Uh

Ta

b

a

ARRA

KG1DC

1

bmmmiPaHdh

tmlpp4l[

tu

1d

Chemical Engineering Journal 179 (2012) 302– 311

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Journal

jo u r n al hom epage: www.elsev ier .com/ locate /ce j

se of delaminated hectorites as supports of copper catalysts for theydrogenolysis of glycerol to 1,2-propanediol

atiana Sáncheza, Pilar Salagrea,∗, Yolanda Cesterosa, Agustin Bueno-Lópezb

Departament de Química Física i Inorgànica, Universitat Rovira i Virgili, C/Marcel·lí Domingo s/n, 43007 Tarragona, SpainDepartamento de Quimica Inorganica, Universidad de Alicante, Ap. 99, E-03080 Alicante, Spain

r t i c l e i n f o

rticle history:eceived 29 July 2011eceived in revised form 28 October 2011ccepted 4 November 2011

a b s t r a c t

Delaminated hectorites were used as supports of copper catalysts for the hydrogenolysis of glycerol.Catalytic results were compared with those obtained for copper catalysts prepared with other supports:hectorites, synthesized by conventional heating and with microwaves, and silicate-bearing hectorite.Copper was introduced by impregnation, solids blend with ultrasounds, and ion-exchange methods.

eywords:lycerol hydrogenolysis,2-Propanediolelaminated hectoritesopper catalysts

Catalytic tests were carried out in liquid phase at 40 bar, 473 K and stirring of 400 rpm. The selectivity to1,2-propanediol was very high for all catalysts (>90%). The highest conversion (61% after 8 h of reaction)was obtained for a delaminated hectorite impregnated with 40 wt.% of copper. This was related to thehighest surface area of this delaminated hectorite together with its lower hydrophilic character (lowerC.E.C.), which made possible to best control copper agglomeration. In fact, copper dispersion increased

talys

during reaction for this ca

. Introduction

The recent increase of biodiesel production as renewable com-ustible has caused some concern oversupply of glycerol in thearket. With the use of biodiesel worldwide, glycerol becomesore readily available while descending price [1]. This has pro-oted the development of novel technologies to convert glycerol

nto high added-value products, such as 1,2-propanediol (1,2-DO), which can be used as humectant, in antifreeze mixtures,s brake fluid or as component of polyesters and alkyl resins [2].owever, 1,2-propanediol is currently obtained from petroleumerivatives. Therefore, its production from renewable resources isighly desired [2].

Catalytic hydrogenolysis of glycerol is one of the most attrac-ive green routes to obtain selectively glycols. Supported noble

etals, such as Ru, Rh, and Pt are well known active cata-ysts for this reaction [3–14]. Unfortunately, these catalysts oftenromote excessive C–C cleavage, resulting in low selectivity toropanediols. The best result at mild reaction conditions (30 bar,23 K) was recently achieved with a Ru/bentonite catalyst, which

ed to 67% of selectivity to 1,2-PDO for a 72.5% of conversion13].

A less expensive and more effective alternative for cracking con-rol is the use of copper catalysts. Cu/ZnO have been the mostsed catalysts for this reaction [15–19]. In the first studies, high

∗ Corresponding author. Tel.: +34 977559571; fax: +34 977559563.E-mail address: pilar.salagre@urv.cat (P. Salagre).

385-8947/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.cej.2011.11.011

t.© 2011 Elsevier B.V. All rights reserved.

selectivity to 1,2-PDO (80–100%) was reported but working athard reaction conditions (80–150 bar, 453–545 K) [15,16]. Morerecently, Balaraju et al., working at softer conditions (20 bar, 473 K),achieved 37% of conversion with 92% of selectivity to 1,2-PDO after16 h [17] whereas Claus et al. obtained 55% of conversion and 86% ofselectivity to 1,2-PDO, at 50 bar and 473 K, after 7 h of reaction. Theuse of an organic solvent during reaction instead of water reducedthe agglomeration of metal particles, the main responsible for cat-alyst deactivation [18]. These authors proposed that the presenceof Ga2O3 in Cu/ZnO catalysts can control deactivation yielding highconversion (99%) and high selectivity to 1,2-PDO (80%) working at50 bar and 493 K [19]. Vasiliadou and Lemonidou attributed deacti-vation to a collapse of mesoporous network with agglomeration ofmetal particles when using monometallic Cu (20 wt.%) and bimetal-lic (5 wt.% Ru–Cu) catalysts supported in mesoporous silica for thisreaction [20].

Other supported-copper catalysts were used for this reaction.Huang et al. reported selectivity to 1,2-PDO of 99% for a 29% ofconversion with Cu/SiO2 after 12 h of reaction, working at 90 barand 453 K [21]. These authors suggested that the mechanism toobtain 1,2-PDO proceeded through the acetol route, as previouslyproposed by Suppes et al. [22]. This acetol route is favored by thepresence of acid sites, but it can also proceed via metal sites thatcatalyze dehydration of glycerol [14,23]. Lastly, other recent inter-esting paper by Claus et al., using different oxides (ZnO, SiO2, CaO,

Al2O3) as supports of copper catalysts, pointed out that the catalystwith the highest activity (52% of conversion and 98% of selectivityto 1,2-PDO after 7 h at 473 K and 50 bar) was that with the highestmetallic area, Cu/ZnO [24].

T. Sánchez et al. / Chemical Engineerin

hepstehlasslqacadscum[oohp

2

2

2

mStwfwpih1i

t[roauw1amL

G

Scheme 1. Hydrogenolysis of glycerol 1,2-propanediol.

Hectorites (Mn+x/n·yH2O[Mg6−xLix](Si8)O20(OH·F)4), and related

ectorite compounds, are layered materials with high cationxchange capacity (C.E.C.), and with interesting surface and acidicroperties [25], which make them suitable to be used as catalyticupports for this reaction. The main disadvantage of hectorite ishe collapse of its structure at relatively low temperatures. Thisncouraged the design of new methodologies to synthesize stableectorites, such as pillared clays or delaminated hectorites. Pil-

ared clays are mostly prepared using bulky cations (e.g. Al137+)

s pillaring species [26–31]. Besides, Torii and Iwasaki reported theynthesis of silicate-bearing hectorites [32], which presented, athort time of hydrothermal treatment, silicate pillars in the inter-amellar space. Delaminated hectorites can be synthesized usinguaternary ammonium salts during the hydrothermal treatment,s proposed by Iwasaki [33], or by preparing polymer-hectoritelays [34]. The mesoporous of delaminated hectorites are obtainedfter removing the quaternary salt and the polymer, respectively,uring calcination. They correspond to antiparticle space of disk-haped crystallites aggregated by edge-to-face bonding. Hectoritesan be used as supports of copper catalysts. Typical methodssed for preparing catalysts include impregnation, solid blend andodification of hectorite by exchange or with copper complexes

35].The aim of this study was to compare the catalytic activityf several Cu/delaminated hectorites with the catalytic behaviorf Cu/hectorites and Cu/silicate-bearing hectorite for the glycerolydrogenolysis to obtain 1,2-propanediol (Scheme 1). For this pro-osal, different wt.% and dispersion of copper were used.

. Experimental

.1. Preparation of catalysts

.1.1. Synthesis of supportsTwo delaminated hectorites were prepared according to the

ethod reported by Iwasaki [33]. The molar ratio of reagents wasi:Mg:Li = 8:5.2:0.8. An acidified silicate solution was mixed withhe appropriate amounts of MgCl2 and LiF. Then, a LiOH solutionas added until pH 12. The resulting suspension was maintained

or 15 min in an ultrasound bath, filtered, washed with deionizedater, and finally, dried overnight at 353 K. This solid was sus-ended in a trimethyldodecylammonium chloride solution (AQ)

n a molar ratio Li:AQ = 1:1. The suspension was submitted toydrothermal treatment in a conventional autoclave at 453 K for

h, or 2 h. After that, they were calcined at 893 K for 75 min obtain-ng HD1 and HD2, respectively.

One silicate-bearing hectorite was prepared by hydrothermalreatment following the procedure proposed by Torii and Iwasaki32] but using a different quaternary salt. The molar ratio ofeagents was Si:Mg:Li = 8:5.4:0.6. A sodium silicate solution, previ-usly acidified with nitric acid 1 M, was mixed with the appropriatemount of MgCl2. Then, a sodium hydroxide solution was addedntil pH 12. The slurry obtained was dispersed in water and mixedith the appropriate amount of LiF, stirring was maintained for

h and heated hydrothermally at 453 K for 2 h, filtered and driedt 353 K. Finally, this solid was mixed with trimethyldodecylam-

onium chloride, by refluxing at 353 K for 1 h, in molar ratio

i:AQ = 1:1, and then filtered and calcined at 893 K for 75 min (HB).Two traditional hectorites were prepared following the

ranquist and Pollack method [36], and the Vicente et al. method

g Journal 179 (2012) 302– 311 303

[37], respectively. The preparation was carried out as follows:135 mL of a slurry composed by SiO2, fresh precipitate Mg(OH)2and LiF in a molar ratio of Si:Mg:Li = 6.9:5:2:1.6 was vigorouslystirred for 1 h. The SiO2 amount used was lower than the stoichio-metric value to avoid the presence of residual SiO2 in the resultinghectorite whereas the amount of LiF was chosen considering thisreagent as source of Li for the lamellars (0.8) and for the interlamel-lar space (0.8). In one method the sample was aged in a conventionalautoclave at 393 K for 8 days (HC), and in the other the sample wasautoclaved in a laboratory microwave oven at 393 K for 8 h (HMw).Samples were filtered and dried at 353 K.

2.1.2. Preparation of supported-copper catalystsSupported-copper catalysts were prepared by 3 methods:

impregnation, solids blend with ultrasounds, and ion-exchange.Impregnated catalysts were prepared mixing 1.5 g of support

with different volumes of a 15 wt.% copper nitrate ethanol solutionunder ultrasounds for 15 min. Different charges of copper (12.5,20 and 40 wt.% Cu) were obtained after solvent rotaevaporation,calcination at 723 K for 2 h, and reduction under pure hydrogen(2 mL/s) at 573 K for 2 h. Impregnated catalysts were designated asCu/support nameimp1, Cu/support nameimp2, Cu/support nameimp3,where 1, 2 and 3 represent 12.5, 20 and 40 wt.% Cu, respectively.

Solids blend with ultrasounds method consisted of an effectivemixture of support and copper nitrate (15 wt.% Cu) until the obten-tion of a homogeneous solid with a copper charge of 40 wt.%. Then,decane was added, and the mixture was submitted to ultrasoundsfor 15 min. The solid was separated by decantation, dried, calcinedat 723 K for 2 h, and reduced under pure hydrogen (2 mL/s) at 573 Kfor 2 h. These catalysts were named as Cu/support namesb.

Lastly, exchanged copper catalysts were obtained by stirring1.5 g of support in an aqueous 0.01 M copper nitrate solution for30 min. Samples were filtered, and washed with deionized water.The reduction step was performed “in situ” under catalytic reactionconditions. Exchanged catalysts were called as Cu/support nameex.

2.2. X-ray diffraction (XRD)

XRD measurements were made using a Siemens D5000 diffrac-tometer (Bragg–Brentano parafocusing geometry and vertical�–� goniometer) fitted with a curved graphite diffracted-beammonochromator and diffracted-beam Soller slits, a 0.06◦ receivingslit, and scintillation counter as a detector. The angular 2� diffrac-tion range was between 5◦ and 70◦. Sample was dusted on to a lowbackground Si(5 1 0) sample holder. The data were collected withan angular step of 0.05◦ at 3 s per step and sample rotation. Cu K�radiation was obtained from a copper X-ray tube operated at 40 kVand 30 mA.

The X-ray pattern was analyzed implementing the programTOPAS 3.0 [38]. This approach calculates the contribution to thereflection width produced by a specific instrument configuration.The crystallite size was calculated from the net integral breadthof the reflections, ˇi [39], according to the following formula thatcomes from the Scherrer expression: ˇi = �/εcos �, where � is theX-ray wavelength, ε is the crystallite size and � is the Bragg angle.0 6 0 reflection was used to calculate the crystallite size of the hec-torites.

2.3. N2 physisorption

N2-adsorption–desorption isotherms were recorded at 77 Kusing a Micromeritics ASAP 2000 surface analyzer. Prior to anal-ysis samples were outgassed at 393 K. Specific surface areas werecalculated from the BET method.

304 T. Sánchez et al. / Chemical Engineering Journal 179 (2012) 302– 311

Table 1Characterization of the supports.

Clay Nomenclature BET (m2/g) Si/Mga ratio C.E.C. (mequiv./100 g sample) Hectorite crystallite size (0 6 0) (nm)

Delaminated hectorite 1 HD1 353 1.12 (1.53) 44 6.0Delaminated hectorite 2 HD2 328 1.10 (1.53) 60 7.4Silicate-bearing hectorite HB 220 0.72 (1.48) 47 6.7Conventional hectorite HC 207 1.4 (1.33) 70 7.9

33)

2

PA

2

t2r

2

m

2

F

6

2

pdc1aftpCrtee

o

H

o

D

versi

ber

Microwaved hectorite HMw 199 1.6 (1.

a Values between parentheses correspond to the theoretical Si/Mg ratio.

.4. X-ray fluorescence (XRF)

Elemental analyses of the samples were obtained with a PhilipsW-2400 sequential XRF analyzer with Phillips Super Q software.nalyses were made in triplicate for each sample.

.5. Transmission electronic microscopy (TEM)

TEM images were collected using a JEOL 1011 Transmission Elec-ron Microscope operating at 80 kV and magnification values of00–350k. Samples were dispersed in hexane, and a drop of theesultant suspensions was poured on carbon coated-copper grids.

.6. Cation exchange capacity (C.E.C.)

The cation exchange capacity (C.E.C.) was determined by theethod reported by Bergaya and Vayer [40].

.7. FT-IR

Infrared spectra were recorded on a Bruker-Equinox-55TIR spectrometer. The spectra were acquired by accumulating

4 scans at 4 cm−1 resolution in the range of 400–4000 cm−1.

.8. Temperature-programmed reduction (TPR)

Copper dispersion was determined by selective temperature-rogrammed reduction of surface copper following the methodescribed by Gervasini and Bennici [41]. These experiments werearried out in a Micromeritics device, model Pulse ChemiSorb 2705.00 mg of catalyst were heated at 10 K/min from 298 to 673 K under

5% H2/Ar flow (15 mL/min), holding the maximum temperatureor 30 min. The H2 consumption was monitored with a TCD detec-or. Then, the selective oxidation of the copper surface to Cu2O waserformed under 0.53% N2O/Ar flow (15 mL/min) at 323 K for 1 h.u2O surface was further reduced with 5% H2/Ar (15 mL/min) byaising the temperature at 20 K/min from 298 to 1173 K, followinghe H2 consumption with the TCD detector. The potential interfer-nce of H2 consumption by the supports was ruled out in blackxperiments performed with the supports (without copper).

Surface copper was determined considering the stoichiometryf the reaction:

2 + Cu2Osurface → 2Cusurface + H2O

“Dispersion (%)” was calculated as the ratio between the amount

Con

Selectivity (%) = num

f surface copper and total copper in the catalyst.

ispersion (%) = Cusurface

Cutotal× 100

75 10.5

A CuO sample (supplied by Micromeritics) was used as referenceto quantify H2 consumption.

2.9. Catalytic activity tests

Glycerol hydrogenolysis was carried out in a 50 mL stainlesssteel autoclave reactor at a stirring speed of 400 rpm. An aque-ous solution of glycerol (60 wt.%) was prepared with pure glycerol(≥99% Glycerol ReagentPlus (GC) Sigma–Aldrich) and deionizedwater to be used as feed. 30 mL of glycerol solution and 1.5 g ofcatalyst were loaded into the reactor for every run. The reactorwas purged four times with N2 and pressurized with H2 to 40 bar.The mixture was heated up to 473 K and maintained for 8 or 24 hdepending on the catalyst. Hydrogen was fed on demand so asto keep the pressure at 40 bar during 8 or 24 h. The liquid phaseproducts were analyzed by gas chromatography using a chromato-graph model Shimadzu GC-2010 equipped with a 60 m capillarycolumn SupraWAX-280 coated with polyethylene glycol and a FIDdetector.

The parameters used to evaluate the catalytic activity of the cat-alysts were conversion of glycerol, and selectivity to the reactionproducts, calculated as follows:

on (%) = number of moles of converted glyceroltotal number of moles of glycerol in the feed

× 100%

of moles of glycerol converted to the reaction productnumber of moles of converted glycerol

× 100%

The number of moles of converted glycerol was calculated from thetotal amount of carbon-based species formed during reaction.

TOF was calculated, before and after reaction (TOF1 and TOF2,respectively) for the most representative catalysts, as follows:

TOF = moles of glycerol converted to 1,2-PDOsurface copper atoms × hour

The number of moles of glycerol converted to 1,2-PDO was calcu-lated from the yield to 1,2-PDO (Conversion × Selectivity).

3. Results and discussion

3.1. Characterization of supports

Table 1 summarizes the characteristics of the synthesized hec-torites.

XRD patterns of all samples were typical of clays materials(Figs. 1 and 2). However, 0 0 1 reflection, related to the layer stack-ing, was only observed in traditional hectorites (HC and HMw). Inorder to confirm this fact, additional diffractograms were recordedat lower angles (2� = 0–10◦) for samples HD1, HD2 and HB but anynew reflection was observed. The low degree of order stacking inmesoporous hectorites could be associated with delamination in

HD hectorites, and with a very limited ordering in the c-axis for HB.In contrast, 0 6 0 reflection, which appeared more or less defined inall samples, was related to the crystallinity of their layers, whichwas evaluated from the Scherrer equation. Higher crystallite sizes

T. Sánchez et al. / Chemical Engineering Journal 179 (2012) 302– 311 305

erns o

ithapodtbt

hHmosa(

sif

Fig. 1. X-ray diffraction patt

nvolve higher sample crystallinity. Regarding synthesized hec-orites, the highest crystallite size was obtained for the traditionalectorite synthesized using microwaves, HMw (10.5 nm). This is ingreement with the increase of crystallinity expected for materialsrepared with microwaves [37,42]. The crystallite size of the restf hectorites, prepared by conventional hydrothermal treatments,id not show significant differences, except for delaminated hec-orite HD1, which had the lowest crystallite size (6.0 nm). This cane explained by the shorter time of hydrothermal treatment usedo prepare this sample (1 h).

TEM images showed lamellar morphology for all synthesizedectorites (Figs. 3 and 4). For delaminated mesoporous hectorites,D1 and HD2, we can see nano-sized layers aggregated to makeesoporous (Fig. 3). This explains why 0 0 1 reflection was not

bserved by XRD for these samples. HD1 had smaller lamellarizes than HD2 according to XRD results. Silicate-bearing hectoritend traditional hectorites were mainly formed by oriented layersFig. 4).

In order to evaluate the presence of amorphous phases, FTIR

pectra were performed for all samples. The FTIR spectrum shownn Fig. 5 corresponds to sample HD1. Similar spectra were obtainedor HD2 and HB hectorites. The band around 1024 cm−1 has

Fig. 2. X-ray diffraction patterns of delam

f samples HB, HC and HMw.

been assigned to the Si–O–Si ordered in the clay mineral bycomparison with the data reported in the literature [43]. The pres-ence of amorphous silica was discarded for all samples since thebands characteristic of this compound, which appear at 795 cm−1,1105 cm−1 with a shoulder at 1200 cm−1 [43], were not observed.Other band at 688 cm−1 was associated with vibrations related toMg–OH librations.

The Si/Mg ratio was determined by XRF for all samples (Table 1).HD1, HD2 and HB presented lower Si/Mg ratio values than the the-oretical ones. This could be related to the high pH used in thepreparation of these materials (pH 12) that could favor the for-mation of soluble silicates. This variation was more significant forHB, probably due to the presence of interlamellar silicates, whichcould be more easily solubilized during the treatment with thequaternary ammine in the second part of its synthesis. For tradi-tional hectorites, higher Si/Mg ratio values than the expected oneswere obtained in spite of the lower amounts of SiO2 used. Thiscould be associated with some magnesium solubilization. This isin agreement with the mechanism proposed by Baird et al. for the

construction of the structure in traditional hectorites [44].

Higher BET areas were obtained for mesoporous delami-nated hectorites HD1 and HD2 with values of 353 and 328 m2/g,

inated hectorites: HD1 and HD2.

306 T. Sánchez et al. / Chemical Engineering Journal 179 (2012) 302– 311

inated

raHs(tBrTtss

st

Fig. 3. TEM micrographs of delam

espectively (Table 1). These values agree with the delaminationnd lamellar size observed by TEM for these samples (Fig. 3), sinceD1 presented smaller lamellar size. Silicate-bearing hectorite (HB)

howed just slightly higher BET area than traditional hectoritesaround 220 m2/g in front of 200 m2/g). It is important to note thathe silicate-bearing hectorites prepared by Tori at 453 K presentedET areas around 560 m2/g after calcination. This high value waselated to the presence of silicate pillars in the interlamellar space.he lower BET area of our HB sample can be attributed to the qua-ernary salt used for its synthesis, which seems to favor the silicateolubilization. The higher layer crystallinity of HMw explains its

lightly lower BET area with respect to HC.

With respect to porosity, it is important to remark that allynthesized hectorites exhibited adsorption isotherms with con-ribution of mesoporosity (type IV), but with different hysteresis

Fig. 4. TEM micrographs of samples

hectorites: (a) HD1 and (b) HD2.

loops. For HC and HMw, hysteresis was type B (following Boer clas-sification [45]) (Fig. 6). Hysteresis type B is associated with lamellarparticles that are packed together with formation of open slit-shaped capillaries with parallel walls, whereas hysteresis type Dis more characteristic of lamellar particles that are packed withoutparallel orientation. The hysteresis loop of the silicate-bearing hec-torite (HB) and delaminated hectorites (HD1, HD2) were identifiedas a mix of B and D types, with differences in their contribu-tion (Fig. 7). Thus, in HB, we observed higher contribution of Bwhereas in delaminated hectorites, D was the main contribution.This is in agreement with the disorder observed in the lamellar

distribution for delaminated samples (Fig. 3). The contribution ofmicroporosity in HD and HB hectorites was not negligible, since weobserved significant values of nitrogen adsorption at lower pres-sures.

: (a) HB, (b) HC, and (c) HMw.

T. Sánchez et al. / Chemical Engineerin

Fig. 5. FTIR spectra of HD1imp3.

Fig. 6. N2 adsorption/desorption isotherms for samples HC and HMw.

Fig. 7. N2 adsorption/desorption isotherms for samples HB, HD1 and HD2.

g Journal 179 (2012) 302– 311 307

C.E.C. values, calculated for all samples, are indicated in Table 1.The higher C.E.C. values were obtained for traditional hectorites(HC and HMw). This can be related to their best constructed struc-ture. Regarding mesoporous hectorites, the highest C.E.C. valuewas obtained for HD2 (60 mequiv./100 g sample). This could bejustified by the longer time of hydrothermal treatment of HD2respect to HD1. The values are lower than the expected ones(around 100 mequiv./100 g sample). To understand this behavior,it is important to have in mind the short times of hydrother-mal treatment that we used in the synthesis of these materials,since probably in this time the expected stoichiometry was notcompletely achieved (Table 1). HB showed lower C.E.C. value(47 mequiv./100 g sample) than the delaminated sample preparedat the same temperature and time of hydrothermal synthesis (HD2,60 mequiv./100 g sample). This is in agreement with the highermagnesium content obtained for HB (Table 1).

3.2. Catalytic behavior of Cu/hectorites in the hydrogenolysis ofglycerol

Table 2 presents the results obtained with different percent-ages of copper impregnated in delaminated hectorite prepared in1 h of hydrothermal treatment (HD1). The percentages of copperused were 12.5 wt.%, 20 wt.% and 40 wt.%. High selectivity valuesto 1,2-PDO (>90%) were obtained with all catalysts. At higher Cuwt.%, conversion increased whereas selectivity to 1,2-PDO slightlydecreased from 99% to 93% with the formation of cracking products,such as 1-propanol, 2-propanol, acetol and ethyleneglycol.

New experiments were performed with copper supportedin delaminated mesoporous hectorites (HD1, HD2). The resultsare shown in Table 3. The copper content, determined by XRFfor exchanged samples, was 1.4 wt.% for Cu/HD1ex, 1.9 wt.% forCu/HD2ex, and around 40 wt.% for the rest of samples, values thatcorrespond to the amount used in the preparation.

All catalysts showed very high selectivity to 1,2-propanediol(93–99%) but different conversion values. Higher conversions wereobtained when using support HD1, with a maximum of 61.4%, at 8 hof reaction, for the catalyst prepared by impregnation with the 40%of copper. A better copper dispersion in the delaminated hectoritewith higher surface area (HD1) could explain this result. It is impor-tant to note that significant differences in the copper particle sizebetween the used Cu/HD1 catalysts, prepared by different meth-ods, were observed (Fig. 8). After reaction, the catalyst with thesmallest copper particles (around 12 nm) was Cu/HD1imp3, whilefor Cu/HD1sb and Cu/HD1ex, copper particle sizes were around20 and 90 nm, respectively. Claus et al. concluded that the mostimportant problem in this reaction is copper agglomeration thatinduces deactivation [18,19]. Interestingly, the sequence of con-version for Cu/HD1 catalysts (Cu/HD1imp3 > Cu/HD1sb > Cu/HD1ex)was in reverse order than their copper particle size after reaction(Cu/HD1imp3 < Cu/HD1sb < Cu/HD1ex).

In contrast, for Cu/HD2 catalysts, the sequence of conversionwas different: Cu/HD2sb > Cu/HD2imp3 > Cu/HD2ex. This sequenceand the low activity of Cu/HD2ex (9.8% after 24 h) can be relatedto the higher C.E.C. of HD2 than of HD1, since higher supportagglomeration can be expected during catalyst preparation dueto an increase in the hydrophilic capacity when the number ofinterlamellar cations neutralizing the layer charges increases. Thissupport agglomeration also favors the agglomeration of copperparticles. Support agglomeration should be less important in thecatalyst prepared by the blend solid method since metal particlesare located in the interparticles support space in this sample.

In order to have more information about the influence of thesupport, the catalytic behavior of Cu/HD catalysts was comparedwith that of copper catalysts supported on traditional hectoritesand mesoporous silicate-bearing hectorite (Table 4). The content

308 T. Sánchez et al. / Chemical Engineering Journal 179 (2012) 302– 311

Table 2Activity results of different percentages of copper catalysts supported on delaminated hectorites by impregnation for the hydrogenolysis of glycerol after 8 h of reaction.

Catalyst Copper impregnated (%) Conversion (%) Selectivity (%)

1,2-PDO 2-PRO 1-PRO Others

Cu/HD1imp1 12.5 37.7 99.0 0.1 n.d. 0.9Cu/HD1imp2 20 41.2 98.1 n.d. 0.1 1.8Cu/HD1imp3 40 61.4 93.0 0.2 0.1 6.7

1,2-PDO: 1,2-propanediol; 2-PRO: 2-propanol; 1-PRO: 1-propanol; n.d.: not detected.

Table 3Activity results of copper catalysts supported on delaminated hectorites for the hydrogenolysis of glycerol after 8 h of reaction.

Catalyst Copper introduction method Conversion (%) Selectivity (%)

1,2-PDO 2-PRO 1-PRO Others

Cu/HD1exa Exchange 21.0 99.3 n.d. n.d. 0.7

Cu/HD1imp3 Impregnation 61.4 93.0 0.2 0.1 6.7Cu/HD1sb Ultrasounds solids blend 43.6 96.0 n.d. n.d. 4.0Cu/HD2ex

a Exchange 9.8 97.8 n.d. n.d. 2.2Cu/HD2 Impregnation 11.2 98.1 n.d. n.d. 1.9

opano

oXFc

ot

imp3

Cu/HD2sb Ultrasounds solids blend 24.3

a Reaction time: 24 h. 1,2-PDO: 1,2-propanediol; 2-PRO: 2-propanol; 1-PRO: 1-pr

f copper in the catalysts obtained by exchange, determined byRF, was 1.5 wt.% for HB, 2.2 wt.% for HC and 2.4 wt.% for HMw.or impregnated and ultrasounds solids blended catalysts, copper

ontent was around 40 wt.%.

Conversion of Cu/HC and Cu/HMw catalysts dependedn the method used for copper introduction, and followedhe same sequence than that observed for Cu/HD2 catalysts

Fig. 8. TEM micrographs of used catalysts: (a) Cu/HD1ex,

98.0 1.4 n.d. 0.6

l; n.d.: not detected.

(exchange < impregnation < ultrasounds solid blend). However,their activity was lower than those of the catalysts preparedwith delaminated hectorites (HD1, HD2). For this reason, higher

reaction time (24 h) was needed for this study. The lower activitycould be related to the lower surface areas and higher C.E.C.values of traditional hectorites, since copper agglomeration shouldbe favored at these conditions. When traditional hectorite was

(b) Cu/HD1imp3, (c) Cu/HD1sb and (d) Cu/HD2imp3.

T. Sánchez et al. / Chemical Engineering Journal 179 (2012) 302– 311 309

Table 4Activity of several copper/clay catalysts for the hydrogenolysis of glycerol after 24 h of reaction.

Catalyst Copper introduction method Conversion (%) Selectivity (%)

1,2-PDO 2-PRO 1-PRO Others

Cu/HBex Exchange 13.3 97.0 n.d. 0.1 2.9Cu/HBimp3 Impregnation 50.2 76.0 n.d. n.d. 24.0Cu/HBsb Ultrasounds solids blend 60.2 91.0 0.3 0.2 8.5Cu/HCex Exchange 5.4 85.5 5.6 n.d. 8.9Cu/HCimp3 Impregnation 13.0 96.7 n.d. 0.4 2.9Cu/HCsb Ultrasounds solids blend 35.0 94.6 0.1 n.d. 5.3Cu/HMwex Exchange 14.0 85.9 12.7 n.d. 1.4Cu/HMwimp3 Impregnation 31.8 97.1 n.d. n.d. 2.9

1 d.

sacsml

o2siHl

3

rcstHc

haaaeftncatc

r(1awtwactspdR

that the catalyst prepared by solid blend led to lower amount ofmetal sites but more active. However, it is important do not forgetthat copper agglomeration took place in this catalyst during reac-tion. This can explain its lower conversion when compared with

0

500000

1000000

1500000

2000000

2500000

3000000

3500000

620570520470420

TC D

sig

nal (

arbi

trar

y un

its)

Temperature K

Cu/HD1 imp3Used

Cu/HD1imp 3Fresh

200000

400000

600000

800000

1000000

1200000

1400000

1600000

1800000

TC D

sig

nal (

arbi

trar

y un

its)

Cu/HD1 sb fresh

Cu/HD1 sb used

Cu/HMwsb Ultrasounds solids blend 39.3

,2-PDO: 1,2-propanediol; 2-PRO: 2-propanol; 1-PRO: 1-propanol; n.d.: not detecte

ynthesized with microwaves, higher conversion was obtainedfter 24 h, independently of the method used for introducingopper, while selectivity to 1,2-PDO was high (85–97%) for bothystems. The higher crystallinity of hectorite prepared withicrowaves can decrease its water affinity and, consequently,

ower support agglomeration should occur.Catalytic activity of Cu/HB catalysts was also lower than those

btained for Cu/delaminated hectorites since it was necessary4 h to obtain significant results. The sequence of conver-ion was similar than that of Cu/HD2 catalysts (exchange <mpregnation < ultrasounds solid blend). The lower surface area ofB could explain the higher copper agglomeration of these cata-

ysts, and therefore, the lower conversion values obtained.

.3. Determination of copper dispersion

In order to correlate copper agglomeration with the catalyticesults obtained, copper dispersion of representative fresh and usedatalysts was determined by H2-TPR of selectively N2O-oxidizedurface copper. Figs. 9 and 10 show the H2 consumption profiles forhe experiments performed with fresh and used catalysts HD1imp3,D1sb and HD1ex, HD2imp3 whereas the quantitative values areompiled in Table 5.

The H2 consumption profiles provide information about theomogeneity of copper in the different samples. As a general trend,ll profiles presented a single H2 consumption peak centered atround 523–573 K. The fresh catalysts prepared by ionic exchangend impregnation exhibit quite symmetric H2-consumption peaks,videncing homogeneity on the copper species. In contrast, theresh catalyst prepared by solids blending showed a main peak withwo shoulders. This profile can be attributed to a more heteroge-eous distribution of copper, related to the presence of differentopper species. The H2-consumption profiles changed after the cat-lytic tests due, in one hand, to copper sintering, and on the other,o modifications of the support that can affect the accessibility ofertain copper species.

More detailed information was obtained from the quantitativeesults included in Table 5. In the case of the most active catalystCu/HD1imp3) an increase in the dispersion of copper, from 8% to1%, was observed after reaction. The dispersion increase suggests

low tendency to copper agglomeration during reaction togetherith a certain layer disaggregation, which favors the accessibility

o copper. The same behavior was observed for catalyst Cu/HD2imp3ith an increase in copper dispersion from 2% to 4% during the cat-

lytic tests. However, the lower dispersion of copper in the freshatalyst (2%) and the higher size of the copper particles (25–50 nm)han in Cu/HD1imp3 in the used catalysts, explains its low conver-

ion at 8 h of reaction. The copper dispersion in the fresh catalystrepared by solid blend, Cu/HD1sb, was 9%. After reaction, this valueecreased to 4% with the subsequent loss of copper surface sites.egarding the catalyst prepared by exchange, Cu/HD1ex, higher

93.8 0.1 0.1 6.0

copper dispersion was observed in the fresh catalyst (98%). How-ever, after reaction we observed a clear loss of copper surface atomsreducing the dispersion until 63%. This loss of copper dispersionagrees with the increase of copper particle sizes observed by TEMafter reaction (90 nm, Fig. 8a).

In order to study the influence of the method used to intro-duce copper in the hectorites on the activity characteristics of themetal sites, TOFs values were calculated taking into account thedispersion results before reaction (TOF1) and after reaction (TOF2).The higher TOF was obtained for Cu/HD1sol catalyst after reaction(TOF2 = 67.5 × 10−24 h−1) Having in mind that this catalyst pre-sented lower copper dispersion than Cu/HD1imp3, we can conclude

Temperature K

0670620570520470420

Fig. 9. Cu/HD1imp3 and Cu/HD1sb characterization by TPR-H2.

310 T. Sánchez et al. / Chemical Engineering Journal 179 (2012) 302– 311

Table 5Results of copper surface dispersion on the different supports and TOFs before and after reaction.

Catalyst Cu content (wt.%) DCu(br) (%) DCu(ar) (%) TOF1a (×10−24) TOF2b (×10−24)

Cu/HD1imp3 40.1 8 11 37.4 27.2Cu/HD1sb 32.4 9 4 30.0 67.5Cu/HD1ex 1.4 98 63 10.5 16.4Cu/HD2imp3 40.1 2 4 28.6 14.3Cu/HBimp3 40.1 10 5 6.6 13.2Cu/HCimp3 38.7 6 n.d. 3.8 n.d.Cu/HMwimp3 39.2 4 2 13.8 27.6

b

Crtllaapt

Facso(t

r: before reaction; ar: after reaction; n.d.: not detected.a TOF1 (h−1) referred to copper surface atoms before reaction.b TOF2 (h−1) referred to copper surface atoms after reaction.

u/HD1imp3. The catalyst that presented the lowest TOFs valuesespect to the other Cu/HD1 catalysts was Cu/HD1ex. This meanshat the metal sites of the catalyst prepared by exchange were theowest active. However, the lower activity of the exchanged cata-yst with respect to the impregnated and solid blend ones is also

consequence of the lower amount of surface copper atoms avail-ble, mainly after reaction in the exchanged catalyst, together withroblems of accessibility to the active sites that can be located inhe interlamellar space.

Other impregnated catalysts were analyzed with this technique.or silicate-bearing hectorite Cu/HBimp3, the fresh catalyst had

good dispersion (10%), but after the catalytic test, significantopper sintering was observed (final dispersion = 5%). The loss ofurface copper explains the longer reaction time (24 h) required to

btain higher conversion (50%). In the case of traditional hectoritesCu/HCimp3 and Cu/HMwimp3) the dispersion obtained before reac-ion was very similar (Table 5). However, after reaction the catalyst

0

500000

1000000

1500000

2000000

2500000

3000000

3500000

670620570520470420

TC D

sig

nal (

arbi

trar

y un

its)

Temperature K

Temperature K

Cu/HD1 ex used

Cu/HD1 ex fresh

0

500000

1000000

1500000

2000000

2500000

3000000

3500000

680630580530480430

TC D

sig

nal (

arbi

trar

y un

its)

Cu/HD2 imp3 used

Cu/HD2 imp3 fresh

Fig. 10. Cu/HD1ex characterization by TPR-H2.

whose support was synthesized by conventional method showeda total loss of copper dispersion while that whose support wassynthesized by the microwave-assisted method only showed adecrease in the dispersion but a certain amount of active sites(surface copper) was maintained at the end of reaction. The TOFsof these impregnated catalysts, calculated respect to the coppersurface atoms obtained after reaction (TOF2), were comparable tothose of the other impregnated catalysts tested.

3.4. Study of support influence on mechanism of reaction

Different glycerol hydrogenolysis mechanisms have been pro-posed but, on the whole, a two steps mechanism has been accepted[14,22,23,6]. The first step consists of glycerol dehydration on acidand/or metal sites with the formation of acetol. Then, a hydro-genation step occurs on metal sites resulting in the formation of1,2-propanodiol. In our catalysts, the acid sites should be providedby the hectorites.

In order to study the effect of the accessibility of acid sites onthe reaction, additional catalytic experiments were performed withCu/HD1 and Cu/HD2 catalysts, prepared by exchange, impregnationand ultrasounds solids blend, using the same reaction conditionsthan in previous experiments but with N2 instead of H2, exceptthe first 15 min that was used H2 for an in situ catalysts reduction,after this time all the H2 were eliminated and substituted by N2. Theactivity results are shown in Fig. 11. Conversion was very low forthe six catalysts (Cu/HD1ex, Cu/HD1imp3, Cu/HD1sb and Cu/HD2ex,Cu/HD2imp3, Cu/HD2sb). The highest conversion value was obtainedfor the catalyst Cu/HD2imp3, prepared by impregnation method

(14.6%). With respect to the selectivity, the main product was ace-tol for all catalysts, in higher amounts for Cu/HD1 (60–85%). Thehigher selectivity to acetol of Cu/HD1 catalysts could be explainedby the presence of higher number of accessible acid sites taking into

Fig. 11. Catalytic results of delaminated hectorites-supported copper catalysts atthe same reaction conditions than for glycerol hydrogenolysis but under N2 insteadof H2.

ineerin

ate

rprshtytcac

4

sbc(pciwto(tmapthpttgt

A

M(

R

[

[

[[

[

[[

[

[[

[[

[

[

[[[

[

[

[[[

[[[[

[[

[[[[[

[

T. Sánchez et al. / Chemical Eng

ccount the higher surface area of HD1 support. The hydrogena-ion product, 1,2-PDO, was observed in very low amounts (<7%), asxpected.

The lower conversion obtained for Cu/HD catalysts under N2espect to the values obtained under H2 agree with the idea pro-osed by Sato et al. who reported that hydrogenation is favoredespect to dehydration at temperatures lower than 483 K [46]. Con-equently, dehydration is possible at higher temperature, but noydrogenation since this is an exothermic reaction. Once confirmedhe higher contribution of the direct hydrogenation to the reactionield in the glycerol hydrogenolysis at 473 K, we can conclude thathe main role of the support in our catalytic systems, at the reactiononditions tested, is the copper dispersion, the control of coppergglomeration and its interaction with copper to obtain specificharacteristics.

. Conclusions

Copper supported on a mesoporous delaminated hectoriteynthesized by hydrothermal treatment for 1 h (HD1) was theest catalytic system for the hydrogenation of glycerol, espe-ially when a 40 wt.% of copper was introduced by impregnationCu/HD1imp3), obtaining, in this case, 93% of selectivity to 1,2-ropanediol for a 61% of conversion after 8 h of reaction at moderateonditions (40 bar, 473 K) without deactivation. This high activ-ty has been related to the higher area of the HD1 hectorite and

ith the control of copper agglomeration during catalyst prepara-ion and during reaction. This fact was confirmed by the increase,f copper dispersion, observed for the impregnated Cu catalystCu/HD1imp3) after reaction (from 8% to 11%). The agglomera-ion control in this catalyst is in agreement with the small copper

etal particles (12 nm) observed by TEM after reaction. The coppergglomeration could be associated with the support agglomerationroduced by the hydrophilic capacity of hectorites, which is relatedo the amount of interlamellar cations (C.E.C). The lower C.E.C of theectorite HD1 could justifier the best results of their derived sup-ort copper catalysts. The use of the delaminated hectorite HD2 orhe mesoporous silicate-bearing hectorite or the traditional hec-orites as supports of copper catalysts for the hydrogenolysis oflycerol, resulted in worse catalytic results. This has been relatedo the lower surface area and higher C.E.C. of these supports.

cknowledgments

The authors are grateful for the financial support of theinisterio de Educación y Ciencia of Spain and FEDER funds

CTQ2008-04433/PPQ).

eferences

[1] M. Pagliaro, R. Ciriminna, H. Kimura, M. Rossi, C.D. Pina, Angew. Chem. Int. Ed.46 (2007) 4434–4440.

[[

[

g Journal 179 (2012) 302– 311 311

[2] J. Van Haveren, E.L. Scott, J. Sanders, Biofuels Bioprod. Biorefin. 2 (2008) 41–57.[3] L. Ma, D. He, Catal. Today 149 (2010) 148–156.[4] Y. Kusunoki, T. Miyazawa, K. Kunimori, K. Tomishige, Catal. Commun. 6 (2005)

645–649.[5] A.J. Byrd, K.K. Pant, R.B. Gupta, Fuel 87 (2008) 2956–2960.[6] T. Miyazawa, S. Koso, K. Kunimori, K. Tomishigue, Appl. Catal. A 318 (2007)

244–251.[7] L. Ma, D. He, Z. Li, Catal. Commun. 9 (2008) 2489–2495.[8] T. Kurosaka, H. Maruyama, I. Naribayashi, Y. Sasaki, Catal. Commun. 9 (2008)

1360–1363.[9] E.P. Maris, R.J. Davis, J. Catal. 249 (2007) 328–337.10] E.P. Maris, W.C. Ketchie, M. Murayama, R.J. Davis, J. Catal. 251 (2007)

281–294.11] T. Miyazawa, Y. Kusunoki, K. Kunimori, K. Tomishigue, J. Catal. 240 (2006)

213–221.12] A. Alias, N. Hamzah, M.A. Yarmo, Adv. Mater. Res. 173 (2011) 49–54.13] N. Hamzah, A. Alias, W.Z. Samad, M. Kassim, M.A. Yarmo, Adv. Mater. Res. 173

(2011) 134–139.14] I. Gandarias, P.L. arias, J. Requies, M.B. Güemez, J.L.G. Fierro, Appl. Catal. B 97

(2010) 248–256.15] US Pat., 5,276,181 (1994).16] J. Chaminand, L. Djakovitch, P. Gallezot, P. Marion, C. Pinel, C. Rosier, Green

Chem. 6 (2004) 359–361.17] M. Balaraju, V. Rekha, P.S. Sal Prasad, R.B.N. Prasad, N. Lingaiah, Catal. Lett. 126

(2008) 119–124.18] A. Bienholz, F. Schwab, P. Claus, Green Chem. 12 (2010) 290–295.19] A. Bienholz, R. Blume, A. Knop-Gericke, F. Girgsdies, M. Behrens, P. Claus, J. Phys.

Chem. C 115 (2011) 999–1005.20] E.S. Vasiliadou, A.A. Lemonidou, Appl. Catal. A 396 (2011) 177–185.21] Z. Huang, F. Cui, H. Kang, J. Chen, C. Xia, Appl. Catal. A 366 (2009)

288–298.22] M.A. Dasari, P.-P. Kiatsimkul, W.R. Sutterlin, G.J. Suppes, Appl. Catal. A 281

(2005) 225–231.23] S. Sato, M. Akiyama, R. Takahashi, T. Hara, K. Inui, M. Yokota, Appl. Catal. A 347

(2008) 186–191.24] P. Claus, H. Hofmann, A. Bienholz, Appl. Catal. A 391 (2011) 153–157.25] K.A. Carrado, Appl. Clay Sci. 17 (2000) 1–23.26] S.P. Katdare, V. Ramaswamy, A.V. Ramaswamy, Micropor. Mesopor. Mater. 37

(2000) 329–336.27] F. González, C. Pesquera, I. Benito, E. Herrero, C. Poncio, S. Casuscelli, Appl. Catal.

A 188 (1999) 71–76.28] A. Gil, S.A. Korili, R. Trujillano, M.A. Vicente, Appl. Clay Sci. 53 (2011)

97–105.29] J. Guo Mei, S. Ming Yu, J. Cheng, Catal. Commun. 5 (2004) 437–440.30] P. Cool, E.F. Vansant, Micropor. Mater. 6 (1996) 27–36.31] J.T. Kloprogge, L.V. Duong, R.L. Frost, A review of the synthesis and characteri-

zation of pillared clays and related porous materials for cracking of vegetableoil to produce biofuels, Environ. Geol. (March) (2005) 3.

32] K. Torii, T. Iwasaki, Chem. Lett. (1988) 2045–2048.33] T. Iwasaki, Appl. Surf. Sci. 130–132 (1998) 845–850.34] K.A. Carrado, X. Langqiu, Micropor. Mesopor. Mater. 27 (1999) 87–94.35] A. Meister, N. Takano, T. Chuard, M. Graf, K. Bernauer, H. Stoeckli-evans, G.

Süss-Fink, Z. Anorg. Allg. Chem. 621 (1995) 117–121.36] W.T. Granquist, S.S. Pollack, Clays Clay Miner. 8 (1959) 150–169.37] I. Vicente, P. Salagre, Y. Cesteros, F. Guirado, F. Medina, J.E. Sueiras, Appl. Clay

Sci. 43 (2009) 103–107.38] A.A. Coelho, TOPAS v3.0. Bruker AXS, 2005.39] A.R. Stokes, A.J.C. Wilson, Proc. Camb. Philol. Soc. 38 (1942) 313–322.40] F. Bergaya, M. Vayer, Appl. Catal. Sci. 12 (1997) 275–280.41] A. Gervasini, S. Bennici, Appl. Catal. A 281 (2005) 199–205.42] O. Bergada, I. Vicente, P. Salagre, Y. Cesteros, F. Medina, J.E. Sueiras, Micropor.

Mesopor. Mater. 101 (2007) 363–373.43] J.T. Kloprogge, R.L. Frost, L. Hickey, Thermochim. Acta 345 (2000) 145–156.

44] T. Baird, A.G. Cairns-Smith, D.W. MacKenzie, Clay Miner. 18 (1973) 17–26.45] J.H. of Boer, The Structure and Properties of Porous Materials, Butterwoth,

London, 1958.46] M. Akiyama, S. Sato, R. Takahashi, K. Inui, M. Yokota, Appl. Catal. A 371 (2009)

60–66.