å, Ni2+=0.69 Å, Fe2+=0.61 Å, Cr2+=0.73 Å, Pd2+=0.86 Å, Cu2+= 0.73 Å and Mn3+/4+=0.53-0.64 Å). The A-site cation has been changed between La3+, Pr3+, Nd3+ and Sm3+ to tune the structural and electronic properties. Therefore, twenty-one purely phase-pure perovskites, among which fourteen present the general formula (La,Pr)Mn(1-x)CuxO3 (with x=0-0.4) and other five the general formula AMn0.7B’0.3O3 (with A=Pr, Nd and Sm and B’=Co, Zn, Ni, Fe, Cr) and (La,Pr)2CuO4 perovskite-related structures have been synthesized, tested in CO and propane oxidation and subjected to different characterization methods in order to obtain numerical values of some properties (also known as basic descriptors) for artificial intelligence analysis. As first step, the (La,Pr)Mn(1-x)CuxO3 series has been investigated in detail. From the combined analysis of the characterization data with those of catalysis, it emerges the high complexity generated by the synergistic effect of a double B-site into the catalytic dynamic. Furthermore, the catalytic results highlight that also the A-site indirectly influences the catalytic scenario, in particular in CO oxidation. This is in contrast to what is generally reported in the literature, where it is claimed that the B-site is the responsible element for the catalytic performance. The Artificial Intelligence analysis performed over the (La,Pr)Mn(1-x)CuxO3 series have highlighted how a single primary feature based on chemical intuition is not sufficient to fully describe the properties linked to the catalytic performance in CO oxidation. Additionally, the found optimal model has confirmed the hypothesis and expectations formulated for the (La,Pr)-based series, where the A-site is indirectly involved in the catalytic performance in CO oxidation. Also the AI performed over the whole perovskite matrix ((La,Pr)Mn(1-x)CuxO3 series + AMn0.7B’0.3O3 series) confirms the results discussed for the (La,Pr)-series in CO oxidation. Many different conclusions can be drawn from the AI analysis results of the data deriving from propane oxidation. For instance, they highlight how the addition of steam into the reaction feed could affect the catalytic scenario to different extents depending on the involved samples. A detailed summary and the final overview of the herein presented project are reported in the conclusion part of this work.
