Catalyst Data With Rates

Applied Catalysis A: General 204 (2000) 5987

Oxide catalysts for ammonia oxidation innitric acid production: properties and perspectives

V.A. Sadykov a;, L.A. Isupova a, I.A. Zolotarskii a, L.N. Bobrova a,A.S. Noskov a, V.N. Parmon a, E.A. Brushtein b,

T.V. Telyatnikova b, V.I. Chernyshev b, V.V. Lunin ca The Boreskov Institute of Catalysis, Siberian Branch of the Russian Academy of Sciences,

Prospekt Akademika Lavrentieva 5, 630090 Novosibirsk, Russiab Joint Stock Venture State Institute of Nitrogen Industry (GIAP), 109544, Zemlyanoy Val Street, Moscow, Russia

c The Lomonosov Moscow State University, 119899, Vorobyevy Gory, Moscow, Russia

Received 28 April 1999; received in revised form 28 February 2000; accepted 29 February 2000

Abstract

This paper generalizes the results of long-term efforts aimed at research and development of industrial oxide catalysts forammonia oxidation in the nitric acid production within two-bed (Pt gauzesCmonolithic oxide layers) technology of the highpressure process.

Main factors determining performance of precious metals and oxides in the high-temperature ammonia oxidation areconsidered. The surface oxygen bonding strength determined by the surface atomic structure appears to be the most important.From this point of view, existing approaches to synthesis of mixed oxide systems including perovskites with controlled nitricoxide selectivity and good stability in the high-temperature process of ammonia oxidation are analyzed. Main features of thebulk oxide monolithic catalysts production technology and principles of a two-bed system design based upon the processmathematical modeling are briefly outlined. Proven economic benefits of this technology recently commercialized in Russiaat nitric acid plants are debated. 2000 Elsevier Science B.V. All rights reserved.

Keywords: Ammonia oxidation; Nitric acid production; Oxide catalysts; Honeycomb monoliths; Perovskites

1. Introduction

Nitric acid production is one of the large-scaleprocesses in chemical industry. The process involvesthe catalytic oxidation of ammonia by air (oxygen)yielding nitrogen oxide then oxidized into nitrogendioxide and absorbed in water. The desired product

Corresponding author. Tel.: C7-3832-344682;fax: C7-3832-343056.E-mail address: [email protected] (V.A. Sadykov).

yield essentially depends on the catalyst selectivityand of course on the operating conditions.

Ammonia usually comprises ca. 90% of the nitricacid production cost [1]. Therefore, the ammonia oxi-dation efficiency is a key factor for the nitric acidproduction.

In the first industrial process of nitric acid produc-tion by ammonia oxidation designed by Ostwald [2]at the beginning of the 19th century, pure platinumgauzes served as the catalyst. Later pure platinum wasdisplaced by platinumrhodium alloys, in which therhodium content varied from 5 to 10%. Despite many

0926-860X/00/$ see front matter 2000 Elsevier Science B.V. All rights reserved.PII: S0926 - 860 X (00 )00506 -8

60 V.A. Sadykov et al. / Applied Catalysis A: General 204 (2000) 5987

Table 1Optimum operating data for ammonia combustion [3]Pressure Gauze temperature (C) NH3 content (vol.%) Yield (%) Pt loss (g/t HNO3) Operating time (months)Atmospheric 810850 12.012.5 97.098.0 0.040.05 812Medium 35 atm 870890 10.511.0 96.596.5 0.100.11 46High 79 atm 920940 10.310.5 94.595.0 0.250.30 1.53

attempts of other type catalyst design to reduce thenitric acid production cost, platinumrhodium andplatinumpalladiumrhodium alloys still serve as themain catalysts.

Table 1 lists the optimum process parameters un-der various pressures [3]. In commercial plants, theconversion efficiency for nitrogen oxides productionranges within 9298% allowing no essential improve-ment. The plant operation is reasonably safe and reli-able, and process parameters are well optimized [4,5].

However, there is one serious operation problemrelated to platinum losses caused by the reactionmedium impact on the catalyst especially in high andmoderate pressure plants. For new gauzes, reactioninduced surface roughening is accompanied by anincrease of the catalyst activity [6]. For example, inhigh pressure (810 atm) converters, the maximumnitrogen oxide yield is attained within 35 days, thusresulting in a significant loss in the acid productionduring start-up. As was shown by Farrauto and Lee[6], the increase of the gauze surface by a specialpretreatment during its manufacture shortens thetime of approach to the optimum regime. However,in all cases, the nitrogen oxides yield continuouslydecreases during the run. Therefore, one needs toreplace the inactivated platinum gauzes by new ones,when reaction selectivity falls below the commer-cially acceptable level. Moreover, some gauzes neverprovide high selectivity, since they completely losetheir activity in a short time run [3].

In the USA, nitric acid is mainly produced in highpressure (810 atm) converters. In Russia as well asin other countries using diluted nitric acid produc-tion technology developed in the USSR, the share ofhigh and moderate pressure converters is not less than50%. Typically, from 10 to 30 gauzes made of thePt(Pd)Rh alloy are arranged on a supporting grid ina reactor chamber with diameters up to 5 m. There-fore, the catalyst loading is quite expensive.

Further, as ammonia is oxidized, Pt is slowly lostfrom the gauze, mainly in the form of more volatileoxides, though small Pt particles can also detach fromgauzes strongly corroded to the end of the run (me-chanical losses) [7,8]. The rate of loss depends uponthe type of plant being higher for high-pressure plants(more than 1 g/t of ammonia converted). The cost oflost Pt is the second largest expense of the opera-tion, exceeded only by the cost of ammonia feed-stock [9]. To recover some of Pt, various approacheswere suggested including filters made of various ma-terials to mechanically catch and retain solid particlesof Pt, chemical absorbents on the base of supportedmolten salts or alkaline earth compounds [7,8] etc.The most successfully used technology of Pt recov-ery is based upon the application of gauzes made ofPd alloys [10]. The recovery alloys are usually usedin the form of multiple sheets of woven gauze placedimmediately downstream from the catalyst gauze. Therecovery reactions consist in the interaction betweenPtO2 and Pd on the getter gauze surface followed bythe solid-state diffusion of Pt into the bulk of wire thusforming a PtPd alloy. Early technology [10] ensured3540% recoveries, while subsequent application ofgauzes with a higher Pd content and improved geome-tries of the individual sheets practically doubled therecovery efficiency [11]. However, the recent surge inthe Pd price now exceeding that of Pt makes the ap-plication of this technology too expensive.

These facts stimulate the research towards at leastpartial substitution of Pt gauzes by some less expen-sive catalytic materials.

With this regard, it is of interest to analyze and gen-eralize the results of recent works aiming at designof essentially less expensive oxide catalysts for thispurpose. Thorough systematic studies were performedmostly in the USSR (especially in the State Instituteof Nitrogen Industry (GIAP), Moscow) and in China.Starting from the early 1960s, at almost all nitric acid

V.A. Sadykov et al. / Applied Catalysis A: General 204 (2000) 5987 61

plants in the (former) USSR operating under atmo-spheric pressure, a two-bed ammonia oxidation system(one platinumpalladiumrhodium gauzeCpelletizedoxide catalyst bed) is used in practice [7,12].

For high-pressure plants, design of two-bed cat-alytic systems meets more difficult problems owingto complexity of the ammonia oxidation processes,the impact of flow hydrodynamics, and to evolutionof both platinum and oxide catalysts under the re-action conditions. These problems were overcomedue to united efforts of GIAP, Boreskov Institute ofCatalysis (Novosibirsk) and Chemical Department ofLomonosov Moscow State University in designinghoneycomb catalysts for the high temperature oxida-tion processes occurring at short contact times [13].

Due to a complex interrelation of all aspects of theproblem, it seems reasonable to consider them succes-sively. So, we shall consider: (1) main peculiarities ofammonia oxidation on platinum gauzes; (2) factors de-termining oxides efficiency in this reaction; (3) mainapproaches to technology of oxide honeycomb mono-lith catalyst production; (4) principles for designingthe two-bed catalytic system based on mathematicalmodeling of ammonia oxidation on gauzes and honey-comb oxide catalysts and (5) the industrial applicationof two-bed systems for ammonia oxidation.

2. Kinetics of ammonia oxidation to NO onplatinum metal gauzes

Despite numerous studies related to this importantlarge-scale process, there is no generally acceptedkinetic law to describe it. Ammonia oxidation on plat-inum is considered as a classic example of stronglyexothermic heterogeneous catalytic reaction with crit-ical regimes of heat ignition and extinction [1416].All attempts to suppress the critical regimes and to ob-tain intrinsic kinetics for concentrations and tempera-tures used in industry failed [17,18]. Moreover, for thisprocess, a set of isothermal critical phenomena causedby ignition and extinction were also discovered [16].

At the same time, these specific kinetic problemsdisappear at temperatures below 300C, when molec-ular nitrogen and N2O are the main products of am-monia oxidation. For example, the chemical kineticsof the low temperature ammonia oxidation on plat-inum wire at ammonia pressures ranging from 0.05

to 0.2 atm and at oxygen pressure ranging from 0.2 to0.95 atm was studied in [19]. According to this study,the experimental data are well described by the firstorder kinetics with respect to the surface coverageby oxygen and ammonia pressure, activation energybeing around 30 kcal/mol (1 kcalD4.18 kJ). For moreoxidized product nitrous oxide, the selectivity isproportional to the surface coverage by oxygen. Withincreasing contact time, the nitrous oxide yield de-creases, which is explained by N2O interaction withammonia on the catalyst surface.

Under industrial operating temperatures (800950C), the oxidation of ammonia on platinum gauzesproduces mainly molecular nitrogen and NO. Theseproducts are formed on the catalyst surface by reac-tions (1) and (2)NH3C1:25 O2DNOC1:5 H2O .54 kcal=mol/ (1)NH3C0:75 O2DN2C1:5 H2O .76 kcal=mol/ (2)

As for the side processes which affect the pro-cess selectivity, there is no agreement between theresearches. Most researchers believe that NO decom-position to N2 and O2 and ammonia interaction withnitrogen oxide yielding N2 and water provides thelargest contribution (reactions (3) and (4))NO D 0:5 N2 C 0:5 O2 .22 kcal=mol/ (3)

NH3 C 1:5 NOD 1:25 N2 C 1:5 H2O .104 kcal=mol/ (4)

At high temperatures, the total ammonia oxidationrate is limited by the rate of ammonia diffusion tothe catalyst, which is typical for the processes withexternal mass transfer limitation [20,21]. However, theratio of rates of reactions (1) and (2) is determined bythe reaction mechanism, and essentially depends onthe composition of platinum gauzes (first of all, on theRh content), on ammonia/oxygen ratio in the feed andmass transfer intensity [2024].

Most researchers consider the basic reaction as apure heterogeneous one, being mass-transfer limitedat atmospheric or elevated pressures [4,5,1728].

Apelbaum and Temkin studied the reaction underkinetic control at low (about 0.01 Torr) pressures onan electrically heated platinum wire 10 cm long and0.1 mm in diameter [29]. They found that NO and N2


desorb from the platinum surface but are never formedfrom other primary products in the hot gas mixture aswas considered earlier [3032]. More detailed analysisof experiments [29] shows that the reaction order inammonia is 1.37 and 1 for oxidation towards N2 andNO, respectively.

Fogel et al. [26,27] used mass-spectroscopy to studythe ammonia oxidation at a pressure 104 Torr.The authors believe the reaction to proceed via theLangmuirHinshelwood mechanism without stableintermediates

NH3.S/ C O2.S/ ! NO C H2O C H.S/ (5)and/or

NH3.S/ C O.S/ ! NO C H2 C H.S/ (6)NH3.S/ C NO.S/ ! N2 C H2O C H.S/ (7)

According to the data obtained by Kuchaev andTemkin [33], there are two distinct kinetic regions forammonia consumption rate. In the first region, the re-action rate is proportional to the ammonia partial pres-sure and does not depend on the oxygen content. Inthe second region, the reaction rate is proportional tothe oxygen partial pressure and does not depend on theammonia content. The authors conclude that ammo-nia molecules react in both ways upon the impactfrom the gas phase or in the adsorbed state.

For ammonia oxidation on Pt, Rh and Pd wiresat 0.11.0 Torr, other authors [34,35] concluded thatwithin the 200900C temperature range, the reac-tion rate may be described by a model based onschemes (5)(7) suggested by Fogel et al. However,as the reaction temperature increases to 1500C, itis necessary to take into account the direct ammoniadecomposition followed by the hydrogen oxidation.

The literature data related to the nitrogen oxidedecomposition (3) are contradicting as well. Someauthors report that the NO yield passes through themaximum at a definite gas flow rate or respectivecontact time [4,28,36,37]. The usual explanation isthat nitrogen oxide is unstable, and decomposes atlong contact times. However, Apelbaum and Temkin[20,21] proved that decomposition of NO does not oc-cur under the usual operation conditions of ammoniaoxidation, since oxygen suppresses this reaction.

As for reaction (4), it may occur via both homoge-neous (gas phase) or heterogeneous (catalyst surface)

mechanisms. The rate of heterogeneous ammonia oxi-dation far exceeds that of the homogeneous route [38].

According to literature [3941], the homogeneousreduction of nitrogen oxides by ammonia is mostly ef-ficient at 9001000C. Hence, the temperature rangesfor the ammonia oxidation and homogeneous reaction(4) overlap. Reaction (4) was found to be of a chaintype involving many stages and intermediates [42] be-ing well described by the second order equation withrespect to reagents [43].

To describe the apparent reaction rate dependenceon reaction parameters in industrial conditions, am-monia diffusion rates are used [4,7,28,4449]. Empir-ical expressions with numerical coefficients obtainedfrom the experimental data are used to estimate theyield of nitrogen oxide [4,7].

Among the papers on the ammonia oxidation mod-eling, those guided by Beskov are to be distinguished[5053]. In these papers, the steady state rates ofammonia consumption are described on the bases ofroutes (1)(4). All reactions are assumed to be hetero-geneous. The equations and kinetic model parameterswere derived from the literature data and from theexperimental results of Atroshchenko and Kargin[4,28,38] obtained at conditions of the mass transferlimitation. According to estimations by Beskov et al.[52], in the temperature range 8001000C, the rateof NO decomposition into N2CO2 is only 23 or-ders of magnitude lower than the rate of reaction (1).Taken into account the results of [20,21], the rate ofreaction (3) appears to be overestimated. In contrast,authors [52] clearly underestimated the reaction (2) asbeing 56 orders of magnitude slower as comparedto the reaction (1).

3. Ammonia oxidation on oxide catalysts

3.1. Kinetics peculiarities

3.1.1. Low temperature reactionThe ammonia oxidation on oxide catalysts in an

oxygen excess was studied in detail at temperaturesbelow 380C. In this case, there is no mass trans-fer limitation, and molecular nitrogen and nitrousoxide are the main reaction products [5458]. Theresults obtained and reaction mechanisms consideredare reviewed [59,60]. Among simple transition metal


oxides, in the ammonia oxidation, the most active areCo3O4, MnO2 (Mn2O3), Cr2O3, CuO. Moderatelyactive are NiO, Fe2O3, V2O5, ZrO2, CeO2, La2O3.The activity and selectivity towards nitrous oxideusually correlate, both decreasing as the surface oxy-gen bonding strength increases and the oxygen andammonia pressures decrease. The reaction kineticsis rather well described by the reaction mechanismsuggested for ammonia oxidation on platinum [19].However, this mechanism does not take into accountthat oxygen species on the surface of oxide catalystsare not uniform by bonding strength and reactivity,which also depend on the sample genesis. There wereno data on the concentrations of intermediates, theirrelative stability and reactivity.

In the recent decade, mechanisms of ammoniaoxidative transformations were essentially revised,when selective reduction of nitrogen oxides (SCR) byammonia in an oxygen excess was studied in detailsfor environmental applications. This process is nowwidely used for NOx abatement from the flue gases ofpower plants and industrial sources. Here, ammoniaoxidation to nitrogen oxides is an undesired side reac-tion, attracting thus a keen research attention. Severalreviews on the subject were published [6163] notmentioning a lot of original papers clearly beyond thescope of our present analysis. These experimental dataappear to be quite reliable, since they were obtainedusing in situ IRS, transient response technique, andisotope methods. However, for various oxides, dif-ferent routes for the surface ammonia transformationwere suggested. This may be caused by the differentsurface chemistry of oxides as well as by inherentdifficulties in identification of surface intermediatesand estimating their reactivity. Thus, for supportedvanadium oxide systems, the route yielding molecu-lar nitrogen is usually ascribed to the interaction ofweakly bound or gas phase NO (NO2) molecules withstrongly adsorbed ammonia species such as NH4C,NH2 [63] or even with the products of their interactionsuch as hydrazine [64]. At the same time, there aremany facts in favor of strongly bound nitritenitratecomplexes be responsible for N2 generation via theirinteraction with ammonia [6570].

Despite the differences in the suggested mecha-nisms, various authors agree that strongly bound sur-face intermediates are required for molecular nitrogento form on the oxide catalyst surface. A temperature

increase reduces the catalyst coverage by such inter-mediate species, and the routes of molecular nitrogenand nitrous oxide formation are suppressed, while am-monia oxidation to NO dominates.

3.1.2. High temperature reactionHigh temperature ammonia oxidation on pure and

promoted bulk oxides of transition metals is discussedin many papers [71100]. The catalytic performanceof mixed oxides of transition, rare earth and alkalineearth metals is considered in [99113].

The high temperature ammonia oxidation on ox-ide catalysts is well analyzed [7]. Note that underindustrial conditions, when NO dominates among thereaction products, the ammonia oxidation rate on theoxide systems as well as on platinum gauzes is lim-ited by ammonia diffusion from the gas phase to thecatalyst surface [7173,89,90]. This means that thereaction selectivity towards NO attains its maximum,when oxygen nearly completely covers the catalystsurface, and there is almost no ammonia or its partialoxidation products on the surface.

To estimate the ratio of the ammonia oxidation rateson oxides and platinum, the ammonia threshold loadscan be compared. A load is defined as the ratio of thevolume gas velocity (referred to 0C and 1 atm) to thecatalyst cross-section. A threshold load corresponds toa load when the process extinction is observed [71,72].As the linear rate of the gas flow increases, the ap-parent thickness of the diffusion layer decreases, andat some value, the heat evolution rate becomes lessthan the heat removal rate, and, thus, the catalyst coolsdown and extinguishes. The threshold load depends onthe surface reaction rate constant, and, thus, on the cat-alyst nature. According to [71], at 1 atm and 10 vol.%NH3 in air, the threshold load for a 0.09 mm diame-ter platinum wire is close to 30 000 L/cm2 h. For the4 mm cobalt oxide pellet, it is 10 000 L/cm2 h, otherconditions being the same [72].

The dependence of the threshold load on the Co3O4catalyst grain size, ammonia and oxygen concen-tration and total pressure studied by the authors of[84] revealed the following equation for the ammoniaoxidation rate:

r D K[NH3]0:36[O2]0:14 (8)

where K is a constant, [NH3] and [O2] are the


concentrations of ammonia and oxygen near the cat-alyst surface, respectively. The reaction activationenergy is close to 9 kcal/mol. This equation character-izes the overall ammonia conversion to all productssuch as nitrogen oxide and molecular nitrogen.

A similar method was used [74,89] for the catalystsbased on iron oxides. According to [89], the lower re-action ignition temperature, the higher maximum NOselectivity estimated by the standard procedure [4],the lower effective activation energy and reaction or-der, hence, the external diffusion limitations are moresevere. So, the reaction selectivity towards NO underindustrial conditions may serve as a measure for thecatalyst activity. Practically in all studies related to ox-ides, the performance of catalysts was estimated usingthis approach.

According to the standard procedure [4], suchstudies are usually carried out under atmosphericpressure in a 40 mm bed, grains size ranging within1.52 mm. At 750800C (low-temperature samples)and at 800900C (high-temperature samples), thelinear gas flow rate ranges from 0.2 to 2 m/s, contacttime from 103 to 6102 s, ammonia content from1.0 to 1.5 vol.%. Under such conditions, the ammoniaconversion is complete, but NO selectivity stronglydepends on the oxide nature and samples genesis.As a rule, selectivity towards NO goes through themaximum depending on the temperature.

Almost all authors agree that maximum selectiv-ity is shown by such oxides as Co3O4 (94% yield),a-Fe2O3 (90% yield), Bi2O3 (9093%), manganeseoxides (up to 80% yield). Moderate selectivity isshown by NiO (3050%), CuO (4050%), PbO2(50%), rare-earth metal oxides (1050%). Oxides ofaluminum, tungsten, molybdenum, tin are not active.

Similar selectivity series for oxides were obtainedon ammonia oxidation under pressure up to 8 atm[7882,85].

Taking into account that under similar operationconditions, on the most active oxides the ammoniaconversion is complete at contact times of 102 s,while on platinum gauzes it is achieved at contacttimes of 104 s, the specific catalytic activity of oxidecatalysts appears to be at least 2 orders of magnitudelower than that of the platinum group metals. Indeed,according to [7], at 900C, the effective first orderrate constant (in s1) is 24 600 for the platinum cata-lyst and 73 for the iron oxide catalyst. This estimation

agrees well with the same scale of the reaction rate dif-ferences estimated for oxides and metals in ammoniaoxidation to nitrous oxide under the kinetic control attemperatures around 200C [59]. An essentially higheractivity of metals in comparison to that of oxides atthe same strength of the oxygen surface bond impliesa significant role of the ammonia NH bond activation(its cleavage or weakening) on the metal surface.

3.2. Factors determining catalytic performance oftransition metal oxides

All earlier attempts to find relationships betweenthe oxides activity (selectivity towards NO) and theirphysical properties such as melting temperature, bandgap, conduction type or electron work function failed[7]. At the same time [76], a good correlation was re-vealed between the activity, the activation energy ofoxygen homo-exchange, and the surface oxygen bond-ing strength estimated by the method of thermochem-ical cycles [59]. The iron oxide, however, stands apartfrom these relationships. Nevertheless, all these effortsrevealed the significance of oxygen bonding with thesurface for ammonia oxidation on oxides.

An important breakthrough in understanding themolecular-scale factors controlling the oxides activityand selectivity in the ammonia oxidation was achievedin works of Ryabchun and co-workers [9598]. Forthe industrial iron oxide catalyst, the detailed studiesusing such methods as thermal desorption, SIMS,X-ray diffraction, epitaxial decoration, laser spec-trometry, unsteady state kinetic experiments with amicro-reactor attached to the high-vacuum surfacescience installation have been carried out. The mostimportant conclusion from all these studies was theevidence for a micro-heterogeneity of the workingcatalyst surface under the reaction conditions. In un-steady state conditions of the surface reduction byammonia, the maximum selectivity towards NO isattained, when 35% of the oxygen monolayer areremoved from the surface accompanied by its recon-struction. This reduction process is of a topochemicalcharacter, and proceeds through the expansion of re-duced domains from the points of their nucleation.Nitrogen dissolved in the catalyst may play somerole in the modification of reduced domains. Oxygenadsorbed on such domains is characterized by theactivation energy of desorption 50 kcal/mol, which


coincides with that for the most active catalyst platinum. In the oxygen excess, the catalyst surface ismostly covered by such an oxygen, though there arealso less (adsorption heat, 2030 kcal/mol) and morestrongly bound oxygen species. A weakly bound oxy-gen favors the oxidation of ammonia to molecularnitrogen, while too strongly bound oxygen is inactive.Independently measured rates of the surface reduc-tion and reoxidation were found to be less than thesteady state rate of the overall catalytic process, thussuggesting the reaction to proceed by an associativemechanism.

Hence, these results help to elucidate the factors de-termining the activity and selectivity of oxide catalystsin the high temperature ammonia oxidation to NO. In-deed, to provide an almost complete surface coveragewith oxygen at operating temperatures ranging within800900C, the oxygen-surface bond must be strongenough. Note that a reasonably high oxygen bondingstrength (the adsorption heat 5060 kcal/mol O2) isusually required not for deep oxidation catalysts, butfor the catalysts of partial (mild) oxidation such astransition metals molybdates or vanadates [114,115].Apparently, the optimum oxygen adsorption heat mustdecrease, if operating temperature decreases.

If at a given temperature the strength of oxygenbonding to the surface is lower than the optimumvalue, the oxygen coverage will decline generating co-ordinatively unsaturated surface sites such as reducedcations of transition metals, which will stabilize theadsorbed ammonia species opening the route to molec-ular nitrogen generation.

At the same time, the oxygen bond strength mustnot be too strong. The surface oxygen must be able toreact with the ammonia molecules with a rate exceed-ing the rate of their adsorption under given conditions,thus providing the transfer of the reaction into the ex-ternal diffusion regime, when the surface coverage byammonia or its intermediates is small. The strength-ening of the oxygen bond with the surface will alsofavor stabilization of strongly bound nitritenitratecomplexes [116]. As a result, the oxygen-surfacebond strength exceeding the optimum one will pushreaction to routes yielding molecular nitrogen.

This idea allows to understand why for all oxidesystems as well as for platinum group metals, the tem-perature dependence of NO yield goes through themaximum [7]. Further, the specificity of the platinum

group metals performance can be explained. Thus, forexample, at 850900C (the typical operating temper-atures for platinum), on bulk palladium, ammonia ismainly oxidized to molecular nitrogen, which is ex-plained by a too low surface oxygen coverage dueto easy dissociation of the surface PdO [117,118]. Atthe same time, the rhodium oxide, covering the sur-face of deactivated platinum gauzes, is practically in-active [3], which may be explained by a too strongrhodiumoxygen bond.

Note that besides the vital importance of theoxygen-surface bonding, a specific activation of theammonia molecule (vide supra) may also play asignificant role. Probably, for oxides, high-chargelow-coordinated surface cations Lewis acid sites,including those produced due to the oxygen and/orwater desorption from the surface, are of essentialimportance. At present, however, an analysis of thisfactor is not possible due to the absence of reliabledata on the properties of such sites on the small sur-face area oxide catalysts characterized in the high-temperature ammonia oxidation.

Analysis of the atomic-scale reasons for variousoxygen species appearing on the catalyst surfaceshould be based upon the discrete nature of the en-ergy spectrum of oxygen species on the oxide surface.This means that there are always several surface oxy-gen species differing by the strength of their bonding(heat of adsorption). These species appear as separatemaxima in the spectra of the oxygen thermal desorp-tion [119121] or as steps on the dependencies ofthe heat of oxygen adsorption versus coverage regis-tered using isosteric methods [122,123], calorimetryor solid electrolyte potentiometry [114,121].

For a given oxide, the existence of different surfaceoxygen species is caused by the atomic structure of itsfaces, by the appearance of point (vacancies) and ex-tended (steps) surface defects, including those emerg-ing in the vicinity of outlets of bulk extended defects[124128].

In the last decades, a lot of data on the atomicstructure of the oxide single crystal faces were accu-mulated using such methods as LEED, STM, AFM,photoelectron diffraction (see, e.g.[129131]). Forfinely dispersed oxides, the atomic structure of theirsurfaces was studied with the help of such methodsas IRS of adsorbed test molecules [132], ESR of ad-sorbed nitroxides [133], TEM [134136], ISS [137],


Table 2Calculated and experimental oxygen desorption heats (kcal/mol) [124]Oxide 1H calculated 1H experimental

Regular sites Defective sites

CuO MO: 9 MO near (0 0 1) twin outlet onto (1 0 1) face: 3 10; 25M2O: 60 50

Co3O4 MO(Oh): 4050 MO near outlet of (1 1 0) stacking fault onto (1 1 0) face: 15 15MO(Td): 6070 40M2O: 130 120

Fe2O3 MO: 3540 (Oh) MO near outlet of (0 0 0 1) stacking faultonto (0 0 0 1) and (1 1 2 0) face: 1830

20; 40

6070 (Td) 60M2O: 130 120

MnO2 MO: 3240 MO on (1 1 0) face, relaxation of surface MnObonds caused by oxygen atom removal: 14

16; 30

M2O: 60 45

neutron diffraction in the argon adlayer [138,139]. Thedevelopment of quantum-chemical semi-empirical[125,126,128,140142] as well as ab initio [140150]methods of the surface energy and adsorption heatcalculation allowed analysis of possible structures ofadsorption centers capable to retain oxygen with thestrength required for the process considered. Someresults are given in Table 2.

According to [124], in an oxidative medium, thewell developed densely packed (1 1 1) faces of spineloxides and (0 0 0 1) faces of corundum-type oxidesare mainly covered by strongly bound non-reactivebridged M2O oxygen species, whose adsorption heatsrange within 100120 kcal/mol. The chemically inertstoichiometric (0 0 0 1) face of hematite is not able todissociate even water molecules to produce hydroxyls[151].

The structures of such faces annealed in high vac-uum and thus weakly reduced, most likely correspondto localization of tri-coordinated cations in the upperlayer [152158]. Theoretical analysis of these struc-tures using non-empirical and semi-empirical methods[143149,159] has revealed that they must experiencea strong relaxation forcing the cation to sink into thesurface oxygen layer. At the same time, STM data forthe (0 0 0 1) face of thin hematite layer grown epitaxi-ally on the Pt (1 1 1) suggest the existence of domainswith the upper layer represented by both oxygen andiron atoms [150]. For the oxidized surface of corun-dum, AFM [160] shows that the upper layer ends by

the hexagonal package of the oxygen atoms with in-teratomic distances close to those in the bulk.

For oxide samples prepared from solutions of cor-responding salts, IRS of adsorbed CO [124,125,127,132,161166], IRS of hydroxyls [167169], neutronscattering in the argon adlayer [138,139], 4He ionsscattering on the surface [137], ESR of nitroxideradicals applied to diamagnetic oxides (a-Al2O3)[133] show that the densely packed faces of spinelor corundum-like oxides do not contain tri- ortetra-coordinated cations in the regular positions.Cations in such coordination are usually present insmall amounts, which strongly depend on the sam-ple genesis, allowing to assign them to such surfacedefects as steps or subsurface stacking faults. For sto-ichiometric iron and aluminum oxides, configurationsof these faces appear to end with the oxygen layercomprising the oxygen ions and hydroxyl groups[167,169]. Some experimental data [170173] andtheoretical analysis [130141] reveal the hydroxylgroups to play a significant role stabilizing denselypacked faces in corundum-like oxides (basal face)and in oxides of rock salt structure (the (1 1 1) face).

In CuO (tenorite) and MnO2 (pyrolusite), M2Obridged oxygen species bound to the regular sitesof the most developed faces have almost optimumbonding strength (5060 kcal/mol) [124,125,128].

Under oxidative conditions, for corundum-likechromium oxide, the densely packed basal (0 0 0 1)face is usually covered by chromate groups (tetrahe-


drally coordinated Cr6C cations), where the terminaloxygen has the adsorption heat close to 60 kcal/mol[124,174176]. Note that in this case, the basalface cations are tetra-coordinated, that agrees wellwith predictions based on the data obtained withhigh-vacuum methods (vide supra). Apparently, un-der oxidative conditions, the chromium cation easilychanges its oxidation state thus favoring the thermo-dynamic stability of such configuration, which is notpossible in the case of iron or aluminum oxides.

For oxides with the close-packed oxygen sublattice,the terminal (MO) oxygen species with the adsorp-tion heat close to the optimum value for ammonia ox-idation to NO (4060 kcal/mol) are mostly located onopen faces: (1 0 0) and (1 1 0) types in spinel structures(1 1 2 3) and (1 0 1 2) in corundum ones. In the caseof oxides obtained via the thermal decomposition ofprecursors (salts, hydroxides) synthesized from watersolutions, these faces are not well represented, thoughtheir exposition increases with an increase of the cal-cination temperature. Thus, according to [178], for asample of cobalt oxide prepared from the nitrate salt,as the calcination temperature increases from 400 to800C, the shape of particles changes from plateletswith the mostly developed (1 1 1) faces to pyramidswith the mostly developed (1 1 0) faces.

Rhombohedral (1 0 1 2) and pyramidal (1 1 2 3)faces of hematite are the natural growth faces [148],and in disperse samples, their share also increaseswith increasing calcination temperature, since theirsurface energies are lower than that of the basalface [126]. On rhombohedral and pyramid faces, theterminal oxygen species are linked to cations in oc-tahedral or tetrahedral coordination [126,129131].Moreover, the terminal oxygen species with a similarbonding strength may appear near steps formed dueto the high-temperature reconstruction of prismaticand basal faces [124,130,160,179].

In stoichiometric MeO oxides of the rock salt struc-ture, the (1 0 0) faces are not able to adsorb oxygenwithout changing the cation charge, though they docontain penta-coordinated cations required for suchan adsorption. As a result, the stoichiometric perfectface of NiO is practically inert to reduction by CO orhydrogen at moderate temperatures [129131,180].For dispersed nickel oxide, such faces are well de-veloped only in samples obtained via plasma ther-molysis of nitrate solution [181]. At the same time,

in samples obtained via traditional thermal decom-position of nitrate solution, the particles morphologyis mainly determined by (1 1 1) faces covered bystrongly bound oxygen species probably stabilized byhydroxyl groups [182].

In oxides with a broad range of the surface/bulknon-stoichiometry (cobalt and manganese spineloxides, nickel oxide), the terminal oxygen speciesappear on the densely packed faces when cationvacancies are generated due to the oxygen excess[124,183186]. This excess oxygen is removed at400500C. Therefore, at temperatures typical forthe ammonia oxidation process, such generation ofterminal oxygen species will hardly be of importancefor the application in concern.

In transition metal oxides, the most weakly boundoxygen species (heat of adsorption 1030 kcal/mol)are mainly located on coordinatively unsaturated clus-ter sites appearing in the vicinity of outlets of extendeddefects such as dislocations, twins, stacking faults andintergrain boundaries [124128]. Usually, the densityof extended defects is determined by the oxides prepa-ration procedure. The dislocation network in the sur-face layers of oxide particles is also generated duringphase transitions, when, e. g. magnetite phase nucleiemerge during the hematite reduction [124,125,127],or when Co3O4 transforms to CoO at high tempera-tures [178]. Such cluster sites may also appear due toreconstruction of the (1 0 0) and (1 1 0) faces of spinel[187190], and (1 1 1) face of rock salt-like oxides[191] caused by a loss of oxygen and/or hydroxyls.Though weakly bound oxygen species determine theactivity of some transition metal oxides [124] andperovskites [192199] in oxidation of CO and hy-drocarbons at temperatures up to 400C, in the hightemperature ammonia oxidation those sites obviouslywill cause a lower NO selectivity owing to their in-ability to keep oxygen at operating temperatures. In-deed, after oxygen removal, clusters of coordinativelyunsaturated cations will appear at the surface beingable to retain the intermediate complexes of ammoniaoxidation, and, hence, to accelerate the N2 generation(vide supra). Moreover, extended defects are knownto be the potential sites for the nucleation of reducedphases [200], which are less selective towards NO [7].

For perovskite systems, the effect of the synthe-sis procedure on the development of certain facesis far less studied than for simple transition metal


oxides. Depending on the system and synthesis con-ditions, the perovskite particles morphology may bepresented by spheres comprised of the cubic micro-grains [195,201,202], platelets including hexagonalones [195,203], prisms with sharpened tops, pyramids,octahedrons [203,204]. These data combined with theresults of the selective electron diffraction [201205]and HREM [204,205] imply that the surface of par-ticles may be represented by the (1 0 0), (1 1 0) and(1 1 1) faces as indexed in the cubic space group.

Since the perovskite structure is formed by thedense packing of mixed lanthanum-oxygen layers,transition metal cations being located between thelayers in oxygen octahedrons, each type face mayhave a different termination. Thus, along the (1 0 0)direction, alternated are layers LaO and MeO2, along(1 1 0) layers O4, MeO and La4Mn2O2, along(1 1 1) densely packed lanthanum-oxygen layers inthe cubic sequence abc with the layers of transitionmetal cations in between.

In general, there are no data which face terminationcomprise the surface. Considering the XPS and SIMSdata [195,199,201,206,207], one may conclude thatin most cases the ratio between the lanthanum cations(or sum of lanthanum and alkaline-earth cations forsamples with such a substitution in the lanthanum sub-lattice) and transition metals in the subsurface layerdiffers insignificantly from that in the bulk, thoughmay vary depending on the conditions of sample syn-thesis and pretreatment. Thus, a reductive treatmentfavors segregation of transition metal cations, whilean oxidative treatment in humid atmospheres enrichesthe surface with lanthanum and alkaline-earth cations[208,209]. Hence, the most developed faces of per-ovskites appear to be represented by all possibleterminations with a nearly equal probability.

On any perovskite face, the oxygen atoms boundto lanthanum cations in the regular positions seem tobe chemically inert at all temperatures of catalysis.The faces of the (1 1 0) type shall be mainly coveredby bridged oxygen species bound to either two man-ganese cations or manganese and lanthanum cations.The (1 0 0) faces with MeO2 composition may containpenta-coordinated cations of transitions metals, whichare sites for oxygen adsorption in the terminal MOform. From stoichiometric lanthanum cobaltites, fer-rites, nickelates and manganites, such oxygen speciesare usually desorbed at 700800C [210213].

In the perovskite matrix, owing to diluting ef-fect of lanthanum cations, clustered transition metalcations-sites for ammonia oxidation to N2 (vide supra)will hardly appear in the vicinity of extended defects.Indeed, in the rock salt-like oxides, at a mono-atomicstep on the (1 0 0) face, the lifting of screening byoxygen anions allows interaction between cations ly-ing at the top and bottom of step. On the contrary, inthe perovskite structure, the step bottom is comprisedby the lanthanumoxygen layer allowing no interac-tion between the transition metal cations [129,131].Clustering becomes possible only in the vicinity ofsuch extended defects as clusters of oxygen vacancies,twins, or intergrowth structures where the local en-richment by transition metal cations occurs [125,196].

If in perovskite-like oxides some lanthanum cationsare replaced by the alkaline-earth cations, variousdefects are generated depending on the degree ofsubstitution and the nature of transition metal cation.Among them, cation and/or anion vacancies, orderingof modifying cations in some planes, microstrains andother extended defects like micro-domain interfaces,intergrowth of phases with different composition,cooperative distortion of lattice, which may evenproduce phases of other symmetry and/or structurewere revealed [113,192,199,201,205,206,211219].Oxygen vacancies appearing due to this partial sub-stitution will affect the catalyst performance only ifthey emerge on the surface. It certainly takes place forpartially substituted lanthanum manganites, ferrites,cobaltites and nickelates, where appearance of oxy-gen species desorbing at temperatures below 500Cand probably related to ion-radical O species boundwith the surface oxygen vacancies is well documented[102,105,113,211,220]. These radical species can beassociated with a new O1s peak in the XPS spec-tra of substituted samples lying at 2 eV above thelevel typical for oxides and non-promoted perovskites[113,207,215,217,220]. However, at least in part, thispeak may also correspond to surface hydroxyls boundto lanthanum cations [216].

If oxygen vacancies appear in the coordinationsphere of transition metal cations in perovskites, theoxygen octahedron easily transforms into anotherpolyhedron (tetrahedron, bipyramid, etc.) thus remov-ing the vacancy [204,205,214]. Therefore, one mayassume that relatively stable surface oxygen vacanciesare most likely present in the lanthanum-containing


faces. This means that oxygen ion-radicals appears tobe mainly linked to the subsurface transition metalcations. Hence, a cation substitution produces theactive oxygen species on rather inert lanthanum-containing faces of perovskites. The ion-radical formof oxygen is coordinated to subsurface transitionmetal cations in part screened by the lattice oxygenanions. Hence, the removal of ion-radical due to des-orption at temperatures of ammonia oxidation will notlead to undesired stabilization of ammonia oxidationintermediates.

From the point of view of the surface oxygen bond-ing strength, the NO selectivity pattern in the hightemperature ammonia oxidation on oxide systems canbe considered.

Since the heat of oxygen adsorption in on-top po-sitions on the regular centers of cobalt oxide is lowerthan that for platinum, the optimum temperature of theammonia oxidation into NO (650C) is also lower. Anincrease of the temperature of maximum NO yield ingoing from Co3O4 (650C) to a-Fe2O3 (700800C)and to a-Cr2O3 (850C) [7,76] apparently correlateswith the increasing strength of MO form bonding withthe regular surface cations [121,124,126].

The maximum NO selectivity on cobalt oxide (96%)and chromium oxide (94%) is higher than that onhematite (8993%). This qualitatively agrees with thefact that on the former two oxides we have a high(up to a half of monolayer) surface coverage by theterminal oxygen form with the adsorption heat ca.4060 kcal/mol [121,176,177], whereas on hematitethis coverage hardly attains 20% of a monolayer [124].

High maximum NO yields on perovskites systems,which competes or even exceeds those on pure transi-tion metal oxides [7,103113], are well explained bythe fact that the strength of the terminal oxygen bond-ing to a transition metal cation in former systems ishigher than that for simple oxides.

A specific behavior of the oxides faces towards sta-bilization of various adsorbed oxygen species as wellas an undesired effect of extended defects explains thedependence of NO selectivity on the synthesis pro-cedure for different samples of the same oxide phase[7]. Beside the effect of some non-controlled alka-line cation admixtures (vide infra), the oxide particlesmay also change their shape and defect structure whenthe nature of the precursor phase or conditions of itsdecomposition are varied.

Fig. 1. Typical TEM image of the hematite particles prepared bythermal decomposition of goethite.

Thus, when hydroxides, oxohydrates or basic metalcarbonates, whose surface is usually represented bythe densely packed hydroxyl layers, are used as precur-sors, the oxides surface is also mainly formed by thedensely packed faces [121,124,127,178,179]. More-over, the precursor phase lattice transforms to that ofthe oxide phase in a topotactic manner, which helps togenerate a great number of extended defects such astwins, dislocations and stacking faults [182,221223].Fig. 1 illustrates such highly defect iron oxide parti-cles obtained through the goethite decomposition. Inthis case, the particles surface is mainly composed bybasal and prismatic faces characterized by a low reac-tivity of regular centers.

If iron oxides are obtained through decomposi-tion of a nitrate solution, their particles are usuallyelongated, roundly shaped (Fig. 2) and contain fewextended defects. The particles roundness impliespresence of high index stepped faces, on which theterminal oxygen species bound to the regular cationsmay appear. Experiments [7,82,83] show that indeedNO selectivity is higher for samples of iron oxidesprepared from the iron nitrate, which appears to cor-relates with a higher surface coverage by oxygenwhose adsorption heat is about 60 kcal/mol [221].

Round-shaped particles are usually formed whenthe non-equilibrium methods are used for the oxidesynthesis. Among these methods are spray-drying andcalcination in the heat carrier flow, the plasma thermol-ysis of salt solution [195,196] or mechanochemicalactivation [193]. Fig. 3 shows that iron oxide particles


Fig. 2. Typical TEM image of hematite particles prepared by thermal decomposition of nitrate solutions.

Fig. 3. Typical TEM image of hematite particles prepared by spray-drying of chloride solutions.


Fig. 4. Typical TEM image of lanthanum manganite particles synthesized via arc plasma thermolysis of mixed nitrate solution (a) andcorresponding microdiffraction in the h1 0 0i zone (b).

obtained through the iron chloride solution decompo-sition in the spraying columns (a commercial hematitesample, the West Siberia smelter) followed by calcina-tion at 800C are indeed round-shaped and show highselectivity in the desired reaction [99,206,224,225].The lanthanum manganite particles obtained throughthe arc plasma thermolysis of mixed nitrate solutionsare also spherical (Fig. 4) [201,202,226]. In this case,the particles are comprised of micrograins, whosesurface is represented by the (1 0 0) faces. At the sametime, the lanthanum manganite particles prepared

by the thermal decomposition of co-precipitated hy-droxides are usually platelets with mostly developed(1 1 0) and, probably, (1 1 1) faces. As the result,these samples are less active in redox reactions[195].

From the practical point of view, the oxides, whichcan be used for synthesis of commercial catalysts,must not transform to lower oxides at the temperaturesof catalytic process in a slightly reducing medium.With this regard, pure manganese oxides MnO2 andMn2O3 though initially selective, are inevitably con-


verted to less selective Mn3O4 spinel oxide due totheir reduction.

Similarly, the phase transition CuO!Cu2O, whichoccurs in the surface layer of oxide particles earlierthan in their bulk and yields a dislocation network[227], also does not allow to use a pure copper oxideas a base for the commercial catalyst. A moderateselectivity of CuO may be explained by such a phasetransition, though copper oxide has a lot of surfaceoxygen with the appropriate bonding strength. A lowselectivity of Cu2O [7] may be caused by a highconcentration of regular coordinatively unsaturatedCuCcations on the most developed stoichiometric(1 0 0) and (1 1 0) faces [228,229], which stabilize theammonia oxidation intermediates.

Pure oxide Co3O4, showing the maximum selec-tivity at 650C [7], can be used only under the at-mospheric pressure and at lowered ammonia content,otherwise it transforms to less selective CoO. Purea-Fe2O3 is more stable catalyst than spinel cobalt ox-ide. However, it is also reduced to less selective Fe3O4(or even FeO) under the industrial conditions, if theoxygen content is below the optimum value [230].

Another problem can be a volatility of metal ox-ides under the reaction conditions. Thus, one of themost selective oxides, bismuth oxide, easily subli-mates at operating temperatures of ammonia oxida-tion into NO. Therefore, bismuth oxide is used onlyas an admixture to other transition metal oxides [7].Chromium oxide also shows a noticeable volatility, es-pecially under high-pressure conditions, because CrO3sublimates from the surface [7].

Hence, all the most selective simple transition metaloxides cannot be used in the industrial high-pressureprocess due to their poor stability in the reaction con-ditions. Therefore, it is necessary to modify the basicoxides by introducing promoters able to increase theirthermal stability, suppress reduction and improve (orat least not worsen) their selectivity.

3.3. Control of NO selectivity for catalysts based ontransition metal oxides

For basic oxides of cobalt, iron and chromium, a lotof other oxides were tested as additives. Mainly, thesewere non-reducible oxides of alkaline, alkaline-earthand rare-earth metals. Moreover, oxides of other tran-sition metals were used to promote the iron and cobalt

Fig. 5. Dependence of NO yield on the content of additives inCo3O4 samples [7]. 650C, GHSV (610)104 h1.

oxides. As a rule, promoters were introduced into theoxide bulk using decomposition of a mixed nitrate so-lution or co-precipitation of hydroxides (hydroxycar-bonates) from such solutions [7].

Most papers report no data on the promotion-inducedchange of the phase composition, defect structure andsurface chemistry of oxides or strength of the surfaceoxygen bonding. Hence, those results can be inter-preted only regarding similar systems characterizedby modern physical and chemical methods in details.

Almost all promoters decrease the NO selectivityof oxide systems in comparison with pure oxidestested under the same conditions [7] (see Figs. 510for illustration). The alkaline cations provide the mostnegative effect on selectivity. In this case, the terminaloxygen bond with the surface transition metal cationbecomes stronger due to electron donating effect ofalkaline cations in the neighboring positions. Thus,

Fig. 6. NO yield vs. radius of added cation for modified Co3O4samples [7]. 650C, GHSV (610)104 h1, promoter content 5%.


Fig. 7. Dependence of NO yield on the content of additives inFe2O3 samples [7]. 750C, GHSV (610)104 h1.

Fig. 8. NO yield vs. radius of added cation for modified Fe2O3samples [7]. 750C, GHSV (28)104 h1, promoter content 5%.

Fig. 9. Dependence of NO yield on the content of additives inCr2O3 samples [7]. 850C, GHSV (610)104 h1.

Fig. 10. NO yield vs. radius of added cation for modified Cr2O3samples [7]. 850C, GHSV (312)104 h1, promoter content 5%.

according to [123], introduction of only 2 wt.% ofpotassium into hematite substantially increases theisosteric heat of oxygen adsorption from the optimumvalue (5060 kcal/mol) (Fig. 11). Probably, the sharp-ness of the effect is caused by the tendency of potas-sium cations to segregate on the surface of transitionmetal oxides due to the difference between the radiiof transition metal and potassium cations [231233].Such a segregation intensifies with the increase ofcalcination temperature [221], thus appearing to bethe main factor strongly decreasing the NO selectivityof alkaline promoted oxides.

Although there are no reliable data showing thesegregation of alkaline metal cations on the sur-face of cobalt and chromium oxides, most likely themechanism of their effect on NO selectivity in thehigh-temperature ammonia oxidation is the same. In

Fig. 11. Heat of oxygen adsorption at the surface ofpotassium-modified iron oxide samples vs. potassium content. De-gree of surface reduction 2% of monolayer.


any case, micro-admixtures of alkaline metal cationsin cobalt and chromium oxides are known to decreasethe oxides activity in CO oxidation [121,176]. Withchromium oxide, the alkaline admixtures can also sta-bilize the surface phase of alkaline metal chromate,which provides a stronger oxygen bonding and lessreactivity in comparison to those of chromate groupson pure chromium oxide [176,234,235].

Alkaline earth cations and Me3C cations like alu-minum, whose radii are comparable to those ofcobalt cations, are relatively easy dissolved in Co3O4forming a mixed spinel, in which Co2C cations mayoccupy tetrahedra, while Me3C cations occupy octa-hedra [235,236]. Since reactive terminal oxygen onthe (1 0 0) and (1 1 0) faces is bound to both tetra-andocta-coordinated cobalt cations, their substitution bynon-reactive species bound to admixed cations de-creases the oxide NO selectivity. In agreement withthis model, the NO selectivity of Co3O4 modifiedby such admixtures depends only weakly on the ad-mixture content (Fig. 5). At the same concentration,more bulky La3C and Y3C cations decrease the NOselectivity of cobalt oxide sharper as compared withAl3C cations, most likely due to their higher tendencyto segregation. According to [237,238], large promot-ing cations introduced into the cobalt oxide increasethe sample dispersion as well as concentration of mi-crostrains. Hence, at least in part, the observed NOselectivity decline for cobalt oxide promoted by largecations, may be assigned to generation of extendeddefects and associated weakly bound surface oxygenspecies localized on cluster centers.

Alkaline-earth and aluminum cations introduced inhematite produce a less strong effect on the reactionselectivity towards NO than in the case of cobalt oxide.The reason is that the hematite structure contains a lotof vacant octahedra. According to [239,240], the ad-mixed cations of different charges (2C, 4C) are prefer-ably located as chains in the regular and interstitialpositions, while some neighboring Fe3C cations areshifted into interstices. Bulky rare-earth cations cansegregate as microinclusions or even X-ray detectedseparate phases, such as perovskites LnFeO3 [92]. Asa rule, the rare-earth admixtures increase the specificsurface area of modified hematite samples most likelyowing to suppression of sintering.

Introduction of alumina into the iron oxide producesa solid solution thus changing the lattice parameters

[241]. The X-ray energy-dispersive analysis [179] re-vealed that aluminum cations prefer to segregate on theextended defects (twins etc.) in the bulk of iron oxide,thus decreasing their surface concentration. The segre-gation of aluminum cations near the extended defectsin transition metal oxides supported on alumina sup-presses the oxides reduction by hydrogen [242244]and CO [128], first of all, due to a stronger bonding ofoxygen located in vicinity of these defects. Since thehematite reduction to magnetite proceeds topochemi-cally, the nuclei of magnetite forming near the outletsof extended defects [245], the alumina addition im-proves stability of hematite-based oxide catalysts inthe reaction of high-pressure ammonium oxidation [7].

On hematite, the maximum selectivity towards NOeither decreases, when aluminum or alkaline earthcations are introduced, or increases, when cerium orlanthanum cations are added [7]. In the latter case,the optimum temperature of the catalyst operation in-creases indicating a stronger bonding of oxygen withthe surface.

For chromium oxide, which, as hematite, has acorundum-type structure, the same admixtures im-pose a far stronger effect on the selectivity towardsNO (Figs. 9 and 10). Moreover, chromium oxide andhematite show the opposite dependencies of NO yieldon the radius of promoter cation. This distinction,which earlier found no explanation [7], can resultfrom the difference in the nature of oxygen species re-sponsible for selective ammonia oxidation to nitrogenoxide, and, in particular, from the specific features ofchromate groups stabilized on the Cr2O3 surface. Inaccordance with the rules of the acid-base interac-tion, basic cations, appearing on the chromia surface,produce salt-like surface chromates phases, where theoxygen bonding becomes stronger. The more basic iscation, the stronger is its interaction with the chro-mate anion, and, hence, the chromium-oxygen bond.The basicity of cation increases, when its radius in-creases, charge being the same, as well as when thecation charge decreases, radius being the same. Thisfact explains the observed features.

Another interesting effect, not explained earlier,is that for mixed oxide systems Co3O4Cr2O3 andFe2O3Cr2O3, NO selectivity is lower than for pureoxides. This effect was observed for a rather broadrange of temperatures (700850C) [230]. The de-cline of NO selectivity for cobalt and iron oxides due


to chromium addition can be ascribed to appearanceof the surface chromate phase, which strengthens theoxygen bonding. However, such a simple explanationdoes not work when Fe and Co cations are intro-duced into the chromium oxide. In this case, admixedcations with a lower strength of oxygen bonding areexpected to increase the NO selectivity, at least, attemperatures close to optimum ones for iron andcobalt oxides, which does not take place. The mostplausible explanation stems from the fact that evensmall admixtures of guest cations affect the decom-position of metal salts thus increasing the density ofextended defects-twins and stacking faults [246,247].Thus, cobalt oxide nuclei formed in the course ofdecomposition of the cobalt nitrate solution with asmall admixture of the iron nitrate, were shown to besurface-enriched by iron [178]. Stacking of such nu-clei produces a stacking fault with the local hexagonalsequence of the oxygen layers typical for hematite (acorundum-like structure) but not spinel structure. Theadmixed cations present in the vicinity of extendeddefects stabilize these defects by preventing theirannealing on calcination at elevated temperatures.

Even in the case when the host and guest cationshave the same charge and provide the same type of theoxide structure, this mechanism of defects generationcontinues to operate, as shown for the chromium ox-ide with a small (1%) admixture of Fe cations. In thiscase, stacking faults in the prismatic planes dominate(Fig. 12) [124]. At the same temperatures of calcina-tion, samples modified with iron cations contain moredefects than pure samples (Fig. 13).

From the molecular point of view, the increase ofthe density of extended defects implies the increaseof the concentration of surface centers where weaklybound oxygen is adsorbed (vide supra). At operatingtemperatures of ammonia oxidation, such an oxygen isremoved from these sites, making them coordinativelyunsaturated and capable to stabilize the intermediateproducts of ammonia oxidation. As the result, the routeof ammonia oxidation to molecular nitrogen becomesmore efficient.

3.4. Systems based on complex oxides of perovskitestructure

Complex oxides with a perovskite structure showa high selectivity towards NO at operating tempera-

Fig. 12. High resolution TEM image of the chromium oxideparticles prepared by thermal decomposition of nitrate solution(a) and corresponding microdiffraction in the h1 1 2 0i zone (b)[124].

tures around 900C, even when composing transitionmetal oxides such as copper and nickel oxides are notselective by itself (Table 3).

Obviously, manganese, iron and cobalt-containingperovskites, as compared with pure transition metaloxides, show a higher maximum selectivity at highertemperatures. In contrary, in the reactions of deepoxidation of hydrocarbons and CO, cobalt, nickel


Fig. 13. Density of microstrains in pure and promoted samples ofchromium oxide vs. calcination temperature.

and copper-containing perovskites exhibit a far loweractivity than individual oxides [196]. This fact wasexplained by a lower probability for cluster activesites to form in the perovskite matrix (vide supra).Hence, the beneficial effect of perovskite structureon the catalytic properties of transition metal cationsin the high temperature ammonia oxidation into NOappears to stem from stabilization of these cations inthe octahedral coordination. This provides the opti-mum strength of the terminal oxygen bonding on the(1 0 0) faces, thus ensuring a high NO selectivity.

Table 3Selectivity of some perovskite systems towards NO

Basic system, opti-mum composition

NO yield,% (at T, C)

Reference

La1xSrxNiO3LaNiO2:92 97.3 (700) [106]La1xCexCo1yByO3C(BDV, Cu, Fe, Ni, W)La0:8Ce0:2Co0:9Fe0:1O3C 99.5 (700) [110]La11:33xThxNiO3La0:6Th0:3NiO2:95 98.7 (700) [113]CaxLa1xMnO3CCa0:9La0:1MnO2:97 97.8 (800) [113]SrxLa1xCoO3CSr0:09La0:91CoO2:986 98.8 (700) [113]SrxLa1xFeO3CSr0:4La0:6FeO2:954 96.9 (800) [113](LaCe)MnO3a 93.8 (900) [13]CaMnO3a 88.2 (920) [13]Ca2Fe2O5a 84.7 (940) [13](LaCe)FeO3a 86.9 (920) [13](DyY)MnO3a 86.8 (925) [13](DyY)FeO3a 83.5 (920) [13](DyY)Mn0:3Fe0:7O3a 83.1 (920) [13]

60 (900)a Samples contain up to 20% of alumina binder.

Mixed oxides with a K2NiF4 structure such asLa2CoO4, La2CuO4, La2NiO4, which are composedof alternating perovskite (ABO3) and rock salt (AO)layers, are also rather selective [248,249]. Selectiveenough are structures like calcium manganite, stron-tium cobaltite (defect perovskite) and calcium ferrite(braunmillerite) (see Table 3). A partial substitutionof one transition metal cation for another (cobalt formanganese, iron for cobalt, etc.) also improves NOselectivity [103105].

A very attractive feature of perovskites is thatthey meet practice demands to retain their structure,composition and catalytic properties at the operatingtemperatures of industrial ammonia oxidation process[113].

In general, perovskites or perovskite-like systemsare selective even in the case when their surfacecontains no weakly bound forms of oxygen desorb-ing below 700C [113,250]. Nevertheless, a partialsubstitution of lanthanum by other cations, produc-ing weakly bound oxygen species desorbing within200600C range, is usually accompanied by a smallincrease of selectivity, at least to some degree ofsubstitution dependent upon the nature of perovskite[113]. However, it does not prove that a weakly boundoxygen is responsible for ammonia oxidation intoNO on perovskite systems. Thus, for La1-xSrxCoO3,the amount of strongly bound oxygen species des-orbing at 800C increases with the strontium content,correlating with NO selectivity as well [212]. Ac-companying increase of the surface concentration ofcobalt [199] implies these oxygen species are boundwith cobalt cations.

In some cases, experiment shows the NO selec-tivity of perovskites to change contrariwise to thedensity of extended defects (microstrains and inter-grain boundaries) [107,113]. New phases evolvingat some substitution degree (mixed oxides of braun-millerite structure, transition metal oxides, promotingoxides, lanthanum carbonates) unambiguously causethe decrease of NO selectivity [106,113]. As a rule,heterophase oxide systems have a well developed dis-location network at the interfaces. This agrees wellwith the assumption that in the oxide systems ex-tended defects decrease the NO selectivity in ammoniaoxidation (vide supra).

For ammonia oxidation to nitrogen oxide onperovskites, the ammonia molecule is assumed to


be activated on the surface Me3C cations then inter-acting with the ion-radical oxygen species adsorbedon vacancies [106,107,113]. Ammonia may also di-rectly interact with activated oxygen species [113].Although such stages may participate in the routesof NO generation, they hardly are dominating, sincea high NO selectivity was obtained for unpromotedperovskites as well.

In lanthanum ferrites and cobaltites, at some de-gree of La substitution, the braunmillerite structurefragments appear [205,214], in which transition metalcations are in the tetrahedral coordination. This pro-vides the terminal oxygen species more stronglybound to those cations, while bonding of ion-radicaloxygen species becomes stronger as well [113]. Asthe result, calcium substituted ferrites and lanthanumcobaltites decrease their NO selectivity estimatedat the same temperature [113]. This model is sup-ported by the fact that calcium substituted lanthanummanganite, in which there is no change of transitionmetal cation coordination, gains in selectivity un-til a defect calcium manganite phase starts to form[113].

4. Industrial oxide catalysts for ammoniaoxidation to nitrogen oxides and technologies fortheir production

4.1. Catalyst compositions

Search for oxide catalysts for ammonia oxidationstarted at the beginning of the 20th century. Thus, in1902 Ostvald patented catalysts based on transitionmetal oxides. During 19141916, for NO production,a German chemical company BASF used the catalystcontaining iron oxide doped with oxides of bismuth,chromium, manganese, uranium. At 750800C, theNO yield on this catalyst was 9294% at an ammoniacontent 7.5%.

In the 1960s, intensive studies resulted in develop-ment of some catalyst compositions, which operated inindustry. Research attention was focused on the cobaltoxide systems, since they allowed operation withoutplatinum gauzes [83,251].

Hedlers catalyst [62] is cobalt oxide stabilized byvarious admixtures. It allowed a 9394% NO yieldand worked for 1 year.

African Explosives [252,253] patented a catalystbased on cobalt oxide promoted with scandium,yttrium and/or other rare earth metals, which at 800850C and 1000 m3/m2 h provided NO yield mount-ing to 9596%.

The ICI catalyst [83] consists of cobalt oxide sup-ported on magnesiumaluminum spinel with specificsurface 350 m2/g. This catalyst works at 7001000Cand provides NO yields close to that obtained onindustrial platinum catalysts. Note, however, that ac-cording to [94], a large surface and developed porestructure of cobalt oxide catalysts cause the NOselectivity to decrease.

The Stamicarbon catalyst [83] may work at temper-atures not higher than 850C at a volume gas flow rateof up to 100 000 h1, ammonia content mounting to10%.

Some data on the oxide catalyst compositions sug-gested in the recent years are given in Table 4.

If only oxide catalysts are used, the contact timeusually ranges from 102 to 101 s, which is essen-tially longer than on platinum gauzes. Moreover, theoxide catalysts selectivity towards NO is always lowerthan that of Pt gauzes, which is ascribed to a largerchemical sorption capacity of oxides [7]. In our opin-ion, it may also be caused by a larger impact of thehomogeneous reaction between ammonia and NO.

With regard to industrial process economics, whichis primary determined by the nitric acid yield andpossibility to use the same reactors without seriousrewamping, it is better to use oxide catalyst togetherwith a fewer number of platinum gauzes [7].

The possibility to use two catalysts beds for ammo-nia oxidation to nitrogen oxides was first mentionedin patent [252,253]. In the 1950s, GIAP designed theironchromium oxide catalyst KN-2 for the secondbed of ammonia oxidation system and determinedoperating parameters for the two-bed system for am-monia oxidation (single Pt gauzeCoxide catalyst bed)under atmospheric pressure [12].

Later some patents suggested ammonia oxidationby oxygen containing gas over the catalyst comprisedby platinum (or its alloy) gauzes with granulated oxidecatalysts loaded over gauzes (iron oxide, manganeseoxide, bismuth oxide [254], cobalt oxide [255], cata-lysts containing iron oxide, chromium oxide, bismuthoxide, lead oxides and other promoters [256,257],catalysts containing iron oxide, cobalt oxide nickel


Table 4Composition and NO selectivity of some oxide catalysts

Optimum composition NO yield, % (at T, C) ReferenceLaNiO3 99 (663) US Patent 4812300La0:75Sr0:21MnO3 98 (663)MeO/support MeDCo, Fe, Cr, Mn, Bi 55 (650) E Patent 0562567 A1MeO/support MeDCo, Pt 95 (760800) US Patent 5242882a-Fe2O3 9198 (620720) DE Patent 3126675 A1a-Fe2O3CGa2O3(Ce2O3)C12 Pt gauzes 9297 SU Patent 856540; 1148151a-Fe2O3C34 Pt gauzes 95.5 (900950) SU Patent 1153981 AMixed oxides of Y, Be, Ba, Cu 92 (900950) SU Patent 1759446 A1a-Fe2O3, Cr2O3 98 (800) SU Patent 300057a-Fe2O3CZrO2 94.3 (850) SU Patent 805514a-Fe2O3, Al2O3, MgOC1 Pt gauze 98.1 (900950) SU Patent 1676141 A1a-Fe2O3, Al2O3, MgOC1 Pt gauze 98 (850920) SU Patent 771958Co3O4Calumina (515%), thoria, ceria, zincand calcium oxides pellets or extrudates

Up to 97 (850900) SU Patent 325761; US Patent3962138, 4036935

Co3O4C0.110% Li2O pellets, grains, monoliths 97 (850900) US patent 3931391Co3O4 or Bi2O3 promoted by rare earth element orthorium, may also contain at least one of oxides ofMn, Fe, Mg, Cr or Nb

97 (850900) US patent 3985681

a-Fe2O3MAl2O4Bi2O3Ce2O3 (MDMg, Mn, Ca, Sr, Ba) 97 (850900) US patent 3985681CoOx , Pt/support 9597 (850900) US patent 5336656a-Fe2O3 9597 (850900) RU Patent 1383563a-Fe2O3, Al2O3, MgO 9597 (850900) RU Patent 1541833Mixed Fe2O3, Al2O3, Cr2O3CPt gauze 97.698.2 (795810), 1 atm Bu Patent 31092a-Fe2O3 RU Patent 1626495

oxide [258]). For these systems, the nitrogen oxideyield at atmospheric pressure was within 9298%.Some other compositions are listed in Table 4. Notethat the two-bed system is used only in NIS and inother countries where the ammonia oxidation technol-ogy developed in the USSR is used (Poland, Bulgaria,etc.) [7,88,93,100].

When a non-platinum catalyst is installed down-stream from the platinum gauzes, its operating condi-tions differ significantly from those experienced if itis used alone. The gas mixture exiting the platinumgauzes contains NO, water, less oxygen (59%) andammonia (12%). As a result, the demand towardscatalyst selectivity increases (that is to minimize theoccurrence of reaction between NO and ammonia onthe catalyst surface), but the demand in activity is farless severe, since platinum gauzes make the processauto-thermal, and not more than 15% of the inlet am-monia concentration comes to the oxide catalyst bedinlet.

The thermal stability of the second bed catalystsis of key importance. For catalyst working under theatmospheric pressure it means temperatures around

800C. Under 710 atm, the catalyst must resist heat-ing up to 900950C. As a result, the oxide ironchromium second bed catalyst, which provides 6 yearsstable operation under the atmospheric pressure [7], isquickly deactivated on operation under elevated pres-sures due to its reduction to magnetite, wustite andeven metal iron [230].

This obstacle predetermines requirements towardsthe composition of oxide catalysts operated underdifferent pressures. For elevated pressures, amongthe most selective are iron-aluminum oxide catalystspromoted by other oxides (ceria, vanadia, etc.) [7,83].Their selectivity is quite stable in high pressure con-verters, though on operation, they loose their specificsurface from 1025 to 1 m2/g, and their pore struc-ture changes (small pores disappear and large poresappear) [241]. Such changes may reduce the mechan-ical strength of the catalyst granules regarding theirpossible attrition and weight loss [259]. Chemicallyunbound highly dispersed corundum segregating onthe surface of some ironaluminum oxide catalystsmay increase the dust generation thus reducing thecatalyst selectivity [260].


4.2. Catalyst synthesis methods

Though modified iron oxides prepared from mixednitrates solutions [261,262] are mostly selective, ni-trates shortage and high prices stimulated a search foralternative raw materials and synthesis methods. Goodresults were obtained by the acid kneading of com-mercially available iron oxide (e.g., semi-product forthe middle-temperature water gas shift ironchromiumoxide catalyst) with alumina and other additives. Themechanical activation of this paste is used to improvethe interaction between the components [263269].The paste is either extruded as cylinders or tubes ordecomposed to produce powders then pressed and sin-tered [270].

The dispersed powders of perovskites are usuallyproduced via such routes as sintering of the oxidesmixture, co-precipitation of mixed hydroxides, ox-alates or other non-soluble compounds followed bycalcination [271]; spray-drying or arc plasma thermol-ysis of mixed nitrates solution [181,195,196].

Mechanochemical activation of solid reagents inhigh power ball mills is also used for perovskitessynthesis [193,196198]. Unlike traditional ceramicprocedures, sintering of mixed activated reagents atreasonably low (600800C) temperatures rapidlyproduces highly dispersed monophase powders ofcomplex oxides. The rate of solids interaction duringactivation and/or subsequent annealing was shown tobe proportional to the difference of their acidbaseproperties determined by the cation charge and anioniccomposition [272]. The particles of perovskites thusobtained are spherical or almost spherical. Since thesingle phase composition and rounded particle shapeare of critical importance for perovskites exhibitinghigh NO selectivity, this synthesis technology seemsto be rather promising for production of catalysts to beused for ammonia oxidation to nitrogen oxides. Thespecific surface area of complex oxides produced viathe mechanochemical method is 1020 m2/g, whereassamples obtained through the ceramic technology havethe specific surface area values not exceeding 1 m2/g.

4.3. Honeycomb oxide catalysts for the second bedof ammonia oxidation to nitrogen oxides

In USSR in 1970s, the catalyst based on iron ox-ides and designed for the second bed of ammonia

oxidation to NO was successfully operated in highpressure (up to 10 atm) converters. A bed of tablets (56 mm56 mm) or extrudates (15 mm long, 56 mmin diameter) was loaded into special baskets.

However, granulated oxide catalyst met some op-erating problems, since it sometimes contained dust,and its loading required a specially designed basket.The pressure drop in the granular catalyst bed was toolarge (up to 1000 mm of water). Moreover, the granu-lar bed is not uniform in thickness providing by-passfor the gas flow and a non-uniform gas flow rate withinitself as well as in platinum gauzes [13].

All these problems stimulated development of thetwo-bed catalytic system for ammonia oxidation usinga honeycomb monolith oxide catalyst. The honeycombcatalyst bed has the following advantages over the bedof granular catalyst:1. no special basket is required for the bed;2. a high mechanical strength, resistance to thermal

shocks, the absence of dust provide the uniform fil-tering of the ammoniaair flow through the catalystbed;

3. a uniform gas permeability of the honeycomb cat-alyst equalizes the gas flow through the platinumgauzes;

4. the honeycomb catalyst bed has a low pressuredrop.

The technology for honeycomb oxide ammo-nia oxidation catalysts was developed in the 1990s[99,196198,210,224226,273277]. It is based onthe wet kneading procedure allowing a waste-lesscatalyst production [278,279]. The active componentpowder is first mixed with organic and inorganic ad-ditives and a binder. Then water and electrolytes areadded to produce pastes, which then are extrudedthrough proprietary extruding heads (spinnerets).Then monoliths are dried and calcined.

Reinforcing aluminasilica fibers are added into thepastes on mixing to improve the catalyst resistanceto thermal shocks. Aluminum oxynitrate or pseu-doboehmite serves as the aluminum binder also allow-ing alumina introduction into the iron oxide to improveits thermal stability. For a better rheology of pastes andpore structure, various surfactants were added such asethylene glycol, polyethylene oxide, carboxymethylcellulose, polyvinyl alcohol, glycerin, etc.

Optimized regimes of drying in controlled humidityconditions and calcination up to 9501000C suppress


cracking and provide monoliths with good mechanicalstrength.

Optimum composition monoliths with active com-ponents based on perovskites and iron oxide are 75 mm75 mm in cross-section and 50100 mm long withdesired channel size and wall thickness. After cal-cination at 900C, the monoliths have the follow-ing parameters: specific surface 30 m2/g, integral porevolume-up to 0.3 cm3/g, crushing strength-8 10 MPa.The best catalysts survive up to 100 thermal cyclesfrom room temperature to 800C without cracking.

Pilot-industrial batches of the honeycomb oxide cat-alysts for the second bed of ammonia oxidation areproduced using the facilities of Boreskov Institute ofCatalysis [13].

5. Two-bed ammonia oxidation: processpeculiarities

Detailed studies regarding the two-bed ammoniaoxidation with granulated catalyst [4,12,81,280,281]allowed the estimation of main process parametersunder various operation conditions (load and pres-sure). These parameters were then used in attemptsto optimize the catalyst loading and bed geometry.However, in fact, all parameters for the two-bed sys-tems with granulated oxide catalyst bed were obtainedempirically with no physical model for the processdescription.

A high pressure drop within the oxide catalystbed stimulated its geometric optimization. Convertersworking under 7.16 atm with platinum gauzes and theoxide catalyst disposed on a cone or a semi-sphericalsurface show a several fold lower pressure drop at thesame yield of nitrogen oxide. Moreover, the platinumloading is reduced by 22.5 kg, and platinum lossesare by 50% lower [281].

With the honeycomb oxide catalyst, all techno-logical parameters (such as the number of platinumgauzes, optimum configuration and size of honey-comb catalyst channels, the bed height and number ofbeds, the distance between the beds) were determinedusing the mathematical modeling [282].

A two-phase model of the ammonia oxidation pro-cess on the platinum gauzes and honeycomb catalystwas developed for this purpose. According to themodel, the process is limited by the mass transfer

Fig. 14. Calculated ammonia conversion and NO yield alongthe monolith length with regard to inlet ammonia concentrationaccounting for (X, Y) and without accounting for (X, Y) thehomogeneous reaction. Process parameters: pressure 0.7 MPa, su-perficial gas velocity 7 m/s, inlet temperature 915C. Monolithcharacteristics: square channels of 2 mm, wall thickness 1 mm.

between the gas phase and catalysts. The processselectivity towards NO is provided by ammonia oxi-dation to NO and N2 on the catalyst surface, and bythe side homogeneous reaction of NO and ammoniayielding N2 thus decreasing the process efficiency.

Fig. 14 shows how ammonia conversion and NOyield change along the monolith length. At the channelinlet, the ammonia oxidation rate is high due to veryfast mass transfer, and conversion along the monolithquickly increases. As the gas velocity profile stabilizesalong the channel, the mass transfer rate decreases,and the homogeneous reaction contribution increases.The homogeneous reaction rate is far lower than thatof the desired reaction, but it has a great effect on achoice of the monolith channel configuration and sec-ond bed geometry: channel size, wall thickness, lengthand number of monolith layers and their mutual dis-position. These parameters should be optimized witha restriction imposed by a permissible pressure drop.

The pressure drop of the second bed is an im-portant factor effecting the efficiency of the wholereactor. Modeling results demonstrate that honey-comb monoliths used as the second bed perform notonly as catalysts. Owing to their regular structure, the


Table 5Operation parameters of two-bed system working under 7.3 atm [13]Month NH3 conversion (%) NH3/air ratio (%) T (C) NH3 load (nm3/h) HNO3 (t) Capacity (HNO3 t/h) Run (h)March 94.10 9.36 866 6000 8406 14.8 568April 94.07 9.15 866 5800 8114 13.5 601May 93.64 9.15 868 5700 9299 13.0 715June 93.77 9.08 865 5650 9819 14.4 684July 92.95 9.13 862 5550 9145 12.9 708

gas flow distribution inside the gauzes pad becomesmore uniform. This results in more uniform and lowermechanical and chemical losses of platinum, thusimproving the gauzes efficiency and increasing theiroperation life.

Computational fluid dynamics study was performedto investigate the effect of monolith catalyst geometryand disposition [283]. The results demonstrated thatthe optimal distance between two catalyst beds can bedefined.

6. Industrial operation of the two-bed catalyticsystem

A long term experience in industrial operation ofthe two-bed catalytic system with granulated catalystsis summarized in [4,7,13,28].

Since 1985 till 1990, in two-bed systems, granulatedironaluminum oxide catalyst was used in UKL-7 con-verters at seven nitric acid plants. The number of UKLconverters varied from 7 to 20 depending on the cat-alyst and baskets supply. If all design and operationrequirements were strictly followed, the NO yield wasthe same as in the case of the standard loading of plat-inum gauzes. The platinum metal loading was thus re-duced by 2025%, and the irreversible platinum losseswere cut for 15% [7,13].

Two pilot batches of the honeycomb oxide catalystwere produced in 1995 and loaded into the UKL-7converters at JSC Azot in Cherepovets (Russia) andJSC Azot in Berezniki (Russia). The catalyst wasrepresented by rectangular prisms 50 mm high with thesquare section 70 mm70 mm. The catalyst (100 L)was loaded in a single layer across the converter ona Nichrome screen covering a support grid. The cat-alyst was covered by another Nichrome screen, andthen pad of one used and nine new gauzes made of

GIAP5 alloy [7] were installed. The average nitro-gen oxides yield was 94%, and no ammonia slip wasobserved.

The full-scale resource testing of the honeycombcatalyst was performed in 1996 in Berezniki. Table 5shows the averaged operation data of thus arrangedtwo-bed system. During the normal life time (3000 h)of gauzes, the ammonia conversion stays at the originallevel.

The platinum metal losses attained 0.1237 g/t ofacid at a norm of 0.157 g, while neither activitynor mechanical properties of the monolith catalystchanged. Loaded in March 1996, the honeycombcatalyst worked without reloading for 1.5 years [13].

At present, the honeycomb catalyst is successfullyoperated in 10 high pressure converters in Berezniki,Cherepovets, Nevinnomyssk, Novgorod and otherplaces in Russia.

7. Comments and future plans

Process economics improvement lies in the furtherreduction of the platinum metal weight via gauzes re-placement by the honeycomb oxide catalyst. Keepinga high NO yields intact demands catalyst composi-tion optimization and better geometry of monoliths.Though analysis of oxygen bonding to the oxidecatalyst surface appears to be rather fruitful for exper-imental data systematization, the energy spectrum ofthe surface oxygen as a function of the surface mor-phology and defect structure is not yet studied enough.There is a lack of data on reaction kinetics under theindustrial conditions. Though all these problems meetsevere experimental difficulties, advanced methods de-signed to study reactions at millisecond contact timesmay provide a breakthrough in understanding thereaction mechanism.


Modified iron oxides and perovskites seems to bethe most promising for the practical application. Fur-ther, optimization of their chemical composition andsynthesis procedure will improve their selectivity andthermal stability.

Engineering problems relate to a more optimizedarrangement of the two-bed system regarding mono-liths (configuration and spatial disposition ) as well asgauzes pack (their number and parameters).

References

[1] P. Pickwell, Chem. Ind. 4 (1981) 114.[2] W. Ostwald, Berg- und Httenm. Rsch. 20 (1906) 12.[3] F. Sperner, W. Hohmann, Platinum Metals Rev. 20 (1976)

12.[4] V.I. Atroshchenko, S.I. Kargin, Nitric Acid Technology,

Khimia, Moscow, 1970, p. 496.[5] V.I. Chernyshov, V.A. Gertsovskii, T.G. Sharova, Ammonia

Combustion Plants (Review), Vol. N9, NIITEchim, Moscow,1976, p. 70.

[6] R.J. Farrauto, H.C. Lee, Ind. Eng. Chem. Res. 29 (1990)1125.

[7] M.M. Karavayev, A.P. Zasorin, N.F. Kleshchev, Catalyticoxidation of ammonia, Khimia, Moscow, 1983.

[8] T.H. Chilton, Chem. Eng. Progr. Monogr. Ser. 3 (1960) 56.[9] H.C. Lee, R.J. Farrauto, Ind. Eng. Chem. Res. 28 (1989) 1.

[10] H. Holtzman, Platinum Met. Rev. 13 (1969) 2.[11] W.R. Hatfield, R.M. Heck, T. Hsiung, US Patent 4 412 859

(1983).[12] D.A. Epshtein, N.M. Tkachenko, M.A. Miniovich, Dokl.

AN SSSR 122 (1958) 874.[13] V.A. Sadykov, E.A. Brushtein, L.A. Isupova, T.V.

Telyatnikova, A.A. Kirchanov, I.A. Zolotarskii, A.S. Noskov,N.G. Kozevnikova, B.Yu. Kruglyakov, O.I. Snegurenko,Yu.N. Gibbadulin, A.A. Hazanov, Khim. Prom. 12 (1997)812.

[14] D.A. Frank-Kamenetskii, Zh. Teor. Phiz. 9 (1939) 1457.[15] D.A. Frank-Kamenetskii, Dokl. AN SSSR 30 (1941) 729.[16] N.Ya. Buben, Zh. Phys. Chem. Add. Vol. (1946) 1948.[17] V.V. Barelko, Kinetika i Kataliz 14 (1973) 196.[18] V.V. Barelko, Dokl. AN SSSR 211 (1973) 1373.[19] N.I. Ilchenko, G.I. Golodets, I.M. Avilova, Teor. Experim.

Khim. 11 (1975) 56.[20] L. Apelbaum, M. Temkin, Zh. Phiz. Khim. 22 (1948) 179.[21] L. Apelbaum, M. Temkin, Zh. Phiz. Khim. 22 (1948) 195.[22] M. Bodenstein, Z. Elektrochem. 41 (1935) 466.[23] L. Andrussow, Z. Angew. Chem. 63 (1951) 350.[24] L. Andr

Catalyst Data With Rates

Documents

Transcript of Catalyst Data With Rates