Journal of Natural Gas Chemistry 19(2010)503–508

Fischer-Tropsch synthesis over ruthenium-promoted Co/Al2O3catalyst with different reduction procedures

Ali Karimi, Ali Nakhaei Pour∗, Farshad Torabi, Behnam Hatami,Ahmad Tavasoli, Mohammad Reza Alaei, Mohammad Irani

Research Institute of Petroleum Industry (RIPI), Gas Research Division, P. O. Box: 14665-137, Tehran, Iran[Manuscript received January 15, 2010; revised February 22, 2010 ]

AbstractThe effect of reduction procedure on catalyst properties, activity and products selectivity of ruthenium-promoted Co/γ-Al2O3 catalyst inFischer-Tropsch synthesis (FTS) was investigated. Catalyst samples were reduced with different reduction gas compositions and passivatedbefore being characterized by TPR and XRD techniques. Different activity and product selectivity analyses were also performed. These resultsshowed that the catalyst dispersion, particle size, and the degree of reduction changed with different reduction gas compositions, which wereresulted from the water partial pressures in reduction process that give varying degrees of interaction with the support. It has been suggestedthat the FTS activity of cobalt catalyst was directly dependent on the catalyst reducibility. A reduction gas with a molar ratio of H2/He = 1 wasused to prevent the formation of Co-support compound during catalyst reduction.

Key wordsFischer-Tropsch synthesis; cobalt; reducibility; activity; C5+ selectivity

1. Introduction

Current development aims at improved Fischer-TropshSynthesis (FTS) technology for the production of high-molecular-weight waxes followed by their hydrogenation toliquid fuels [1−9]. FTS catalysts with high volumetric pro-ductivity decrease reactor volume requirements and improvesignificantly the process economics. In general, volumetricproductivity is controlled by varying the density of active siteand its turnover frequency [10−12]. Due to high activity andlong longevity, cobalt-based FT catalyst is currently the cata-lyst of choice for the conversion of syngas derived from natu-ral gas to liquid fuels. Cobalt catalysts provide best compro-mise between reduced costs and high CO conversion, besides,they offer favorable C5+ selectivities as well as low water gasshift (WGS) activity for the synthesis of liquid fuels from nat-ural gas. Supported Co catalysts with high specific rates re-quire the synthesis of small metal crystallites at high local sur-face densities on support and the use of supports or alloys thatincrease the rate per surface Co (turnover rate) [13,14].

Cobalt catalysts are normally pretreated separately with

hydrogen before being used in the FTS reactions. During cat-alyst reduction oxygen atoms in Co active crystallites, espe-cially Co3O4, are predominately removed as H2O, so watervapor is a byproduct of metal oxide catalyst reduction [13,14].Hydrogen reduction is carried out in the range from 250 to400 ◦C preferably at low pressures and high linear gas veloc-ities to minimize the vapor pressure of the producing waterwhich enhances sintering of the reduced metal [15−17]. Ithas been suggested that water vapor decreases the reducibilityof cobalt in possibly two ways: i) inhibiting the reduction ofwell dispersed CoO on alumina support possibly by increasingthe cobalt-alumina interactions and ii) facilitating the migra-tion of cobalt ions into probable tetrahedral sites of γ-Al2O3to form a non-reducible entity (at T< 827 ◦C), which resultsin a decrease in the amount of cobalt to be reduced using con-ventional reduction procedures [18−20].

One important issue in the development of this processis the improvement of the catalyst activity by increasing thenumber of active Co metal sites that are stable during reduc-tion and under reaction conditions. Therefore, it is importantto understand which parameters can influence the activity andstability of catalysts.

∗ Corresponding author. Tel: +98-21-44739782; Fax: +98-21-44739716; E-mail: nakhaeipoura@ripi.ir; nakhaeipoura@yahoo.com

Copyright©2010, Dalian Institute of Chemical Physics, Chinese Academy of Sciences. All rights reserved.doi:10.1016/S1003-9953(09)60111-0

504 Ali Karimi et al./ Journal of Natural Gas Chemistry Vol. 19 No. 5 2010

2. Experimental

2.1. Catalyst preparation

Ruthenium-promoted cobalt catalyst (Ru/Co = 0.01 byweight) were prepared with 30 wt% cobalt on ConndeaVista Catalox B γ-Al2O3 support. The support was cal-cined at 500 ◦C for 10 h prior to impregnation. The cat-alyst was prepared by sequential aqueous impregnation ofthe support with appropriate solutions of cobalt nitrate(Co(NO3)2·6H2O 99.0%Merck) and ruthenium (III) nitrosyl-nitrate (Ru(NO)(NO3)3). After each step, the catalyst samplewas dried at 120 ◦C and then calcined at 400 ◦C for 3.5 h. Thecobalt and promoter loadings were analyzed by an inductivelycoupled plasma atomic absorption (ICP-AES) system.

2.2. Catalyst pretreatment

In general, 1 g of the catalyst was charged into a 1/4”stainless steel tube reactor. The reactor was placed in a moltensalt bath with a stirrer to ensure a uniform temperature regionalong the catalyst bed. The bath temperature was controlledvia a PID temperature controller. Standard reduction of thecalcinated catalyst was carried out in the reactor at 1 atm witha heating rate of 1 ◦C/min to a final temperature of 365 ◦C andthe sample was held at this temperature for 12 h with differentreduction gas compositions (RGC) of H2/He. Helium wasused as the balance gas. The nomenclature for the pretreatedcatalyst samples used in this study is given in Table 1.

Table 1. Catalyst nomenclature

Catalyst Reduction gas compositions (RGC,mol%)H100 100% H2H66 66% H2/HeH50 50% H2/HeH33 33% H2/He

2.3. Catalyst characterization

The surface area, pore volume, and average pore radiusof the support and fresh catalyst were measured by an ASAP-2010 system fromMicromeritics. The samples were degassedat 200 ◦C for 4 h under 50 mTorr and their BET areas, porevolumes, and average pore radii were determined.

The fresh and reduced samples were characterized byXRD and TPR techniques. The reduced samples were pas-sivated at room temperature for 2 h with a mixture of O2/He(5.2 vol% O2) according to a standard procedure describedelsewhere [21,22]. XRD measurements were conducted on aPhilips PW1840 X-ray diffractometer with monochromatizedCu Kα radiation. The average cobalt oxide crystallite thick-ness was calculated from Scherrer equation using the (311)Co3O4 peak located at 2θ = 36.9o. A K factor of 0.89 was

used in the Scherrer formula. The Co3O4 particle size wasconverted to the corresponding cobalt metal particle size ac-cording to the relative molar volumes of metallic cobalt andCo3O4. The resulting conversion factor for the diameter ofa given Co3O4 particle d(Co3O4) being reduced to metalliccobalt is:

d(Co0)(nm) = 0.75×d(Co3O4) (1)

TPR spectra of the fresh and reduced catalysts were ob-tained using a Micromeritics TPD-TPRModel 2900 equippedwith a thermal conductivity detector. The catalyst sampleswere first purged in a flow of argon at 300 ◦C to removetraces of water, and then cooled to 50 ◦C. The TPR spectraof each sample (50 mg) were obtained using 5.1% hydrogen-argon mixtures with a flow rate of 40 cm3/min. The sampleswere heated from 50 to 827 ◦C with a ramp of 10 ◦C/min.

The amount of chemisorbed hydrogen was measured us-ing the Micromeritics TPD-TPR 2900 system. 0.22 g of thecatalysts were reduced at 365 ◦C for 12 h and then cooled to100 ◦C under a hydrogen flow. Then the flow of hydrogenwas switched to argon at the same temperature, and lasted forabout 30 min in order to remove the weakly adsorbed hydro-gen. Afterwards temperature-programmed desorption (TPD)of the samples was obtained by increasing the temperatureof the samples to 400 ◦C under the argon flow with a ramprate of 10 ◦C/min. The hydrogen TPD spectrum was used todetermine the cobalt dispersion and its average surface crys-tallite size. Furthermore, to calculate the dispersion, it wasassumed that two cobalt surface atoms are covered by one hy-drogenmolecule and that ruthenium does not contribute to theamount of adsorbed hydrogen. The cobalt metal particle sizewas calculated from the cobalt metal dispersion by assumingspherical uniform cobalt metal particles with a site densityof 14.6 at/nm2 [11]. These assumptions give the followingformula:

d(Co0)(nm) = 96/D(%)×DOR (2)

where,D is the dispersion, d is the particle diameter and DORis the degree of reduction or fraction reduced. The dispersionof the catalysts was given by the following formula:

D(%) =Number of Co0 atoms on surfaceNumber of Co atoms in sample

×100 (3)

After the TPD of hydrogen, the sample was re-oxidized at400 ◦C by pulses of 10% oxygen in helium to determine theextent of reduction. The amount of oxygen consumed by thesample was calculated from the known pulse volume and thenumber of pulses reacting with the sample. The degree of re-duction was calculated by assuming that all cobalt in metallicform was oxidised to Co3O4. Any oxidation of Ru to RuO2was not considered in the calculations. The degree of reduc-tion or fraction reduced can be calculated from oxygen con-sumed by the samples via the following formula:

Fraction reduced =O2 uptake (mol/gCat) ×atomic weight×2/3

Metal weight (%)(4)

Journal of Natural Gas Chemistry Vol. 19 No. 5 2010 505

To compare Co3O4 particle sizes calculated from XRDwith the dispersion obtained from chemisorption, the Co3O4particle sizes were converted to the corresponding Co parti-cle sizes according to the relative molar volumes of metalliccobalt and Co3O4.

2.4. Catalytic measurements

A series of experiments were designed to evaluate cat-alysts in terms of their FTS activities (gCH/(gCat·h)) and se-lectivities (The percentage of the converted CO that appearsas a given product) in a tubular fixed-bed micro-reactor. Thecatalyst (1 g) was charged in a 1/4” tubular fixed-bed micro-reactor. The reactor was placed in a molten salt bath with astirrer to ensure a uniform temperature along the catalyst bed.The bath temperature was controlled via a PID temperaturecontroller. Separate Brooks 5850 mass flow controllers wereused to add H2 and CO to a mixing vessel at desired rates.It was preceded by a lead oxide-alumina containing vessel toremove carbonyls before entering the reactor. A schematicdiagram of the experimental setup is shown in Figure 1. Af-ter pretreatment the FTS tests were carried out under 220 ◦C,1 atm, H2/CO = 2 and GHSV = 1200 h−1. The effluent (CO,CO2, and C1–C20 hydrocarbons) from the reactor was ana-lyzed using an on-line Varian 3800 gas chromatograph. COconversion and different product selectivities were calculatedbased on the GC analyses. Anderson-Schultz-Flory (ASF)distribution line was plotted for C3+ products to determinethe chain growth probability α.

Figure 1. Schematic diagram of the experimental setup

3. Results and discussion

3.1. Catalyst characterization

The results of ICP and N2-adsorption tests for support and

fresh catalyst after calcinations are listed in Table 2. This tableshows that the pore volume, BET surface area, pore volumeand average pore radius did not show a significant difference.However, the BET surface area of the fresh catalyst was muchlower than that of the γ-Al2O3 which indicates pore blockagedue to the loading of cobalt on the support.

Table 2. ICP and N2-adsorption

Support/ Co loading Surface area Pore volume Average porecatalyst (wt%) (m2/g) (cm3/g) radius (nm)γ-Al2O3 — 270 0.64 4.7

Fresh catalyst 30 160 0.45 4.3

XRD patterns for representative fresh catalyst and sam-ples reduced at different reduction gas compositions (RGC)are shown in Figure 2. Support peaks for this catalyst(γ-Al2O3) appeared at 46.1o and 66.5o, while the peaksat 36.8o, 56o, 60o and 66o can be attributed to the (311),(422), (511) and (440) crystal planes of Co3O4, respectively[10,11,23]. Using the Scherrer equation, the average size ofthe Co3O4 crystallites in the catalysts could be estimated fromthe line broadening of the (311) Co3O4 crystallites plane at 2θof 36.8o. The peaks at 2θ values of 49o and 78o for H100 andH66 samples in Figure 2, correspond to CoAl2O4 spinel [23].

Figure 2. XRD patterns of the fresh and pretreated catalysts

It is known that significant amount of water was formedduring the catalyst reduction process and FTS reaction. Wa-ter vapor has impacts on the reduction of alumina-supportedcobalt catalyst, i.e., increasing the mobility of cobalt ions onthe support and therefore enhancing the Co-Al interaction toproduce Co support compound formation (Co-SCF) species[20,23]. Trivalent cobalt and aluminum have rather similarPauling’s ionic radii (0.063 nm for Co3+ and 0.054 nm forAl3+). It is known that Co3+ of Co3O4 can be gradually re-placed by Al3+ to produce a series of Co3−s AlsO4 (0<s<2)spinels [10]. The interaction between Co3O4 and aluminacould result in partial substitution of Co3+ ions in Co3O4spinel by Al3+ ions. These Co-SCF series include CoAl2O4,

506 Ali Karimi et al./ Journal of Natural Gas Chemistry Vol. 19 No. 5 2010

Co2AlO4, Co3O4 etc. Zhang et al. [19] reported catalyst de-activation for the Re-Co/Al2O3 catalyst, which was attributedto the interaction between Co and Re with the support in thepresence of water vapor, decreasing the extent of reduction ofcobalt oxides. Tavasoli et al. [23] also reported that the maindeactivation mechanisms of Ru-Co/Al2O3 catalyst at high wa-ter partial pressure are: cobalt re-oxidation, metal-support in-teractions and aluminates formation.

As shown in Figure 2 when H2 partial pressure in reduc-tion gas compositions (RGC) decreases, the cobalt aluminatepeak from XRD spectra disappears, which shows that the par-tial pressure of water on the catalyst surface is also decreasing.

Figure 3. TPR profiles of the fresh and pretreated catalysts

TPR profiles of the catalyst samples for various pretreat-ment conditions are shown in Figure 3. In this Figure, sev-eral hydrogen consumption peaks can be seen in the TPRcurves of the fresh and pretreated catalysts. The first low-temperature peak of the reduction process occurred between327 and 450 ◦C for the fresh catalyst can be attributed tothe reduction of supported cobalt nitrate [10,11,24]. Becausecalcination temperatures above 450 ◦C are needed to com-pletely decompose the supported cobalt nitrate, some nitratewill always be present in the calcined samples. This low-temperature reduction peak was eliminated in the pretreatedcatalysts because of complete decomposition of the supportedcobalt nitrate in these samples. The last two peaks most likelyrepresent a two-step reduction of Co3O4 to cobalt metal withCoO as an intermediate species:

Co3O4+H2 −→ 3CoO+H2O (5)

3CoO+3H2−→ 3Co0+3H2O (6)A two-step reduction fits well with the hydrogen con-

sumption ratios in the two separate peaks for all of the cat-alysts in this study. It should be mentioned that the broadreduction range at 427−727 ◦C seems to comprise more thanone peak. Bulk Co3O4 can be completely reduced at temper-atures below 527 ◦C, while Co2AlO4 and CoAl2O4 speciescan be reduced at higher temperatures and are therefore rec-ognized as irreducible phases [23]. As shown in Figure 3,

different shoulders appear in TPR profiles of H100 and H66samples. These different TPR patterns for H100 and H66samples may be related to higher water partial pressures inthe reduction process that give varying degrees of interactionbetween cobalt species and the support, thus retarding the re-ducibility [10−12].

The degree of reduction is an important parameter, con-taining information about the interaction with the support. Acorrected dispersion can also be calculated based on the de-gree of reduction [24]. When TPR is used to estimate thedegree of reduction, reduction of Co3O4 to Co0, requiring2.67 mol of H per mol of cobalt, is assumed. For the calcinedγ-Al2O3-supported catalysts, some of the cobalt remains ascobalt nitrate after calcination, which is not taken into accountin the calculations from TPR. The degree of reduction fromO2 titration is based on the assumption of complete oxidationfrom Co0 to Co3O4.

Catalyst dispersion, average particle size, and the degreeof reduction measured using oxygen titration and TPR for allof the catalysts are given in Table 3. As shown in Table 3,the extent of reduction calculated from oxygen titration datawas significantly lower than that calculated from TPR. BorgØet al. [25] suggested that incomplete oxidation of the re-duced catalysts with oxygen pulses is the cause. Khodakovet al. [26] also indicated that oxygen titration might underes-timate the extent of reduction. They showed that in an inertatmosphere at temperatures above 350 ◦C, the supported CoOphase could be more stable than Co3O4 [27]. Thus, oxygentitration conducted in helium at 400 ◦C could result in oxida-tion of cobalt metal phases to CoO or a mixture of CoO andCo3O4 instead of Co3O4 [26].

The chemisorption results given in Table 3 have been cal-culated with the assumptions that the adsorption stoichiome-try is H : Co = 1 and that Ru does not contribute to the amountof chemisorbed hydrogen. As shown in Table 3, volumet-ric chemisorption gives a higher dispersion for all catalysts.According to previous studies, this is due to the promotioneffect of rutheniumon the reduction of highly dispersed cobaltphases [16,17,23]. The same promoting effect of Ru is notobserved from the XRD measurements. The main reason isprobably that the highly dispersed phases present poor crys-tallinity which is beyond the detection limit of XRD. Also theXRD measurements were done on oxide catalysts instead ofon reduced catalysts as in volumetric chemisorption.

As shown in Table 3, H33 and H50 catalysts have higherdispersion of cobalt and larger particle diameter than the othercatalysts. The larger cobalt particle diameters for H33 andH50 samples might be related to lower water partial pressuresin the reduction process that give varying degrees of interac-tion with the support. Literature results showed that the largecobalt particles located in wide pores of alumina support weremore easily reduced than small particles present in narrowpores because of weaker metal-support interactions for largeparticles [24−27]. As shown in Table 3 the degree of reduc-tion for H50 catalyst is higher than those of the others. In H33catalyst, the diffusion of H2 in catalyst pores is minimized dueto low water vapor partial pressure [13]. So, the degrees of

Journal of Natural Gas Chemistry Vol. 19 No. 5 2010 507

Table 3. H2 temperature-programmed desorption (TPD) and re-oxidation of catalysts

Reduction (%) Dispersion (%) dp Co (nm)CatalystTPR O2 titration XRD (H2-TPD) XRD H2-TPD

H100 71 58 5.3 5.7 18 17H66 77 61 4.8 5.2 20 19H50 82 64 4.4 4.9 22 20H33 75 60 4.3 4.9 22 20

reduction become low and the number of Co metal active sitesis decreasing.

3.2. Fischer-Tropsch synthesis

After reducing the catalysts with different RGC, FTS testswere carried out under 220 ◦C, H2/CO = 2, GHSV = 1200 h−1and 1 atm. The effluent of the reactor was analyzed by an on-line Varian 3800 gas chromatograph for the calculation of CO

conversion and C1–C20 hydrocarbons selectivity. The resultsfor CO conversion and C5+ selectivities are illustrated in Ta-ble 4. It has been suggested that the FTS activity of cobaltcatalyst was directly dependent on the catalyst reducibility[10−18]. The cluster size of cobalt in alumina-supported cat-alyst influences its stability and reducibility, and thus its ac-tivity for FTS. In case of highly dispersed cobalt on support(lower cobalt cluster size), cobalt aluminate can be formedduring pretreatment, reduction and catalytic reaction, whichcannot be easily reduced to cobalt metal [10−17].

Table 4. Catalysts activity (after 8 h time-on-stream)

Product selectivity (%)Catalyst CO conversion (%) FTS rate/gCH/(gCat·h) αCO2 CH4 C2–C4 C5+

H100 76 0.31 0.91 1.5 9.1 5.2 84.2H66 82 0.33 0.90 1.3 9.8 5.8 84.4H50 85 0.35 0.90 1.3 10.2 6.3 83.5H33 82 0.33 0.91 1.4 9.6 5.1 85.3

As shown in Table 3, the pretreatment conditionsinfluence the catalyst reducibility, cobalt dispersion and cobaltcluster size. The pretreated H100 catalyst has lower catalystreducibility and smaller cobalt cluster size because of higherH2O partial pressure in reduction atmosphere. As shown inTable 4, H100 catalyst has lower catalytic activity becauseof lower catalyst reducibility and smaller cobalt cluster size.Also higher FTS activity produced higher H2O byproduct andchanged the catalytic activity. Bulk oxidation of cobalt bywater is not permitted thermodynamically under FT condi-tions; only small Co clusters interact with support and are re-oxidized in the presence of water. Commercial FTS practicerequires that cobalt catalysts withstand long-term use at highCO conversions, in which water concentrations approach sat-uration levels during reaction conditions and may even con-dense within catalyst support pores. In case of a scale upstrategy, it is necessary to determine a specific procedure forcatalyst reduction.

4. Conclusions

The important point in the development of FTS is the im-provement of the catalyst activity and selectivity by increas-ing the number of active Co metal sites that are stable duringreduction and under reaction conditions. The reducibility andH2 consumption of H50 catalyst sample are higher and its Co-SCF is lower than those of the other ones, hence, H50 catalystcould stand long-term use at high CO conversions.

These results show that catalyst reduction under higher

H2 partial pressure changes the FTS activity in two routes:(i) decrease the catalyst reducibility under pretreatment con-ditions which decreases the FTS activity, (ii) decrease thecobalt cluster size because of stronger cobalt-support inter-action. The cobalt may be oxidized under FTS condition byH2O byproduct.

References

[1] Anderson R B. The Fischer-Tropsch Synthesis. Orlando: Aca-demic Press, 1984

[2] Dry M E. In: Anderson J R, Boudart M ed. Catalysis-Scienceand Technology, New York: Springer-Verlag, 1981. 159

[3] Bartholomew C H. Recent Developments in Fischer-TropschCatalysis, New Trends in CO Activation, Studies in Surface Sci-ence and Catalysis. Amsterdam: Elsevier, 1991. 64

[4] van der Laan G P, Beenackers A A C M. Catal Rev Sci Eng,1999, 41: 255

[5] Eliason S A, Bartholomew C H, Fuentes G A. Catalyst Deacti-vation. Amsterdam: Elsevier, 1997. 517

[6] Pour A N, Shahri S M K, Zamani Y, Irani M, Tehrani S. J NaturGas Chem, 2008, 17: 242

[7] Pour A N, Shahri S M K, Bozorgzadeh H R, Zamani Y, TavasoliA, Marvast M A. Appl Catal A, 2008, 348: 201

[8] Pour A N, Zamani Y, Tavasoli A, Shahri S M K, Taheri S A.Fuel, 2008, 87: 2004

[9] Pour A N, Taghipoor S, Shekarriz M, Shahri S M K, Zamani Y.J Nanosci Nanotech, 2009, 9: 4425

[10] Chu W, Chernavskii P A, Gengembre L, Pankina G A, Fongar-land P, Khodakov A Y. J Catal, 2007, 252: 215

508 Ali Karimi et al./ Journal of Natural Gas Chemistry Vol. 19 No. 5 2010

[11] BorgØ, Eri S, Blekkan E A, Storsæter S, Wigum H, Rytter E,Holmen A. J Catal, 2007, 248: 89

[12] Khodakov A Y. Catal Today, 2009, 144: 251[13] Iglesia E. Appl Catal A, 1997, 161: 59[14] Dalai A K, Davis B H. Appl Catal A, 2008, 348: 1[15] Tavasoli A, Karimi A, Sadaghiani K, Khodadadi A, Mortazavi

Y, Sharifnia S. US 7 226 954B2. 2006[16] Tavasoli A, Mortazavi Y, Khodadadi A, Sadagiani K. I J Chem

Chem Eng, 2005, 35: 9[17] Tavasoli A, Mortazavi Y, Khodadadi A, Karimi A. I J Chem

Chem Eng, 2005, 36: 25[18] Jacobs G, Patterson P M, Zhang Y Q, Das T, Li J L, Davis B H.

Appl Catal A, 2002, 233: 215[19] Zhang Y L, Wei D G, Hammache S, Goodwin Jr J G. J Catal,

1999, 188: 281

[20] Jongsomjit B, Sakdamnuson C, Praserthdam P. Mater ChemPhys, 2005, 89: 395

[21] Shroff M D, Datye A K. Catal Lett, 1996, 37: 101[22] Pour A N, Housaindokht M R, Tayyari S F, Zarkesh J. J Natur

Gas Chem, 2010, 19: 333[23] Tavasoli A, Abbaslou R M M, Dalai A K. Appl Catal A, 2008,

346: 58[24] BorgØ, Hammer N, Eri S, Lindvag O A, Myrstad R, Blekkan

E A, Rønning M, Rytter E, Holmen A. Catal Today, 2009, 142:70

[25] BorgØ, Rønning M, Storsæter S, van Beek W, Holmen A. StudSurf Sci Catal, 2007, 163: 255

[26] Khodakov A Y, Griboval-Constant A, Bechara R, ZholobenkoV L. J Catal, 2002, 206: 230

[27] Khodakov A Y, Lynch J, Bazin D, Rebours B, Zanier N, Mois-son B, Chaumette P. J Catal, 1997, 168: 16