# Fileset

[64_MT-MH2022006.pdf](https://mdr.nims.go.jp/filesets/8a051411-1708-4947-b1fb-d2d522a4023a/download)

## Creator

Haruka Yoshikawa, Farid Labib, [Ya Xu](https://orcid.org/0000-0001-9067-5244), Ryuji Tamura

## Rights

[In Copyright](http://rightsstatements.org/vocab/InC/1.0/)

## Other metadata

[Catalytic Properties and Their Relation with Adsorption Energies Calculated by Density Functional Theory in Pd-Containing 1/1 Approximant Crystals](https://mdr.nims.go.jp/datasets/90852739-bbc6-4e25-b22c-ac0d1a807f78)

## Fulltext

Catalytic Properties and Their Relation with Adsorption Energies Calculated by Density Functional Theory in Pd-Containing 1/1 Approximant CrystalsCatalytic Properties and Their Relation with Adsorption Energies Calculated byDensity Functional Theory in Pd-Containing 1/1 Approximant CrystalsHaruka Yoshikawa1,2,+1, Farid Labib1, Ya Xu2,+2 and Ryuji Tamura11Department of Materials Science and Technology, Tokyo University of Science, Tokyo 125-8585, Japan2Center for Green Research on Energy and Environmental Materials, National Institute for Materials Science, Tsukuba 305-0003, JapanQuasicrystals and approximant crystals (ACs) have a unique complex structure with many crystallographically non-equivalent sites. Inorder to apply this characteristic potential to catalysts, we investigated catalytic properties of Pd-containing Tsai-type 1/1 ACs, i.e., Al–Pd–Scand Ga–Pd–Sc, in the acetylene hydrogenation reaction and also performed density functional theory calculations of adsorption energies ofreactants and products. The catalytic properties are found to significantly depend on the kind of the semimetal element such as Al and Ga, wherethe Al–Pd–Sc 1/1 AC shows higher catalytic activity and selectivity. The adsorption energy of reactant acetylene is smaller in the Al–Pd–Sc1/1 AC whereas the amount of product ethylene are comparable for both ACs. Therefore, the adsorption rate of reactants is increased whilethe desorption rate of products remains almost the same in the Al–Pd–Sc 1/1 AC. Furthermore, the adsorption energies are found to differsignificantly from site to site, implying a superior potential of ACs for designation of active sites using many non-equivalent crystallographicsites for high catalytic performance. [doi:10.2320/matertrans.MT-MH2022006](Received April 5, 2023; Accepted June 1, 2023; Published June 16, 2023)Keywords: catalyst, Tsai-type 1/1 approximant crystals, acetylene hydrogenation, density functional theory, catalytic activity, selectivity, non-equivalent crystallographic site1. IntroductionThe ligand and ensemble effects are of significantimportance for the improvement of the catalytic propertiesof metal catalysts.1–4) These effects are related to the localstructure of the surface and therefore are considered to becontrollable by the atomic arrangement. However, it is highlychallenging to design catalysts by manipulating the localstructure at the atomic level even today. In this study, weattempt to design the local structure by using metalliccompounds that have many crystallographically non-equiv-alent sites that can be substituted with other elements.Typical examples of such materials are quasicrystals andapproximant crystals (ACs), which are a group of materialscomposed of an icosahedral cluster made of concentricpolyhedral atomic shells.5,6) In quasicrystals, the icosahedralclusters are quasiperiodically arranged whereas they areperiodically arranged in ACs. For example, in 1/1 ACs, theicosahedral clusters are arranged in a body centered cubiclattice. Due to the unique complex cluster structure, thesematerials have many non-equivalent crystallographic sites.Several studies on the catalytic properties of quasicrystalsand related materials have been performed so far, and muchattention has been paid to new catalysts originating fromtheir unique structures.7–16) In this work, materials composedof the Tsai-type cluster are chosen for the investigation ofcatalytic properties: The Tsai-type cluster is a type oficosahedral cluster where rare earth elements occupy thevertices of the inner icosahedron shell among variousconcentric shells, and many compounds composed of theTsai-type cluster have been discovered in recent years.17)The objective of this study is to evaluate the catalyticproperties of Pd-containing Al–Pd–Sc and Ga–Pd–Sccompounds, which are classified as Tsai-type 1/1 ACs.18,19)and to investigate the relationship between the characteristicstructure and catalytic properties by calculating adsorptionenergies using density functional theory (DFT). For thispurpose, the C2H2 hydrogenation reaction (eq. (1)) waschosen since the reaction can be performed at relativelylow temperatures even with bulk metal and hence sampledegradation during the reaction is suppressed due to theabsence of oxygen, and, furthermore, the mechanism ofethylene hydrogenation as a side reaction is simple, makingit easy to evaluate the selectivity of the catalyst.20)C2H2 þ H2 ! C2H4 �H ¼ �173 kJ=mol ð298KÞ ð1Þ2. Experimental Procedure2.1 Sample preparationIngots weighing approximately 0.7 g of polycrystalline1/1 ACs with nominal compositions of Ga55Pd30Sc15 andAl55Pd30Sc15 were prepared from highly pure elements usingarc-melting technique under Ar atmosphere. The alloy ingotswere crushed to powder with a particle size of 75 µm or lessunder argon atmosphere as catalysts sample. The amount ofpower catalyst for reaction was approximately 0.1 g.2.2 Sample characterizationFor phase identification, X-Ray diffraction (XRD)measurement was carried out using Rigaku SmartLab SEand Rigaku MiniFlex600 diffractometers with Cu K¡.The Brunauer–Emmett–Teller (BET) specific surfaceareas measurement of the crushed powder samples wereperformed using specific-surface-area analyzer (Micromet-rics, ASAP2020). Since the metal powder has a small surfacearea, Kr was used as the adsorption gas.2.3 Catalytic performanceThe C2H2 hydrogenation reaction test was conducted ina fixed-bed flow reactor. The catalyst powder sample was+1Graduate Student, Tokyo University of Science+2Corresponding author, E-mail: XU.Ya@nims.go.jpMaterials Transactions, Vol. 64, No. 10 (2023) pp. 2425 to 2430Special Issue on Metallurgy for Advanced Catalytic Materials©2023 The Japan Institute of Metals and Materialshttps://doi.org/10.2320/matertrans.MT-MH2022006loaded into quartz reaction tube with an inner diameter of8mm and thermocouple inserted just beneath the sample tocontrol sample temperature. Prior to the reaction, the samplewas reduced under hydrogen flow (30.0mL/min) at 430°Cfor 1 h. The reaction gas was supplied from a mixture gas(2.19 vol% C2H2–N2) cylinder and a pure hydrogen gascylinder, the feed composition of C2H2, hydrogen, and N2,were 0.5, 5.0, and 23.8mL/min, respectively. The reactionwas performed by increasing the temperature from 120 to200°C at 20°C intervals, and the incubation time for eachtemperature was 30 minutes. The composition of the productsobtained during the reaction was analyzed several times ateach temperature using a online gas chromatography system(GL Science, GC323, GL Science, GC-4000 Plus). A zeoliteadsorbent column (13X molecular sieve) was used to separateH2 and N2. A DVB-EVB-Ethyleneglycol dimethacrylatecolumn (Polapak N 80/100) was used to separate C2H2, C2H4and C2H6. The total flow rate of the outlet gases wasmeasured using a flow meter (MesaLabs, Definer 220). C2H2reaction rate and C2H4 selectivity was calculated usingeq. (2) and eq. (3).C2H2 reaction rate¼ FC2H2in � FC2H2outsurface areaðmol�min�1�m�2Þ ð2ÞC2H4 selectivity ¼ FC2H4outFC2H2in � FC2H2out� 100 ð%Þ ð3ÞWhere FC2H2in, FC2H2out and FC2H2out represent the C2H2 flowrates at the reactor entrance and C2H2 and C2H4 flow rates atoutlet, respectively.To access activation energy, an additional reaction testwas carried out in a relatively low temperature range whichis suitable to obtain lower C2H2 conversions (<20%) forcalculating the activation energy of the reaction fromArrhenius plot.2.4 DFT calculationDFT calculations were performed using the first-principleselectronic structure calculation program PHASE/0.21) Self-inconsistent iterations stopped when the convergence energywas less than 2 © 10¹6 eV. The geometry of the optimizedsurface and adsorbate-substrate composite was determined bystatic relaxation. In the structure optimization and transitionstate calculations, the convergence criterion was set to 2 ©10¹6 eV for the total energy and 2 © 10¹2 eV/¡ for the forceacting on the atoms. In this study, a surface structurecontaining many local environments is preferable since weinvestigate the effect of non-equivalent sites on the catalyticproperties. In addition, the 1/1 ACs contains a very largenumber of atoms (184 atoms in Al–Pd–Sc and Ga–Pd–Sc) inthe unit cell, resulting in large slab sizes and convergencedifficulties on high index surface. For this reason, weprepared the slab models of a (100) surface which contains7 non-equivalent sites for bulk symmetry as well as all threeconstituent elements and is a low index surface based onthe result of the HAADF structure analysis of Al–Pd–Sc andGa–Pd–Sc reported by So et al.18,22) (Fig. 1(a)). The slabcontaining 188 atoms and a vacuum layer of 16¡. Since thebridge sites were reported to have a smaller C2H2 adsorptionenergy than the on-top sites in quasicrystals,12) the bridgesites were selected as the starting state for adsorption in thepresent study. Structural stabilization calculations wereperformed by setting C2H2 and C2H4 molecules at 2¡ fromthe surface on the bridge site between Al, Ga, and Pd onthe slab surface (Fig. 1(b)). Calculations were performedby setting 2 © 2 © 1 k-points, and the adsorption energiesof C2H2 (EC2H2 adsorption) and C2H4 (EC2H4 adsorption) wereevaluated using the equation eq. (4) and eq. (5),EC2H2 adsorption ¼ Etotal � Eslab � EC2H2ð4ÞEC2H4 adsorption ¼ Etotal � Eslab � EC2H4ð5Þwhere Etotal is the total energy of the slab with the absorbedC2H2 or C2H4 molecule, Eslab is the total energy of the bareslab, and EC2H2and EC2H4represent the total energy of theisolated C2H2 and C2H4 molecules, respectively.3. Results and Discussions3.1 CharacterizationFigure 2 presents powder XRD patterns of theGa55Pd30Sc15 and Al55Pd30Sc15 1/1 ACs along with theresults of Le Bail fittings23) obtained by assuming the spacegroups Im�3 using the Jana 2006 software suite.24) The redand black lines in the figure represent measured (Iobs) andcalculated (Ical) peak intensities, respectively, while thedifference between the two and the expected Bragg peakpositions are shown by blue line and green vertical bars,respectively. As shown, the experimental peak positions andtheir intensities are consistent with the calculation confirminghigh purity of the synthesized 1/1 ACs with the space groupsIm�3.Fig. 1 (a) Al–Pd–Sc and Ga–Pd–Sc 1/1 ACs (100) slab structure model based on the results of HAADF-STEM structure analysis.21,22) (b)Surface layer of the slab and adsorption points on the bridge site from 1 to 15.H. Yoshikawa, F. Labib, Y. Xu and R. Tamura2426The BET specific surface area was low level (0.09–0.11m2·g¹1) for both samples and the difference were small(Table 1). The crushed surface does not have a high specificsurface area, and it can be assumed that a surface similar tothat of the bulk is formed macroscopically.3.2 Catalytic performanceFigure 3 shows the C2H2 reaction rate (Fig. 3(a)) and C2H4selectivity (Fig. 3(b)) of the Al–Pd–Sc and Ga–Pd–Sc Tsai-type 1/1 ACs for the C2H2 hydrogenation reaction as afunction of the reaction temperature. The reaction rate ofAl–Pd–Sc sample increased significantly with over the entiretemperature range, whereas that of Ga–Pd–Sc samplegradually increased with temperature. The values of reactionrate of Al–Pd–Sc Sample were much higher than that ofGa–Pd–Sc sample at all the test temperatures. The reactionrate of Al–Pd–Sc sample was 1.63 © 10¹3mol·min¹1·m¹2(50.9% in terms of conversion rate) at 200°C, which wasapproximately 13 times of the corresponding value of Ga–Pd–Sc (1.09 © 10¹6mol·min¹1). The selectivity was higherthan 80% at all temperatures for both samples and thedifference between Al–Pd–Sc and Ga–Pd–Sc was small.These results suggest that the elemental substitution of Ga forAl was effective to enhance the catalytic activity and retain arelatively high selectivity even with increased activity.Figure 4 shows the Arrhenius plot of C2H2 reaction rateand activation energy of Al–Pd–Sc and Ga–Pd–Sc samples.The activation energy of Al–Pd–Sc sample is estimated as53.9 kJ/mol, which is lower than that of Ga–Pd–Sc (65.8kJ/mol). For comparison, the activation energies of severalPd-based catalysts for acetylene hydrogenation reportedFig. 2 Le Bail fitting of the powder x-ray diffraction (XRD) patterns of(a) Ga55Pd30Sc15 and (b) Al55Pd30Sc15 1/1 ACs. The measured (Iobs),calculated (Iobs) peak intensities, the difference between the two and theexpected Bragg peak positions are represented by red, black, blue, andgreen lines, respectively.Table 1 BET surface area of prepared catalysts sample. These samplescrushed to powder with a particle size of 75 µm or less under argonatmosphere.Fig. 3 (a) C2H2 reaction rate and (b) C2H4 selectivity of Al–Pd–Sc and Ga–Pd–Sc 1/1 ACs. The feed gas composition of C2H2, hydrogen,and N2, were 0.5, 5.0, and 23.8mL/min, respectively.Fig. 4 Arrhenius plot of C2H2 reaction rate and activation energy of Al–Pd–Sc and Ga–Pd–Sc 1/1 ACs. Approximate straight lines are shown asdotted lines, and activation energies were calculated from the slopes.Catalytic Properties of Pd-Containing 1/1 Approximants Crystals 2427previously were summarized in Table 2.2,25,26) The value ofactivation energy of Al–Pd–Sc 1/1 AC catalyst prepared inthis study is smaller than the values of the CuPd0.09/SiO2 andPdAg/K+¢¹ zeolite catalysts. Compared with Pd/SiO2 andPd foil catalysts which have been reported to exhibitexcellent activity, the activation energy of Al–Pd–Sc washigher. However, these catalysts have poor selectivity,25,27,28)whereas Al–Pd–Sc 1/1 AC combines both activity andselectivity to some extent.3.3 DFT calculationThe C2H2 and C2H4 adsorption energies of each adsorptionsites are shown in Table 3. The C2H2 adsorption energies ofAl–Pd–Sc were lower than that of Ga–Pd–Sc at each site.These results reveal that the adsorption of the reactantacetylene is more likely to occur on the (100) surface of Al–Pd–Sc than on that of Ga–Pd–Sc, which is in good agreementwith the experimental results revealing a higher activity ofAl–Pd–Sc than that of Ga–Pd–Sc. In addition, the DFTresults show that the adsorption energies differed significantlyfor each adsorption site even when the combination ofneighboring atoms was the same. Krajci and Hafner reportedthat the area around the triangle composed of two atoms ofAl or Ga and one atom of Pd is most likely to adsorb C2H2in B20 structure Al–Pd and Ga–Pd.10,11,16) However, in thisstudy, there was no particular tendency for these sites nearthis triangle to have small adsorption energies. For example,the adsorption energies at sites 11 and 13 were not small evenin the Al–Pd–Sc system.The structures of the four sites with particularly smalladsorption energies (No. 1, No. 5, No. 8, and No. 12) in thefinal state are shown in Fig. 5 and Fig. 6. The adsorptionenergies are in decreasing order as No. 8 µ No. 12 <Table 2 Calculated activation energies from Arrhenius plot of Al–Pd–Scand Ga–Pd–Sc 1/1ACs and the previously reported values of several Pd-based catalysts.2,25,26)Table 3 Adsorption energies of C2H2 and C2H4 at each site of (100) surface of Al–Pd–Sc and Ga–Pd–Sc.H. Yoshikawa, F. Labib, Y. Xu and R. Tamura2428No. 1 < No. 5. The initial sites of No. 8 and No. 12 arealmost identical in both adsorption energy and adsorptionstructure and are considered to have converged to the samefinal position. In all four initial structures, acetylene wasadsorbed on the hollow site in the relaxed structure.Furthermore, three of the sites with the lowest adsorptionenergies (No. 1, No. 8, and No. 12) adsorbed on the sitesadjacent to Sc, suggesting that the sites near Sc may adsorbacetylene more easily. The distance of the C–C bond in thefinal state was about 1.4¡ for all sites as shown in Fig. 6.This is closer to the double bond length of ethylene(1.34¡)29) than the triple bond length of acetylene(1.20¡).30) This result suggests that the reactivity differsfrom site to site in the Tsai type 1/1 ACs as we originallyassumed. In both samples, there was a large difference inadsorption energy for each site, but the local structure thatcauses particularly strong adsorption in the 1/1 ACs is notyet understood. Further calculations of the electron densitydistribution and the electronic states of the electrons involvedin binding are needed to utilize 1/1 AC structure for catalystdesign.In contrast, the adsorption energy of C2H4 did not differsignificantly between Al–Pd–Sc and Ga–Pd–Sc. This resultsuggests that the desorption rate of C2H4 on both Al–Pd–Scand Ga–Pd–Sc samples is close since the adsorption energyof C2H4 corresponds to the desorption rate of C2H4. In theC2H2 hydrogenation reaction, C2H4 adsorbed on the surfaceafter the reaction undergoes further hydrogenation reactionto form C2H6, which reduces the selectivity.31) Al–Pd–Scexhibited high activity because it easily adsorbs C2H2,whereas the adsorption energy of C2H4 did not changeobviously with element substitution of Ga by Al. Therefore,C2H4 was desorbed before the hydrogenation reactionoccurred in both Al–Pd–Sc and Ga–Pd–Sc, resulting in Al–Pd–Sc showed high C2H2 reaction rate and C2H4 selectivity.4. ConclusionThe catalytic performance in acetylene hydrogenation wasinvestigated for two kinds of Pd-containing Tsai-type 1/1ACs, Al–Pd–Sc and Ga–Pd–Sc. The Al–Pd–Sc 1/1 ACwas found to exhibit a high activity and selectivity in theacetylene hydrogenation reaction. This catalyst shows evenhigher activity compared to the previously reported Pd-basedbimetallic catalyst, suggesting that Tsai-type 1/1 ACs haveexcellent potential for catalysts.The DFT calculations of C2H2 and C2H4 adsorption energyon (100) surface of Al–Pd–Sc and Ga–Pd–Sc revealed thatthe Al–Pd–Sc 1/1 AC adsorbs reactant C2H2 more strongly,while the adsorption of product C2H4 is comparable to thatof the Ga–Pd–Sc 1/1 AC, which is likely responsible for thehigh activity and selectivity of the Al–Pd–Sc 1/1 AC. In bothACs, large difference in adsorption energy is observed amongnon-equivalent sites, which suggests superior potential todesign catalysts by controlling the local structure at theatomic level using quasicrystals and ACs.AcknowledgmentsThis work supported by New Energy and IndustrialTechnology Development Organization (NEDO), JapanSociety for the Promotion of Science through Grants-in-Aid for Scientific Research (Grants No. JP19H05817,No. JP19H05818) and JST, CREST Grant No.JPMJCR22O3, Japan. The calculations in this study wereperformed on the Numerical Materials Simulator at NIMS.H. Yoshikawa appreciates the support of the MaterialScience Human Resource Development Fellowship of TokyoUniversity of Science.REFERENCES1) I.-C. Oğuz, T. Mineva and H. Guesmi: J. Chem. Phys. 148 (2018)024701.2) Y.S. Ma, T. Diemant, J. Bansmann and R.J. Behm: Phys. Chem. Chem.Phys. 13 (2011) 10741–10754.3) D.J. Gorin, B.D. Sherry and F.D. Toste: Chem. Rev. 108 (2008) 3351–3378.4) J.R. Kitchin, J.K. Nørskov, M.A. Barteau and J.G. Chen: Phys. Rev.Lett. 93 (2004) 156801.Fig. 5 Top view of Al–Pd–Sc 1/1 AC (100) with adsorbed C2H2 afterstructural relaxation started from various initial adsorption sites (a) No. 1,(b) No. 5, (c) No. 8, (d) No. 12. C–C bond and C–H bond were markedwith green and blue lines, respectively.Fig. 6 Side view of Al–Pd–Sc 1/1 AC (100) with adsorbed C2H2 afterstructural relaxation started from various initial adsorption sites (a) No. 1,(b) No. 5, (c) No. 8, (d) No. 12. C–C bond and C–H bond were markedwith green and blue lines, respectively.Catalytic Properties of Pd-Containing 1/1 Approximants Crystals 2429https://doi.org/10.1063/1.5007247https://doi.org/10.1063/1.5007247https://doi.org/10.1039/c1cp00009hhttps://doi.org/10.1039/c1cp00009hhttps://doi.org/10.1021/cr068430ghttps://doi.org/10.1021/cr068430ghttps://doi.org/10.1103/PhysRevLett.93.156801https://doi.org/10.1103/PhysRevLett.93.1568015) S. Takeuchi, K. Edagawa, A.P. Tsai and K. Kimura: Physics ofQuasicrystals, (Asakura-Syoten, Tokyo, 2012) pp. 30–68.6) A.P. Tsai and A. Yamamoto: Philos. Mag. 87 (2007) 2599–2600.7) K. Abe, R. Tsukuda, N. Fujita and S. Kameoka: RSC Adv. 11 (2021)15296–15300.8) M. Krajčí and J. Hafner: ChemCatChem 8 (2016) 34–48.9) M. Krajčí, A.-P. Tsai and J. Hafner: J. Catal. 330 (2015) 6–18.10) M. Krajčí and J. Hafner: J. Phys. Chem. C 116 (2012) 6307–6319.11) M. Krajčí and J. Hafner: J. Catal. 295 (2012) 70–80.12) M. Krajčí and J. Hafner: J. Catal. 278 (2011) 200–207.13) M. Krajčí and J. Hafner: Phys. Rev. B 84 (2011) 115410.14) M. Yoshimura and A.P. Tsai: J. Alloy. Compd. 342 (2002) 451–454.15) A.P. Tsai and M. Yoshimura: Appl. Catal. A Gen. 214 (2001) 237–241.16) M. Krajčí and J. Hafner: Phys. Rev. B 87 (2013) 035436.17) A.-P. Tsai: J. Non-Cryst. Solids 334­335 (2004) 317–322.18) Y.G. So, A. Katagiri, R. Tamura and K. Edagawa: Philos. Mag. Lett. 98(2018) 292–300.19) Y.G. So and K. Edagawa: Mater. Trans. 50 (2009) 948–951.20) A.J. McCue, C.J. McRitchie, A.M. Shepherd and J.A. Anderson:J. Catal. 319 (2014) 127–135.21) T. Yamasaki, A. Kuroda, T. Kato, J. Nara, J. Koga, T. Uda, K. Minamiand T. Ohno: Comput. Phys. Commun. 244 (2019) 264–276.22) Y.G. So, M. Nagao, T. Nagai and K. Kimoto: J. Alloy. Compd. 543(2012) 7–11.23) A. Le Bail, H. Duroy and J.L. Fourquet: Mater. Res. Bull. 23 (1988)447–452.24) V. Petříček, M. Dušek and L. Palatinus: Z. Krist.-Cryst. Mater. 229(2014) 345–352.25) M.R. Ball, K.R. Rivera-Dones, E.B. Gilcher, S.F. Ausman, C.W.Hullfish, E.A. Lebron and J.A. Dumesic: ACS Catal. 10 (2020) 8567–8581.26) H. Molero, B.F. Bartlett and W.T. Tysoe: J. Catal. 181 (1999) 49–56.27) J. Osswald, K. Kovnir, M. Armbruster, R. Giedigleit, R.E. Jentoft, U.Wild, Y. Grin and R. Schlogl: J. Catal. 258 (2008) 219–227.28) J. Osswald, R. Giedigkeit, R.E. Jentoft, M. Armbruster, F. Girgsdies,K. Kovnir, T. Ressler, Y. Grin and R. Schlogl: J. Catal. 258 (2008) 210–218.29) W. Majer, P. Lutzman and W. Hüttner: Mol. Phys. 83 (1994) 567–578.30) D.J. Gearhart, J.F. Harrison and K.L.C. Hunt: Int. J. Quantum Chem.95 (2003) 697–705.31) D.H. Mei, M. Neurock and C.M. Smith: J. Catal. 268 (2009) 181–195.H. Yoshikawa, F. Labib, Y. Xu and R. Tamura2430https://doi.org/10.1080/14786430701397945https://doi.org/10.1039/D1RA01958Ahttps://doi.org/10.1039/D1RA01958Ahttps://doi.org/10.1002/cctc.201500733https://doi.org/10.1016/j.jcat.2015.06.020https://doi.org/10.1021/jp212317uhttps://doi.org/10.1016/j.jcat.2012.07.025https://doi.org/10.1016/j.jcat.2010.12.004https://doi.org/10.1103/PhysRevB.84.115410https://doi.org/10.1016/S0925-8388(02)00274-8https://doi.org/10.1016/S0926-860X(01)00500-2https://doi.org/10.1103/PhysRevB.87.035436https://doi.org/10.1016/j.jnoncrysol.2003.11.065https://doi.org/10.1080/09500839.2018.1538578https://doi.org/10.1080/09500839.2018.1538578https://doi.org/10.2320/matertrans.MC200827https://doi.org/10.1016/j.jcat.2014.08.016https://doi.org/10.1016/j.cpc.2019.04.008https://doi.org/10.1016/j.jallcom.2012.07.119https://doi.org/10.1016/j.jallcom.2012.07.119https://doi.org/10.1016/0025-5408(88)90019-0https://doi.org/10.1016/0025-5408(88)90019-0https://doi.org/10.1515/zkri-2014-1737https://doi.org/10.1515/zkri-2014-1737https://doi.org/10.1021/acscatal.0c01536https://doi.org/10.1021/acscatal.0c01536https://doi.org/10.1006/jcat.1998.2294https://doi.org/10.1016/j.jcat.2008.06.014https://doi.org/10.1016/j.jcat.2008.06.013https://doi.org/10.1016/j.jcat.2008.06.013https://doi.org/10.1080/00268979400101431https://doi.org/10.1002/qua.10586https://doi.org/10.1002/qua.10586https://doi.org/10.1016/j.jcat.2009.09.004