1 DGMK liagungsbericht & 9803

Proceedings of the DGMK-Conference “Selective Oxidations in Petrochemistry" October 8-9,1998, Hamburg-Reinbek, Germany

f

DGMK

German Society for Petroleum and Coal Science and Technology

Petrochemistry Division Alle Rechte, auch die der Ubersetzung, des auszugsweisen Nachdrucks, der Herstellung von Mikrofilmen und der fotomechanischen Wiedergabe, nur mit ausdrucklicher schriftlicher Genehmigung der DGMK.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of DGMK.

Als Manuskript gedruckt.

ISSN 1433-9013 ISBN 3-931850-44-7

Preis: DM 150,- zuziigl. ges. MwSt.

Verbreitung und Verkauf nur durch: /s DGMK

Deutsche Wissenschaftliche Gesellschaft fur Erdol, Erdgas und Kohle e.V.

Kapstadtring 2, 22297 Hamburg Postfach 60 05 49, 22205 Hamburg Telefon: (040) 63 90 04-11/33 Telefax: (040) 63 00 736 Bankverbindung: Dresdner Bank AG, Hamburg, Konto 9 085 166 00 (BLZ 200 800 00) Amtsgericht Hamburg 69 VR 6898 DISCLAIMER

Portions of this document may be illegible in electronic image products. Images are produced from the best available original document. /N DGMK

German Society for Petroleum and Coal Science and Technology

Tagungsbericht 9803

Proceedings of the DGMK-Conference “Selective Oxidations in Petrochemistry" October 8-9,1998, Hamburg-Reinbek, Germany

(Authors Manuscripts)

edited by G. Emig, C. Kohlpaintner & B. Liicke

DE99G2168

CONTENTS

Page

NEW VISTAS IN SELECTIVE OXIDATIONS R. Sheldon

SELECTIVE CATALYTIC OXIDATIONS OF ALKYLAROMATIC COMPOUNDS R.W. Fischer and F. Rohrscheid

PROCESS ENGINEERING IN SELECTIVE ALKANE OXIDATIONS 27 J.J. Lerou

PARTIAL OXIDATION OF n-AND i-PENTANE OVER PROMOTED VANADIUM-PHOSPHORUS OXIDE CATALYSTS V.A. Zazhigalov, B.D. Mikhajluk and G.A. Komashko

ACTIVE CATALYTIC SITES IN THE AMMOXIDATION OF (37) PROPANE AND PROPENE OVER V-Sb-0 CATALYSTS S. A. Buchholz and H. W. Zanthoff

A NEW KINETIC MODEL BASED ON THE REMOTE CONTROL /4s) MECHANISM TO FIT EXPERIMENTAL DATA IN THE SELECTIVE OXIDATION OF PROPENE INTO ACROLEIN ON BIPHASIC CATALYSTS H. M. Abdeldayem, P. Ruiz, F. C. Thyrion and B. Delmon

THE LIQUID PHASE OXIDATION OF n-BUTANE: A SEARCH FOR PLAUSIBLE MECHANISMS C.C. Hobbs

MODELLING OF THE PARTIAL OXIDATION OF a,(3- UNSATURATED ALDEHYDES ON Mo-V-OXIDES BASED CATALYSTS H. Bohnke, J.C. Petzoldt, B. Stein, C. Weimer and J. W. Gaube

THE MORPHOLOGICAL MODIFICATION OF ELECTROLYTIC SILVER DURING THE OCM REACTION AND IT'S EFFECT ON 0 CATALYSIS A. J. Nagy, G. Weinberg, E. Kitzelmann and G. Mestl

STRATEGIES FOR CATALYST DEVELOPMENT: POSSIBILITIES © OF THE “RATIONAL APPROACH” ILLUSTRATED WITH PARTIAL OXIDATION REACTIONS W. Weiss, Th. Schedel-Niedrig and R. Schlogl II

Page

A PREDICTIVE TOOL FOR SELECTIVE OXIDATION OF HYDROCARBONS: OPTICAL BASICITY OF CATALYSTS P. Moriceau, A. Lebouteiller, E. Bordes and P. Courtine

COMPARISON OF THE REDOX ACTIVITIES OF SOL-GEL AND 103 ) CONVENTIONALLY PREPARED Bi-Mo-Ti MIXED OXIDES M. Wildberger, J.-D. Grundwaldt, T. Mallat and A. Baiker

CATALYTIC OXIDATIVE CONVERSION OF ALKANES TO OLEFINS AND OXYGENATES M. Baerns

HETEROGENEOUS CATALYTIC OXIDATIVE DEHYDROGENATION OF ETHYLBENZENE TO STYRENE WITH CARBON DIOXIDE T. Badstube, H. Papp, P. Kustrowski and R. Dziembaj

V2O5-ZO2 CATALYSTS FOR THE OXIDATIVE @ DEHYDROGENATION OF PROPANE - INFLUENCE OF THE NIOBIUM OXIDE DOPING S. Albrecht, K.-H. Hallmeier, G. Wendt and G. Lippold

POSTERSESSION

OXIDATIVE DEHYDROGENATION OF ETHANE ON RARE-EARTH OXIDE-BASED CATALYSTS O. Bayevskaya and M. Baerns

SELECTIVE EPOXIDATION OF ALLYLIC ALCOHOLS WITH A TITANIA-SILICA AEROGEL M. Dusi, T. Mallat and A. Baiker

THE SILVER CATALYST PROCESS FOR CONVERTING METHANOL TO FORMALDEYHDE - KINETIC INVESTIGATIONS E. Panzer and G. Emig

PHENOL BY DIRECT HYDROXYLATION OF BENZENE WITH \163 NITROUS OXIDE - ROLE OF SURFACE OXYGEN SPECIES IN THE REACTION PATHWAYS A. Reitzmann, E. Klemm, G. Emig, S. A. Buchholz and H. W. Zanthoff

PARTIAL OXIDATION OF N-BUTANE TO MALEIC ANHYDRIDE 171 OVER A VANADIUMPYROPHOSPHATE CATALYST IN THE RISER REGENERATOR SYSTEM St. HeB, M. Liauw and G. Emig Page

OXIDATION OF AROMATIC ALCOHOLS ON ZEOLITE- ( 173 ] ENCAPSULATED COPPER AMINO ACID COMPLEXES S. Ernst and J. M. Teixeira Florencio a-Sb204- Induced Improvements of the catalytic behavior of MoO,-(010) in the oxygen-assisted dehydration of 2-butanol: 3 Implications in selective oxidation E. M. Gaigneaux, M.-L. Naeye, O. Dupont, M. Gallant, B. Kartheuser, P. Ruiz and B. Delmon

SELECTIVE OXIDATIONS ON VANADIUMOXIDE CONTAINING AMORPHOUS MIXED OXIDES (AMM-V) WITH TERT.- BUTYLHYDROPEROXIDE Y. Deng, M. Hunnius, S. Storck and W. F. Maier

ON THE CATALYTIC GAS PHASE OXIDATION OF BUTADIENE Z) TO FURAN B. Kubias, U. Rodemerck, F. Ritschl and M. Meisel

SUPERCRITICAL CARBON DIOXIDE AS AN INNOVATIVE REACTION MEDIUM FOR SELECTIVE OXIDATION F. Loeker and W. Leitner

CHARACTERIZATION OF VPO AMMOXIDATION CATALYSTS BY IN SITU METHODS © A. Martin, B. Lucke, A. BrQckner, U. Steinike, K.-W. Brzezinka and M. Meisel

PARTIAL OXIDATION OF 2-PROPANOL ON PEROVSKITES R. Sumathi, B. Viswanathan and T. K. Varadarajan

CHEMO-ENZYMATIC EPOXIDATION OF OLEFINS BY I CARBOXYLIC ACID ESTERS AND HYDROGEN PEROXIDE M. RQsch gen. Klaas and S. Warwel I STUDY OF PROPANE PARTIAL OXIDATION ON VANADIUM- 241, CONTAINING CATALYSTS f G. A. Komashko, S. V. Khalamejda and V. A. Zazhigalov

ACTIVE GROUPS FOR OXIDATIVE ACTIVATION OF C-H BOND IN 249/ C2-C5 PARAFFINS ON V-P-0 CATALYSTS V. A. Zazhigalov

ADSORPTION AND REACTION ON A POLAR SINGLE 257 CRYSTALLINE CHROMIA SURFACE O. Seiferth, M. Bender, B. Dillmann, D. Ehrlich, I. Hemmerich, F. Rohr, K. Wolter, C. Xu, H. Kuhlenbeck and H.-J. Freund

s

7

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

R. Sheldon Delft University of Technology, Faculty of Chemical Technology, P.O. Box 5045, NL-2600 GA Delft, The Netherlands

NEW VISTAS IN SELECTIVE OXIDATIONS

Manuscript was not available

DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 8 9

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*DE012198096* R. W. Fischer"’, F. Rohrscheid" ’ Celanese GmbH, P.O. Box 13 01 60 D-46147 Oberhausen, Germany Central Research & Technology, Hoechst AG, Frankfurt a. M., Germany

SELECTIVE CATALYTIC OXIDATIONS OF ALKYLAROMATIC COMPOUNDS

Focused to the guidelines of ‘Sustainable Development* ‘Responsible Care' and Customer Satisfaction ’, modern production processes are critically assessed on their balance between their ecological benefits and their economical parameters as well as their value to the community. Also in the area of fine chemicals, it is obvious that more and more processes are devolved which save feedstock, reduce emissions and minimize the potential for safety hazards: Less additive but more integrated protection of the environment yielding ecologically highly valuable processes. The described production of aromatic carboxylic acids is an ideal example for such a modern process. Nowadays the synthesis of derivatives of benzoic acid utilizes air as Ideal oxidant and acetic acid as environmental unquestionable solvent. The major byproduct of the oxidation reaction is water in some cases, dependend on the substrate also carbon dioxidej

Major advantages of the process are • High environmental acceptance • low manufacturing costs • minimization of risks and potential hazards • high flexibility in product change • high space time yields and thus high capacities realized in small operating units • high quality of the products in connection with easy work up.

Aromatic Carboxylic Acids are highly important, synthetically useful fine chemicals and building blocks for different types of polymers. The industrial importance of terephthalic acid or DMT (dimethyl terephthalate), both building blocks for PET, and phthalic acid anhydride (PTA) is obvious. PET is a basic monomer for polyester polymers used for the production of end-user products such as bottles, video tapes or fine fashions, as well as environmentally beneficial packaging materials. Aromatic polyamides (aramides) were the first in the series of high-performance specialty thermoplastics. Such polymers named Kevlar® are used in bulletproof vests or the first commercially available LCPs. Nomex® is used as substitute for asbestos. In the area of high performance polymers there is a still growing market interest in sophisticated building blocks bearing multiple carboxylic functionality's on aromatic systems, the latter often being connected by fluorinated spacer groups as in the high price, electronic grade polyimide monomers 2,2- bis(3',4'-anhydrodicarboxyphenyl)hexaflouropropane, both produced via oxidation of the corresponding o-xylene precursors. —------

DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 DE99G2093 10

As fine chemicals they are mostly used as intermediates to pharmaceuticals, agrochemicals or pigments. Here, the current market volume is in the range of 16 kto/a corresponding to ca. 220 Mio DM. The obvious commercial importance of the further developed Amoco-MC technology has thus stimulated an enormous amount of research activity which has resulted in more than 350 different types of aromatic substrates that have been oxidized using this method. A review of the literature since the mid 1980s concerning the oxidation of alkyl-, acyl-, or formyl-substituted aromatic compounds with air as oxidant and transition metal salts as catalysts reveals trends which may be tentatively summarized as follows: On the basis of production volume and number of patents, the main target is the development of new or improved routes to polycarboxylic acids such as terephthalic, phthalic and isophthalic acid and 2,6-naphthaline dicarboxylic acid, formed by the direct oxidation of methyl, ethyl or isopropyl substituted precursors. There is an increasing effort in the field of high price polyimid monomers bearing four carboxy groups in one molecule. Besides the development of new polymer feedstock, there are several targets in the area of fine chemicals: thus all kinds of o-nitro configured systems - especially nitro- substituted derivatives, are highly interesting. However they are difficult to synthesize if no nitric acid-supported reaction is applied. Halogen- and alkyl-, alkylamino- and aminosulfonyl-substituted benzoic acids are of increasing importance for pharmaceuticals and agrochemicals often bearing comparable substitution patterns. Polyhalogenated aromatic acids are important targets, due to their various substitution possibilities.

According to the presented slides (see selection in the attachment below) the given lecture will cover the principals of oxidation of aromatic compounds with air via radical mechanisms mediated by transition metal catalysts. The broadness of the process and its general applicability will be outlined. An insight on mechanistic pathways and general parameters concerning the choice of the right catalyst for special substrates as well futre aspects of this type of oxidation technology will be given.

The last part of the lecture will discuss new possibilities for the synthesis of aromatic quinones. A summary is given on the next page. 11

Oxidation of arenes to Quinones: Besides the catalytic oxidation of alkyl side chains of aromatic systems, based on radical reaction mechanisms, the direct oxidation of the arene system to yield hydroxy substituted aromatic compounds or - after full oxidation - quinones is the second type of important catalytic oxidation of arene systems. Applied as a novel catalyst system alkylrhenium oxides turned out to be very active and selective by the formation of quinones. At comparable mild reaction conditions, (acetic acid/HaCVCHaReOa, 20 - 40 “C), naphthalene derivatives like 2-methyl-naphthalene are oxidized with high chemo- and especially regio-selectivity (85-98%) preferentially to the 1,4-quinone, in the case of 2-methylnaphthalene to menadione (vitamin Ka). o CHl [CHjReOjl

Ac0H/H202

O

This reaction can be smoothly applied to a broad variety of alkyl-susbtituted benzene and naphthalene derivatives: 2,3-dimethyl-, 2,5-dimethyl-, or 2,3,5,8- tetramethylnaphthalene and 1 -hydroxy-2-methylnaphthalene are converted by MTO/H2O2 in good to excellent yields (60 - 100 %). The high regioselectivity of the MTO/H2O2 system is particularly noteworthy: in the industrial synthesis of menadione with CrOa, producing 18 kg of chromium waste per kg of product, selectivity of only 40 - 60 % are reported. Thus the method described allows a novel economic as well as ecologically synthesis of quinone derivatives preferably starting, from electron rich aromatic systems.

The solvent of choice for the reaction and the work up procedure is concentrated acetic acid. Mixtures of acetic acid and acetic acid anhydride are advantageous concerning catalyst efficiency and life time. Similarly to alkyl-substituted rhenium oxides, their simple inorganic counterparts Re2C>7 and Re03 work also as selective catalysts in the transformation of arenes to the corresponding hydroxy compounds or quinones. This turns out to be advantageous, because such oxides are more easily accessible than the alkylrheniumoxides. Electron poorer systems like benzene or toluene are oxidized to the corresponding phenols under certain co-catalyst conditions. As discussed above, also simple inorganic rhenium oxides such as Re207 and Re03 become active in oxidation chemistry upon treatment with concentrated hydrogen peroxide. The catalytic active species are described as side-on bound bis(peroxo) complexes [H4Re20i3] and [HOReOfCbhhbO] respectively.

\ O 12

References:

[1] R. W. Fischer in 'Applied Homogeneous Catalysis with Organometallic Compounds' (editors: B. Comils, W. A. Herrmann), p. 430, VCH, 1996. [2] R. W. Fischer, F. Rohrscheid in ‘Applied Homogeneous Catalysis with Organometallic Compounds' (editors: B. Cornils, W. A. Herrmann), p. 439, VCH, 1996. Selective & Catalytic Oxidation of Alkyl Aromatic Compounds

Celanese GmbH

Dr. Richard W. Fischer Introduction Challenge of Oxidizing Aromatic Compounds: • oxidize as selective as possible • in high yields and conversion • without using expansive or questionable oxidants • implementing integrated protection of the environment • with high margins of operating profit

Oxidation of CH beeing part of the aromatic system to yield phenols, naphthols or quinones i

Oxidation of the alkyl group to yield: aryl alcohols, aldehydes & acids Oxidation of Alkylaromatic Compounds:

One of the largest industrial-scale application of homogeneous catalysis is represented by the oxidation of p-xylene to terephthalic acid or its esters as basic monomer for polyesters:

Co/Mn/Br acetic acid 190 - 205 °C 15 - 30 bar

AMOCO MC Process Oxidation of Alkylaromatic Compounds:

Benzoic acid derivatives produced via oxidation of toluenes using air as oxidant:

- /7-nitro benzoic acid - halogenated benzoic acids (chloro-, dichloro derivatives, bromo, fluoro-derivatives) - special benzoic acids as polyimide monomers or as intermediates for agrochemicals - terephthalic acid - naphthaline dicarboxylic acids

Most of these acids precipitate in high purity from the reaction media (glacial acetic acid). The mother layer, saturated with product, is recycled up to 40 times until it is exchanged completely. Substituted Benzoic Acid Derivative Market, 1997 Market Volume in t/a (volume: 15.3 kt/a, 204 Mio DM

Di-Cl-BA, 600 t/a

Cl-Nitro-BA, 2000 t/a

F-BA, 450 t/a

Chloro-derivatives 600 t/a

t.-Butyl-i-PT, 200 t/a

4-Nitro-BA, 5800 t/a

4-t.-Butyl-BA, 2500 t/a Oxidation of Alkylaromatic Compounds: Catalysts, Solvents, Reaction Conditions & General Processing

Catalysts: First and rate determine step: formation of bencylic radicals

CH

Catalyst metals: Co, Mn, Ce, Hf, Mo, Ni, Pd, Ti, V, Zr Co-Catalyst: HBr active species: MII’III[Br(02CH3)12] Oxidation of Alkylaromatic Compounds: Scope & Limitations: Catalyst effects

[Co/Mn/Br]

s 100-250 °C 5-30 bar air acetic acid

[Co (OAc)2]

S Product selectivity is cleanly steered by catalyst composition. Oxidation of Alkylaromatic Compounds: Catalyst effects

co 2h o, < 10 %

Co 3*, MEK Co 2* 0 'CH,

> 90 %

electron transfer mechanism via the co-oxidant MEK Oxidation of Alkylaromatic Compounds:

Solvent of Choice: Acetic Acid

- high solubility for water, the catalyst and for a broad variety of substrates - very low solubility for most of the oxidation products - active role in the catalytic cycle as ligand - perfect medium for catalyst recycle - low solvent oxidation, major byproducts: H20 & C02 - explosion limits can be handled easily - ecologically unquestionable solvent - usable over a wide temperature range - co-oxidation with ethenal or methyl-ethyl ketone yields solvent Oxidation of Alkylaromatic Compounds: Ligand effects

Total radical consumption Side Reactions Complete Conversion no yield of oxidation product yield 30 - 70 % yield 90 - 98 %

-I (<>. m, p) -OCOCI-I3 (m, p) -F, -Cl, -Br (0, 111, p) l 5 O Q

-Oil (o, in, p) p ' -OCH3 (p) w

-NM2 (

-OPh (o) -C02H (0) -N02 (/», P)

-co 2i-i -N02 (e>) -Ph (0) (/”, p)

-S02NH2 (0) -Ph (m, p) i - no substrate conversion, -C(=0)Ph (0, in, p) i also when traces of such higher yields if ■ derivatives are present co-oxidation is -S02NI-I2(-R2 ) ("?, P) - ortho effect applied -S02R(-Ph) (in, p)

-P(=0)R2 (0, 111, p) Oxidation of Alkylaromatic Compounds

Typical Oxidation Parameters:

Catalyst concentration: 1.5 - 10 mmol / mol substrate Substrate concentration: 30 - 40 % T 160 - 220 °C P 20 - 30 bar

U > 99.5 % S >95%

0X1STY max* • 1.8 kgI/%-1 Oxidation of Alkylaromatic Compounds: Co/Mn-ratio i

Limits of Catalyst Composition

8

s 100

ratio Mn/Co Subtrate Conversion Formation of Intermediates Limits Oxidation 90-

of

Catalyst

of Limits

Alkylaromatic Co+Mn

of

Catalyst Concentration

[mmol/mol]

Concentration

Compounds: — f

Oxidation of Alkylaromatic Compounds: Future Trends

Selective and efficient oxidation of o-nitro toluenes • development of sophisticated co-oxidation techniques

Regeneration of used, inactive catalyst salts • development of efficient separation & recycle techniques

New catalyst systems • Further development of catalysts performing with increased selectivity and increased reactivity • Development of heterogeneous catalysts • Investigation of ligand modified catalyst systems 27

i DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998 |

J. J. Lerou DuPont Nylon Intermediates R & D, Sabine River Laboratory, P.O. Box 1089, Orange, TX 77630-1089, USA

PROCESS ENGINEERING IN SELECTIVE ALKANE OXIDATIONS

Manuscript was not available

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 j 28 29

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*DE012198102* V. A. Zazhigalov, B. D. Mikhajluk, G. A. Komashko _____ Ukrainian-Polish Laboratory of Catalysis, Institute of Physical Chemistry, National Academy of Sciences of Ukraine, pr. Nauki 31, Kyiv-22, 252022, Ukraine

PARTIAL OXIDATION OF N-AND l-PENTANE OVER PROMOTED VANADIUM- PHOSPHORUS OXIDE CATALYSTS

V It is known, that the cost of raw materials for catalytic oxidation processes is about 60 % of the products price. Cheap initial compounds to produce variety of products and to replace olefins and aromatic hydrocarbons are paraffins. That is why catalytic systems which could be possibly rather efficient in selective oxidation of paraffin hydrocarbons are under very close investigation now. One of such processes is n-pentane oxidation. The obtained results [4-2J" on n- pentane oxidation over VPO catalysts were quite encouraging in respect of possible reach high selectivity and yield of phthalic anhydride. However, in our workJ3] it was shown that the main product of n-pentane oxidation in the presence of VPO catalytic system as well as VPMeO [^was maleic anhydride. Some later our results were confirmed in where to grow the selectivity towards phthalic anhydride the Co ­ additive was introduced. On the basis of the proposal made before on the mechanism of paraffins conversion over the vanadyl pyrophosphate surface 5&P[ with their activation at the first and fourth carbon atoms, we assumed possible methylmaleic (citraconic) anhydride forming at n- and i-pentane oxidation. This assumption has been recently supported by both our [3^8], and other researchers’ .[9,4 Of experimental results. In \Pf it was also hypothized possible mechanistic features for phthalic anhydride forming from n-pentane. The present work deals with the results of n- and i-pentane oxidation over VPO catalysts promoted with Bi.'Cs, Te, Zivj

Experimental The catalysts were prepared by the procedure described earlier [11] in organic medium. Additives of the elements were introduced directly into the reaction mixture simultaneously with V2O5. Atomic ratios of the components in the prepared samples were following: PN = 1.15; BiN = 0.05, 0.075, 0.10, 0.15, 0.20, 0.30; Cs/V = 0.05, 0.10; TeAZ = 0.05, 0.10; ZrAZ = 0.05, 0.10, 0.20, 0.30. Study of the catalysts by XRD, XPS and TPD NH3 was performed analogously with [11], Before the catalytic tests were carried out in the reaction of pentane oxidation the fresh catalysts had been activated in n-butane oxidation reaction at 440 °C for 96 h. N- and i-pentane oxidation was investigated in the stainless steel reactor with internal diameter 6 mm, the catalyst load was 0.5 c.c. The feedstock of two different compositions was used: 1.6 % vol. n-pentane (Fluka) and 1.8 % vol. i-pentane (Aldrich) in air. It has been also studied 1-butanol (Aldrich) oxidation (0.9 % vol. in air) in the same reaction conditions. Initial compounds and the reaction products

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 30

were analized by means of GC using two on-line Chrom-5 chromatographies with data collection and processing with PC. Zeolites of NaX type were used for analysis of 02, CO (the column length 2 m, 10-11 °C). silicagel LSK-2.5 - for analysis of C02 and hydrocarbons (3 m. 50-150 °C), two columns: Poropak Q (3 m, 100-220 °C) and F-50 (5%) deposited on Chromosorb (2.5 m, 100-220 °C) for analysis of oxygenated products. Provision was made at the reactor outlet to condense oxygen-containing products, which then were dissolved in acetone (acetonitrile) and analized on Varian 3400 chromatography equipped with capillary column DB-1 and mass-spectrometer lncos-50 Finnigan.

Results and discussion According to XRD data all the investigated samples before catalysis contain in their composition VOHPO4 0.5H2O phase having the most intense reflections at d = 0.570 (100 %), 0.451 (52-61 %), 0.367 (39-45 %), 0.329 (50-53 %), 0.310 (31-35 %), 0.294 (89-97 %), 0.278 (19-23 %), 0.265 (17-23 %), 0.185 (23-30 %). The samples promoted with Bi at BiA/ > 0.05 show additionally XRD lines at d = 0.442, 0.303, 0.286, 0.216 nm and, besides them, 0.352 and 0.245 nm at BiA/ > 0.2. All these reflections belong to BFO4 phase. Their intensity grows with an increase of Bi content in the sample. As for the Zr-containing samples, only that with ZrA/ = 0.3 shows additional reflections which could be attributed to Zr02 phase. The remaining samples do not show any additional peaks on their XRD pattern.

Table 1 The catalysts physico-chemical properties

Sample SSA XPS TPD NH3 m2/g Binding Energy, eV PAZ A, C/A V 2p P2p 01s ml/m2

VPO 15.1 517.3 133.9 531.9 1.32 0.11 0.56 VPBiOOS 12.2 517.3 133.7 532.0 1.32 - - VPBIO10 12.5 517.2 133.8 531.8 1.34 0.20 0.65 VPBIO20 12.2 517.4 133.9 531.9 1.43 0.22 0.70 VPBIO30 12.8 517.5 133.8 532.0 1.55 - - VPCsOOS 10.5 517.5 134.0 531.6 1.35 - - VPCsOlO 10.1 517.4 134.0 531.5 1.44 0.09 0.34 VPTeOOS 9.4 517.7 133.9 532.1 1.33 - - VPTeOlO 5.6 517.6 133.8 532.2 1.39 0.26 0.66 VPZrOlO 17.4 517.7 133.9 532.0 1.44 0.18 0.60 VPZrO20 21.3 517.7 133.8 532.1 1.53 0.25 0.56 VPZrOSO 28.4 517.7 133.9 532.2 1.64 -

* The catalyst figures mean the ratio MeAZ.100

Some physico-chemical properties of the prepared catalysts are summarized in Table 1. One can see from the data in, that additives introduction, except Zr, leads to 31

% 0.4 PhA £

Temperature, C Temperature, C

Figure 1. Effect of the reaction temperature on the products concentration (maleic anhydride - MA, phthalic anhydride - PhA, citraconic anhydride - CA) in n-pentane (A) and i-pentane (B) oxidation over VPBiO20 catalyst. SV = 2640 h"1. some decrease in the specific surface area of the sample. The binding energy of P 2p electrons remains practically unchanged, but that of V 2p and 0 1s electrons slightly increases when introducing Te or Zr. On the contrary, some decrease of V 2p binding energy occurs at Cs introduction. In all the cases promotion of VPO catalyst leads to growth in the surface PAZ ratio. TPD NH; data show that the promoter, except Cs, provokes an increased number of acidic centers on the surface. At the same time, a portion of strong acidic centers (C/A) also grows. An exception here is the same Cs. The main products of n- and i-pentane partial oxidation are maleic, phthalic and methylmaleic anhydrides. Typical curves for thermal dependence of the products concentration change over VPBiO catalyst are presented in Figure 1 (n-pentane oxidation - 1A, i-pentane oxidation - 1B). As it can be seen from Figure 1A, maleic anhydride concentration is higher than that of phthalic and citraconic anhydrides within all the reaction temperature range. The curves for the latter products pass through the maximum value and the highest concentration of citraconic anhydride is observed at the lower temperature. At i-pentane oxidation (Figure 1B) the concentration of maleic anhydride is also higher than that of phthalic anhydride but their total concentration is lower than that in the case of n-pentane oxidation. In so doing, maximum formation of phthalic anhydride shifts to the lower temperatures. At low reaction temperature the major product of the process is citraconic anhydride and its concentration exceeds that obtained at n-pentane oxidation. The maximum content of citraconic anhydride in the reaction products is also shifted to the lower temperatures. Inspection of the complete oxidation products (CO and CO2) forming in both processes reveals some common features with the reaction of n-butane oxidation 1.0 —1 0.8 —i

■■5 0.5 — 0.4 —

400 Temperature, C Temperature, C

Figure 2. Effect of the reaction tempe- Figure 5. Effect of the reaction tempe­ rature on the products concentration rature on the products concentration (butenes- BT, butadiene - BD, maleic (maleic anhydride - MA, phthalic anhyd- anhydride - MA, phthalic anhydride - ride - PhA, citraconic anhydride - CA) PhA) in 1-butanol oxidation over in n-pentane (A) and i-pentane (B) oxi- VPBiOOS catalyst. SV = 2640 h"1. dation over VPZrOlO catalyst. SV = 2640 h'1.

[11]. In all cases carbon oxide can be found in the reaction products at the higher temperatures than carbon dioxide. Within almost all the studied temperature range CO concentration remains lower than CO2 one. Among the products there were also found acetic, acrylic acids, butenes, butadiene (GC), tetrahydrophthalic anhydride, benzoic acid (MS). It should be noted that the products of pentane dehydrogenation were not found by both methods of analysis. Low concentration of C4 olefins does not allow to determine exactly the thermal maximum for their comcentration. It has been hypothized by us in [3], that the main pathway for phthalic anhydride to be formed could be the Diels-Alder reaction between C4= and maleic anhydride, which runs quite easily [12]. In this respect, our results on 1-butanol oxidation over VPBiO catalysts are of a certain interest. It is seen from Figure 2, that the reaction starts at low temperatures and C4= are the only products in a tangible amounts. The reaction temperature increase leads to maleic anhydride forming and the higher temperature the larger its concentration is and the lower content of C4= becomes. Only after an appearance of maleic anhydride phthalic anhydride can be found in the products and its concentration is passing through maximum with the reaction temperature rise. It is quite interesting, that in this case the observed CO2 concentration exceeds that of CO, too. All the results clearly show that the right way for phthalic anhydride formation is the Diels-Alder reaction. 33

In accordance with this, the mechanism of n-pentane oxidation can be presented as following. An activation of the paraffin molecule at the first and fourth carbon atoms is accompanied with methyl group abstraction. Hydrocarbon fragment

C4H9 is oxidized primarily into maleic anhydride. At the same time, if the strong acidic sites are available on the surface, isomerization of the methyl group can happen and i-pentane forms. In its turn, an activation of the latter at the first and fourth carbon atoms leads to methylmaleic abhydride forming. Participation of the strong acidic centers in this process is evidenced by data presented in [3] and those obtained in the present study. Over the samples promoted with Cs and having little number of such acidic centers (Table 1) citraconic anhydride cannot be formed from n-pentane. From the literature data [13] it is known that C4Hg radical is not stable enough and transforms into butene. Thus, for the hydrocarbon fragment remaining after methyl group abstraction two competitive reaction pathways are possible: i) its oxidation to maleic abhydride and ii) its dehydration to butene. On the other hand, butenes are known to be easily oxidized over VPO catalysts into maleic anhydride through butadiene forming stage [14,15], It should be taken into account that maleic anhydride can be produced from butenes over vanadyl pyrophosphate with lower selectivity than from n-butane [16,17]. These facts can serve as an explanation for essentially lower selectivity towards maleic anhydride at n-pentane oxidation as compared to n-butane oxidation over the same VPMeO catalysts. The presence of maleic anhydride amd C4= in the reaction mixture leads to the Diels-Alder reaction run and eventually phthalic anhydride forms. It is absolutely obvious, that the rate of phthalic anhydride formation is limited by the concentration of C4=, which is oxidizing into maleic anhydride though competitive reaction path. That is why the maximum phthalic anhydride concentration is observed at quite low temperatures (the more lower for i-pentane oxidation). It is very difficult to explain this fact by further oxidation of phthalic anhydride as it is done in some papers [5,9,10] because phthalic anhydride can be produced from o-xylene with high selectivity even at much higher temperatures [18]. Close study of n-pentane interaction with vanadyl pyrophosphate surface in TAP system [19] has shown that phthalic anhydride is formed some later than maleic anhydride and while the concentration of the former is constantly increasing the content of the latter is decreasing from impulse to impulse. If the reaction mixture contains 2-pentene, the concentrations of maleic and phthalic anhydrides grow practically simultaneously. The presence of olefin in the reaction mixture predetermines early proceeding of the Diels-Alder reaction. This is also the reason for the larger concentration of phthalic anhydride in the olefin oxidation products as compared with paraffin oxidation. Activation of i-pentane at the first and fourth carbon atoms leads to formation of methylmaleic anhydride as a main product at low temperatures (Figure 1B). An absence of C4= in the reaction mixture does not allow to obtain noticeable amount of phthalic anhydride. The temperature rise provokes partial abstraction of methyl group and formation of maleic anhydride and C4=, that in turn leads to increased concentration of phthalic anhydride. The thermal dependences for maleic and phthalic anhydrides forming similar to those given in Figure 1 are observed for n-pentane oxidation over VPO and the promoted VPBiO, VPCsO, VPTeO catalysts. Figures 3 and 4 present the data on n-pentane conversion and the selectivity towards maleic and phthalic anhydrides in this reaction. One can see, that bismuth

I 34

X hc .%

- 1.4

ii m __LLI |_j __ill ill In Q-5 VPO VPBiOS VPBilO VPBi20 VPBi30 VPCsIO VPTelO VPZrlO

Figure 3. n-Pentane conversion (X) and specific rate (W) of its oxidation over VPMeO catalysts.

□ -PhA

40—_

0 -W. VPO VPBiOS VPBilO VPBI20 VPBI30 VPCsIO VPTelO VPZrlO

Figure 4. Selectivity towards maleic and phthalic anhydrides in the reaction of n- pentane oxidation over VPMeO catalysts.

introduction leads to increased activity of the catalyst. Zirconium and tellurium action as promoters causes decrease in both hydrocarbon conversion and its specific oxidation rate. Maximum value for n-pentane oxidation rate is observed over Cs-promoted catalysts. At the same time, Figure 4 shows that Cs introduction leads to decrease of the partial products yield. Such behaviour of Cs-containing samples can be connected with their physico-chemical properties. Thus, the binding energy of O 1s electrons is the lowest one among VPMeO samples, that points out on enlarged effective negative charge at the oxygen atom. This is a favourable factor for more efficient activation and the higher conversion rate of the paraffin molecule [20]. At the same time, low concentration and strength of the acidic centers dictate low selectivity towards oxygenated products. As for VPTeO catalysts featuring high concentration of acidic centers and enhanced surface acidity at rather high O 1s binding energy, it is observed low activity and quite high selectivity. Moreover, the selectivity towards phthalic anhydride is higher than that of other samples (except VPZrO) at the cost of better dehydrogenative ability of the catalyst as compared to its oxidating activity. The catalysts containing Zr additive stand out because of their unusual selectivity to phthalic anhydride that is even higher than that to maleic anhydride. A little amount of 1,2,3,4-tetrahydronaphthalene, naphthalene and traces of naphthoquinone have been also found among the reaction products by means of mass-spectrometry. The thermal curves for the products accumulation (Figure 5) differ from those presented in Figure 1. It is seen, that phthalic anhydride concentration is always higher than that of maleic anhydride. Their amass is going on concurrently and reaches the maximum concentration at the same temperature. Taking into account our results together with literature data [21] on zirconium dioxide ability to catalize paraffins condensation to form cycles it can be assumed that the reaction mechanism over VPZrO catalysts is different. n-Pentane condensation leads to polycyclic compounds forming followed by their oxidation in two parallel directions yielding maleic and phthalic anhydrides.

Conclusions It can be concluded from our results, that phthalic anhydride formation over VPO catalysts occurs by the Diels-Alder reaction. As this takes place, it is difficult to expect reaching high selectivity because of competitive reaction of C4= oxidation. To enhance the selectivity towards phthalic anhydride it could be reasonable to introduce some additives (for example, Zr) capable to change the paraffin activation mechanism.

References 1. Cent! G., Burattini M„ Trifiro F. Appl. Catal., 32, 353 (1987). 2. Cent! G., Trifiro F. Chem. Eng. Sci., 45, 2589 (1990). 3. Zazhigalov V.A., Haber J„ Stoch J., Mikhajluk B.D., Pyatnitskaya A.I., Komashko G.A., Bacherikova I.V., Catalysis Letters, 37, 95 (1996). 4. Honicke D., Griesbaum K„ Yang Y., Chem.-Ing.-Tech., 59, 222 (1987). 5. Cavani F„ Colombo A., Trifiro F., Sananez Schulz M.T., Volta J.C., Hutchings G.J., Catalysis Letters, 43, 241 (1997). 6. Pyatnitskaya A.I., Komashko G.A., Zazhigalov V.A., Belousov V.M., Bacherikova I.V., Seeboth H., Luke B„ Wolf H., Ladwig G., All-Union Conference on Mechanism Catal. Reactions, Nauka, Moscow, 2, 286 (1978). 7. Zazhigalov V.A., Heterogeneous Hydrocarbon Oxidation, 211th National Meeting, American Chemical Society, New Orleans, 40, 207 (1996). 8 . Zazhigalov V.A., Haber J„ Stoch J., Komashko G.A., Pyatnitskaya A.I., Bacherikova I.V., Polish Pat. 301781 (1996). 36

9. Cavani F„ Colombo A., Guintoli F., Gobbi E., Trifiro F„ Vazquez P., Catal. Today, 32, 125 (1996). 10. Ozkan U.S., Harris T.A., Schilf B.T., Catal. Today, 33, 57 (1997). 11. Zazhigalov V.A., Haber J., Stoch J„ Pyatnitskaya A.I., Komashko G.A., Belousov V.M., Appl. Catal., 96, 135 (1993). 12. Robinson W.D., Mount R.A., Encuclopedia of Chemical Technology. J .Wiley & Sons, Inc., N.-Y., 14, 770 (1981). 13. Chakir A., Cathonnet M., Boettner J.C., Gaillard F„ Combust. Sci. And Tech., 65, 207 (1989). 14. Centi G., Trifiro F„ Ebner J.R., Franchetti V.M., Chem. Rev., 88 , 55 (1988)/ 15. Seeboth H., Kubias B„ Wolf H., Lucke B., Chem. Techn., 28, 730 (1978). 16. Seeboth H„ Ladwig G., Kubias B., Wolf H., Lucke B„ Ukr. Khim. Zhurn., 43, 842 (1977). 17. Miyamoto K., Nitadori T„ Mizuno N., Okuhara T., Misono M., Chem. Lett., 2, 303 (1988). 18. Nikolov V., Klissurski D., Anastasov A., Catal. Rev.-Sci. Eng., 33, 319 (1991). 19. Golinelli G„ Cleaves J.T., J. Molec. Catal., 73, 353 (1992). 20. Zazhigalov V.A., Haber J., Stoch J., Bacherikova I.V., Komashko G.A., Pyatnitskaya A.I., Appl. Catal., 134, 225 (1996). 21. Arata K„ Appl. Catal., 146, 3 (1996). 37

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*05012198111* S. A. Buchholz and H. W. Zanthoff // Ruhr-Universitat Bochum, Lehrstuhl fQrTechnische Chemie, D-44780 Bochum, Germany

ACTIVE CATALYTIC SITES IN THE AMMOXIDATION OF PROPANE AND PROPENE OVER V-Sb-O CATALYSTS

DE99G2091 Abstract The ammoxidation of propane over VSbyOx catalysts (y = 1, 2, 5) was investigated with respect to the role of different oxygen species in the selective and non selective reaction steps using transient experiments in the Temporal Analysis of Products (TAP) reactor. Only lattice oxygen is involved in the oxidation reactions. Using iso ­ topic labelled oxygen it is shown that two different active sites exist on the surface. On site A, which can be reoxidized faster by gas phase oxygen compared to site B, mainly CO is formed. On site B C02 and acrolein as well as NO and NzO in the pres­ ence of ammonia in the feed gas are formed and reoxidation mainly occurs with bulk lattice oxygen ^

Introduction The direct ammoxidation of propane to acrylonitrile has recently attained much at­ tention in industrial and scientific research as an alternative route to the presently used SOHIO-process starting from propene. Catalysts have been developed with maximum yields close to economically attractive values [1, 2]. Complex metal oxides based on vanadium and antimony were found to be promising catalysts for a techni­ cal application of this reaction. However, further improvement of the catalyst is nec­ essary to make this process economically feasible. Basic research work was per­ formed to investigate the catalyst structure of V-Sb-oxide catalysts [3], their catalytic performance [4] and the reaction mechanism [3,5] in order to gain valuable informa ­ tion for possible catalyst improvements. However, only little is known so far about the role of oxygen in the different reaction steps. In a previous work we demon ­ strated that the reaction of propane to acrylonitrile occurs by a redox mechanism with participation of lattice oxygen only [6]. In the present study we investigated the role of

DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 38

different lattice oxygen species in selective and non-selective reactions steps. For this purpose transient experiments with isotopic labelled oxygen (1S02) were per­ formed in the Temporal Analysis of Products reactor (TAP).

Experimental VSbyO, catalysts of different stoichiometry were prepared by redox reaction from Sb203 and NH4VO3 in an aqueous ammonia solution [4] and by solid state reaction of SbaOs and V2Os (index s) [7]. Additionally Sb2C>4 was prepared by calcining Sb203 in air. The catalyst were characterised by surface methods (ISS and XPS) and by bulk methods (XRD, Raman, FTIR, TEM/EDX). The various characterisation meth­ ods reveal that the catalyst consist of amorphous surface VOx , crystalline =VSb04 and Sb204. At Sb:V ratios < 2 also crystalline V2O5 was detected. The TAP reactor has been described previously in detail [5, 8]. Only a brief summary of the experimental data is given here. The VSby Ox catalysts were placed in the iso ­ thermal zone of the microreactor (mcat.= 150-200 mg)), heated to 823 K and kept there for 30 minutes to desorb impurities from the catalyst surface. Subsequently, the catalysts were brought to the desired reaction temperature. Two experimental modes of the TAP reactor were employed: In continuous flow experiments at re­ duced pressure (p < 1 torr), a constant gas stream of (C3H8 /C3H6)/18 02/(Ne/He) or NH3/18 02/(Ne/He) at a flow rate of 0.4 ml-min"1 was fed to the catalyst. The product formation and isotopic distribution in the products were monitored at the reactor out ­ let by continuously recording mass spectra in the range of 1-60 with a quadrupole mass spectrometer. Additionally, vacuum pulse experiments (p = 10'6 Torr; pulse size: 1016 molecules/pulse) were performed using a CgHs/Ne mixture (1:4). Quantifi ­ cation was conducted using an inert gas (Helium/Neon) as internal standard. Fur- therdetails on the reaction conditions of the experiments are given in the results sec­ tion.

Results and discussion Interaction of with V-Sb-Oxides: A CsH

I

lysts, initially, a high conversion of propane was monitored (VSb2Ox: 52 %, VSb20x(s): 28 %, VSbsO x: 35 %). With increasing reaction time conversions de­

creased strongly within the first 20 to 30 minutes and decreased only slightly for

longer times on stream. After 70 (VSb2Ox (s)), 110 (VSb5Ox), 160 (VSb2Ox) minutes on stream the propane conversion decreased to 16 % (VSb2Ox), 9 % (VSb2Ox(s)) and 8 % (VSbsOx). The oxygen conversion increased from 44 % to 54 % for the VSb2Ox, but for the VSb2Ox(s) and VSb5Ox it decreased from 42 % to 21 % and 27 % to 14 %, respectively. No conversion was observed over Sb204. Propene, CO and C02 were formed as main products (yields in the range of 3 % to 25 %) and yields to acrolein of about (0.6 to 1.5 %) % were monitored. Initially, in all oxygen containing products only the isotope 160 originating from the catalyst lattice was detected. With increasing time-on-stream the part of 18 0 detected in the prod ­ ucts increased due to reduction of the catalyst by the hydrocarbon and reoxidation with 18 02 from the gas phase. The distribution of oxygen isotopes in the product C02

overVSb 2Ox is exemplary shown in Figure 1.

—■— c16 o. —A—C O O

Molecules 0o reacted /10

Figure 1 Distribution of oxygen isotopes in the product C02 during experiments employing a continuous gas flow of CaHa/^Oa/He = 1:1:5,5 over VSb2Ox under reduced pressure in the TAP reactor (meat. = 200 mg, T = 940 K, V$rp = 0,41 ml min '). 40

With increasing amount of oxygen converted into the catalyst (= reaction time) the part of 18 0 in CO2 increases. A total content of 50 % 18 0 in the product CO2 is ob ­ served after 9.2-1 019 molecules 18 02 reacted. The part of C16018 0 runs through a

maximum value of 43 % after 5.8-1 019 molecules. The distribution of the oxygen iso ­ topes in C02 is not statistical (expected distribution for the C16018 0 maximum: 25% C16Oz, 50% C1S02, 25% C16018 0). The higher amount of 1sO observed in the prod ­ ucts therefore has to be attributed to additional diffusion of 160 from the catalyst bulk to the surface. A similar behaviour was observed for the other two catalysts. The above results show that the oxidation of propane occurs by lattice oxygen only, in accordance with our previous pulse experiments [6]. A quantification of the number of oxygen species participating in the reaction was difficult, due to overlapping ef­ fects of i) decrease in catalytic activity with reaction time and ii) oxygen diffusion from the bulk at the reaction temperatures applied. However the 50 % exchange of ,8 0 after about 102° reacted oxygen molecules show, that only a few near-surface layers of oxygen are involved in the reaction.

Interaction of CiH k /18 Q? with the catalyst: Continuos flow experiments were per­ formed at a lower reaction temperature in order to minimise oxygen bulk diffusion using the primary reaction product propene as hydrocarbon. A CaHe/^Oz/Ne = 1:1:8.8 gas mixture was supplied at 663 K (VSb,Ox, VSb2Ox) and at 713 K (VSb5Ox). Again, the conversion time-on-stream dependence showed a decrease over the first minutes of reaction. Initially, only 160 containing products (acrolein, CO and C02) were observed. However, with increasing reaction time the part of 18 0 containing products increased. Table 1 summarises the isotopic distributions observed at differ­

ent stages of the catalyst treatment: a) at t = 0 s, b) after about 1 h continuous flow of C3H6/18 02/Ne, c) after subsequent vacuum treatment at reaction temperature and d) after subsequent vacuum treatment at higher temperature. After approx. 1 h treatment with reaction mixture a content of 18 0 in the different products of 54 to 75

% is observed. After 1 h interim time, leaving the catalyst in the vacuum without gas- flow, a decrease of the 18 0 content in CC2 and acrolein for the observed. No change is observed for the 18 0 content in CO except for the VSb,Ox. After a further interim

heating up of the catalyst in the vacuum (AT = VSb,Ox, VSb2Ox: 100 K; VSb5Ox: 50 K) and cooling down to reaction temperature again, the ,8 0 content in all products 41

strongly decreased, especially in C02/acrolein. It can be further deduced from Table 1 that the 18 0 content in acrolein at the different reaction conditions behaves similar to that of C02, while it is significantly different from that in CO for VSb2Ox and

VSb5Ox.

Table 118 0 content in the oxidation products of Caty^CVNe = 1:1:8.8 over VSb,Ox a) at t = 0 s, b) after about 1 h continuous flow, c) after subsequent vacuum treatment at reaction temperature and d) after subsequent vacuum treatment at T = 773 K(mcaL = 150 mg, Vstp = 0,45 mi min'1 for all catalysts, VSb,Oxt VSb2Ox: T = 663 K, VSb5Ox: T = 713 K). Catalyst treatment 18 0 in acrolein / 18 0 in C02/ 18 0 in CO/ % % % VSbi.2.sOx a) t = 0 min 0 0 0 VSb,Ox b) t= 56 min 62 61 59 c) 40 min vacuum 61 58 56 d) after heating 34 33 34 VSb2Ox b) t = 68 min 66 68 73 c) 40 min vacuum 59 59 73 d) after heating 31 33 42 VSbsOx b) t = 60 min 54 56 75 c) 35 min vacuum 48 50 75 d) after heating 38 39 59

The presence of only 160 containing products at t = 0 min experiments again reveals that only lattice oxygen is involved in the product formation. Acrolein and C02 exhibit similar 18 0 contents under all reaction conditions applied. Therefore, it is assumed that these products are formed at the same sites (type A). Site A exchanges its oxy­ gen easily with lattice oxygen from subsurface layers (oxygen diffusion between bulk/surface leads to a decrease in 1sO in the products after some time in the vac­ uum). In contrast the different isotopic behaviour of CO reveals that this product is apparently formed on a different active site (type B) which is not exchanging its oxy ­ gen with subsurface layers in a high rate, but is easily reoxidised with gas phase oxygen.

Interaction of NH t/18Op with the catalyst: Similar experiments to those described above with propane and propene were performed using a NH3/18 02/Ne = 1:1:13 re­

action mixture. N2 and nitric oxides (NO and N20) are formed as has been reported previously [9]. Initially, the NO* products contained only 160 from the catalyst lattice. 42

With increasing time the *18 0 content in NO and N20 increased. The continuous flow of NH3/18 02/Ne was stopped several times and pulse experiments using C3Hs/Ne were performed during the breaks in order to get information about the distribution of isotopes in the oxidation products of propene. Table 2 summarises the conversion of propene, the yields to C02 and acrolein, the 18 0 content of acrolein, C02 and CO during pulse experiments and in N0/N20 during the continuous flow experiment im­ mediately before the CaHs pulse experiments

Table 2 Yields and 18 0 content in acrolein, C02, CO (and N20, NO)* during pulses of C3H6 over VSb50„ treated with NHV^Oz/Neon (meat =150 mg, T = 663K, C3H6/Ne = 1:4, NHs/^Oa/Ne = 1:1:13).

Time X(C3H6)** Y(ACR) Y(C02) Y(CO) 1sO content % min % % % % ACR C02 CO NO* N20 0 78.1 2.7 25.8 49.6 0 0 0 0 0 10 73.5 38.3 16.0 19.2 14 14 14 13 17 22 67.0 42.0 12.6 12.4 28 26 22 29 34 31 52.8 35.7 9.0 8.1 37 37 25 37 42 after 1 h 38.5 24.3 8.0 6.2 28 27 24 25 29 vacuum * during treatment of the catalyst with a continuous gas flow of NH:/“02/Neon "additional 10 to 20 % of propene were irreversibly adsorbed on the catalyst

From Table 2 it becomes obvious that 1sO lattice oxygen, originating from the reac­ tion of NHa/18 02 with the catalyst, is detected also in the products of propene oxida ­ tion. The content of isotopes is comparable in C02/Acro!ein and N0/N20. With the NHa/'8 02 treatment the 18 0 content in the products increased from 0 to 37 % in NO and C02 and acrolein and to 42 % in N20. For CO a smaller content of 18 0 is ob ­ served, after 31 minutes only 25 % are found in the product. After 1 h in vacuum the 18 0 content decreased to 27+2 % for C02, acrolein, NO and N20. For CO it re­ mained nearly unchanged at 24 %. The analysis of the isotopic distribution of N0/N2O and C02/acrolein shows that the formation of these products proceeds with participation of lattice oxygen and that the content of 18 0 increases with an identical rate in those products. Furthermore an identical decrease of the 18 0 content is observed when oxygen bulk diffusion is en­ hanced by increasing the reaction temperature. This indicates that these products 43

are most probably formed on similar sites. In contrast, the 18 0 content in CO is dif­ ferent, supporting the finding that CO is formed on a different site.

Conclusions In the present study the role of different lattice oxygen species in different reaction steps of the ammoxidation of propane over V-Sb-0 catalysts was investigated. The results reveal that all reaction steps proceed with participation of lattice oxygen only. Oxygen originating from the gas phase only reoxidises the active sites reduced by ammonia or the hydrocarbon. This finding is in accordance with results reported in literature for propane conversion over other vanadium containing catalysts, e.g. V- Mg-oxides [10], Ag 0,otBio.ssVo^Moo^sOx/AlaOa [11] and V-P-oxides [12,13 and refer­ ences therein]. The different time-on-stream and temperature behaviour of the 18 0 content of CO and COa/acrolein revealed that two different active sites exist. On site A, which is reoxidised fast with gas-phase oxygen but only slowly with lattice oxygen, mainly CO is formed. With increasing Sb:V-ratio the number of site of Type A (CO formation) decreases as indicated by a lower selectivity towards CO at similar conversions. On site B mainly CO* and acrolein are formed and reoxidation occurs easily with bulk lattice oxygen. Since Sb204 behaves like an inert, it is concluded that both active sites contain vanadium as key element. The assumption of two different active sites is in accordance with recent FTIR-investigations of the reaction of Ca-hydrocarbons and oxygen over V-Sb-0 catalysts by Cent! et. al. [14]. The authors observed two different absorption bands which were assigned to different surface vanadium spe ­ cies. One species (v = 1145 cm'1) is assigned to an amorphous vanadium oxide (VOx) which is easily reduced and slowly reoxidised with bulk lattice oxygen. A sec­ ond band (v = 1060 cm"1) is assigned to a vanadium species surrounded by anti­ mony. This species is reoxidised by both adsorbed gas phase and bulk oxygen spe ­ cies. The combination of the results presented in this work an the FTIR-investigations of Centi et al. allows us to assign the type A sites of to a free vanadium oxide species (VOx), whereas type B sites can be assigned to vanadium in the non-stoichiometric =VSb04 phase. The surface and bulk redox processes during reaction of propene and/or ammonia with the catalyst can be summarised as shown in Figure 2. 44

Figure 2 Proposed participation of surface and bulk oxygen species in partial oxidation of propene over V-Sb-oxides

Acknowledgement The present work was financially supported by the German Federal Ministry of Edu­ cation, Science, Research and Technology, project no. 030 0059 B8 .

References

[1] Moro-Oka, Y.; Ueda, W.; in: Catalysis Vol. 11, Royal Chem. Soc., Athenaum Press, 1994 [2] Hou, Z.-Y.; Dai, Q.; Wu, X.-Q., Chen, G.T.; Appl. Catal. 161 (1997), 183 [3] Centi, G.; Perathoner, S.; Trifiro, F.; Appl. Catal. A: Generali 57 (1997), 143 [4] Guttman, A.T.; Grasseli, R.K.; Brazdil, J.F.; Suresh, D.D.; U.S. Pat. US 4,788,317 [5] Buchholz, S.A.; Zanthoff, H.W.; ACS Symp. Ser. 638 (1996), 259 [6] Zanthoff, H.W.; Buchholz, S.A.; Catal. Letters 49 (1997), 213 [7] Birchall, J.; Sleight, A.W.; Inorg. Chem. 15 (1976), 868 [8 ] Buyevskaya, O.; Rothaemel, M.; Zanthoff, H.; Baems, M., J. Catal. 146 (1994), 336 [9] Zanthoff, H.W.; Buchholz, S.A.; Ovsitser, O.Y.; Catal. Today 32 (1996), 291 [10] Michalakos, P.M.; Kung, M.C.; Jaha, J., Kung, N.H., J. Catal. 140 (1993), 226 [11] Baems, M.; Buyevskaya, O.V.; Kubik, M.; Maiti, G.; Ovsitser, O.; Seel, O.; Catal. Today 33(31 (19971 85 [12] Kung, H.H: Kung, M.C.; Appl. Catal. A:General 157 (1997) 105 [13] Grzybowska-Swierkosz, B.; Appl. Catal. A:General, 157 (1997) 409 [14] Centi, G.; March!, F.; Perathoner, S.; J. Chem. Soc., Faraday Trans. 92(241 (1996), 5151 DE99G2090

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 199°

I ^-*DE012198120*

H. M. Abdeldayem*, P. Ruiz*, F. C. Thyrion**, B. Delmon* ------* Unite de Catalyse et Chimie des Materiaux Divises, Universite Catholique de Louvain, Place Croix du Sud 2/17, B-1348 Louvain-La-Neuve, Belgium ** Unite des Precedes Faculte des Sciences Appliquees, Universite Catholique de ! Louvain, Louvain-La-Neuve, Belgium

; A NEW KINETIC MODEL BASED ON THE REMOTE CONTROL MECHANISM TO \ ! FIT EXPERIMENTAL DATA IN THE SELECTIVE OXIDATION OF PROPENE INTO | ACROLEIN ON BIPHASIC CATALYSTS ! Abstract ! ! (A new kinetic model for a more accurate and detailed fitting of the experimental | ‘ data is proposed. The model is based on the remote control mechanism (RCM). The j RCM assumes that some oxides (called "donors") are able to activate molecular I ! oxygen transforming it to very active mobile species (spillover oxygen(Oos)). 0# ; ' migrates onto the surface of the other oxide (called "acceptor") where it creates j j and/or regenerates the active sites during the reaction. The model contains two terms, « ■ one considering the creation of selective sites and the other the catalytic reaction at j each site. The model has been tested in the selective oxidation of propene into ! acrolein (T= 380,400, 420°C; oxygen and propene partial pressures between 38 and I 152 Torr). Catalysts were prepared as pure M0O3 (acceptor) and their mechanical I mixtures with a-Sb204 (donor) in different proportions. The presence of a-Sb204 ! changes the reaction order, the activation energy of the reaction and the number of | active sites of M0O3 produced by oxygen spillover. These changes are consistent with i j a modification in the degree of irrigation of the surface by oxygen spillover. The fitting j ! of the model to experimental results shows that the number of sites created by Oso i increases with the amount of a-Sb204- \ ; j 1-Introduction ! In the field of selective oxidation, kinetic parameters (reaction orders, energy of i activation, reduction and oxidation constants) derived from the conventional models j turn out to be exceedingly sensitive to the proportion of the various elements and to experimental conditions [1-3]. The reasons why experimental results depart from the corresponding equations are not clear yet Multiphase catalysts always exhibit better performances than those containing only one phase particularly in terms of selectivity and resistance to ageing. One of the increasingly more popular explanations of the superiority of multiphase oxide catalyst j is the occurrence of a remote control mechanism (RCM) [4]. A RCM operates when the first phase, which we call "donor", dissociates molecular oxygen 02 to spillover Oxygen Oo$ and when this Oos, having migrated to the acceptor, reacts with its surface j to give more active and more selective sites. This results in a steady state situation in which selective sites are continuously created by the effect of 0« on the acceptor, but are also slowly deactivated because of harmful side reactions occurring in parallel with the main one. During the selective oxidation of the hydrocarbon, spillover oxygen controls the dynamic surface phenomena of the continuous oxido-reduction process. Investigations done in our laboratory in the last years suggest new lines for a more accurate modeling of the kinetics of selective oxidation reactions. Melo Faus et al. [5] have developed a mathematical model of the remote control effect in cooperation between a-Sb204 (spillover oxygen donor) and M0O3 (acceptor) and j DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 ! 46

calculated the catalytic activity as a function of proportion of donor a-Sb204 and acceptor M0O3 in the mechanical mixtures, at different reaction conditions. The calculated results match very well with results obtained experimentally in the dehydration of N-ethyl formamide, an oxygen-aided catalytic reaction. On the basis of pseudo-steady state equations of this model, Rebitsky et al. constructed a complete kinetic model of the remote control effect in selective oxidation, including transient effects [6], This predicts isothermal bistability in certain conditions. Such bistabilities have been found experimentally in special reactions [7]. Recently Vande-putte et al [8 ] have proposed a model based on the remote control mechanism and a classical Mars-van Krevelen equation [9] in the selective oxidation of isobutene to methacrolein. For mechanical mixtures of a-Sb204 and M0O3 in different proportions, the model gives the fraction of the potential active sites which are really active as a function of the catalyst composition and experimental conditions. The fraction of the surface of M0O3 which selectivity realises the oxidation via the oxidoreduction processes depends on the degree of irrigation of M0O3 by spillover oxygen produced on a-Sb204 and cannot be considered as constant. As in the case of hydrotreatment [10], we aimed at developing the model that had been proposed by Vand-Putte et al. in order to estimate the number of active sites produced by 0M as a function of the reaction conditions and catalyst composition namely. The new model is more detailed, in particular with respect to the creation of new actives sites.

2-Basis for a new kinetic model based on the RCM The models based on the Mars-Van Krevelen (MVK) picture do reflect the local oxido-reduction process, but they generally fail to reflect the changes of selectivity observed when the surface gets more or less reduced, as a function of the reaction conditions. They actually neglect the influence of the donor phase and the spillover oxygen it emits. The new kinetic model is based on the oxido-reduction process but it incorporates both the mechanism of site creation (via RCM) and the catalytic reaction at each sites given by the classical MVK (1): rate = - —re^3Hc_^o _xPg)( (1) rvK‘ red rPHCn K„p: The overall kinetic equation then takes the form: rate = Fre(Donor, Acceptor, PQX, Phc . T) X (Catalytic reaction) (2)

According to RCM, the creation of active sites is a function, Fre, of the proportion of donor and acceptor, and the experimental reaction conditions, where Pox and PHc represent, respectively, the oxygen and hydrocarbon pressures and T the temperature of the reaction. The catalytic reaction in the created sites is given by the classical MVK mechanism. The difficulty in the oxidation reaction is that the oxygen and hydrocarbon pressures are present as variables in both parts of the equation (RCM on the one hand and MVK on the other one). This general case has been treated theoretically, describing the time-dependent transient effect when the experimental conditions are modified [6], The aim of the present work is limited to 3 objectives: (1) to show that significant changes in the apparent reaction order and activation energy of the reaction are 47

indeed observed when the RCM operates and that these changes can be described When taking into account the irrigation of the acceptor by Oos, (2) to compare the fitting of the experimental results using our modelwith that obtained when using theclassical Mars-Van Krevelen model and (3) to show that the parameters used, which in theory have a physical meaning, correspond to acceptable values. • If the number of active sites is constant, the overall reaction rate for a total number N of catalytic sites is: Pn K fed ‘ HC k p: rate = N (3) KredPJc + K=P2 Where K^d and Kox are the rate constants (for individual catalytic site) of the reduction and reoxidation steps respectively. The reaction order corresponding to oxygen and alkene (n, m) is usually taken as 0.5 to 1.0. As the catalytic sites are created through the remote control, the quantity N in eqn. (3) should vary as a function of the amount of spillover oxygen produced by the donor. If the catalytic activity of the sites created by spillover oxygen is the same as those initially existing on the acceptor phase, the modified equation will simply be obtained by changing the total number N in eqn (3) by (Naccept + Nos), where Naccept represents the sites contained in the acceptor and Nos the sites created on the acceptor by spillover oxygen. As indicated before, Nos would depend not only on the structure of the catalyst (donor, contacts between donor and acceptor) but also on the reaction conditions (T, Phc , Pox , etc.):

Nos» ((Donor, contacts, PHc, Pox-) (4)

Then, in the general form, the reaction rate will be written as: rate = K«,„, + N„(Ponor,contacts,T,P HC,Pjj - ,5, *\ed ‘He + *'ox rox

Calculated in this way, the rate measured under given conditions (Phc , Pox , T) will reflect the variation of Nos due to action of spillover oxygen and, namely, the variation of the function Frc of eqn (1). The present work concerns the selective oxidation of propene into acrolein. Catalysts were constituted of i) pure M0O3, as typical acceptor ii) pure a-Sb204 as typical donor and iii) mechanical mixtures of M0O3 and a-Sb204 in different proportions. The donor ct-Sb204 is completely inert in the oxidation of propene to acrolein. No decomposition of propene or reaction products were observed when a- Sb2C>4 was used as catalyst. No contamination has been observed in samples containing both a-Sb2C>4 (donor) and M0O3 (acceptor) by mixed mechanically [11]. This justifies the assumption that the synergy observed and change in the kinetics parameters are due to a RCM, namely that the single role of a-Sb204 is to control the catalytic activity of M0O3 thanks to the action of spillover oxygen. The relative values of Km*, and Kox can be determined from the variation of the reaction rate as a function of PHc and Pox when working with pure M0O3, because the number of active sites in the acceptor is constant and is not influenced by remote control effect (pure M0O3 has only acceptor properties). If our model is correct, the values of Nos which reflect the fraction of active sites produced by spillover oxygen 48

should increase when the proportion of the donor increases. This is one of the points we wish to demonstrate in this paper. On the other hand, as shown above, the presence of donor can reoxidize the acceptor via spillover oxygen in a more efficient way. This would keep the acceptor phase in a more oxidized form compared with pure acceptor phase; the state of the acceptor (M0O3) with respect to oxygen partial pressure as reflected by the reaction order, should be depending on the a-Sb204 amount in the mixture and actually increase when the a-Sb204 content increases. Kinetically, this double dependence should lead to apparent (empirical) order which may vary when the reaction parameters are modified. With this exception, we have calculated the empirical reaction order using eqn (A):

rate (A) as a first approximation, the empirical activation energy was also calculated by plotting the logarithm of the reaction rate as a function of the inverse of temperature.

3-Experimental 3.1. Preparation of the catalyst oxides Pure oxides were prepared separately. M0O3 was prepared starting from an aqueous solution (82 g/l) of ammonium heptamolybdate (Merck, extra pure powder) complexed with the same number of equivalents of oxalic acid (Janssen Chimica, 99 + %). The mixture was stirred at 40°C until obtaining a homogeneous limpid solution. After having removed the solvent under vacuum, the obtained solid was dried overnight at 110°C, decomposed at 330°C during 20 h and calcined at 400°C during 20 h and ground. a-Sb204 was prepared by dispersing 50g Sb203(99 +%) ( Aldrich, extra pure powder ) in 100 ml concentrated nitric acid (65%) at 50°C during 1h. After having removed the nitric acid under vacuum, the obtained solid was washed carefully with distilled water until the filtrate reached pH = 6 The obtained solid was dried in air overnight at 110°C and then at 500°C during 24 h. All phases have been checked by X - ray diffraction. This fits, respectively, the patterns of a-Sb204 cervantite and M0O3 molybdite phases reported in the ASTM files. Specific surface area of a-Sb204 and M0O3 were respectively 2.4 m2/g and 6.2 m2/g.

3.2. Mechanical mixtures of acceptor and donor The mechanical mixtures a-Sb204 + M0O3, which we shall also call biphasic catalysts, were prepared by dispersing adequate amounts of both powders in 200 ml of n-pentane. The suspension was agitated vigorously with an Ultra-Turrax equipment at 60001 /min during lOmin and subsequently ultrasonically for 3 min. The n-pentane was evaporated under reduced pressure, while continuing the agitation at 25°C. The solid was dried in air at 100°C overnight. The corresponding mechanical mixtures were not subjected to further calcination. The mixture was pressed into wafers, then gently broken. The fraction between 100 and 800 pm was used. The granulometry of the catalyst is thus not too small, and permits easy flow of the reacting gas. The pure oxides were treated exactly in the same way as the mixtures. The composition of the mixtures was expressed as mass fraction : 49

K ______Weight of Mo03______m Weight of MoOi + Weight of a - Sb204 Mechanical mixtures with Rm = 0.0 (a-SbzO^; 0.25; 0.5; 0.75 and 1.0 (pure Mo03) were prepared.

3.3. Catalytic test 3.3.1. Test equipment and general test conditions Catalytic measurements were performed in a conventional fixed-bed reactor system at atmospheric pressure. The reactor was made of a Pyrex U-tube of 8 mm internal diameter into which the catalyst was packed. In order to obtain directly unambiguous results; the rates measured should be based on results obtained The amount of catalyst was varied from experiment to experiment in order to have conversion in the range 5 to 15% (amount of catalyst between 250-2000mg). The catalytic bed was deposited over a bed of small glass spheres (sphere diameter (0.75- 0.5 mm ) of a height of 10 cm and under another bed of 3 cm high. The catalytic tests were realized at atmospheric pressure. The reactant mixture contained propene and oxygen diluted with helium. The total feed rate was 30 ml Z min in all cases. It was checked that there was no diffusional limitation into the pores of catalyst and intra particle diffusion had no influence on the rate observed. Acrolein was the main selective product. No other partially oxygenated product was observed. The carbon mass balance has been calculated and was always better than 90%, in spite of the analytical difficulties due to the low conversion.

3.3.2. Standard test varying the partial pressure of oxygen or the partial pressure of propene The procedure for a test was as follows: the reactor was fed with a mixture corresponding to a partial pressure of oxygen of 76 Torr, a partial pressure of propene of 76 Torr, a total pressure of 760 Torr. The diluent was helium and the reactor stabilized to the desired temperature. Analyses of reactants and products were realised during 3h. The test was repeated by changing the starting feed of the gas mixture for other partial pressures of oxygen (152 Torr and 38 Torr) maintaining the partial pressure of propene constant at 76 Torr. The partial pressure of helium was changed to maintain the total pressure at atmospheric condition. An identical procedure was repeated for other partial pressures of propene (152 Torr and 38 Torr). The tests were made at 380,400 and 420°C. For each composition of the gas mixture and each temperature the catalytic tests were carried out on fresh catalyst sample. The fitting of the parameters to the theoretical models was based on the 'least square* method. This method provides the set of variable parameters of the chosen model which minimizes the sum of the squares difference between the experimental rate (rateeXp) and those given by the theoretical model (ratem0del).

4. Results and Discussion 4.1. Empirical Reactionorder for oxygen and propene As indicated above, we calculated the values of reaction orders using the empirical formula eqn (A). In the case of pure M0O3 the rate of oxidation of propene to acrolein was first order in both oxygen and propene (Table 1). In the case of mixtures, the rate of partial oxidation of propene into acrolein was nearly first order in proplyene 50

(a result similar to that obtained with M0O3). But significant changes in the oxygen order were observed as a function of the a-Sb204 amount in the mixtures.

Table 1 Empirical parameters, oxygen and propene order at400°C and activation energy. Catalyst propene order (n) oxygen order (m) E (KJ/mol) MoO 3(100%) 1.0 0.9 137.0 Mo0 3(75%)+a-Sb204(25%) 0.7 0.1 91.9 Mo0 3(50%)+a-Sb204(50%) 0.8 0.4 103.3 Mo0 3(25%)+ a-Sb204 (75%) 1.0 0.0 84.2

Referring to Table 1, if we accept that the changes of reaction orders reflect changes in the MVK mechanism, the results would lead to an absurd conclusion, namely that the oxidoreduction of a same phase, namely M0O3, would exhibit different kinetics of reduction or oxidation according to experimental conditions. It would be even more difficult to accept that the reaction orders of both reduction and oxidation change so drastically (e.g., for oxygen nearly 0 for Rm = 0.25 to 0.9 for Rm =1.0). In addition, the fractional order found cannot be related to any mechanistic feature. In this perspective, the results would rather suggest that the MVK mechanism is not valid. The RCM provides the explanation of the change of the reaction order: this not due to any modification of the kinetics of reduction of the surface by the hydrocarbon or reoxidation by molecular oxygen. This simply reflects the modification of the number of selective active sites which undergo the reduction-oxidation process. When M0O3 is alone, the dependence with respect to the oxygen partial pressure as measured by reaction orders is high because the oxidation step is difficult. Conversely, the reoxidation is easier when a-Sb204 is present, giving a lower order with respect to the partial pressure of oxygen due to irrigation of the surface of M0O3 by oxygen spillover (Oos), permitting in this way higher reoxidation rates.

4.2. Empirical activation energy The empirical activation energy obtained for catalysts containing pure M0O3 and various contents of a-Sb204 are present also in Table 1. The activation energy of oxidation of propene into acrolein decreases quite significantly when M0O3 is mixed with

involved, and that the classical MVK model is not sufficient to interpret the results. This confirms the fact that when two phases are present in the catalyts, the classical MVK model is not sufficient for accounting for the complexresults obtained.

Table 2. Values of and K<,x obtained by fitting experimental reaction rate at 400°C to MVK eqn. (1)______'______Catalyst K«, (Torr* 1) KnrfTon-h MoO 3(100%) 0.4 0.62 Mo0 3(75%)+a-Sb204(25%) 0.33 2.27 MoO 3(50%)+a- Sb204 (50%) 0.38 1.82 M0O3 (25%)+a- Sb204 (75%) 0.46 4.53

4. 3. 2. Fitting of the new model Table 3 also presents the values of Nos obtained by fitting the experimental rates of mixtures as a function of Phc and Pox with model eqn. 5 considering Naccept =1 • The values of Kred and Ko X have been taken of pure M0O3 given in Table 2. As expected, the number of active sites produced by oxygen spillover (namely Nos) in the reduction-oxidation mechanisms increases with the proportion of a-SbzO, in the mixtures. In the context of the remote control mechanism, the increase of Nos with the amount of a-Sb204 in the mixture is due to a better irrigation of the surface of M0O3 by Oso. Similar correlation between Nos and the amount of a-SbzO* in the mixtures was obtained at the other different temperatures (380 and 420°C).

Table 3 Values of Nos obtained by fitting experimental reaction rate at 400°C to eqn.(5) Catalyst Kred (Torr" 1) Kq X (Torr' 1) N q S MoO 3(100%) 0.4 0.62 0 Mo0 3(75%)+a-Sb204(25%) 0.4 0.62 0.26 Mo03(50%)+a-Sb20 4(50%) 0.4 0.62 0.36 Mo0 3(25%)+a-Sb204(75%) 0.4 0.62 0.86 In a simpler approach, the values of Nos presented in Table 3 can be used to model the influence of Oso on the Fre during the reaction. Actually this approach is insufficient, however, because the amount of Oso flowing onto the surface of M0O3 depends on in particular oxygen and hydrocarbon pressures, number of contacts between a-Sb204 and MoOa, temperature, etc. (eqn. 2) and also the catalytic reaction term namely MVK equation depends on the experimental conditions (oxygen and hydrocarbon pressure). A substantially better agreement is obtained when separating the influence of oxygen in increasing the amount of O50 produced when the amount of donor increases, from the influence of an increase of 62 on the KoxP<* term of MVK equation. Let us assume, as first approximation, that Nos is proportional to P“ (taking into account the fact that the production of Oso on the donor necessitates dissociation of 02): N „ - N P “ . The parameter N* s should reflect more precisely the influence of the amount of a-Sb204. The values of N* s in Table 4 increases when increasing the a-Sb204 amount in the mixtures. The trends are exactly these expected. 52

Table 4 Values of obtained by fitting experimental reaction rate at400°C to model eqn.(5) assuming N M = N ", P 5 considering Naccept =1, the values of K«a and Kox have been taken identical to those of pure MoQ 3 (Table 2).______Catalyst n ;s

Mo 03(75%) + a-Sb204(25%) 0.02 Mo 03 (50%)+ a-Sb204 (50%) 0.03 Mo 03 (25%)+ a-Sb204 (75%) 0.07

5-Conclusions The essential feature of the new model is the separation of the kinetic equation in two terms, one constituted of the classical MVK and the other one given by the RCM. Our results prove that the new model based on this mechanism (eqn. (2)) simultaneously (i) fits the experimental results in a self-consistent way and (ii) fully respects the universally accepted features of the MVK mechanism. This model fits more adequately experimental data. The new model offers good prospects for a satisfactory modeling of experimental data. It is compatible with the remote control mechanism. All the parameters of the model have a precise Physical meaning.

References [1] F. Benyahia and A.M. Meams, Appl. Catal., 70,149 [1991]. [2] F. Benyahia and A.M. Meams, Appl. Catal., 66,383 [1990]. [3] D.L. Hucknal, Selective Oxidation of Hydrocarbons, Academic Press, London, 1974. [4] L.T. Weng and B. Delmon, Appl. Catal. A., 81,141 [1992]. [5] F. Melo Faus, B. Zhou, H, Matralis and B. Delmon, J. Catal, 132,200 [1991]. [6] T. Rebitzki, B. Delmon and J.H. Block, AlChE J„ 41(6) [1995]. [7] A. Gil, P. Ruiz, B. Delmon, Canad. J. Chem. Eng., 74,600 [1996]. [8] D. Vande Putte, S. Hoomaerts, F.C. Thyrion, P.Ruiz and B.Delmon, Catal. Today, 32, 255 [1995]. [9] P.Mars and D.W. Van Krevelen, Chem. Eng. Sci., Supll., 3,41 [1954]. [10] Y.- W. Li, B. Delmon, Proc.4th. Intern. Conf. On Spillover, Dalian, China, Sept. 15-18,1997, in press. [11] L. T. Weng B. Zhou, B. Yasse, B. Doumain, P. Ruiz and B. Delmon, in M.Phillips and Teman (Editors). Proc. of the Int Congr. Catal., Calgary, Canada, Chemical Institute of Canada, Ottawa, Vol. 4,1988, P. 1609. 53

DGMK-Conference "Selective Oxidations in Petrochemistry'

| DE99G2089 C. C. Hobbs ! Celanese Ltd., Corpus Christ! Technical Center, 1901 Clarkwood Road, i P.O. Box 9077, Corpus Christi, TX 78469, USA

THE LIQUID PHASE OXIDATION OF N-BUTANE: A SEARCH FOR PLAUSIBLE MECHANISMS

The self-sustaining oxidation of paraffin hydrocarbons is an extremely complex process involving multiple series and parallel oxidation reactions. It is further complicated by inde­ pendent concurrent reactions among the products; e.g., esterification of acids and alcohols. Each product in a lower oxidation state than carbon dioxide is subject to subsequent attack by the free radical flux, leading to still further oxidation. A free radical oxidation is, inevitably, a cooxidation of reactants and products. Usually, the more significant final products come from other isolable intermediates rather than directly from the hydrocarbon feeds. Additional complications arise when some part of the products formed can be stripped into a gaseous vent stream and/or when more than one liquid phase is formed in the recovery process. In such cases, the ratios of the product components in the product stream can be quite different from the ratios in the reaction zone. These complications can cause products to have different histories with respect to exposure to reaction conditions. Application of ki­ netic equations to such systems requires that such differences be taken into account. Yet another potential difficulty is related to the kinetics of branching chain reactions. Starting such reactions usually requires more than just mixing the reactants. Moreover, such reactions are subject to inhibition by even trace amounts of effective inhibitors. The specific reaction rates of such reactions may change noticeably, even by orders of magnitude, without a significant change in the concentrations of the measured intermediates. Some processes, such as production of synthetic fatty acids, have been in existence for more than 100 years. Beginning in the 1930s and continuing in recent times, a series of re­ markable investigations by some very able workers has given real insight into the mecha­ nisms of many of the individual reactions. However, because of the problems previously mentioned, the knowledge of the inner workings of practical processes can still be called fragmentary. Knowledge of the relationships between products and operating conditions fre­ ■h quently appears to be rather empirical (or even mythical). . . IvjS A , ■■^oday rI wo uld-liko-to-teH-yeu-aboijt an approach that f-belicvf has given pjs some key in­ {ao.es formation about the mechanisms of the liquid phase oxidation of butane to acetic acid. This . procedure has been developed over the last 34 years; however, much of whatfl wilfJaiscussGw jjtedajr represents a synthesis of previous insights. Many of the observations are relatively recent and have not been previously published. In principle, this approach should be applica ­ ble to many oxidation processes. ) Let me begin by noting that Dr. Cheves Walling, in a very illuminating paper published in JACS in December 1969, reported that limiting rate reactions tend to have short chain lengths and that hydroperoxides tend to be transient intermediates in such reactions'. Another important conclusion was that, based on a simple kinetic scheme, limiting rates are not a function of hydroperoxide concentration or catalyst concentration.

DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 54

This description appears to apply very well to most aspects of butane LPO. On this ba­ sis, we may describe the first isolable non-peroxidic components which come directly from the hydrocarbon as “primary." Non-peroxidic products which come only from primaries can be called “secondaries, ” etc. It is quite possible for an intermediate to be both primary and secondary (and even higher). For example, ethanol appears to be primary in the sense de­ scribed. Acetaldehyde, on the other hand, is both primary and secondary; that is, some of it comes directly from butane and some of it is produced by secondary attack on primary etha­ nol:

r,-CtHl0-iSU-ElOH—^AcH I[O]I

The radical flux initiates the reaction with each substrate by abstracting hydrogen from it. Subsequent reaction of the resulting carbon centered radical with oxygen is very rapid; i.e., the hydrogen abstraction is rate limiting. The consumption of hydrocarbon is described by:

(2> at where HC= hydrocarbon = average radical lcHC = oxidation rate constant for the hydrocarbon [/] = molar concentration of i in reactor solvent and the rate of accumulation of a primary is given by:

~ = a*^*k'f*[ HC]* [A]-C *[P]* [Jt.] (3) at rir where a = fraction of carbon in consumed HC going to P n _ number of carbon atoms in a molecule of i.

/t' _ oxidation rate constant for i.

It is possible to divide Equation 3 by Equation 2 and cancel the radical terms:

—?P- = a *2i!£.—(4) -dHC n, k“c [HC]

Equation 4 is a departure into relative kinetics. Not only does it remove terms about which we have little or no information, it also eliminates or greatly reduces the problem of describing the effects of unknown trace inhibitors. Since there is good reason to believe the butane LPO reactor is thoroughly backmixed (in a material sense), we may convert differential Equation 4 to its algebraic analog: 55

(5) -AHC nP C [HC]

Equation 5 is directly usable. If we assume that a * n»r and are relatively constant, a plot of _[£]_ vs ^ should give a straight line with a slope of and an intercept of [HC] -AHC k'f

nP

The utility of Equation 5 can be expanded as follows:

AP . [Pj .nHC (6 ) -AHC k™ [HC] nP

kj S-AHQ*[P] (7) -AHC C AP*[HC]

The term ( AHC) [-P] jg a dimensionless number that is, in effect, a turnover ratio. I have AP*[HC] chosen to call this number oxidation exposure (OXEXP)*. It is a measure of the relative ex­ posure of the surviving primary product to oxidation conditions compared to the exposure of the consumed hydrocarbon. For example, consider the case where the concentration of hy ­ drocarbon in the solvent is 1 mole/L. If, during the period where 1 mole of hydrocarbon is consumed for each liter of reaction solvent, 2 moles of P are isolated for each mole/L of P in the reaction solvent ( AP = 2 * [P]) > the oxidation exposure of P is:

(-D*m (8 ) Now note that: CEFF(P) AP*n p (9) 100 ~ -AHC*n llc where CEFF{P) = 0/° °f carbon in consumed HC isolated as P in the product From Equations 7 and 9:

CEFF(P) (10) 100 kr 1 + jW* OXEXP(P)

100 1 —+ * OXEXP(P) (11) CEFF(P) a*kt 56

100 From Equation 11, a plot of OXEXP(P) vs should be a straight line with an in- CEFF(P) 1 kl‘ tercept of — and a slope of __ m . a a* k"‘ Equation 10 can be expanded to cover any relationships of intermediates. The general equation is:

CEFF(x) '-I u-1 (12) 100 1 + ^r *OXEXP(x)

where x = product component ij = all precursors ofx, both direct and indirect a(i j) = fraction of carbon in consumed i which is converted to j when i oxi ­ dizes — fraction of i that is oxidized %%r*OXEXP(i) h* ------(i3) l + ^r *OXEXP(i)

The EST term in Equation 12 is required for components that are alcohols, acids or esters. The term is important because alcohols, in general, oxidize much faster than the acids with which they react or the esters that are formed and also give different products. The EST term is given by: £rr s * (kc%, * [ale] * [acid] - kM * [es/] * [H2Q])*n„ (H) -AHC*n., c*RF where s = sign selection term: it = 0 for components that are not alcohols, acids or esters. It = +1 for esters and -1 for acids or alcohols, jt = esterification rate constant k = hydrolysis rate constant RF=Rousing Factor, L of total volume/L of liquid volume

In a system where fractionation of components occurs between the reactor solvent and the product, each component can have its own individual OXEXP. Fortunately, however, the ratios of product OXEXPs tend to remain relatively constant over moderate ranges of OXEXPs. As a result, one can designate some major component as the standard and refer to its OXEXP as the system oxidation exposure (SOXP). We have chosen to use MEK as the standard. A usable approximation of the OXEXP of an individual component then becomes:

OXEXP(i) = OXPR(i)*SOXP (15)

where OXPR(i) = oxidation exposure ratio of i relative to MEK. 57

OXEXP is a completely general term that is independent of whether or not fractionation occurs in product isolation. In the special case where there is no fractionation between reac­ tor solvent and product, OXEXP is related to conversion as follows:

%CQNVERS1QN OXEXP= de) 100 - %CON VERSION

%CQNVERS1QN OXEXP (17) 100 1 +OXEXP

As conversion goes from 0 to 100, OXEXP goes from 0 to co ­

A plot for ethanol is shown in Figure 1. This is a very important intermediate in butane LPO. Several studies, including *

CH3—CH2—CH—CH3 —>■ CH3—CH2- + CH3—C ;o2 H CH3—CH2—00" CHj—CH2—O’ CHj CH2OH

♦ R. Although some of the s-butoxy radicals may abstract hydrogen to form s-butyl alcohol, the major fate, by far, must be the jg-scission reaction (Equation 18). In contrast, the ethoxy radical precursor of ethanol must show a large preference for hydrogen abstraction as op ­ posed to yg-scission. These proposals are supported by literature reports 3-4. Ethanol oxidizes quickly and efficiently to acetaldehyde. Acetaldehyde, which has a few additional sources in butane LPO, is obviously the major isolable intermediate (almost half of the carbon in consumed n-butane goes to or through AcH). It oxidizes so rapidly, however, that very little can actually be isolated. Figure 1 shows data for several years of recent plant operation. One important point is immediately clear; the points available are inadequate to define the y-intercept within ac­ ceptably narrow limits. The intercept shown is based on the aforementioned '4C data as well as runs conducted quite some time ago in pilot plant and bench-scale units. In the previous runs, we were able to reduce the SOXP to below 0.3 and even less. The resulting plots were compatible with Figure 1. An additional consideration proves useful in assigning a(ij) factors to product compo ­ nents and their precursors: All a(ij) values for each j must sum to exactly 1.000 in order to maintain a material balance. 58

Ethanol, of course, esterifies to some extent in butane LPO. This has been taken into ac­ count in Figure 1 (which is based on extensions of Equations 12 and 14). Apparently about 15 to 20% of the ethanol formed is esterified before it can be oxidized; about 65 to 70% is oxidized (almost exclusively to acetaldehyde) and the rest is isolated. Ethanol appears to be about 6 to 7 times as susceptible to oxidation as n-butane. This is consistent with literature reports ’. While it has been proposed that some esters are formed directly by oxidation reactions 6-7, in particular by the Bayer-Villiger Oxidation of ketones by peracids, more persuasive reports indicate that esters are not formed in significant amounts until after alcohols and acids ap- pearw.w. The latter findings are supported by the report that the expected peracids do not react rapidly with the ketones present 11-12. Moreover, the expected peracid concentration would be very low in a catalytic, limiting rate system 1. Lactones appear to be the only sig ­ nificant exceptions to the rule of no appreciable production of esters by free radical paraffin oxidation reactions. Perhaps the strongest evidence that the Bayer-Villiger reaction with ali­ phatic ketones is not a significant factor in butane LPO comes from our unpublished "C studies. When 4-'4C-2-butanone was fed to a butane LPO reaction, the specific activities of ethanol and ethyl acetate were vanishingly small, accounting for only about 0.3% of C atom 4 in the consumed methyl ethyl ketone (MEK). This accounts for only about 0.1 to 0.2% of the ethanol (free plus esterified) formed. s-Butyl alcohol is another important primary product in butane LPO (see Figure 2). About one-fifth of the n-butane appears to go to s-butyl alcohol as the first isolable interme­ diate. It appears to oxidize about 15 times as fast as n-butane. It probably arises by two mechanisms: 1) abstraction of hydrogen by s-butoxy radicals and 2), the Russell mecha­ nism13.

R R R R (19) R—C R—C >9 —o H ° °G H ' o c—o C—O R R

Two alkylperoxy radicals participating in the Russell mechanism need not be identical; it is only necessary that one of them have an a -hydrogen. It is not obvious, therefore, how much s-butyl alcohol is made by hydrogen abstraction and how much by the Russell mechanism. Two alkylperoxy radicals can also react in a non-terminating step that produces alkoxy radicals:

R, COO. •OOCR3 ------02 + 2R3CO. (20)

For two tertiary alkylperoxy radicals, Equation 20 is essentially the only bimolecular radical reaction possible. Even when the Russell mechanism can occur, alkoxy radical production can still be a major pathway. Alkoxy radicals can also be produced by reactions of hydroper ­ oxides and catalyst ions:

R3 COOH + M+n ------»- R3CO. + OH-+ M+n+1 (21) 59

i

MEK is a mixed primary-secondary intermediate. In a primary sense, it comes directly from s-butylperoxy radicals via the Russell mechanism. The secondary portion comes by the oxidation of s-butyl alcohol. This oxidation is very efficient and quite fast. It is so fast that, above rather low oxidation exposures, MEK can be treated as a “pseudo-primary" (see Figure 3). About one-eighth of the carbon in consumed butane generates primary MEK. Including the carbon that is converted to MEK through s-butyl alcohol, almost one-third of the carbon j in the consumed n-butane generates MEK; often only about one-half or less of this survives in the product. So far we have discussed the major primary and secondary intermediates that arise from initial attack on the secondary hydrogens of n-butane. This appears to be about 85% of the total. The other 15% of the initial attack occurs on the primary hydrogens of n-butane. This relatively high ratio of primary/secondary attack indicates that alkoxy radicals are major con- i tributors to hydrogen abstraction". Since acylperoxy radicals are much stronger hydrogen abstractors that alkylperoxy radicals':, they are likely to be in this group as well. The n-butyl radicals produced by initial attack on methyl groups will react rapidly with oxygen to produce n-butylperoxy radicals. When these participate in the Russell mechanism, n-butyl alcohol and/or n-butyraldehyde can be produced. An intriguing minor product, y -butyrolactone (GBL), seems to provide evidence that a large part of the traffic generated in terminal attack may pass through n-butoxy radicals. GBL is substituted in the 1,4 positions. It appears very unlikely that this could be the result of two independent attacks on the same butane molecule 4. Such attacks are possible but the methylene groups should be favored over the remaining methyl group at least about 90% of the time. It is strongly implicit that the second attack occurs through a backbiting mecha­ nism. It is unlikely that the backbiting would occur largely through n-butylperoxy radicals because 1) alkylperoxy radicals are not vigorous hydrogen abstractors and 2) to the extent n- butylperoxy radicals do backbite, the number 3 position should be favored because methylene groups are more reactive and that position is favored by a six-membered ring intermediate, n- Butoxy radicals, on the other hand, are vigorous hydrogen abstractors and the number 4 posi ­ tion is favored by a six-membered ring intermediate. Once the backbiting has occurred, a path to 4-hydroxybutanal is fairly obvious. The ring hemiacetal of this component should be easily converted to GBL. Two observations lend support to this proposal: 1) GBL correlates well with primary and pseudo-primary intermediates (i.e., its non-hydrocarbon precursors are rapidly consumed; see Figure 4) and 2) it has been reported that the generation of n-butoxy radicals (albeit in the vapor phase) in the presence of oxygen results in some formation of GBL". Primary alkoxy radicals containing more than one carbon atom are eligible to participate in the ^-scission reaction but, based on the efficiency of ethanol formation, do not appear to do so readily. n-Butoxy radicals should undergo ^-scission more easily than ethoxy radicals but, based on mass balance considerations and the proposed mechanisms, hydrogen abstrac­ tion would still be expected to predominate. This proposal is supported by literature re­ ports' 4. Both * 4C studies and kinetic plots indicate that primary and secondary alcohols are rap­ idly converted to the corresponding carbonyl compounds in very high efficiency (>90%). The initial attack on such alcohols is directed toward the hydrogen connected to the carbinol carbon atom. When oxygen adds to the resulting radicals, 1-hydroxyalkylperoxy radicals are produced. In water solvent, these peroxy radicals are reported to decompose readily to gener ­ ate a carbonyl compound and hydroperoxy radicals".'8 : 60

OH OH R +H+ R-C-OO. + H+ C -O R O. R + HOO.

(22)

It seems likely that a similar reaction would occur in other proton containing solvents. The HOO* radicals produced are known to be powerful inhibitors of free radical oxidations. This seems to explain why alcohols cooxidize rapidly in hydrocarbon oxidations but actually inhibit rates, sometimes even stopping the reactions. n-Butyl alcohol, produced as a result of terminal attack, ester!fies to a minor extent but most of it is oxidized to butyraldehyde (perhaps -90%). Butyraldehyde is rapidly oxidized to butyric acid in perhaps 60 to 70% efficiency. Degradation products include n-propyl alcohol, CO and CO^. < Acetone is an interesting intermediate in butane LPO. Some of it can readily be ex­ plained as coming from the isobutane present in all commercial butane. This, however, can­ not account for more than about one-fourth of the amount seen. Some acetone seems to come from the oxidation of butyric acid via secondary attack on the number 3 carbon atom. On the basis of abundance and reactivity, this can explain no more than about an additional one-third of the amount seen; about 40% or more is still to be accounted for. The most likely sources seem to be backbiting reactions of n-butylperoxy and/or n-butyrylperoxy radicals. Reactivity considerations may favor the butyrylperoxy case. As noted, aldehydes are major intermediates. They are the predominant precursors of the corresponding acids. They appear to be about 50 to 75 times more susceptible to oxidation than n-butane. Acetaldehyde is on the highest traffic pathway. It accounts for almost 50% of the consumed n-butane carbon and is the precursor of about 85% of the acetic acid. It pro ­ duces acetic acid in about 80-85% efficiency. Acetic acid is remarkably resistant to oxidation; it is only about 4% as susceptible to oxi ­ dation as n-butane. Acetic acid is resistant to attack by oxygen centered radicals because of polarity effects15. The carboxyl group appears to exert a retarding effect on the a -position. Based on relative methanol and formic acid specific activities in products from butane LPO conducted in the presence of 2-”C-acetic acid, roughly about 80% of the attack on acetic acid occurs on the methyl group while less than about 20% occurs on the carboxyl group. Attack on the methyl group could produce formaldehyde (and, subsequently, formic acid) without going through methanol. Attack on the carboxyl group would be expected to produce metha­ nol (or, at least, methyl radicals) first. In a number of l4C studies, formic acid and methanol have almost equal specific activities indicating an important common precursor (probably methyl radicals). Propionic acid is a minor product that seems to come from propyl alcohol, propionalde- hyde and MEK. It is about 5-10 times as susceptible to oxidation as acetic acid. This indi­ cates the suppressing effect of the carboxyl group is less effective on the yy-position. In addition, of course, the a -methylene would be more reactive than the a -methyl in acetic acid. Actually, it is possible to argue that the yy-position may even be somewhat activated, but this question should still be regarded as open. 61

Butyric acid seems to be about 1 to 1.1 times as susceptible to oxidation as n-butane. This is even stronger support for the proposal that the yg-position may be somewhat activated. Such activation would favor the production of acetone by oxidation of butyric acid, but the production of acetone does not necessarily establish the activation. Other products from the oxidation of butyric acid are believed to include acetic acid, carbon dioxide and propionalde- hyde. Some literature reports indicate that attack on both a and p positions by alkylperoxy radicals is suppressed by the presence of a carboxy or a carboalkoxy group with p attack being more suppressed than a attack15-20. This conclusion is open to question, however, since the experimental approach does not distinguish between a attack and attack by oxida ­ tion of the carboxy group. As a result, the indicated a attack may be inflated. Out data sup­ port the possibility that a attack by alkoxy and acylperoxy radicals is suppressed by perhaps a factor of about 2 while p attack is enhanced by a factor near 2. This is concordant with the reported enhancement of p attack by hydroxy radicals21. Attack at positions more distant than p is apparently unaffected by the presence of a carboxy or carboalkoxy group 19 -20. The immediate precursor of formic acid appears to be formaldehyde. Formaldehyde, to a large extent, comes from methanol and/or methyl radicals; very little formaldehyde is found in the product, so it seems to oxidize rapidly. Formic acid esterifies about ten times as fast as acetic acid so a larger portion of the formic acid produced appears as esters than is the case for acetic acid. Formic acid and formates are perhaps 10 to 20 times more susceptible to oxi ­ dation in the butane LPO system than are acetic acid and acetates. Since, however, they are more volatile and extractable than acetic acid and acetates, they also have lower exposure to oxidation. Of the acetic acid (free plus esterified) formed, about 96% survives in the product. For formic acid, the comparable figure appears to be about 55-60%. The liquid phase oxidation of butane to produce acetic acid proceeds through an entan­ gled labyrinth of series and parallel paths. Under practical process conditions, the reaction appears to proceed by a limiting rate mechanism in which peroxides can be treated as tran­ sient intermediates. Kinetic chains are very short and the major hydrogen abstracting radicals appear to be alkoxy and acylperoxy radicals. Isolable intermediates can be described in terms of primary, secondary etc. Recovered products are fractionated with respect to the reaction solvent by phasing processes that occur during recovery. Relative kinetics and a dimension ­ less turnover ratio, which is a measure of exposure to oxidation conditions, permit the appli ­ cation of relatively simple kinetic expressions. This has made possible the development of a mathematical model that has given new insight into the complex processes of paraffin oxida ­ tion.

REFERENCES

1. C. Walling, j Am. Chem. Soc ■ 91,7590 (1969). 2. C. C. Hobbs, T. Horlenko, H. R. Gerberich, F. G. Mesich, R. L. Van Duyne, J. A. Bed­ ford, D. L. Storm and J. H. Weber, i„± Eng. Chem. Proc. Des. Dev. H, 59 (1972). 3. G. Franz and R. A. Sheldon, in Ullmann’s Encyclopedia of Industrial Chemistry, 5th ed., Vol. A18, 261-311 (1991). 4. R. K. Jensen, S. Korcek, L. R. Mahoney and M. Zinbo, j Am Chem Soc 103(7), 1742- 1749(1981). 5. B. D. Boss and R. N. Hazlett, jncj £ng. Chem. Prod. Res. Dev. 14(2), 135-138 (1975). 6. B. D. Boss and R. N. Hazlett, Can. J. Chem. 47, 4175-4182 (1969). 62

7. F. R. Mayo, prcpr. Div. Pel. Chem. Am Chem. Soc ■ 19(4), 627(1974). 8. S. Blaine and P. E. Savage, jn± Eng. Chem Res. 30(9), 2185-2191 (1991). 9. S. Blaine and P. E. Savage, jnj_ Eng. Chem. Res. 31(1), 69-75 (1992). 10. S. Blaine and P. E. Savage, Prepr Div. Pet. Chem., Am. Chem. Soc. 35,239-244 (1990). 11. G. R. Krow, Org. React. (New York) 1993,45,251. 12. C. C. Hobbs, Applied Homogeneous Catalysis with Organometallic Compounds> B. Comils and W. A Herrmann eds., Vol. 1,521-540, VCH (1996). 13. G. A. Russell, j_ Am. Chem. Soc. 79 > 3871 (1957). 14. C. Walling, personal communication, 1998. 15. C. Walling, Active Oxygen in Chemistry> C. S. Foote, J. S. Valentine, A. Greenberg and J. F. Liebman eds., Blackie Academic & Professional, New York (1995), p.32. 16. A. Heiss and K. Sahetchian, jnt_ j_ Chem. Kin. 28 > 531-544 (1996). 17. E. Bothe, D. Schulte-Frohlinde and C. v. Sonntag, j. Chem. Soc., Perkin Trans. 2 1978(5), 416. 18. E. Bothe, M. N. Schuchmann, N. Man, D. Schulte-Frohlinde and C. v. Sonntag, photo- chem. Photobiol. 1978,18(4-5), 639. 19. W. Pritzkow and V. Voerckel, Qxid. Commun. 4(1-4), 223-228 (1983). 20. W. Pritzkow and V. Voerckel, j pra kt. Chem. 326(4), 572-578 (1984). 21. F. R. Hewgill and G. M. Proudfoot, Aust. J. Chem. 34(2), 335-342 (1981).

FIGURES 1-4 EtOH Tuning Plot October 23,1995, to March 30,1998 C. C. Hobbs - Hamburg Presentation 30-f

25-;-

c is - :* to 10-i-

> 5-

0.5 1.0 1.5 System Oxidation Exposure Figure 1. 63

s-Butyl Alcohol Tuning Plot October 23,1995, to March 30,1998 C. C. Hobbs - Hacburg presentation 200 -f

= ISO­

'S 100

I 50

System Oxidation Exposure Figure 2

MEK Kinetic Plot October 23,1995, to March 30,1998 C. C. Hobbs - Ha-ibtiro Presentation

n Mixed Primary SeC(Mi3SiyPJ$ o 5

"Primary "Plot

System Oxidation Exposure s-Butyl Alcohol Esterification Ignored Figure 3

GBL Ceff vs Butyric Acid Ceff October 23,1995, to March 30,1998 1.1-f 5. - S; 7 r f I

1.0-

O 0.9-

Pseudo-Prinjary Kinetic Plot

0.7-L

Butyric Acid Ceff, % Figure 4 64 65 DE99G2088 ^*DH012198149* i DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

H. Bohnke, J. C. Petzoldt, B. Stein, C. Weimer, J. W. Gaube Institut fur Chemische Technologie der TU Darmstadt, Petersenstr. 20, D-64287 Darmstadt, Germany

MODELLING OF THE PARTIAL OXIDATION OF a, P-UNSATURATED ALDEHYDES ON Mo-V-OXIDES BASED CATALYSTS

Abstract A kinetic model based on the Mars-van Krevelen mechanism ptf that allows to describe the microkinetics of the heterogeneously catalysed partial oxidation of a, p- unsaturated aldehydes is presented. This conversion is represented by a network, composed of the oxidation of the a, p-unsaturated aldehyde towards the a, p- unsaturated carboxylic acid and the consecutive oxidation of the acid as well as the parallel reaction of the aldehyde to products of deeper oxidation. The reaction steps of aldehyde respectively acid oxidation and catalyst reoxidation have been investigated separately in transient experiments.

The combination of steady state and transient experiments has led to an improved understanding of the interaction of the catalyst with the aldehyde and the carboxylic acid as well as to a support of the kinetic model assumptions. ^

1. Introduction The development of a kinetic model is usually focussed on the goal to provide a suitable tool for the optimization of a production process. The set of kinetic equations I describes the particular operation state as a function of the partial pressure of reactants and the temperature. On the other hand a detailed kinetic model can provide an improved insight into the reaction mechanism and the specific action of the catalyst particularly if steady state and transient experiments are combined.

2. Experimental 2.1 Steady State Experiments A simplified flow diagram of the experimental unit for the steady state experiments is shown in figure 1. For a constant and exact mass flow acrolein (ACR) and acrylic acid (AA) are dosed with microprecison pumps and all gas streams are adjusted by mass flow controlers. Acrolein and acrylic acid are directly injected into an evaporator

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 66

which serves also as a mixer for all gas streams. The presence of water vapor is necessary to prevent the monomers from polymerisation at the elevated temperature in the evaporator. The gas stream is conducted to the differential recycle reactor, realized by a jet loop reactor [2], which behaves like a CSTR. A small part of the gas leaving the reactor passes through the UN1VAP precision gas sampling system of the on-line GC [3], By this way of sample injection chromatograms of remarkable reproducibility are obtained. The main flow passes a cooling trap and is analysed with respect to oxygen, carbon monoxide and carbon dioxide. The pressure in the unit is kept constant by a pressure controler. acrolein

jet loop reactor Ni ■ condensation

O, evaporation V acrylic acid GC~.~Fip Oi/CO/COT . HiO Nt saturation —fJL— continuous analysis and data registration Figure 1: Experimental unit for steady state kinetic measurements

2.2 Transient Experiments Transient experiments and sorption studies were carried out in a static adsorption apparatus, as schematically shown in figure 2.

Dosing of 0%, H2. H20 .acrolein, XjC p|R acrylic-acid and Ar

; Gas- 'multiposition :—Chromatograph lH valve sample- y loops .mass- : spectrometer oven , online MS adsorption

vacuum

Figure2: Experimental setup for adsorption and transient kinetic measurements 67

The apparatus has been designed to render possible both sorption studies and transient kinetic experiments under reaction conditions, similar to that of the industrial process. Furthermore, the reduction degree of the catalyst can be adjusted and evaluated (by reduction and back-titration with oxygen, respectively). The gas composition in the adsorption/reaction cell is determined by GC/MS. Carbon equivalents adsorbed on the catalyst are calculated from the carbon balance. By use of a trapping type multiposition valve, small samples are taken at short time intervals to be analysed consecutively by GC/MS. This allows to take advantage of the high accuracy of GC/MS analysis and avoiding large time gaps between data points. Stability of the sample during storage in the sample loop has been proved. Moreover, switching to on-line MS (by using a direct transfer-line to the MS) allows a quasi continuous measurement of the gas composition in the cell.

2.3 Preparation of the catalyst A Mo-V-oxide based catalyst was prepared according to the patent specification EP 17000 [4], To avoid mass transfer limitation pellets of egg shell type with a thin active layer of about 200 pm thickness were used for the presented experiments.

3. Results As the jet loop reactor behaves like a CSTR the rates of consumption and respectively the rates of formation of products can be calculated by: ri mca The rate of oxygen consumption is calculated on the basis of the oxygen balance:

1 - flAOt) + (”.U —flAm** ~ i>IIOAc ~ ~ «CO ~ ”("0, j + lb, —----

1 2 Mica, The rate of water formation obtained from the hydrogen balance is: [a°i,o ~”//,o) = [3 £(#.«•* —”3ck ) + (»Ui -«!«) + «,<«/■»*]+ 2n//afcj The kinetics of the acrolein oxidation can be described on the basis of the Mars-van Krevelen mechanism which expresses the separated reduction of the catalyst by the compound which is oxidized and the subsequent reoxidation of the catalyst by supplied oxygen. Figure 3 shows the rate of acrolein consumption as a function of oxygen partial pressure. At low oxygen partial pressures the reoxidation is rate determining whereas in the range of elevated oxygen partial pressures the rate of 68

conversion is nearly independent of the oxygen pressure, indicating that the oxidation of the substrate, respectively the reduction of the catalyst, is the rate determining step.

4.0 kPa 7 = 280 *C, p°200kP;

2.6 kPa

1.4 kPa

variation p(acfotein)

Figure 3: Oxidation of acrolein: -rACR = f(p02)‘. parameter p ACR; solid lines: calculated The oxidation of acrolein was found to be independent of the concentration of acrylic acid also the rate of by-product formation which is approximately given by the rate of consecutive oxidation of acrylic acid, turned out to be independent of the concentration of acrolein. For further interpretation of the results of steady state kinetic studies transient experiments were carried out in which the oxidation of acrolein was observed in absence of molecular oxygen. The course of acrolein concentration at 230 °C is shown in Figure 4 A. T = 230 *C unloaded catalyst preloaded 10 catalyst 5 with acrylic acid 4 B 9- 3 ' 2 1 0. 0 2 4 6 8 1012141618 20 0 2 4 6 8 101214 1618 20 time (s) time [sl 10 12 8 * 6^10 6- "\* 8 ■ a _-T ▼ 6 4- - .4 2 2 0-w T 0 40 80 120 780 810 840 0 40 80 120 600 900 1200 time [si time [sl acrylic acid Figure4: Transient kinetics of acrolein oxidation (total pressure and partial pressure) A: partly reduced catalyst without pre-adsorption B: catalyst preloaded with acrylic acid 69

After a very fast drop in acrolein concentration and total pressure, that may be attributed to the adsorption of acrolein, the slope is strongly decreased and the appearence of acrylic acid in the gas phase is observed. In order to elucidate the nature of these two sections, experiments with a second pulse of acrolein (figure 5) after building up the adsorption layer or with pre-adsorption of acrylic acid (figure 4 B) were performed. When the second pulse of acrolein is introduced into the cell the first section of fast adsorption becomes almost negligible and only the slow reaction accompanied with the formation of acrylic acid takes place.

—acrolein r acrylic add

Figure 5: Transient kinetics of acrolein oxidation; partial pressure after a second pulse of acrolein (T = 230 °C) Therefore it can be assumed that at first an extended part of the surface is densely covered by an adsorbate. Subsequently this part of the surface is not achievable for acrolein, which then reacts on a smaller part of the catalyst surface with high selectivity towards acrylic acid. The acrolein oxidation carried out after pre-adsorption of acrylic acid (fig. 4 B) shows the same course of acrolein consumption as in the preceeding experiment. This result allows the conclusion, that the adsorbate on the major part of the active catalyst surface is a kind of acrylic acid intermediate which can either be formed by the oxidation of acrolein or by pre-adsorption of acrylic acid. The derivation of the kinetic model is based on this well founded assumption that the catalyst surface shows two domains. The smaller part is highly active with respect to the oxidation of acrolein to acrylic acrylic which itself is easily desorbed from these sites. On the other hand a greater part of the surface is covered by a strongly adsorbed acrylic acid species, as discussed by Davydov et al. [5], which is slowly oxidized to by-products. The steady state experiments reveal that the rate of acrolein consumption passes through a maximum as the acrolein concentration is raised (fig. 6 A). This maximum 70

T= 280 °c 37 kPa 19 kPa 20 kPa

variation p(H 20) Variation p(0 2)

6kPa 8.5 kPa

p (acrolein) [kPa]

Figure 6: Oxidation of acrolein, steady state kinetics A: -r = f(PAca) parameter p(0 2); solid lines calculated B: -r = f(PAC(0 parameter p(H 20); solid lines: calculated appears in the range where the reoxidation of the catalyst is rate determining so that this effect suggests an adsorptive blocking of reduced sites of the catalyst by acrolein and thus a hindrance of the reoxidation of the catalyst. The transient experiments have shown that the reaction rate of acrolein consumption remains nearly constant even though the amount of adsorbed species reaches saturation. That means that the oxidation of acrolein by the catalyst remains unchanged while the reduced sites may be blocked by adsorbed species. Figure 6 B shows that in the range where the reoxidation of the catalyst is rate determining the rate of acrolein consumption increases with elevated water vapour pressure.

Figure 7: Amount of adsorbed water as function of p(H20) and reduction degree (R) 71

This result leads to the conclusion that adsorbed water enhances the activation of oxygen on reduced sites. This is in line with the observation that the water adsorption increases strongly with the degree of reduction of the catalyst (figure 7). Novakova et al. have found that in the oxidation of acrolein oxygen from water molecules, namely H2018 , gets incorporated into acrylic acid [6]. Therefore oxygen from water molecules becomes inserted into the oxygen vacancies, i.e. the reduced sites of the catalyst.

The reaction network of the acrolein oxidation includes the main reaction towards acrylic acid (r,), the negligible parallel reaction (r2) and the consecutive reaction (r3), which both contribute to the formation of by-products. The presented kinetic model takes into account the observed independence of the two reaction steps, i.e. the oxidation of acrolein and the consecutive oxidation of acrylic acid taking place at two different reactive sites. Therefore two oxidation degrees, ®ACR and M, are defined that stand for certain degrees of coverage of the catalyst surface with reactive oxygen following the theory of Mars and van Krevelen [1], The rate of reoxidation is a function of the oxygen partial pressure and the degree of reduction. In the case of acrolein oxidation the reoxidation is hindered by acrolein. As it was shown in figure 6 B water vapour has an accelerating effect on the rate of reoxidation of acrolein sites which is included in the rate constant ktfR . Oxidation of acrolein r, =krPACx-G>ACK r2=k2-pACR-®AC* -rtf* = ktf*-tf 0\ -p7 CR (1 - O'") = 1/2 -r, - v2 2 with

fast adsorption and reaction of acrolein the catalyst is completely covered by reaction products. Nevertheless the oxidation of acrolein towards acrylic acid occurs continuously unaffected by acrylic acid. The reaction rate is the same observed for a catalyst with pre-adsorbed acrylic acid. The immediate pressure drop observed for the unloaded catalyst is much higher than in the case of the pre-loaded catalyst. This observation leads to the conclusion that the domain on which in the steady state the oxidation of acrolein occurs is much less extended than the domains covered with acrylic acid respectively with precursors such as acrylate ions, as discussed by Andrushkevich [7], on which the consecutive oxidation towards by-products occur with a very low rate. The different behaviour of these domains suggests different structures or even different phase composition. XRD and TEM studies on the structure of MoA/-oxide catalysts by Schlogl et al. [8] have shown rather complicated phase compositions and distorted domains. Since the kinetic studies show that the domain of acrolein oxidation forms only a small part of the catalyst surface the assignment of the active domain to a specific phase or distorted region will be a difficult task. The results of the steady state experiments (fig. 3,6) could be well represented by the kinetic model (acrylic acid consumption is not shown here). The effect of water on the rate of reoxidation can be traced back to a direct role of water in the mechanism of reoxidation or/and the displacement of species adsorbed on reduced sites by water. Both interpretations find support by transient experiments.

References

[1] P. Mars, D.W. van Krevelen, Spec. Supp. Chem. Eng. Sc/., 3,41 (1954). [2] G. Luft, 0. Schermuly, Chem.-Ing.-Tech., 49,907, (1977). [3] W. David, Labo, 7-8, 62 (1994). [4] BASF AG, European-Patent 0017000 A1 (1980). [5] A. Davydov, T. Andrushkevich, React. Kinet. Catal. Lett., 25,1175 (1984). [6] J. Novakova, K. Habersberger, React. Kinet. Catal. Lett., 4, 389, (1976). [7] T. Andrushkevich, Catal. Rev.-Sci. Eng., 35, 213 (1993). [8] R. Schlogl, H. Werner, O. Timpe, D. Herein, Catal. Lett., 44,153 (1997). 73 DE99G2087

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

H *05012198158* A. J. Nagy, G. Weinberg, E. Kitzelmann, G. Mestl " Fritz-Haber-lnstitut der Max-Planck-Gesellschaft, Faradayweg 4-6, D-14195 Berlin, Germany , >. 7 /$

THE MORPHOLOGICAL MODIFICATION OF ELECTROLYTIC SILVER DURING THE OCM REACTION AND IT'S EFFECT ON CATALYSIS

Abstract Strong support for the existence of catalytically-active, sub-surface oxygen is proposed. The existence of two types of sub-surface oxygen was determined by thermal-desorption spectroscopy (TDS). The first is termed Op and is characterized by a broad thermal- desorption peak centered at approximately 773K. The second is referred to as O? and is characterized by an unsaturated thermal desorption signal beginning at approximately 873K. Or is assigned to oxygen which is incorporated in the uppermost layers of low-indexed terminating crystal structures such as (111) formed as a result of thermal reordering. Op diffusion occurs most likely via interstitial jumping and 0Y via interstitralcy diffusion. Both CH< conversion and C% selectivity improve with time on stream. This improvement in catalytic performance correlates well with surface facetting and particle rupture, which occurs as a result of stress buildup from oxygen dissolution in the silver crystallite

II. Introduction Silver has long been recognized as a partial oxidation catalyst of immense industrial importance. A great deal of research has been directed, therefore, to the elucidation of the oxidation of methanol to formaldehyde and the ethylene epoxidation reactions. 1,13 Despite years of intensive study, questions about the fundamental catalytic mechanisms of these reactions are still being debated. Despite differences of opinion, the general consensus is that a solid understanding of the silver-oxygen interaction is critical to uncovering the secret behind the exceptional performance of silver as partial oxidation catalyst. A large body of literature regarding this interaction has also accumulated over the course of the last 2 decades.4A6The hypothesis, suggested, in this study differentiates itself from the majority of published work in that the importance of a sub-surface species which plays a dynamic role in catalysis is proposed. A number of authors have ascertained that sub-surface oxygen is critical for both the ethylene epoxidation and the partial oxidation of formaldehyde. 3,4^!! has been suggested, in the overwhelming majority of these studies, that the modification of the surface electronic properties of silver and various adsorbed species by sub-surface oxygen is necessary to selective oxidation. We suggest here that oxygen plays a dynamic role in the oxidative coupling of methane where it is continuously formed and consumed in reaction. This is believed to be largely due to the extremely high mobility of oxygen in the silver lattice under these reaction conditions 7,8,9 . Both the OCM reaction and the methanol oxidation reaction are active under similar reaction conditions T>923K. Both are dehydrogenation reactions where a thermally-unstable end product is desired. The advantage of the OCM reaction is that the low reaction conversions enable a reliable determination of the reaction kinetics. The extremely high conversions obtained for the methanol oxidation reaction make it impossible to study the reaction kinetics.'-

We have proposed previously that the formation of subsurface oxygen is essential for the methanol oxidation reaction 9 '10,11. In particular, it was suggested that an oxygen species incorporated in the uppermost layers of (111) terminating crystal structures was the selective

DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 74

partial oxidation species. This species was coined Oy . The formation of Oy in significant amounts was found to be possible only at elevated temperatures (T>773K) and oxygen dosing pressures (P>0.01mbar). It’s presence and location in the (111) surface was confirmed using a variety of surface-science spectroscopies. 9 "10'11 Having established the nature and location of this species, the question of determining the mechanism of its formation arises. In particular, a rationalization of the observation that this species only forms in the uppermost layers of close-packed terminating crystal structures was desired. Finally, a clearer understanding of the dynamics of the formation of Or under reaction conditions was needed to supplement the body of data obtained under UHV conditions. The work presented here originates from a number of studies which were carried out with the goal answering these important questions.

III. Results The silver-oxygen interaction: Temperature-desorption spectroscopy is a well known tool for determining the presence and relative bond strengths of various surface-adsorbed species. The TD spectra shown here distinguish themselves from classical thermal- desorption spectroscopy in that the analysis of sub-surface species is considered. One must be wary of the fact, therefore, that the rate limiting process for desorption of a sub-surface species is likely limited by the rate of surface segregation of the dissolved species. This is evidenced by the fact that the frequency factors for bulk diffusion (lCf'-lO"5 s"1) typically lie many orders of magnitude lower than surface recombination and desorption processes (107-1015 s"1). The information obtained, therefore, from the TD spectra shown here are interpreted with the assumption of limiting bulk diffusion.

Figures la-c show a series of 0% TD spectra obtained after having dosed silver with Oi at a variety of temperatures and pressures. The desorption behavior is clearly a strong function of the dosing 400 600 800 1000 1200 conditions. Two desorption peaks are T (K) visible and are labeled Op and 0Y. The Figures la-c (Top-Bottom): O2 TDS of O2 desorption presence of surface-bound atomic oxygen from Ig Electrolytic Ag after dosing at various pressures Oa can be effectively neglected as the and temperatures. 75

dosing temperatures exceed the thermal desorption temperature of this species (623K)3-6-7-12. The large FWHM and assymetry of the Op peak are attributed to the fact that polycrystalline offers a variety of diffusion paths for oxygen (i.e. grain-boundary, crystallites of various packing density, etc.).

Figure la shows that the Op desorption temperature shows no dependence on the dosing pressure at 573 K. It does, however, shift progressively to lower temperatures with increasing dosing pressures for pretreatments made at 773 and 973K (Figures lb,c). A decrease in the desorption temperature implies a decrease in the resistance to bulk diffusion for the Op species. Again, it must be emphasized that the rate limiting step in the thermal desorption of a sub-surface species should be the bulk diffusion to the surface. The assignment of Op to a bulk-dissolved oxygen species is supported by three facts. Firstly, the maximum uptake of 6xl0'7 mol OVg Ag corresponds to approximately 30 multilayers of oxygen. Multilayer adsorption of oxygen is known not to occur at these temperatures 3-6'7,12. Secondly, even at the high-dosing pressures used here the Op is not saturated. Thirdly, Op desorbs at temperatures lower than the dosing temperatures used in Figures lb and c. This is only possible for the instance where oxygen is frozen into the bulk during the rapid cooling of the sample at the end of the oxygen pretreament. The shift to lower desorption temperatures indicates that increasing the dosing pressure results in morphological changes of the silver which result in decreased diffusion resistance. The dependence of the Op desorption temperature on the dosing temperature shows that the desorption temperature of Op increases with increasing dosing temperature. A comparison of the O.Olmbar 0% dose shows that the desorption temperature increases from 673K to 850K to 1050K for doses made at 573K, 773K and 973K.

The Oy desorption peak behaves entirely different. It increases, in all cases, only slightly with increasing dosing pressures. This implies that the Oy peak is confined only to the uppermost silver layers. Furthermore, the peak maximum appears to remain constant at 1000K. The quantity of desorbing Oy does, however, increase significantly with increasing dosing temperature.

The calculated activation energies of desorption based on Figures la-c are shown in Figure 2. It is important to note that these values do not represent the activation energies of individual mechanistic steps. They represent, rather the total energetic barrier to surface segregation, recombination and desorption. A minimum at approximately lOmbar Oz is found for all dosing temperatures. The relatively small variation in Eapp for doses made at Figure 2. Calculated values ofE^, for O2 desorption 573K correlates well with the fact that the after dosintt at various oressures and lemocralures. 76

desorption temperatures in Figure la remain more or less constant for different dosing pressures. The kinetics of silver transport at 573K are obviously not fast enough to allow significant restructuring on the time scale of a typical run. The case is changed, however, for dosing at 773K. Here, oxygen desorption goes through a pronounced minimum. 773K is apparently sufficiently high to allow efficient mass transfer of silver, but is not so high as to allow pure thermal reordering to dominate. The morphological reconstruction is essentially guided by the oxygen-induced reduction in surface energy which is kinetically prohibited for temperatures lower that 773K. This temperature corresponds excellently with the Tamann temperature of silver (753K) at which bulk-silver diffusion is activated. The increased values of Eapp for doses made at 973K indicate that thermal reordering and the formation of closely- packed surface structures dominate at this temperature resulting in surface annealing and subsequent diffusion inhibition. The decrease in E,pp for dosing pressures up to lOmbar O2 agrees excellently with the shift of the desorption maximum to lower temperatures seen in Figures la-c. The increase in Eapp above lOmbar agrees with the transformation of the facets to a (331) orientation as observed by in-situ XRD.9 This explains the reason why the desorption temperatures do not continue to shift to lower temperatures for doses made above lOmbar O2 (Figures la-c).

The behavior of both Op and 0Y may be explained by considering the high- temperature changes in silver morphology. Figure 3 shows the surface of a silver following pretreatment for 3 days in a stream of 10% O2 in He. The surface is extensively facetted. Facetting is not observed for samples pretreated in flowing He only. The facetting clearly arises as a result of the silver-oxygen interaction. Facetting occurs as a result of the desire of a solid surface to obtain a minimum in the surface free energy 12. Pure thermal reordering favors, therefore, the creation of densely-packed crystal structures which exhibit the lowest surface free energy. It is known however, that gas adsorption causes a decrease in the surface energy. Oxygen adsorption is highest on rough surfaces which are, thermodynamically speaking, energetically unfavorable. These surfaces exhibit, however, the highest sticking coefficients for oxygen adsorption ((111)<(100)<(100-))3,6,12. Treatment in oxygen, therefore, results in an adsorbate- induced decrease in the surface energy of loosely-packed terminating crystal structures. The result is the formation of surface facets composed most likely of (111) terminating terraces and (110) terminating steps 8,9 ‘12. The formation of these steps is favored at increasing dosing pressures. Higher dosing temperatures, in contrast, tend to result in pure thermal reordering and a decrease in the step height and number. It has been previously shown with in-situ XRD that oxygen diffuses preferentially in the [110] direction through the silver bulk.9 The (110) steps exhibit, therefore, not only one of the highest sticking coefficients for oxygen, but also provide a “window" for oxygen diffusion to the bulk. Equimolar counterdiffusion of oxygen to and from the bulk occurs. At elevated temperatures, close-packed terraces are 77

predominantly formed and present a diffusion barrier to Op segregation. The peak desorption temperatures in Figures la-c, therefore, shift to higher temperatures.

The fact that Oy has been observed with STM to EAg =45.95kJ/mol Ag incorporate into lattice sites in the -s (111) terraces10,11 implies that Op —O— Dxg (HcHman-Ref 20) undergoes interstitialcy diffusion —O—— D qj (Outlaw-Ref 6) *. This is supported by the data g _]0 presented in Figure 4, which is a c comparison of data on the ^_^Eo 2=64kJATiol Ag kinetics of oxygen diffusion in silver presented by Outlaw et al.13 -is ■Eo 2=465kJ/md Ag and silver interdiffusion by I EAg =26.4kJ/mcl Ag Hoffman et al14. In both cases, an , 923K f v625K inflection point in the temperature 0.8 1 1.2 1.4 1.6 ______1000/T(K)______dependence of the diffusion coefficient was observed. This transition Figure 4. A comparison of published dam regarding the diffusion kinetics of O2 diffusion in silver and agrees excellently with the onset of Oy silver interdiffusion. desorption seen in Figures la-c. Furthermore, the point at which the kinetics of oxygen diffusion in silver exceed the silver- interdiffusion kinetics at 573K agrees excellently with the onset of the Op peaks shown in Figures la-c. The similarity of the activation energies of diffusion for the species diffusing above 923K for both silver interdiffusion and oxygen diffusion in silver is excellent evidence for an interstitialcy diffusion mechanism. Both silver atoms and oxygen undergo identical diffusion steps for interstitialcy diffusion. In contrast to this, oxygen atoms are highly mobile relative to Ag for the case of interstitial diffusion, resulting in different activation energies of diffusion. It is, therefore, suggested that 0Y diffuses primarily by interstitialcy diffusion and Op by grain-boundary diffusion and interstitial diffusion.

The Kinetics of the OCM Reaction:

Figure 5 shows a series of Arrhenius plots obtained for the OCM reaction over silver. The results clearly show the differences in activity between fresh and used catalysts. There is an inflection point at 1020K in the plot for reaction over a fresh silver catalyst. The extremely high activation energy of 457kJ/mol agrees well with Figure 5. Comparison of Arrhenius plots for the OCM the gas-phase C-H bond strength of over fresh and used Ag with and without thinning in SiO>. CH4 (439kJ/mol). Fresh silver is apparently inactive below 1023K. The decreased activation energy of 67 kJ/mol above 78

1023K attributed to the simultaneous occurrence of a number of solid-state transformations which ultimately result in catalyst activation. The first and most dominant of these is catalyst sintering. The effect of sintering is seen by comparing the pre-exponential factors for reaction with and without inclusion of catalytically-inactive SiOi in the reactor bed. The pre-exponential factor is 2.76 times higher for the sample containing Si02 but exhibits the same activation energy. Catalyst sintering obviously results in the destruction of active sites for reaction. The activation of the catalyst is attributed to the facetting of the crystal surface resulting in Or formation and the rapture of the catalyst particles which results in an increased total surface area for catalysis. Facetting has already been discussed in the previous section in terms of the oxygen-silver interaction. The SEM micrograph shown in Figure 6 shows that the silver surface subsequent to 3 days treatment in the OCM reaction is nearly identical to the surface shown in Figure 3 after treatment in oxygen. Particle rapture also occurs as a result of the mechanical stress caused by the incorporation of oxygen into the silver bulk9 . The mean particle size was observed to decrease from 50-100pm for the fresh catalyst to 10-20pm after reaction (data not shown). The activation energy of reaction (138kJ/moI) and the Eapp calculated from the TDS data in Figure la-c (145kJ/mol) agree excellently. This is an excellent indicator that the rate-limiting step in reaction is the diffusion of Op in the bulk.

Figure 7 shows the results for reactant conversion and product selectivity for an isothermal ran made at 1023K over a fresh electrolytic silver sample. The initial oxygen conversion of 100% drops off exponentially to 15% over the course of 50 hours. The decrease in activity is complimented by a drastic change in selectivity. The initial 75% selectivity to CO: decreases rapidly and the selectivity to C2He increases, The slow variation of the conversion and selectivity are indicative of a 80 100 catalyst undergoing a slow, solid- state transformation. A .§• O comparison of SEM micrographs .2 60 ■ Oa Conversion 80 "5 of the surface after 2 hours on line • CO Selectivity and after 3 days on line show the ■ C02 Selectivity § 2 40 h 60 presence of holes exclusively at ■ C2H4 Selectivity the facet steps for the 2 hour ■ C2Hg Selectivity 40 ! sample (not shown). No holes are 20 found in the sample treated 3 days on line. The presence of holes ; 20 indicates the reaction of Op with dissolved hydrogen forming 10 20 30 40 50 water which raptures the surface. Time on Stream (hours) Op apparently forms Figure 7.02 Conversion and selectivity to selected products preferentially in the steps and for the OCM over fresh electrolytic silver. 79

initially reacts to complete oxidation products. The absence of holes correlates well with the improvement in selectivity after long times on line hinting that the orientation of the steps may change over time. It is, unfortunately, impossible to unambiguously determine the step and terrace orientation with SEM. It is clear, however, that surface facetting is necessary for the selective coupling of methane to Ci hydrocarbons. 10 20 30 40 Time on Stream (hours) The initial high selectivity to CO2 is not due to local overheating of the Figure 8 .02 Conversion and selectivity to selected products catalyst bed. This was proven by for reaction over fresh electrolytic Ag mixed with SiCh. cooling the bed and performing the same experiment. It was found that that the initial high selectivity to CO2 disappears and that the conversion immediately obtains the steady state value measured at the end of the previous experiment (data not shown). Local heating effects would be expected to be enhanced for the used catalyst which has sintered and shows a higher packing density.

Figure 8 shows the results obtained after mixing the catalyst with SiCfo. The behavior of the catalyst is dramatically altered. The initially low O2 conversion of 10% was not expected. The reason for this behavior was found only after performing a detailed SEM analysis. Figure 9 shows the extensive formation of surface facets by pinning at SiC>2 particles on the surface. Clearly, incorporation of SiCL in the Ag surface nucleates the formation of the previously mentioned surface facets. Absolutely no hole formation was observed for this sample. This correlates well with the exceptionally good selectivity to C2 products. The exact reason for the lack of hole formation in these facets is unclear. It may be a result of the steps having a different orientation than in the facets formed in the pure silver sample for short times on line shown in Figure 8. This is supported by the comparatively low conversion to COx products for short times on line. It would be naive to assume that all of the facets show identical terminating crystal structures. The data clearly show, however, that the improvement in catalyst selectivity is connected with facet formation. The decrease in activity likely results from the rapid formation of (111) terraces which show an exceptionally low sticking coefficient for O2 (<10'7) and hence tend not to form Oa which avors complete oxidation. More research is

Figure 9. Electron micrograph of surface obviously required to determine the exact pinning resulting from the incorporation of SiOj orientation of these facets as a function of particles in the surface after 2 hours in the OCM reaction conditions. 80

Figure 10 shows a suggested scheme for the high-temperature oxidehydrogenation of hydrocarbons over silver. This scheme is by no means intended to be representative of the mechanism as a whole. There are surely a large number of intermediates and side reactions taking place. The point is that reaction with Oa, Op and Oy leads, ultimately, to the formation of the products indicated. A more detailed understanding of silver-catalyzed reactions requires still more research. The work here has shown, however, that the silver catalyst is in a dynamic state of change during reaction. The reaction-induced morphological changes are intrinsically tied with diffusion phenomena leading to the formation of Op and Ot. This, in turn, has a direct impact on the catalytic activity.

III. Conclusions

It has been shown that the oxidative coupling of methane over pure electrolytic silver is an extremely structure-sensitive reaction. The formation of facets appears to be necessary for the selective dehydrogenation of CH4 to Ci products. This is attributed to the formation of an oxygen species which is incorporated in the uppermost layers of (111) terraces most likely via an interstitialcy diffusion mechanism. Fresh silver samples show a preferential tendency to form complete oxidation products. The dramatic change in silver selectivity and activity with time on stream is attributed to the oxygen-induced facetting and rupture of the silver particles. The excellent agreement between the activation energy of reaction and the activation energy of oxygen diffusion in silver suggests that the diffusion of bulk-dissolved oxygen is the rate- limiting step of reaction. Further work is necessary to determine the exact orientation of the facets formed during reaction. This work provides, however, the necessary foundation for understanding the structure sensitivity of high-temperature oxi-dehydrogenation reactions catalyzed by silver.

1 Ullman, in "Oilman's Encyclopedia of Indusirail Chemistry", 5th edition, Vol A11,619-651 (1988) 2 Spcrber H„ Chemie-lng.-Technology, 41(17), 962 (1969) 3 van Santen R.A and Kuipers H.P.C.E., Adv. Calal. 35,265 (1987) 4 Rovida G., Pratsi F„ Maglietta M. and Fetroni E., Surf. Sci. 43,230 (1974) 5 Backx C., de Groot C.P.M. and Bilocn P., Surf. Sci. 104,300 (1981) 6 Czandema A., Chen S. and Biegen J., J. Calal. 33,163 (1974) 7 Eichenauer W. and Pebelcr A., Z Metallkade. 48,373 (1957) 8 Schmalzried H„ in "Chemical Kinetics of Solids ”, VCH Verlagsgcsellschafl Weinheim (1995) 9 Herein D., Nagy A., Schubert H„ Weinberg G„ KitelmannE. and Schlogl R„ Z JurPhys. Chem, 197,67 (1996) 10 Bao X., Muhler M„ Schcdel-Niedrig Th. and Schlogl R.. Phys. Rev. B, 54,2249 (1995) 11 Bao X„ Muhler M., Schlogl R„ Ertl G., Cat. Lett. 32 285 (1995) 12 Wei Ta-Chin and Phillips J., Adv. Calal. 41,359 (1997) 13 Outlaw R.A., Wu D., Davidson M.R. and Hoflund G.B., J. Vac. Sci. Tech. A 10(4), 1497-1502 14 Hoffman R.E. and Turnbull D., J. Appl. Phys. 22,634 (1951) 81

DGMK-Conference, "Selective Oxidations in Petrochemistry", Hamburg 1998 ill *DE012198167* W. Weiss, Th. Schedel-Niedrig, R. Schlogl Fritz-Haber-lnstitut der Max-Planck-Gesellschaft, Faradayweg 4-6, D-14195 Berlin, Germany

STRATEGIES FOR CATALYST DEVELOPMENT: POSSIBILITIES OF THE “RATIONAL APPROACH" ILLUSTRATED WITH PARTIAL OXIDATION REACTIONS

Abstract DE99G2086 I The paper discusses two petrochemical selective oxidation reactions namely" the practised • formation of styrene (STY) and the desired oxidative functionalisation of propane. The present knowledge about the mode of operation of oxide catalysts is critically considered. The dehydrogenation of ethylbenzene (EB) should be described by an oxidehydration with water acting as oxidant. The potential role of the coke formed during catalytic reaction as co-catalyst will be discussed. Selective oxidation is connected with the participation of lattice oxygen mechanism which transforms unselective gas phase oxygen into selective oxygen. The atomistic description of this process is still quite unclear as well as the electron structural properties of the activated oxygen atom. The Role of solid state acidity as compared to the role of lattice oxygen is much less well investigated modem multiphase-multielement oxide (MMO) catalysts. The rationale is that the significant efforts made to improve current MMO systems by chemical modifications can be very much more fruitful when in a first step the mode of action of a catalyst is clarified on the basis of suitable experiments. Such time-consuming experiments at the beginning of a campaign for catalyst improvement pay back their investment in later stages of the project when strategies of chemical development can be derived on grounds of understanding^

Introduction

In recent years there are strong pleas that combinatorial chemistry could be applied to heterogeneous catalysis [1] and that such a spin-off from a successful research strategy in drug chemistry could be the desired strategy for the discovery of novel or greatly improved catalytic processes. In a critical analysis of the potential of combinatorial chemistry for heterogeneous catalysis [2] it was pointed out that several pre-requisites for the application of the combinatorial technique are not fulfilled in heterogeneous catalysis. One of the most serious problems is that we do not know in almost all practical cases which chemical reactions occur during a desired transformation of a feedstock molecule. We know educt and products but rely on speculations and chemical intuition when it comes to the description of the reactions necessary to achieve the educt-to-product conversions. It is thus difficult to conceive how a targeted catalyst development strategy should be applied if we have no sound ideas of what reactions we want to modify in their kinetics.

A less critical view of our understanding of heterogeneous catalysis leads to the enumeration of a series of chemical rules of what we have to do in order to achieve a desired modification of given uncatalysed reaction kinetics. In selective oxidation the lattice oxygen concept and catalyst promotion by auxiliary oxides according to the remote control hypothesis or the addition of alkali to the catalyst and of steam to the feed are such rules. They have proven useful to develop

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 82

heterogeneous catalysis to the high standard which we see today. These rules are, on the other hand, not founded on thorough understanding but merely the concentrate of the practical experience gained in over 70 years of empirical catalyst development. For this reason it can be foreseen that their uncritical further application will bring about the danger that earlier unsuccessful development pathways which were not documented for their negative nature will be repeated more and more often.

The present contribution illustrates how the application of targeted surface science which fully takes into account the possibilities of modem in-situ high pressure experiments can reveal new information on the mode of action of the iron-oxide catalysed dehydrogenation of EB to STY. As a consequence, a new strategy for considering this reaction can be derived which emerges in several development directions of this well-established petrochemical process. A second example will be the selective oxidation of a small alkane molecule Here we see that the understanding of the oxygen activation process and of the involvement of solid acid properties of the unknown surface structures of MMO catalysts are still inadequately developed.

The conversion of ethylbenzene to styrene

The reaction

C6H5CH2CH3C6H5CHCH2+ H2 (1)

is formally an endothermic dehydrogenation as no oxidant for the hydrogen is used in the process. The catalyst is principally a potassium-promoted iron oxide although a variety of other oxides are also active in the process [3]. During catalyst improvement additional promoter oxides with various ascribed functions have been added to the catalyst formulation. The catalyst function was initially thought to be a bi-phasic strong Broensted base consisting of liquid KOH supported on porous iron oxide of debated nature. In a later multi-method study [4,5] devoted to the structure of the active phase of the technical iron oxide catalyst it was found that this initial model termed "supported liquid phase" (SLP) describes the deactivated state of the phase- separated system KOH-magnetite-additional oxides. The active phase was found to be composed of two ternary potassium-iron oxides with different crystal structures. One of them KFe02 was found on the surface of the iron oxide crystals whereas the other ternary oxide, K2Fe22Ow which is isostructural to the spinel-phases of bare iron oxides, was found to be present in solid solution with a matrix phase of magnetite. The activity of the catalyst was found to scale with the surface abundance of KFe02 and with the absence of hydroxyl species ascribed to decomposed KOH. The mode of action of this MMO catalyst was, however, tentatively assumed identical with that of the SLP model to provide strongly basic surface sites which should activate the non-acidic C- H bonds of the ethyl groups. Better than in the SLP model was the explanation of the function of the iron III oxide as redox centre supporting the electron exchange from the organic substrate to the product hydrogen. For a catalytic cycle it was necessary, however, to assume a reaction

2 Fe(l,,-0-H -> 2 Fe(,m=0 + H2 (2)

Although reaction (2) is formally possible, it is chemically quite unlikely that at high temperatures .under chemically reducing conditions a hydroxilated oxide should produce molecular hydrogen and not water with a deeper reduced oxide. 83

i

None of these considerations take into account that any physically plausible reaction model must be based upon the surface structure of the active catalyst which is not known until now. The in- situ XPS studies were consistent with a surface termination structure of the bulk phase in the same composition as the bulk unit cell. This would call for a Fe3+ :K+ :012 " 3 surface 4 5 6 7 abundance in the ratio of 1:1:1. As only very little information is available about the ways how oxide surfaces solve the problem of electroneutrality at the surface, it is not useful to assume unrelaxed cuts through the bulk unit cell as basis for mechanistic considerations about for example the existence of strongly basic oxygen atoms in the form of ’’Fe-O-K" groups.

A significant advance in our understanding was only possible with an experimental clarification about the relation between oxide structure, adsorption geometry and surface electronic structure and the catalytic activity of respective model systems. The questions to be asked require the application of single crystalline surface qualities allowing spectroscopic techniques to be applied despite of the non-conducting bulk electronic structure of iron oxides and allowing a full structural characterisation with atomic resolution. To keep this goal achievable, the catalyst had to be simplified to a binary iron oxide surface. Fortunately, there are kinetic indications[6] that this simplification is valid. The apparent activation energy for STY formation over promoted and unpromoted hematite iron oxide were reported to be identical. This gives rise to the hope that the reaction pathway is identical on the model and on the technical catalysts and that promotion enhances merely the density of the active sites. It should be noted that the experimental fact of the similarity of the activation energies and the description of the surface composition of the promoted iron oxide catalysts are difficult to understand in one model.

The advent of technologies [7] for synthesising thin supported iron oxide films with LEED single crystalline qualities [8] and proven spectroscopic characteristics [9] allowed to combine the reproducible synthesis with a controllable surface defect disposition with detailed structural data of magnetite and hematite surfaces [10,11,8]. The construction of experimental facilities allowing ultra-high-vacuum characterisation and reaction studies at realistic pressures by the admission of EB and steam in the practically used ratio of 1:5 to 1:10 gave the possibility [12] to probe directly the structure-reactivity relation [13] for different iron oxides in the EB dehydrogenation.

The results [14,15] can be summarised in the following statements They refer only to the binary oxide model system.

1. Chemisorption of the educt occurs with comparable sorption energies on magnetite and hematite with a clear differentiation between high energy defect sites and low energy perfect terrace sites. 2. Conversion to STY occurs only in the presence of water. 3. Conversion to STY occurs only on defective hematite surfaces, perfect iron oxides are all inactive, defective magnetite surfaces are also inactive. 4. The type of defects, point or line defects (step edges, screw dislocations, mesoscopic roughness) is less important as compared to the density of defects. 5. The size of the vast majority of the defects is small compared to the size of the EB molecule which makes it plausible that an active site is a defect plus some flat terrace surface in the proximity of the defect. 6. The EB molecule is under technical reaction pressures and temperatures chemisorbed with a geometry of the benzene ring flat on the oxide surface. 7. It is bonded via a pi-electron interaction with iron sites in crude similarity with low-valent 84

iron-arene complexes. 8. Only the catalyst surfaces active for EB dehydrogenation are able to split water into molecular hydrogen which desorbs and oxygen which is most probably trapped at defect sites. 9. The catalytic activity is inevitably correlated with the formation of a coke deposit consisting of aromatic structures and a residual hydrogen content allowing a carbonisation stage to occur prior to combustion in oxygen.

These experimental facts allow to draw the following conclusions about the mode of action of the catalyst. The ensemble of atoms forming an active site is a section of iron-terminated iron oxide [16] at which the EB molecule is flat adsorbed. As the ethyl group, which must react, is bent relative to the aromatic plane, steps or defects must provide deviations from the flat geometry in order to allow the intimate contact of the ethyl group with the catalyst. At the defects oxygen atoms are exposed which react as "basic sites" with the ethyl hydrogens and form two hydroxyl groups. Sorption kinetics and the modified iron atoms (reduced to Fe11) regulate the residence time of the STY molecule such that no further reduction of the iron oxide occurs. The STY desorbs and at reaction temperature the hydroxilated iron oxide restructures by liberation of a water molecule and by enlarging the defect due to the loss of one oxygen atom. The stoichiometric partial reduction of an oxide into a lower valent oxide and water becomes only a catalytic cycle, if the lost oxygen atom is replaced in an independent process. In the absence of gas phase oxygen this can only be achieved by oxidation of water into hydrogen gas and an oxygen atom which oxidises the two Fe" sites and becomes reduced to the missing oxo-anion. Such a mode of operation renders the whole process into an oxidehydration of EB with the participation of "lattice oxygen" which is replenished by water as very mild oxidising agent.

A critical part in this scenario is the control of the residence time of the organic molecule over the oxide surface. The sorption site could be the carbonaceous deposit and not the bare oxide. The carbon surface would be inert enough not to cause further reaction of the STY molecule. The reaction path would then comprise the generation of an oxygen atom from water and its transport onto the carbon surface where it could react with the EB substrate. The well-proven fact that the reaction of EB to STY can be accomplished by oxidehydration [17,18] over e.g. activated carbon catalysts [19] is a very strong indication that this reaction pathway may well contribute to the overall conversion. The reaction of EB over iron oxide would be a parallel process and is required as initial reaction path to create the carbon catalyst. The required water, the high reaction temperature and the choice of the hematite catalyst would find a logical explanation as ingredients for a co-catalytic process providing the oxidising agent for the EB without burning much of the organic substrate as it would inevitably occur with the presence of gas phase oxygen at 900 K.

This mode of action can be written formally in the following way:

Fe[] + H20-»Fe(0) + H2 (3) Fe(O) +Cy -> Fe[] + Cy (0) (4) Cy (0) + EB -> Cy + STY + H20 (5)

Equation (5) symbolises a complicated sequence of elementary step reactions. It describes the oxidation of a saturated alkane side chain by an activated oxygen. In order to better understand this process, we first need to know much more about the "activated" oxygen species. In extreme 85

terms such a species can either be a,,surface radical" [20] which should be better referred to as hyperoxo-anion in the oxide surface or a Bijerrum base (such as in molten alkali oxide systems).

Returning to our issue of strategic considerations it is clear that the picture about the mode of action of the MMO catalyst completely can change the scope of STY catalyst development. We can define three major targets of further research, in order to significantly improve the performance of the process.

The first one is the improvement of the MMO system to activate water. For this purpose the presence of Fem is as important as the presence of stable defects at which the water molecule can be activated. The stabilisation of defects at 900 K in water and reducing atmospheres is a significant problem of solid state chemistry which has apparently been solved by the addition of all the promoters in the technical catalysts. Optimisation of the promoters with the rationale of defect stabilisation and stabilisation of a high abundance of Fem is one new direction of research.

The second direction is the understanding of the structural requirements of the carbon substrate and the proof that activated oxygen can migrate onto carbon [21,22], It is quite possible [19] that the active carbon is not the initial material but the in-situ deposit from the STY. An alternative scenario could be that normal oxygen from water splitting is activated by carbon, as we see it happen in the oxidehydration process described in the literature [23].

The third area is the activation of an alkane function by oxygen. This process is of immense relevance to a wide area of basic chemicals functionalisation and occurs not only in all processes using saturated alkanes as feedstock.

Oxidative Functionalisation of Alkanes

Oxidative functionalisation of alkanes [24] transforms them into alkenes, alcohols, aldehydes or carboxylic saturated or unsaturated acids. All these compounds are more reactive towards total oxidation than the precursor alkanes. The case of EB is fortunate as the benzene ring exerts an activating influence on the C-H bonds and facilitates the functionalisation. A typical difficult reaction is the conversion of propane to acrylic acid [25]. In general, the understanding of reaction pathways in allylic oxidation [26,27] and in C-4 selective oxidation [28] is more advanced than our insight into the activation of unreactive alkane C-H bonds with maybe the exception of methane activation [29,30,31].

The general problem with direct functionalisation is to achieve a sufficient chemo- and regioselectivity of the oxygen attack in order to obtain useful molecules. It is less important to optimise the yield of these processes rather than maintain a high selectivity for not simply burning the feedstock to CO2 and water. As this is the thermodynamically preferred reaction it is difficult to suppress it with any catalyst which is active enough to activate the nonreactive alkane. A positive influence of an optimised kinetics is also limited as selective redox reactions tend to be slow except for the very simplest cases (methane activation) where radical chemistry is involved.

Alkanes may be functionalised with strong acids leading to intermediate carbocations which can react with oxygen. Most MMO catalysts with maybe the exception of heteropoly molybdates and tungstates exhibit unclear acid - base properties [32,33]. The oxidation state of the catalyst affects 86

via the metal-oxygen bonding properties the strength of the Broensted acidity of OH groups and the Lewis acidity of the metal centres. Solid acid-base titration experiments as a function of the oxidation state are still complicated and were not performed widely [34,35]. Conventional probe molecules for acidity [36,37] are on MMO systems oxidised upon treatment of the adsorbed state to even moderate temperatures. The direct determination of acidity on MMO is still much in its infancy [38,39,40] and is additionally hampered by the presence of lattice hydroxyl functions

Another pathway with a direct attack of “basic oxygen" could lead to insertion of oxygen into one C-H bond and dehydration of an intermediate alcohol function to an olefin which can undergo successive functionalisation with more activated oxygen. The problem in such a reaction scenario is the control of the extent of functionalisation which can easily end up as CO2 and water taking into account the strength of the basicity (or the highly electrophilic character) of the initial oxygen species required to activate alkane C-H bonds. A successful catalyst development would require several oxygen atom species with different basicity or oxidising power. A detailed management of the relative surface abundance which would have to match the fractional coverage of the initial alkane (very low) and of the intermediate olefin (higher) could provide a means of selective functionalisation. Conceptually simpler are dehydrogenation reactions such as the formation of STY discussed above or the formation of formaldehyde from methanol [41], the selective oxidation of ethanol [42] or oxidative coupling of methane. The latter reactions can be carried out on bare coin metal surfaces [43,44] allowing to study the different species of active oxygen without the problem of a large abundance of nonreactive matrix oxide. Common to these reactions is the only action of active oxygen as basic hydrogen acceptor. The oxidising function is undesired in these reactions.

Recent developments in homogeneously catalysed selective oxidation [45] involve strong acids and metal organic oxidants. Although such reaction paths are of very limited practical use because of the price of the oxidant, these systems are useful for model studies. Practically useful are systems in which oxygen from air is activated in one homogeneous cycle (involving main group element compounds which act simultaneously as proton acceptors) which is coupled to another homogeneous cycle which does the acid catalysed activation of alkanes. An early forerunner of this principle is the lead chamber process for sulfuric acid synthesis using NO as catalytic medium.

MMO catalysts are neither only solid acids nor exclusively suppliers of activated oxygen. The dual functions require a partly reduced oxide to provide electron conductivity and the presence of sufficient OH groups to allow for proton exchange reactions. Insufficient reactivity of a MMO catalyst can thus either be cured by adjusting the oxidation state using a strong reductant or enhancing the oxygen partial pressure (often on the expense of selectivity) or by adding water to allow a sufficient coverage of the surface with OH groups. Systematic studies of measuring activity and selectivity on alkane activation under control of oxidation state and solid state acidity simultaneously have not yet been performed. They require advanced in-situ spectroscopy coupled with a suitable reactor design to achieve meaningful kinetic data (not as differential rector and not in transport limitation conditions). Such instrumentation is not readily available but will be needed to clarify the multiple role of MMO catalysts. In order to develop multi-functional oxidation catalysts which might have occurred unintentionally with the admixture of many oxides of different valence states and with differing solid state reactivity [46], it is of paramount importance to identify the chemical tools (composition, real structure) of fine-tuning the oxidising power and simultaneously the basicity 87

of an oxygen atom which is active enough to react with chemisorbed alkane molecule. Our present understanding discriminates "unselective" gas phase oxygen from "selective" lattice oxygen [47] without knowing the exact electronic structure of any of the two species.

The implementation of the in-situ high pressure X-ray absorption spectroscopy of oxygen at its K-edge allows a direct access to the electronic structure of an active oxygen species [48]. Under conditions of selective oxidation at several millibar total pressure the electronic structure of oxygen atoms was probed [49] which correlate in their surface abundance with the independently determined conversion of methanol to formaldehyde. The catalyst was elemental copper which contains oxygen under reaction conditions as oxide, as adsorbate, as surface embedded- and as bulk-dissolved species [41]. It was possible to discriminate the chemical bonding states of all species and to identify the oxygen segregating from the bulk to the surface as the selectively acting catalyst. It exhibited a unique electronic structure characterised by an interaction of oxygen atomic s-p states with exclusively the s-p states of the metallic copper. Oxygen species involving in their bonding the copper d-states were found to catalyse the total oxidation of methanol. This in comparison to MMO simple system catalysts allows to test our present ability in discriminating details of the electronic structures of active atomic oxygen species. Preliminary experiments with a real MMO catalyst [50] in the form of a vanadium-molybdate system reveal that the method is capable of identifying the electronic structure of the active oxygen species which seems, however, not to be identical to that found in the copper catalyst [51].

The geometric real structure of the active catalyst will be of dominating influence on the reaction kinetics of the surface-to-subsurface exchange of oxygen. An MMO catalyst needs to. be an ion conductor for oxygen anions [26] to allow the efficient operation of the lattice oxygen mechanism. This conductivity depends on the average electronic structure of the active solid and also quite fundamentally on the defect structure [52]. Modem high resolution transmission electron microscopy on molybdate catalysts has revealed [53] that not any arbitrary type of defects is required but rather specific isolated point defects are useful for selective oxidation kinetics. Extended defects such as well-ordered shear structures are detrimental for the anion conductance and reduce thus the selective oxidation function. This information is widely disregarded in synthetic attempts to find novel phases of MMO catalysts which often are derivatives of shear structures in the ReOj structure family. There are indications [54] that a bulk structure which is amorphous in long range ordering but contains clusters of well-ordered metal- oxygen octahedra interconnected by non-octahedral bridge units [55] are particularly active as selective oxidation catalyst. It is necessary to combine polyvalent matrix components with structure forming linker components in a similar way as it is done in inorganic glass synthesis to achieve the synthesis of such oxides. It is important to note that the Zachariasen theory [56] for glass formation predicts the efficiency of promoter oxides enumerated in the remote control theory [57]. This theory which describes the structure of an oxide glass by a continues random network (CRN) of polyhedra predicts the operation of three rules which can be used as guides to design nanocrystalline networks with a significant proportion of local defects yielding the CRN: • A proportion of network-forming cations are surrounded by oxygen tetrahedra or triangles. • The oxygen polyhedra only share comers which each other • Some oxygen atoms are linked only to two cations and do not form any additional bonds. If we relax these rules allowing clusters of oxides to form more densely packed structures and includes these clusters as “large cations ” into the CRN concept structural motives similar to that of Moj O|4 [54] are obtained. 88

Such a concept may explain the "synergetic" effect in MMO catalysts [58] by the formation of a metastable glass phase which is absent in post-mortem studies but operative under the special conditions of selective oxidation catalysis (simultaneous presence of oxygen and of reducing organic components together with steam and COi). These conditions are not applied in oxide synthesis studies and so we know little about the phase diagrams of multi-component oxides under catalyiically relevant conditions.

Conclusions

The functionalisation of C-H bonds with activated molecular oxygen requires two basically different functions of the oxygen. It acts as proton acceptor forming water molecules and as substituent which is transferred from the catalyst into the product molecule. It is important that both functions can be controlled independently in order to achieve the necessary chemo- and regioselectivity of the process.

An advanced catalyst has thus to meet these dual operation conditions and provide several functions: it acts as electron conductor in the activation of oxygen (source) and oxidation of the substrate (sink) steps of the reaction and it provides ionic conductivity for oxygen anions (source) and protons (sink). These requirements are needed simultaneously in all situations where no free radicals occur in the catalytic process. Good performing catalysts provide these functions with small but controllable activation barriers. By adaptation of these activation barriers to the energetics of the C-H bonds in educt and product it can be expected to solve the problem of selectivity in partial oxidation. Unfortunately, the activation barriers are not independent from each other. The average oxidation state of the working catalyst is the leading parameter determining the other barriers as long as the system is a single phase under reaction conditions. It is pointed out that the determination of the number of active phases under reaction conditions is at present a widely unsolved problem for MMO catalysts and that the detection of a multi-phase nature ex-situ is, even when it is in agreement with thermodynamic considerations, by no means a proof of the multiphase nature under reaction conditions [59,60]. For proven multi-phase systems (e.g. the carbon covered iron oxide in STY synthesis) the diffusion processes of active species across the phase boundaries have to considered additionally.

To modify the activation barriers for the basic functions in a rational way it is essential to know several physicochemical quantities of the starting MMO phase. These data are:

• the surface coverage with reactive OH groups in the working oxidation state • the surface coverage with atomic oxygen in the working oxidation state (sticking coefficients of di-oxygen under catalytic conditions) • the surface-to subsurface diffusion equilibrium in the working oxidation state • the activation kinetics of hydrogen and oxygen as seen in isotope exchange data in the working oxidation state • the working oxidation state as a function of the structure of the organic substrate.

These data can only be obtained by performing the relevant physicochemical measurements under catalytic action and spectroscopic in-situ control. Comparisons of kinetic data or even of only catalytic performances are here of little value as they will inevitably be done under varying oxidation states of the MMO catalyst. The resulting data can thus not be directly related to the 89

parameter under study as in each experiment a catalyst with a different electronic structure is used.

These considerations call for a change in the strategy of research. The numerous attempts to find new MMO catalysis or to promote existing systems with additional phases of debatable function should be reduced for the benefit of the analysis of the bottlenecks of existing catalysts by determining the above mentioned quantities. From these data it is possible to identify the limiting function and to improve on that in a rational fashion.

The chemical intuition in designing new catalyst structures is limited by the practical necessity of creating a stable and sustained catalytic function. As the functions described above will always lead to less stable MMO phases than the less-active well-ordered and close-packed phases, it is not to be expected to increase the selectivity of an MMO catalyst without reducing its solid state stability. Reactor design and control of the operation parameters of loading, auxiliary gas compositions (water vapour) and gas flow conditions will have to be adapted very carefully after a chemical modification of an MMO catalyst in order to maintain its stability and so observe the true consequences of the chemical modification .

It becomes obvious that in-situ experiments are no longer an exotic tool for fundamental science. They should rather be considered as targeted working tools for the catalyst practitioner in order to increase productivity of research as it has occurred in the past with the standard techniques of ex- situ characterisation.

Acknowledgements

The results discussed here were obtained by a large group of motivated co-workers mentioned in the respective references. Only due to their successful work it was possible to reach these conclusions. Two joint research projects with BASF and funded through the catalysis programme of the Bundesminsterium fiir Bildung und Forschung provided the experimental base on which many of the considerations are built upon. 90

Literature

[1] R. D. Gall C. L. Hill. J. Mol. Calal. A: Chemical. 1996 . 114. 103-111.

[2] R. Schlogl, An sew. Chem.. in press 1998 .

[3] E. H. Lee. Catal. Rev.. 1973. 8.285.

M J. Schiitze, M. Wesemann, T. Rayment, A. Dent, R. Schlogl, G. Ertl M. Muhler, J. Catal..

1990.126.339-360. [5] R. Schlogl, G. Ertl M. Muhler, J. Catal.. 1222,138.413-444.

M T. Hirano, Anpl. Catal.. 1986 .26, 65.

[7] G. A.Somorjai W. Weiss, J. Vac. Sci.Technol. . 1993 . All.2138.

[8] W. Weiss, Surf. Sci-1222, 377-379. 943.

[9] Th. Schedel-Niedrig, W. Weiss, R. Schlogl, Phvs.Rev.B , 1225,52(24), 17449-17460.

[1°] M. Ritter, W. Ranke, W. Weiss, Phvs. Rev. B. 1998 .57(12), 7240-7251. H1] A. Barbieri, M. A. Van Hove, G. A. Somorjai W. Weiss, Phvs. Rev. Lett.. 1993 .21,1848.

[12] w. Weiss, M. Ritter, D. Zscherpel, M. Swoboda, R. Schlogl, J. Vac. Sci. Technol. A. 1998 .

16(D. 21-29. [13] W. Weiss, D. Zscherpel, R. Schlogl, Catal. Lett.. 122§,52,215-220.

[14] W. Ranke, W. Weiss, Surf. Sci.. in press 1998 .

[15] D. Zscherpel, W. Ranke, W. Weiss, R. Schlogl, J. Chem. Phvs.. ,L22§, 108,9506-9515.

[16] X.-G. Wang, W. Weiss, Sh. K. Shaikhutdinov, M. Ritter, M. Petersen, F. Wagner, R.

Schlogl, M. Scheffler, Phvs. Rev. Lett- J22§,81(5), 1038-1041. [17] G. Emig, H. Hofmann, J. Catal.. J5S2,84,15-26.

[18] A. Schraut, G. Emig, H. Hofmann, J. Catal.. 1988 . 112.221-228.

[19] A. Guerrero-Ruiz, I. Rodriguez-Ramos, Carbon . 1994 . 32(1), 23-29.

[20] a . B. Anderson, J. F. Brazdil, R. K. Grasselli S. P. Mehandru, J. Phvs. Chem.. 1987 .91,

2930-2934. [21] V. A. Garten, D. E. Weiss, Aust. J. Chem.. 1957 . IQ, 309-328. 91

[22] R. Schldgl in Carbons (Eds.: G. Ertl, H. Knozinger, J. Weitkamp), VCH

Verlagsgesellschaft. 1997 . vol. 1, p. 138. [23] J. Haber, M. Wojciechowska, W. Gut, Bull. Polish Acad. Sci. Chem.. 1996 .44(1), 55-62.

[24] J. Lebars, A. Auroux, M. Forissier, J. C. Vedrine, J. Catal.. 1996 .162(2). 250-259.

[25] W. Ueda, Y. Suzuki, Chemistry Lett.. 1995 .541-542.

[26] J. Haber A. Bielanski, Catal. Rev.-Sci. Eng. . 1979 .19(1). 1-41.

[27] R. K. Grasselli, J. D. Burrington, Adv. in Catal.. 1981 .30. 133-163.

[28] u. S. Ozkan, S. A. Driscoll, L. Zhang, K. L. Ault, Journal of Catalysis . 1990 . 124. 183-193.

[29] s. S. Bharadwaj, L. D. Schmidt, J. Catal.. 1994 . 146.11-21. [30] A. W. Sexton, B. Kartheuser, C. Batiot, H. W. Zanthoff, B. K. Hodnett, Catal. Tod. . 1998 .

40(2-3), 245-250. [31] F. Arena, N. Giordano, A. Parmaliana, J. Catal.. 1997 .167(1). 66-76.

[32] M. Davidovic, N. Tjapkin, Ph. Colomban, A. Novak U. Mioc, Solid State Ionics . 1991 .46.

103-109. [33] p. Concepcion, A. Galli, J. M. L. Nieto, A. Dejoz, M. I. Vazquez, Catal. Lett.. 1996 .43(3-

4), 451-460. [34] u. Mioc, M. Davidovic, N. Tjapkin, Ph. Colomban, A. Novak, Solid State Ionics . 1991 .46,

103-109. [35] A. Bielanski, A. Cichowlas, D. Kostrzewa, A. Malecka, Z. Phvs. Chem. NF.. 1990 . 167.

93-103. [36] V. M. Mastikhin, I. L. Mudrakovsky, S. V. Filimonova, Chem. Phvs. Lett.. 1988 . J49(2),

175-179. [37] v. M. Mastikhin, V. V. Terskikh, M. N. Timofeeva, O. P. Krivoruchko, J. Molecular

Catalysis A . 1995. 95.135-140. [38] J. Lebars, J. C. Vedrine, A. Auroux, S. Trautmann, M. Baems, Appl. Catal. A. 1994 .

119(2). 341-354. [39] H. KnSzinger G. Mestl in Vibrational Spectroscopies (Ed.: H. Knozinger, J. Weitkamp G.

Ertl), 1997 . vol. 2, p. 539. 92

[40] h . Knozinger in Characterization of Heterogeneous Catalysis bv Vibrational

Spectroscopies (Ed.: R. A. van Santen H. H. Brongersma), Plenum Press, 1991 . p. 167. [41] D. Herein, H. Werner, Th. Schedel-Niedrig, Th. Neisius, A. Nagy, S. Bemdt, R. Schlogl in

The Selective Oxidation of Methanol: A Comparison of the Mode of Action of Metal and

Oxide Catalysts (Ed.: S. T. Oyama, A. M. Gaffney, J. E. Lyons R. K. Grasselli), Elsevier

Science B. V., 1997 . vol. 110, p. 103. [42] w. Zhang S. T. Ovama. J. Am. Chem. Soc. . 118 . 1996 .7173-7177.

[43] X. Bao, M. Muhler, R. Schlogl, G. Ertl, Catal. Lett.. 1995. 32(1-2), 185-194.

[44] D. Herein, A. Nagy, H. Schubert, G. Weinberg, E. Kitzelmann, R. Schlogl, Z. Phvs. Chem..

12^,191,67-96.

[45] S. S. Stahl, J. A. Labinger, J. E. Bercaw, Angew. Chem.. 1998 . 110.2298-2311. [46] S. Albonetti, F. Cavani, F. Trifiro, Catal. Rev. Sci, Eng. . 1996 .38(4), 413-438. [47] (Ed.: J. Haber A. Bielanski), Marcel Dekker, Inc., 1991 . p. 1.

[48] a . Knop-Gericke, M. Havecker, Th. Neisius, Th. Schedel-Niedrig, Nucl. Instr. Meth. A.

1998.406.311-322. [49] M. Havecker, A. Knop-Gericke, Th. Schedel-Niedrig, R. Schlogl, Angew. Chem. Int. Ed..

222B, 110(13/14), 2049-2053. [50] Th. Schedel-Niedrig, Fresenius' J. Anal. Chem.. 1998 . 361.680-682.

[51] Th. Schedel-Niedrig, X. Bao, M. Muhler, R. Schlogl, Ber. Bunsenges. Phvs. Chem.. 1997 .

101(7), 994-1006. [52] J. Maier, Prog. Solid St. Chem.. 23, 171-263.

[53] p. L. Gai-Boyes, Catal. Rev.-Sci. Eng. . 1922,34(1-2), 1-54.

[54] H. Werner, O. Timpe, D. Herein, Y. Uchida, N. Plunder, U. Wild, R., Hibst, H. Schlogl,

Catal. Lett.. 1997.44.153-163. 93

[55] x. V. Andrushkevich, V. M. Bondareva, G. Ya. Popova, L. M. Plyasova in New

Compounds of the Vanadium-Molybdenum Oxide System. In Situ Investigation of the

Mechanism of Acrolein Oxidation to Acrylic Acid. The Role of the Structure and Bond

Energy of the Intermediate Compounds /Surface Science and Catalysis: New Developments

in Selective Oxidation bv Heterogeneous Catalysis (Eds.: P. Ruiz, B. Delmon), Elsevier

Sci. Publishers B. V., 1992 . vol. 72, p. 91.

[56] W. H. Zachariasen, J. Am. Chem. Soc. . 1932 .54(4), 3841-3851.

[57] l. T. Weng, P. Ruiz, B. Delmon in Classification of the Roles of Oxides as Catalysts for

Selective Oxidation of Olefins (Eds.: P. Ruiz, B. Delmon), Elsevier Sci. Publishers B. V.,

1992. vol. 72, p. 399.

[58] o. Legendere, Ph. Jaeger, J. P. Brunelle in Strong Evidence of Synergetic Effects between

Cobalt. Iron and Bismuth Molybdates in Propene Oxidation to Acrolein (Eds.: P. Ruiz, B.

Delmon), Elsevier Sci. Publishers B. V., 1992 . vol. 72, p. 387. [59] B. Delmon. Surf. Rev, and Lett- 1995 .2(11.25-41.

[60] e. M. Gaigneaux, M. Dieterle, P. Ruiz, G. Mestl, B. Delmon, J. Phvs. Chem- submitted

J22S- 94 95 *DE012198176*

DGMK-Conference. "Selective Oxidations in Petrochemistiy ”; Hamburg 1998 '

P. Moriceau, A. Lebouteiller, E. Bordes, P. Courtine Departement de Genie Chimique, Universite de Technologie de Compiegne, B.P. 20529, F-60205 Compiegne, France

A PREDICTIVE TOOL FOR SELECTIVE OXIDATION OF HYDROCARBONS: OPTICAL BASICITY OF CATALYSTS

INTRODUCTION The selectivity to a given product obtained in the mild oxidation (MOX) or in the oxidative dehydrogenation (ODH) of hydrocarbons is known to be driven by a large number of intricatingly related properties 1 of the catalytic solid. Thanks to the development of computer facilities, expert systems just begin to be available23. Recently, Hodnett et al.4 tried to account for the limited selectivity attained in MOX of light alkanes by answering the question: How well an active site distinguishes between a C-H bond in a reactant and a similar C-H bond in a product. We have tried another approach which consists in correlating properties of the gaseous molecules and of the solid. Ai has already studied the influence on activity/selectivity of the basicity-acidity of both molecules of reactant and product, and of the catalyst 5,6. The basicity of the catalyst oxygen has long been recognized as a suitable parameter 7,8,9 , but difficult to quantify. Recent advances in inorganic chemistry allow now to predict various kinds of chemical transformations 10,11,12. We have used the Optical Basicity parameter, An, which accounts for the electron donor power of the lattice oxygen. MOX and ODH reactions proceed by Mars and van Krevelen's mechanism, so that the active-selective oxygens of the solid are responsible for the attack of the C-H bond of the hydrocarbon (activity), and are inserted into the intermediate complex (MOX), or in the water formed (ODH) (selectivity). Very recently, the values of A& have been computed for several transition metal oxides, most of them being known as catalysts 13.

In order to see if the An parameter of a given oxide catalyst could be related to catalytic parameters in a given MOX or ODH reaction, data have been taken from literature and plotted against the calculated An of the catalyst. Examples will be given for the selective ODH of isobutyric acid to methaciylic acid on polyoxometallates H4.xMxPVMon0 4o. z and on another, Fe-P-0 catalyst, as well as for a series of catalysts for the ODH of 1-butene to butadiene. The validity of the approach being demonstrated, the second part of the paper deals with semi-empirical relationships between the electron donor power of the catalyst oxygen

! DE99G2085 DGMK-Tagungsbericht 9803,3-931850-44-7,1998 96

(An,) and its equivalent for gaseous molecules, which is the energy of ionization i.e. the electron donor power of the molecules.

SCALE OF OPTICAL BASICITY FOR OXIDIC CATALYSTS

The Optical Basicity, An,, obtained by UV spectroscopic measurements, has been proposed by Duffy in studying oxide glasses and other compounds 16-11. Au, characterizes the Lewis basicity of the solid since it is the electron donor power of the lattice oxygen. The Optical Basicity of a mixed oxide is calculated as the weighted mean of the component simple oxides, according to equation [1]: A/a =(^A v 4o o/2 + ArjgA Bo 6/2+„.) [1] where XA and XB are the fraction of negative charge neutralized by the cations A and B, components of the simple oxides AO, and BOb. To determine Au, of highly covalent transition metal oxides, such as V205, M0O3, WO3, etc., calibration curves have been drawn between Au, and ICP, the Ionic Covalent Parameter as defined by Portier et al.12 (Fig. 1).

1.2 * dlO • dl-d9

0.8 --

0.6 '

!Mo 66

Figure 1: Correlations ICP=f(A) for various electronic configurations. 97

ICP accounts for the influence of the ionic-covalent character on the acid strength of the cation in oxides and is calculated by equations [2] and [3]: ICP = log P - 1.38% + 2.07 [2]

% = 0.274 z - 0.15 zr;-0.01 n+H-a [3]

P is the polarizing power of the cation (P= zZ r2 with z = valence, r, = Shannon ionic radius, with r(02") = 1.40 A (7); % is the electronegativity (in Pauling-type unit); a is a correcting term, depending on valence and coordination of the cation 14. The zero scale is set for Au+ in six- coordination. By plotting ICP against An, values known for simple oxides, five straight lines were obtained 13, depending on the electronic configuration of the component cation : s-p, d°, d'-d9 , d10, d10s2 (Fig. 1). From these calibration curves, the calculation of An, for any cation with any valence and coordination is possible, provided its ICP is known 13. For example, using equations 1-3, An, values are 0.63 and 0.68 for V205 and VO2, respectively, where Vs* and V4* are 6-coordinated.

CORRELATION BETWEEN An, AND CATALYTIC PARAMETERS '

1. Selectivity correlated with Optical Basicity The example chosen to show that the optical basicity of the catalyst is indeed related to its selectivity is the ODH of isobutyric acid to methacrylic acid. Two catalysts are active and selective for this reaction, polyoxometallates and phases of the Fe-P-0 system. Among other authors, we have shown that the polyoxoacid HiPVMouO-ro is active and selective in this reaction 14,15. The selectivity varies when the protons of the heteropolyacid are completely or partially substituted by a series %Sel. of cations. At the same conversion, the selectivity to methacrylic acid Be (0.5; (mol%) increases for the M2* series: Co

The explanation lies in the basic properties of oxygen, those needed to activate isobutyric acid, and those needed to "deactivate" the adsorbed product, i.e. to allow the desorption of methacrylic acid. During the oxidation of C-C to C=C, the basicity of the oxygen after activation of the reactant must increase in order to favour the desorption of the product. Indeed, the Lewis basicity is higher for the o-7t bond in methacrylic acid than for the o bond in isobutyric acid. Therefore, the optimum value of oxygen basicity obtained with

H2CUPVM011O40 corresponds to a good compromise. Phases of the Fe-P-0 system have been found to exhibit similar levels of activity and selectivity in this reaction in the same operating conditions (except that an excess of water vapour must be added to the feed)16* 17. At the steady state, the structure is made up with trimers of edge-sharing octahedra containing iron, with Fe2*/Fe 3* = 2/1n. Calculating Au, for the mean composition FePO^FezPzO? gives An, = 0.51, which is similar to the values obtained with polyoxometallates. Although the layered (infinite) structure of Fe-P-0 is very different from that (molecular, Keggin-type units) of polyoxometallates, similar values of the optical basicity of oxygen are strikingly found for catalysts exhibiting similar catalytic performance.

2. Adsorption Parameters correlated with Optical Basicity The data proposed by Matsuura for the ODH of 1-butene to butadiene on several catalysts 18 gives an example of a relationship between adsorption and optical basicity. Two kinds of catalysts were examined, more or less active and/or selective iron salts (FeMQi, M = P, As, Sb), FefflizO; and FesCL, and others known to be selective in the reaction, BizMoOs, USbsOio or SbzCL-SnOz. A linear correlation was observed between the entropy and the enthalpy of adsorption of 1- Qads (KJ/mol) butene at the same temperature, a 90 Fe3Q4* definite range of (AS, AH) being defined 80 -• ,...FfiiiBi209>... for selective catalysts 18 . By plotting the 70 - USbsOio* I enthalpy of adsorption AH against the 60 • • ♦ Sb2CM/Si;n<>2 BtiMoOs, calculated Au, value of these catalysts, a 50 - ♦♦FeSbOt linear relationship is observed (Fig. 3). 40- FeAsOt The most selective catalysts are 30 FePOt 20 observed in the range 6(AH) =45-72 10 kJ/mol) and S(Au,) =0.8-1.0. Although 0.4 0.6 0.8 1.2 Ath(Fe304) =0.82, this oxide is not selective but is very active because its Figure 3 : Linear correlation between Q^, measured enthalpy of adsorption is too high 18 . by Matsuura18 and calculated Au, values of oxide catalysts. 99

This remark gives the opportunity of emphasize the fact that the theoretical optical basicity A& is a bulk parameter, so that it should be corrected to account for surface where catalytic reactions proceed. Such corrections 13 will be described in a forthcoming paper.

CORRELATIONS BETWEEN IONIZATION POTENTIALS AND OPTICAL BASICITY

The above examples show that the optical basicity of a catalytic solid is related to its catalytic properties. We have tried to apply this concept to other oxidation reactions proceeding on catalysts known to be selective, and have looked for an acid-base property of the gaseous molecules. Let us propose an oversimplified view of catalytic reaction. The ionization energy, usually called ionization potential (IP), is the ability of a molecule to give an electron. If this molecule is the reactant, the electron will be received by the catalyst. This is a simple consideration, supposing that the first broken bond (e g., C-H) is ionic, which is unlikely when dealing with organic molecules. IP has already been used as a measure of the acid-base character6,19 . Conversely, the adsorbed product will take an electron from the catalyst to desorb as a neutral molecule, which can be accounted for by the ionization potential of, that time, the product. This difference, A(IP) = llP^cum - IPpnxi^tl, is assumed to be related to selectivity. Moreover, for oxidation reactions the net number of exchanged electrons will be provided by the catalyst: In other words, there should be a relationship between A(IP) and the electron donor power of the selective catalyst, which is accounted for by its Au, value. Plotting A(IP) against A& for various [reaction-product/selective catalyst] couples lead to figure 4. The IPs of several Cz-C; hydrocarbons are taken from a database20. When the number of carbon atoms varies during the transformation, A(IP) is multiplied by np/n R (where nR and nP stand for reactant and product, respectively). The A& of each selective catalyst was calculated with equations 1-3, as explained above. Only catalysts exhibiting high selectivities (> 50 mol%) were considered. Simple, two-electrons reactions, like epoxidation of ethylene or propylene, as well as complex, 14-e reaction, like n-butane to maleic anhydride, are considered. The most common oxidation reactions are situated in the range A(IP)=0.15-1.6 eV. The selective catalysts range from acid (A=0.45, e.g. for alumina) to basic (A=1.25 for AgzO) according to the reaction. Two distinct, linear A(IP) = f (A) relationships are obtained by linear regression and hold within a limit of error of A=0.01 to 0.05. 100

No Catalyst / Reaction / [Reference! 1.6 - 1 0.7I5A1P04/0.285A1 20, / Paraffins Ethvlbcnzcne-'Styrcnc |22| 2 (VO)2P2Oj /n-Butane-*MA (23] 3 FePOVFczPA ZIBA-MAA [24] 1.4- 4 H4PVM o „04o /IBA-MAA [25[ ♦14 5 Sbo. 5P1.09 Wo 9 M012O42.675 / MACO-*MAA |26)

a : 6 P1.2MO12VCS1 j0412$ / X 15^16-18-19 12- Isobulane-*MAA f 271 7 CU0.24CS2.O8 PMO10VO36 28 / Isobutanc-*MAA f281 8 NiW04 /Cyclohexane-*Benzcne [29] 9 V2Oj /Tolucnc-*Bcnzoic acid [30] 1- 10 V2O5 /Toluene-»BenzaIdchyde [30]

11 C0M0O4 /n-Butane-*butenc-l [31] ih\ 12 0.5Mo 50,V0.5Mo 03 / £ Ethane->Acetic acid [32] E30.8- 13 Mo ,8 052 /Ethane-*Ethylenc [32] 14 0.7Cr2OVCcO2 / / 70 \ isobutane-*Isobutene [331 1*9 \ 15 Mg 2V20j ZPropanc-*Propcnc [34-36] 0.6- 16 CdMoPOs.iy /Propane-*Propcnc [21] P \ f 17 0.05V2Oj/TiO2/orthoxylcnc-*PA[37-40] ; eA = % 18 0.04V205/Ti02 /propanc-*Propene [41] 6; d\ AgMo 3P20|3j,5 /Propane-*Propcnc [21] 0.4- a V0P04/(V0)2P20,/ A But-l-cnc-*MA [42] Z:*5 xg\ b 0.7V20$/0.3Mo 03 /Benzenc-»MA [1]

1 \ c Bi2MoO« /But-l-enc-*Butadicnc [43] d FeSb04 /But-l-cne-*Butadicnc [18] 0.2- :* 4 —v— ? A aefin? c Sb20.i/Sn02 /Propcnc-*Propcnal [43-44] \ f USb3O|0/But-l-cne-»Butadiene [43] g Pd02-6 /Pd02-4 / A _ , , Ethylene-* Acetaldehyde [451 V h TI.O3 /Propcnc-*Mcthyloxiranc [44] 0.45 0.65 0.85 1.05 1.25 i Ag 20/Al203 / EthyIene-*Ethylene oxide [46]

Figure 4 : Correlations between the ionization potentials difference and optical basicity of selective catalysts in various oxidation reactions. MA maleic anhydride ; IBA: isobutyric acid; MAA: methacrylic acid; MACO : methacrolein ; PA : phthalic anhydride.

Fourteen couples in ODH and MOX of alkanes fit statistically well the first line with a positive slope, while nine others fit the other line with a negative slope. The latter corresponds to the oxidation of unsaturated hydrocarbons (olefins, aromatics). However, the points representative of reactants like ethylbenzene (reaction 1), toluene (reactions 9, 10), o-xylene (reaction 17), are found to fit the "paraffins" curve. At first, this could be surprising when one remembers the 101

important role played by the jt-allylic intermediate. But, as such, the correlation is sensitive only to the groups concerned by the transformation, which are the saturated radicals (ethyl, methyl) : ethylbenzene is transformed into vinylbenzene (styrene), toluene is transformed into tolualdehyde, etc. Conversely, the point representative of the formation of maleic anhydride from benzene lays down on the "olefins" curve.

CONCLUSION

^Whatever the composition of the catalyst (promoted, supported, multicomponent, etc.) is, it is possible to calculate its electron donor capacity A. However, one important question remains: How are the surface and the bulk values of A related? Most oxidation catalysts exhibit either a layered structure as V2O5, and approximately A& ~A««r, or a molecular structure as polyoxometallates, and no correction seems to be needed. Work is in progress on that point. Of great importance is also the actual oxidation and coordination states of cations at the steady state: As jif Nos. 16 and H) (Fjg>-4^have been calculated from the composition determined by XANES and XPS*T Finally, the model is able to discriminate between "paraffins" and olefins as reactants. These calibration curves should help to find new cataly sts. \

Acknowledgement: C. Pham and Elf Atochem are thanked for helpful discussions and financial support of this work.

1 P Courtine., in Solid Slate Chemistry in Catalysis, ACS 279. 186"" Meeting of the Am.Chem.Soc., Washington D.C. 37 (1985) 2 M. Baems and E. Korting, Handbook of Heterogeneous Catalysis, Eds. G. Ertl, H. Knozinger and J. Weitkamp, Wiley-VCH, 1,419 (1997) 3 M. Baems, N. Guan, E. Korting, U. Lindner, M. Lohrengel, H. Papp, Int. J. Energy Res., 18, 197(1994) 4 C. Batiot, B.K. Hodnett, Appl. Catal. A, General, 137, 179 (1996) 5 M. Ai, Bull. Chem. Soc. Jpn., 49, 1328 (1976) 6 M. Ai, J. Catal., 98,401 (1986) 7 J. Haber, in Perspectives in Catalysis, J.M. Thomas and K.I. Zamaraev Eds, Blackwell Sci. Pub, 371 (1994) 8 H.H. Kung, Transition Metal Oxides : Surface Chemistry and Catalysis, Stud. Surf. Sci. Catal., Elsevier, Amsterdam, 45 (1989) 9 V.D. Sokolovskii, Catal. Rev.-Sci. Eng., 32,1 (1990). 10 J.A. Duffy and Ingram, J. Non-Cryst. Solids, 144,76 (1992) 11 J.A. Duffy, Geochim. Cosmochim. Acta, 57,3961 (1993) 12 J. Pettier, G. Campet, J. Etoumeau, M.C.R. Shastry, B. Tanguy, J. Compounds Alloys, 209, 59 (1994) ■ 102

13 A. Lebouteiller and P. Courtine, J. Sol. Stale Chem. 137, 94 (1998) ; A. Lebouteiller, Thesis, Compiegne (1997) 14 C. Desquilles, M.-J. Bartoli, E. Bordes, G. Hecquet, P. Courtine, Erdohl ErdCas Kohle, Petrochem., 109, 130 (1993); C. Desquilles, Thesis, Compiegne (1993) 15 V. Tessier, Thesis, Compiegne (1996) 16 J. Belkouch, B. Taouk, L. Monceaux, E. Bordes, P. Courtine and G. Hecquet, Stud. Surf. Sci. CataL, 82, 819 (1994) 17 J.M.M. Millet, J. Vedrine, Appl. CataL, 76,209 (1991) 18 1. Matsuura, 6thInt. Cong. Catalysis. London, Bl, 819 (1976) 19 J.T. Richardson, J. Catal., 9, 172 (1967) 20 Handbook ofPhysics, C.R.C. Press (1996) 21 L. Savary, G. Costentin, F. Mauge, J.-C. Lavalley, J. El Fallah, F. Sluder, A. Guesdon, H. Ponceblanc, J. Catal., 169,287 (1997); L. Savary, Thesis, Caen (1994) 22 S.M.Bautista, J.M.Campelo, A.Garcia, D.Luna, J.M.Marinas, R.A.Quiros, Stud. Surf. Sci. Catal., 82, 759 (1994) 23 F.Cavani, F.Trifiro, Appl. Catal. A General, 88, 115 (1992) 24 J.Belkouch, Thesis, Compiegne (1991) 25 M.J. Bartoli, Thesis, Compiegne (1990) 26 Air products and Chemicals (C. Daniel), US. 404397 (1983) 27 Sumitomo Chemicals, EP. 418657A (1989) 28 J.L.Cazaudumec, Thesis, Compiegne (1994) 29 a. L.L.Van Reijex, W.M.H.Sachtler, P.Cossee, D.Bronwer, 3rd Int. Congr. Catalysis, WMH Sachtler, GCA Schuit, P. Zwietering Eds., II, 829 (1965) b. G.M.Dixon, D.Nicholls, H.Steiner, ibid, II, 815 (1965) 30 J.E.Germain, R.Laugier, Bull. Soc. Chim. Fr, 2, 541 (1972) 31 J.C Daumas., Thesis, Paris (1970) 32 M. Merzouki,, B. Taouk, L. Monceaux, E. Bordes, and P.Courtine, Stud. Surf. Sci. Catal., P.Ruiz and B.Delmon Eds, Elsevier, 72, 165 (1992) 33 P.Moriceau, Thesis n°2106, Lille (1997) 34 D. Slew Hew Saw, V. Soenen, and J.-C. Volta, J. Catal., 123,417 (1990) 35 MAChaar, D.Patel, M.C.Kung, H.H.Kung, J. Catal., 105,483 (1987) 36 X.Gao, PRuiz, Q.Xin, X.Guo, B.Delmon, J. Catal., 148, 56 (1994) 37 M.Gasior, I. Gasior, and B. Grzybowska, Appl. Catal. 10, 87 (1984) 38 B.Grzybowska, , Catal. Today 1, 341 (1987) 39 A.Vejux, P.Courtine, J. Solid State Chem., 23, 93 (1978) 40 P.Courtine, E.Bordes, Appl. Catal. A : General, 157, 45 (1997) 41 N.Boisdron, Thesis, Lille (1991) 42E.Bordes, P.Courtine, J. Catal., 57,236 (1979) 43B.C. Gates, J.R. Katzer, and G CA Schuit, , in "Chemistry in Catalytic Processes", McGraw-Hill, 366 (1979) 44 J.Askander, Thesis, Compiegne (1982) 45 J.L.Seoane, Thesis, Paris (1970) ^P.A.Kilty, N.C.Rol, W.M.H.Sachtler, Proc. 5,h Int. Corn. Catal., North Holland, Amsterdam, 929 (1973) 103

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*DE012198185* M. Wildberger, J.-D. Grundwaldt, T. Mallat, A. Baiker ' n-'• Laboratory of Technical Chemistry, Swiss Federal Institute of Technology, ETH- Zentrum, CH-8092 Zurich, Switzerland

COMPARISON OF THE REDOX ACTIVITIES OF SOL-GEL AND CONVENTIONALLY PREPARED Bi-Mo-Ti MIXED OXIDES

I Novel sol-gel Bi-Mo-Ti oxides have been prepared and characterized by XRD, XPS, FT-Raman and HRTEM. The surface Bi3+ and Mo 6* species of some xerogels and an aerogel could be reduced and oxidized at room temperature, whereas the conventionally prepared reference materials were not reduced by H2 belowSOO °C. The unusual redox properties, under very mild conditions, are likely due to the unique morphology of Bi-Mo-oxides stabilized by titania. During butadiene oxidation to furan at above 400 °C the sol-gel mixed oxides restructured considerably and their performance was barely better than that of titania-supported Bi-Mo oxides^J

1. Introduction

Sol-gel mixed oxides can possess unique catalytic properties under mild conditions. Their acidity and catalytic activity are usually attributed to the molecular scale homogeneity, achieved by the appropriate control of the hydrolysis and condensation steps. A recent example is the fast and highly selective epoxidation of bulky cyclic olefins with titania-silica aerogels and f-butylhydroperoxide, at 60-90 °C [1 ]. The redox properties of various aerogels and xerogels at elevated temperatures have received much less attention and - to the best of our knowledge - their redox behavior at ambient and elevated temperatures have not been compared [2]. Bi-Mo oxides in the presence of titania as the major component were chosen for the present study. Bismuth molybdates are industrially important catalysts for the oxidation of propene to acrolein, oxidative dehydrogenation of butene to butadiene and ammonoxidation of propene to acrylonitrile [3,4]. Bi- and Mo-containing catalysts have also been shown to be active and selective in the partial oxidation of butadiene to furan at or above 400 °C, affording up to 20 % furan yields [5,6].

The aim of our study was to compare the redox properties of sol-gel derived and t conventionally prepared (supported) Bi-Mo-Ti oxides. The surface reduction and re- I

$ I DE99G2084 i DGMK-Tagungsbericht 9803,3-931850-44-7,1998 104

oxidation with hydrogen and oxygen were studied at room temperature, and the catalytic properties at elevated temperatures were determinedby the partial oxidation of butadiene to furan.

2. Experimental

All Bi-Mo-Ti mixed oxides contained 10 wt% nominal amount of bismuth molybdate (Bi203-2Mo0 3) and 90 wt% Ti02. The Mo and Bi precursors {Bi(N03)3-5H20, (NH4)6M07024-4H20, BiCI3 and MoOCI 4}, together with the hydrolysant aqueous HN03, were added to a solution of titanium(IV)tetraisopropoxide in isopropanol. The first number of the catalyst abbreviation indicates the nominal amount of bismuth molybdate in wt-%. Cl denotes the use of Cl-contaning Bi and Mo precursors, P indicates prehydrolysis of the Bi and Mo precursors prior to the addition to the Ti-alkoxide solution. Xerogels (X) were obtained by removing the solvent in vacuo at 100 °C. For the low temperature (LT) aerogel the liquid was removed from the gel by semicontinuous extraction with supercritical C02 at 40 °C and 240 bar. Finally, the catalysts were calcined in an 02 flow at 500 °C. The supported oxide 10BiMoO/TiO 2 was obtained by coprecipitation of (Bi(N03)3-5H20 and (NH4)6Mo 7024-4H20 in a slurry of Ti02 (P25, Degussa, 58 m2g"*). More details of the preparation can be found elsewhere [7]. y-Bi 2Mo0 6 was prepared by grinding Bi203 and MoO a and heating to 600 °Cfor 10 h. XRD patterns were recorded on a Siemens 6/0 D 5000 powder diffractometer (CuKg radiation, Ni filter). BET surface areas were determined by N2 physisorption. Thermoanalytical investigations were carried out using a Netzsch STA 409 thermoanalyzer. A Philips CM 30 ST electron microscope was used for the HRTEM investigations. FTRaman spectroscopy was carried out on a Perkin Elmer 2000 instrument using the 1033 nm line of a Ni YAG laser with 200 mW power. XPS analysis was performed on a Leybold Heraeus LHS11 MCD instrument equipped with load lock, transfer, high pressure, preparation and analysis chambers. The samples were outgassed at room temperature till 10"7 -10"6 mbar in the load lock. The base pressure of the other chambers was below 5-1 O'10 mbar. After treatment and evacuation in the reactor, the samples could be quickly transferred to the analysis chamber via the transfer chamber which served as a second evacuation unit. XPS analysis was performed at a pressure of 1-3 102 mbar with Mg radiation (12 kV, 20 A). Photoelectrons were detected by a hemispherical analyzer with 150 eV constant pass energy (for high resolution spectra at a pass energy of 37.8 eV). Steady state charging effects were eliminated by corrections of the energy shift (C 1s line of adsorbed hydrocarbons at 285.0 eV). The surface composition of the mixed oxides was determined from the areas of the Mo 3d, Bi 4f, O 1s, Ti 2p and C 1s peaks [8-10]. For the time-resolved reduction and re-oxidation of the oxide surfaces, the sample was 105

treated in the reactor in H2 or Oz (10"6 -10"4 mbar). After quick evacuation (<30 s, <5-10"8 mbar) the sample was transferred to the analysis chamber. The amount of reduced Bi was calculated by computing the peak areas of the four Bi3* and Bi° peaks. The oxidation of 1,3-butadiene was carried out in a computer controlled microreactorsystem equipped with a GC/MS analytical facility (HP 5890/HP 5972). The stainless steel tubes were heated to 200 °C. The feed gas flow was set by Brooks 5850 TR mass flow controllers. Typically, 100 mg catalyst (180 - 400 pm) was loaded to the quartz glass U-tube reactor (4 mm i.d.). The total flow rate was 1.67 cm3s'1. The reaction temperature was decreased from 550 to 400 °C in steps of 10 °C. At each step the product composition was measured after 1 h equilibration time. Kr was used as an inner standard for GC analysis. The MS signals at m/z = 54,68 and 84 were used to determine the amounts of butadiene, furan and Kr, respectively. The yields of light hydrocarbons, CO and C02 were determined with a second GC (Porapack QS column, TCD). The C-balance was in the range of 97-103 %.

3. Results

3.1 Structure and composition of mixed oxides

All uncalcined sol-gel catalysts were X-ray amorphous and possessed high surface area (up to 535 m2g" ’). However, 45 - 95 % of the original surface area was “lost ” during calcination in 02 at 500 °C. The average pore diameter increased considerably and well developed anatase crystallites were formed, but no crystalline Bi- or Mo-containing phase was detectable. After calcination the BET surface areas of sol-gel and conventional catalysts were rather similar (Table 1).

Table 1: Some characteristics of Bi-Mo-Ti mixed oxides determined by N2 chemisorption, XRD and XPS analysis. Catalyst Type Sbet Phases Surface composition (Crystallite Mo/Bi Mo/Ti [mV] size, [nm]) [at/at] [at/at] 10BiMoTiO-X Xerogel 55 Anatase (13) 3.7 0.13 10BiMoTiO-XCI Xerogel 65 Anatase (13) 0.4 0.08 10BiMoTiO-XCIP Xerogel 56 Anatase (11) 1.3 0.10 10BiMoTiO-LT Aerogel 33 Anatase (8) 1.9 0.08 10BiMoO/Ti0 2 Supported 32 Anatase (23), 0.9 0.16 Rutile (30) 106

The amorphous structure of uncalcined xerogels and the aerogel, and the restructuring at elevated temperature resulting in crystalline titania were confirmed by HRTEM measurements. The FTRaman spectra of the sol-gel materials after calcination possessed only a broad, low intensity unstructured band between 700 and 950 cm-1 which could be due to the presence of very small bismuth molybdate particles [11]. These results indicate that titania as a major component can properly stabilize the Bl­ and Mo-containing phases, despite the calcination at high temperature. The surface Mo/Bi and Mo/Ti ratios varied over a broad range (Table 1), demonstrating that the parameters altered in the preparation of sol-gel mixed oxides had a major influence on the surface composition.

3.2 Surface reduction and re-oxidation at ambient temperature

All sol-gel derived catalysts, except the prehydrolyzed sample 10BiMoTiO-XCIP, showed a reduction of Bi3* to Bi° at room temperature after evacuation in the load lock due to the reducing conditions (reducing gases in the zeolite trap, H2 from the ionization gauge or pump oil vapor). The possible influence of charging effects due to inhomogeneities in the sample could be excluded by applying a tubular voltage to the analyzer. The extent of reduction was strongly dependent on the outgassing time in the load lock, as illustrated in Fig. 1 on the example of the 10BiMoTiO-LT aerogel. It is also shown that the surface layer of the aerogel could easily be re-oxidized at room temperature by exposing the sample to air. The Mo/Bi surface ratio had no detectable influence on the reducibility of the Bi3* species in the mixed oxides. For comparison, after evacuation in the load lock for 48 h, no reduced species could be detected on the surface of crystalline y-Bi 2Mo0 6 and Bi203, which is in accordance with former observations [11 -13j. Similarly, no change was observed after a prolonged treatment in H2 at room temperature. Significant reduction of the Bi3* and Mo 6* species occurred only at 300 °C or above. The surface C-content of the sol-gel catalysts was between 7.5 and 9.5 at% which was similar to that found in conventionally prepared samples. The predominant part of the C 1s peak at 285.0 eV probably stemmed from hydrocarbons (pump oil). A peak fitting procedure with the Mo 6* (232.3 eV) and Mo 5* (231.4 eV) species revealed that the amount of reduced molybdenum was minor, always below 10 %. In contrast, a more facile reduction and re-oxidation to Mo 6* was found in conventional bismuth molybdates (at significantly higher temperatures [12,13]). No reduction of Bi3* and Mo 6* in the sol-gel catalysts was observed under UHV conditions (p<10 -9 mbar). This observation was used for studying the dynamic behavior of the surface redox processes. The time-resolved redox behavior of the surface could be followed quasi in situ by performing the reduction - oxidation in the reactor chamber and the analysis in the XPS chamber. A typical cycle is shown in Fig. 2. The reaction 107

rates strongly depended on the partial pressures in the range 10"6 to 10"4 mbar. Further experiments are required to reveal the mechanism of the reduction - oxidation processes.

166 164 162 160 158 156 154 Binding energy [eV]

Figure 1: High resolution XP spectra of the 10BiMoTiO-LT aerogel in the Bi 4f region after outgassing for 2.5 h (a), after outgassing for 12 h (b) and after re-oxidation in air for 2 h (c), at room temperature.

Time [h] 0 10 20 30 40

Time [min]

Figure 2: Relative intensity of Bi° species present in the 10BiMoTiO- LT aerogel during reduction in 10"6 mbar H2 (♦, top axis) and during re­ oxidation in 10"4 mbar 02 (o, bottom axis), at room temperature. 108

Temperature programmed reduction followed by TG and MS detection could not display any H2 consumption below 180 °C, indicating that only the outer surface layer of the sol-gel mixed oxides could be reduced at room temperature in the XPS chamber. Moreover, the extent of surface reduction seems to be rather small when compared to the total amount of bismuth molybdate in the catalysts.

3.3 Catalytic oxidation of butadiene to furan

In preliminary experiments the influence of some reaction parameters was investigated using the 10BiMoTiO-XCIP xerogel. The furan selectivity increased with decreasing conversion and reaction temperature, and decreasing 02 and C4 concentrations. Above 15 % conversion, the maximum selectivity was achieved with an 02: C4 ratio of 1. (No optimization of the reaction parameters has been attempted.) It emerged from calculations based on product composition that at 25 % conversion or above practically all 02 in the feed was consumed when the 02: C4 ratio was unity or lower. Employing excess 02 in the feed enhanced the conversion and also the CO and C02 formation, but decreased the furan selectivity. The main products of butadiene oxidation were CO, C02 and furan. Other minor oxygenated products such as acrolein, 2,5-dihydrofuran, and benzene were also identified.

Table 2 Performance of Bi-Mo-Ti mixed oxides in the oxidation of 1,3-butadiene to furan. Feed composition: 5 voi-% butadiene, 5 vol-% 02,2 vol-% Kr and 88 vol-% He.

Catalyst at 10 % conversion at 25 % conversion Temp. Selectivity Temp. Selectivity [°C] [%] [°C] [%] 10BiMoTiO-X 520 39 550 24 10BiMoTiO-XCI 450 37 510 32 10BiMoTiO-XCIP 430 48 530 37 10BiMoTiO-LT 410 28 530 30 10BiMoO/TiO 2 450 45 520 29

The catalytic performance of various sol-gel derived and conventionally prepared Bi-Mo-Ti oxide catalysts were compared under identical conditions except for temperature. The conversion was altered by decreasing the temperature in steps of 10 °C. The selectivities at 10 and 25 % conversions are collected in Table 2. The 109

10BiMoTiO-XCIP xerogel prepared by prehydrolysis of the Cl-containing precursors afforded the highest furan selectivity. The low temperature aerogel 10BiMoTiO-LT was the most active but an unselective catalyst. The four sol-gel derived catalysts listed in Table 2 represent rather different materials, due to the variation of the preparation conditions. Still, none of them exhibited markedly better performance in the butadiene to furan transformation than the conventionally prepared supported mixed oxide 10BIMoO/TiO 2.

3.4 Restructuring during butadiene oxidation

XRD, HRTEM and XPS analysis evidenced that exposure to reaction conditions resulted in considerable restructuring of the sol-gel mixed oxides. The changes in crystallinity of the most selective catalyst 10BiMoTiO-XCIP is illustrated in Fig. 3. During reaction the mean diameter of anatase crystallites increased to 30 nm and rutile crystallites (=50 nm) could also be observed. The small reflections at 32.1°, 45.2°, 45.9°, 53.1°, 54.2° and 56.3° indicate the appearance of crystalline y-Bi 2Mo0 6. Similarly, HRTEM analysis of the xerogel 10BiMoTiO-XCIP showed strikingly different catalyst morphology and pronounced crystallinity after reaction.

20 30 40 50 60 70 20

Figure 3: X-ray diffraction patterns of 10BiMoTiO-XC!P xerogel after drying (a), after calcination at 500 °C (b), and after butadiene oxidation at 400-550 °C for 20 h; ♦: rutile, o: anatase. 110

A further indication of restructuring is that, according to XPS analysis of the same sample, ca. 25 % of Bi was present as Bi° after reaction; Mo and Ti remained in their fully oxidized state. (Note that 10BiMoTiO-XCIP was the only catalyst which was not reduced in the load lock during evacuation at room temperature.)

4. Conclusions

The sol-gel derived Bi-Mo-Ti mixed oxides possessed a special morphology: titania as the major component stabilized the Bi- and Mo-containing phases in an X-ray amorphous state. The facile reduction and re-oxidation of Bi3* observed by XPS may be attributed to the increased activity of the lattice oxygen, due to the small size of the bismuth molybdate “crystallites". Application in catalytic oxidation at elevated temperatures resulted in rapid restructuring and loss of the outstanding redox properties at ambient conditions.

Acknowledgement

Financial support by Du Pont de Nemours, Wilmington (DE) is kindly acknowledged.

References

[1] R. Butter, T. Mallat and A. Baiker, J. Catal., 153,177 (1995). [2] G.M. Pajonk, Catal. Today, 35,319 (1997). [3] Y. Moro-Oka and W. Ueda, Adv. Catal., 40,233 (1994). [4] R.K. Grasselli, in “Handbook of Heterogeneous Catalysis ”, Eds. G. Ertl, H. Knozinger and J. Weitkamp, Vol. 5, p. 2302, Wiley-VCH, Weinheim, 1997. [5] N.G. Glukhovskii, I.L.Beriostotskaya and I.S. Vses, Neftekhimia, 26,89 (1986). [6] H.-G. Lintz and A. Quasi, Catal. Lett., 46,255 (1997). [7] M.D. Wildberger, M. Maciejewski, J.D. Grunwaldt, T. Mallat and A. Baiker, Appl. Catal. A, in press. [8] T. Notermann, G.W. Keulks, A. Skiliarov, Y. Maximov, L.Y. Margolis and O.V. Krilov, J. Catal., 39, 286 (1975). [9] J.Y. McGilp, P. Wightmar, and E.Y. McGuire, J. Phys. C: Solid State Phys., 10, 3445 (1977). [10] V.S. Dharmadhikari, S.R. Sainkar, S. Badrinarayan and A. Goswami, J. Electr. Spectr. Rel. Phenom., 25,181 (1982). [11] I. Matsuura, R. Shut and K. Hirakawa, J. Catal., 63,152 (1980). [12] B. Grzybowska, J. Haber, W. Marczewski and L. Ungier, J. Catal., 42, 327 (1976). [13] K. Uchida and A. Ayame, Surf. Sci., 357-358,170 (1996). 111

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

i in inn unit imif *DE012198194* M. Baerns r _ Institut fur Angewandte Chemie Berlin Adlershof e.V., Rudower Chaussee 5 D-12484 Berlin, Germany

CATALYTIC OXIDATIVE CONVERSION OF ALKANES TO OLEFINS AND OXYGENATES

INTRODUCTION DE9 9G2083 Light alkanes, i.e., methane, ethane, propane and n- as well as i-bufane, are readily available as constituents of natural gas and as side products in refinery op ­ eration. They may be considered as a feedstock in chemical industry when directly transformed to petrochemical intermediates. Methane when converted to ethylene by its oxidative coupling and to methanol or formaldehyde may serve as a Ct building block beside of its role for producing synthesis gas by steam reforming or partial oxi­ dation. Light alkanes might replace their olefinic counterparts for oxygenates or other petrochemicals production [1], Autothermal oxidative dehydrogenation of light al­ kanes to the respective olefins is of present interest in research and development; presently alkane dehydrogenation is carried out by thermal dehydrogenation requiring an appreciable amount of heat input into the reactor. A more economic and envi­ ronmentally friendly process version may be achieved by oxidative dehydrogenation. No industrial endeavours have become known by which light alkanes can be transformed to valuable petrochemicals or feedstocks by either oxidative coupling of methane to ethylene, oxidative dehydrogenation (ODH) or by oxidative functionaliza ­ tion of the alkanes except for the commercially applied synthesis of maleic anhydride (MA) and acetic acid from butane. For the reaction of butane to MA by which in par­ ticular benzene and to a minor degree butene were replaced as feedstocks, vanadyl pyrophosphate has unique catalytic properties which are unfortunately not transferable to the oxidative functionalization of either methane, ethane or propane. Presently, the oxidation of ethane to acetic acid and the ammoxidation of propane to acrylonitrile look most promising although the transfer of these developments into an industrial process still remains to be proven. Against the above background it is obvious that there still exist challenging tasks to develop new catalytic materials and engineering solutions to make proc­ esses viable for the reactions mentioned. The spectrum of valuable petrochemicals which could be derived from C, to C„ alkanes by their selective oxidation is summa­ rized below: Alkane Products Methane Ethylene, methanol, formaldehyde, synthesis gas Ethane Ethylene, acetic acid Propane Propylene, acrolein, acrylic acid Butane Butenes, maleic anhydride, acetic acid Details on the oxidative coupling of methane (OCM), its partial oxidation to methanol and formaldehyde, the oxidative dehydrogenation of ethane and propane as well as

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 112

their functionalization are dealt with in this communication; results from literature and partly also from our own work are communicated. Research activities are going on in all the fields mentioned above although em­ phasis is different for the various subjects as may be derived from an assessment of world-wide publications (ca 400) on light alkane oxidation refereed in Chemical Ab­ stracts from January 1997 to July 1998; the following percentages for the various subjects were obtained: Methane: 67 % (oxidative coupling 44 %, formaldehyde 31 %, methanol 5 %, miscel­ laneous without high-temperature combustion and NOx reduction 9%) Ethane: 8 % (ethylene 87 %, acetic acid 13 %) Propane: 16 % (propene 48 %, acrolein 17 %, acrylic acid 11 %, acrylonitrile 24 %) i-Butane: 8 % (i-butene 75 %, oxygenates 25 %)

METHANE CONVERSION The OCM reaction as well as the partial oxidation of methane to methanol and formaldehyde are dealt with. The autothermal catalytic partial oxidation of methane to syngas which has not been included in petrochemical feedstocks and which has been extensively studied from a fundamental point of view (see e.g. [2 to 10]) is not dealt with in the present context.

OCM Reaction Traditional OCM catalysis has reached a mature state in the sense that many catalytic materials have been tested for their performance and no further significant improvements are foreseeable. Selectivities of 80 to 85 % for C2 hydrocarbons have been achieved at methane conversions of up to approximately 10 to 15 % depending on conditions chosen. Increasing the degree of methane conversion which requires an increased oxygen concentration in the feed leads to a loss in selectivity. The highest C2 yields achieved vary between 20 to 25 % applying a one-pass through- the-reactor operation. Recycle of methane is only possible after extensive and expensive separation procedures by which process economics are deteriorating. Therefore, for improving process economics new schemes in reactor operation and effluent separation are needed; e.g. distributed oxygen feed [11, 12], chroma ­ tographic effluent separation [13] and electrocatalytic reactor-separator [14] have been suggested. Comprehensive kinetics of the OCM reaction have also recently become available [15] which makes it possible to model and simulate innovative reaction engineering concepts. Many of the fundamentals of the OCM reaction are now understood. High tem­ peratures (> 1000 K) and basic metal-oxide-type catalysts are required to generate methyl radicals and simultaneously avoiding total oxidation which occurs via methoxy species preferentially formed at low temperature. The formation of methyl radicals and their recombination have been first discovered by Lunsford and co-workers [16] and later confirmed by independent means by high-vacuum transient experiments [17]. It was further shown that the electronic properties of the catalytic solid material play a major role. High oxygen conductivity [18] and p-type conductivity [19] favour the selectivity towards C2 formation. 113

From contact potential differences between the catalyst surface and a refer­ ence electrode conclusions with respect to the type of surface oxygen species could be derived [20]. The present authors believe that there will be only marginal improvements in C2 selectivity within the next future; the fundamental problem that the products are more reactive than methane which then leads to their total oxidation has to be over ­ come. Hence, new catalytic reaction engineering means have to be implemented in any viable OCM process.

Direct conversion of Methane to Methanol and/or Formaldehyde Methanol Direct conversion of methane to methanol had been studied as homogene ­ ously and heterogeneously catalysed gas-phase reaction since many decades with­ out any breakthrough with respect to selectivity and yield of methanol respectively. Recently Gesser and co-workers [21] published data on the homogeneous partial oxidation of methane in a high-pressure tubular reactor (65 bar, 450 °C, contact time 2 min); methanol selectivities of 70 to 80 % at methane conversions of 8 to 10 % per pass were achieved. The same group had studied the effect of solid catalysts (Sn02, Mo0 3/Si02, Sm203, Bi203, Li/ZnO, Mn203/Mn304) at 30 bar and various tem­ peratures [22]; no improvement was observed. To Cesser’s opinion better results should be obtained by using a catalyst facilitating high-rate production of methyl radicals which then react subsequently in the oxygen-containing gasphase to methanol [23]. Periana et al. from Catalytica Inc. reported in 1998 on the oxidation of methane to a methanol derivative CH30S03H which then can be hydrolysed to methanol [24]. The reaction takes place at 3400 kPa at 493 K in concentrated sulphuric acid (102%) containing a 50 mM of a dichloro(q-2-{2,2 ’-bipyrimidyl}platinum(M)[(bpym)PtCI 2] complex concentration. After 2.5 hours 90 % of the methane fed to the reactor were converted with 81 % selectivity to the above methanol derivative. An overall volumetric productivity of 10"6 mol cm"3 s"1 was demonstrated. The method avoids former shortcomings [25] and it appears to be interesting although the productivity of the system is low and should be increased by about one order of magnitude to become competitive to usual catalytic technologies. Another novel approach for the catalytic oxidation of methane to methanol is the conversion of a methane-oxygen-hydrogen mixture as put forward by Wang and Otsuka [26]; depending on temperature selectivities of 89 % (623 K) and 36 % (696 K) were obtained for an iron phosphate catalyst at a low degree of methane conver ­ sion (1 to 8%) [26b]. Russian pilot-plant work has shown that for the homogeneous non-catalytic oxidation up to 60 % selectivities at 3 to 4 % methane conversion for methanol plus formaldehyde were achieved; this result was considered satisfactory for industrial use under the given constraints. Similar results were recently reported by Aoki et al. [27] for the sol/gel Mo0 3/Si02 catalysed partial oxidation of methane at 873 K; total selectivities of methanol plus formaldehyd amounted to about 60 to 80 % formalde- hyd being, however, the major product. 114

Formaldehyde As catalytic material metal oxides, e.g. Mo/Si0 2, B203/Be0/Si02l V20g/Si0 2, and Fe203/Mo0 3, have been used in oxidising methane by either oxygen or nitrous oxide. Methane-to-oxidant ratios varied between 0.3 to 3 at temperatures between 723 and 973 K; particularly, in the high temperature range (>600 °C) total oxidation became dominant. In general, maximum yields of below 5% have been reported. For detailed references see [28]. Mechanistic studies are mainly concerned with the nature of the active surface structure and the surface species and intermediates formed in the reaction. For Mo0 3/Si02 catalysts it is believed that the formation of silicomolybdates causes an increase of the formaldehyde selectivity [29]. Parmaliana and co-workers reported that in selective redox catalysis gasphase oxygen interacts with methane-derived in­ termediates while the participation of bulk-lattice oxygen leads to carbon oxides [30]. Kinetic studies of Sexton and Hodnett [31] indicate that oxygen is adsorbed on silica- supported vanadia catalysts and takes part in the reaction according to an Eley- Rideal or Mars-van-Krevelen type mechanism; the nature of the interaction depends on feed composition, i.e., methane-rich or -lean conditions. The results further show that COx is a consecutive product of HCHO and that the intermediate formation of methanol can be excluded. Spencer [32], however, pointed out that HCHO and CO2 are primary products while CO is formed consecutively. Summarising it can be concluded that the decisive features of a catalytic mate­ rial leading to high formaldehyde selectivity are still unknown. Similar to OCM and methanol formation, the consecutive reaction of the desired product to COx occurring particularly at increasing methane conversion has to be minimised.

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES TO OLEFINS Catalytic ODH of ethane, propane and iso-butane to olefins is an alternative to present thermal dehydrogenation technology. The present state of research for ODH of ethane and propane is presented. Previous works have been summarised e.g. by M.V. Landau et al. [33] and Albonetti et al.[1J. The catalytic ODH of light alkanes is carried out in a temperature range from about 620 to 1450 K. In the case of autothermal operation, the reaction is ignited at low temperature (> 600 K) and proceeds then at a higher temperature level. The heat management has to be tuned in such a way that only as much heat is produced by oxidation as required for maintaining the simultaneous and/or subsequent endo ­ thermic dehydrogenation. Several solutions have been suggested for solving this task. The reaction can be carried out autothermally in a catalytic fixed-bed reactor or in a monolith-type reactor or in a fluidized-bed reactor containing a catalytic packing. Results on the catalysis of ODH of ethane and propane are communicated separately since there has no fully common catalytic scheme for those alkanes emerged yet. General reactor concepts for ODH are discussed thereafter.

ODH of Ethane (ODE) Numerous catalysts have become known: e.g. V-Mo-Nb-0 [34], Li-Mg-0 [35], rare earth oxides [36], Pt- and Rh-coated ceramic foams [37], hydroxyapatites with incorporated lead [38] and complex metal-halide/RE-oxides [39]. Some of the more 115

recent studies which deal with special inorganic structures and which are in some cases promoted by halides - are discussed with respect to performance and funda­ mental aspects for illustration. Calcium hydroxyapatites with incorporated lead have been studied by Moffat and co-workers [38] for ODE; partly, tetrachloromethane (TCM), had been added to the feed gas. A special feature of the catalytic material is its bifunctionality including acidic and basic properties. The catalytic experiments were carried out at 773, 873 and 973 K, p(C2H6)=0.27 kPa and p(02)=6.7 kPa. At 773 K the addition of TCM led to an increase in activity and ethylene selectivity which was ascribed to the formation of chloroapatites on the catalytic surface. With TCM there was evidence for the intrusion of homogeneous gasphase reactions at 873 K; this result may be somehow misleading since homogeneous ethane dehydrogenation starts above 873 K anyhow. In the absence of TCM the activity for the ODE reaction was only little affected at 973 K by the lead content as opposed to lower temperatures; however, selectivity was improved which was attributed to an effect of lead on the reaction pathways of ethyl radicals present in the gasphase. From these results it is concluded that the influence of TCM on selectivity was minor while the addition of lead resulted in an increase of ethylene selectivity from 60 % (no Pb) to about 80 % (5 to 35 % Pb) at complete oxy ­ gen conversion; the balance in selectivity was about 10 % methane and 10 % COx. Unfortunately, the authors do not comment on the above mentioned bifunctionality of the hydroxyapatite with respect to the catalytic results. Beside the frequently described mixed metal oxides, special inorganic struc­ tures, in some cases promoted by halides, have been applied in ODE. Wan and collaborators [39a,b] investigated RE compounds containing fluorine for ODE (a: LaF3-Ce02, b: Sm203-LaF3 promoted by BaF2). The fee-gas composition applied consisted of ethane:oxygen:nitrogen equal to 2:1:7 and the reaction tempera­ ture amounted to 953-973 K (a) and 873 K (6); no data have been provided on possible hot spots occurring in the reactor. - For system "a" ethylene selectivities between 90 to 95 % at 25 to 50 % oxygen conversion were obtained which dropped, however, to ca 30 % for complete oxygen conversion. The authors found that an ionic exchange between the two compounds took place during the calcination process of the solid material leading to the formation of O' ions, anion vacancies and partially reduced Ce4+ centres to which the selectivity behaviour of the catalyst is related. The "inert" fluorides on the surface are assumed to be beneficial for the isolation of active centres and herewith avoiding deep oxidation; (this argument may be related to a si­ milar one for FeP04 catalysts in the methane-to-methanol oxidation). - For system "b" a maximum ethylene selectivity of 84 % was obtained at X(Oz) = 42 % while at X(02) = 94 % selectivity decreased to 42 %. The dependence of catalyst performance is proposed to depend on structural defects and on active site isolation by .inert" fluorides. Au and Zhou have comprehensively characterised ODE SmOF catalysts pro ­ moted by SrF2 and BaF2 [39c]. Special attention was given to the activation of oxygen by applying laser-Raman spectroscopy, TPD of 02, EPR and XRD. Based on the re­ sults reported above they discussed the possibilities of producing O*. 022", and 0‘2 species. For catalytic testing they applied similar conditions and their results were similar to those communicated by Wan and collaborators [39a,b], - It was found that at 893 K the principal oxygen species on the promoted catalysts was O' while over the unpromoted one (SmOF) a certain amount of dioxygen species existed besides 116

mono-oxygen species. It was suggested that due to dioxygen species complete oxi ­ dation of ethane is favoured on unpromoted SmOF. - Similar catalytic results com ­ pared to [39a,b,c] were obtained by Ueda and co-workers [39d,e] using a layered complex metal oxy-chloride (SrBi304CI3). They postulate that the ODE reaction oc ­ curs through a direct interaction with the chloride-containing surface of the catalyst. It is considered that the high selectivity to ethylene can be ascribed to a surface-radical mechanism and that the surface chlorine blocks those active sites which might pro ­ vide active oxygen for total oxidation. Results on the oxidative dehydrogenation of ethane on rare-earth-oxide (REO) based catalysts (Na-P-Sm-O, Sm-Sr(Ca)-0, La-Sr-0 and Nd-Sr-O) are described in [40, 28]. Oxygen adsorption was found to be a key factor which determines the activ­ ity of this type of catalysts. Due to high activity, ignition of the reaction mixture takes place resulting in non-isothermal operation. A catalyst layer of 1 mm was sufficient to ignite the preheated reaction mixture and to sustain the reactions. Contrary to non- catalytic oxidative dehydrogenation, reaction temperatures above 700°C can be achieved without significant external heat input. Ethylene yields of up to 34-45 % (8=66-73 %) were obtained on REO-based catalysts under non-isothermal conditions (Tmax = 810 - 865°C) at contact times in the order of 30 - 40 ms. In conclusion, it can be said that significant progress has been made in under­ standing the catalysis of ODE. Nevertheless, an important problem, i.e., maintaining high ethylene selectivities at complete oxygen conversion and high ethane conver ­ sions, remains still to be solved.

ODH of Propane (ODP) Catalysts based on vanadia, especially V-Mg-0 have been intensively studied [41-43], For V-Mg-0 the pure phases Mg 2V208 , Mg 2V207 and MgV 206 as well as their mixtures are formed in the preparation process depending on the ratio of V-to-Mg applied [41]. It was concluded that the catalytic performance depended on anion va­ cancies which participate in the redox catalytic cycle in which the reoxidation step is rate limiting and on the proper adjustment of strong Lewis acidity and mild basicity. In later work Pantazidis [44a] showed that nucleophilic oxygen takes part in ODP to pro- pene via a Mars-van-Krevelen mechanism while adsorbed electrophilic oxygen ad- species, originating from gasphase oxygen lead to deep propane oxidation; both types of oxygen are assumed to contribute to total consecutive oxidation of propene. It was further postulated that the different forms of reactive oxygen are associated to the same site but different site arrangements (oxidised and superoxidized). In a re­ cent contribution Pantazidis et al. [44b] specified the active site. They assume that much of the catalytic surface consists of dispersed vanadia surface units having inti­ mate crystallographic interaction with magnesia crystallites. A monolayer of mono ­ meric and polymeric species stabilises an unusual polar (111) orientation of MgO up to 1073 K. There exists a totally reversible order/disorder restructuration of this va­ nadia overlayer which is related to the redox state of the surface depending on the reductive/oxidative characteristics of the surrounding atmosphere. The phenomena described determine the elementary reaction steps in ODP. - Fang et al. identified V4+ as the active site of this type of catalyst by EPR [45]. For V20g/Zr0 2 catalysts of different compositions [46] a maximum of activity was observed for vanadia contents between 3 to 5 mol% to which monomeric va­ nadyl and polyvanadate surface species were ascribed; (C3H8 /02/N2 = 1:9:10; 700 K; 117

catalytic fixed-bed reactor). Activity and selectivity depended on the reducibility of the various catalysts. Selectivities between 40 and 80% at propane conversions from 5 to 15 % were obtained; unfortunately, any information about the dependence of selec­ tivity on conversion is missing which makes a complete understanding of the results impossible. Khodakov et al. [47] found in a related study that the structure of the dis­ persed vanadia on zirconia depends on its surface density and preparation condi ­ tions. At low loadings presumably monomeric VOx species exist while at higher loadings polymeric species of polyvanadate prevail. From catalytic and spectroscopic evidence it is concluded that polyvanadates species highly dispersed on zirconia are active sites in OOP. Special attention has been also given to Mg-V-Sb-oxide catalysts (e.g. [48]). The composition Mg 4V2Sb2Ox was identified as the best catalytic material among various alternatives [48a]: Propane is assumed to be oxidised by nucleophilic lattice oxygen; the resulting lattice vacancy is then replenished by gasphase oxygen. An isolated catalytic site appears to play an important role for alkane activation. Vana­ dium oxide forms an inorganic radical (V-O- or V*=0) by which at least one hydrogen is abstracted; the second hydrogen is likely abstracted by lattice oxygen being associated with antimony in Mg 4V2Sb2Ox. Also RE-oxide catalysts have been applied for OOP. Au and Zhang [49] re­ ported on RE orthovanadates (RE: Pr, Gd, Dy, Ho, Er, Nd, Tb, and Lu). At low pro ­ pane conversion they obtained e.g. for TbVO^ (773 K, C3H8 :02:He=20:10:70) a pro ­ pane selectivity of about 80 % (X(Ca)=2 %) which dropped to about 50 % at X(C3)=10 %. They confirm the preposition of other authors (see above) that low valence va­ nadates (V4*) play an important role in ODH. In our own work [50, 51], catalyst properties determining propene formation on various metal oxide catalysts were identified based on different assumptions on the type of oxygen species available for the performance of ODH [52]. The respective materials were selected by means of an expert-system [52]. Results on the oxidative conversion of propane on three types of catalysts on which different mechanisms of propane activation prevail are summarized in [51]. Redox systems operating at tem­ peratures <500°C can produce propylene selectively especially at a high degree of catalyst reduction at which, however, only low propane conversion levels can be achieved. On rare-earth oxide-based catalysts, adsorbed oxygen is involved in pro ­ pane activation. Surface reaction of propane with adsorbed oxygen leads to the for ­ mation not only of propylene and COx but causes also significant formation of ethyl ­ ene, methane and C4 hydrocarbons due to C-C cleavage. Although this type of cata­ lysts is promising for the oxidative dehydrogenation of ethane, their use in the case of propane is limited due to significant formation of ethylene besides propene. On B203- Al203 catalysts, propane is assumed to be activated on Lewis acid sites leading to propyl radicals which then desorb in the gas-phase [51, 53]. The highest propene yield amounted to 22 % (S = 45 - 48 %) on B2O3(30 wt. %)ZAI203; in addition, yields to C,-C3 oxygenates of up to 8 % were achieved [5]. It was shown that the coordination of boron influences catalyst activity; trigonal B03 species being present in both crys ­ talline and amorphous phases are active in the dehydrogenation of propane to pro ­ pene. In summarising, one might draw a similar conclusion as for ODE. Our under­ standing of OOP catalysis has certainly increased; nevertheless, the further suppres ­ sion of non-selective reaction steps depending on the different oxygen species or 118

sites respectively needs still to be resolved. In OOP as compared to ODE there is an additional selectivity loss due to C-C bond fission leading to methane which has to be addressed in further work.

Reactor concepts Isothermal operation which is usually applied in catalyst testing is hard to ac­ complish in industrial practice due to the exothermic and endothermic reactions tak­ ing place consecutively as well as simultaneously. For the ODH reactions often short- contact-time reactors are required. Monolith reactors were suggested by Schmidt and collaborators [see e.g.[55]) and Capannelli et al. [56]. In [55] the reactor consists of a ceramic-foam on which noble metals were deposited and in [56] the V2O5/AI2O3 catalyst was coated as a thin layer on a cylindrical ceramic support; in such reactors diffusions! limitations of the reaction are assumed to be avoidable and particularly in [56] the pressure drop is low. Another solution for carrying out the ODH reactions is a fluidized-bed reactor with an internal catalytic packing [57]. The packing serves the purpose of converting part of the alkane via the exothermic catalytic ODH reaction; the endothermic catalytic or non-catalytic thermal dehydrogenation of unconverted alkane is continued in the fluidized bed to which the heat generated in ODH is trans­ ferred. In all the reactors an autothermal reactor-operating concept prevails as has been described by Choudary [58]. A comparison of different reactor configurations (catalytic packed-bed, monolith-like and membrane reactors) has been reported by Capanelli et al [59]. Summarizing at this point it appears that the knowledge and experience exist­ ing on catalysis and catalytic reaction engineering of ODH could be the basis for pilot- plant work required for any final judgement with respect to transferring ODH into practice. The industrial interest in ODH becomes obvious by the patents and patent applications in this area (see e.g. [57]).

PARTIAL OXIDATION OF ETHANE AND PROPANE TO OXYGENATES The partial oxidation of ethane and propane suffers in a similar manner as methane from the increased reactivity of the desired product for consecutive oxida ­ tion steps as compared to the alkane feed. That is to say, extended contact times with the catalyst corresponding to increased conversion of the alkane leads to de­ creasing product selectivity in favour of COx. Among the light alkanes only butane leads to a rather stable product, i.e., maleic anhydride, which is less subjected to total oxidation. The main task of present research is to find catalysts designed to reduce consecutive total oxidation of the desired oxygenates derived from ethane (acetic acid) and from propane (acrolein and acrylic acid as well as of acrylonitrile by am- moxidation). Trifiro [1] has recently described the chemistry and the implications in preparing selected catalysts based on various molybdates of Bi, Mn, Co, Fe, Te, and Ce as well as titania-supported vanadia and V-P-O; these methods may be trans­ ferred to other oxide systems. He considered such characteristics as redox proper ­ ties, Lewis and Bronsted acidity, basic properties, surface coordination, types of neighbours near the transition element, lattice oxygen mobility and the presence of defects for activation of molecular oxygen. In the following, more recent results on the partial oxidation of ethane and propane are communicated. 119

Ethane to Acetic Acid The catalytic oxidation of ethane to acetic acid was first reported by Thorn- steinson et at. [61]. Using an unsupported Mo-V-Nb-O catalyst (BOOK; 2 MPa) acetic acid selectivities of 26 % were obtained at low ethane conversion (5 %). In a patent issued to BP [62] in 1993 selectivities of 78 % (XC2 = 14 %, 550 K; 2,8 MPa) were reported for a Mo-Re-V-Nb catalyst when water vapor was present. In more recent patent applications advanced catalysts of better performance have been claimed. Selectivities of 70 % (XC2 = 8 %) were communicated by Bordes et at. [63] for vana­ dium containing heteropoly acids (Ha^PMOt^ViO^; i = 1,2,3) supported on Ti02. Bor- chert et al. [64] claimed that adding palladium enhances the selectivity towards acetic acid of Mo-Re-V-Nb catalysts (S = 91 %, X# = 4 %) and of Mo-V-Nb catalysts (S = 80 %, XC2 = 10 %) [65]. In general, the reaction was carried out at elevated pressures and in an excess of ethane. Adding water increases the selectivity towards acetic acid. The present data suggest that such an ethane-to-acetic acid process might eventually be put into practice. From a fundamental point of view results of Ruth et al. [66] and Merzouki et al. [67], Tessier et al. [68] and Roy et al. [69] are of importance. Ruth etal. [70] studied the catalytic performance of Mo-V-Nb oxides and of the individual phases therein which are obviously an essential part of (nearly) all catalysts used so far. The crystal ­ line phases found were Mo 6Vg O40, Mo 3Nb2011, and Mo0 3; furthermore, an amor ­ phous part was identified of the approximate composition Mo MV13Nb2Ox to which the selectivity towards acetic acid is ascribed based on a comparison with the catalytic performance of the pure phases. Tessier et al. [71] studied pure VPO and titania- supported VPOx and VOx. They reported that polyvanadates are the active specie responsible for the formation of acetic acid. Roy et al. [72] reported for titania sup ­ ported MoVPO catalysts that a different Mo A/ ratio does not change the selectivity towards acetic acid at iso-conversion and concluded that only vanadium sites partici­ pate in acetic-acid formation. The role of molybdenum is assumed to influence the electronic density around vanadium sites From scientific publications and patents it appears that for an ethane-to-acetic acid process either a (multitubular) fixed-bed [62, 64, 65, 70] or a fluidized-bed rector [62,64,. 65,70,71] may be used. Higher selectivities and yields respectively depend on further catalyst devel­ opment but also on reaction engineering improvements in carrying out the reaction.

Propane to Acrolein Compared to current technology consisting of two process stages when using propane, i.e., its dehydrogenation to propene and subsequent conventional oxidation to acrolein, a direct one-stage route could become an attractive alternative if selec­ tivities and yields respectively are high. Various multicomponent oxide catalysts have been studied (VPO, BiMoO, BiMoVAgO, BiBaTeO, and BPO) [72-74]. The perform ­ ance data published indicate that the reaction leads to non-satisfactory selectivities. Maximal yields of 13 % were reported by Kim [72a] which were obtained under con ­ ditions where a primary gasphase dehydrogenation of propane occurred [72b]. Lit­ erature [72, 73] and our own data [74] show that maximum acrolein selectivities did not exceed 60 % even at low propane conversions. Total oxidation of propane as well 120

as consecutive oxidation of acrolein contribute to a loss in selectivity. Yields reported are generally below 3 to 5 %. The fundamental requirements for suitable catalysts are still not well under­ stood although knowledge has been gained with respect to some isolated aspects. The redox potential of the catalytic material certainly plays an important role; specifi­ cally for an AgMo0 3P2014 catalyst the ratio between Mo'' and Mo'' 1 should be around 2 [75]; the silver cations seem to migrate from the bulk to the surface inducing elec­ tronic exchanges in Mo 3015 units in this way stabilizing the assumed active phase [75]. As a more general conclusion it was stated that e.g. in Me-Bi-Mo-0 catalysts (Me: Ca, Mg, Zn) partly promoted with K4P207 different types of lattice oxygen are involved in propane dehydrogenation and in its consecutive selective and non-selec- tive oxidation to acrolein and COx, respectively [74]. The catalyst design approach by combining two active phases (dehydrogenation and oxygen insertion) had severe shortcomings since the two active phases did not act co-operatively to any large ex­ tent. A two layer fixed-bed reactor was required to obtain an acrolein yield of 7.4 % (S=20 %) [74]. One problem to be overcome is the problem of the required high temperatures in oxidative propane dehydrogenation which lead to oxidative degradation of acrolein formed. No commercialisation of this reaction is presently likely even if there are in­ dustrial patent applications (see e.g.[76]).

Oxidation of Propane to Acrylic Acid In 1986 Ai [77] reported on the partial oxidation of propane to acrylic acid using V205-P205-based catalysts. Selectivities of about 60 % were obtained for propane conversion below 10 %; with increasing conversion selectivity decreased due to con ­ secutive oxidation of acrylic acid to COx. Selectivity could be improved by feeding water vapor along with the propane-oxygen feed. Maximum yields amounted for op ­ timum conditions to ca. 10 % (catalysts: Te/PA/ = 0.1-0.15/1.5/1). Mizuno et al. [78] published a slightly improved 13%-yield of acrylic acid in 1995 using a substituted heteropoly acid, i.e., Cs2 5Fe0 08 Hl 26PVMo 11O40. The crucial phases of the catalytic material were later identified by Volta ’s group [79]: A VOP04 phase (mainly delta) associated with poorly crystallized (V0)2P207 was observed in the selective catalysts; the catalytic performance of these phases is ascribed to their Bransted- and Lewis- acidic sites. As an important side product acidic acid is formed under all conditions. No real progress has been reported in the scientific literature; in a very recent publi ­ cation of Ai only acrylic acid yields of 7.5 % were described. Nevertheless, the indus­ trial interest appears to be high in the reaction as may be derived from the numerous patent applications [80].

Ammoxidation of Propane to Acrylonitrile Ammoxidation of propane dates back to patents of Guttermann et at. [81a,b] which initiated extensive fundamental and applied research work. Cent! et al. [82] using Sb-V oxide catalysts obtained selectivities of about 60 % (optimum Sb:V ratio 2 to 3) corresponding to a yield of about 40 %. This result is a progress against past work. The performance of these catalysts was related to the Sb/V ratio, the Lewis-acid sites and the different vanadium sites resulting from the different phases prevailing [83]. A very comprehensive review on V-Sb oxide cata­ . 121

lysts for propane ammoxidation covering many of the more fundamental aspects was recently published by Cent! et al. [84]; supplementing mechanistic work from the groups of Anderson [85], Zanthoff (e.g. [86,87]) and Albonetti [88] is of interest. A further step forward towards an industrially viable process are results obtained by Ushikubo et al. (1997b) for Mo-V-Nb-Te oxide catalysts with yields of about 45 % (693 K, C3H8 /NH3/air = 1:(0.7 to 1.5): 15) during a stable 2000-h experiment. - The progress made in the ammoxidation of propane to acrylonitrile gives hope that there will be further improvement in the performance of the catalytic materials which will be certainly supported by advances in catalytic reaction engineering for this reaction.

CONCLUSIONS |ah of the direct reaction schemes described and the corresponding process schemes are still in an exploratory state. Ethylene by oxidative coupling of methane could become competitive if process schemes are developed with significantly less expenditures for separation of the product from unconverted feed. No encouragement for formaldehyde from methane can be presently derived from the existing knowl ­ edge. Liquid-phase oxidation of methane to methanol appears to be attractive but no final judgement is possible at present. Oxidative dehydrogenation of ethylene and propane look promising although further catalyst improvement is required. Acetic acid from ethane and acrylonitrile from propane have a certain potential as an alternative to present technology. The outlook for acrolein and acrylic acid from propane is less favourable; new concepts for catalyst design are necess

REFERENCES [1] Albonetti, S„ Cavani, F„ Trifiro, F.; Stud. Surf. Sci. Catal. 110.19 (1198) [2] Dissanayake, D., Rosynek, M.P., Kharas, K.C.C., Lunsford, J.H.; J. Catal. 132. 117 (1991). [3] Buyevskaya, O.V., Wolf, D„ Baems, M.;Catal. Lett., 23, 249 (1994). [4] Tsang, S.C., Claridge, J.B., Green, M.L.H.; Catal. Today 23, 3 (1995). [5] Wallens, E.P.J., Hoebink, J.H.B.J., Marin, G.B.; Catal. Lett. 23. 291 (1995). [6] Buyevskaya, O.V., Walter, K„ Wolf, D. Baems, M.; Catal. Lett. 23, 81 (1996). [7] Heitnes-Hofstad, K., Hoebing, J.H.B.J., Holmen, A., Marin, G.B.; Catal. Today 4Q, 157 (1998). [8] Tang, S„ Lin, J„ Tan, K.L.; Catal. Lett. 51(3,4) 169 (1998). [9] Au, C.T., Liao, M.S., Ng„ C.F.; Phys. Chem. A 132(22), 3959 (1998). [10] Miao, Q„ Xiong, G„ Sheng, S., Cui, W., Xu, L. Guo, X; Appl. Catal., A 154(1-2, 17 (1997). [11] Hinsen, W„ Baems, M.; US 4, 608,449 (1986). Baems, M„ Buyevskaya, O.V., Mleczko, L., Wolf, D.; Stud. Surf. Sci. Catal. 12Z (Natural Gas Conversion IV, 421 (1997). [12] Santamaria, J.A., Memendez, M., Pena, J.A., Barahona, J.I.; Catal. Today 12, 353 (1992). [13] Tonkovich, A.L., Carr. R.W., Arts, R.; Science 262,221 (1993) [14] Yenekakis, I.V., Makri, M„ Jian, Y., Vayenas, C.G.; Prep. ACS Division of Petroleum Chemistry, In., 41(1). 119 (1996). [15] Stansch, Z., Mleczko, L., Baems, M.; Ind. Eng. Chem. Res. 26.2568 (1997). [16] Driscoll; D.J., Martin, W„ Wang, J.-X., Lunsford, J.H.; J. Am. Chem. Soc. 1QZ. 58 (1985). 122

[17] Buyevskaya, O.V., Rothaemel, M„ Zanthoff, H.W., Baems, M.; J. Catal. 145, 346 (1994). [18] Borchert, H., Baems, M.; J. Catal. 168 . 315 (1997). [19] Dubois, J.-L.; Rebours, B., Cameron, C.J.; Appl. Catal. 6Z, 73 (1990) [20] Gayko, G„ Wolf., D„ Kondratenko, E.V., Baems, M.; J. Catal. (accepted for publication). [21] Yariagadda, P.S., Morton, L.A., Hunter, N.R., Gesser, H.D.; Ind. Eng. Chem. Res. 2Z, 252 (1998). [22] Hunter, N.R., Gesser, H.D., Morton, L.A., Yariagadda, P.S.; Appl. Catal. §Z, 45 (1990). [23] Gesser, H.D.; Third Workshop .CrC3 Hydrocarbon Conversion", Krasnoyarsk, July 14- 17 1997, oral contribution. [24] Periana, R.A., Taube, D.J., Gamble, S., Taube, H., Satoh, T., Fujii, H.; Science 250, 560 (1998). [25] (a) Periana, R.A., Taube, p.J., Taube, H„ Evitt, E.R.; US 5,306,855 A (1994). (b) Periana, R.A., Taube, D.J., Douglas, J., Evitt, E.R., Loftier, D.G., Wentrcek, P.R., Voss, G., Masuda, T.; Stud. Surf. Sci. Catal. fil, 533 (1994). (c) Periana, R.A.; Adv. Chem. Ser. 252,61 (1997). [26] (a) Wang, Y„ Otsuka, K.; J. Catal. 155, 256 (1995). (b) Wang, Y„ Otsuka, K.; J. Catal. 1Z1, 106 (1997). (c) Wang, Y.. Otsuka, K.; J. Mol. Catal. A: Chem. HI, 341 (1996). [27] Aoki, K., Ohmae, M., Nanba, T., Takeishi, K., Ueno, A., Ohfune, H., Hayashi, H., Udagawa, Y.; Proceedings of JECAT, Tsukuba, Japan, Nov. 25-28,1997, 33 [28] Baems, M.; Buyevskaya, O.; Proceedings of JECAT, Tskukuba, Japan, Nov. 25-28, 1997, 21.; Catal. Today, in print. [29] Barbaux, Y„ Elmrani, A.R., Payen, E., Gengembre, L., Bonnele, J.-P., Grzybowska, B.; Appl. Catal., 44,117(1988). [30] (a) Parmaliana, A., Sokolovskii, V., Miceli, D., Arena, F., Giordano, N.; J. Catal. 148 . 514 (1994). (b) Parmaliana, A., Sokolovskii, V., Arena, F„ Frusteri, F„ Miceli, D.; Catal. Lett. 4Q, 105 (1996). (c) Parmaliana, C.A., Sokolovskii, V, Miceli, D, Giordano, N.; Appl. Catal. A: General 135. L1 (1996). [31] Sexton, A.W., Hodnett, B.K.; Stud. Surf. Sci. Catal. 110,1129 (1997). [32] Spencer, N.D.; J. Catal. m 187 (1988). [33] (a) Landau, M.V., Kaliya, M.L., Herskowitz, M., van den Oosterkamp, P.V., Pocque, P.S.G.; CHEMTECH February 1996 24. (b) Albonetti, S., Cavani, F., Trifiro, F.; Stud. Surf. Sci. Catal. 115,19 (1198) [34] Thorsteinson, E.M., Wilson, T.O., Young, F.G., Kasai, P.H.; J. Catal., 52.116 (1978). [35] (a) Morales, E„ Lunsford, J.H.; J. Catal., 115,255 (1989). (b) Conway, S.J., Lunsford, J.H.; J. Catal., 121, 513 (1991). (c) Conway, S.J., Wang, D.J., Lunsford, J.H.; Appl. Catal. A, Z2, L1 (1991). (d) Swaan, H.M., Toebes, A., Seshan, K„ Ommen, J.G.V., Ross, J.R.H.; Catal. Today 12, 629 (1992). [36] (a) Korf, S.J., Roos, JA, Diphoom, J.M., Veehof, R.H.J., Van Ommen, J.Q., Roos, J.R.H.; Catal. Today 4, 279 (1989). (b) Bernal, S„ Martin, G.A., Moral, P., Perrichon, V.; Catal. Lett. 6, 231 (1990). (c) Kennedy, E.M., Cant, N.W.; Appl. Catal. Z5. 321 (1991). [37] (a) Tong, Y„ Rosynek, M.P., Lunsford, J.H.; J. Catal. 125, 291 (1990). (b) Huff, M„ Schmidt, L.D.; J. Phys. Chem. 2Z. 11815 (1993) [38] Sugiyama, S.. Abe, K„ Miyamoto, T., Hayashi, H., Moffat, J.B.; J. Mol. Catal. 120, 297 (1998). [39] (a) Luo, J.Z., Wan, H.L.; Appl. Catal. A: General 152 137 (1997). (b) Luo, J.Z., Zhou, X.P., Chao, Z.S., Wan, H.L.; Appl. Catal. A: General 15S, 9 (1997). (c) Au, C.T., Zhou, X.; J. Chem. Soc, Faraday Trans., 22 (3), 485 (1997). (d) Ueda, W., Lin, S.W., Tohomoto, I.; Catal. Lett. 44,241 (1977). (e) Lin, S.W., Kim, Y.C., Ueda, W.; Bull. Chem. Soc. Jpn. Zl, 1089 (1998). 123

[40] O.V. Buyevskaya and M. Baems, DGMK-Tagungsbericht (1998) .Selective Oxidations in Petrochemistry", in press. [41] Kung, M.C., Kung, H.H.; J. Catal. 134, 668 (1992). [42] Guerrero-Ruiz, A., Rodriguez-Ramos, I., Fierro, J.L.G., Soenen, V., Hermann, J.M., Volta, J.C.; Stud. Surf. Sci. Catal. Z2.203 (1992). [43] Pantazidis, A., Auroux, A., Herrmann, J.-M., Mirodatos, C.; Catal. Today 22, 81 (1996). [44] (a) Pantazidis, A., Bucholz, S.A., Zanthoff, H.W., Schuurman, Y„ Mirodatos, C.; Catal. Today 40, 207 (1998). (b) Pantazidis, A., Burrows, A., Kiely, C.J., Mirodatos, C.; J. Catal. HZ, 325 (1998). [45] Fang, Z., Weng, W., Wan, H., Cai, Q.; Chinese Science Bulletin 42(2), 172 (1997). [46] Albrecht, S., Wendt, G„ Lippold, G., Adamski, A., Dyrek, K.; Solid State Ionics 101-103. 909 (1997). [47] Khodakov, A., Yang, J., Su, S„ Iglesia, E„ Bell, A.; J. Catal. 1ZZ, 343 (1998). [48] (a) Michaels, J.N., Stem, D.L., Grasselli, R.K.; Catal. Lett. 42.139 (1996). (b) Carrazan, S.R.G., Peres, C., Bernard, J.P., Ruwet, M., Ruiz, P., Delmon, B.; J. Catal. 15S, 452 (1996). [49] Au, C.T., Zhang, W.D.; J. Chem. Soc., Faraday Trans. 93(6), 1195 (1997). ... [50] Baems, M., Buevskaya, O.V., Korting, E. Proc. 15th North American Society Meeting, Chicago, May 18-22 (1997). [51] Baems, M., Buyevskaya, O.V., Korting, E.; Proc. 15th North American Catalysis O. V. Buyevskaya, M. Baems; Catal. Today, 42. 315 (1998). [52] Korting, E. and Baems, M., Proceedings of the MRS 1996 Fall meeting, Boston, Dec. 02-05 (Symposium S), Materials Research Society, Pittsburgh (1996). [53] Buyevskaya, O.V., Kubik, M„ Baems, M.; ACS Symp. Ser. 62B. 155 (1996). [54] Buyevskaya, O.V., MOIIer, D. Pitsch I., Baems, M.; Stud. Surf. Sci .Catal., in print. [55] Yokoyama, C., Bharadwaj, S.S., Schmidt, L.D.; Catal. Lett. 22,181 (1996). [56] Capannelli, G, Carosini, E., Monticelli, O., Cavani, F., Tritiro, F.; Catal. Lett. 39,241 (1996). [57] Baems, M., Buyevskaya, O.V.; Deutsche Patentanmeldung 197 345 41.7 (1997). [58] Choudary, V.R., Uphade, B.S., Mulla, S.A.R.; Angew. Chem. Int. Ed. Engl., 24(6), 665 (1995). [59] Capannelli, G„ Carosini, E„ Cavani, F„ Monticelli, O., Trifiro, F.; Chem. Eng. Sci. 51(10), 1817(1996) [60] (a) Kolts, J.H., Guillory, J.P.; US 4,672,145 (1987). (b) Kung, H.H., Chaar, M.A., US 4,777,319 (1988). (c) Kolts, J.H., Guilleroy, P.; US 4,895,823 (1990). (d) Agaskar, P.A., Grasselli, R.K., Michaels, J.N., Reischmann, P.T., Stem, D.L., Tsikoyiannis, J.G.; US 5,430,209 (1995). (e) Umansky, B.S., Boyer, K.A., Hsu, C.-Y.; US 5,510,558 (1996). (f) Agaskar, P.A., Grasselli, R.K., Michaels, J.N., Reischman, P.T., Stem, D.L., Tsikoyiannis, J.G.; US 5,527,979 (1996). (g) Agaskar, PA, Grasselli, R.K., Michaels, J.N., Reischman, P.T., Stem, D.L., Tsikoyiannis, J.G.; US 5,563,314 (1996). (h) Agaskar, P.A., Grasselli, R.K., Michaels, J.N., Reischman, P.T., Stem, D.L., Tsikoyiannis, J.G.; US 5,530,171 (1996). (i) Bharadwaj, S.S., Schmidt, L.D.; US 5,639,929 (1997). (k) lezzi, R„ Bartolini, A; US 5,633,421 (1997). [61] Thomsteinson, E.M., Wilson, T.P., Young, F.G., Kasai, P.H.; J. Catal. 52,116 (19978). [62] (a) Kitson, M„ US 5,210,293 (1993); (b) Kitson, M„ EP 0 407 091 (1991). [62] Kitson, M„ US 5,210,293 (1993; (b) Kitson, M., EP 0 407 091 (1991). [63] Bordes, E., Gubelmann.Bonneau, M., Tessier, L.; EP 0 683153 (1995]. [64] Borchert, H., Dingerdissen, U., Weiguny, J.; Ger. Often. DE19 620 542 (1997). [65] Borchert, H., Dingerdissen, U.,; Ger. Often. DE 19 630 832 (1998). [66] (a) Ruth, K:, Kieffer, R„ Burch, R.; J. Catal. H5.27 (1998); (b) Ruth, K., Kieffer, R.. Burch, R.,; J. Catal. 1Z5, 16 (1998). [67] Merzouki, M., Taouk, B., Monceaux, L., Bordes, E., Courtine, P.,; Stud. Surf. Sci. Catal. Z2, 165 (1992). 124

[68] Tessier, L, Bordes, E., Gubelmann.Bonneau, M.,; Catal. Today 24, 335 (1995), [69] Roy, E., Gubelmann-Bonneau, M., Ponceblanc, H„ Volta, J.-M.; Catal. Lett. 42, 93 (1996). [70] Hallet, C„; EP 0 480 594 (1992) [71] Benkalowycz, N.C., Blum, P.R., Wagner, D.R.; US 5,300,684 (1994). [72] (a) Kim, Y.-C., Ueda, W„ Moro-oka, Y.; Appl. Catal. ZQ, 175 (1991). (b) Kim, Y.-C., Ueda, W„ Moro-oka, Y.; Catal. Today IS, 673 (1992). [73] (a) Takita, Y., Yamashita, H., Moritaka, K.; Chem. Lett. 1733 (1989). (b) Matsuura, I., Kimura, N.; Stud. Surf. Sci. Catal. 82,271 (1994). (c) Barrault, J„ Magaud, L.; Stud. Surf. Sci. Catal. 82, 305 (1994). [74] Baems, M., Buyevskaya, O.V., Kubik, M., Maiti, G., Ovsitser, O., Seel, O.; Catal Today 23. 85 (1997). [75] Savary, L„ Costentin, G„ Mauge, F„ Lavalley, J.C., El Fallah, J., Studer, F„ Guesdon, A., Ponceblanc, H.; J. Catal. 168,287 (1997). [76] Tenten A., Schildberg, H.P. (BASF), Ger. Often. DE 19,622,331 (1996). [77] Mamoru, A:, J. Catal. 101, 283 (1986). [78] Mizuno, N., Tateishi, M., Iwamoto, M.; Appl. Catl. A General 128, L165 (1995) [79] Hightower, J.W., Delgass, W.N., Iglesia, E„ Bell, A.T.; Stud. Surf. Sci. 161,1205 (1996). [80] Mamoru, A.; Cat. Today 42, 297 (1998). [81] (a) Guttermann, A.T., Grasselli, R.K., Bradzil, J.F.; US Patent 4,746,641 (1998). (b) Guttermann, A.T., Grasselli, R.K., Bradzil, J.F.; US Patent 4,788,317 (1998). [82] Centi, G„ March!, F., Perathoner, S.; J. Chem. Soc., Faraday Trans., 22(24), 5151 (1996). [83] Centi, G„ Mazzoli, P„ Perathoner, S.; Appl. Cat. A: General 165. 273 (1997). [84] Centi, G„ Perathoner, S., Trifiro, F.; Appl. Cat. A: General 15Z, 143 (1997). [85] Nilsson, R., Andersson, A.; Ind. Eng. Chem. Res. 26, 5207 (1997). [86] Zanthoff, H.W., Schaefer, S„ Wolf, G.-U.; Appl. Catal. A: General 164,105 (1997). [87] Zanthoff, H.-W.. Buchholz, S.A.; Catal. Lett. 43, 213 (1997). [88] Albonetti, S., Blanchard, G., Burattin, P„ Cassidy, T.J., Masetti, S., Trifiro, F.; Catal. Lett.45,118(1997). 125 ffofrt. I/GO

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*05012198200* T. Badstube*, H. Rapp*, P. Kustrowski ”, R. Dziembaj** * Institut fur Technische Chemie, Universitat Leipzig, Linn6str. 3, Germany ** Faculty of Chemistry, Jagiellonian University, Crakow, Poland

HETEROGENEOUS CATALYTIC OXIDATIVE DEHYDROGENATION OF ETHYLBENZENE TO STYRENE WITH CARBON DIOXIDE

introduction: DE99G2082

About 90% of the world production of styrene is commercially produced by thermal dehydrogenation of ethylbenzene using iron oxide based catalysts. This process is usually realized at about 600°C with a conversion of approximately 50% of ethylbenzene and a selectivity of styrene up to 90 %. This dehydrogenation is strongly endothermic, and it demands large amounts of energy needed to preheat huge amounts of water (watenethylbenzene « 10-15) up to at least 650°C. The excess of water is necessary to realize this endothermic dehydrogenation, to shift the equilibrium owing to diminishing partial pressure of reactants, and to gasify coke formed on deactivated catalyst. Since the working catalysts is always covered by a certain amount of carbon deposits it can be assumed that water prevents the transformation of active into inactive deposits [1], The shift of thermodynamic equilibrium of this strongly endothermic reaction towards styrene needs either a sufficient increase in the reaction temperature or a lowering of the pressure or a removal of hydrogen from the reaction system. The latter may be achieved by oxidation of hydrogen to water. This can provide both the necessary energy for the otherwise endothermic reaction and for removal of hydrogen. Oxygen addition to the ethylbenzene dehydrogenation system [2], led to a high amount of ethylbenzene was burnt to carbon oxides. The idea of our work was to apply carbon dioxide as an oxidative reactant instead of oxygen, the latter being also responsible for the total oxidation of ethylbenzene as well as styrene and the carbonaceous deposits. At present, it seem to be proven that the active carbonaceous deposit on the acidic catalysts form active centers for the oxidative dehydrogenation [3], Addition of small amounts of carbon dioxide into the traditional process, i.e. into the water/ethylbenzene mixture, gave negative results. The styrene yield decreased and the lifetime of the catalysts was shortened [4]. Only Sugino etal. [5] published

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 126

experimental results of ethylbenzene dehydrogenation in present of excess of CO2 without water. They used active carbon catalysts impregnated by iron, alkali and alkaline earth nitrates and proposed a mechanism with ferrites acting as the active phase. In the present paper we will deal with the dehydrogenation of ethylbenzene to give styrene catalyzed by iron oxide supported on an activated carbon in the presence of carbon dioxide and in absence of water.

EXPERIMENTAL:

Active carbon from Chemviron-Carbon with an ash content of 0.2 wt. % was used as support. It was initially activated with concentrated HN03 at 90°C for 1h. Then it was impregnated with an aqueous solution of iron(lll)nitrate to obtain a loading of 5 wt. % of Fe on the sample and dried in air at 180°C for 2h. Then the samples were impregnated with an aqueous solution of alkali nitrate (Li, Na, K.) or alkaline earth nitrate (Be, Mg, Ca). The obtained samples were dried again at 180°C for 2h. Iron, nickel and cobalt loaded carbonized PPAN, PC, inorganic supports like AI2O3, Si02/Zr02 or Ti02 respectively and commercial iron catalysts applied for styrene production were also used in our tests. Preliminary tests showed the beginning of ethylbenzene cracking in the gasphase noticeable above 580°C. Therefore all the catalytic screening tests were carried out within a temperature range of 350°-550°C in a microcatalytic fixed bed reactor. Prior to each experiment the catalysts were outgassed in a helium stream at 350°C. The catalytic tests were performed under the following conditions: T=350°- 550°C, 100 mg catalyst, total flow 100 ml/min, 3 vol. % ethylbenzene, molar ratio carbon dioxide/ethylbenzene 10:1, and helium as carrier gas. The more promising catalysts were tested under different conditions with respect to total flow, the catalyst mass, the ratio of carbon dioxide/ethylbenzene and the molar ratio of iron to promoter. The effluent of the reactor was analyzed using a Shimadzu GC system consisting of a FID for hydrocarbons and a TCD for permanent gases. Helium was applied as carrier gas. A computer aided process control unit was used for controlling and data acquisition.

RESULTS AND DISCUSSION:

The catalytic activity of active carbon supported iron catalysts was clearly confirmed. The yield of styrene obtained on pure active carbon was equal to 9.5 % at 550°C. The yields increased to 48 % for Fe-loaded active carbon up to 71 % for Li-promoted 5 % Fe/active carbon catalysts (Fig. 1). 127

I

-*-5% Fe/D90

-tV-5% Fe/D-90+Li

S 30 -

450 temperature [°C]

Fig. 1: Comparison of the yield of styrene obtained on differently promoted catalysts as a function of the reaction time

Alkali metals were better promoters than alkaline earth metals. The activity of the catalysts promoted with different alkali metals did not differ much. High catalytic conversion (> 60 %) of ethylbenzene with high selectivity towards styrene (> 90 %) was obtained at 550°C with carbon supported iron catalysts promoted with alkali metals (Fig. 2).

-OK

temperature [°C]

Fig. 2: Comparison of the yield of styrene obtained on alkali metal or alkaline earth metal promoted 5 % wl Fe/active carbon catalysts respectively

i 128

The influence of pretreatment with a mixture of carbon dioxide in helium and pure helium respectively was examined, but no significant effect of pretreatment period was observed in the range between 15 to 60 min (Fig. 3). The catalytic activity after pretreatment were comparable with the activity of the fresh catalyst. Pretreatment only in He-atmosphere led to a decrease in the catalytic activity after 30 min. however with a mixture of He and COa the yield of styrene was stable.

abed

□ after 30 min of reaction time ■ after 60 min of reaction time

Fig. 3: Influence of pre-treatment of the catalytic activity on 5 % Fe/active carbon+K catalyst at 550°C in He a) 15 min b) 30 min in He + C02 c) 15 min d) 30 min

The yield of styrene on the Li-promoted 5% Fe/active carbon catalyst as a function of temperature at different molar ratios of carbon dioxide to ethylbenzene is shown in Fig. 4. All three alkali promoted catalysts were tested with the following COztethylbenzene molar ratios: 0, 1, 5, 10, 20. All three catalysts showed a similar behavior. A strong increase in the yield of styrene was observed with an increase in the amount of carbon dioxide in the reaction mixture. A ratio of 10:1 led to the highest activity. At a CO2/EB ratio of 20 a decrease in catalytic activities of all tested samples was observed. The reason for this fact may be the formation of inactive metal carbonates on the surface of the catalysts in presence of large amounts of CO2. 129

without C02 -e- 1:1

-x- 10:1

450 temperature [*C]

Fig. 4. Yield of styrene vs. reaction temperature on Li-promoted 5 % Fe/active carbon catalyst at different molar ratio of C02 to ethylbenzene

The molar ratio of alkali metal to iron of 1:10 led to the highest catalytic activity. After varying the molar ratio to 1:5 or 1:20 respectively the yield of styrene decreased in both cases (Fig. 5). In addition to styrene, benzene, toluene, carbon monoxide and water were formed as products.

-o-K -o-Na

molar ratio iron-promoter

Fig. 5: Yield of styrene vs. molar ratio of iron to promoter on alkali metal promoted 5 % wt Fe/active carbon catalysts at 550°C 130

CONCLUSIONS:

Alkaline promoted active carbon supported iron catalysts are very active in the oxidative dehydrogenation of ethylbenzene to styrene in the presence of carbon dioxide. The best results were obtained at 550°C for a Li-promoted catalyst with a conversion of ethylbenzene of 75% and a selectivity towards styrene of nearly 95%. These results are better than those obtained with industrial catalysts which perform the dehydrogenation process with an excess of water. The main product of the dehydrogenation reaction with COz was styrene, but the following by-products were detected - benzene and toluene. The selectivity towards toluene was always higher than towards benzene. We observed also the formation of carbon monoxide and water, which were produced with a constant molar ratio of about 0.8. The weight of the catalysts increased up to 20% during the reaction due to deposition of carbon. Using a too large excess of CO2 (COz/EB >10) was harmful for the styrene yield. The most favorable molar ratio of CO2 to EB was 10:1. No correlation between the molar ratios of reactants and the amount of deposited coke on the surface of catalysts was observed. The highest catalytic activity showed iron loaded D-90 catalysts which were promoted with alkali metals in a molar ratio of 1:10.

Iron, nickel and cobalt loaded carbonized PPAN, PC, inorganic supports like AI2O3, Si02/Zr02 or TiOz respectively and commercial iron catalysts applied for styrene production did not show comparable catalytic activity in similar conditions^

REFERENCES: [1] F. Cavani and F. Trifiro, Appl. Catal. A: General 133 (1995) 219 [2] J. Iwasawa, H. Nobe and S. Ogasawa; J. Catal. 31 (1973) 444 [3] T.G. Alkhazov, A.E. Lisovskii, M.G. Safarov and A.M. Dadasheva, Kinet. Katal. 13 (1972) 509 [4] Hirano, T.; Appl. Catal. 26(1986)65 [5] M. Sugino, H. Shimada, T. Turuda, H. Miura, N. Ikenaga and T. Suzuki, Appl. Catal. A: General 121 (1995) 125

ACKNOWLEDGMENT:

The work has been financially supported by the Volkswagen-Foundation of Germany. 131

DGMK-Conference "Selective Oxidations in Petrochemistry ”, Hamburg 1998

*DE01219821X* S. Albrecht*, K.-H. Hallmeier*, G. Lippold**, G. Wendt* /]u * Universitat Leipzig, Fakultat fQr Chemie und Mineralogie, Institut fur Technische Chemie und Wilhelm-Ostwald-lnstitut fiir Physikalische und Theoretische Chemie, Linnestr. 3, D-04103 Leipzig, Germany ** Universitat Leipzig, Fakultat fur Physik und Geowissenschaften, Institut fur Experimentalphysik I, Leipzig, Germany

V20s-Zr02 CATALYSTS FOR THE OXIDATIVE DEHYDROGENATION OF PROPANE - INFLUENCE OF THE NIOBIUM OXIDE DOPING

Introduction DE99G208 1

e oxidative dehydrogenation (ODH) of light alkanes is an alternative way for the production of olefins. A wide variety of catalytic systems has been investigated. Vanadium oxide based catalysts were described in the literature as effective catalysts for the ODH of propane. The catalytic activity and selectivity depend on the kind of support material, the kind of dopants and the formation of complex metal oxide phases. In recent papers it was claimed that both orthovanadate and/or pyrovanadate species are selective for the ODH of propane IV. Niobia based materials were investigated as catalysts for acidic and selective oxidation type reactions 12,3/. In the ODH of propane niobia exhibited a high selectivity to propene but the conversion of propane was low. V20s-Nb20s catalysts proved to be catalytically active and selective and showed no formation of oxygenates /4-10/. In the present study the influence of the niobia dopant on the catalytic properties of V205-Zr02 catalysts in the ODH of propane was examined. The structural and textural properties of the catalysts were investigated using several methods. )

Experimental

Preparation of the catalysts. The support material Zr02 [S(BET) = 86 m2/g] was prepared by precipitation from an aqueous solution of ZrO(NO), and a NH3 solution. The dried precipitate was calcined at 773 K for 6 h in air. V20s-Zr02 catalysts and V2Os- Nb205-Zr02 catalysts were obtained by wet impregnation of Zr02 with an aqueous

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 132

solution containing the appropriate amounts of ammonium metavanadate and vanadium oxalate/niobium oxalate, respectively. The catalyst precursors were dried and calcined at 773 and 873 K in air for 6 h. Catalytic measurements. The ODH of propane was performed in a computercontrolled fixed bed flow reactor at reaction temperatures between 673 and 773 K. A reaction gas mixture containing propane, air and nitrogen in a volume ratio of 1: 9:10 was applied. The residence time was varied between 0.009 to 0.045 g s/ml. No deactivation phenomena were observed within a reaction time of 30 h. The.reported catalytic activities are steady-state data. Characterization of the catalysts. The following methods were used: X-ray diffraction (XRD), textural measurements by nitrogen adsorption-desorption, temperature programmed reduction (TPR), Laser Raman spectroscopy (LRS) and X-ray absorption spectroscopy (XANES). The details of the measurements are discussed elsewhere 711,12/.

Results and Discussion

The main reaction products in the ODH of propane on V205-Zr02 catalysts were propene, CO and C02. Only traces of oxygenates and ethene were detected. A general view about the propane conversion in dependence on the catalyst composition and the residence time is shown in Fig. 1. A comparison between the different catalysts reveals the highest propane conversion for catalysts with a vanadia content of 3 and 5 mol%. Vanadia and zirconia were catalytically inactive under the chosen reaction conditions. The specific conversion rate (mol propane converted per hour times surface area) increases up to a vanadia content of 10 mol%. At higher vanadia contents a nearly constant conversion level was observed. The selectivity to propene was measured in the range of 50 and 90 % depending on the conversion degree. Comparing the catalytic behaviour of V2Os-Nb2Os catalysts (at low vanadia contents) in the ODH of propane with those of V205-Zr02 it is evident that the catalytic activity of the former catalyst system is lower because of their low surface areas /11/. 133

a)

residence time / g-s/ml

Zr02 c-1 c<3 c=5 c=10 c=20 c=40 V205 V2O5 content / mol%

b)

residence time / trs/ml

ZrQ2 c=1 e=3 c=5 c=10 c=20 c=40 V205

V2O5 content/mol% Figure 1: Propane conversion (a) and specific conversion rate (b) as a function of the V2Os content and the residence time for V20rZr02 catalysts, reaction temperature: 698 K

Undoped niobia catalysts have a low catalytic activity. The niobia dopant influences the catalytic activity of V205-Zr02 catalysts in different ways. In the case of the catalyst system with 3 mol% V205 content a decrease in the catalytic activity was observed with increasing NbN atomic ratio. A maximum of the catalytic activity was found for the catalyst system with 5 mol% V205 content at NbAZ atomic ratios between 1 and 3. Because of textural effects (see below) the specific activities of the latter catalysts do not change between NbAZ ratios of 0 and 1. At higher niobia contents the specific activities decrease (Fig. 2). 134

a)

b)

residence time / If gs/ml 0.045 0.015 0.009 0.03 0.5 1 3 Nb/V atomic ratio

Figure 2: Propane conversion (a) and specific conversion rate (b) as a function of the Nb/V atomic ratio and the residence time, 5 mol% V205-Nb205-Zr02 catalysts, reaction temperature: 698 K

In the case of V205-Zr02 catalysts a decrease in the surface area was observed with increasing vanadia content (Table 1). Niobia does not influence the values of the surface area of samples with low vanadia contents significantly. A remarkable enhancement of the specific surface area was observed for the series with 5 mol% vanadia content at Nb/V atomic ratios between 1 and 10.

Table 1: Specific surface areas of V205-Nb205-Zr02 catalysts in dependence on vanadia content (cfmol%1) and Nb/V atomic ratio, rasp. Nb/V atomic ratio specific surface area / m2/g c= 1 c = 3 c= 5 Nb/V = 0 79 74 21 Nb/V = 0.03 72 66 17 Nb/V = 0.5 74 52 24 Nb/V = 1 72 57 30 Nb/V = 3 - 62 40 Nb/V= 10 64 - 38 135

The Laser Raman spectra were recorded to assign the vanadate and niobate species formed by the interaction of the catalyst components. Crystalline V205 shows a band at 992 cm* 1 due to the V=0 stretching mode. For the V205-Zr02 catalysts with low vanadia contents a V=0 stretching mode at 1028 cm* 1 for monomeric and at 982 and 778 cm* 1 for polymeric vanadate species were found. Additionally, the increase in vanadia

content led to the formation of crystalline V205 711,12/. Crystalline niobia is characterized by bands at 990, 896, 842 and 756 cm'1 which are assigned to Nb=0 stretching modes in distorted octahedral structures Z13Z. Thus, the mode assignments for the V20s-Nb205-Zr02 samples was difficult because of the superimposition of the V=0 and Nb=0 stretching modes. Fig. 3 shows the Laser Raman spectra of the V20s-Nb205-Zr02samples with 5 mol% vanadia content in dependence on the Nb/V atomic ratio. For a comparison the spectra of vanadia and niobia are presented too. In the spectra of the samples with Nb/V of 0, 0.03 and 0.05 the V=0 stretching modes for the monomeric (1020 cm* 1) and the polymeric (984-986,779 cm* 1) vanadate species were detected. For the sample with a Nb/V atomic ratio of 3 the intensity of the band for the monomeric vanadate species is en­

Nb O 842 756 hanced and a new mode at 965 cm'1 arises which can be attributed to the formation of V-O-Nb species. In the V2Os-Nb2Os catalyst system a mode at 965 cm'1 was found which could be caused by the formation of the B- Nb/V = 3 (Nb,V)205 phase Z11Z. XANES investi­ gations on these samples showed the fourfold coordination for the vanadium ions 714/. 1200 1100 1000 900 800 The reduction behaviour of the wavenumber / cm-1 catalysts was investigated using TPR. Figure 3: Laser Raman spectra of 5 mol% Preliminary studies indicated that the V205-Nb205-Zr02 catalysts in dependence on the Nb/V atomic ratio TPR results are not effected by heat 136

and mass transfer 7157. Crystalline Vanadia was reduced in four steps as follows (Fig. 4): V205 V60i, - V02 - VnO^ -

The reduction of niobia started at temperatures above 900 K whereas zirconia adsorbed hydrogen in the same temperature range 711/. The reduction profiles of V205-Zr02 catalysts with low vanadia contents are characterized by one peak at considerably lower temperatures compared with crystalline vanadia. At temperature 7 K vanadia contents & 5 mol% the peak Figure 4: TPR profiles of 3 mol% VjCVNbjOs-ZrOj number increased and the peak catalysts in dependence on the Nb/V atomic ratio shifted to higher tern- peratures because of the formation of crystalline vanadia. The reduction degrees (formally calculated for the reduction of Vs* to V3*) reveal that a certain part of V4t ions are present A direct evidence for the V4* ions was shown by XPS 711/ and ERR 7127. The TPR profiles of the V20s-Nb205-Zr02 samples are documented in Fig. 4. It is recognized that the samples are reduced at considerably lower temperatures compared with crystalline vanadia and niobia. With increasing Nb/V atomic ratio a shift of the peak maxima to higher temperatures was observed for the samples with a vanadia content of 3 mol% and less significant for the series with a vanadia content of 5 mol%. With respect to the investigations on the V2Os-Nb2Os catalyst system it is concluded that the decrease in the reducibility of the vanadium oxide component is caused by vanadia - niobia interactions on the zirconia support.

Conclusions

The interaction of vanadia with zirconia and vanadia with niobia influences the catalytic and the structural properties of the investigated catalyst system. By spectroscopic measurements (LRS, XANES) it was established that in dependence 137

on the vanadia content V20s-Zr02 catalysts contain monomeric vanadate species with a single terminal V=0 bond and bridging V-O-Zr bonds as well as polymeric vanadate species with V=0, V-O-V and V-O-Zr bonds; crystalline V205was found at vanadia contents a 5 mol%. In the case of the vanadia-niobia catalyst system monomeric and polymeric vanadate species were detected. The formation of B-(Nb,V)205 was established at vanadia contents % 5 mol% by XRD. In comparison to crystalline vanadia the reducibility of the supported vanadate species was increased shown by TPR measurements. A maximum of the catalytic activity in the ODH of propane was obtained for V20s-Zr02 catalysts with vanadia contents between 3 and 5 moI%; in the range where monomeric and polymeric vanadate species were observed. Doping of V20s-Zr02 catalysts with niobia at low vanadia contents led to an increase in the amount of monomeric vanadate species. However, the reducibility of the vanadate species decreased. It is assumed that the strong interaction between niobia and vanadia influences the interaction between vanadia and zirconia. Niobium ions are incorporated in the vanadia surface species under formation of V-O-Nb species. In general, doping of V205-Zr02 catalysts with niobia caused a decrease in the catalytic activity in the ODH of propane for catalysts with a vanadia content of 3 mol%. A textural effect was observed for the catalyst series with 5 mol% V205 content It was found that the specific surface area increased by doping of V205-Zr02 with niobia. A maximum of the catalytic activity was obtained for catalysts with a Nb/V atomic ratio between 1 and 3. In conclusion, it is suggested that oxygen in the V-O-Zr, V-O-V and V-O-Nb configura ­ tion in the surface polymeric vanadate species is involved in the reaction of ODH of propane. The metal-oxygen bond strength influences the catalytic activity and selec­ tivity.

The financial support by the "Deutsche Forschungsgemeinschaft", by the "Graduiertenkolleg der Physikalischen Chemie der Grenzflachen" in Leipzig and by the "Ponds der Chemischen Industrie" is gratefully acknowledged.

References

71/ EA Mamedov, V. Cortes Corberan; Appl. Catal. A General, 1271 (1995) 121 K. Tanabe, S. Okazaki; Appl. Catal. A General, 133 191 (1995) 131 J.C. Vedrine, G. Coudurier, A Ouqour, P.G. Pries de Oliveira, J.C. Volta; Catal. Today 28 3 (1996) 138

141 R.H.H. Smits, K. Seshan, J.R.H. Ross; J. Chem. Soc., Chem. Common. 8 558 (1991) 151 J.R.H. Ross, R.H.H. Smits, K. Seshan; Catal. Today 16 503 (1993) 161 R.H.H. Smits, K. Seshan, J.R.H. Ross; Stud. Surf. Sci. Catal. 72 221 (1992) m R.H.H. Smits, K. Seshan, J.R.H. Ross, L.C.A. van den Oetelaar, J.H.J.M. Hellwegen, M.R. Anantharaman, H.H. Brongersma; J. Catal. 157 584 (1995) 181 R.H.H. Smits, K. Seshan, J.R.H. Ross, A.P.M. Kentgens; J. Phys. Chem. 99 9169 (1995) 19/ K. Seshan, R.H.H. Smits, J.R.H. Ross; Proceedings DGMK-Conference Selective Hydrogenation and Dehydrogenation, Kassel 1993, Tagungsbericht 9305, p. 211 /10/ T.C. Watling, G.Deo, KSeshan, I.E. Wachs, JA Lercher; Catal. Today 28139 (1996) 711/ S. Albrecht; Dissertation Universitat Leipzig, 1998 712/ S. Albrecht, G. Wendt, G. Lippold, A Adamski, K. Dyrek; Solid State Ionics 101-103 909(1997) 713/ J.-M. Jehng, I.E. Wachs; Chem. Mater. 3 100 (1991) 714/ K.-H. Hallmeier, S. Albrecht, D. Mayer, G. Wendt, R. Szargan; HASYLAB, Jahresbericht 1997, p. 827 715/ S. Albrecht; Diplomarbeit Universitat Leipzig, 1994 716/ H. Bosch, B.J. Kip, J.G. van Ommen, P.J. Ceilings; J. Chem. Soc., Faraday Trans. I 80 2479(1984) 139

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

O. Buyevskaya, M. Baems Institute for Applied Chemistry, Berlin-Adlershof, Rudower Chaussee 5, D-12484 Berlin, Germany

OXIDATIVE DEHYDROGENATION OF ETHANE ON RARE-EARTH OXIDE- BASED CATALYSTS

*DE01f 2198229* Abstract ‘Results on the oxidative dehydrogenation of ethane on rare-earth oxide (REO) based catalysts (Na-P-Sm-O, Sm-Sr(Ca)-0, La-Sr-0 and Nd-Sr-O) are described. Oxygen adsorption was found to be a key factor which determines the activity of this type of catalysts. Continuous flow experiments in the presence of catalysts which reveal strong oxygen adsorption showed that the reaction mixture is ignited resulting in an enhanced heat generation at the reactor inlet. The heat produced by the oxidative reactions was sufficient under the conditions chosen for the endothermic thermal pyrolysis which takes place preferentially in the gas phase. Ignition of the reaction mixture is an important catalyst function. Contrary to non-catalytie oxidative dehydrogenation, reaction temperatures above 700°C could be achieved without significant external heat input Ethylene yields of up to 34 - 45 % (S = 66 - 73 %) were obtained on REO-based catalysts under non-isothermal conditions (T™* = 810 - 865°C) at contact times in the order of 30 to 40

Introduction

Oxidative dehydrogenation of ethane being an alternative for the highly endothermic thermal pyrolysis has been the subject of many studies in which different catalysts were tested in the temperature range from 350 to 850°C [1 - 9], ForT < 600°C, high ethylene selectivities of 89 and 91 % at ethane conversions of 19 and 15 % were reported on B2O3(30wt%)/AI2O3 at 550eC and on K2Pi2MoioWiSb l-Fe1Cro^Ceo. 750„ at 470°C, respectively [1, 2]. Mo-V-Nb, Mo-V-Sb and Mo-V-Ta oxide systems showed high activity at low temperatures (350 - 400eC); an ethylene yield of 38 % (S = 65 %) was obtained on Mo 0.73Vo. 1eNb0.09 Ox catalyst at 350°C [3]. High selectivities and yields to ethylene can be achieved when chlorine-containing compounds are fed into the reactor or when the catalyst is doped with halides [4-6]. Conway et al. [5] obtained an ethylene yield of 57 % (S = 76 %) over a Dy 203/Li*-Mg 0-CI" catalyst at 570°C after 25 h on stream. To obtain high ethylene yields on catalysts consisting of rare-earth metal oxides without doping with halides reaction temperatures mostly

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 DE99G2080 140 above 600°C were applied [7-9]. An ethylene yield of approx. 49 % (S = 70 %) was obtained by Velle et al. [7] on SrCeo.sYbo. 5O2.75 at 700°C. A lithium-doped lanthanum- calcium oxide (Li:La:Ca = 1:1:2.5) was described as an effective catalyst resulting in C2H4 yields of 48.2 % (S = 89.2 %) at 620°C and GHSV = 1000 h‘1 [8], The dehydro ­ genation of ethane at T = 650 - 850°C was studied by Choudhary et al. [9] using a diluted Sr0/La203 catalyst; ethane conversions of approx. 20 % (Sc2H4 = 75 %) and 68 % (Sc2H4 = 82 %) were obtained at 700 and 850°C, respectively. Although the authors proposed that the reaction is initiated on the catalyst no results on ethane oxidation without a catalyst under the same conditions are given. This is an impor ­ tant issue, especially concerning this particular study since high ethylene yields were achieved only at very high tempeature of 850°C at which non-catalytic reac­ tions might prevail. As shown by Burch and Crabb [10], a significant non-catalytic oxidative dehydrogenation occurs already at 600°C; the authors obtained an ethylene yield of 33.2 % (S = 73.7 %) in an empty reactor (0% = 5 mm). The present contribution refers to both applied and fundamental aspects of the oxidative dehydrogenation of ethane in the presence of rare-earth oxide (REO) based catalysts. Particular attention is given to the operation under non-isothermal conditions caused by ignition of the reaction in the presence of REO-based catalysts and to solid properties determining the catalyst activity and the ignition temperature.

Experimental

LnSr0jO, (Ln = La, Nd, Sm) catalysts were prepared by addition of a strontium hydroxide solution to an aqueous lanthanide-nitrate solution. The mixture was stirred and heated to evaporate the water. The dried sample was calcined in a flow of oxygen at 700°C for 2 h. SmNao.02eP0.014O, and SmNao.oo2ePo.ooi40, catalysts were prepared by impregnation of Sm203 with Na2HPC>4T2 H20 solution. After drying at 120°C the catalyst was calcined in air at 700°C for 4h. For preparation of a SmCa50, catalyst, a samarium nitrate solution was added to a suspension of Ca(OH)2 in water. The mixture was stirred and heated to evaporate the water; the catalyst precursor was calcined at 700°C in a flow of air for 2 h. Catalytic experiments were carried out using an U-form quartz reactor (0% = 6 mm - catalyst (quartz) zone; 0^ = 2 mm - outlet tube) operated at ambient pressure. A quartz capillary (0«,= 3 mm) containing a movable thermocouple for measuring the temperature profiles was placed into the center of the catalytic layer. The catalyst (m,* = 0.2 or 0.5 g, dp = 250 - 355pm) was packed between two layers of quartz particles of the same size. The oven was filled with sand in which the reactor was immersed. After preheating the catalyst to 300°C in air the reaction mixture consisting of C2H6, 02 and N2 was continuously passed over the catalyst at a total flow rate of 330 ml/min; the reactor was heated to the desired temperature. Partial 141

pressure of ethane was 45.5 (55.5) kPa, the ethane-to-oxygen ratios were varied from 2.5 to 3.75. The products were analyzed by gas chromatography. Oxygen adsorption on Sm203 and Na-P/Sm203 surfaces was studied by means of 02 pulse experiments at 550°C in vacuo (P = 10-4 Pa) in a TAP reactor [11]. The cata­ lyst (me* = 0.3 g) was first exposed to oxygen pulses at reaction temperature and kept under vacuum for 10 min before the pulse experiments. Surface areas were measured by the BET method applying nitrogen adsorption at liquid nitrogen temperature. Surface compositions of Na-P/Sm203 samples were obtained from XPS spectra recorded with an ESCALAB 220i-XL spectrometer (Fisons Instruments) using AIKa radiation (1486.6 eV).

Results and Discussion 1. Continuous flow experiments Blank experiments performed at 700eC in the reactor filled with quartz (dp = 250 - 355 pm) using a total flow rate of 330 ml/min, p(C 2H6) = 45.5 kPa and C2Hs/02 = 2.5 resulted in an ethane conversion of only 2.9 % (SC2H4= 99 %). On heating the reactor to 796°C, ethane conversion increased to 67.7 % (Sc2m= 66 %, Scm = 6.9 %, So* = 4.8 %) resulting in an ethylene yield of 44.7 %; oxygen conversion amounted to 89.5 %. No ignition was observed. The temperature in the sand bath (803°C) was slightly higher than that in the reactor. The total flow rate used corresponds to a superficial velocity u * 200 cm/s at 800 eC in the quartz bed. Thus, the non-catalytic oxidative dehydrogention occurs effectively at T s 800°C even at high superficial velocity. On Sm203-, La203- and Nd203- based catalysts, the oxidative dehydrogenation of ethane was studied under the same reaction conditions as in the experiments with quartz. When a catalyst was used, the ignition of the reaction mixture took place above a certain inlet temperature. Depending on the catalyst and the reaction mix­ ture applied the reaction was ignited between 440 and 710 °C. After ignition, the temperature in the catalyst bed rose rapidly within a few minutes. Temperature increased mostly to well above 700 °C being the border line between catalytic and homogeneous gas phase oxidation. The temperature can be then stabilized at the desired level (TR) without extinction of the reaction by adjusting the temperature in the sand bed which, in turn, determined the preheating of the gas mixture (T0); it is emphasized that the quantitative data on the temperature rise depends on heat exchange between the reactor and the sand bath, i.e., this is a design-depending phenomenon. A characteristic temperature profile in the reactor after ignition is shown in Figure 1. The ignition temperature was strongly affected by catalyst composition. For SmSrosO* SmCa5Ox and SmNao.o2sPo.ouO x, ignition temperatures amounted to 450, 590 and 710 °C, respectively. 142

L /mm

Figure 1. Temperature profile in the reactor after ignition of the reaction mixture on

a SmSfoO, catalyst (p(C 2H6) = 45.5 kPa, C2H e/02 = 2.5, x = 0.036 g s/ml).

Results on ethane conversion in the presence of various catalysts under non- isothermal conditions are summarized in Table 1. Ethylene, methane, CO, C02, C3 and C< hydrocarbons were detected as products. On LnSr0.2O, (Ln= Sm, Nd, La), SmCasOx and SmNa0.oo 28 Po.ooi 40x catalysts, ethylene yields of 28.0 - 34.1 % (S = 63 - 67 %) were obtained at Tr = 750 - 865 eC and a short contact time of 0.036 g s ml* 1. When using a SmNao.o 28 Po.ouOx catalyst showing the high ignition temperature the yield of ethylene amounted to 45.8 % (S = 67 %) at TR = 867°C. Although high ethylene yields were obtained at 800°C in the absence of any catalyst due to non-catalytic oxidation, it was not possible to ignite the reaction in this case. From the results obtained it follows that an important catalyst function is the ignition of the reaction. Hereby, reaction temperatures of 600 - 870 eC can be achieved without significant external heat input. The increased ethylene yields are partly obtained via thermal dehydrogenation due to heat produced by catalytic oxidation which would correspond to autothermal reactor operation. The C2Hs-to-0 2 ratio influenced the ethylene yield and selectivity as well as the difference between TR and T». The respective dependencies for a Sm2Oa/SrO catalyst are shown in Figure 2. 143

Table 1. Ethane and oxygen conversions, product selectivites, ethylene yield as well as preheating and maximal temperatures for different catalysts

Catalyst Sbet T T„ Tmix. Conv. / % Selectivity Z % Yc2H4

m2/g gsmf 1 eC °c C2H8 02 C2H4 CH, C3+ % SmSr0^Ox 8.0 0.036 380 788 43.0 100 68.9 5.6 3.4 29.6 560 855 51.9 100 65.7 6.4 2.6 34.1

NdSr02Ox 7.2 0.036 276 621 38.8 100 59.5 3.7 4.6 23.1 529 800 45.6 100 63.0 6.7 2.9 28.7

LaSro^Ox 4.1 0.09 262 600 34.2 99.5 56.0 2.5 5.0 19.2

SmCa5Ox 1.6 0.036 548 829 45.4 99.2 66.8 3.7 2.4 30.3 624 865 49.7 100 66.7 5.3 2.2 33.2

SmNao.oo28Po.ooi40x 4.6 0.036 330 750 42.5 99.3 65.9 3.3 2.3 28.0

SmNao. 02ePo. 014O, 3.3 0.036 693 867 67.0 98.9 68.3 11.1 2.9 45.8

Quartz 803 796 67.7 89.5 66.0 6.9 4.8 44.7 reaction conditions: p(C 2H6) = 45.5 kPa, C2H&/02 = 2.5, total flow rate = 330 ml/min

S(C h )

■ 600

S(CO,)

Figure 2. Dependence of ethane conversion, product selectivities and preheating temperature (T„) on the C2Hs-to-0 2 ratio on SmSr02Ox catalyst (TR = 850 eC, p(C 2Hs) = 55.5 kPa, nw = 0.2 g, x = 0.044 g s ml"1). 144

The effect of the catalyst-bed height on the process parameters was studied using 0.2 g of the SmSr0.2Ox catalyst at TR = 750 and 850°C. Results are given in Table 2.

Table 2. Results on ethane dehydrogenation on a SmSr0^Ox catalyst with different height of the catalyst bed.

Bed height 1 mm* 7 mm TR/eC 750 850 750 850 X(C2H s) / % 40.5 54.9 40.4 52.6 S(C2H„)/% 61.9 68.2 60.8 65.4 nu= 0.2 g; p(C 2H6) = 45.5 kPa, CzHe/Oz = 2.5, Ft*. = 330 ml/min; * reactior with i.d. = 12 mm

No significant difference in the ignition temperature, ethane conversion and ethylene selectivity was observed in the experiments with bed heights of 1 and 7 mm. Thus, a layer of 1 mm was sufficient to ignite the reaction mixture; further reactions occur most probably in the gas phase. For better utilization of the heat produced by the oxidative ethane conversion in the non-catalytic zone, a fluidized bed of the inert material following the fixed bed of catalyst was used. This allowed sufficient heat transfer from the ignition zone (catalyst layer) to the post-catalytic zone in which the thermal dehydrogenation occured. Ethylene yields of 43 -45 % (S = 73 - 76 %) were obtained. Compared to the reactor in which fixed beds of both the catalyst and inert material were used, the latter reactor concept allowed the maintenance of the gradientless temperature level of 700 to 800°C required for the thermal dehydrogenation within the whole post-catalytic zone. Oxidative dehydrogenation of ethane in the presence of oxygen over Ft- and Rh- coated ceramic foam monoliths in an autothermal reactor at contact times of the order of milliseconds was reported by Schmidt et at. [12-13], A C2H4 yield of 52.3 % (S = 69 %) was obtained over a Pt-Sn (7:1) catalyst at 920°C using C2H b/02 = 1.9 and a feed flow rate of 5 standard liters per minute [13]. The authors believe that heterogeneous reactions dominate under these conditions. As follows from present experiments, the oxidative dehydrogenation of ethane on REO-based catalysts at contact times in the millisecond range implies the ignition of the reaction on the catalyst; due to the heat generated the part of ethane is thermally dehydrogenated.

2. Transient studies As shown in our previous works [14, 15], the nature of the oxygen species involved in alkane activation is an important parameter influencing the choice of the optimal mode of operation in the oxidative dehydrogenation of low alkanes. 145

Oxygen adsorption on Sm2Orbased catalysts was investigated by means of pulse experiments in vacuum (approx. 10"* Pa) applying the Temporal-Analysis-of-Prod- ucts (TAP) reactor [11]. Evaluation of kinetic data from the TAP experiments was performed using a reactor model and software described elsewhere [16, 17]. The following reaction model implying molecular and subsequent dissociative 02 adsorp ­ tion was used for determining the rate constants: (1) 02 + [] - [02] (2) [0*1+ [] — 2[0] The comparison between experimental and calculated oxygen response on Oz pul­ sing over SmNao.oo28Po.ooi40x is illustrated in Figure 3; the adequate description of the experimental signal was obtained using the above model. The estimated rate con ­ stants together with the results on surface composition are given in Table 3.

experiment model

time / s Figure 3. Comparison between simulated and measured oxygen responses at the reactor outlet obtained on pulsing 02 over SmNao.oo 26Po.ooMO, at 550eC.

Table 3. Results on surface composition and kinetics of oxygen adsorption Catalyst Sbet Surface composition Oxygen adsorption at (XPS) 550°C (TAP) m2/g Na/Sm P/Sm kad.(t) / S'* Sm203 5.6 - - 710* SmNao.oo26Po.ooMO, 4.6 0.21 0.02 2102 SmNao.o2ePo.oMO, 3.3 2.37 0.12 no adsorption

Sodium-doped samaria showed different ignition temperatures depending on the amount of dopant. Na/Sm atomic ratios in the surface layer derived from XPS for Na- SmNa0.002eP0.00MO, and SmNao.o 28 Po.oMO, samples amounted to 0.21 and 2.37, 146

respectively (cf. Table 3). For the latter catalyst with the surface enrichment of sodium, the ignition of the reaction occurred only at high temperature of 710°C. Results of the oxygen pulse experiments showed that doping with small amount of Na-phosphate resulted in an increase of the rate constant of oxygen adsorption compared to undoped Sm203 (cf. Table 3). No oxygen adsorption was detected on the SmNao.o 28 Po.ouO x catalyst with a high content of Na-phosphate at 550°C. Thus, oxygen adsorption, determining the catalyst activity and ignition temperature on REO-based catalysts is significantly suppressed by high alkali doping.

Conclusions

REO-based catalysts revealing oxygen adsorption are very active towards ethane conversion. Due to high activity ignition of the reaction mixture takes place resulting in non-isothermal operation. Contrary to non-catalytic oxidative dehydrogenation, reaction temperatures above 700°C can be achieved without significant external heat input The addition of strontium or small amounts of alkali metals to rare earth oxides resulted in low ignition temperatures of 440 - 485°C. A catalyst layer of 1 mm was sufficient to ignite the reaction mixture and to sustain the reactions using preheating the reaction mixture. Ethylene yields of up to 45 % were obtained under non-isothermal conditions at contact times of the order of 30 - 40 ms.

Literature 1. G. Colorio, J.C. Vddrine, A. Auroux, B. Bonnetot, Appl. Catal. A137 (1996) 55. 2. F. Cavani, M. Koutyrev, F. Trifiro, Catal.Today 28 (1996) 319. 3. E.M. Thorsteinson, T.P. Wilson, F.G. Young, P.H. Kasai, J. Catal. 52 (1978) 116. 4. R. Burch and S.C. Tsang, Appl. Catal. 65 (1990) 259. 5. S.J. Conway, D.J. Wang and J. H. Lunsford, Appl. Catal. 79 (1991) L1. 6. W. Ueda, S.-W. Lin and I. Tohmoto, Catal. Lett. 44 (1997) 241. 7. O.J. Velle, A. Andersen and K.-J. Jens, Catal. Today 6 (1990) 567. 8. L. Ji and J. Liu, Chem. Commun. (1996) 1203. 9. V.R. Choudhary, B.S. Uphade and S.AR. Mulla, Angew. Chem. Int. Ed. Engl. 34 (1996)665. 10. R. Burch, E.M. Crabb, Appl. Catal. 97 (1993) 49. 11. Y. Schuurman and J.T. Cleaves, Ind. Eng. Chem. Res. 33 (1994) 2935. 12. M. Huff and L.D. Schmidt, J. Phys. Chem. 97 (1993) 11815. 13. C. Yokoyama, S.S. Bharadwaj and L.D. Schmidt, Catal. Lett. 38 (1996) 181. 14. O.V. Buyevskaya and M. Baems, Catal. Today 42 (1998) 315. 15. M. Baems and O.V. Buyevskaya, Proceed.Third Japan-EU Joint Workshop (JECAT97), November 25-28 (1997) 21. 16. M. Rothaemel and M. Baems, Ind. Eng. Chem. Res., 35 (1996) 1556. 17. M. Soick, O. Buyevskaya, M. Hdhenberger and D. Wolf, Catal. Today, 32 (1996) 163. 147

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*DE012198238* M. Dusi, T. Mallat, A. Baiker C? Laboratory of Technical Chemistry, Swiss Federal Institute of Technology, ETH-Zentrum, CH-8092 Zurich, Switzerland

SELECTIVE EPOXIDATION OF ALLYLIC ALCOHOLS WITH A TITANIA-SILICA AEROGEL

ABSTRACT

.n amorphous mesoporous titania-silica aerogel (20 wt% - 80 wt% SiO^ and

ferf.-butylhydroperoxide (TBHP) have been used for the epoxidation of various allylic alco ­ hols. Allylic alcohols possessing an internal double bond were more reactive than those with a terminal C=C bond. Epoxide selectivities could be improved by addition of (basic) zeolite 4A

and NaHC03 to the reaction mixture. 1

INTRODUCTION

Epoxides are desirable as useful intermediates as well as target molecules. In the past years, considerable effort has been spent on substituting homogeneous transition metal cata­ lysts by solids [1]. The development of an amorphous silica-supported titania by Shell researchers [2] represented a major breakthrough in the heterogeneously catalyzed epoxida- tions. This catalyst has found industrial application in propylene epoxidation, and has been shown to be active and selective also in the epoxidation of allyl alcohol and cyclohexenol. The idea of preparing well dispersed titania-on-silica by reacting a Ti-precursor with the surface silanol groups has been applied also by some other groups: silica gel treated with

Ti(0'Pr)4 [3] and an amorphous TlO^-SiO; [4], prepared by hydrolysis of tetraethoxytitanium and tetraethoxysilicon, were used for the epoxidation of allylic alcohols. However, the activity and selectivity of these catalysts were rather moderate, and oxidation of the allylic alcohols to the corresponding unsaturated ketones was an important side reaction. There is a wealth of data available now which demonstrate that the nanoscale engineer ­ ing of titania-silica xerogels and aerogels can provide outstanding epoxidation catalysts [5-9]. Even bulky cyclic olefins and a-isophorone, possessing an electron deficient C=C double

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 148

bond, could be epoxidized with high rate and selectivity using alkylhydroperoxides and an aerogel containing 20 wt% Ti02 [10,11], The same catalyst afforded 87 % yield in the epoxi- dation of Zrans-2-hexen-l-ol [12]. The reaction was carried out with TBHP and the in situ dried aerogel. Water, formed in small amount during the reaction, was eliminated by some zeolite 4A additive. It has been shown that the activity of titania-silica aerogel in the epoxidation of cyclic olefins is 2-3 orders of magnitude higher than those of Ti-|3 and Ti-MCM-41 [9]. The aim of the present work was to extend the application range of our titania-silica aerogel and study the epoxidation of various allylic alcohols possessing a terminal or an inter­ nal double bond. The general scheme of the main and side reactions is depicted below, in Sbjieme 1.

R1 R3 R2 -R2 XiA>c R1 R3 Titania-silica 3 R2 OH R*C^oh 1 TBHP % R2 OH ------R2 OH 2 acid / base 4 acid / base 1-4 dimers, oligomers R1, R2, R3 = H, alkyl, cycloalkyl

Scheme 1

EXPERIMENTAL

The titania-silica aerogel containing 20wt% Ti02 was prepared using the sol-gel method [8]. The solvent was semicontinuously extracted with supercritical C02 at 313 K and

24 MPa (low-temperature aerogel). The raw aerogel was calcined in a flow of dry air at 673 K for 5 h. In the standard epoxidation procedure, 70 mg catalyst was heated in an argon flow to

473 K for 16 h. To the in situ dried catalyst were then added 20 mmol olefin, 0.5 ml ethylben ­ zene (internal standard) and toluene (solvent). The reaction was started by the addition of 5 mmol TBHP as a 5.5 M solution in nonane at the desired temperature. The total reaction vol ­ ume was 6 ml. Depending on the method, zeolite 4A and / or NaHC03 were used as additives.

The reaction products were analyzed by GC. 149

RESULTS

Oxidation of primary allylic alcohols

In a previous study the oxidation of trans- 2-hexen-l-oI was achieved with high reaction rate and epoxide selectivity (Table 1, entry 1) [12]. The dominant side reaction was the acid- or base-catalyzed ring opening of the epoxide by the reactant alcohol to give a hydroxy-ether (8 %), whereas oxidation of the -OH function to a carbonyl group was minor (1 %) (Scheme 1). Some other alkyl substituted, open chain allylic alcohols have been epoxidized under the same conditions. The important conversion and selectivity data are collected in Table 1. High

Table 1: Epoxidation of primary allylic alcohols*

Reactant Additive(s) T Olefin Epoxide selectivity (g) (K) conversion b(%) (%) related to

Olefin TBHP

lc 4A(1.5g) 363 96 91 91

2 , 4A (0.5 g) 363 82 80 75

3 4A (0.5 g) 363 70 97 86

4 4A (0.5 g) 363 109 34 43

5= 4A (1.5 g) 363 97 86 n.d. NaHC03 (0.34)

6° 4A(1.5g) 333 32 94 n.d. NaHC03 (0.34) a. Standard reaction conditions; reaction time: 1 h. b. Olefin conversion is related to TBHP (olefin : TBHP = 4:1) c. Reaction conditions: 200 mg catalyst (dried in situ at 473 K), 60 mmol olefin, 13.5 mmol TBHP, 1 ml cumene (internal standard), toluene (solvent); total volume: 21 ml; reaction time: lh.

conversions were achieved in 1 h at 363 K. The prenol conversion over 100 % (entry 4) was due to considerable oligomerization of the olefin. Seemingly, the conditions applied were good for the epoxidation of hexenols and pentenol, but not suitable for prenol possessing a trisubsti- 150

luted double bond. In the latter case the epoxide was only a minor product (34 %), and hydroxy-ether (13 %) and unsaturated aldehyde (5 %) were also formed. Lower reaction rates (based on conversion) were achieved for cis-allylic alcohols (entries 2 and 3) as compared to trans-hexenol. In case of cis-pentenol (entry 3) epoxide ring opening was a minor reaction (ca. 3 %), and no oxidation of the -OH function to a carbonyl group was observed. It is noteworthy that the cis- and trans-isomers of the reactants retained their stereospecificity in the products.

Influence of reaction additives on the epoxidation ofprenol The role of reaction additives is illustrated by the oxidation of prenol. The remarkable effect of NaHC03 additive (besides zeolite 4A) is shown in Table 1 (entries 4-6). NaHC03 suppressed oligomerization and increased the epoxide selectivity by a factor of 2.5, under oth ­ erwise identical conditions. A similar striking effect of various basic additives on the product composition has been observed recently in the epoxidation of P-isophorone [13]. The kinetics of the prenol oxidation reaction, using the standard procedure and either zeolite 4A or 4A and NaHC03 together as additives, is shown in Fig. 1. In presence of zeolite 4A only, the reaction

14 12 10 8 6 4 2 KXXX X 0

Time (min) Time (min)

Fig. 1: Time resolved product formation in the oxidation of prenol using 1.5 g zeolite 4A (a) or 1.5 g zeolite 4A and 4 mmol NaHCOj (b) as additives. Reaction conditions: 0.2 g aerogel (driedin situ at 473 K), 60 mmol prenol, 13.5 mmol TBHP, 1 ml cumene (internal standard), toluene (solvent), total volume: 21 ml; T=363 K. (2) Epoxide; (3) unsaturated aldehyde; (4) hydroxy-ether. was very fast at low conversion, but after ca. 30 min the decomposition of the epoxide (mainly by hydroxy-ether formation) became faster than its formation (Fig. la). The influence of reac- 151

lion additives on the rate and selectivity can be interpreted on the basis of our former detailed study of hexenol oxidation [12]. Zeolite 4A eliminates the water, which forms during oxida ­ tion by the oxidative dehydrogenation of the -OH group and by the decomposition of the per­ oxide [I] and maintains dry conditions, affording excellent performance of the aerogel in the oxidation of hexenols and pentenol. However, in the oxidation of prenol the Na-form of zeolite 4A seemingly acted as a catalyst of the ring opening and oligomerization side reactions (Scheme 1). In these instance the addition of some weakly basic NaHC03 (besides 4A) afforded higher epoxide selectivities and suppressed the hydroxy-ether formation substantialy (Fig. lb).

Diastereoselective epoxidation of open chain and cyclic allylic alcohols

The data in Table 2, entries 1-3, illustrate that the conversion of open chain alcohols possessing a terminal double bond required longer reaction times, as compared to allylic alco ­ hols with an internal double bond (Table 1) or cyclohexenol (Table 2, entry 4). The higher reac-

Table 2: Diastereoselective epoxidation of open chain and cyclic allylic alcohols 8

Reactant t Olefin conversion* 1 Epoxide selectivity (%) Diastereomeric (h) (%) related to ratio

Olefin TBHP

1 5 95 37 65 49:51 JCU erythro : three

2 5 60 83 77 45:55 erythro : threo

3 OH 5 74 71 67 _=

4 1 85 65 67 70:30 Q-°h cis: trans a. Standard reaction conditions; 0.5 g zeolite 4A; T=363 K. b. Olefin conversion is related to TBHP (olefin : TBHP=4:1) c. The diastereomers could not be separated by the gas-chromatographic method.

tivity of allylic alcohols possessing an internal C=C double bond is likely due to the higher 152

electron density resulting from the electron releasing effect of the alkyl group. For comparison, the same titania-silica catalyst epoxidized 2-hexene three times faster than 1-hexene, under otherwise identical conditions [14]. The highest epoxide selectivity was obtained with hexen-3-ol (entry 2), which com ­ pound was the least reactive. No diastereoselectivity (buten-3-ok 49 / 51) or only poor threo- selectivity (hexen-3-ol: 45 / 55) was observed. Note that in the metal-catalyzed epoxidation of acyclic ally lie alcohols possessing a terminal or a frame-substituted double bond, 1,2- and 1,3- allylic strains are weak and the otherwise subordinate steric and electronic factors lead to erythro-selectivity [15, 16]. The moderate /Area-selectivity in theTS-1 catalyzed epoxidation of this type of allylic alcohols has been attributed to a transition state structure similar to that which had been established for peracid epoxidations [17]. Cyclohexenol was much more reactive under the conditions applied, and a reasonable epoxide selectivity was achieved (Table 2, entry 4). The c/s-epoxide was produced preferen­ tially. The stereochemical outcome of the epoxidation of cyclic allylic alcohols with Ti-substi- tuted zeolites TS-1 and Ti-p has been explained by interaction of the -OH functional group with the active site or the epoxidizing agent in the transition state [17-19].

DISCUSSION

Allylic alcohols could only be efficiently epoxidized by the titania-silica aerogel when

the reaction conditions were properly chosen. A careful in situ predrying of the catalyst pro ­ vided high rate and selectivity in the oxidation of open chain, unbranched allylic alcohols. Removal of acidic sites by condensation of -Si-OH to Si-O-Si connections has been reported [13,20] and is supposed to play an important role during catalyst predrying. It has been shown for prenol that the presence of zeolite 4A and NaHC03 additives has a substantial influence on

the reaction rate and product distribution. The positive influence of solid bases is attributed to their minor solubility in the reaction medium and to the neutralization of the acidic sites on the aerogel. Prenol is an example of demanding epoxidation reactions. The formation of side prod ­ ucts could be minimized by the addition of NaHC03 and 86 % epoxide selectivity was

obtained at 97 % conversion. In general, the activity of the titania-silica aerogel is rather low in the oxidation of alcohols. Accordingly, the competing formation of carbonyl compounds from allylic alcohols is usually a minor side reaction, and the corresponding epoxy alcohols can be 153

synthesized with good selectivity. In contrast, TS-1 catalyzed preferentially the oxidation of the alcoholic function of prenol to form the unsaturated aldehyde [4], Due to the high activity of TS-1 in the oxidation of various functional groups, the selective formation of epoxy alcohols is limited to the oxidation of some unsaturated alcohols possessing a terminal C=C double bond [4,21].

ACKNOWLEDGEMENT Financial support of this work by Hoffmann-La Roche AG, Switzerland and the “Kom- mission fur Technologic und Innovation" (KTI) is gratefully acknowledged.

REFERENCES

[1] R.A. Sheldon in: Aspects of homogeneous catalysis; R. Ugo, Ed.; D. Reidel; Dordrecht; Vol. 4(1981). [2] Brit. Pat., 1'249 079 (1971), C.A. reference 74,12981M. [3] C. Cativiela, J.M. Fraile, J.I. Garcia and J.A. Mayoral, J. Mol. Catal. A: Chemical, 112, 259 (1996). [4] T. Tatsumi, M. Yako, M. Nakamura, Y. Yuhara and H. Tominaga, J. Mol. Catal. A: Chemical, 78, L41 (1993). [5] S. Klein, J.A. Martens, R. Parton, K. Vercruysse, P.A. Jacobs and W.F. Maier, Catal. Lett., 38,209 (1996). [6] S. Thorimbert, S. Klein and W.F. Maier, Tetrahedron, 51,3787 (1995).

[7] H. Kochkar and F. Figueras, J. Catal., 171,420 (1997). [8] D.C.M. Dutoit, M. Schneider and A. Baiker, J. Catal., 153,165 (1995). [9] R. Hutter, T. Mallat, D. Dutoit and A. Baiker, Top. Catal., 3,421 (1996). [10] R. Hutter, T. Mallat and A. Baiker, J. Catal., 153,177 (1995). [11] R. Hutter, T. Mallat and A. Baiker, J. Catal., 157,665 (1995). [12] M. Dusi, T. Mallat and A. Baiker, J. Catal., 173,423 (1998). [13] R. Hutter, T. Mallat, A. Peterhans and A. Baiker, J. Catal., 172,427 (1997).

[14] M. Dusi, T. Mallat and A. Baiker, unpublished data. [15] A.H. Hoveyda, D A. Evans and G.C. Fu, Chem. Rev., 93,1307 (1993). [16] L. Palombi, F. Bonadies and A. Scettri, Tetrahedron, 53,11369 (1997). [17] W. Adam, A. Corma, T.I. Reddy and M. Renz, J. Org. Chem., 62,3631 (1997). 154

[18] R. Kumar, G.C.G. Pais, B. Pandey and P. Kumar, J. Chem. Soc., Chem. Commun., 1315(1995). [19] W. Adam, R. Kumar, T.I. Reddy and M. Renz, Angew. Chem. Int. Ed. Engl., 35, 880 (1996). [20] S. Imamura, T. Nakai, H. Kanai and T. Ito, J. Chem. Soc. Faraday Trans., 91, 1261 (1995). [21] B. Notari, Adv. Catal., 41,253 (1996). 155

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*05012198247* E. Panzer, G. Emig / Lehrstuhl fur Technische Chemie I der Universitat Erlangen-NOmberg, Egerlandstr. 3, D-91058 Erlangen, Germany ,

THE SILVER CATALYST PROCESS FOR CONVERTING METHANOL TO FORMALDEHYDE - KINETIC INVESTIGATIONS

Abstract

{In pre-experiments a-tubular reactor was checked whether it is suitable for kinetic measurement on the system of the silver-catalysed partial oxidation of methanol to formaldehyde. Detrimental effects of heat-transfer and mass-transfer on the exper ­ imental results were ruled out. Investigations on the characteristics of the reaction showed that it is possible to manipulate the composition of the product mixture by changing the inlet concentration of the reactants. A modified power-law model was established to describe the reaction kinetics. It considers the preadsorption step of oxygen on the catalysts surface and fits the experimental data quite well. During the rapid oxidation the catalysts surface undergoes a drastic change. It gets coarse and has an adsorption capacity of 11 m2/g after being exposed to the reaction mixture.

1 Introduction DE99G20 78 ______Formaldehyde is an important intermediate for the mass-production of polymers. It is used for the manufacture of condensates with urea, melamine and phenol. The largest amount of formaldehyde is produced according to the silver catalyst process which is well-tried. It gives a high yield and has a good selectivity towards the desired product. However an improvement of the products chemical composition by a better control of the secondary reactions to water and hydrogen may still be achieved. Furthermore we have not sufficient knowledge about the mechanism of this reaction under industrial conditions yet. The extreme temperature of 600-650 °C and the high reactant concentrations make it difficult to obtain reliable data. Therefore the kinetics of the highly exothermic oxidation have not been measured in this reaction region so far.

2 The Manufacture of Formaldehyde

The typical product mixture of the partial oxidation of methanol consists of formaldehyde as main product and considerable amounts of the by-products water and hydrogen. The occurrence of these main components can be explained by the following reaction equations.

DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 156

The endothermic dehydrogenation of methanol (Eq. 1):

CH3OH <—> CH20 + H2 (1)

The exothermic oxidation of methanol (Eq. 2):

CH3OH + I/2O2 -> CH20 + H20 (2)

Other by-products in the reaction mixture are carbon dioxide, carbon monoxide from the thermal decomposition of formaldehyde, and formic acid.

Early investigations showed that the preadsorption of oxygen on the silver surface is an important reaction step. Methanol can react only on those activated adsorption centers [1]. Schlogl and coworkers [2] characterize three adsorbed species of oxygen. The a- oxygen is dissociatively adsorbed on the silver. It causes the exothermic oxidation to formaldehyde with water as side-product. The 7 -species is called subsurface-oxygen. It is integrated into the first layer of the silver bulk and changes the orientation of its surface. This adsorbate is supposed to dehydrogenate methanol to formaldehyde and hydrogen. The /3-species is dissolved in the bulk and supplies the oxygen for the other two species.

3 Experimental

The conversion of methanol to formaldehyde on a silver catalyst produces a reaction mixture which consists of the reactants methanol and oxygen. Furthermore it contains 3 formaldehyde, water, hydrogen, methyl formate (which is a consecutive product of formic 6 acid), carbon dioxide, carbon monoxide, and nitrogen as inert gas.

For the analysis of this mixture we used a gas chromatograph (Hewlett-Packard, HP 5890). Carbon dioxide, water, methyl formate, methanol and formaldehyde were separated on a fused silica capillary column of the type Poraplot-Q (d;=0.53mm,L=45m). The other components were analyzed on mole-sieves (5 A).

3.1 Reactor

The partial oxidation of methanol is a rapid and highly exothermic reaction. Heat and mass transfer limitations should be minimized. The reactor and the conditions of opera ­ tions should hence be carefully chosen. This is a basic condition for the later simplification of the reactor balance equations. A tubular alumina reactor was used with an inner diameter of 10mm and a length of 500mm. The reaction region can be heated up to about 700°C. After this a heat-exchanger cools down the gas mixture to about 150°C in order to prevent the consecutive thermal de­ composition of the formaldehyde to carbon monoxide and hydrogen. The diluted catalyst bed is fixed between two layers of inert material (alumina or quartz particles). 157

3.2 The Reactor Model

A detailed mathematical description of the reactor has to take into consideration the in­ fluence of the heat-transfer and mass-transfer from the continuous phase to the catalyst particle. A heterogeneous, two-dimensional reactor model can describe this fact. It re­ quires mass-balances and heat-balances for both the particle and the fluid phase. It is quite obvious that this set of equations contains too many parameters and that there is no way to get reliable kinetic parameters using them for the regression analysis. A drastic simplification of the model is inevitable. In order to assess the effect of the mass-transfer the gas velocity was increased in propor ­ tion to the catalysts initial weight in order to get the same hydrodynamic residence time but different flow conditions in the particle bed. The measured values showed no signifi ­ cant change in the conversion of methanol. This shows clearly that the mass-transfer may be neglected. The influence of heat-transfer was estimated according to an equation given by Meats [3]. With typical values for the reaction the condition was just fulfilled. Due to this the experimental results should be discussed carefully. Completely isothermal behaviour of the reactor could not be achieved, but under normal working conditions the gas flow in the reactor has an ideal plug-flow characteristic. The Bodenstein-number >50 showed this clearly [4j. For a first approximation the tubular reactor can be considered to be an ideal plug-flow reactor with restrictions concerning the thermal properties. But its experimental results can surely be used for a qualitative view on the characteristic behaviour and the kinetics of the system. The reactor can be described by the simplified mass-balance (Eq. 3):

(3) i

Alumina was used for the dilution of the catalyst bed and for filling the reactor. While it was considered to be completely inert under reaction conditions, experimental results showed a remarkable influence of the alumina on the reaction. This effect should also be kept in mind when the measuring results are discussed.

4 Experimental results

4.1 The Reaction System

A plot of yield and conversion versus temperature demonstrates that a minimum temper­ ature must be exceeded until the reactivity of the system gets significant (Fig. 1). Beyond 280°C , a drastic increase of the conversion can be observed. With decreasing tempera­ ture the concentration of methyl formate in the reaction mixture grows. Methyl formate 158

• conversion oxygen *yeld formaldehyde •yeld hydrogen AyeUtveier o yield carbon dioxide x yeld carbon moooade oyeti methyl formate

mole fraction otypetc 007

500 temperature ["Cl

Figure 1: Effect of the reaction temperature is probably a consecutive product of the reaction of formic acid with methanol on the GC capillary Poraplot-Q column. Due to this the amount of methyl formate represents the formic-acid in the reaction product. In the large-scale production formic acid has to be removed from the product because it promotes the polymerisation of the aqueous formaldehyde solution. Lefferts [5] describes a similar behaviour with one difference. He detects much smaller concentrations of this product.

At about 600°C a typical increase of the carbon monoxide and hydrogen concentration is observed. This can be explained by the thermal decomposition of formaldehyde

In the further experiments the concentration of oxygen and methanol in the reaction mixture was kept in a range which is close to the conditions of the industrial process. The typical mole fractions vary around xo2 = 0.07; X mc OH = 0.35.

Fig. 2 and 3 show characteristic experimental values plotted versus the concentration of one of the reactants. The yield and selectivity of formaldehyde as well as the conversion of methanol are of the same order as in the manufacturing process. (These values should be compared with those of one pass through the reactor of a large-scale reactor working according to the "methanol-rich" process.)

The ratio of hydrogen to water compares the dehydrogenation course induced by the 7 - oxygen with the oxidation caused by the a-oxygen. From both plots it can be clearly seen that it is possible to change the chemical composition of the product mixture by a variation of the reactant concentration. A promotion of the dehydrogenation would be desirable because the waste gas of the reaction should have a large heat of reaction for a good energy-balance of the process. A small concentration of water in the product mixture is advantageous too because consecutive manufacturing processes need water- free formaldehyde which must be distilled from the aqueous reaction product. The total oxidation of the reactants to carbon dioxide is negligible. 159

3 0,6 *-• selectivity to formaldehyde 2 0,6 •7171 conversion methanol -♦—yield formaldehyde 5 0.3 -** — yield cartoon dead© • 1.0

temperature: 60CTC motefracsoocxypgrt QQ7

0,35 0.4I mole fraction methanol H

Figure 2: Effect of the methanol inlet concentration on conversion, yield and selectivity

If we focus on Fig. 2 in detail we can see that the increase of the inlet concentration of methanol reduces the conversion of methanol as well as the yield and selectivity to formaldehyde. Furthermore the total oxidation to carbon dioxide decreases. But a high methanol concentration promotes the desirable dehydrogenation of methanol (Eq. 1) which can be seen from an increased hydrogen /water ratio in the reaction mixture.

selectivity to formaldehyde —•—conversion methanol e ye>d formaldehyde —*t —yteld cartoon tio»de ratio H2/H20

temperature 600*C

0.10 0.12 0. mete fraction oxygen [♦]

Figure 3: Effect of the oxygen inlet concentration on conversion, yield and selectivity

Fig. 3 shows that an increasing oxygen concentration in the reaction mixture causes more drastic changes in the composition of the product. More oxygen leads to a significant increase of the conversion of methanol and the yield and selectivity to formaldehyde. In 160

contrast to the effect of methanol a higher inlet flow of oxygen causes an increase of the total oxidation to carbon dioxide. Furthermore the oxidation of methanol at the a-oxygen centers with the production of water is promoted (Eq. 2). These experimental results can be explained quite well with the commonly used theory of the reaction mechanism [6 ]. The conversion of methanol needs preadsorbed centers on the silver surface. Due to this requirement the effect of a variation of the methanol inlet pressure must be limited by the preadsorption of the silver with oxygen. According to this fact the influence of the methanol concentration on the composition of the product mixture is smaller too.

4.2 Kinetic Measurements

The first discussion of the experimental results in a kinetic model is based upon a sug ­ gestion by Wiesgickl [7]. He uses the main reaction courses of the system and describes them with power-law equations. The description of the dehydrogenation of methanol takes into account the preadsorption of water. The model assumes a dissociative adsorption of oxygen and integrates it into the rate equation (Eq. 4). This leads to the following equations for the dehydrogenation:

CHzOH A CH20 + H2 n = kxPcihOHPol (4)

CHzOH M CH20 + H2 n/z = kmpcihoPih (5)

The oxidation is described by:

CHzOH +1/202 CH20 + H20 r2 = k2pCHz0HPl£ (6 )

The equation for the complete oxidation of methanol takes into account that methanol is an excess species. Because of this there is only a dependence on the oxygen concentration in the rate equation (Eq. 7):

CH3OH + 3/202 A C02 + 2H20 r3 = h3p l£ (7)

With this set of rate equations and the simplified reactor model (Eq. 3) the mass balance for each species of the reaction can be written down. Based on them the kinetic parameters of the reaction system are estimated The experimental data for the optimization of the parameters are series of measurements with a variable residence time of the reaction mixture in the catalyst bed. The optimiza ­ tion is done by the program Simusolv. The optimization yields a set of parameter values which are given in Table 1. 161

Parameter Value Standarddeviation ki ■ 5.3196 0.148 km 0.18865 0.0036 k2 1.1555 0.0093 k3 2.7266 0.15

Table 1: Parameters of the rate equations

All parameters are significant. The model fits the concentration curve of methanol quite well. Formaldehyde is fitted with a larger fault. The concentration of oxygen is poorly predicted by this set of parameters. This may be caused by the small concentration of oxygen in the product stream which leads to a larger measuring error for this species and affects the quality of the experimental data.

4.3 SEM-Investigations on the Silver Catalyst

The surface of the silver catalyst undergoes a drastic change during the reaction. Our experiments showed that there is a difference how the reaction was controlled.We had two general reaction conditions: The kinetic reaction conditions are characterized by a small initial catalyst weight and an incomplete conversion of oxygen. It was used for kinetic measurements. The industrial reaction conditions are characterized by an excess of catalyst and a complete conversion of oxygen like it is used in industrial manufacture. A fresh silver sample shows a plain metallic surface under the SBM (Fig. 4a). The industrial sample has small holes of about 1 pm diameter (Fig. 4b). They may be explained by the desorption of subsurface 7-oxygen caused by its reaction with adsorbed surface species. The edges of the silver particles are slightly smoothed. This might be due to sintering of the metal.

The silver sample used under kinetic conditions shows a drastically changed surface (Fig. 4c). The metal surface of the grains is now coarse. Small holes like on the industrial sample are rarely found. A measurement of the BBT surface showed that the silver has an adsorption surface of about 11 m2/g. The surface of the fresh silver sample could not be measured this way.

An explanation of the difference between the kinetic and the industrial sample might be the different catalyst mass. In the kinetic sample there was only a small amount of catalyst. Bach catalyst particle was part of the reaction and was subjected to the thermal and chemical stress of the rapid oxidation reaction. In the industrial sample only the first layer of silver crystals was involved in the reaction. The catalyst experienced a much smaller thermal and chemical stress.

References

[1] I. N. Vlodavets, S.Y. Pshezhetskii. Zh. Fiz. Khim., 25:612,1951. 162

[2] B. Pettinger, R. Schlogl, G. Ertl X. Bao, M. Muhler. Catalysis Letters, 22:215-225, 1993.

[3] D.E. Meats. Ind. Eng. Chem. Process Des. Dev., 10:541,1971.

[4] E. Fitzer, W. Fritz, G. Emig. Technische Chemie - Einfuhrung in die chemische Reaktionstechnik. Springer Verlag, 1991. [5] L. Lefferts. PhD thesis, University of Twente, 1987.

[6] H. Schubert, U. Tegtmeyer, R. Schlogl. Catalysis Letters, 28:383-395,1994. [7] G. Wiesgickl. Master’s thesis, Friedrich-Alexander Universitat Erlangen-Niirnberg, 1989. 163

DGMK-Confereoce "Selective Oxidations in Petrochemistry", Hamburg 1998

A. Reitzmann*. E. Klemm*, G. Emig*, S. A. Buchholz**, H. W. Zanthoff** * Institute of Technical Chemistry I, University of Erlangen-Numberg, Egerlandstr. 3, D-91058 Erlangen, Germany ** Institute of Technical Chemistry, Ruhr-University Bochum, Bochum, Germany

PHENOL BY DIRECT HYDROXYLATION OF BENZENE WITH NITROUS OXIDE - ROLE OF SURFACE OXYGEN SPECIES IN THE REACTION PATHWAYS

*DE012J98256* Abstract ^Transient experiments in a Temporal Analysis of Products (TAP) Reactor were per­ formed to elucidate the role of surface oxygen species in the oxidation of benzene to phenol on ZSM-5 type zeolites with nitrous oxide as a selective oxidant. It was shown by puls experiments with nitrous oxide that the mean lifetime of the generated surface oxygen species is between 0.2 s at 500 °C and about 4.2 s at 400 °C. After­ wards the surface oxygen species desorb as molecular oxygen into the gas phase where total oxidation will take place if hydrocarbons are present. Dual puls experi ­ ments consisting of a nitrous oxide puls followed by a benzene puls allowed studying the reactivity of the surface oxygen species formed during the first puls. The obser ­ vation of the phenol formation was impeded due to the strong sorption of phenol. Multipulse experiments were necessary to reach a pseudo steady state phenol yield.

Introduction

The one-stage oxidation of benzene to phenol is of outstanding economic interest due to the well-known disadvantages of the conventional cumene route [1], Up to now only nitrous oxide was found to be a sufficiently selective oxidant for the one- stage route in gas phase on ZSM-5 type zeolites. Panov et al. [2] postulate a specific form of surface oxygen formed upon decomposition of nitrous oxide on ZSM-5 type zeolites containing extraframework iron clusters. The existence of such surface oxy ­ gen species was proved by temperature programmed decomposition of nitrous oxide in a closed system and monitoring the relevant species nitrous oxide, nitrogen and

DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 164

oxygen mass-spectrometrically. They found that at temperatures lower than 300°C nitrous oxide decomposes with release of molecular nitrogen but without appearance of molecular oxygen in the gas phase. The desorption of the chemisorbed oxygen started at 300 °C.

These observations were also confirmed by [3] when performing temperature pro ­ grammed continuous flow experiments in a Temporal Analysis of Products (TAP) reactor. Thus, the formation of such surface oxygen upon decomposition of nitrous oxide seems to be a proven fact. However, the knowlege about the active center it­ self is still a matter of discussion. As already mentioned above, Panov et al. [4] fa­ vour iron or even iron impurities as the only active centers, whereas some other authors stress the role of extra lattice alumina [5,3]. Kazanski et al. [6] consider Lewis sites due to defects of the zeolite lattice to be relevant. In the present paper we will not focus on the question of the nature of the active site. The goal of the pre­ sent paper is to elucidate the properties of the surface oxygen species like its lifetime and its reactivity in order to understand the crucial steps affecting activity and selec­ tivity of the direct oxidation of benzene to phenol with nitrous oxide. As an appropri ­ ate method transient experiments in a Temporal Analysis of Products (TAP) reactor were performed.

Experimental As an appropriate catalyst for maximum selectivity (70 %) at high conversion (30 %) a ZSM-5 with a high fraction of extra lattice alumina (20 % measured by 27AI NMR) and very low iron oxide content (0.05 wt. %) was chosen [3]. The ZSM-5 was syn ­ thesized in our laboratory according to the german patent DE2831611 [7], The over ­ all SiOz/AI^O; modulus was about 50. The TAP reactor has been described previously in detail [8], Thus, only a brief sum­ mary of the experimental data is given here. The zeolite catalyst is placed in the iso ­ thermal zone of the microreactor (mraL= 50 mg), heated to 823 K to desorb impurities and water from the catalyst surface. Afterwards the catalysts is brought to the de­ sired reaction temperature. In principal, two experimental modes of the TAP reactor are possible: continuous flow experiments (p = 1 torr) and vacuum puls experiments 165

(p = 10'® Torr; pulse size: 10" molecules/pulse). In the present paper only the pulse experiments will be discussed since the continuous flow experiments have been al­ ready described in [3], The pulses of the reactants and of the products are monitored at the reactor outlet by recording a characteristic mass value of the interesting spe ­ cies depending on time with a quadrupole mass spectrometer.

Results and discussion A) Single Pulse Experiments with NzO Single N20 pulses of about 10" molecules have been injected at the reactor inlet at different temperatures. In figure 1 the corresponding outlet signals are depicted. It can be seen that the release of molecular oxygen occurs delayed in time which can not be explained by an interaction between molecular oxygen and the catalyst sur­ face proved by pulsing molecular oxygen. Thus, it can be concluded that a chemi­ sorbed oxygen species has been formed upon decomposition of N20. The mean life time (first moment of the desorption peak) of that species is increasing if the tem­ perature is decreasing: 0.2 s at 500 °C, 0.4 s at 458 °C, 1.6 s at 420 °C and 4.2 s at 390 °C (not depicted in Figure 1). At temperatures lower than 390 °C the oxygen re­ lease is no more detectable because it is superimposed by the analytical noise. When conducting a series of single N20 pulses at 500 °C with different time intervals between two pulses following each other it was observed that the released amount of molecular oxygen depends on the period of the multi pulse experiment: with a period of 1 minute only half of the oxygen release was observed than with a period of 3 minutes. Thus, 3 minutes seem to be sufficient to regain a clean catalytic surface since a period of 10 minutes results in the same amount of oxygen release. This may allow the conclusion that parallel or subsequent to the mass-spectrometrically moni ­ tored oxygen release (depicted in Figure 1) a slow desorption process takes place which can not be detected directly. This can be due to the two-center recombination of adsorbed atomic oxygen or due to a further adsorbed oxygen species with a longer lifetime. The interaction between NzO and the catalytic surface must be weak at the chosen temperatures since no significant spreading compared to the Ne signal was ob ­ served (see Figure 1). This is consistent with the static experiments of Panov et al. 166

[4] who observed that NzO adsorbed at 50 °C is completely desorbed if the tem­ perature exceeds 200 °C. Thus, the decomposition of the N20 must be a very rapid process. This can be also deduced from Figure 1 because there is no time shift be­ tween the nitrogen signal (contribution of N20 to AMU=28 is subtracted) and the N20 signal.

NnO-Z N, + O-Z

O-Z + O-Z

0.0 0,5 1.0 1,5 2,0 2.5 3,0 3,5 4,0 time [s]

Figure 1: Single puls experiments with N20 at various temperatures.

B) Dual Puls Experiments with N20 and benzene (at 430 °C)

In the dual puls experiment at 430 °C first a N20 pulse is injected to form the surface oxygen species and with an offset of 0.4 s a benzene pulse is introduced. The offset of 0.4 s ensures that no more N2Q is present in the gas phase (see Fig. 1) and al­ lows studying the interaction between benzene and oxygen. In Figure 2 the corres ­ ponding signal of molecular oxygen is depicted. It can be seen that with the injection of benzene at 0.4 s offset the amount of molecular oxygen in the gas phase is abruptly reduced towards zero. Since no C02 is detected and only a very small amount of CO which corresponds to 1 % of the observed oxygen consumption it can 167

be excluded that the oxygen is consumed by total oxidation (see Figure 2). Thus, the adsorbed surface oxygen species must react with benzene to adsorbed oxidized hy ­ drocarbon species. From experiments on the same catalyst and at the same tem­ perature in an integral reactor apparatus it is known that only phenol is formed ini­ tially and no other oxidized compounds [9], Unfortunately, in the TAP experiments no phenol production could be detected in the gas phase when injecting benzene (Fig. 2). By multi pulse experiments described in subsection C) it could be shown that this can be explained with a very strong adsorption of phenol.

Consumed amount of oxygen

PHENOL

time [s] time [sj

Figure 2 Dual puls experiments with N20 and benzene at 430 °C. Left figure: 1: Oxygen release in the Single Puls Experiment with NzO 2: Oxygen release in the Dual Puls Experiment with N20 and benzene (offset 0,4s) shadded area: oxygen consumption due to benzene injection Right figure: Products observed in the gas phase in the Dual Puls Experiment with N20 and benzene (corresponds to curve 2 in the left figure) 168

C) Multipuls Experiments with N20 (at 400 °C) In the dual puls experiments an oxygen consumption was observed without detecting an adequate amount of product molecules in the gas phase. It is most likely that the phenol formed by selective oxidation is strongly adsorbed on the catalyst surface. In order to prove that assumption, multi puls experiments with a 1:6 mixture of benzene and N20 are performed at 400 °C and the characteristic mass of phenol is monitored. Figure 3 shows that about 30 pulses are necessary to detect phenol for the first time in the gas phase. After about 4000 pulses a pseudo steady state is observed, i.e. sorption equilibrium has been reached. It has to be noted that each multi puls ex­ periment comprises 1000 pulses so that between two subsequent experiments a short break occurs in which phenol desorbs to a certain amount into the high vac­ uum. This is the reason why one multi pulse experiment does not continue at the phenol production level of the preceding experiment. The strong sorption of phenol compared to that of benzene influences the activity and the selectivity of the phenol production [10].

catalyst pretreated with 5000 pulses

catalyst pretreated with 4000 pulses'

catalyst pretreated with 1000 pulses

fresh catalyst

number of pulses [-]

Figure 3: Multi puls experiment with a N20/benzene mixture at 400 °C [11] 169

Conclusions Single puls experiments with N20, dual pulse experiments with N20 and benzene, and multipuls experiments with a N20/benzene mixture were performed in the Tem­ poral Analysis of Products (TAP) reactor at various temperatures using a HZSM-5 catalyst with 20 % extraframework alumina and 0,05 wt% iron oxide. These experi ­ ments allow the following conclusions concerning the reaction mechanism of the one-stage synthesis of phenol from benzene using N20 as a selective oxidant: • A chemisorbed surface oxygen species is formed upon N20 decomposition.

• The surface oxygen species desorbs as molecular oxygen into the gas phase

resulting in a mean life time on the surface between 0.2 s at 500 °C and 4.2 s at

390 °C. • If benzene is present the surface oxygen species reacts to 99 % with benzene to form phenol (at 430 °C). • Phenol accumulates on the catalyst surface until a sorption equilibrium is re­ ached.

Furthermore, it can be concluded that a catalyst is highly selective in phenol forma ­ tion if the formation of molecular oxygen due the desorption of the surface oxygen species is much slower compared to the reaction of the surface oxygen with benze­ ne to form phenol. Phenol should desorb as fast as possible to avoid product inhibiti­ on and further oxidation. 170

References

[1] Weissermel, K„ Arpe, H.-J., Industrial Organic Chemistry, 3rd edition, VCH, Weinheim 1997. [2] Panov, G.I., Sobolev, V.I., Kharitonov, A.S., J. Molec. Catal. 61 (1990), 85. [3] Klemm, E„ Reitzmann, A., Buchholz, S., Zanthoff, H., Chem.-Ing.-Tech. 70, in press. [4] Panov, G.I., Uriarte, A.K., Rodkin, M.A., Sobolev, V.I., Catal. Today, 41 (1998), 365 [5] Motz, J.L., PhD Thesis, RWTH Aachen, 1995. [6] Zhobolenko, G., Kazanskii, V., Kinetika iKataliz 32_(1991), 151. [7] BASF AG, Patent DE 19634406,1980. [8] Buchholz, S.A., Zanthoff, H.W., ACS Symp. Ser. 638 (1996), 259. [9] Hafele, M„ Reitzmann, A., Klemm, E., Emig, G., Stud. Surf. Sci. Catal. 110 (1997), 844. [10] Klemm, E., Wang, J.-G., Emig, G., Microporous and Mesoporous Materials, in press. [11] Hafele, M„ PhD Thesis, University of Erlangen-Nuremberg , 1997. 171

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

St. HeR, M. Liauw, G. Emig Lehrstuhl furTechnische Chemie I der UniversitSt Erlangen-Numberg, Egerlandstr. 3, D-91058 Erlangen, Germany

PARTIAL OXIDATION OF N-BUTANE TO MALEIC ANHYDRIDE OVER A VANADIUMPYROPHOSPHATE CATALYST IN THE RISER REGENERATOR SYSTEM

Manuscript not available

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 172 173

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*DE012198265* S. Ernst, J. M. Teixeira Florencio * Department of Chemistry, Chemical Technology, University of Kaiserslautern, P.O. Box 3049, D-67653 Kaiserslautern, Germany

OXIDATION OF AROMATIC ALCOHOLS ON Zf -m/'a net ii ATED COPPER AMINO ACID COMPLEXES

^Abstract / Copper complexes of the amino acids histidine, arginine and lysine have been introduced into the supercages of zeolite Y and, for the first time, into the large intracrystalline cavities of zeolites EMT and MCM-22. The resulting host/guest compounds are characterized by X-ray powder diffraction, UVAZIS- spectroscopy in the diffuse reflectance mode and by catalytic tests in the liquid- phase oxidation of aromatic alcohols (viz. benzyl alcohol, 2- and 3-methylbenzyl alcohol and 2,5-dimethylbenzyl alcohol) with tertiary-butylhydroperoxide as oxidant. It was observed that intracrystalline copper-amino acid complexes possess remarkable catalytic activity, yielding the corresponding aromatic aldehydes and acids^j

Introduction The preparation and catalytic properties of transition metal complexes immobilized in zeolites or zeolite-like microporous or mesoporous materials is currently under intense study [e.g., 1-3]. Pertinent examples for such host/guest materials are phthalocyanine-, bipyridine- and salen-complexes of metal ions like Fe2*, Co2+or Mn2+ immobilized in the intracrystalline cavities of zeolites Y or EMT [4-6]. One major aspect in this context is the possibility of heterogenizing homogeneous catalysts and thereby combining the advantages of both catalytic principles, viz. high catalytic activity which allows to perform reactions under relatively mild conditions, combined with an easy separation of the catalyst from reactants and products in solution. It is another aspect of these host/guest compounds that they can be looked upon as models for enzymes: the encaged complexes represent mimics of the active centers, while the inorganic zeolite matrix plays the part of the protein mantle [7]. More recently, Weckhuysen et al. [8,9] reported the incorporation of copper(ll)-amino acid complexes, where one copper ion is complexed by two amino acid ligands, in zeolite Y via ion exchange of the preformed complexes from aqueous solution into the zeolite pores and cavities. Furthermore, the resulting host/guest compounds were explored for their catalytic activity in the selective liquid-phase oxidation of substrates like cyclohexane, cyclohexene, 1-pentanol and benzyl alcohol with tertiary-butylhydroperoxide (TBHP) as oxidant [8,9]. It was found that the

DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 174

immobilized amino acid complexes possess high catalytic activities already at relatively mild reaction temperatures (i. e., at 60 °C).

In the present contribution, we report our results on the introduction of copper(ll) complexes of the amino acids histidine, arginine and lysine (cf. Fig. 1) into the intracrystalline voids of two large-pore zeolites, viz. Y and EMT (large intracrystalline cavities with twelve-membered ring pore openings), and into the large cavities of zeolite MCM-22, which are only accessible via smaller ten-membered ring windows. The catalytic properties of the prepared host/guest compounds were explored in the liquid-phase oxidation of selected a aromatic alcohols differing in size and shape (viz. benzyl alcohol, 2- and 3- n methylbenzyl alcohol, and 2,5-dimethylbenzyl alcohol) in order to detect influences of the size of the zeolite pores or the reactants on the activity and/or selectivity of the immobilized complexes.

Histidine Arginine Lysine

T OH 0^ ^NH2

NH2

NH N Jm HN=== h2n nh 2

Fig. 1: Structures of the amino acids histidine, arginine and lysine used as ligands for the formation of intrazeolitic copper(ll)-amino acid complexes.

Experimental Section Zeolite NaY was a commercial product (nsJnK = 2.6; obtained from Union Carbide Corp., Tarrytown, N. Y., USA). Zeolites EMT OVn^ = 3.8; [10]) and MCM-22 (nsi/n/y = 20; [11]) were synthesized according to published procedures. Prior to any further modification steps, the as-synthesized zeolites were calcined for 16 hours at 540 °C in a purge of 5 vol.-% oxygen in nitrogen to remove the organic templates. Subsequently, an ion exchange with a diluted aqueous solution of NaCI was performed. Our first attempts to introduce preformed copper(ll)-histidine complexes into zeolite Y via ion exchange techniques (as previously described in [8,9]) failed. Therefore, the flexible- ligand method was applied in the case of copper-histidine complexes. For this 175 purpose, the zeolites were ion exchanged with aqueous solutions of Cu(N03)2 such as to obtain copper loadings of one Cu2+ in every second large cage of the zeolite. The copper-exchanged zeolites were then suspended in a solution of a surplus of L-lysine (Fluka; nUySine/ncuz* = 5) in bidistilled water for 12 hours at room temperature. During this procedure, the pH of the suspension was kept constant at 7.25. Afterwards, the solid was recovered by filtration, washed with distilled water and dried at 60 °C in air. The introduction of copper(ll)-arginine and copper(ll>lysine complexes into the zeolitic host materials occurred via the method described by Weckhuysen et al. [8,9], i.e., the Na-exchanged forms of the zeolites were suspended in aqueous solutions containing Cu(N03)2 and the respective amino acid (L-arginine or L-lysine; Fluka) in a molar ratio of 1 : 5. The pH of the suspension was adjusted to 10. It was stirred for 18 hours at room temperature. Afterwards the zeolite was filtered, washed with distilled water and dried in air at 60 °C. The prepared host/guest compounds were characterized by X-ray powder diffraction and diffuse reflectance UV/VIS- spectroscopy. The catalytic tests were conducted in a batch-type continuously stirred reactor thermostated at 60 °C. In each experiment, 10 mmol of reactant and 20 mmol of oxidant (tertiary-butylhydroperoxide, TBHP, 70 wt.-% in water) were mixed with 30 cnr of acetonitrile as solvent and 0.1 g of catalyst were added. The composition of the reaction mixture was followed in dependence of the reaction time using temperature programmed capillary gas chromatography. Turnover numbers (TON) were calculated as moles of reactant converted, divided by the number of copper complexes present in the catalyst.

Results and Discussion X-ray powder diffraction of the samples before and after the introduction of the complexes revealed that the structures of the zeolitic host materials remained virtually unchanged upon incorporation of the copper-amino acid complexes. UV/VIS-spectra of the (light-blue colored) copper-histidine containing zeolites were very similar to those reported by Weckhuysen et al. [8,9], viz. they show two specific absorption bands at wavenumbers around ca. 40000 cm'1 and ca. 16000 cm* 1, respectively. While the former can be attributed to a ligand-to- metal chargetransfer, the latter is commonly assigned to a d-d absorption [8,9]. The UV/VIS-spectra for the copper-arginine- and the copper-lysine-containing zeolites are depicted in Figs. 2 and 3. It can be seen that all prepared materials exhibit the typical spectrum with a ligand-to-metal chargetranfer band occurring at ca. 43000 cm'1 and the d-d-transition band at ca. 16000 cm* 1. Note, however, that there is a small but significant shift in the maximum of the d-d-transition band for arginine and lysine immobilized in MCM-22 to lower wavenumbers as compared to zeolites Y and EMT. This shift can be attributed to changes in the steric and/or electronic environment in the different types of host materials. As a whole, the results of solid-state UV/VIS-spectroscopy show that 176

- * [Cu(Argmin6yMCM*22 ...... [Cu

X100

50000 40000 30000 20000 10000

Wavenumber / cm'

Fig. 2: UV/VIS-spectra of copper-arginine-loaded zeolites Y, EMT and MCM-22.

------lCu(LysineyMCM-22 ...... [Cu(LysineyEMT ------[Cu(Lysine) 2]Y X100

50000 40000 30000 10000

Wavenumber / cm Fig. 3: UV/VIS-spectra of copper-lysine-loaded zeolites Y, EMT and MCM-22. 177

the copper-amino acid complexes can be introduced not only into the intracrystalline cavities of the large pore (twelve-membered ring) zeolites Y and EMT, but also into the cages of the MCM-22 structure.

The catalytic properties of the host/guest compounds were first explored in the liquid-phase oxidation of benzyl alcohol. Typical results obtained with copper- arginine- and copper-lysine complexes in zeolite MCM-22 are depicted in Fig. 4. Both catalysts are active in the oxidation of benzyl alcohol to benzaldehyde and further to benzoic acid. Additional results obtained in the oxidation of benzyl alcohol over different catalysts are summarized in Table 1. By contrast to the reports of Weckhuysen et al. [8,9], copper-histidine complexes in zeolite Y (but also in zeolites EMT and MCM-22) are not very active for the oxidation of benzyl alcohol under the conditions applied in the present study. One tentative explanation for this experimental observation is probably the much higher complex content (4 per unit cell of zeolite Y as compared to 0.25 per unit cell, cf. [8,9]) of our catalysts, which could lead to a partial pore blockage, in particular with the relatively large histidine as ligand.

|Cu(Arginine),]MCM-22 [Cu(Lysine) 2lMCM-22

£

▼ Benzyl alcohol a . • Benzaldehyde >- ■ Benzoic acid

REACTION TIME / h

Fig. 4: Oxidation of benzyl alcohol on copper-arginine- and copper-lysine complexes in zeolite MCM-22 as host material. 178

Table 1: Results of the oxidation of benzyl alcohol on copper-amino acid complexes in zeolites with different pore architectures after ca. two days of reaction (Bale: benzyl alcohol; Bald: benzaldehyde; Bac: benzoic acid; TON: turnover number).

Catalyst Xealc.% YBald.% YBac,% TON

CuHisY CuHisEMT 9 6 3 5 CuHisMCM-22 3 3 0 49

CuArgY 43 31 12 230 CuArgEMT 65 34 31 368 CuArgMCM-22 82 20 62 1457

CuLysY 86 27 59 457 CuLysEMT 76 29 47 431 CuLysMCM-22 90 23 67 1361

It also follows from Table 1, that copper-arginine- and copper-lysine-complexes immobilized in zeolites Y, EMT and MCM-22 are quite active in the oxidation of benzyl alcohol to benzaldehyde and benzoic acid. Moreover, it can be seen that the same complexes, when deposited in zeolite MCM-22, are much more active than in the large pore zeolites Y and EMT. Probably, the somewhat elongated shape of the large cages in MCM-22 (as compared to the almost spherical cages in zeolites Y and EMT) results in a different orientation of the encaged complexes with a better accessibility of the active sites for the reactant molecules. Indeed, a different orientation and/or different electronic interactions have been suggested based on the results of the UV/VIS-spectroscopic characterization (see above).

Pertinent results for the oxidation of substituted benzyl alcohols which differ in their molecular sizes are depicted in Fig. 5. Note that only oxidation of the hydroxy group was observed, possible products from an oxidation of the methyl groups were absent under the conditions applied in the present study. It can be seen that the highest conversion over copper-arginine-Y is observed for 3- methylbenzyl alcohol, while the reactivity of 2-methylbenzyl alcohol and 2,5- dimethylbenzyl alcohol is significantly smaller. These data can be interpreted as 179

[Cu(ArginineyMCM-22 '

• 2-Mcthytbenzyl alcohol v 34/ethyt benzyl alcohol ■ 2,5-DimelhylbenzylaJcohol

10 20 30 40 50 10 20 30 40 50

REACTION TIME / h

Fig. 5: Oxidation of 2- and 3-methylbenzyl alcohol and of 2,5- dimethylbenzyl alcohol on copper-arginine complexes in zeolites Y and MCM-22. resulting from increased difficulties for the reactant molecules to approach the active sites with the hydroxy group. In principle, the same reactivity sequence as with copper-arginine-Y is also observed with copper-arginine in zeolite MCM-22 (cf. Fig. 5). Although expected based on their sizes, no exclusion effects were obviously excerted by the relatively narrow ten-membered ring windows on the conversion of the more bulky reactants (i.e., reactant shape selectivity). Again, the same type of complexes encaged in the cavities of zeolite MCM-22 are more active as if they were immobilized in zeolite Y.

Conclusions Copper complexes of the amino acids histidine, arginine and lysine can be immobilized in the large intracrystalline cavities of zeolites Y, EMT and MCM- 22. The resulting host/guest compounds (except for the case of copper-histidine complexes) possess high catalytic activities for the oxidation of aromatic alcohols with tertiary-butylhydroperoxide as oxidant and under relatively mild reaction conditions. The sole products are the corresponding aldehydes and acids. Oxidation of the benzene ring or of the methyl substituents could not be observed. The relative reactivities of the different reactants seem to be governed by steric constraints on the level of the interaction of the hydroxy group with the active sites, rather than by the window size of the zeolite host. 180

Acknowledgements The authors gratefully acknowledge financial support by Deutsche Forschungsgemeinschaft, Ponds der Chemischen Industrie and Max-Buchner- Forschungsstiftung.

References [1] D. E. De Vos, F. Thibault-Starzyk, P. P. Knops-Gerrits, R. F. Parton and P. A. Jacobs, Macromolecular Symposia 80,157 (1994). [2] G. Schulz-Ekloff and S. Ernst, in: "Handbook of Heterogeneous Catalysis", G. Ertl, H. Knfizinger and J. Weitkamp, Eds., Wiley-VCH, Weinheim, 1997, pp. 374-387. [3] K. J. Balkus, Jr. and A. G. Gabrielov, J. Inch Phenom. 21,159 (1995). [4] P. P. Knops-Gerrits, F. Thibault-Starzyk and P. A. Jacobs, Nature 369, 543(1994). [5] C. Bowers and P. K. Dutta, J. Catal. 122,271 (1990). [6] S. Ernst, Y. Traa and U. Deeg, in: "Zeolites and Related Microporous Materials: State of the Art 1994", J. Weitkamp, H. G. Karge, H. Pfeifer and W. Htilderich, Eds., Studies in Surface Science and Catalysis, Vol. 84 B, Elsevier, Amsterdam, 1994, pp. 925-932. [7] R. F. Parton, D. De Vos and P. A. Jacobs, in: "Zeolite Microporous Solids: Synthesis, Structure and Reactivity", E. G. Derouane, F. Lemos, C. Naccache and F. Ramoa Ribeiro, Eds., Kluwer Academic Publishers, Dordrecht, The Netherlands, 1992, pp. 555-578. [8 ] B. M. Weckhuysen, A. A. Verberckmoes, I. P. Vannijvel, J. A. Pelgrims, P. L. Buskens, P. A. Jacobs and R. A. Schoonheydt, Angew. Chem. 107, 2868 (1995). [9] B. M. Weckhuysen, A. A. Verberckmoes, L. Fu and R. A. Schoonheydt, J. Phys. Chem. 100,9456 (1996). [10] J. Weitkamp, R. Schumacher and U. Weiss, Chem.-lng.-Tech.64, 1109 (1992). [11] S. Unvem'cht, M. Hunger, S. Ernst, H. G. Karge and J. Weitkamp, in: "Zeolites and Related Microporous Materials: State of the Art 1994", J. Weitkamp, H. G. Karge, H. Pfeifer and W. Htilderich, Eds., Studies in Surface Science and Catalysis, Vol. 84 A, Elsevier, Amsterdam, 1994, pp. 37-44. DGMK-Conference “Selective Oxidations in Petrochemistry", Hamburg 1

*DE012 198274* "7? E. M. Gaigneaux' 1, M.-L. Naeye, 0. Dupont, M. Gallant"1, B. Kartheuser"1, P. Ruiz and B. Delmon Unite de catalyse et chimie des materiaux divis§s, Universite catholique de Louvain, Place Croix du Sud 2/17, B-1348 Louvain-La-Neuve, Belgium “Charg 6 de recherche" fellow for the Ponds National de la Recherche Scientifique of Belgium "•> CERTECH - Pole CTAS. Zone industrials C, B-7180 Senefffe (Belgium)

a-Sb204-INDUCED IMPROVEMENTS OF THE CATALYTIC BEHAVIOR OF Mo0,-(010) IN THE OXYGEN-ASSISTED DEHYDRATION OF 2-BUTANOL: IMPLICATIONS IN SELECTIVE OXIDATION

Abstract /This article concerns the synergetic effects between an M0O3 sample composed of crystallites exposing preferentially the (010) basal faces and a-Sbg 04 in the oxygen-assisted dehydration of 2-butanol at 220 °C. The conversion of 2-butanol and the yield to butene improved when M0O3 was reacted in the presence of a-Sb204. The origin of the synergism is discussed. When reacted in the absence of a-Sb204, M0O3 got over-reduced and fragmented to M0O2. M0O2 is intrinsically less active than M0O3 thus explaining that the deep reduction of M0O3 corresponds to its tendency to deactivate. In the presence of a-Sb204, the formation of M0O2 is inhibited with, as a consequence, the absence of deactivation. This leads to the synergetic effects obtained with the mechanical mixture of M0O3 with a-Sb204. )

1. Introduction The cooperation effects between M0O3 and a-Sb204 have been recently investigated in the simultaneous dehydrogenation-dehydration of 2-butanol to methyl-ethyl-ketone (MEK) and butene [1]. Synergetic increases in the conversion of 2-butanol, in the yields to butene and to MEK and in the selectivity to MEK were measured : the performances obtained using physical mixtures containing the two phases were higher than those of the constituting oxides tested individually. As no indication of mutual contamination between the simple oxides during the reaction was detected, the synergism was attributed to the occurrence of a remote control mechanism. The model considers than a-Sb204 activates O2 from the gas phase to spillover oxygen (Oso). This migrates onto the surface of M0O3 with which it reacts, so creating and regenerating the active sites [2], As a result, the performances of M0O3 are higher when tested in the presence of a-Sb204 (i.e. when Oso is present). The main finding of the study was the discovery that a remote control mechanism could operate at relatively low temperatures, namely below 200 °C [1]. The practical effects of Oso at the surface of M0O3 to improve its performances were discussed (creation of new active acid sites, inhibition of the deactivation of redox sites and inhibition of coke formation) but the detailed mechanism by which Oso promotes the synergy could not be identified experimentally [1], Until recently, we were facing a very similar situation with the synergetic effects between M0O3 and a-Sb204 in the selective oxidation of isobutene to

DGMK-Tagungsbericht 9803, 3-931850-44-7,1998 182

methacrolein. The synergetic creation of active sites at the surface of M0O3 was attributed to the reaction of of its surface with Oso produced by (%-Sb2O4. In attempts to determine the mechanism explaining this creation of sites, we had decided to use special M0O3 samples with crystals exposing preferentially the (010) basal faces (instead of those used previously possessing an isotropic morphology). This strategy led us to demonstrate that the creation of new sites under the influence of Oso corresponded to the creation of new active (100) steps in the middle and on the edges of the (010) faces. The approach in the experiments presented in this contribution was to use the same strategy in the reaction of dehydration-dehydrogenation of 2-butanol. We therefore used here oriented M0O3 crystals exposing the (010) basal faces to a larger extent than those used in our previous experiments [3,4], In addition to the catalytic tests, we report on the characterization of the samples before and after reaction using X-ray diffraction (XRD), specific area measurements and environmental scanning electron microscopy (ESEM). The mechanism by which the deactivation of M0O3 is avoided and by which its catalytic performances are improved when this oxide is reacted in the presence of a-Sb204 (i.e. when Oso is present) is demonstrated experimentally. Because there were some indications that a deep reduction of M0O3 to lower molybdenum oxides may occur during the 2-butanol reaction, we additionally evaluated the catalytic performances of M0O2 in the same conditions. 2. Experiments 2.1. Preparation of the samples An a-Sb204 sample was synthesized by calcination in air at 450 °C during 20 hours of Sb203 (Aldrich, 99%) previously dispersed in concentrated nitric acid (UCB, 70+%) at 50 °C during 1 hour, washed with distilled water and dried overnight in air at 110 °C. The XRD pattern of the sample obtained was that of the cervantite phase standard [5] and its specific area was 1.7 nr g . An M0O3 sample with crystals developing preferentially the basal (010) faces was synthesized by recrystallizing a commercial molybdenum trioxide (BDH Chemicals, 99.5+%) under a flow of pure 02 (Air Liquids, 99.995%) at 600 °C during 12 hours. The obtained yellowish powder, hereafter noted "MoO 3-(010)", exhibited all the XRD peaks of the molybdite phase standard [5]. The ratio X IhOO / X lokO. where X IhOO and X lokO are the sum of the intensities of the XRD peaks of the (hOO) and (OkO) series respectively, was 0.03 (compared to ratios around 0.1 Tor M0O3 with isotropic crystals [6-8 ]). The specific area of MoC> 3-(010) was 0.2 nr g* 1. A commercial M0O2 toldrich, 99%) was used as received from the supplier. The specific area was 3.1 nr g . A physical mixture, called hereafter"mechanical mixture", of Mo03-(010) and a-Sb204 was prepared by i) interdispersion of the two powders in n-pentane (Aldrich, 98% - 100 ml for 1 g of powder) using ultrasounds during 10 minutes, ii) evaporation in vacuo of the n-pentane in a rotavapor maintaining the sample under agitation, iii) drying in air at 80 °C overnight. The mixture contained 72%wt. of Mo C3-(010) and 28%wt. of a-Sb2C>4. The interdispersion of the two oxides and the absence of morphological perturbation for the MoO 3-(010) crystals after the mixing were checked by ESEM and XRD. The XRD pattern of the mixture fitted with 183

that theoretically calculated on the basis of the patterns of the pure oxides and of their respective mass fractions in the mixture. In order to rigorously compare the catalytic performances of the mixture with those of the pure oxides, these were submitted exactly to the same "mixing" procedure before being catalytically tested. The XRD patterns of the pure oxides after the mixing procedure fitted perfectly with those of the fresh oxides. ESEM did not indicate any modification of the MoO 3-(010) crystals after the mixing procedure. Figure 1 shows ESEM micrographs of the pure MoO 3-(010) after the mixing procedure (Figure 1 left) and of the mechanical mixture of MoC>3-(010) with a-Sb204 (Figure 1 right).

Figure 1. ESEM micrographs of pure MoO3-(010) after the mixing procedure (left), and of the mechanical mixture of M0O3 (010) with a-Sb204 (right). Some particles of a-Sbz04 are pointed with arrows. The bars represent 10 pm.

2.2. Catalytic activity measurements The dehydration of 2-butanol was performed at atmospheric pressure in a fixed bed microreactor. The reaction gas was dry air (Air Liquids, S) saturated with 176 Torr of 2-butanol (Aldrich, 99%). The flow was 90 ml min"1. The reaction temperature was 220 °C. The masses of catalysts used were selected so that the area developed by each phase (when present) in the reactor was identical when tested alone and in the mixture : namely, 1250 mg for MoO 3-(010), 480 mg for a-Sb204 and 1730 mg for the physical mixture. The catalytic test with M0O2 was carried out with 100 mg. In order not to perturb the orientation of the crystals of Mo 03-(010) (when present), all the catalysts were used as powders without pressing them into pellets. The volume of the catalytic bed was kept constant for all the tests by diluting the catalysts in glass balls (diameter between 0.49 and 0.72 mm) previously checked to be inactive. The duration of the tests was usually 12 hours. An additional test during 6 hours was run with the pure MoO 3-(010). The non-transformed 2-butanol and the products of the reaction were analyzed at the reactor outlet by on-line gas chromatography with a flame ionization detector. The catalytic performances reported here after were obtained when the steady state was reached. The only product of reaction detected was butene. No methyl-ethyl-ketone was observed in the present experiments. The catalytic performances were expressed in terms of conversion of 2-butanol (%C), 184

yield (%Y) and selectivity to butene (%S). To allow the comparison of the performances of MoO3-(010)- with those of M0O2 (which have different specific areas), %C and %Y were reported to the area of the fresh molybdenum oxide in the reactor. These normalized values were called hereafter "intrinsic performances" and noted "%C.m"^" and "%Y.rrf^" respectively. This normalization procedure is valid under the realistic assumption that the reaction order is one and that the conversions are low. The detection and the evaluation of the intensity of a possible synergism between the oxides in the mechanical mixture was made by comparing the observed performances with theoretical values calculated assuming that no cooperation occurred, namely as if the two phases in the mixture behaved independently (as if they were alone in the reactor). Equations 1a, 1b and 1c show how these theoretical values were estimated.

Theoretical conversion of 2-butanol : (1a) %Clh = %C obtained on Sb204 + %C obtained on M0O3

Theoretical yield inbutene : (1b) %Yth = %Y obtained on Sb204 + %Y obtained on MoOa

Theoretical selectivity to butene : (1c) %S* = (%Ylh/%Clh) x 100

2.3. Characterization Most of the samples were characterized before and after catalytic test by X-ray diffraction (XRD), environmental scanning electron microscopy (ESEM) and specific area measurements. XRD was performed in the continuous coupled 8/28 reflection mode on a Kristalloflex Siemens D5000 diffractometer using the Kai ,2 radiation of Cu (X=1,5418 A) for 28 angles varying from 10 deg to 80 deg at a scan rate of 0.4 deg min"1 (step size = 0.04 deg, step time = 6 s). ESEM was performed on a Philips XL30 microscope using a field emission gun and a gaseous secondary electron detector. The observation was made on uncoated samples with 2 to 5 Torr of water in the cell and an acceleration voltage of 15 kV. An X-ray spectrometer EDAX was attached to the microscope. The specific area measurements were made on a Micromeritics ASAP2000 device. The analysis was based on the adsorption isotherms of Kr at 77 K. Specific areas, hereafter noted "SBET", were calculated according to the B.E.T. equation.

3. Results 3.1. Catalytic activity measurements Table 1 summarizes the results of the catalytic tests. The performances obtained on the pure oxides matched those already reported [1]. In particular, the total inactivity of cc-Sb2C>4 is consistent with the observation that it never possesses Bronsted acid sites to perform dehydration reactions [9]. The particularity of these results was that MoOs-(010) did not produce any methyl-ethyl-ketone as was observed elsewhere on other M0O3 samples [1,10]. This difference is explained by the fact that the previous investigations used isotropic M0O3 with crystals developing an important proportion of (100) lateral faces. These are known to be active in redox reactions. On the contrary, (010) faces, which constitute the major 185

part of the surface area exposed by MoO 3-(010), are not able to significantly perform a partial oxidition of hydrocarbons [7,8]. The mechanical mixture of MoO3-(010) with a-Sb 2C>4 exhibited a synergetic behavior : the conversion of 2-butanol and the yield to butene measured were higher than the corresponding theoretical values in the absence of cooperation. The selectivity to butene on the mixture, even if slightly lower, remained in the range of the theoretical one. After 12 hours of catalytic reaction, the color of the Mo Q 3-(010) tested alone had turned from light yellow to dark blue. After 6 hours of reaction, only the half of the catalytic bed at the reactor inlet side had turned to blue. The mixture of Mo 03-(010) and a-Sb204 did not present any significant change of color even after 12 hours of reaction. Table 2 shows the intrinsic performances of MoC> 3-(010) and of M0O2. The intrinsic conversion of 2-butanol and intrinsic yield to butene obtained on M0O2 were lower than those obtained on MoO 3-(010). The selectivities to butene obtained on the two samples were in the same range.

Table 1 - Observed and theoretical values at 220 °C of the conversion of 2-butanol (%C), yield (%Y) and selectivity to butene (%S) - The theoretical values assuming the absence of cooperation are shown in parentheses. Catalyst %C %Y %S MoOs-(OIO) 34 12 35 Mechanical Mixture 57 (34) 18 (12) 32 (35) a-Sb?04 0 0 -

Table 2 - Intrinsic conversion of 2-butanol (%C.m"'2), intrinsic yield to butene- (%Y.m* 2) and selectivity to butene for MoOn-(OfO) and MoO? at 220 °C. Catalyst %C.m"2 %Y.m"2 %S MoO3-(010) 181 64 35 MoO? 116 42 36

3.2. Characterization results 3.2.1. X-ray diffraction In addition to the peaks of the fresh sample, the XRD pattern of the pure Mo O3-(010) reacted during 12 hours presented two new peaks at 36.9 deg and 53.6 deg. These indicated the presence of M0O2 tungarinovite, with the peak at 36.9 deg assigned to the combination of the (200), (111) and (-211) major lines of the JCPDS M0O2 standard and the peak at 53.6 deg assigned to the combination of the (-220), (-312) and (-222) standard lines [5j. The other peaks of M0O2 were not distinguishable because they overlap with the lines of M0O3. Figure 2 shows the XRD spectra of the pure MoO 3-(010) before and after 12 hours of catalytic test. For MoC> 3-(010) reacted during 6 hours, only the blue part of the catalytic bed exhibited the M0O2 peaks. The XRD patterns measured after 12 hours of reaction for the mechanical mixture of MoO 3-(010) with a-Sb204, for the pure a-Sb2C>4 and for M0O2 did not exhibit any modification with respect to those of the corresponding fresh samples. In particular for the reacted mechanical mixture, no additional lines at 36.9 deg nor at 53.6 deg were recorded. 186

26 (deg.) Figure 2. XRD patterns (b) of fresh pure MoC> 3-(010), and (a) after 12 hours of reaction.

3.2.2. Environmental scanning electron microscopy After being tested alone, MoO 3-(010) exhibited, in addition to the big M0O3 crystals, numerous and small "cotton ball-like" particles. These were observed at places where the M0O3 crystals were deeply fragmented (Figure 3). This suggested that the fragmentation and the formation of particles were intimately related phenomena. After 6 hours of reaction, the particles and the fragmentation were only observed for MoO 3-(010) from the blue part of the catalytic bed. When MoOs-(OIO) was reacted in the presence of a-Sb2C>4, neither the "cotton ball-like" particles nor the fragmentation of the M0O3 crystals were observed even after 12 hours of reaction. But the M0O3 crystals exhibited a new morphology that was not observed for the fresh material nor for Mo 03-(010) reacted alone. Instead of having a sharp intersection as in the fresh sample, the edges between the (010) and the (100) faces acquired a facetted structure with (100) steps developed on the (010) faces (Figure 3). Even if apparently similar to the fragmentation observed for Mo 03-(010) reacted alone, the facetting occurred differently : the steps it induced were smoother and more regular than those originating from the fragmentation. Figure 3 gives views of the "cotton ball-like" particles on the M0O3 crystals reacted alone (left), of a fragmented MoOs-(010) crystal reacted alone (middle), and of the facetted edge between the (010) and (100) face of an M0O3 crystal reacted in the presence of a-Sb2C>4 (right).

Figure 3. ESEM views (left) of "cotton ball-like" particles (pointed with arrows) on pure MoOg-fOlO) after 12 h of reaction (bar = 5 pm), (middle) of a fragmented MoOs-(010) crystal after 12 h of reaction alone (bar = 1 pm), and (right) of the facetted (010)-(100) edge (new steps pointed with arrows) of a M0O3 crystal reacted in the presence of a-Sba04 (bar = 1 pm). 187

The "cotton ball-like" particles were identified by X-ray microanalysis as molybdenum oxide. Hence, and because there was a clear parallelism between the presence of particles and the detection of XRD peaks of tungarinovite, the particles were identified as M0O2. A more refined direct characterization of the structure of the "cotton ball-like" particles could not be achieved because of their small size.

3.2.3. Specific area measurements (SBET) Table 3 presents the SBET values obtained before and after 12 hours of reaction. Pure MoO 3-(010) exhibited a marked increase of the SBET value after the reaction. This very likely originates from the formation of the small M0O2 particles. The mixture of MoO 3-(010) and a-Sb204 also exhibited a higher SBET after reaction than before, but the increase was smaller than that for the pure Mo 03-(010). a-Sb204 had nearly the same SBET values before and after reaction.

Catalyst SBET before reaction (m^ q' 1) SBET after reaction (nf q" 1) Mo O3-(010) 0.16 0.30 Mechanical Mixture 0.48 0.67 a-Sb?C>4 1.63 1.31

4. Discussion 4.1. Synergetic effects between separate M0O3 and a-Sb204 The mechanical mixture of MoO 3-(010) with a-Sb204 gave a conversion of 2-butanol and a yield to butene higher than those theoretically calculated assuming the absence of cooperation between the phases. One of the hypotheses classically proposed to explain such a synergism is that the specific areas of the oxide phases increase to a larger extent in the mixture - e.g. as a result of a more intense attrition - than when reacted alone. Our observation was that the SBET increase for the mixture actually was smaller - an increase of only 39 % - than that of pure Mo 03-(010) - an increase of 87 %. Hence, this hypothesis can be discarded. Another classical explanation for the synergetic effects between simple oxides considers that a more active binary compound is formed during the reaction as the result of a mutual contamination between the pure oxides. The objective of this contribution is not to further discuss this hypothesis. The comparison of the characterization of the mechanical mixture before and after reaction led to conclusions identical to those of previous publications dealing with the same catalytic system, or other ones very similar [1,2,6,9-12]. This applies to the comparison of the XRD patterns of the mixture before and after test (after reaction, there was no XRD peak that was not assigned either to pure M0O3 or to pure a-Sb204), of the ESEM views of the mixture before and after test (there was no crystal domain that could not be assigned either to an M0O3 crystal or to an a-Sb204 particle) and also of XPS results to be presented elsewhere [12]. Hence, the conclusion drawn in the previous papers also applies here, namely that there is no mutual contamination occurring between M0O3 and a-Sb 204 during the catalytic reaction. This leads us to attribute the synergetic effect to a cooperation 188

between M0O3 and a-Sb204 phases remaining completely separate and mutually non-contaminated throughout the reaction.

4.2. Formation of M0O2 and deactivation of MoO 3-(010) Mo 03-(010) underwent a deep reduction to M0O2 during the catalytic reaction. The phenomenon was clearly demonstrated by XRD (appearance after reaction of new peaks assigned to tungarinovite) and was confirmed by XPS results to be shown in detail elsewhere [12]. The ESEM picture of MoOs-(010) after test also confirmed that a new type of Mo oxide crystallites was formed. The dehydration of 2-butanol does not involve the removal of an oxygen atom from the catalyst but only requires the presence of Bronsted acid site on its surface. One could thus wonder what could be the origin of the reduction of M0O3. The selectivity to butene only reached 35 %. But 65 % of the converted 2-butanol underwent a total oxidation, leading to the formation of CO, CO2 and H2O. As for the dehydration reaction, the total oxidation of 2-butanol only occurs on M0O3 and not on a-Sb204 (inactive in the 2-butanol reaction). The formation of total oxidation products follows the Mars and van Krevelen mechanism and includes the removal of oxygen atoms from the surface of M0O3. The progressive formation of M0O2 thus indicates the difficulty of pure Mo 03-(010) to get reoxidized by the molecular oxygen from the gas phase during the catalytic reaction. This behavior is well-known for selective oxidation reactions [2]. The reduction of M0O3 reflects the imbalance between the kinetics of its reduction and reoxidation (the reduction occurs faster than the reoxidation), it is therefore logical that the quantity of M0O2 formed in the system increased with time (as observed when comparing MoO 3-(010) reacted during 6 hours and during 12 hours). The comparison of the intrinsic activities of the two Mo oxides demonstrates that M0O2 is less active than M0O3. The reduction of the latter to M0O2 during the catalytic reaction corresponds to a certain tendency of the catalyst to deactivate. It is therefore surprising to observe a steady-state stability (always reached before 6 hours of reaction) of the activity observed for MoOs-(OIO) in spite of the fact that the quantity of M0O3 transforming to M0O2 continuously increases during the test. One would rather expect a continous decrease of activity. This apparent inconsistency finds an explanation in two different observations. First, the fragmentation of M0O3 to M0O2 particles is accompanied by an increase of the active area of the catalyst (as shown by the values of the SBET of MoOs-(OIO) before and after reaction). This increase of SBET would compensate the loss of intrinsic activity associated with the transformation of M0O3 to M0O2. A second explanation of the apparent inconsistency, which could potentially account for a longer term stability of the activity of Mo 03-(010) than reasonably expected in the previous hypothesis, relates to the mechanism of formation of M0O2 from the M0O3 crystals. Even if this is not fully supported experimentally, one can nevertheless imagine that the segregation of the M0O2 crystallites occurs very rapidly after the moment where its nucleation takes place. Particularly, the very rough appearance of the steps on the M0O3 crystals resulting from the detachment of the M0O2 crystallites suggests that the phenomenon is uncontrolled "explosive". XPS results obtained elsewhere strongly support this view [12]. Hence, when a particle of M0O2 detaches from an M0O3 crystal, one can reasonably assume that the new surface of M0O3 left behind is nearly fully oxidized. Its activity should then be similar to that of the fresh catalyst. 189

Under this assumption, a loss of activity would thus only occur when the overall dimensions of the M0O3 crystals start to decrease as a consequence of the progress of the detachment of M0O2. As can be seen on the ESEM pictures, this situation was not yet reached after the 12 hours of test (the size of the M0O3 crystals was in the same range before and after test). The observation of a real decrease of activity would likely require a much longer reaction time.

4.3. Superficial suboxidic state of MoC> 3-(010) in the presence of a-Sb204 For the reacted mixture, we did not observe any fragmentation of the M0O3 crystals or any formation of M0O2 particles. One can then presume that the reduction occurring at the surface of M0O3 during the catalytic reaction in the presence of a-Sb204 is a smoother and more controlled phenomenon than that observed for pure Mo 03-(010). This is confirmed by the XPS results presented in [12]. Summarizing, the reacted mixture of MoO 3-(010) and a-Sb204 exhibited a relatively important amount of Mo 5* at the surface compared to that of Mo 4*, while the MoO 3-(010) reacted alone only exhibited a small proportion of Mo 5* compared to a more important amount of Mo 4*. This shows that, in the mixture with a-St>204, the surface of M0O3 is stabilized in a reduced intermediary M0O3-X suboxidic state. The smooth and regular stepped morphology observed by ESEM for the crystals of M0O3 in the mixture with a-Sb204, which perfectly fits with that typical of suboxides like M018 O52 [13,14], argues in favor of this interpretation.

4.4. Origin of the synergetic effects between M0O3 and a-Sb204" In view of the difference of behavior of MoO 3-(010) when reacted in the presence and in the absence of a-Sb204, one can understand the origin of the synergetic effects detected with the mixture. Two intimately related aspects can be cited. The first aspect is that the reduction of M0O3 to intrinsically less active M0O2 is inhibited in the presence of a-Sb204. The overall performances of Mo 03-(010) are thus higher in mixture with a-Sb204. The second aspect is that in the presence of a-Sb204, the surface of M0O3 is maintained in a state less reduced than M0O2. On the one hand, we have presently no experimental proofs yet that the Mo suboxides exhibit a higher Bronsted acidity which would explain the enhancement of the performances in the dehydration of 2-butanol. But, on the other hand, the Mo suboxides have already been reported to exhibit remarkable catalytic performances in selective oxidation reactions, namely higher than those of fully oxidized M0O3. The enhanced activity in oxidation concerns M018 O52 in the oxidation of isobutene to methacrolein [3,11,15] and M05O14 in the oxidation of methanol to formaldehyde [14]. In these systems, the higher activity of the suboxides originates from their ability to easily undergo the reoxidation step of the Mars and van Krevelen mechanism [3]. In the catalytic conditions used here, this easier reoxidation of the suboxides leads to the absence of over-reduction of the surface of M0O3 in spite of the occurrence of considerable total oxidation. The consequence is that the formation of M0O2 and, thus, the deactivation related to its formation, are avoided. The superficial suboxidic state of M0O3 is only maintained when M0O3 is reacted in the presence of a-Sb204. This special behavior of the surface of M0O3 thus accounts for the synergetic effects it showed in mixture with a-Sb204. 190

4.5. Mechanism of the synergy: role of a-Sb204 The role of a-Sb204 is very likely dictated by a remote control mechanism (ROM) [2]. The model considers that a-Sb204 (the "donor phase") activates 02 from the gas phase to monoatomic spillover oxygen (Oso) species. Oso migrates onto M0O3 (the "acceptor") where it helps the reoxidation of the surface after that the latter was involved (and got reduced) in the total oxidation of 2-butanol. Thanks to this additional supply of oxygen, the M0O3 surface thus avoids to over-reduce to M0O2 and is dynamically stabilized in the suboxidic state.

5. Conclusion During the catalyticoxygen-assisted dehydration of 2-butanol at 220 °C, pure M0O3 undergoes a deep reduction to M0O2. The phenomenon results from the non selective total oxidation of 2-butanol and from the difficulty of M0O3 to reoxidize its surface using O2 from the gas phase. When reacted in the presence of a-Sb204, the reduction of M0O3 to M0O2 is inhibited and some special M0O3 -X suboxidic state is dynamically stabilized at its surface. As M0O2 is intrinsically less active than M0O3 , the consequence is the improvement of the catalytic performances of M0O3 which leads to a synergism between the two phases. As in other cases where mixtures of M0O3 with a-Sb204 were used, the most reasonable explanation is that the influence of a-Sb204 is dictated by the remote control mechanism : spillover oxygen produced by a-Sb204 dynamically assists the reoxidation of M0O3 during the catalytic reaction. The results point to the possibility to benefit from the effects of spillover oxygen and RCM - the reoxidation of the surface and the related creation and regeneration of active sites - in selective oxidation reactions carried out at temperatures as low as 220 °C.

References 1. EM. Gaigneaux, P.E Tsiakaras, D. Heria, L. Ghenne, P. Ruiz and B. Delmon. Catal. Today, 33 (1997) 151. 2. LT. Weng and B. Delmon. Appl. Catal. A, 81 (1992) 141. 3. EM. Gaigneaux, P. Ruiz and B. Delmon. Appl. Surf. Sci., 121/122 (1997) 552. 4. EM. Gaigneaux, J. Naud, P. Ruiz and B. Delmon. Stud. Surf. Sd. Catal., 110 (1997) 185. 5. @ 1996 JCPDS - International Centre for Diffraction Data: a) 05-0508 molybdite standard, b) 11-0694 cervantite standard, c) 32-0671 tungarinovite standard. 6. E.M. Gaigneaux, P. Ruiz and B. Delmon. Stud. Surf. Sd. Catal., 112 (1997) 179. 7. RA. Hernandez and U.S. Ozkan. Ind. Eng. Chem. Res., 29 (1990) 1454. 8. M.R. Smith and U.S. Ozkan. J. Catal., 141 (1993) 124. 9. B. Zhou, T. Machej, P. Ruiz and B. Delmon. J. Catal., 132 (1991) 183. 10. E.M. Gaigneaux, D. Heria, P. Tsiakaras, U. Roland, P. Ruiz and B. Delmon. ACS Symp. Ser., 638 (1996) 330. 11. E. Gaigneaux. Ph.D. thesis, University catholique de Louvain (Belgium), 1997. 12. EM. Gaigneaux, P. Ruiz and B. Delmon. To be submitted. 13. J.C. Volta, O. Bertrand and N. Floquet. J. Chem. Soc., Chem. Common., (1985) 1283. 14. R. Gottschail, G. Mesti, M. Dieterie, U. Wild, N. PfSnder, G. Weinberg, Ch. Unsmeier and R. Schlfigl. Dechema-Jahrestagungen •98, 26-28 Mai 1998, Wiesbaden . Rhein-Main-Hallen (Germany). 15. E.M. Gaigneaux, P. Ruiz and B. Delmon. To be submitted. DE99G2074 DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

1L 111 Ulll lit 1111 IMIILIJII *DEO 12198283* Y. Deng, M. Hunnius, S. Storck, W. F. Maier Max-Planck-lnstitut ftir Kohlenforschung, Kaiser-Wilhelm-Platz 1, D-45470 MQIheim an der Ruhr, Germany

SELECTIVE OXIDATIONS ON VANADIUMOXIDE CONTAINING AMORPHOUS MIXED OXIDES (AMM-V) WITH TERT.-BUTYLHYDROPEROXIDE

1. Introduction In the past stoichiometric oxidation reagents have dominated the manufacture of fine and bulk chemicals. Today environmental economic concerns require new selective catalytic oxidation technologies. Due to their availability, low costand environmentally safe by-products air and peroxides are the most attractive oxidants.

Research in this field focusses on the development and application of metal- substituted molecular sieves, zeolites and aluminophosphates as catalysts for heterogeneous redox reactions.ii] We have been active in the preparation of amorphous mixed oxides with a stable monomodal narrow micropore distribution by an optimized acid catalyzed sol-gel process. This template-free method allows the formation of amorphous microporous oxides based on SiOa, TiOa, ZrO2 and AI2O3 under ambient conditions. (2) Through hydrolysis and cocondensation of metal alkoxides, the materials composition can be varied to a large extend. The implementation of non-hydrolyzable methyl groups by cocondensation with methyltriethoxysilane (MTES) allows to tune the surface polarity of the materials and to design tailor made catalysts that meet the reaction requirements with respect to the polarity of the substrate and reagent.13] The materials described here are further denoted as AMM-AbM0SicD, (amorphous microporous mixed oxides), with b indicating the molar % of the active center A, c the molar % of methyl group containing silicon and D is the porous base oxide. AMM-TixSi have in the past been applied in selective oxidations with organic hydroperoxides or H2O2 and in shape selective hydro- cracking.P. 4-6] More recent results describe the oxidative dimerization of propene with air and AMM-lnxSi catalysts and the selective oxidation of toluene with AMM-MnxSi materials.!?. 8 ]

uThe catalytic oxygen transfer properties of vanadium containing zeolites and ’ vanadium based sol-gel catalysts with hydrogen peroxides are well known.lt&uj The severe problem of vanadium leaching caused by the presence of the by-product water has been addressed.^ To avoid any interference with homogeneously catalyzed reactions, our study focusses on selective oxidations in a moisture-free medium with fert.-butylhydroperoxide. We have investigated the catalytic properties of amorphous microporous materials based on Si02, TIO2, Z1O2 and AI2O3 as matrix material and studied the effects of surface polarity on the oxidation of 1-octene and cyclohexane^

2. Preparation The monolithic amorphous oxides are prepared by an acid catalyzed sol-gel process.

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 192

The molar composition of the Si02 based sols for the preparation of the AMM- VbMeSicSi is given by: x OV(OPh)3: y Si(OEt)4: z MeSi(OEt)3: 3(x+y+z) EtOH : 2(x+y+z) H2C:0.35(x+y+z) HCI The preparation of Ti02 based materials AMM-VbMeSicTi followed the formulation [i3]; x OV(OPh)3: y Ti(OPri)4: z MeSi(OEt)3: 25(x+y+z) EtOH :0.20(x+y+z) HCI To reduce the hydrolysis and condensation rate of the Zr- and Al-alkoxides for the preparation of AMM-VbMeSicZr and AMM-VbMeSicAI based materials, 4-hydroxy-4- methyl-2-pentanone was chosen as complexing agent (CA) in the formulations: x OV(OPri)3: y AI(OsecBu)3: z MeSi(OEt)3:3(y) CA: 65(x+y+z) PriOH : 1.5(x+y+z) H20:0.06(x+y+z) HCI

x OV(OPri)3: y Zr(OPri)4: z MeSi(OEt)3:3(y) CA: 65(x+y+z) PrOH: 1(x+y+z) H20:0.06(x+y+z) HCI After gelation, all materials have been dried at room temperature for 1 week and calcined by heating to 338 K with 0.2 K/min. This temperature was kept constant for 3 h. With a rate of 0.5 K/min the temperature has been increased to 523 K and kept there for another 3 h. Prior to the catalytic application the materials have been ground in a ball mill. The preparation of a mesoporous V-containing MCM-41 material has been described before.lM

3. Characterization Physisorption Specific surface area and pore size distribution were determined by argon adsorption at 87.5 K with an Omnisorp 360. Pore size calculations are based on the Horvath Kawazoe model for micropores and the BJH formalism for mesopores.MS] Table 1 summarizes the results. For the materials with Ti-, Zr- and Al-oxide as base material the specific surface areas are significantly lower than for the Si-oxide based materials. The catalysts also differ in their pore structure. Pressure transitions in a plot adsorbed volume vs. log p/p 0 that indicate a narrow micropore distribution can only be observed for the Si02 based materials.116]

Table 1 Physisorption experiments Catalyst Specific Surface Cumulative Pore Max. Pore Mean Pore Area (m2/g) volume (ml/g) Diameter (nm) Diameter (nm) VS-1 462 0.22 0.62 0.70 VIMCM-41 673 0.93 2.90 5.38 AMM-V5Si 495 0.22 0.63 0.99

AMM-VsTi 169 0.11 0.58 1.65

AMM-VsZr 46 0.04 0.59 2.57 AMM-VSAI 276 0.22 0.65 2.87 AMM-VgMoSigoSiOl) 675 0.63 3.17 4.51 AMM-Vg^SiggSi (I) 683 0.35 0.66 1.23 193

Hydrophobicity Index To determine the influence of non-hydrolyzable methyl groups on the surface polarity the hydrophobicity index HI was measured. This index is determined by competitive adsorption of octane and water on the materiaLM Materials with a HI smaller 1 are hydrophilic, with a HI larger 1 hydrophobic. AMM-VbMeSicD catalysts were prepared with different amounts of non-hydrolizable methyl groups. It can clearly be seen in Table 2 that with an increasing amount of methyl groups the hydrophilic character changes to hydrophobic. This trend is in agreement with previous results of AMM-catalysts with titanium as active metal. With an increasing amount of methyl groups the BET-surface increases from 500 m2/gto 780 mz/g for the AMM-V5MeSix Si catalysts. Similar results were obtained for the AMM-VsMeSixTi materials. With an increasing amount of methyl groups the hydrophobicity index increases. Since the AMM-VgTi shows a slightly higher HI than the AMM-VgSi, TiOa seem to be more hydrophobic than SiOa.

Table 2 Hydrophobicity index of AMM-VsMeSicSi/n materials Catalyst Hydrophobicity Index (HI) AMM-V5Si 0.20 AMM-Vg^Si^Si 0.43 AMM-VgMeSigoSi 0.60 AMM-V5MeSi50Si 2.10 AMM-VsTi 0.49 AMM-Vg^SigoTi 0.55 AMM-VgMeSigoTi 0.60 AMM-V5MsSi40Ti 1.99

4. Epoxidation of 1-Octene The epoxidation of olefins is an attractive reaction for the selective oxyfunctionalization of organic substrates. This heterogeneously catalyzed reaction has been studied in detail with hydroperoxides.fia-ao] The reaction is highly dependent on the catalysts metal center, its oxidation state and Lewis acidity. High oxidation states and high Lewis acidity seem to lead to the best catalysts, commonly based on metal ions incorporated in the matrix or surface of porous silicas. Autoretardation of this reaction due to pore and active site blocking caused by by ­ product water or alcohol is a common problem (product inhibition). Little is known about the effect of alternative matrix materials. To study the influence of different matrix materials on the active vanadium center and the effect of a more hydrophobic catalyst surface, 1-octene was chosen as substrate for the epoxidation with fert.-butylhydroperoxide. Figure 1 shows the reaction profile of the SiOa based materials in this reaction. The modification with methyl groups slows down product inhibition by the fert.-butanol and leads to higher octene conversions. i 194

AMM-V5Si

Time (h) Figure 1 Reaction profile (with respect to TBHP) of 1-octene epoxidation on AMM-VsMaSixSi catalysts and a crystalline VS-1 reference.

The selectivites to the 1-octene oxide are between 85 and 93 % for the AMM catalysts as indicated in Table 3. A crystalline VS-1 reference with a vanadium content of 1.2% has been tested for comparison. Lower conversions and selectivities have been observed.

Table 3 1-Octene epoxidation on AMM-VsM®SixSi catalysts and a crystalline VS-1 reference. Catalyst Reaction time Conversion (%) Selectivity (%) TOP (mmol olefin (min) / (mol V s) VS-1 300 14.7 65.0 13.0 AMM-VgSi 300 33.5 84.6 7.2 AMM-V^Si-joSi 300 36.0 92.7 7.8 AMM-V^Si-joSi 300 41.1 90.1 8.9 AMM-VgMeSisoSi 300 45.0 85.1 9.7 Reaction conditions: 15.8 mmol 1-octene, 3 mmol TBHP (3m in iso-octane), T= 353 K, catalyst: 50 mg.

An extended drying period during the preparation of the AMM-V3MeSi50Si resulted in a mesoporous material. The influence of the pore size on the reaction is demonstrated in Figure 2. The microporous catalyst (AMM-VgMeSisoSi (I), 0.66 nm diameter) is clearly less active in the epoxidation than the mesoporous material (AMM-VaMeSigoSi (II), 3.17 nm diameter) indicating diffusion limitation. No difference in the selectivities to 1-octene oxide were observed for both materials (78% after 5h) 195

AMM-V3Si 50% Methyl (I) AMM-V3SI50% Methyl (II)

V------1------1------1------1------0 5 10 15 20 25 Time (h) Figure 2 Reaction profile (with respect to TBHP) of 1-octene epoxidation on AMM-Vg^SisoSi catalysts; (I) material with pore diameter of 0.66 nm, (II) material with pore diameter of 3.17 nm.

The activities ofTK Zr- and Al-oxide based AMM-V5D catalysts are significantly lower than the SiOa based materials and do not exceed conversions of 30% in 20h. For the AMM-VsMeSixTi catalysts, methylation of the matrix has a reverse effect on the catalytic behavior. Figure 3 shows that the activity is lower for materials with higher HI. This is unusual, but it may be due to the formation of acid sites in the SiOa/TiOa mixed oxide that can strongly interfere in epoxidation reactions.#]

Table 4 1-Octene epoxidation on AMM-VsMeSixTi. Catalyst Reaction time Conversion (%) Selectivity (%) TOP (mmol olefin (min) /(mol Vs) AMM-VsTi 300 16.6 38.0 4.7 AMM-V5MeSi10Ti 300 10.7 42.9 3.0 AMM-Vg^SizgTi 300 8.78 33.9 2.5 AMM-Vg^Si^Ti 300 8.49 44.9 2.4 Reaction conditions: 15.8 mmol 1-octene, 3 mmol TBHP (3m in iso-octane), T= 353 K, catalyst: 50 mg. 196

AMM-V5Ti 20% Methyl AMM-V5TI40% Methyl

T I I I 0 5 10 15 20 Time (h) Figure 3 Reaction profile (with respect to TBHP) of 1-octene epoxidation on AMM-V5MeSixTi catalysts.

AMM-V5AI

AMM-V5AI40% Methyl

Time (h) Figure 4 Reaction profile (with respect to TBHP) of 1-octene epoxidation on AMM-V5MesixAI catalysts. 197

Figure 4 depicts that the concept of material modification to enhance the surface hydrophobicity can be extended to other matrix materials as well. In the case of the AMM-VsMeSixAI catalysts higher conversions due to less product inhibition are found with increasing methyl group content. The selectivities to the epoxide reached are between 75 and 63% as shown in Table 5.

Table 5 1-0ctene epoxidation on AMM-V5MeSixAI catalysts. Catalyst Reaction time Conversion (%) Selectivity (%) TOP (mmol olefin (min) /(mol Vs) AMM-VSA! 300 3.37 63.5 0.6 AMM-V5M6Si2oAI 300 5.75 74.8 1.1 AMM-V5MeSi4oAI 300 9.03 72.8 1.7 AMM-V5MaSi60AI 300 15.1 70.3 2.8 Reaction conditions: 15.8 mmol 1-octene, 3 mmol TBHP (3m in iso-octane), T= 353 K, catalyst: 50 mg.

With AMM-VsMeSixZr catalysts no clear correlation between suface polarity and activity in the 1-octene epoxidation could be established. Conversions reached 20% and the selectivites to the epoxide did not exceed 20%.

5. Oxidation of Cyclohexane The two oxidation products of cyclohexane, cyclohexanol and cyclohexanone, are important precursors for many bulk chemicals including nylon. Cyclohexanol/ cyclohexanone are produced in homogeneously catalyzed reactions with chromium or cobalt salts under pressure.#?] Since there are no heterogeneous processes yet, we have studied the selective oxidation of cyclohexane with TBHP with selected AMM-catalysts.[22] To investigate the effect of pore size on actvitiy and selectivity two special catalysts, a microporous (AMM-V5MeSi5oSi) and a mesoporous (VlMCM-41 (4%)), have been prepared. The catalysts had similar specific surface areas, narrow pores size distributions, surface polarity, active centers and identical oxidation states of the active metal, and differed only in pore diameter.M In the following the catalytic behavior of these mesoporous and microporous catalysts are compared. The catalysts were used as fine powders (particle sizes 1-2 pm) to avoid particle size effects. The major difference in catalytic performance is the initial rate of the reaction shown by turnover frequencies (mmol alkane/(mol V s)) of 1.7 for the AMM-VsMeSisoSi and of 3.8 for the VIMCM-41 (4%) after 300 min reaction time. This difference can be attributed to higher diffusion barriers in the microporous relative to the mesoporous catalyst. Assuming that the active centers in the two materials are of similar concentration and activity, this also confirms, that the diffusion limitation is more severe in the microporous catalyst.

The cyclohexane oxidation with TBHP on vanadium containing AMM-catalysts shows a high selectivity to the desired products cyclohexanol and cyclohexanone. The changes in surface polarity have apparently no effect on the selectivity or activity as summarized in Figure 6. In all cases the conversion of TBHP is nearly complete. 198

Figure 6 Selectivity of the cyclohexane oxidation with TBHP depending on the amount of methyl groups. Conditions: 343 K, cyclohexane : TBHP = 6:1, max. Conversion 17%.

amount of non-hydrolysable methylgroups x Further experiments were carried put to investigate the influence of the matrix oxide on the selectivity of the reaction. Figure 7 shows the catalytic results. Even though the conversion of cyclohexane is rather poor with these other matrix oxides (<1% compared to 10 % with the AMM-VsSi-materials), the detection of activity is remarkable, since these oxides are hexagonal coordinated and not tetrahedral. Despite the low conversion the selectivity is poor with all matrices except silica. The relative amount of by-products increases from the matrix silica via titania and zirconia to alumina, where by-products already dominate the reaction. However, the potential of these new redox-catalysts is not known and further studies have to be conducted.

Figure 7 Selectivity of vanadium containing catalysts with different base oxides. Conditions: 343 K, cyclohexane : TBHP = 6:1, max. Conversion 17%.

□ cyclohexanol □ cyclohexanon M by-product

AMM-VsSi AMM-VJi AMM-V5Zr AMM-V5AI 199

6. Conclusions We have prepared isolated vanadium centers in different microporous/mesoporous oxide materials based on SiOg, TiOg, ZrOg and AI2O3. By copolymerization with MeSi(OEt)a the surface polarity of all materials could be adjusted in the one-step preparation procedure. These new AMM-VbMeSicD materials show promising activities and selectivities as redox catalysts for the epoxidation of 1-octene and in the oxidation of cyclohexane. Although the TiOa, Z1O2 and AI2O3 based V-mixed oxides showed remarkable activities and selectivities, the AMM-Vt>MeSicSi materials were the best catalysts for both reactions.

7. Literature [1] I.W.C.E. Arends, R.A. Sheldon, M. Wallau, U. Schuchardt, Angew. Chem.,109, 1191, (1997). [2] W.F. Maler, I.-C. Tilgner, M. Wisdom, H.-C. Ko, Adv. Mater., 5.726, (1993). [3] S. Klein, W. F. Maler, Angew. Chem., 108,2376, (1996). [4] S. Klein, J.A. Martens, R. Parton, K. Vercruysse, P.A. Jacobs, W.F. Maler, Catal. Lett., 38.209, (1996). [5] S. Klein, S. Thorimbert, W.F. Maier, J. Catal., 163,476, (1996). [6] W.F. Maier, JA. Martens, S. Klein, J. Heilmann, R. Parton, K. Vercruysse, PA. Jacobs, Angew. Chem., 108,222, (1996). [7] F. Konietzni, U. Kolb, U. Dingerdissen, W.F. Maier, J. Catal., 176,527, (1989). [8] S. Bukeikhanova, H. Orzesek, U. Kolb, K. KOhlein, W.F. Maier, Catal. Lett., 50,93, (1998). [9] P. Kumar, R. Kumar, B. Pandey, Synlett, 4,289, (1995). [10] R. Neumann, M. Chava, M. Levin, J. Chem. Soc., Chem. Commun., 1685, (1993). [11] R. Neumann, M. Levin-Elad, Appl. Catal. A, 122,85, (1995). [12] B.I. Whittington, J.R. Anderson, J. Phys. Chem., 97,1032, (1993). [13] C. Lange, S. Storck, B. Tesche, W.F. Maier, J. Catal., 175,280, (1998). [14] R.D. Oldroyd, G. Sankar, J.M. Thomas, M. Hunnius, W.F. Maier, JL Chem. Soc., Faraday Trans., submitted, (1998). [15] J. Seifert, G. Emig, Chem. Ing. Tech., 59.475 (1987). [16] S. Storck, H. Bretinger, W.F. Maier, Appl. Catal. A, in press (1998). [17] C.H. Berke, A. Kiss, P. Kleinschmitt, J. Weitkamp, Chem. Ing. Tech., 63,623, (1991). [18] R.A. Sheldon, JA. Van Doom, C.W.A. Schram, A.J. De Jong, J. Catal., 31,438, (1973). [19] R.A. Sheldon, JA. Van Doom, J. Catal., 31,427, (1973). [20] R.A. Sheldon, J. Mol. Catal., 7,107, (1980). [21] K. Weissermel, H.J. Arpe, Industrielle Organische Chemie, VCH, Weinheim 1994. [22] M.G. Clerici, G. Bellusi, U. Romano, J. Catal., 129,159, (1991). 200 201

DGMK-Conference "Selective Oxidations in Petrochemistry ”, Hamburg 1998

B, Kubias*, U. Rodemerck*, F. Ritschl**, M. Meisel** *DE012198292* * Institut fur Angewandte Chemie Berlin Adlershof e.V., ’ J Rudower Chaussee 5, D-12484 Berlin, Germany ** Institut fOr Chemie, Humboldt-Universitat zu Berlin, Berlin, Germany

ON THE CATALYTIC GAS PHASE OXIDATION OF BUTADIENE TO FURAN

Abstract i Applying the thermochemical selectivity criterion of Hodnett et al. [1] it is shown that ^ the selectivity of the furan formation is not limited by a too low strength of the C-H

bonds in furan when compared with the C-H bond dissociation energy in the educt molecule butadiene. In the oxidation of butadiene on a CSH2PM012O40 catalyst a maximum yield of 22 mol % furan has been obtained. To improve this comparatively low furan yield the oxidation activity of the catalyst must be lowered to prevent the consecutive reaction to maleic anhydridoj

DE99G2073 Introduction Butadiene is a cheap and available feedstock for producing synthetic rubbers and polymer resins [2], Its application in valuable oxygenated specialities such as tetrahy- drofuran or butanediole, however, is limited and is not reached by catalytic gas phase oxidation until now. Both products could be obtained via furan provided that furan is accessible in high selectivity by the oxidation of butadiene. Such a route would be an attractive one due to the lower cost of the hydrogenation of furan in comparison with the hydrogenation of maleic anhydride (MA) being also an alternative source for tet- rahydrofuran and butanediol production, respectively. However, up to now it has been impossible to obtain furan in high yield by partial oxi­ dation of butadiene due to the formation of MA in a consecutive reaction step. A furan yield of about 30 % on a (VO^PzOy catalyst as published in [3] has proven to be non- reproducible [4], A selectivity of 94 % at 14 % butadiene conversion reported for an Ag-Mo oxide catalyst has turned out to be not stable under reaction conditions [5], Furan and 2,5-dinydrofuran can also be formed by the reaction of atomically adsorbed oxygen with coadsorbed butadiene on Ag(110) as shown by Madix et al. [6]. This re-

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 202

action type, however, is not commercially relevant. On the promoted industrial Ag catalyst tne catalytic oxidation of butadiene proceeds exclusively to 3,4-epoxy-i- Dutene ana not to 2,5-dihydrofuran and furan, respectively [7], Thus, the question is whether tne selectivity of the furan formation might De limited tor reasons concerning the reaction mechanism (e.g. parallel reaction path from butadi­ ene to MA) or by thermochemical properties of the molecules involved. A limitation by the mechanism seems to be unlikely as we have shown by a comparative study of me oxidation of butadiene, dihydrofuran and furan (4). To answer the question whether too low bond strengths in furan could be responsible for the limited furan selectivity the thermochemical selectivity criterion (TSC) of Hodnett et al. [1] can be applied. Hodnett et al. analyzed the selectivity/conversion data of more than 20 heterogene ­ ously catalyzed selective oxidation reactions and correlated the selectivities at given conversion degrees with the differences of the dissociation enthalpies of selected bonds in tne products and the reactants. They found that these reactions are highly selective at higher conversion degrees only men if the differences between tne small­ est C-H bond enthalpy in the reactant and the smallest C-H or C-C bond enthalpies in the product are lower than 30 - 40 kJ/mol (0 - 50 % conversion). If the difference is larger man 70 kJ/mol usually the reactions are completely unselective at a conversion degree higher than 10 %. The aim of this work was to study the oxidation of butadiene to furan and to the less desired product of the consecutive oxidation, maleic anhydnde, applying this criterion and to evaluate the catalytic potential of promising oxidic catalysts such as heteropoly compounds and typical catalysts for the allylic oxidation as descnbed in [9] and (10).

Experimental Preparation of catalysts The heteropoly compounds were prepared as follows: supported Cs^tPMoizOw] (on pumice, HPA 1, table 2) was obtained according to Ai (9]. Unsupported heteropoly molybdates and their salts (HPA 2 - HPA 5) were prepared by analogy with the method descnbed in [11]. In the case of the HPA 6 additionally appropnate amounts of V2Os besides M0O3 were dissolved. HPA 7-10 were prepared according to [12]. In the case of a Bi-Mo oxide catalyst a typical catalyst for tne ammoxidation of propene similar to that described in [10] was used. U-Sb oxides were prepared according to [13] 203

Catalytic test Reactor and analytical procedure are descnoed in [4]. The catalytic tests were per­ formed as follows - over the catalyst bed (2-10 cmj. diameter of the granules 1 25- 2,5 mm) feeds containing butadiene and oxygen in a ratio of i ; 3-5 and different amounts of water vapour were passed. Mostly, the concentrations of the products were determined immediately after the start of the reaction by on line gas chromato ­ graphy to get information on the initial behaviour and on tne conditioning procedure of the catalysts. In the case of tne heteropoly compounds the reaction temperatures were increased step by step up to their decomposition temperatures and above it in order to study the catalytic properties of the decomposition products, too.

Results and discussions Application of the thermochemical selectivity criterion (TSC) In order to investigate whether the furan selectivity is limited by thermochemical prop ­ erties of the molecules involved the following working hypothesis was denved accord ­ ing to Hodnett et al. [1]: the selectivity of furan (or MA) is then limited if tne difference of the bond dissociation enthalpies

AD"H = D'Hc-HIbutadiene) * D"Hc-H.C-C or C-O (furan. MAI is larger than 30 - 40 kJ/mol. In this working hypothesis also C-O bonds (in furan and MA) are included which were not considered by Hodnett et al.. The necessary D"H values and also the heat of formation values, AH",. of the rele­ vant molecules and their radicals are scarcely found in literature and, if present, show often large differences. Therefore, we calculated AH", by quantum chemical methods and determined D"H values assuming radical bond dissociations: D"H(X-Y) = AH-,(X-) + AH“,(Y') - AH", (X-Y) (X*, Y*: radicals of the compound X-Y) From the applied semiempirical quantum chemical methods MNDO, MIND03, AM1, and PM3 the MNDO method has given the best AH", values for butadiene in compari ­ son to literature data. However, it is known [14] that MNDO fails to calculate AH", of small radicals. Therefore, we finally used the program DMOL (15) which applies the density functional theory (DFT). This program delivers the total energy , Et. which can be transformed into AH", by the help of atom equivalents. 204

The calculated differences A(D”H) of the bond dissociation enthalpies of the weakest C-H bonds in butadiene, furan, and MA are summarized in the top half of table 1.

Table 1: Differences A(D'H)/ kJ/ mol for the relevant educt-product couples Upper part: A = D"H(educt, weakest C-H) - D“H( product, weakest C-H) Lower part: A = D"H(educt, weakest C-H) - ring fission enthalpy (product ) Ring fission enthalpy = l(Et/ AH”f of all fission products) - Et/AHuf(ring)

I educt I product | A(D°H) /kJ/mol (DFT method) I butadiene I furan I -39.6 j butadiene [MA j -30.2 I furan I iviA | +9.6 I [ ring fission enthalpy (product) A(D°H) | | butadiene | furan I+66.6 (Liu[16]) +92.6 butadiene ]ma j +S3 to +143 +131 to +357 I furan |MA I+88 to +148 +221 to +407 i

As can be seen, none of the a (D'H) values is larger than +30 kJ/moi. Thus, all of the cited reactions from the thermochemical point of view should proceed with high product seiectivities at high conversion degrees of the reactants. This is in agreement with experimental results in the case of the oxidation of butadiene and turan to MA where high seiectivities at high conversion degrees have been found. From the above results it may be also concluded that the selectivity of the butadiene conversion to furan should not be limited for thermochemical reasons. In tne lower half of table 1 ring fission enthalpies of furan and MA are used for the estimation of A(D“H). According to Hodnett et al. (1), TSC rests only on gas phase reaction enthalpies. Consequently, the enthalpies of the gas phase fission reactions of furan and MA, resp., must be introduced into the cntenon. The only known infor­ mation about the fission of the furan ring in the gas phase stem from shock wave tube expenments (see [16]). This fission proceeds via a non-radical, peri cyclic mecnanism with a very low fission reaction enthalpy of about 67 kJ/mol [16] (radical C-H bond dissociation enthalpies are in an order of magnitude of about 400 kJ/moll). From thus low fission enthalpy a A(D"H) of 92,6 kJ/mol follows indicating a strong restriction on turan formation from butadiene. The same picture was obtained by us in some tenta- 205

tive DFT simulations of the fission of MA (of. table 1). These limitations, however, can only be apparent ones due to the highly selective formation of MA from butadiene via furan as the reaction intermediate. Therefore, and for another reason (see para­ graph .conclusion") nng fission enthalpies basing on non-radical mechanisms are not suitable to obtain information from TSC while C-H bond dissociation enthalpies de­ rived from a radical bond dissociation lead to reasonable results.

Oxidation of butadiene on heteropoly compounds and catalysts for allylic oxi­ dation To evaluate the catalytic potential of heteropoly compounds different molybdato phosphates and their caesium salts (HPA 1-6) and, additionally, some other hetero ­ poly acids with different central ions (HPA 7-10) were tested. Furthermore, Bi-Mo and U-Sb oxides were used as catalysts. Figure 1 shows furan yield and selectivity, respectively, vs. butadiene conversion plots of the catalyst HPA 2 which are typical for all HPA catalysts tested. The maxi­ mum furan yields are obtained at butadiene conversion degrees between 60 and 80 %. At 100 % butadiene conversion the furan yield decreases to zero and MA is formed as the product of the consecutive reaction in a yield of 30 mol%.

x/% x/% Fig. 1: Oxidation of butadiene on HPA 2: furan (■) and MA (□) yields and selectivi- ties, respectively, versus butadiene conversion, CC4H6 = 0.9 %, 02: C4H6 : H20 = 4.5 :1:16

The main side products besides the total oxidation products were acrolein, acrylic acid and aromatic compounds (BTX). Carbon balances have turned out to be always less than 100 % showing that some products could not be detected. Brown deposits 206

were found at the reactor outlet which led us to the conclusion that cycloaddition re­ actions forming mgher molecular products had occurred. Indeed, one of tnese com ­ pounds, phthalic anhydride, was identified. It was astonishing that the maximum se­ lectivity for furan has not been found at low but at medium conversion degrees of butadiene. The reason for this may be competitive reactions (Diels-Alder condensa ­ tion) which could more influence the selectivity at low conversion (high butadiene concentration). Indeed, relatively high yields of aromatic compounds were observed at low conversion degrees of butadiene.

In table 2 the maximum furan yields at the respective butadiene conversions and the reaction conditions are summanzed. Additionally, the MA yields obtained under the conditions of tne maximum furan yields are listed.

Table 2: Maximum yields or furan (F) and MA obtained from butadiene oxidation on heteropoly compounds and catalysts for allylic oxidation, m„. = 1.5 - 15 g 02: CcHe = 3.5 . GHSV = 300 - 6000 h"1 (STP)

j T/K catalyst | ypmax /% yMA'"/o 2) | Xc4H6^°o CH2cV°o | I I I HPA 1 CsH2[Fivio, 204oj :'' I 5 I DO | 600 13 | I HPA 2 H3[PiviOi204oj I 19 19 I 11 | 670 14 I | HPA 3 CsH2[PMoi 204o] I 22 18 1 77 | 650 14 I | HPA 4 Cs2H[PM o 1204o ] | 18 16 1 75 | 670 14 | I HPA 5 Cs3(PMO|204o] I 12 8 I 60 1 670 14 I I HPA 6 CsH 3[PVMOii04o ] I 13 10 1 65 1 670 14 I IHPA7 H 3[PWi204o ] I 10 5 1 65 1 670 14 I I HPA 8 H^SiMoiaOin] I 7 9 i 50 I 670 14 i I HPA 9 H4[SiW12O40] I 8 5 ! 60 1 690 14 I I HPA10 Ks[aSiMoWoCr?rH 20)?0-,?,] I 6 1 i 45 ! 790 14 I ] BiFesCOj.ENh.sPo.EMOtoSiicOx I 6 10 1 35 I 620 78 I USbOs/USb.O.o I 4 <"1 1 40 1 750 o l ; USb^Oto | 5 1 ! 45 ! 790 0 | ! 20 I 780 ! USkO 0 ! 8 0 83 I on pumice • oetermmea at me same conversion degrees wnere maximum furan yielas were ootamea On HPA 1 the same furan yield was found as described by Ai (7). On the pure mono ­ caesium salt of the molybdato phosphoric acid (HPA 3) the maximum furan yield amounts to 22 mol %. This is the highest furan yield described up to now considering the open literature and the fact that the results in [3) are non-reproducible [4], At the 207 reaction conditions mentioned 18 mol % of MA as the second selective oxidation product have been formed. None of the other heteropoly compounds was able to produce higher furan yields tnan 20 mol % indicating that the catalytic potential of this catalyst type in the conversion of butadiene to furan is limited. A variation of the redox potential of the feed (02:C4H6 ratio) and of the concentration of water vapour did not result in a further increase in furan yield and selectivity either. On the catalysts usually used for allylic oxidation only less than 10 mol % furan were obtained in contrast to the results descnbed in [10]. The presence of a large concen ­ tration of water vapour led to the formation of up to 10 mol % MA, obviously favonng the desorption of this product from the surface of the multimetal molybdate catalyst. In contrast to this, on U-Sb oxides the formation of MA is fully suppressed by the addi­ tion of large amounts of water vapour. It is interesting to note that under tnese condi ­ tions a comparatively high furan selectivity of 40 % can be observed at a 20 % con ­ version of butadiene.

Conclusions From the difference of the calculated lowest C-H bond dissociation enthalpies in bu­ tadiene and furan being less than 30 kJ/mol it is concluded that furan selectivity is not limited by the thermochemical properties of both molecules according to the cnterion of Hoonett et al. [1], From this result it follows that a further search for more suitable catalysts for the partial oxidation of butadiene makes sense. The other bonds in the cyclic compound furan should only then be considered if, as in the case of C-H bonds, a correlation between the thermochemical and the activation entnalpies of these bonds would exist on the basis of similar mechanisms of bond fissions. Obvi­ ously, in the case of a ring fission reaction in the gas phase obeying a non-radical, pencyclic mechanism the fission reaction mentioned does not meet this demand. On heteropoly compounds the maximum yield of furan amounts to 22 mol % at a con ­ version degree of butadiene of about 80 %. Efforts to lower the oxidation activity of these catalysts by changing the composition of the gas phase failed as in the case of VPO catalysts (see [4]). Neither the variation of the number of cations nor the type of the hetero atoms in the heteropoly compounds tested led to an increase in furan yield. From tne results obtained with the catalysts for allylic oxidation it is concluded that these oxides are less suitable catalysts for the oxidation of a substrate with exciu- 208

sively vinylic C-H bonds. The medium yields and selectivities of the furan formation ootained with Doth catalyst types may be explained by a too high redox activity of tne chosen oxides. Obviously, improved catalysts should have reduced oxidation ability combined with optimum acid-base properties.

Acknowledgements The authors are gratefully to D. Fratzky for the preparation of the U-Sb oxide cata­ lysts.

Literature (1) C. Batiot, F.E. Cassidy, A.M. Doyle, B.K. Hodnett, Stud. Surf. Sci. Catal., 110, 1909, (1997) C. Batiot, B.K. Hodnett, Appl. Catal. A, 137,179, (1996) F.E. Cassidy, C. Batiot, B.K. Hodnett, Spec. Publ.-R. Soc. Chem., 216, 237, (1998) F. E. Cassidy, B.K. Hodnett, Erdol, Ergas, Kohle, 114, 256, (1998) (2) M.L. Morgan, Chemistry & Industry, 1998, 90 (3) G. Centi, F. Trifiro, J. Mol. Catal., 35, 255, (1986) (4) U. Rodemerck, B. Kubias, M. Meisel, D. Fratzky, A. Krepel, DGMK- Tagungsbericht 9705, 297,(1997) (5) M. Wildberger, Dissertation ETH Zurich, 1998 (6) J.T. Roberts, A.J. Capote, R.J. Madix, J. Am. Chem. Soc., 113, 9849, (1991) (7) J.R. Monmer, Stud. Surf. Sci. Catal., 110,135, (1997) (8 ) B. Kubias, U. Rodemerck, M. Meisel, S. Ridi, H. Hibst, 31. Jahrestreffen deut- scher Katalytiker , Leipzig, 1998, Tagungsband S.209 (9) M. Ai, J. Catal., 67, 110, (1981) (10) N.G. Glukhovskii, I.L. Belostotskaya, L.S. Vorobeva, Neftechimia, 26, 89, (1986) (11) S. Berndt, D. Herein, F. Zemlin, E. Beckmann, G. Weinberg, J. SchUtze G. Mestl, R. Schlogl, Ber. Bunsenges., 102, 763, (1998) (12) M. Meisel, S. Ridi, unpublished results (13) R.K. Grasselli, D.D. Suresh, J.Catal., 25, 273, (1972) (14) J.P.P. Stewart, Manual of the MOPAC'93 Code, Fujitsu, Tokyo, 1993 (15) Program DMOL, Version 950, from Byosym Technologies, Inc., 1995 and Ver­ sion 960 from Molecular Simulation MSI, Inc. 1996 (16) R. Liu, X. Zhan, L. Zhei, J. Comput. Chem., 19, 240, (1998) 209

DGMK-Conference inrq 1998

DE99G2072 F. Loeker, W. Leitner Max-Planck-lnstitut fur Kohlenforschung, P.O. Box 10 13 53, D-45466 MOlheim an der Ruhr, Germany

SUPERCRITICAL CARBON DIOXIDE AS AN INNOVATIVE REACTION MEDIUM FOR SELECTIVE OXIDATION

1. Basic properties of supercritical carbon dioxide as a solvent

Supercritical carbon dioxide (scCOJ is of great interest as a new reaction medium for i chemical processes catalyzed by transition metal compounds' 11. In contrast to many common i i organic solvents, scC02 is not toxic, not flammable and its use is not associated with any immediate environmental hazards. Further advantages arise from the "gas-like" properties of I

the supercritical fluid, i. e. high miscibility with reactant gases, absence of a liquid/gas phase ! boundary, high diffusitivity/low viscosity and high compressibility. Beyond its critical point

(Tc = 31.0°C, p c = 73.8 bar, Figure 1) the solvent properties of compressed C02 can be regulated considerably by varying the reaction parameters pressure and temperature.

^supercritical // y region y

pe = 0.467 gem"

temperature

Figure 1: Simplified phase-diagram of carbon dioxide.

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 210

The use of scC02 for extraction processes is technically well established in food industry. For example the destraction of green coffee beans using scC02 technology was developed in the

Max-Planck-Institute for Coal Research in Miilheim in the late sixties'21 and is nowadays used worldwide for the production of approximately 100 000 t decaffeinated coffee per year.

Similar technology could provide the opportunity to seperate selectively the products or catalysts from a reaction mixture, thus offering a potential solution to one of the key problems of homogeneous catalysis.

2. Metal-catalyzed reactions in supercritical carbon dioxide

Our group at Miilheim has an ongoing interest in the use of scCO, as a solvent for metal catalyzed reactions. One of the major obstacles to a broad use of scC02 in this area has long been the limited solubility of many transition metal catalysts. In particular, complexes bearing aryl substituted phosphorous donor ligands, which have proven extremely useful in homogeneous catalysis, are in general poorly soluble in scC02. For this reason our group developed a generally applicable methodology to increase the solubility of arylphosphorous ligands and their metal complexes in scC02. Fixation of appropriate solubilisers (i. e. long perfluoroalkyl chains) at the aryl moiety keeps structural and electronic changes at the active center to a minimum compared to the unsubstituted parent complexes' 31. The synthetic utility of this new type of arylphosphorous ligands was demonstrated in the rhodium catalyzed hydrofbrmylation of 1-octene: the corresponding C,-aldehydes were formed with an n/iso- ratio of 4.8 in nearly quantitative yields' 31.

Furthermore we investigated the transition metal catalyzed olefin metathesis in collaboration with the group of A. Furstner1* 1. We found that compressed and especially supercritical C02 serves efficiently as a substitute solvent for all kinds of metathesis processes, in particular for 211

the ring opening metathesis polymerisation (ROMP) of cycloolefins like norbomene and for the ring closing metathesis (ROM) of diens to macrocyclic compounds. The extractive properties of the supercritical fluid allow to isolate the pure and solvent free products, which have important applications as fragrances and olfactory compounds; the catalyst is recovered in active form.

Most remarkably, we found that the product distribution during RCM could be controlled by simply adjusting the density of scC02. The desired ring closing products were produced selectively at densities higher than 0.65 g/cm 3 while lower densities led to oligomers via acyclic diene metathesis (ADMET). This unprecedented way of mimicking "high dilution conditions" might be of general use for the synthesis of macrocyclic compounds.

Moreover, the use of scC02 as a reaction medium allows the use of carbene complexes

[(R3P)2C12Ru =C(H)R*)] as catalysts for the metathesis of substrates with unprotected N-H functionalities. The deactivation of the ruthenium carbenes, which is normally observed with these type of substrates, is probably suppressed by C02 acting as a protecting group through reversible formation of the corresponding carbamic acids'41.

These examples demonstrate nicely the unique features of scC02 which can be exploited to result in beneficial effects on metal catalyzed reactions and processes.

3. Oxidation reactions in supercritical carbon dioxide

3.1' Oxidation with peroxides as oxidants

Oxidation reactions are among the most important processes to introduce functional groups into hydrocarbons and many of these reactions are catalyzed by transition metal compounds.

Despite the obvious advantages offered by scC02, only very few attempts to achieve synthetically useful oxidation reactions in this medium have been reported. Tumas et al. 212

described the epoxidation of allylic and homoallylic alcohols with tert.-butylhydroperoxide

(TBHP) in compressed C02 including assymetric epoxidation under Sharpless conditions' 51.

Cyclic olefins like cyclooctene and cyclohexene have been epoxidized in scC02 with TBHP by the group of Walther. The total number of catalytic cycles was still low (< 10 for cyclohexene, < 30 for cyclooctene) but high selectivity for the epoxide was observed in case of cyclooctene' 61.

Furthermore catalytic oxidation in a two phase water/C02 medium has been investigated for the synthesis of adipic acid from cyclohexene'"' 1. In this system, substrate and product are dissolved in the supercritical fluid and the oxidant (e. g. NalOJ resides in the aqueous phase.

The catalyst Ru02 is oxidized to Ru04 in the aqueous phase, which allows oxidation of the substrate presumably at the liquid/supercritial interface. Again, catalyst efficiency was very low (< 5 catalytic cycles), probably due to deactivation by formation of carbonates in the aqueous phase.

3.2 Oxidation with molecular oxygen

One of the biggest challenges in oxidation catalysis is the selective oxidation of hydrocarbons with molecular oxygen 1’1. Supercritical C02 appears to be an ideal solvent for such processes, because its complete miscibility with 02 provides high concentrations of the oxidant and avoids mass transfer limitations. Furthermore, C02 itself is of course inert to oxidation by 02 and consequently offers additional safety compared to common solvents. Finally, supercritical fluids provide heat transport capacities orders of magnitude higher than gaseous phases and therefore are predestined for highly exothermic oxidation reactions.

Koda et al. m studied the catalytic oxidation of cyclohexane with 02 in the presence of acetaldehyde (cyclohexane : acetaldehyde = 4:1) and a Fe-porphyrin catalyst bearing penta- 213

fluorophenyl groups as solubilisers. The main oxidation products were cyclohexanone (2.9 %

yield) and cyclohexanol (2.3 % yield) with optimum yields slightly above the critical point.

An interesting, but hitherto widely neglected feature of C02 in this context is its ability to

chemically interact with metal oxygen complexes, leading to transition metal

peroxocarbonates as illustrated in Figure 2l’,10). These complexes are able to transfer one

oxygen to oxophiles like phosphines, olefins and active methylene groups 1"1, mimicking the

reactivity of monooxygenases.

Figure 2: Formation of transition metal peroxocarbonates

The participation of transition metal peroxocarbonates has been invoked to explain the

influence of gaseous C02 on the rhodium-catalyzed oxidation of tetrahydrofuran (THF) to y-butyrolactone in conventional solvents. This reaction has been studied by the groups of

Aresta,Mbl and Nicholas 1"'1 with THF as both solvent and reactant using a 0JC02 (1:1) gas

mixture as oxidant (Figure 3).

Employing scC02 as solvent, we found that this oxidation occured also effectively despite the

extremely low solubility* 31 of the catalyst [(dcpe)Rh(hfacac)] (dcpe = dicyclohexylphosphino-

ethane, hfacac = hexaflouroacetylacetonate) in scC02. The reaction was carried out in a

window equipped 100 ml stainless steel autoclave to ensure that a homogeneous supercritical

reaction mixture was formed under the reaction conditions (0.45 mol of substrate and a 214 substrate to catalyst ratio of 4500 :1). The turnover numbers (TON, mol product per mol Rh) and the amount of by-products, mainly formic acid, 2-hydroxy-THF and dihydrofuran-2,5- dione were found to depend strongly on the temperature and the partial pressure of O,. The highest TON obtained for the formation of y-butyrolactone was 135, with a maximum turnover frequency (TOP, mol product per mol Rh and hour) of 15 h"' (Figure 3). as® co2/o2 other G# O [Rh] a oxidation products ______lw?f Aresta(1992) 20°C, IbarO/CO; TOF = 50 h"1 i

Nicholas (1992) 60°C,3-30bar02/C02,8d TON =1000

# This work 60 °C, 35 bar 02, scC02,18 h TON = 135, TOF = 15 h"'

Figure 3: Oxidation of THF

Detailed NMR-investigations of the behaviour of the catalyst under reaction conditions showed, that the posphine ligand is oxidized rapidly to the corresponding phosphine oxide, as expected. However, no oxidation of THF was observed using the phosphine free complex

[(cod)Rh(hfacac)] (cod = 1,5-cyclooctadien) as catalyst. Differences in solubility between the two metal species could be ruled out as a possible explanation for their difference in activity.

Therefore the presence of a phosphorus compound (phosphine or phosphine oxide) seems to be essential for the oxidation process. 215

4. Conclusion and outlook ^Although the catalytic efficiency of all catalytic oxidation processes studied in scC02 up to

now is far from being satisfactory, the principle possibility to carry out such reactions in this

medium is clearly evident. Future research in our group will be directed towards the

development of homogeneous and heterogeneous catalysts that are adopted to the special

requirements of both the oxidation process and the supercritical reaction medium. Preliminary

results from these studies regarding the epoxidation of olefins with molecular oxygen as

oxidant will be presented on the conference post er. ^

5. References

[1] Recent reviews: a) P. G. Jessop, T. Ikariya, R. Noyori, Science, 269, 1065 (1995); b) D. A. Morgenstem, R. M. LeLacheur, D. K. Morita, S. L. Borkowsky, S. Feng, G. H. Brown, L. Luan, M. F. Gross, M. J. Burk, W. Tumas in Green Chemistry (Eds.: P. T. Anastas, T. C. Williamson), ACS Symp. Ser. 626, American Chemical Society, Washington DC, , pp. 132 (1996); c) E. Dinjus, R. Fomika, M. Scholz in Chemistry under Extreme or Non-Classical Conditions (Eds.: R. van Eldik, C. D. Hubbard), Wiley, New York, pp. 219 (1996); d) P. G. Jessop in Fine Chemicals Catalysis (Eds.: D. G. Blackmond, W. Leitner), Topics in Catalysis, 5,95 (1998). [2] K. Zosel, Angew. Chent., 90,748 (1978); Angew. Chem. Int. Ed. Engl., 17,702 (1978). [3] S. Kainz, D. Koch, W. Baumann, W. Leitner, Angew. Chem., 109,1699 (1997); Angew. Chem. Int. Ed. Engl., 36, 1628 (1997). [4] A. FQrstner, D. Koch, K. Langemann, W. Leitner, C. Six, Angew. Chem., 109, 2562 (1997); Angew. Chem. Int. Ed. Engl., 36, 2466 (1997). [5] D. R. Pesiri, D. K. Morita, W. Glaze, W. Tumas, Chem. Commun., 9,1015 (1998). [6] U. Kreher, S. Schebesta, D. Walther, Z. anorg. allg. Chem., 624, 602 (1998). [7] a) R. A. Sheldon, J. K. Kochi, Metal-Catalyzed Oxidations of Organic Compounds, Academic Press, New York (1981); b) K. A. Jorgensen, Chem. Rev., 89,431 (1989). [8] X.-W. Wu, Y. Oshima, S. Koda, Chem. Lett., 1045 (1997). [9] M. Aresta, I. Tommasi, E. Quaranta, C. Fragale, J. Mascetti, M. Tranquille, F. Galan, M. Fouassier, Inorg. Chem., 35,4254 (1996). [10] W. Leitner, Coord. Chem. Rev., 153,257 (1996), and references cited therein. [11] a) M. Aresta, E. Quaranta, A. Ciccarese, J. Mol. Catal., 41,355 (1987); b) M. Aresta, C. Fragale, E. Quaranta, I. Tommasi, J. Chem. Soc. Chem. Commun., 315 (1992); c) A. K. Fazlur-Rahman, J.-C. Tsai, K. M. Nicholas, J. Chem. Soc. Chem. Commun., 1334 (1992). DE99G2071 217

DGMK-Conference "Selective Oxidations in Petrochemistry ”, Hamburg 1£

*02012198318*

A. Martin*, B. Liicke*, A. Bruckner*, U. Steinike*, K.-W. Brzezinka**, M. Meisel*** * Institut fur Angewandte Chemie Berlin Adlershof e.V., Rudower Chaussee 5, D-12484 Berlin, Germany ** Bundesanstalt for Materialforschung und -prufung, Berlin, Germany *** Humboldt-Universitat zu Berlin, Institut fur Chemie, Berlin, Germany

CHARACTERIZATION OF VPO AMMOXIDATION CATALYSTS BY IN SITU METHODS Introduction The ammoxidation of substituted toluenes and methylheterocycles to the corresponding nitriles is an industrially important reaction [e.g. 1]. The synthesized nitriles are valuable intermediates in the organic synthesis of different dyestuffs, pharmaceuticals and pesticides. Vanadium phosphates (VPO) of different structure and vanadium valence state are suitable precursor compounds of very active and selective catalysts for the ammoxidation of these aromatic educts [2,3]. If VOHPO4 • x H20 (x = 0, V2, 2 and 4) is heated in the presence of ammonia, air and water vapour (NH4)2(V0)3(P207)2 as XRD-detectable phase is formed [4,5]. Caused by the stoichiometry of the transformation reaction (V/P=l => V/P=0.75) and the determination of the vanadium oxidation state of the transformation product (- 4.11 [4]) a second, mixed-valent (Vrv/Vv) vanadium-rich phase (VxOy ) must be formed. The VxOy phase seems to be responsible for the catalytic activity, i.e. the increase of its proportion could boost the catalyst performance. | In-situ methods are well known as powerful tools in studying catalyst formation processes, their solid state properties under working conditions and the inter­ action with the feed, intermediates and products to reveal reaction mechanisms. This paper gives a short overview on results of intense studies using in-situ techniques to reveal VPO catalyst generation processes, interaction of educts, intermediates and products with VPO catalyst surfaces and mechanistic insights. Catalytic data of the ammoxidation of toluene on different VPOs complete these findings. The precursor-catalyst transformation processes were preferently investigated by in-situ XRD, in-situ Raman and in-situ ESR spectroscopy. The interaction of aromatic molecules and intermediates, resp., and VPO solid surfaces was followed by in-situ ESR and in-situ FTIR spectroscopy. Mechanistic information was mainly obtained using in-situ FTIR spectroscopy and the temporal-analysis-of-products (TAP) technique. Catalytic studies were carried out in a fixed-bed microreactor on pure (NH4)2(V0)3(P20?)2, generated [(NH4)2(V0)3(P207)2 + VxOy ] catalysts, having different VxOy proportions by use of VOHPO4 • V2 H20 (V/P = 1) and recently studied (V0)3(P04)2 • 7 H20 (V/P = 1.5) precursors; the well-known (V0)2P207 was used for comparison^

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 218

Experimental

The precursor compounds VOHPO4 • V2 H20 (VP,/lH) and (V0)3(P04)2 • 7 H20 (VP7H) have been prepared in aqueous solutions as described elsewhere [6], The materials were pelletized and crushed (1-1.25 mm) and then pretreated by heating to 673 K under NH3-air-H20 vapour (molar ratio =1:7:5, total flow = 13 1 h-1) for 5 h. The generated catalysts are denoted as AVPgen ,AH and AVPgen7H, respectively. The catalysts contain different proportions of the mixed-valent VxOy phase (see Equ. 1 and 2, Table 1 ). Pure (NH4)2(V0)3(P207)2 (AVPsyn ) has been synthesized as described in [4], (V0)2P207 (VPP) that has been generated by usual dehydration of VP,AiI precursor is applied as model catalyst for comparison.

4 Viv 0HP0 4 • 1/2 H20 (2 NH3,02) => [(NH4)2(Viv 0)3(P207)2 + VIwvxOy ] + 3 H20 (1)

2 (V^0)3(P04)2 • 7 H20 (2 NH3,02) => [(NH4)2(Viv 0)3(P207)2 + 3 Vlv/vx0y ] + 13 H20 (2)

Table 1 Chemical analyses, BET surface areas and vanadium valence states of the solids used for further characterization and ammoxidation runs

Samples N H H/N V BET surface V valence state (wt.%) (wt.%) molar ratio (wt.%) area / m2 g "1

AVPsyn 4.77 1.37 4.02 25.91 3.04 4.000 AVPgenHH 4.95 1.28 3.62 29.17 2.95 4.111 AVPgen7H 3.81 0.95 3.49 35.24 3.60 4.248 VPP" 32.83 5.31 4.007

The precursor-catalyst transformation was followed by in-situ XRD using a commercial XRK reactor chamber (A. Paar) that could be heated up to 1173 K (sample amount ca. 150 mg). The reactor chamber is coupled with an on-line capillary GC to check catalytic properties during transformation. Similar studies were carried out in an home-made water cooled in-situ Raman cell (T<873 K, sample amount ca. 10-20 mg, 2.5 mW power level) [7], The cell is mounted on a DILOR-XY spectrometer (Ar-Laser, backscatter geometry, micro-Raman technique). ESR measurements were performed with the cw spectrometer ERS 300 (ZWG) equipped for in-situ investigations with a flow reactor (0.4 g catalyst) and a gas as well as liquids (with vaporization) supplying system [e g. 8 ]. The infrared spectra were recorded with a Bruker IPS 66 FTIR spectrophotometer using self-supporting discs mounted in an in-situ cell connected with a gas manifold-evacuation system. The TAP reactor unit was used for the investigation of the effect of a prereduction-preoxidation of a VPO catalyst near-surface area on 219

its catalytic properties [9] and for isotope experiments (15NH3-containing feed) [10] that should reveal the role of NH4+ ions during the catalytic cycle. The catalytic properties of the solids were determined during the ammoxidation of toluene to benzonitrile using a fixed-bed U-tube quartz-glass reactor (ca. 1,5 g catalyst). The following reaction conditions were applied: toluene : NH3: air: H20 vapour = 1 : 4.5 : 32 : 24, atmospheric pressure, W/F = ca. 10 g h mol 1 (total flow). The catalytic runs were performed at ca. 573, 598 and 623 K for 45 min each. Toluene conversion and benzonitrile yield were followed by on-line capillary GC using a FID as detector.

Results and discussion | Fig. 1 depicts the XRD patterns obtained during heating the VPWH precursor up j to 713 K. The transformation proceeds via dehydration, incorporation of NH4+ i. ions and formation of the three-dimensional AVP structure [4,7]. The transfer- matron process is finished after ca. 4-5 h. Cooling under N2 reveals the pattern of | pure AVP, i.e. no additional information on a VxOy phase could be seen. |

713, Oh

2 Theta

Figure 1 in-situ XRD patterns of the VOHPO4 • Vi H20 phase transformation into (NH4)2(V0)3(P207)2 + VxOy (AVP^h ) [11] Otherwise, cooling down the sample under the NH3-containing feed (up to 393 K) generates NH4V03 particles beside AVP as clearly identified by the obtained patterns as shown in Fig. 2 (a). The lines indicate the NHfV03 reflections. For comparison, the pattern of pure AVP^ is shown in Fig. 2 (b). Thus, it seems very likely that NH4V03 could be formed from mixed-valent vanadium oxides, existing under reaction conditions and an excess of closely neighboured NH4+ ions [4], 220

„

mJ lAi t tV'VvJ JiAlikl

b wVui^V lAui io 15 20 25 30 2 Theta

Figure 2 XRD patterns of the (NH4)2(V0)3(P207)2 + VxOy catalyst (AVPgenl 4H) obtained after cooling under NH3-containing feed (NH4V03 marked by lines) (a) in comparison to pure (NH4)2(V0)3(P207)2 (AVPsyn ) (6) The transformation of the orthophosphate precursor VP7H proceeds in a similar manner, but various, more intense reflections in the pattern of the N2 cooled transformation product point to different vanadium oxides; a specific assignment to defined solids is not straightforward and still under study. Recently, the VPWH transformation process into AVPgenWH was also followed by in- situ Raman spectroscopy. The Raman spectra revealed the existence of NH4V03, too [7], Moreover, it was possible to find areas that consist of V205 mainly. NH4V03 and V205 are spatially separated from the AVP surface as detected by the micro-Raman technique [7], Fig. 3 presents Raman spectra of selected surface areas (marked by asterisks) using an Olympus lOx objective (Laser focus ca. 10 pm) in comparison to the pure solids. The search for lower valent vanadium oxides was not successful because V(IV)-containing vanadium oxides are not so easy to detect by Raman spectroscopy. This fact is strengthened by the simultaneous presence of V(V) compounds as in this case. In-situ ESR spectroscopy can be applied as a useful tool for the characterization of the electronic properties of V(IV)-containing solids and their interaction with feed components during catalytic runs [8 ]. 221

30000 I / a.u.

20000 (NH4)2(V0)3(P207)2 10000

NH,VO. 0 0 500 1000 1500 2000 wavenumber / cm'1

Figure 3 Raman spectra of selected surface areas (marked by *; micro-Raman technique) of an (NH4)2(V0)3(P207 )2 + Vx Oy catalyst (AVPgenMH) obtained after \ phase transformation in comparison to spectra of pure compounds

Fig. 4 shows in-situ ESR spectra of the VPWH transformation and the respective quotients of the 4th and the square of the 2nd moment of the ESR signals, characterizing the efficiency of the spin-spin exchange between neighbouring V02t ions. The spectra reveal that the transformation proceeds via a strongly disordered state. The exchange VOHPO< W HjO (NH4),(V0),(P207),+ efficiency is also related to the 14 v,o, catalytic activity [e.g. 12]. Active A 0 0 catalysts contain coupled vanadyl • 0 10 sites and are characterized by a • high degree of the exchange efficiency in comparison to less active ones with mostly isolated <>000 vanadyl centers. The reason 300 400 600 600 12 3 therefore could be that the T/K time-on-stream /h (at 673 K) alteration of the electron density at a discrete surface vanadyl unit caused by the redox process can be easily delocalized via the overlapping d-orbitals of the exchange coupled sites [12]. These moment quotients spin- spin exchange values also change during interaction of the solid yOHP0 4 • V2 H20 transformation into with feed components. For (NH4)2(V0)3(P207)2 + VxOy (AVPgenvai) example, in-situ studies revealed that substituted toluenes with electron-donating substituents (-OCH3) disturb the spin-spin exchange much more than toluene or educts having rather electron- 222

withdrawing groups (-C1) [13]. This is due to a stronger interaction of the more basic ring systems with the vanadyl sites of the surface. These findings are reflected in catalytic runs under comparable conditions; conversion and nitrile selectivity are decreased in the -OCH3 case compared to the reaction of Cl- toluenes. In-situ ESR and in-situ FTIR spectroscopic studies [14] also confirm the reaction mechanism of the ammoxidation of toluene on VPO catalysts that show coupled vanadyl sites as in the case of (VOlzPgO, (Fig. 5). These ideas could be also transferred to vanadium oxide surfaces. Some V-O-V and/or V-O-P links are hydrated under reaction conditions and NH4+ ions are formed. The toluene molecule is chemisorbed on a lewis-acid vanadyl site; the reaction proceeds via methylene species, benzaldehyde, benzylimine to benzonitrile as indicated by in- situ FTIR spectroscopy.

h 2o

Figure 5 Reaction mechanism of the toluene ammoxidation on a vanadyl dioctahedra unit (100 plane) of (VO^PzO? Pulse catalytic experiments using the TAP technique, (VOlgPgO? as catalyst and toluene as educt revealed a distinct dependence of the catalytic activity and product selectivity on the oxidation state of the catalyst surface [9]. The catalyst 223

surface was oxidized by 02 and reduced by NH3 pulses. The investigations showed that an increasing proportion of surface V(V) accelerates the catalytic process but the nitrile selectivity drops by overoxidation, otherwise, increasing surface V(III) portions drastically restricts the catalytic activity. Some TAP followed ammoxidation runs were carried out on AVPsyn to study the role of NH4+ ions during ammoxidation using 15NH3-containing feed [10]. The experiments revealed that no gas phase ammonia reacts and NH4+ ions participate as potential nitrogen source in the nitrile formation. It was found out that during an ammoxidation pulse series only 14N-benzonitrile is formed at first and the portion of 15N-benzonitrile slowly increases with further pulses. It seems veiy likely that the NH4+ ions formed during catalytic runs could be able to intervene in the mechanism as well. Fig. 6 schematically depicts these ideas.

Figure 6 N-insertion during ammoxidation by NH4+ ions of the solid

620 K 595 K 570 K

VPP AVP,genl/2H gen7H

Figure 7 Area-specific rate of toluene conversion during ammoxidation runs using NH4+-containing VPO catalysts and (VO^PgO? (VPP) for comparison Fig. 7 demonstrates area-specific toluene conversion rates of different AVP catalysts and the usual VPP specimen, depending on the reaction temperature. 224

Pure AVPsyn (exposing single vanadyl sites on the surface) seems to be nearly inactive. The conversion rate of the VPP catalyst (edge-shared vanadyl dioctahedra units) looks rather low as well. Owing to the existence of mixed- valent V,Oy (planes of edge-shared vanadyl octahedra) of the AVPgen solids and additionally in increased portions of the orthophosphate derived catalyst, these solids reveal a significant enhancement of the toluene conversion rate at almost equal high nitrile selectivities in comparison to the other VPOs. In conclusion, the studies have shown that in-situ methods could help to reveal and understand catalyst formation processes and the effect of catalyst-feed inter­ action on conversion and selectivity. It could be seen that an active and selective ammoxidation on VPOs requires i) adjacent vanadyl units, ii) an easy and fast change of the surface V-valence that should be +4 on the average and Hi) the presence of NH4+ as structural unit or rapidly generated by hydrolysis of V-O-P links.

References

1 R.G. Rizayev, E.A. Mamedov, V.P. Vislovskii and V.E. Sheinin, Appl. Catal. A- General, 83 (1992) 103. 2 A. Martin, B. Lucke, H. Seeboth, G. Ladwig and E. Fischer, React. Kinet. Catal. Lett., 38 (1989) 33. 3 A. Martin and B. Lucke, Catal. Today, 32 (1996) 279. 4 Y. Zhang, A. Martin, G.-U. Wolf, S. Rabe, H. Worzala, B. Lucke, M. Meisel and K. Witke, Chem. Mater., 8 (1996) 1135. 5 U. Steinike, F. Krumeich, L. Wilde, A. Martin and G.-U. Wolf, Mat. Sci. Forum, 278-281 (1998) 660. 6 A. Martin, G.-U. Wolf, U. Steinike and B. Lucke, J. Chem. Soc., Faraday Trans., in press. 7 Y. Zhang, M. Meisel, A. Martin, B. Lucke, K. Witke and K.-W. Brzezinka, Chem. Mater., 9 (1997) 1086. 8 A Bruckner, B. Kubias and B. Lucke, Catal. Today, 32 (1996) 215. 9 A. Martin, Y. Zhang and M. Meisel, React. Kinet. Catal. Lett., 60 (1997) 3. 10 A Martin, Y. Zhang, H.W. Zanthoff, M. Meisel and M. Baems, Appl. Catal. A- General, 139 (1996) Lll. 11 L. Wilde, unpublished results. 12 A. Bruckner, A Martin, N. Steinfeldt, G.-U. Wolf and B. Lucke, J. Chem. Soc., Faraday Trans., 92 (1996) 4257. 13 A Bruckner, A Martin, B. Lucke and F.K. Hannour, Stud. Surf. Sci. Catal., 110 (1997) 919. 14 Y. Zhang, A Martin, H. Berndt, B. Lucke and M. Meisel, J. Molec. Catal. A Chemical, 118 (1997) 205. 225

DGMK-Conference "Selective Oxidations in Petrochemistry ”, Hamburg 1998

R. Sumathi, B. Viswanathan, T. K. Varadarajan Department of Chemistry, Indian Institute of Technology, Madras 600 036, India

PARTIAL OXIDATION OF 2-PROPANOL ON PEROVSKITES

■Abstract ______Partial oxidation of 2-propanol was carried out on ABi.x B’x03 (A = Ba, B= Pb, Ce,Ti; B’=Bi, Sb and Cu) type perovskite oxides. Acetone was the major product observed on all the catalysts. All the catalysts underwent partial reduction during the reaction depending on the composition of the reactant, nature of the B site cation and the extent of substitution at B site. The catalytic activity has been correlated with the reducibility of the perovskite oxides determined from Temperature Programmed Reduction (TPR) studies. \

Introduction

Transformation of 2-propanol is widely used to characterise acid-base properties of oxide catalysts owing to the simplicity involved in the design of the reactor and analysis of the products [1,2]. On contact with an acidic or basic site 2-propanol can undergo three types of competitive reactions, namely, (I) Intramolecular dehydration to give propene and water (ii) Intermolecular dehydration to give di-isopropyl ether and (iii) dehydrogenation to give acetone and hydrogen. While the surface acid-base properties of the solids are known to be responsible for these reaction types, the specific role of each property in each process remains obscure. The initial rates of dehydration and dehydrogenation have been correlated by many authors with the acidity and basicity respectively of the solid used [3]. Others however, have found interesting correlations between the ratio of both initial rates and the overall basicity of the catalysts employed [4,5]. Calculation of the values of Arrhenius activation energy for the dehydration and dehydrogenation processes allow one to characterise the nature and strength of the surface acid sites.

Experimental

The perovskite oxides were prepared by following the procedure reported in the literature [6], Corresponding carbonates or oxides were mixed in appropriate ratio, ground and then calcined in the range 1073-1173 K in air for 20-24 h with intermittent grinding. Temperature-programmed reduction studies were carried out by following the procedure given in literature [7]. A mixture of 90 vol % nitrogen and 10 vol% hydrogen was used

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 226

as a reducing gas for all TPR experiments. Hydrogen was purified over a copper metal trap at 473 K. For a typical TPR experiment, 100 mg of the sample was loaded in a U tube ( 4 mm id) quartz reactor. Hydrogen uptake was monitored using gas chromatograph (GC) with thermal conductivity detector (TCD) and then passed through the reactor containing the catalyst. The reactor exit was directed through a cold trap to remove product water and then to a second arm of the TCD. The flow rate of the reducing gas was maintained at 32 ml/min. The reactor was heated at a linear rate of 20 K/min from 323 to 1223 K. The consumption of hydrogen due to the reduction of a specific species on the catalyst was monitored by the TCD and the output was recorded on a strip chart recorder. Gas phase catalytic oxidation of 2-propanol was carried out in a fixed bed continuous flow reactor made of pyrex glass with an inner diameter of 15 mm. The temperature of the catalyst bed was measured with the help of a thermocouple placed at the centre of the catalyst bed. 2-propanol was distilled before use and checked for its purity. The alcohol was fed into the reactor by means of syringe infusion pump. Reactions were carried out both in the presence and absence of oxygen. The partial pressure of oxygen was varied from 0-200 torr. Products were collected after allowing about 15 min to attain steady state conditions in ice cold traps. Quantitative analysis was carried out using 20% carbowax column operating at 333 K. After every run the catalyst was regenerated in air at 673 K for 4h.

Results and Discussion

X-ray diffraction studies reveal that BaPbi.xBix 03 and BaCei.xBix 03 form a complete solid solution in the whole composition range (o^x^l). BaPbi.xBixCuy03 forms a perovskite type structure when x = 0.30 and y £0.35. BaPb|.xSbxC>3 also forms a perovskite type structure when x £0.50. On all the catalysts only acetone is observed as the major product along with hydrogen and water. Small amounts of methyl iso-butyl ketone ( condensation product of acetone and 2-propanol) is observed at low temperatures and high contact times. This is because at high temperatures and low contact times acetone formed as a result of dehydrogenation of 2-propanol is desorbed rapidly as a result of which acetone is not available for further condensation reaction. No dehydration product, propene or di-isopropyl ether, is observed. This shows the absence of strong acidic sites on these perovskite oxides. Data for the comparison of the catalytic activity of various perovskite oxides are given in Table 1. It is seen that percentage conversion of 2-propanol on BaCe03 is less when compared to other perovskite oxides. XRD pattern recorded after the reaction for BaTiOs and BaCe03 showed the absence of bulk reduction whereas on all other perovskite oxides partial reduction of the sample has been observed during the reaction. It has been reported that when CO oxidation is carried out on pure CuO which is calcined at 1213 K the catalytic activity was very low when compared to that of uncalcined CuO. This was attributed to lower reducibility of calcined CuO [8], Based on these results it can be inferred that low reducibility ofBaTiOs and BaCeOs accounts for the low activity of these catalysts for partial oxidation of 2-propanol. 227

The observed variation in catalytic activity can be explained on the basis of reducibility of the oxides determined from TPR studies. In the case of BaPb|.xBix 03 with an increase in Bi content reducibility increases which is indicated by a decrease in Tma*. It is found that the catalytic activity follows the same trend. When Cu is substituted for Pb in BaPbo. 8Bio. 2O3 it is seen that activity is high when compared to BaPbo.sBio. 2O3. This can be attributed to the higher reducibility of Cu substituted systems, indicated by the decrease in Tmax- The possible presence of more labile oxygen species on the surface of BaPbi-x.yBi xCuy03 accounts for the high activity exhibited by them. In the case of BaPb|.xSbx03, the Sb rich catalysts were found to be more active with Tm* decreasing with increase in Sb content. The extent of reduction was also found to increase with an increase in Sb content. However, in the case of BaCei.xBixC>3 the reducibility decreases when the Ce content increases. BaCeo. 2Bio. 8 O3 shows a well defined peak while BaCeo.gBio. 2O3 shows a broad TPR profile without having a well defined Tmax. These studies show that the reducibility and catalytic activity of the perovskite oxides depend on the nature as well as on the extent of substitution at B site. Table 2 correlates the metal-oxygen (M-O) bond energy calculated for B site cation and molar free energy of reduction (AG° )for various oxides which occupy B site with T max and catalytic activity. As the M-0 bond energy decreases, Tmax decreases showing that reducibility of the oxides increases, due to the easy removal of lattice oxide ions. This is reflected in the molar free energy of reduction of the corresponding metal ion. The reducibility increases in the order BaTiC>3 = BaCe03

Mechanism

Ouqour et al.[ 1] carried out 2-propanol decomposition on niobic acid and molybdates of cobalt and nickel. Niobic acid and its oxides exhibited only acidic properties resulting in the formation of propene, whereas mixed cobalt and nickel molybdates in their a or p form, exhibited acidic and basic properties resulting in the formation of both peopene and acetone. For propene formation, three mechanisms are considered, namely Ei (Bronsted H+ site ), E2 ( acid-base pair site) and concerted Bid, (acid-base pair site) [9.10]. For acetone formation, two mechanims are invoked, the first with an alkoxy intermediate and the abstraction of an hydrogen atom on Ca as the limiting step and the second with the formation of an enolate intermediate and the abstraction of hydrogen atom on Cp as the limiting step [11], In the present study, formation of propene is not observed on any of the perovskite oxides. Hence, it can be suggested that neither Ei nor E2 mechanism is operating on these oxides since both Ei and E2 require the presence of strong acidic sites. The third alternate mechanism is the Eicb- It has been reported that on catalysts with large number of basic sites both dehydrogenation and dehydration takes place via Eub mechanism, the former predominating over the later. Since only acetone 228

(dehydrogenation product) is observed in the present study it can be inferred that Eicb mechanism is operating. This mechanism requires the presence of both acidic and basic groups or in other words electron pair acceptor (EPA) and electron-pair donor (EPD) groups. According to Vinek er al [12] every cation-anion pair on the surface is capable of acting as an active centre provided it is accessible. The ratio of dehydrogenation/ dehydration activity depends upon the EPD and EPA strengths. The EPD strength can be obtained from O Is binding energy obtained from XPS studies. Vinek et al.. [12] carried out alcohol decomposition on various oxides and showed that oxides which give Ols binding energy between 531 and 533 eV (like AI2O3, SiOz, GeOz and P2O5) promote mainly dehydration via Eior E2 mechanism, whereas oxides, like Y2O3, LazOs, CeOz, SmzOz and DV2O3 which give Ols binding energy below 531 eV are considered to be basic and promote dehydrogenation to give acetone via Eicb mechanism. According to this mechanism, the interaction between a basic site and an alcohol molecule causes a proton to be abstracted from the alcoholic group, thus producing an adsorbed alkoxide species. Yamashita et al. [13] used deuterated 2-propanol and carried out the reaction on fine particle nickel catalysts and found that the formation of acetone is proceeded by the abstraction of hydrogen from the -OH group. Based on the fact that only acetone is observed in the present study, a possible mechanism is proposed as given below. Hydrogen adsorbed on the catalyst surface can take up a lattice oxide to give a "water molecule and this is evidenced by the reduction of the catalyst This depends on he reducibility (mobility of oxide ion ) of the catalyst

h 3cx

0

,n + M

C II 0 229

Table 1 Catalytic performance data for 2-propanol oxidation on various perovskite oxides (W/F=40.0 g.h/mol; p 0i= 0 torr) ______Catalyst Temperature (K) Conversion Tmax Ea (moI%) (K) (kJ/mol) 623 9.6 BaPb03 648 17.8 913 65.3 673 22.0

623 14.2 BaPbo. 5Bio. 4O3 648 24.3 848 55.9 673 29.1

623 20.6 BaBi03 648 29.1 798 48.4 673 38.9

BaPbo.gBio.1Cuo.1O3 623 19.1. 648 26.9 848 48.9 673 31.1

623 25.1 648 36.2 698 33.9 BaPbo.eBio.iCuojOg 673 43.6

623 20.9 648 28.1 823 46.1 BaPbo. 9Sbo.lO 3 673 35.5

623 23.7 BaPbo. 5Sbo. 5O3 648 35.9 773 32.2 673 42.1

623 4.2 BaCeOg 648 6.1 - 90.1 673 7.9

623 7.1 baueo. 8bio. 2U3 648 8.9 _ 81.0 673 10.9

623 15.9 BaCeo. 2Bio.gO 3 648 24.2 806 48.4 673 29.1 230

Table 2 Comparison of M-O bonding energy and molar free energy of reduction (AG°) of pure oxides which occupy B site in the perovskite structure with the Tmax and catalytic activity of the perovskite oxides. (Reaction conditions; temperature 673 K,W/F: 40.0 g.h./mol)

*Bond **AG° perovskite Tm„(K) convers #Rate M-0 energy (kJ/mol) ion (mol h"'m"2) (kJ/mol) (mol%) Cu-0 96.6 -108.9 BaPbo.6 Bio.1Cuo.3O3 698 43.6 0.30

Sb-0 106.3 - 85.3 BaPbo. 5Sbo.sO 3 773 42.1 0.26

Bi-0 108.6 - 50.4 BaBiOs 803 38.9 0.24

Pb-0 119.7 - 25.0 BaPb03 913 22.0 0.15

Ti-0 277.0 +212.0 BaTi03 >1273 7.5 0.05

Ce-0 293.2 +225.0 BaCe03 >1273 7.9 0.04

* M-0 bond energy was calculated according to the procedure given by Shimizu (1980)[14] ** AG° = Free energy of reduction was calculated for the reaction, metal oxide + Fh —> Metal + H2O # Normalised Rate per unit surface area.

References 1. A.Ouqour, G.Coudrier and J.C.Vedrine, J.Chem.Soc.Faraday Trans., 88(16), 3151(1993). 2. D.C.Tomaczak, J.L.AUen and K.R.Poeppelmeir, J.Catal., 146,155 (1994). 3. J.E.Rekoske and M.A.Barteau, J.Catal., 165,57 (1997). 4. A.Gervasini and A.Auroux, J.Catal., 131,190 (1991). 5. M.A.Aramendia, V.Borau, CJimenez, J.Marinas, A.Porras and FJ.Urbano, J.Catal., 161,829 (1996). 6 . A.W.Sleight, G.L.Gillson and P.E.Bietstedt, Solid State Commun., 17,27 (1975). 7. N.W.Hurst, SJ.Gentry, A.Jones and B.D.McNicol, Catal.Rev.Sci.Eng., 24(2), 233 (1982). SIHalasz, A.Brenner, M.Shelef and K.Y.Simon Ng, Catal.Lett., 6,349 (1990). 9. F.Figueras, L.De Mourgues and Y.Trambouze, J.Catal., 14,107 (1969). 10. Y.Imuzi, T.Yamaguchi, H.Hattori and K.Tanabe, Bull.Chem.Soc.Jpn., 50,1040 (1977). 1 l.L.Nondeck and J.Sedlacek, J.Catal., 40,34 (1975). 231

12-H.Vinek, H.NolIer, M.Ebel and K.Schwarz, J.Chem.Soc.Faraday Trans 1,72,734. (1977). 13. M.Yamashita, F.Y.Dai, M.Suzuki and Y.Saito, Bull.Chem.Soc.Jpn., 64,628 (1991). 14. T.Shimizu, Chem.Lett., 1 (1980). 232 233

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*05012198^36* M. ROsch gen. Klaas, S. Warwel —\~" Instituie'for Biochemistry and Technology of Lipids, H.P. Kaufmann-lnstitute, Federal Centre for Cereal, Potato and Lipid Research, Piusallee 68, 0-48147 MQnster, Germany

CHEMO-ENZYMATIC EPOXIDATION OF OLEFINS BY CARBOXYLIC ACID ESTERS AND HYDROGEN PEROXIDE

Countless processes for the epoxidation of C=C-bonds in higher molecular olefins have been proposed until today. However, no method, let it be transition metal catalyzed [1] or using dioxiranes [2] or the Mg-salt of monoperoxy- phthalic acid [3], has been able to challenge meta-chloroperbenzoic acid (mcpba) as the working horse in organic synthesis, although mcpba is neither cheap nor completely harmless. he situation for industrial epoxidations is similar. j^Ethylen and, recently, butadiene can be epoxidized directly with oxygen and for the epoxidation of propylene, the use of heterogeneous transition metals and organic peroxides (Halcon-Process) is the major player. But again, beside from those notable exceptions, all other epoxidations, including large ones like the epoxidation of plant oils as PVC-stabilizers (about 200.000 t/year), are carried out with peroxy acids. Because mcpba is far to expensive for most applications, short chain peracids like peracetic acid are used. Being much less stable than mcpba and thus risky handled in large amounts and high concentrations, these peroxy acids were preferably prepared in-situ. However, conventional in-situ formation of peracids has the serious drawback, that a strong acid is necessary to catalyze peroxy acid formation from the carboxylic acid and hydrogen peroxide^

[H1 R-C-QH—±—Wz@z- R-C-OOH + HzO

The presence of a strong acid in the reaction mixture often results in decreased selectivity because of the formation of undesired by-products by opening of the oxirane ring.

DE99G2069 DGMK-Tagungsbericht 9803,3-931850-44-7,1998 234

Therefore, we propose a new method for epoxidation based on the in-situ preparation of percarboxylic acids from carboxylic acid esters and hydrogen peroxide catalyzed by a commercial, immobilized lipase. J

Since 1990 it is known, that some lipases - an immobilized lipase from Candida antarctica B on polyacrylic resin (Novozym ® 435) is the most active and stable one - are able to catalyze the conversion of middle chain fatty acids with H2O2 to peroxy fatty acids [4-6]. Based on that reaction, Kirk et al. proposed a chemo-enzymatic epoxidation method (figure 1):

[Lipase]

R-C-OH , R-C-00H

Figure 1 Chemo-enzymatic epoxidation of C=C-bonds [4]

The feature of chemo-enzymatic epoxidation that distinguishes it from conventional methods of in-situ peroxy acid generation, is the absence of a strong mineral acid, which makes the reaction highly selective. The reaction is

very useful in oleochemistry [7], where C=C-bond and carboxyl group are situated in the same molecule. However, for the epoxidation of olefins the reaction has some disadvantages: 235

• Fatty acids are necessary as the source of oxidant. Although catalytic amounts are sufficient in principle, stoichiometric amounts have to be used to obtain reasonable reaction rates. These fatty acids must be removed afterwards.

• 60 % H2O2 is applied in e.g. toluene as the solvent. Such a two-phase- system can hardly be realized in a continuous form like a fixed bed reactor, although the enzyme is stable enough for multiple use [8].

• Short chain peroxy acids like peracetic acid can not be prepared with satisfactory results.

Using the same enzyme peracids can also be made from carboxylic acid esters and hydrogen peroxide (so called "perhydrolysis"). We found [9], that perhydrolysis works extremely well, if the ester is applied as solvent and therefore in large excess:

9 [Novozym®435] 9 R-C-OR' + H202------R-C-OOH + R OM

ester as solvent (i.e. large excess)

With this reaction it is possible to use commercial 35% H2O2 without any decrease of activity. Additionally, perhydrolysis has a much broader substrate range (table 1).

There are various ways to analyze percarboxylic acids. They can be quantified by a combination of titrations or isolated by thin-layer chromatography. However, the most convenient and informative method for the comparison of peroxy acids is the use of an epoxidation as a test reaction. We chose the epoxidation of 1-octene:

[Novozym ® 435]

+ H202 in RCOOC2H5 It should be emphasized, that the epoxidation of terminal double bonds is more difficult (and about twenty times slower) than that of internal C=C-bonds. Under 236

the conditions, that were applied for the experiments in table 1, commercial buffered peracetic acid as well as mcpba yield not more than 70 % epoxide.

Table 1 Chemo-enzymaticepoxidation of 1 -octene by Novozym ® 435 catalyzed perhydrolysis of various ethyl esters by H2O2

Ester Presumed oxidizing yield of 1,2-epoxyoctane agenta ) (mol-%, GC) ethyl acetate peracetic acid 61 ethyl isobutyrate perisobutyric acid 70 ethyl laurate perlauric acid 60 ethyl methoxyacetate methoxyperacetic acid 74 ethyl acrylate peracrylic acid 52

S-ethyl-2-bromo- S-2-bromo-perpropionic 30 propionate acid diethyl carbonate monoperoxy carbonic acid 30 monoethylester

preformed buffered peracetic acid b) 65

mcpba b) 72 0.1 mol l"1 1-octene in ester; olefin : H2O2 = 1:5 (molar); enzyme: olefin «1:3300 (molar); 16 h, 40 °C; for experimental details see [9] a) oxidizing agents have not been isolated or positively identified in all cases b) 0.1 mol M 1-octene in C2H4CI2; olefin : CO3H =1:5 (molar); 16 h, 40 °C

By biocatalytic perhydrolysis an impressive choice of peracids can be made in- situ for the epoxidation of olefins. Well known species like peracetic acid [10] can be utilized as well as new oxidants. In most cases yields are comparable to those obtained by conventional means; using new oxidants like peracrylic acid, percarbonic acid derivatives and chiral peracids, the yields are lower. 237

Nevertheless, all these chemo-enzymatic epoxidations have a selectivity >

98 %; there are no by-products, that can be seen in GC.

Table 2 shows the results of the epoxidation of various olefins using ethyl acetate as the source of peroxy acid and as solvent. Yields of 72 - 94 % are achieved with excellent selectivities.

Table 2 Chemo-enzymatic epoxidation of various olefins by acetic acid ethylester / Novozym ® 435 / H2O2

olefin conditions yield of epoxide

(mol-%, GC)

1-octene 40 °C, 16 h, A 61

1-octene 40 °C, 72 h, A 81

4-octene 20 °C, 24 h, B 94

1-tetradecene 40 °C, 16 h, A 71

7-tetradecene 20 °C, 24 h, B 90 1,13-tetradecadiene 40 °C, 72 h, A 77 (diepoxide*)

21 (monoepoxide*) styrene 40 °C, 24 h, B 73 styrene 40 °C, 48 h, B 92 * a-pinene 20 °C, 16 h, B 72

0.1 - 0.5 mol |-1 C=C in acetic acid ethylester

A: C=C : H2O2 = 1:5; 0.01 mol C=C / g Novozym ® 435

B: C=C : H2O2 = 1 :1,5; 0.05 mol C=C / g Novozym ® 435

[*] preparative experiment; isolated yield 238

But, if a single peroxy acid will epoxidize all these olefins with satisfactory results, what is the point of having a range of peroxy acids at hand ?

First of all, the option to choose a particular peroxy acid (and a particular carboxylic acid esteras the solvent) is an advantage for the engineering of the oxidation. The reaction compounds can be selected, so that the reaction product can be isolated most easily (e.g. for the epoxidation of a high boiling olefin a low boiling peracid will be chosen and vice versa). Furthermore, the amount of hydrogen peroxide, that is soluble in the reaction medium, depends on the polarity of the ester.

In some cases the choice of oxidant will influence the selectivity and even the general outcome of the reaction. As we reported recently [11] the chemo- enzymatic epoxidation of unsaturated alcohols using almost every carboxylic acid ester as the source of peroxy acid yields selectively in a three-step-one- pot reaction epoxyalkanolacylates. Figure 2 (left side) shows the direct synthesis of epoxystearylbutyrate from oleyl alcohol and butyric acid ethylester.

A/WWVWV oh

[Novozym ® 435] \ diethylcarbonate H202,16 h

AAAA^/WWo^V O 89 % ° ° 60 % /VW\=AAAA_0.C/V 7% °

OH ° 4%

Figure 2 Chemo-enzymatic epoxidation of unsaturated alcohols by perhydrolsis of carboxylic acid esters and carbonic acid esters 239

However, using diethylcarbonate (figure 2, right side) as the peroxy acid source, the epoxidized alcohol is obtained with excellent selectivity. That may be explained by the insufficient stability of carbonic acid monoesters. In the water containing phase they decompose to CO2 and alcohol and no esterification of the unsaturated alcohol occurs.

In conclusion, lipase-catalyzed perhydrolysis of carboxylic acid esters with hydrogen peroxide is a versatile method to obtain tailor-made oxidants for C=C-epoxidation as well as for other oxidations to be examined in the future.

Acknowledgement

This work was supported by the Deutsche Forschungsgemeinschaft (Schwerpunktprogramm Peroxidchemie / Sauerstofftransfer) und by the Bundesministerium fur Emahrung, Landwirtschaft und Forsten. The authors like to thank Novo Nordisk, Biotechnologie (Mainz) and Soivay-lnterox (Hannover) for donations of materials.

References

[1] a) Review: K. A. Jorgensen, Chem. Rev. 89,431 (1989) b) recent example: J. Rudolph, K.L. Reddy, J.P. Chiang and K.B. Sharpless, J. Am. Chem. Soc. 119, 6189 (1997) [2] W. Adam, J. Bialas and L. Hadjiaragoplou, Chem. Ber. 124,2377 (1991) [3] H. Heaney, Topics Current Chem. 164,3 (1993). [4] F. Bjorkling, S.E. Godtfredsen and O. Kirk, J. Chem. Soc., Chem. Common. 1990,1302 [5] F. Bjorkling, H. Frykman, S.E. Godtfredsen and O. Kirk, Tetrahedron 48, 4587 (1992) [6] O. Kirk, F. Bjorkling, T. Damhus and S.E. Godtfredsen, Biocatalysis 11, 65 (1994) [7] M. ROsch gen. Klaas and S. Warwel, Lipid Technol. 1996,77 [8] S. Warwel and M. Rusch gen. Klaas, J. Mol. Catal. B 1,29 (1995) [9] M. ROsch gen. Klaas and S. Warwel, J. Mol. Catal. A117, 311 (1997) [10] M. Rusch gen. Klaas and S. Warwel, 15. DECHEMA-Jahrestagung der Biotechnologen, Tagungsband, Frankfurt a.M. 1997,311 [11] M. Rusch gen. Klaas and S. Warwel, Synth. Common. 28,251 (1998) 240 241

DGMK-Conference "Selective Oxidations in Petrochemistry ”, Hamburg 1998

G. A. Komashko, S. V. Khalamejda, V. A. Zazhigalov Ukrainian-Polish Laboratory of Catalysis, Institute of Physical Chemistry, National Academy of Sciences of Ukraine, pr Nauki 31, Kyiv-22, 252022, Ukraine

STUDY OF PROPANE PARTIAL OXIDATION ON VANADIUM-CONTAINING CATALYSTS

*DE012198345*

Partial oxidation of propane into acrolein and acrylic acid is an alternative process to industrial one which lies in two-step propylene oxidation. Propylene is rather expensive as compared to propane, and some success reached in n-butane oxidation to maleic anhydride provokes a great interest to the paraffin oxidation reaction. However, it is known that the shorter carbon chain is the lower reactivity of the paraffins becomes. That is why, the appropriate reaction temperature is higher. In its turn, such conditions cause further oxidation of the partial products. Consequently, to solve the main task successfully, it is needed to develop the catalyst, which would be able to activate the hydrocarbon at relatively low temperatures and to slow down the desired products oxidation into COx. There are few publications on propane oxidation into acrolein [1-3] over V-P-0 catalytic systems. It has been shown that unpromoted catalysts are quite active but low efficient for partial products synthesis. More perspective appeared to be some defective mixed oxides containing Bi, V, Mo allowing to obtain the selectivity of 65 % but at low paraffin conversion (13 %). Ag, Li, Na additives enhance acrolein formation. It is assumed, that one of the intermediates in this reaction is propylene. On the other hand, it is reported [4], that propane oxidation over oxide systems of perovskite structure leads to propanal formation and propylene is not found in the products. Both propylene and propanal are the reaction products when the oxidation is performed in the presence of B-P-0 catalyst [5], Propane oxidation into acrylic acid has been observed at rather low temperature i on V-P-0 cataltsts [6]. The presence of surplus phosphorus in the catalyst i composition led to decrease in the selectivity towards the acid, but Te additive and water in gas phase promoted the process [7]. In so doing, maximum yield of acrylic acid is 7 % at 60 % paraffin conversion. Propylene is proposed to be an intermediate product. Much higher yield of the acid can be reached using Cs2.5Fe0.03 H1.26 VM011O40 catalyst. Our conception on paraffin oxidation over V-P-0 catalysts presented in [8] shows, that the main product of propane oxidation must be acrylic acid. The ! improved activity of the catalyst can be attained at the cost of introduction of the additives increasing an effective negative charge on oxygen atom. To reduce probability of acrylic acid further oxidation it is needed strengthenning of the surface acidity, that would favour desorption of the acidic product. The pyrophosphate lattice distortion should also be an advantageous factor as hydrocarbon fragment in such case is limited in its movement within two paired vanadyl octahedra.

DE99G2068

DGMK-Tagungsbericht 9803,3-93185044-7,1998 242

EXPERIMENTAL The following catalytic systems were investigated in the present work: - V-P-0 catalysts promoted with different amounts of Bi, Zr, Te and prepared in organic medium following [9], La additive was introduced by means of mechanochemical treatment of V-P-0 precursor and La203 in ethanol medium similar to [10], using a planetary mill. - V2Os mechanochemically treated in a planetary mill (some properties of the sample were reported in [11]) and in a vibratory mill for 2 h in the presence of water. - V-Si-0 and V-P-Si-0 catalysts prepared by barothermal crystallization from the mixtures of VOfi-OCjHgk, SifOCgHs), and PO(i-OC3H7)3 keeping the reaction conditions close to those described in [12]. - V-Mo-Me-O catalyst developed for propylene oxidation [13]. Catalytic tests were carried out in a flow type reactor made of quartz and having an internal diameter of 6 mm. 2 c.c. of the catalyst diluted with inert material were loaded into the reactor. The reaction mixture contained 1.8 amd 5.1 vol. % propane in air. Analysis of the reaction products was performed by means of on-line gas- chromatographies using molecular sieve (1 m, 25 °C) for analysis of 02 and CO, silicagel LSK (1 m, 25 °C) to analyze C02, dimethylsulfolan (15 %) deposited on chromosorb (3 m, 25 °C) for hydrocarbons determination and Poropak Q (2 m, 130- 170 °C) - for oxygenated products.

RESULTS AND DISCUSSION

V205 catalysts. The presented in Table 1 data show that untreated vanadium oxide is low active in propane oxidation up to 450 °C. Mechanochemical treatment of V205 in the vibratory mill leads to increase of the specific surface area and insignificant reduction of the sample. One can see from XRD data, that the treatment provokes some decrease of (010) reflection intensity. So, the relative content of (010) crystallographic plane having vanadyl oxygen reduces after the treatment. It should be noted that in such conditions the efficiency of the mechanochemical treatment is quite low because rate of the mechanical energy leading up to the sample is commensurate with that of the accumulated energy dissipation. Due to such treatment vanadium oxide becomes much more active catalyst. Thus, at 400 °C propane conversion reaches 40 % though oxygenated products are found in trace amounts. The mechanochemical treatment in the planetary mill is characterized by high rate of the energy leading up and much improved efficiency. Already after 10 min of the treatment (see Table 1) the surface area becomes so enlarged as after 2 h treatment in the vibratory mill, and then, the longer the treatment is the larger surface area grows. Besides, the surface has got more reduced as compared to the sample treated in the vibratory mill. And, on the contrary to the latter case, the concentration of (010) surface becomes larger than that before the treatment. Catalytic activity in propylene oxidation is enhanced not so much as in previous case and conversion is about 30 % at 440 °C. However, acrolein and acrylic acid are formed and the total selectivity towards oxygenated products reaches 13 %. 243

Close inspection of the obtained results leads to a conclusion that the planes of the oxide perpendicular to (010) are active but catalyze propane complete oxidation. (010) crystallographic plane is less active but the selective oxidation of parrafin occurs over it. It is very difficult so far to discuss the delicate mechanism of the reaction running in the planes, elementary acts and the role played by V* + ions.

Table 1 Propane oxidation on V205 after mechanochemical treatment

Treatment R* SSA Hydrocarbon Conversion** Selectivity, mol. %*** m2/g concentration C3H8, % AA AcrA Acr MA Prop

0.7 4.4 1.8 0 0 0 0 0 0 VM120# 0.6 6.2 1.8 63 0.6 0.5 0 0 22 PM10 2.2. 6.4 1.8 16 4.1 6.6 2.8 0 32 PM20 4.1 8.0 1.8 23 4.4 9.1 3.2 0 30 PM30 4.2 8.2 1.8 29 4.2 9.8 3.6 0.5 23

* R -1 (010)/l(110) ** V = 2400 h-1,T = 420 0C *** AA - acetic acid, AcrA - acrylic acid, Acr - acrolein, MA - maleic anhydride, Prop propylene # Time of treatment, min

V-Mo-Me-O catalyst This catalyst is known to be an efficient one in propylene oxidation into acrolein and provides about 90 % yield at the temperatures below 350 °C. However, on propane oxidation the sample acts as a low active catalyst and even at 470 °C propane conversion is only 4 %. At further temperature increase the selectivity towards oxygenated products is low (the main product is acetic acid, selectivity is about 10 %). It means that the oxidation process over this catalyst proceeds through propylene chemisorption by double bond. Its absence in paraffin molecule is the fundamental reason for propane not to be activated and oxidized efficiently.

V-Si-O and V-P-Si-O catalysts. The samples prepared by alkoxide method using barothermal devices have a surprisingly highly developed surface. XRD pattern of the catalysts does not show reflections of any crystalline phase composed them, just wide halo. As one can see from Table 2, V-Si-0 samples are quite active in oxidative conversion of propane, but the selectivity towards acrylic acid is low and does not exceed 2 %. Much higher selectivity is observed towards acetic acid. The major product of the reaction is propylene, although its maximum yield is not high. Simultaneous introduction of V and P into silica matrix leads to some decrease in catalytic activity and growth of the selectivity to acrylic acid. Additionally, maleic anhydride is formed, but propylene and acrolein are not found in the reaction 244

products. Since ethane is also detected as a product, it could be assumed the reaction pathway to produce maleic anhydride. Obviously, disproportional transformations of propane take place to form ethane and butane followed by the latter product oxidation into maleic anhydride. Study of n-butane oxidation over the given catalysts revealed their rather high activity, and the selectivity towards maleic anhydride reached 40 %. The reaction mixture enrichment with propane provokes hydrocarbon conversion decrease (Table 2). In so doing, the selectivity towards acrylic acid increases and that to acetic acid and maleic anhydride grows. Propylene and acrolein are also found in the reaction products. When introducing steam into the reaction mixture the selectivity towards propylene reduces in favour of acrylic acid and acrolein. Acetone is detected in trace amounts.

Table 2 Propane oxidation on V-Si-O and V-P-Si-O catalysts

Catalyst SiA/ SSA Hydrocarbon Conversion* Selectivity, mol. %** ratio m2/g concentration C3H8, % AA AcrA Acr MA Prop

VSiO 20 700 1.8 53 5.3 1.8 0 0 32 , VPSiO 10 350 1.8 48 6.4 5.5 0 8.4 0 VPSiO 10 350 5.1 42 3.2 14.1 3.7 3.9 12 VPSiO 10 350 5.1# 39 2.9 18.3 7.2 2.2 3

* SV = 1800 h-1, T = 380 0C ** AA - acetic acid, AcrA - acrylic acid, Acr - acrolein, MA - maleic anhydride, Prop - propylene # - with steam

V-P-Me-0 catalysts. It is known from the literature [14-16], that an increase of both phosphorus content in the catalyst and hydrocarbon concentration in the reaction mixture as well as steam introduction in it lead to growth of the selectivity towards acrylic acid. Additives of different nature can change the process indexes.

1. V-P-La-0 compositions. The catalysts containing lanthanum oxide are known to be among the most active ones in the reaction of oxidative coupling of methane [17]. It has been shown in our work [18], that La-promoted V-P-0 sample catalyzes n-butane oxidation at the reduced temperature. As it can be seen from the data in Table 3, La additive leads to abrupt improvement of the catalytic activity. At the same time, the selectivity towards acrylic and acetic acids decreases and that to maleic anhydride remains practically unchanged. Enrichment of the reaction mixture with hydrocarbon provokes sharp increase in the selectivity to propylene and moderate growth in acetic and acrylic acids formation at some reduction of the paraffin conversion. When steam is added into gas phase it 245 j is observed further increase in the selectivity towards the oxygenated products and further lowering in both paraffin conversion and propylene forming.

Table 3 Propane oxidation on V-P-La-0 catalysts

Catalyst MeAZ Hydrocarbon Conversion* Selectivity, mol. %“ ratio concentration C3H8, % AA AcrA Acr MA Prop

VPO _ 1.8 14 11.1 9.2 1.1 6.0 0 VPLaO 0.05 1.8 43 3.0 1.3 0.9 5.6 0.8 VPLaO 0.20 1.8 64 trace 0 0 trace trace VPLaO 0.05 5.1 26 9.3 8.4 1.2 7.7 12.1 VPLaO 0.05 5.1# 19 10.1 15.7 2.3 8.9 5.7

*SV= 1800 h-1,T = 380 0C *** AA - acetic acid, AcrA - acrylic acid, Acr - acrolein, MA - maleic anhydride, Prop - propylene # - with steam

2. V-P-Bi-0 and V-P-Bi-Me-0 compositions. One can see from Table 4, that promotion with Bi causes some growth of the activity of V-P-0 catalyst in propane oxidation. In so doing, the paraffin conversion increases with introduction of the larger amount of bismuth. As differentiated from the previous case, the oxygenated products formation is enhanced by introduction of bismuth, the selectivity is passing through its maximum value depending on the bismuth content. Enrichment of the feed mixture with paraffin and steam introduction into it create more favourable conditions for the acids forming over the catalyst of optimum composition (BiA/ = 0.1), but the conversion decreases. Mechanochemical promotion of such V-P-Bi-0 sample with lanthanum leads to some rise in catalytic activity not only in paraffin oxidation but also in further oxidation of the oxygenated products, and the selectivity towards them decreases. Cobalt additive influences on the V-P-Bi-0 catalytic performance in some different way. The promoted with Co catalyst is more active and selective in the reaction of acrylic acid synthesis from propane. Acetic acid is formed in the reduced amount. At the larger concentration of propane in the presence of steam the selectivity towards acrylic acid, and so its yield, still further increases. In all cases enrichment of the reaction mixture with hydrocarbon provokes sharp increase in the selectivity to propylene.

3. V-P-Te-O and V-P-Zr-0 compositions. As it follows from [7], the optimum content of tellurium is TeAZ = 0.1. The catalyst of the same composition has been studied in the present work. It is interesting to note, that tellurium introduction is accompanied with the catalyst surface area reduction by more than 3 times. Nevertheless, the catalytic activity remains at the level of unpromoted V-P-0 sample (see Table 3). The selectivity towards acrylic acid essentially increases and only traces of acetic acid can be 246

Table 4 Propane oxidation on V-P-Bi-0 and V-P-Bi-Me-0 catalysts

Catalyst Me/V Hydrocarbon Conversion* Selectivity, mol. % ratio concentration C3H8, % AA AcrA Acr MA Prop

VPBiO 0.03 1.8 23 13.0 11.3 0 0 0 VPBiO 0.03 5.1 11 24.2 20.8 0 0.2 0 VPBiO 0.05 1.8 31 13.0 12.1 0 0.6 0 VPBiO 0.10 1.8 40 15.2 16.4 0 0.8 0 VPBiO 0.10 5.1 29 21.1 27.2 2.1 3.4 3.3 VPBiO 0.10 5.1# 23 25.2 31.4 4.7 4.8 2.2 VPBiO 0.20 1.8 46 9.2 10.4 0 trace 0 VPBiO 0.30 1.8 47 3.3 8.4 0 0 0 VPBiO 0.30 5.1 34 10.1 15.7 2,3 6.9 4.7 VPBiLaO** 0.10 1.8 53 14.6 16.0 0.6 3.8 0.7 VPBiCoO** 0.10 1.8 50 11.4 21.2 1.2 3.0 2.2 VPBiCoO** 0.10 5.1# 34 17.3 32.2 2.2 3.3 1.8

* SV = 1800 h-1, T = 400 0C ** La/V = Co A/ = 0.05 # - with steam detected. At the higher content of propane it is observed some reduction in its conversion and no change in the selectivity to acrylic acid. Promotion of V-P-O catalyst with zirconium (Table 5) leads to an improvement in its activity and the selectivity towards acrylic acid. Enrichment of the reaction mixture with propane and the presence of steam in it allow to grow the selectivity, but the catalytic activity decreases. The hydrocarbon conversion, depending on zirconium concentration in the catalyst, is going through a maximum at Zr/V = 0.1, and the selectivity to acrylic acid - at Zr/V = 0.2. It should be noted, that over V-P-Zr- O catalyst having ZrA/ = 0.3 at the hydrocarbon concentration of 5.1 vol. % benzene was found in traces amount among the other products of the reaction.

CONCLUSION (The presented results indicate that maximum selectivity to acrylic acid can be reached over V-P-Zr-0 catalysts. When the hydrocarbon concentration is 5.1 vol. % the selectivity is about 30 % at quite high paraffin conversion. Conclusively, some explanations to the observed facts can be given. The V-P-0 catalyst promotion with lanthanum by means of mechanochemical treatment is distinguished by the additive uniform spreading all over the matrix surface. Such two- phase system is highly active in propane conversion (lanthanum oxide) and further oxidation of the desired products. The similar properties are attributed to V-P-Bi-La-0 247

Table 5 Propane oxidation on V-P-Te-0 and V-P-Zr-0 catalysts

Catalyst Me/V Hydrocarbon Conversion* Selectivity, mol. % ratio concentration C3H8, % AA AcrA Acr MA Prop VPTeO 0.10 1.8 15 trace 24.3 0.3 0.7 0 VPTeti 0.10 5.1 9 trace 25.1 0 0.2 6.0 VPZrO 0.05 1.8 32 11.6 20.1 0.8 5.6 0.6 VPZrO 0.10 1.8 55 7.5 21.4 2.2 3.6 2.9 VPZrO 0.10 5.1 49 5.4 29.2 2.9 3.4 10.3 VPZrO 0.10 5.1# 40 15.2 32.4 3.7 3.7 3.2 VPZrO 0.20 1.8 49 9.2 30.2 0.7 11.2 0 VPZrO 0.20 5.1 47 3.3 33.2 2.2 8.6 0 VPZrO 0.30 1.8 35 10.4 18.6 8.7 13.3 5.3 VPZrO 0.30 5.1 22 20.1 18.2 12.9 7.7 17.5

* SV= 1800 h-1, T = 400 0C # - with steam

catalyst. Bismuth, tellurium and zirconium additives having clearly defined acidic properties provoke the surface acidity strengthening and make easier desorption_of the acidic product (acrylic acid) from the surface lowering its further oxidation. Additionally, since bismuth and zirconium are able to form phosphates and, according to to create space limitations for the paraffin molecule movement out of the active group boundaries, this can be one more support in favour of the selectivity increase. With this point of view very infesting results were obtained in It has been shown that the more limited the size of the vanadium unit, the higher the selectivity, is. Monoclinic phase AV2P2O10 which consists in clusters of four vanadium atoms is sensibly more reactive than the orthorhombic phase consists in Voo infinite chains. )

REFERENCES 1. Kim Y.C., Ueda W., Moro-oka Y., Appl. Catal., 70,175 (1991). 2. Kim Y.C., Ueda W„ Moro-oka Y., Catal. Today, 13,673 (1992). 3. Takita Y., Yamashita H., Moritaka K., Chem. Lett., 10,1733 (1989). 4. Conner W.C., Soled S„ Signorelli A., Stud. Surf. Sci. Catal., 57,1224 (1991). 5. Komatsu T., Uragami Y., Otsuka K., Chem. Lett., 11,1903 (1988). 6. Ai M„ Catal. Today, 13,679 (1992). 7. Ai M„ J. Catal., 101,389 (1986). 8. Zazhigalov V.A., In this edition. 9. Zazhigalov V.A, Stoch J., Pyatnitskaya A.I., Komashko G.A., Haber J., Belousov V.M., Appl. Catal., 96,135 (1993). 10. Zazhigalov V.A., Haber J., Stoch J., Bogutskaya L.V., Bacherikova I.V., Appl. Catal., 135,155 (1996). 248

11. Zazhigalov V.A., Haber J., Stoch J., Kharlamov A.I., Bogutskaya L.V., Bacherikova I.V., Kowal A., Solid State Ionics, 101-103,1257 (1997). 12. Zazhigalov V.A., Bacherikova I.V., Yaremenko V.E., Astrelin I.M., Stoch J., Teoret. and Experim. Chem., 31,250 (1995). 13. Zazhigalov V.A., Konovalova N.D., Zaitsev Yu.P., Belousov V.M., Yaremenko E.I., Ukr. Khim.Zhurn., 54,1263 (1988). 14. Michalakos P.M., Kung M.C., Jahan I., Kung H.H., J. Catalysis, 140,226 (1993). 15. Gribot-Perrin N., Volta J.C., Burrows A., Kiely C., Gubelmann-Bonneau M., Stud. Surf. Sci. Catal., 101,1205 (1996). 16. Matsuura I., Kimura N„ Stud. Surf. Sci. Catal., 82,271 (1994). 17. Lee J.S., Oyama S.T., Catal. Rev.- Sci. Eng., 30, 249 (1988). 18. Zazhigalov V.A.Haber J., Stoch J., Komashko GA., Pyatnitskaya A.I., Bacherikova I.V., Stud. Surf. Sci. Catal., 82, 265 (1994). 19. Savary L, Costentin G., Bettahar M.M., Grandin A., Gubelmann-Bonneau, Lavalley J.C., Catal. Today, 32, 305 (1996). 20. Leclaire A., Grandin A., Chardon J., Borel M.M., Raveau B„ Eur. J. Solid State Inorg. Chem., 30,393 (1993). 249

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

*DE012198354* V. A. Zazhigalov Ukrainian-Polish Laboratory of Catalysis, Institute of Physical Chemistry, National Academy of Sciences of Ukraine, pr. Nauky 31, Kyiv-22, 252022, Ukraine

ACTIVE GROUPS FOR OXIDATIVE ACTIVATION OF C-H BOND IN C2-Cs PARAFFINS ON V-P-O CATALYSTS

/por the first time in scientific literature, in our joint work with Dr. G. Lad wig ^ in 1978 it was established phase portraite of the oxide vanadium-phosphorus system within wide range of PA/ ratios from 0.5 to 3.2. Some later those data were confirmed r in-{2]. By investigation of the properties of individual vanadium-phosphorus phases it. was also shown that the active component of such catalysts in n-butane oxidation was vanadyl pyrophosphate phase (VOjaPaO/. From then the conclusion has been evidenced by numerous publications and at present it has been out of doubt practically all over the world $eF-examplersee-Ftefrf3}J. It was hypothized (h-Wthat the unique properties of (VOjzPzO? in the reaction of n-butane oxidation could be explained by the presence of paired vanadyl groups and nearness of the distances between neighbouring vanadyl pairs and that between the first and fourth carbon atoms in n-butane molecule. The molecule activation occured at the latter atoms by proton abstraction. A comparison of the results on n-butane and butenes oxidation over vanadyl pyrophosphate M allowed to conclude that the paraffin oxidation did not take place due to the molecule dehydrogenation process at the first stage of its conversion. Up to now, more than 100 papers related to paraffins oxidation over vanadyl pyrophasphate and the physico-chemical properties of the catalyst have been published. The process of n-butane oxidation is realized in practice. But still, the question about the nature of active sites of the catalyst and the reaction mechanism remains open and provoks further investigations. The present paper deals with our opinion about the problem and the experimental results supporting

MECHANISTIC FEATURES OF n-BUTANE OXIDATION

For n-butane partial oxidation, the important role of (100) crystallographic plane of (VO^PzO? having paired vanadyl octahedra in its structure is generally accepted. It could be seen from dependence of the catalytic properties on the relative content of the plane [4-6], upon its deactivation [7], and the observations of the surface by

' The present contribution is devoted to the light memory of Dr. G. Ladwig, died last year, one of the leading specialists in the field of phosphates chemistry with whom the author had been working for several years.

DE99G2067 DGMK-Tagungsbericht 9803,3-931850-44-7,1998 250

HREM [8,9] as well. The proposed by us scheme is grounded on the following statements: Activation of C-H bond proceeds with proton abstraction. On the contrary, it has been proposed in work [10] homolytic break of the bond and the authors of publications [4, 11] considered such a case simultaneously with H+ and H" abstraction and their fixation on 0°" and V4* (or V5*). It should be noted, that both these assumptions have not been ever proved experimentally. Proton abstraction, contrastingly, is supported by both the results on activation of n-butane oxidation at N20 introduction [12] and the data on dependence of the process rate on an effective negative charge (XPS) on oxygen atom [13]. The similar correlation has been found (Auge-spectroscopy) for propane activation [14]. Dehydrogenation of hydrocarbon with butenes forming does not occur at the first stage of activation. On oxidation of n-butane in the standard conditions (1- 2 % vol. in air) there are no any other partial products formed but only maleic anhydride. Attempts to explain an absence of olefins in the products at the cost of differences in their oxidation rate constants [15] are too weak to warrant any idea. Calculations by equation given in [15] show that while the paraffin conversion is 50- 70 % the olefin content in the products should be 0.01-0.04 % vol. The detection of C4 olefins on oxidation of paraffin rich reaction mixture (30.5 % vol. of C4H io ) can be a sign of the changed reaction mechanism because maleic anhydride is not formed in this case. It has been shown [16] by means of high-sensible system of analysis that even after reducing impulses (n-butane-inert gas) the only maleic anhydride is found. And, eventually, after the surface reduction with butane maleic anhydride does not form and C4 olefins appear. The olefins C4 oxidation gives the products that are intermediates of maleic anhydride formation: crotonaldehyde, 3-buten-2-on and furan [17]. On n-butane oxidation in standard conditions none of them can be found. The results presented in [1,18-20] show, that oxidation of butenes over (VOjzPzO? occurs with the lower selectivity than n-butane. Thus, the selectivity towards maleic anhydride is 70 % if it is produced from 1-C4Hg, 58 % - from C4Hs as compared to 82 % - from n-butane [20]. n-Butane activation in the very beginning involves methyl-groups. Grounded on the conception of primary dehydrogenation of n-butane, the authors of publications [10,11,21] proposed simultaneous hydrogen abstraction from two methylene groups. Later this version has been transformed into parallel activation of methylene and methyl groups [4]. The most serious confirmation of this assumption could be the experimental data published in [10]. As a result of comparison of kinetic parameters for C4H io , C4H6D4, C4H4Dg, C4D«> oxidation, the authors have concluded that butane absorption is irreversible over the surface of V-P-0 catalyst and hydrogen atom is abstracted from methylene group. On their opinion, this stage is a common one for the processes of partial and complete oxidation. Activation at methyl group forms about 5-10 %. The conclusions [10] are based on the assumption of homolytic C-H bond rupture. At the heterolytic break of the bond it should be taken into account that deuterium (more positive than hydrogen) introduction into alkane molecule leads to increase in electron density at C-H bonds [22]. When nucleofilic oxygen of the catalyst surface is abstracting proton from CH3 group one can expect some reduction of the process rate for CH3CD2CD2CH3 as compared to C4H io , and that is observed 251

in fact [10]. On CD3CH2CH2CD3 oxidation the difference in C-D and C-H binding energy is compensated by the higher positivity of deuterium and the conversion rates for C4H10 and C4H4D6 are almost equal. On the contrary to [10], the authors of [23] have shown, that when n-butane interacting with (VO)2P207 surface there is no absorption of hydrocarbon. The absorption takes place only from the mixture of hydrocarbon with air or oxygen. For the formed surface compound IR spectra show vibrations of -CH3 of much lower intensity than those for -CH2- and an absence of vibrational bands attributed to =CH2 and =CH- groups. The found CH3- groups belong to the surface hydrocarbon complexes which are not involved into maleic anhydride forming. The similar conclusion on proton abstraction from methyl groups by nucleofilic oxygen ions of the catalyst has been made in publications [2,4]. A high kinetic mobility of the methyl hydrogen in n-butane molecule as for isotopic exchange with D2 is shown in [25,26], where it runs much faster in CH3- groups (70-95 %) than in CH2. Calculations performed in [27,28] also confirm n-butane activation at the terminal methyl groups and difference in mechanisms of paraffin and olefins oxidation. Scheme for the paraffins conversion over V-P-0 catalysts surface It has been hypothized by us [1] that n-butane oxidation proceeds through formation of the intermediates:

x C4H90H ^ C4H10 v J------► C4H80 —► C4H60 —► C4H403 H0C4H80H

Results obtained on these intermediate products oxidation in the presence of (VO)2P207 [19] showed, that all the compounds had provided high selectivity towards maleic anhydride, except dihydrofuran. More than ten years later these results were confirmed in [29]. Over V-P-0 catalyst having PA/ =1.16 all the given products, except n-butanol, guaranteed the higher selectivity at over 90 % conversion as compared to that at 37 % butane conversion. The proposed by us scheme for n-butane molecule conversion over (100) surface is presented in Fig. 1. Adsorption of the hydrocarbon molecule takes place on two neighbouring groups consisting of paired vanadyl octahedra (Fig. 1a). Taking into account C’...C4 distance in n-butane molecule (0.39 nm) and C-H bond length it becomes obvious, that vanadyl oxygen having substantial negative charge [30] is able to bind hydrogen of the methyl groups, eventually abstracting it as a proton (Fig. 1b). Calculations [31] show that such hydroxyl groups are well retained on the surface. Hydrocarbon fragment freely moves within the same structural units and the terminal carbon atoms charged negatively are coordinating to vanadium atoms with positive charge (Fig. 1b). In so doing, it becomes clear the important role of the vanadium middle valency state larger than (4+) in the stationary catalyst. Vanadium ions (5+) much easier bind negatively charged fragment and being electron acceptor transforming themselves into V4*. It should be mentioned, that practically all the authors, which were defending an idea of two-phase-catalyst (Vs+OP04 and

i 252

a b

Fig. 1. n-Butane transformation on (100) plane of (VO^RzO?

(V^O^PzOy) before, consider now that vanadium ions (5+) do not form any separate phase but are uniformly spreaded around the surface (bulk) of the sample [32-34], Vanadium (4+) ions also can form the analogous Me-C bond but this one would be less strong. At this step, it is possible oxygen molecules incorporation with Me-O- O-C bond forming. Similar bonds could be assumed to appear on the stationary catalyst surface treated with oxygen after passing some impulses of hydrocarbon. Such peroxocomplexes of butane are oxidized at the temperatures of 300-400 °C preferably to CO% but some amount of butenes and tetrahydrofuran also are formed [35]. Another possibility to obtain peroxo-groups at available V4t-V5> pairs on the vanadyl pyrophosphate surface was described in [36]. However, the authors supported the conception of two methylene groups activation. Hydrocarbon fragment, being quite strong retained on the surface through Me-C bonds, is further loosing hydrogen becoming bonded to bridged V-O-V oxygen (Fig. 1b) to form -OH and -OH2 groups (Fig. 1c). The latter desorb from the surface as water and produce oxygen vacancies. If this takes place, a new fragment having double bond C2=C3 forms and it is connected to both vanadium and terminal oxygen (Fig. 1c). It could be supposed, that the further transformations of the fragment to a certain extend is depending on a probability of oxygen vacancies filling up either from 253

gas phase or at the cost of oxygen diffusion from the catalyst bulk. The rate of the diffusion is not high enough [37], that is why gas-phase oxygen is of essential importance for maleic anhydride forming [38]. The water molecules from gas phase can play the similar role and favour the selectivity increase [39]. Carbonyl groups are forming with participation of the terminal oxygen V-O-P. Hydrocarbon fragment changes its configuration and closes the cycle at the cost of terminal oxygen from vanadyl pair not being involved into the process before (Fig. 1d). It follows from the stated above, that elementary act of maleic anhydride formation from butane is closed in space containing three paired vanadyl groups. The limited possibilities for movement of hydrocarbon fragment out of the boundary of this structural unit must lead to the selectivity growth. Otherwise, complete oxidation of the fragment to COx takes place. The role of space limitations, confining the movement of the fragment beyond the boundaries of the structural unit, can be played by overstoichiometric phosphorus of the catalyst which is concentrated mainly on the surface. It connects with POH groups from the broken pyrophosphate groups of the surface [36] to form polyphosphate chains (Fig. 2). That is why it becomes clear the mechanism of the catalyst regeneration with phosphorus-organic compounds vapours and established by us phenomenon of the selectivity increase owing to barothermal treatment of the catalyst with H3PO4 or P2O5 vapours [40]. The similar role can be also played by phosphates of the promoters, that was noted in our work [41] and confirmed in [42,43],

Fig. 2. Vanadyl pyrophosphate (100) plane with overstoichiometric phosphorus

CONSEQUENCES FROM THE PROPOSED MECHANISM

Oxidation of n- and i-pentane. Taking into account the considered above principles and geometric correspondence, one should expect i-pentane oxidation with formation of methylmaleic anhydride, or maleic anhydride at the higher temperature. 254

Our results [44] support this expectation. In its turn, n-pentane oxidation and its activation at the first and fourth carbon atoms should lead to maleic anhydride and methylmaleic anhydride at the cost of methyl group isomerization (see [44,45] for details).

Fig. 3. Propane (a) and ethane (b) transformation on (100) plane.

Propane oxidation. The distance C?...C3 in propane molecule does not allow to get its two-centers activation similar to that in case of n-butane.An accordance with this, just after proton abstraction from one of the carbon atoms hydrocarbon fragment fixation on the surface occurs through V-C bond formation (Fig. 3a) followed by hydrogen abstraction from the second carbon atom and double bond formation. 71- Binding of the hydrocarbon fragment takes place. A probability for its desorption is not high. Rotation of the molecule around its fixation point and oxidation at the third carbon atom is more probable. Due to the presence of the terminal oxygen atoms in sufficient amount within rotation space it should be expected acrylic acid but not acrolein formation. The other direction of the process could be methyl group abstraction and acetic acid and carbon oxides formation. The experimental results presented in [46,47] shows that exactly these products are forming in the reaction of propane oxidation over V-P-0 catalysts. Ethane oxidation. Ethane, being much smaller in size molecule, can be oxidized preferably into carbon oxides and ethylene. Just after proton abstraction and binding of the hydrocarbon fragment with vanadium atom (Fig. 3b) an activation of the second carbon atom occurs. On the other hand, for ethane the activation within one group of paired octahedra can be also expected. In so doing, proton abstraction would take place on vanadyl oxygen and H" abstraction - on vanadium. The most probable product in this case could be ethylene and this really happens experimentally (48). 255

CONCLUSION The proposed model explains all the observed experimental results on n-butane oxidation and allows to foresee the primary products, which could be formed on other paraffins oxidation. On the basis of the model, besides the described reaction of methylmaleic anhydride production from n-pentane, the direct synthesis of tetrahydrofuran from n-butane has been realized.

REFERENCES 1. Pyatnitskaya A.I., Komashko G.A., Zazhigalov V.A., Belousov V.M., Bacherikova I.V., Seeboth H., Luke B., Wolf H„ Ladwig G., All-Union Conference on Mechanism Catal. Reactions (Russ), Nauka: Moscow, 2,286(1978). 2. Bordes E., Courtine P. J. Catalysis, 57,236 (1979). 3. Vanadyl Pyrophosphate Catalysts (Ed. Cent G.) Catal. Today, 16,1-147 (1993). 4. Cent! G„ Trifiro F., Ebner J.R., Franchetti V.M., Chem. Rev., 88, 55 (1988). 5. Kesteman E., Merzouki M., Taouk B., Bordes E„ Contractor R., Stud. Surf. Sci. Catal., 91,707 (1995). 6. Zazhigalov V.A., Belousov V.M., Pyatnitskaya A.I., Komashko G.A., Merkureva Yu.N., Stoch J„ Stud. Surf. Sci. Catal., 55,497 (1990). 7. Inumaru K., OkuharaT., Misono M., Chem. Lett., 10,1955 (1992). 8. Gai P.L., Kourtakis K„ Science, 267,661 (1995). 9. Gai P.L., Kourtakis K., Coulson D.R., Sonnichsen G.C., J.Phys. Chem., 101, 9916 (1997). 10. Pepera M.A., Callahan J.L., Desmond M.J., Milberger E.C., Blum P.R., Bremer N.J., J. Amer. Chem. Soc., 107,4883 (1985). 11. Busca G., Cavani F., Cent! G., Trifiro F., J. Catal., 99,400 (1986). 12. Matsuura I., Proc. 8 th Intern. Congr. Catal., Weinheim e.a„ 4,473 (1984). 13. Zazhigalov V.A., Haber J., Stoch J., Bacherikova I.V., Komashko G.A., Pyatnitskaya A.I., Appl. Catal., 134,225 (1996). 14. Sokolovskii V.D., Catal. Rev.- Sci. and Eng. 32,1 (1990). 15. Cent! G., Fomasari G., Trifiro F„ J. Catal., 89,44 (1984). 16. Gleaves J.T., Ebner J.R., KuechlerT.C., Catal. Rev.- Sci. Eng., 30,49 (1988). 17. Krylova E.K., Yakubson A.M., Blyakhman L.I., Titshenko A.I., Neftekhimia, 17, 286 (1977). 18. Seeboth H., Kubias B., Wolf H.. Lucke B., Chem. Techn., 28,730 (1978). 19. Zazhigalov V.A., Zaitsev Yu.P., Belousov V.M., Mechanism Catal. React., Novosibirsk, Nauka, 1, 82 (1982). 20. Miyamoto K„ Nitadori T„ Mizuno N.. Okuhara T., Misono M., Chem. Lett., 2, 302(1988). 21. Hodnett B.K., Delmon B., Ind. Eng. Chem. Fundam., 23,465 (1984). . 22. Ingold C.K., Structure and Mechanism in Organic Chemistry, Cornell Univ. Press, 1969. 23. Novak Ya., Gerej S., Heterogen. Catalysis, BAS, Sofia, 1,171 (1983). 24. Kruchinin Yu.A„ Mitshenko Yu.A., Nechiporuk P.P., Gel'bshtejn A.I., Kinetika i Kataliz, 25,396 (1984). 25. Bird R„ Kemball C„ Leach H.F., J. Catalysis, 107,424 (1987). 26. Brown R., Kemball C., Sadler I.H., Proc. 9 th Intern. Congr. Catal., 3, 1013 (1988). 27. Ziolkowski J., Bordes E., Courtine P., J. Catal., 122,126 (1990). 256

28. Ziolkowski J., Bordes E., Courtine P., J. Molec. Catal., 84, 307 (1993). 29. Guliants V.V., Benziger J.B., Sundaresan S., Stud. Surf. Sci. Catal., 101, 991 (1996). 30. Thompson M.R., Hess A.C., Nicholas J.B., White J.C., Anchell J., Ebner J.R., Stud. Surf. Sci. Catal., 82,167 (1994). 31. Witko M„ Hermann K., J. Molec. Catal., 81,279 (1993). 32. Zhang-Lin Y., Forissier M., Vedrine J.C., Volta J.C., J. Catal., 145,267 (1994). 33. Rodemerck U„ Kubias B„ Zanthoff H.-W., Wolf G.-U., Baems M., Appl. Catal., 153,217(1997). 34. Abon M„ Volta J.-C., Appl. Catal., 157,173 (1997). 35. Wilk R.D., Cohen R.S., Cemansky N.P., Ind. Eng. Chem. Res., 34,2285 (1995). 36. Agaskar P.A., DeCaul L, Grasselli R.K., Catal. Lett., 23,339 (1994). 37. Ovsitser O.Yu., Sokolovskii V.D., Catal. Lett., 17,239 (1993). 38. Zazhigalov VA, Zaitsev Yu.P., Belousov V.M., Wustnek N.. Wolf H„ React. Kinet. Catal. Lett., 24. 375 (1984). 39. Contractor R.M., Horowitz H.S., Sister G.M., Bordes E., Catal. Today, 37, 51 (1997). 40. Zazhigalov V.A., Kharlamov A.I., Haber J., Stoch J., Yaremenko V.E., Bacherikova I.V., Bogutskaya L.V., Stud. Surf. Sci. Catal., 111,207 (1997). 41. Zazhigalov V.A., Belousov V.M., Komashko G.A., Pyatnitskaya A.I., Merkureva Yu.N., Poznyakevich A.L., Stoch J., Haber J., Proc. 9 th Congr. Catal., 4, 1546 (1988); 5,493 (1988). 42. Matsuura I., Ozaki K., Kimura N., Yutani E., Stud. Surf. Sci. Catal., 92, 185 (1995). 43. Matsuura I., Ishimura T., Hayakawa S., Kimura N., Catal. Today, 28,133 (1996). 44. Zazhigalov V.A., Mikhajluk B.D., Komashko G.A., In this edition. 45. Zazhigalov V.A., Haber J., Stoch J., Mikhajluk B.D., Pyatnitskaya A.I., Komashko G.A., Bacherikova I.V., Catal. Lett., 37, 95 (1996). 46. Ai M„ J. Catal., 101, 389 (1986). 47. Komashko G.A., Khalamejda S.V., Zazhigalov V.A., In this edition 48. Lopez Nieto J.M., Zazhigalov V.A., Will be published. 257

DGMK-Conference "Selective Oxidations in Petrochemistry", Hamburg 1998

O. Seiferth, M. Bender, B. Dillmann, D. Ehrlich, I. Hemmerich, F. Rohr, K. Wolter, C. Xu, H. Kuhlenbeck, H.-J. Freund Fritz-Haber-lnstitut der Max-Planck-Gesellschaft, Faradayweg 4-6, D-14195 Berlin, Germany

ADSORPTION AND REACTION ON A POLAR SINGLE CRYSTALLINE CHROMIA SURFACE

Manuscript not available

I

DGMK-Tagungsbericht 9803,3-931850-44-7,1998 258