Journal of Catalysis 299 (2013) 284–297

Contents lists available at SciVerse ScienceDirect

Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat

Growth mechanism of coke on HBEA zeolite during ethanol transformation

L. Pinard a,⇑, S. Hamieh a, C. Canaff a, F. Ferreira Madeira a, I. Batonneau-Gener a, S. Maury b, O. Delpoux b,
K. Ben Tayeb c, Y. Pouilloux a, H. Vezin c
a
IC2MP, UMR 7285 CNRS, Université de Poitiers, 4 Rue Michel Brunet, 86022 Poitiers Cedex, France
b
IFP Energies Nouvelles, Rond-point de l’échangeur de Solaize, BP3, 69390 Vernaison, France
c
LASIR, UMR 8516 CNRS, Université des Sciences et Technologies de Lille, Bâtiment C4, 59655 Villeneuve d’Ascq Cedex, France

a r t i c l e i n f o a b s t r a c t

Article history: HBEA (11) zeolite was deactivated rapidly by coking during the ethanol transformation into hydrocar-
Received 26 September 2012 bons, at 623 K and 30 bar. The nature of carbonaceous deposit was studied, after zeolite dissolution by
Revised 12 December 2012 hydrofluoric acid both by gas chromatography coupled with mass spectrometry and by matrix-assisted
Accepted 14 December 2012
laser desorption/ionization time of flight mass spectrometry (MALDI-TOF MS). The coke on the external
Available online 29 January 2013
zeolite surface was characterized directly on the spent catalyst by laser desorption/ionization time of
flight mass spectrometry (LDI-TOF MS). The coke species were identified and classified into 17 families
Keywords:
following their unsaturation number (4–23). The coke was polyaromatic, and it was very alkylated
Coke characterization
(MA)LDI-TOF MS
(methyl, ethyl, and propyl groups). It is compounded of alkylbenzenes, mainly hexamethylbenzene
EPR (HMB), and alkyl-pyrenes, located within the zeolite pore, which could migrate to the external zeolite
Ethanol transformation surface and grow into polyaromatic compounds constituted up to eight aromatic rings and up to 40 car-
BEA zeolite bon atoms. Pulsed electron paramagnetic resonance spectroscopy measurements and molecular simula-
tion by using Cerius2 software have shown that HMBs are distant of 0.6 nm inside the zeolite pores.
Ó 2012 Elsevier Inc. All rights reserved.

1. Introduction the pores, channel size, proximity of molecule to wall) allowing

recombination, condensation, cyclization reactions, etc. and conse-
The catalytic conversion of ethanol to higher hydrocarbons than quently to polyaromatic compounds as result.
ethylene (C3+) is a promising way to incorporate a high level of The chemical nature of coke components and precursors can be
renewable carbon to gasoline pool. Ethanol to hydrocarbons specified by spectroscopic characterization of coked zeolites with
(ETH) reaction can occur over different types of zeolites (BEA, various techniques: infrared (IR), Raman [7], UV–visible [8], nucle-
MFI, FAU, etc.) [1]. But, ethanol transformation includes side reac- ar magnetic resonance (NMR) [9], and electron paramagnetic reso-
tions leading to the formation of carbonaceous materials, desig- nance (EPR) [2–5,10]. This last technique allows to characterize
nated as ‘‘coke’’ [1–5]. The ‘‘coke’’ poisons active sites and/or and quantify the nature of radicals generated during coke forma-
blocks their access [6]. It is well demonstrated that these products tion. But, the only way to determine the chemical composition of
are formed and trapped in the zeolite pores, and their formation carbonaceous deposits, that is, the complete distribution of their
can be considered as a ‘‘ship in the bottle’’ process [6]. In other components, is to remove ‘‘coke’’ from the zeolite by dissolving it
words, the growth of these products depends on the shape and in a solution of hydrofluoric acid [11]. A large part of coke compo-
dimension of the zeolite pores. These coke molecules, trapped in nents can be solubilized in organic solvents and analyzed by
the zeolite micropores being relatively simple, could act as active classical techniques such as GC, HPLC, NMR, IR, MS, and especially
species in hydrocarbons transformation. Indeed, on medium pore GC–MS. Unfortunately, the characterization of heavy molecules
zeolite, HZSM-5, the formation of C3+ hydrocarbons was observed (>400 g mol1) is much more restricted. However, significant ad-
even when the catalyst was saturated with coke molecules, mainly vances have been recently achieved with the use of (MA)LDI-TOF
polyalkylbenzenes. Deep characterizations of coke showed the MS (matrix-assisted) and LDI-TOF MS (laser desorption/ionization
existence of radical species [2–5]. Madeira et al. assume that etha- time of flight mass spectroscopy) [12–14]. These techniques, which
nol transformation into hydrocarbons occurs on radical coke mol- differ only by the adding (or not) a matrix (typically aromatic acid)
ecules as active site [4,5]. But, in the case of wide pore zeolites to the sample, were recently applied in the characterization of
(HFAU, HBEA), spatial constraints are different (diffusion inside heavy carbonaceous compounds, either directly on spent catalyst
or on coke recovered from the catalyst. The primary advantage of
the laser desorption/ionization is the relatively low energy input,
⇑ Corresponding author. Fax: +33 (5) 49 45 37 74.
which minimizes the fragmentation. Thus, the measured mass
E-mail address: [email protected] (L. Pinard).

0021-9517/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2012.12.018
 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297 285

(a) 100

80

conversion / yield (%)

60 EtOH Conv.
C2H4 yield
DEE yield
40 C3+ hydroc. yield

20

0
0 4 8 12 16
TOS (h)

(b) (c) 80
TOS = 0.8h
80 TOS = 0.8h
TOS = 16h
TOS = 16h

60
60 Selectivity (%)
Selectivity (%)

40
40

20 20

0 0
C3-C4 C5-C11 C12+ P O N A

Fig. 1. Ethanol conversion over HBEA, at 623 K and 30 bar, DEE, ethylene and C3+ hydrocarbons yield (a); C3+ hydrocarbons distribution by number of carbon atoms (b) and
chemical family (PONA) (c). ‘‘Data from [1]’’.

Table 1
Pore volumes and acidity for fresh and coked HBEA zeolite after the catalytic test (TOS = 16 h, T = 623 K, N2/EtOH = 4, P = 30 bar). Data from [1].

%C (wt.%) Aciditya (lmol g1) Pore volume (cm3 g1)

Brønsted Lewis Total Micro Meso
Fresh 266 249 0.835 0.271 0.563
Coked 13.2 13(95%) 32 (87%) 0.150 (82%) 0.052 (81%) 0.098 (83%)
a
Number of acid sites able to retain pyridine at 423 K (xx%) = loss.

spectrum directly represents the molecular weight distribution of (3.3 L h1) at 773 K and a total pressure of 30 bar. The catalytic test
the analyzed mixture. was carried out in a continuous down flow fixed bed reactor under
This work seeks to deepen the coke composition and location a total pressure of 30 bar and at 623 K. The catalyst (0.3 g) was
formed over HBEA zeolite during ethanol transformation. For that, placed in the middle of the reactor (stainless steel, 40 cm long,
a variety of characterization techniques on the deactivated catalyst 1.3 cm internal diameter, and 1.7 cm external diameter). Ethanol
and on extracted coke is used. (96%, V/V) was fed into the reactor at 2 mL h1, corresponding to
a N2/EtOH molar ratio of 4 and a WHSV of 15 h1. Reaction prod-
ucts were analyzed by online gas chromatography using a VARIAN
2. Experimental 3800 gas chromatograph equipped with two detectors: a FID
detector (J&W PONA capillary column) and a TCD detector (5A
2.1. Ethanol transformation sieve + Porabond Q).

HBEA (Si/Al molar ratio = 11) is a commercial material from 2.2. Spent catalyst characterizations
Zeolyst International, UK. The sample was compacted, crushed,
and sieved to obtain 0.2–0.4 mm homogeneous particles. Prior to The amount of carbon in the spent catalyst was measured using
catalytic testing, it was activated in situ under nitrogen flow a CE instrument NA2100 PROTEIN elementary analyzer.
286 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297

of the desorption/ionization process, and the spectra were the sum

3740

2965
2930
0.1 (a.u) of 40,000 shots. The spectra presented in this paper were from raw

2874
data with baseline correction and no signal smoothing.
Calibration of mass scale was performed with a mixture of

1593
PEG300 and PEG600 recorded under the identical conditions on a

1360
3745
nearest spot. In order to improve the exactitude of mass, the cali-

1565
1455
bration was also done on the same spot by internal calibration.

1500
The LDI-TOF MS analyses were preformed on the same instru-
ment. The spent catalyst was directly spotted onto the target with-
out the matrix. The LDI experiments were conducted using the
3660
3605

same setting as MALDI measurements. It should be noted that

3780

the results of MALDI and LDI-TOF MS were not quantitative. These

techniques are still considered to be a highly unreliable procedure
for the quantification. But, in this work, they allowed both to detect
all the coke components and to analyze directly the coke located
3500 3000 2500 2000 1500 on the external surface.
Pulsed EPR spectroscopy experiments were carried out with an
Wavenumber (cm-1)
X-band ELEXSYS E580E spectrometer. All the experiments were
Fig. 2. Infrared spectra of fresh (dotted line) and deactivated HBEA zeolite (black achieved at room temperature. Echo field sweep experiments were
line) after the catalytic test (TOS = 16 h, T = 623 K, N2/EtOH = 4, P = 30 bar). ‘‘Data performed using standard Hahn echo sequence p/2–s–p with dif-
from [1]’’.
ferent s values in order to separate the different radical species
by T2 relaxation time effect. s values were varying from 0.2 to 2 ls.
Matched HYSCORE experiments were performed according to
Fresh and spent HBEA zeolites were characterized by nitrogen the sequence 16(p/2)–s–72(HTA)–t1–16(p)–t2–72(HTA)–s–echo,
adsorption at 77 K. The fresh zeolite was outgassed overnight at where HTA is the high turning angle value of the pulse matching
623 K before measurement, and to avoid carbon elimination, the proton nuclear frequency (14.7 MHz), and t is set to 136 ns.
spent catalyst only at 363 K during 2 h then at 423 K during 1 h. The experiments were performed using the four-pulse DEER se-
The carbon content on the spent zeolite was the same before and quence (p/2)m1–s1–(p)m1–s–(p)m2–(s1 + s2)–s–(p)m1–s2–echo [16].
after pretreatment. The pore volumes were calculated from the The pump pulse (m2) length was set to 20 ns and applied at the
desorption branch of the N2 isotherms using the t-plot method. maximum of the left part of the spectrum with the optimization of
IR spectroscopy was used to characterize hydrocarbons species the amplitude (maximum of echo inversion). The observer pulses
on the spent catalyst by the individual bands that could be as- (m1) p/2 and p length were set, respectively, to 16 and 28 ns and
signed to certain functional groups. The FT-IR measurements were positioned at a 40 MHz higher frequency corresponding to the sec-
carried out in a Nicolet 750 Magna FTIR 550 spectrometer (resolu- ond radical species in the echo field sweep spectrum. Signal pro-
tion 2 cm1). Samples were pressed into thin wafers and activated cessing was achieved using the DeerAnalysis2011 software
in situ in the IR cell under secondary vacuum (106 mbar) at 623 K package under Matlab [17]. The signal was corrected by subtract-
and 423 K for, respectively, the fresh and spent zeolites. The acidity ing the unmodulated background echo decay by using a homoge-
of the samples was measured by infrared spectroscopy of adsorbed neous three-dimensional spin distribution. The Tikhonov
pyridine. The concentration of the Brønsted and the Lewis acid regularization was applied to the corrected dipolar evolution data
sites was calculated from the integrated area, after pyridine set to obtain the distance distributions [18].
adsorption at 423 K, of the PyH+ and PyL bands at 1545 and
1450 cm1, respectively. The experimental methods were de- 2.3. Molecular simulation
scribed in details in a previous paper [15].
For the analysis of carbonaceous deposits, the spent catalyst The modeling results published herewith were generated, using
was dissolved in 51% fluorhydric acid solution. HF was neutralized the program Cerius2 version 3.8 developed by Molecular Simula-
by an acid boric solution and sodium hydrogen carbonate [11, (see tions Inc. Monte Carlo simulations (MC) in the Grand Canonical
Supplementary information SI-1). The residue was extracted by ensemble (GCMC), followed by energy optimization, have been
CH2Cl2 and characterized by GC–MS and MALDI-TOF MS. In order done to model the adsorption of aromatic molecules in the BEA
to eliminate polar compounds, which are no volatile, the extract zeolite. GCMC simulations have been done at 623 K. The simulation
was cleaned up on SPE cartridge (Silica). GC–MS analyses were per- box consisted in 8 (2  2  2) unit cells of BEA zeolite. The atomic
formed on Thermoelectron DSQ equipped with DB5 ms column. positions for the BEA zeolite were modeled from available X-ray
The temperature was first increased to 313 K for 2 min and then diffraction determination of the BEA polytype A structure [19].
to 573 K (heating rate, 6 K min1). The split/splitless injector was The zeolite lattice was taken rigid during the MC simulation to re-
set at 553 K and operated in split mode (20 mL min1). The quad- duce the computational time. The partial atomic charges of the
rupole mass spectrometer was performed with electron impact zeolite atoms were taken from the literature [20]. The partial
ionization (70 eV, source temperature 523 K) operated under full atomic charges of the aromatic molecules were calculated using
scan monitoring mode. the method developed by Rappé and Goddard III [21]. Their geo-
For MALDI-TOF MS, 10 mg of the extract was suspended in metrical parameters have been optimized using the Dreiding force
500 lL of tetrahydrofuran (THF) and was mixed with Dithranol field [22]. The electrostatic interactions were modeled by a Cou-
as MALDI matrix (5 g L1 in THF). After sonication, about 0.5 lL lombic potential. The intermolecular non-bonded interactions, as
of this mixture was hand spotted onto the stainless steel target well as the non-bonded interactions between guest molecules
and dried at room temperature. MS was performed on a Brüker and zeolite which were dominated by dispersive interactions with
Autoflex Speed mass spectrometer in a reflectron positive mode O framework atoms [23], were described with a Lennard–Jones (LJ)
P r r
where ions were generated by a 337 nm wavelength nitrogen laser. potential: V LJ ðrij Þ ¼ ij 4eij ½ð r ij Þ12  ð r ij Þ6 . The short-range L–J
ij ij
Analyses were achieved using pulsed ion extraction (delay time interactions with Si framework atoms were ignored since they
130 ns). The laser power was adjusted slightly above the threshold were well shielded by the O framework atoms. The LJ potentials
 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297 287

(a) (b) polyaromatic compounds

5 10 15 20 25 30 35 40 45 50 55
Time (min)

(a)

8.5 9.0 9.5 10.0 10.5 11.0 11.5 12.0

Time (min)

(b)

12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5 16.0 16.5
Time (min)

Fig. 3. GC–MS total ion chromatogram of the CH2Cl2 soluble coke obtained from spent HBEA sample ((a) and (b) zooms).
288 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297

were taken from Dreiding force field [22] and literature works aromatic (A). Fig. 1 b and c show the evolution of the fraction selec-
[23,24]. The cross interaction potential parameters were obtained tivity for TOS of 0.8 h and 16 h. Initially, the main fraction was the
through the Lorentz–Berthelot relations. A cutoff radius of 0.8 nm C5AC11 (50%) and after 16 h of reaction, C3AC4 became major. So,
was applied to the Lennard–Jones interactions, and the long-range the chain length is associated with the catalyst deactivation state.
electrostatic interactions were calculated using the Ewald summa- On the catalyst (TOS = 0.8 h), there was an important formation
tion technique. of paraffins and aromatics with a molar ratio of 5. This could be ex-
plained by taking account the formation of more than one aromatic
ring in the hydrogen transfer reactions, for instance, the formation
3. Results and discussion of dimethylnaphthalene using 6 olefins for the hydrogen transfer.
On the coked zeolite (TOS = 16 h), nearly, no aromatics and napht-
3.1. Catalytic ethanol transformation enes were found, the products being almost exclusively butanes
and butenes and more particularly isobutane.
The deactivation of the HBEA zeolite was examined in ethanol
transformation at 623 K under a total pressure of 30 bar. Fig. 1a 3.2. Fresh and spent catalysts characterizations
presents the ethanol conversion and yields into products as a func-
tion of time-on-stream (TOS). Ethanol was converted into water, HBEA zeolite was characterized by FTIR spectroscopy. Four dis-
diethylether (DEE), ethylene which are dehydration products, and tinct OH bands are present in the IR spectrum of the zeolite (Fig. 2)
hydrocarbons ranging from 3 to 18 carbon atoms (C3+). Methane at 3605, 3660, 3745, and 3780 cm1; their assignments were al-
was produced in negligible amounts, and only small amounts of ready proposed by several authors [25–27]. The most intense (at
ethane were detected. No CO, CO2, and H2 were detected at any 3745 cm1) corresponds to the internal and external silanol groups
time. As proposed by Ferreira Madeira et al. [1], C3+ hydrocarbons [26,27]; the high intensity of this band is furthermore in agree-
formation results from the ethylene oligomerization followed by ment with the small crystallite size measured by TEM microscopy
cyclization and aromatization (by hydrogen transfer reactions (40 nm). The three other bands, which are very small, were as-
(HT) for paraffins and aromatics formation). The yields were pre- cribed, respectively, to bridging OH groups (3605 cm1), to hydro-
sented in terms of wt% content in the organic fraction. Initially xyl groups Al(OH) linked to extraframework Al species (EFAL)
(after 0.8 h of reaction), ethanol conversion was complete, and (3660 cm1), and to hydroxylated monomeric EFAL species and
then, it continuously decreased with TOS. As no dealumination oc- framework defects with Lewis acidity (3780 cm1) [28]. Pyridine
curred during reaction (see Supplementary information SI-2); the adsorption at 423 K causes the disappearance of the bridging OH
deactivation was only due to the coke formation, indeed, the car- band, but also the bands corresponding to hydroxylated EFAL spe-
bon content deposited over HBEA zeolite after 16 h was high, cies (3660 and 3780 cm1), which is due to their Lewis acidities. In
13.2 wt% (Table 1). addition, a small part of the external and internal silanol groups
No DEE was initially detected, whereas ethylene and C3+ hydro- (11%) interacts with pyridine. The concentration of the Brønsted
carbons yields were 80% and 20%, respectively. But, after 2 h of TOS and Lewis acid sites, estimated from the IR bands after desorption
DEE yield increased, concomitant with the decrease both of ethyl- at 423 K corresponding to pyridine adsorbed on these sites (1545
ene and of C3+ hydrocarbons yields. After 16 h of reaction, the yield and 1450 cm1), was 266 and 249 lmol g1, respectively (Table 1).
into C3+ hydrocarbons was very low (2%). The coke deposition (13.2 wt%), formed during EtOH transfor-
C3+ hydrocarbons were divided into three fractions according to mation (TOS = 16 h), results in a reduction in the Brønsted and Le-
the carbon number: C3AC4, C5AC11 and C12+, and also according to wis acid sites, 95% and 87%, respectively. The apparent decrease in
their chemical structure: paraffin (P), olefin (O), naphtene (N), and Lewis acidity is probably due to a partial blockage of micropores

Table 2
Molecular soluble coke composition determined by GC–MS after digestion of the coked HBEA zeolite.

Family Formula UNa Carbon number m/z Range Example of molecular structureb Location
A CnH2n6 (nP6) 4 9–17 120–232 Inside/external
Alkyl-benzene surface

B CnH2n8 (nP10) 5 12–15 112–142 R R Inside

Alkyl dihydronaphthalene

C CnH2n12 (nP10) 7 10–22 128–296 R Inside

Alkyl-naphthalene

D CnH2n18 (nP14) 10 17–20 220–262 R Inside

Alkyl-phenanthrene Alkyl- R
anthracene

E CnH2n20 (nP15) 11 20–27 246–386 Inside

4H-Cyclopentaphenanthrene R

F CnH2n22 (nP16) 12 18–31 244–412 R Inside/external

Alkyl-pyrene surface

a
Unsaturation number (UN).
b
R = alkyl groups (CH3, C2H5, C3H7, etc.).
 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297 289

F
328.246

Intens. [a.u.]
1000
F
342.261
F
314.230

F
356.280

F
300.217

G
F
G 370.297

G G
G
382.296

F
G
286.197 F 396.313
E E
E
E H H F
H
E H H F 410.330
274.197 G
H I I H G
I I 422.334
F J J
E J F 438.360
260.188 G H I 458.331
226.089 J
248.186 F I 486.363
510.362 526.387
546.364
572.397

250 300 350 400 450 500 550

Mass per charge (m/z)

Fig. 4. MALDI-TOF mass spectrum of the CH2Cl2 soluble coke obtained from spent HBEA sample digested by HF.

(81%) making impossible the diffusion of pyridine molecules to in- with methylene chloride during 5 h. The extracted coke molecules,
ner Lewis sites. Coke molecules cause both a decrease in the inten- which are located on the external surface and/or in the pore-mouth
sity of silanol groups (80%) and in the mesopore volume (83%), region (see Supplementary information, SI-1), are composed of
meaning that one part of coke molecules was located on the exter- hexamethylbenzene (HMB) (major components) and ethyl-substi-
nal surface of the crystallites. tuted benzene molecules (trace).
The FTIR spectrum of coked zeolite presents several new bands After dissolution of the coked zeolite in hydrofluoric solution,
not present on fresh zeolite, namely in the 2800–3000 cm1 (CAH only one part of coke molecules (no quantify) was soluble in meth-
stretching modes of paraffinic groups) and 1300–1700 cm1 re- ylene chloride. The composition of the soluble coke was deter-
gions (bands of CAH deformation and CAC stretching vibrations). mined by coupling gas chromatography to mass spectrometry
The bands at 2965 and 2874 cm1 are characteristic of the asym- (GC–MS). Fig. 3 shows the GC–MS total ion chromatogram. The
metric and symmetric vibrations of the CH3 group, and the band identified coke species were classified following the unsaturation
at 2930 cm1 corresponds to asymmetric band of CH2 group. The number (UN) into six families (A–F, Table 2):
bands at 1455 and 1380 cm1 are attributed to the asymmetric
and symmetric CH3 deformation vibrations. The bands at 1565  A, UN = 4, alkylbenzenes as HMB (evidenced by the mass frag-
and 1593 cm1 are assigned to the stretching mode of a C@C dou- ments 91, 105, 119 and 139).
ble bond and are characteristic of the alkylaromatic and polyaro-  B, UN = 5, dihydronaphthalene and indene (displayed by mass
matic compounds. It can be noted that no bands were observed, fragments 104, 118, and 145, 159, respectively).
in the 3000–3200 cm1 region corresponding to aromatic CH  C, UN = 7, naphthalene (indicated by the mass fragments 141,
stretching modes, this means that the aromatic coke molecules 155 and 169).
were very alkylated.  D, UN = 10, phenanthrene/anthracene (evidenced by the mass
fragments 191 and 205).
3.3. Soluble coke composition  E, UN = 11, cyclopentaphenanthrene (discernible by the mass
fragments 189 and 203).
Ten percentage of the total coke deposited on catalyst was  F, UN = 12, pyrene (indicated by the mass fragments 215 and
recovered through a direct soxhlet treatment of the spent zeolite 219).
290 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297

Table 3
Molecular composition of soluble coke determined by MALDI-TOF MS after digestion of the spent zeolite, and molecular composition of external coke measured directly on spent
zeolite by LDI-TOF MS.

Family Formula UNa Carbon number m/z Range Example of molecular structuree Location
Part 1
Eb,c CnH2n20 (nP15) 11 20–27 246–386 Inside
4H-Cyclopentaphenanthrene R

Fb,c CnH2n22 (nP16) 12 18–31 244–412 R Inside/external surface

Alkyl-pyrene

Gb,c CnH2n24 (nP19) 13 23–33 270–452 R Inside/external surface

Alkyl-dihydro-cyclopentapyrene R

Hb,c CnH2n26 (nP19) 14 24–32 296–450 R Inside/external surface

Alkyl-cyclopentapyrene R

Ic,d CnH2n28 (nP20) 15 26–32 308–462 Inside/external surface

Alkyl-benzo[a]pyrene R

Part 2
Jc,d CnH2n30 (nP21) 16 26–33 334–432 External surface
R

Kc,d CnH2n32 (nP22) 17 29–37 332–486 R External surface

Ld CnH2n34 (nP25) 18 29–34 358–512 R External surface

Md CnH2n36 (nP25) 19 32–34 384–538 R External surface

Nd CnH2n38 (nP26) 20 32–38 396–536 External surface

R

Od CnH2n40 (nP29) 21 32–38 408–534 External surface

R

Pd CnH2n42 (nP29) 22 33–38 406–532 External surface

R

Qd CnH2n44 (nP30) 23 34–38 446–530 External surface

R

a
Unsaturation number (UN).
b
Soluble coke determined by GC–MS analysis.
c
Soluble coke determined by MALDI-TOF MS analysis.
d
External coke determined by LDI-TOF MS analysis.
e
R = alkyl groups (CH3, C2H5, C3H7, etc.).

The number of alkyl fragments (CH2 group) on alkylbenzene peaks of A, B and C families (Fig. 3a and b) did not allow their rel-
and on polyaromatic species (A–F) was up to 15. This means that ative quantifications.
coke was extremely alkylated both by methyl and ethyl groups. The soluble coke was also analyzed by MALDI-TOF MS using
Moreover, propyl group was also evidenced by the mass fragments Dithranol as matrix. The mass spectrum, displayed in Fig. 4, gave
120 and 134 m/z. a clear Gaussian-curved distribution of masses with peaks in the
This is evidently that some soluble coke molecules were not mass range from 250 to 450 Da and a maximum at 328 Da. The
separated from the GC column. Moreover, the overlapping of the number of soluble coke species was rather limited. One prominent
 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297 291

Table 4
Number of alkyl fragments (CH2 group) on the different coke families determined by MALDI-TOF MS and by LDI-TOF MS.
18 18

16 E 16 F
14 14

12 12

Distribution (%)

Distribution (%)
10 10

8 8

6 6

4 4

2 2

0 0
4 5 6 7 8 9 10 11 12 13 14 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
-CH2-number -CH2-number
18 18

16 G 16 H
14 14

12 12
Distribution (%)

Distribution (%)
10 10

8 8

6 6

4 4

2 2

0 0
2 3 4 5 6 7 8 9 10 11 12 13 14 15 4 5 6 7 8 9 10 11 12 13 14 15
-CH2-number -CH2-number
18 25

16 I J
14 20

12
Distribution (%)

Distribution (%)
15
10

8
10
6

4 5
2

0 0
4 5 6 7 8 9 10 11 12 13 14 15 5 6 7 8 9 10 11 12 13 14 15
-CH2-number -CH2-number
25 18
K 16 L
20 14

12
Distribution (%)

Distribution (%)

15
10

8
10
6

5 4

0 0
4 5 6 7 8 9 10 11 12 13 14 15 3 4 5 6 7 8 9 10 11 12 13 14
-CH2-number -CH2-number
18 18

16 N 16 M
14 14

12 12
Distribution (%)

Distribution (%)

10 10

8 8

6 6

4 4

2 2

0 0
5 6 7 8 9 10 11 12 13 14 15 5 6 7 8 9 10 11 12 13 14 15
-CH2-number -CH2-number
18 20

16 O 18 P
14 16
14
12
Distribution (%)

Distribution (%)

12
10
10
8
8
6
6
4 4
2 2
0 0
3 4 5 6 7 8 9 10 11 12 3 4 5 6 7 8 9 10 11 12
-CH2-number -CH2-number
25
Q
20
Distribution (%)

15

10

0
5 6 7 8 9 10 11
-CH2-number

peak was always accompanied by smaller peaks at +2, 2, 4, and by CH2 group than preceding set. The maximum peak at 328 Da
6 Da. This sequence corresponds to a successive increase in the corresponds in agreement with GC–MS results to family F, pyrene
unsaturation number of coke molecules. Moreover, this pattern (CnH2n22, nP16), with 9 CH2 groups, and in consequence, 330 Da to
was repeated every 14 Da. This is a sign for each set being larger family E (4H cyclopentaphenanthrene, CnH2n20, nP15), 326, 324,
292 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297

Table 5
Location and distance between coke molecules inside the HBEA zeolite at 30 bar and 623 K obtained by molecular simulation.

322 Da to families G (CnH2n24, nP19), H (CnH2n26, nP19) and 3.4. Coke location
I (CnH2n28, nP20), respectively (Table 3). Family D (CnH2n22, nP16),
phenanthrenes–anthracenes, was too light to be detectible by The location of coke molecules inside the zeolite pores was con-
MALDI-TOF MS. From these results, it is possible to determine the firmed by molecular simulation using Cerius2 software. Table 5
molar distribution of alkyl fragment (CH2 group) for each family shows the preferential location in the zeolite pore of hexamethyl-
group number. As shown in Table 4, the number of CH2 group was benzene (HMB, family A), heptamethylnaphthalene (HMN, C) hex-
between 4 and 15 with a maximum around 8–9. amethylanthracene (HMA, D) and hexamethylpyrene (HMP, F).
 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297 293

Other molecules locations in the BEA pore as octamethylnaphtha-

lene were reported in the Supplementary information (SI-2). The
size of HMB (0.65  0.69 nm), in relaxed conformation, is close to
the pore aperture of HBEA zeolite (0.66  0.67 nm), which enables
its migration to the crystallite external surface. At 623 K and
30 bar, the calculated maximum number of HMB by unit cell was
around 3.25; the molecules located both in channels and channel
intersections were remote of 0.6 nm. The steric hindrance of these
molecules blocked them inside the zeolite pores. There were 2
HMN molecules by unit cell located in the channel and in the chan-
nel intersections. In the case of HMA and HMP, the molecules are
only located in the channel intersections, and they were distant
to 0.98 and 1.13 nm, respectively. No interaction occurs between
the two absorbed molecules. Moreover, the molecular simulation
Fig. 6. Proton Matched HYSCORE recorded at room temperature. The match pulse
showed that the HMP was among the biggest polyalkylated polyar-
for 1H nuclei detection was set to 72 ns.
omatic compound which may possibly exist in the BEA pores since
the octamethylpyrene was too big to be present in the BEA frame-
work. Consequently, the molecular simulation allows discriminat- As the echo field sweep experiments clearly show, two peaks far
ing the coke molecules located inside and outside the zeolite pore. from 48 MHz to 10 MHz higher than the highest coupling protons
The adsorption of neutral organic molecules, such as hexameth- measured using HYSCORE techniques. Consequently, the separa-
ylbenzene (HMB), can yield on acidic zeolite radical cation, HMB+; tion arises from two different radical species. Using DEER experi-
the protonated ring may be counterbalanced or stabilized by zeo- ments to refocused electron–electron dipolar interaction of these
litic negative charge [29]. To go further insight toward the struc- species, we can clearly see in Fig. 7 that electron modulation comes
ture and the localization/distribution of the radical inside the from dipolar interaction between radical species. Extraction of the
zeolite, pulsed EPR experiments were performed. The two pulses distance let us show principally three broad peaks at, respectively,
echo field sweep experiment for coked HBEA zeolite shows a very 1.6, 2.2, and 2.8 nm, indicating that radical can be separated of
intense Gaussian line of about 40 MHz linewidth centered at
g = 2.0067 indicating the presence of organic radical included in
the zeolite pores (Fig. 5). By varying the interpulse delay s in the observed
Hahn sequence, we can expect to separate different chemical spe- pump
cies by spin–spin T2 relaxation mechanism. Effectively by increas-
ing these s values until 600 ns, the modification of the line shape
was observed with the appearance of two peaks that come from
the presence of two different species that have small difference
of g value (Dg = 0.003). The species can been distinguished for Δν=48MHz
the optimum s of 600 ns, whereas, as we increase further s until
2 ls, only a drop of signal intensity can be observed. Proton match
HYSCORE experiments have been performed to solve the proton 3400 3420 3440 3460
hyperfine coupling of the radicals species. In the Fig. 6, we can ob- Magnetic Field [G]
served in the (, +) quadrant an antidiagonal cross peak that split
close to 30 MHz, that is, two times of 1H nuclear frequency 1.00
(14.5 MHz). This cross peak is centered at the hyperfine coupling
value A/2 providing a hyperfine coupling of 36 MHz for one proton.
Proton matched experiments allow in the quadrant (+, +), the 0.95
detection of 2m 1H. A strong 1H nuclear pattern in the (+, +) quad-
rant was detected with a maximum coupling for the protons of
27 MHz. This pattern is broad and not solved. Using double quanta 0.90
sequence, the magnetization due to some proton coupling is trans-
fer at 2m 1H (29 MHz). Three pairs of cross peaks were observed
0.85
with a respective hyperfine coupling of 8, 19, and 12.7 MHz.
0.0 0.2 0.4 0.6 0.8 1.0 1.2
t (µs)
200 ns

400 ns

600 ns

1.0 1.5 2.0 2.5 3.0 3.5

3400 3420 3440 3460 Distance (nm)
Magnetic Field [G]
Fig. 7. DEER experiments spectrum scheme excitation (top), dipolar evolution
Fig. 5. Echo field sweep experiments recorded at room temperature. Spectra were background subtracted (middle) and distance distribution using Tikhonov regular-
recorded for three s values, respectively, of 0.2, 0.4 and 0.6 ls. ization (bottom).
294 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297

about 0.6 nm each from others. This result is in agreement with the played in Fig. 8 was more complex than this of MALDI-TOF MS
distance obtained by molecular simulation with HMB. (Fig. 4). The number of coke species was higher, and the mass dis-
The quantification of radical species compared to the total tribution was ranged from 280 to 650 Da. The spectrum was com-
amount of polylalkylated aromatic compounds in the sample posed of an overlapping of three Gaussian-curved distributions of
should be possible if (1) all polyalkylated aromatic compounds mass, characterized by CH2 (14 Da) increment and maxima at
were radical and (2) also if these species were mono-radicalar spe- 314, 388, and 462 Da. This difference of 74 Da between the maxima
cies. Effectively, the EPR spectrum corresponds to all the radical could be assigned to building unit of C6H2 or the sum of C2 + C4H2
species present in the sample. As we have shown, it is possible (24 + 50), attached to parent polyaromatic structure, resulting in
by HYSCORE technique to detect a strong coupling proton due to an enlargement of 2 condensed rings. The first maximum observed
polyalkylated moieties, but such spectroscopy suffers to be non previously in Fig. 4 was attributed to family F (alkyl-pyrene,
quantitative due to electron and nuclear relaxation mechanisms. CnH2n22, nP16), and the second and the third were assigned to
In another hand, it is normally possible via the distance found by families K (alkyl-anthracene, CnH2n32, nP22) and P (CnH2n42,
DEER experiment and corroborated by the molecular simulation nP29), Table 3. Moreover, each prominent peak was accompanied
to determine the spin concentration for each peak correspond to with smaller peaks at +2, 2, 4, 6 Da. Therefore, the external
radical distance, but actually, it remains difficult to extract with coke was constituted of 12 families of polyaromatic compounds
a good accuracy. In the future, the advanced pulsed ENDOR (Elec- (F–Q, Table 3, part 2), up to 9 aromatic rings and 40 carbon atoms.
tron Nuclear double resonance) measurements will be able to give These polyaromatic molecules were more or less alkylated. The
the concentration of all radical varieties present in the sample. distribution of the number of alkyl fragments (CH2 group)
established for each family (Table 4) was between 4 and 15 with
3.5. Insoluble coke composition a maximum around 7 and 8.
The presence of pyrene-based structure on external surface
For the characterization of carbonaceous residues exclusively shows that some of molecules formed inside the channels of HBEA
deposited on the external surface, LDI-TOF MS was performed di- zeolite, migrate to the external surface, and continue to grow into
rectly on the spent zeolite [14]. The LDI-TOF mass spectrum dis- polyaromatic molecules covering external acid sites (Fig. 2).

F
314.212
Intens. [a.u.]

1000
F
328.227

F P
462.253
300.199
P
K
388.229
K
374.207

F K
342.242
402.242 P
G K 434.221 P
360.193 O 476.274

H K O
416.263
G J O
H
J L M
G L P
490.293
F
286.184 I M O
I Q 502.295
F Q
Q
G H QP
504.307
536.281
G
G 562.297

156.141 564.324
578.315

200.069

200 300 400 500 600 700

Mass per charge (m/z)

Fig. 8. LDI-TOF mass spectrum obtained directly from coked HBEA zeolite.
 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297 295

Alk. Sull. Auto-Alk.

2 EtOH B
A A
Isom

Sull. Alk. HT
B
Auto- C C C
Alk.

Isom HT Alk.

D D

Auto-
Alk.

Alk. HT

F F E

Fig. 9. Schematic illustrations buildup of polyaromatic deposits formed on HBEA zeolite pores.

Indeed, Guisnet and Magnoux [30] have shown that pyrene formed
inside the channels of ZSM-5 (5.1  5.7 M 5.3  5.6) migrates to C¼8 þ Rþ ! Cþ8 þ RH ! þ Hþ
the crystallite surface at high temperature (T > 650 K).

3.6. Proposal buildup of coke

The MALDI-TOF MS results of extracted coke molecules allowed þ 2 olefins ! þ 2 paraffins

to complete the identification of the coke molecules obtained by
GC–MS. Moreover, the combination of the molecular simulation
Multiple alkylations of mono-aromatic molecules could take
and LDI-TOF MS techniques discriminated the coke molecules lo-
place leading to polyalkylbenzene (family A), with a number of al-
cated inside and outside of the BEA pore. Moreover, as the UN of
kyl fragment (CH2 group) up to 15. Polyalkylbenzenes, frequently
each coke molecule increased by one (Tables 2 and 3), this means
referred to as hydrocarbon pool species, in particular hexamethyl-
that the growth of coke occurs only step by step and not by ring
benzene and heptamethylbenzene [32–33], could undergo molec-
condensation. Moreover, the combination of the molecular simula-
ular rearrangement via the ‘‘paring reaction’’ [34–36]. This
tion and LDI-TOF MS techniques discriminated the coke molecules
mechanism originally proposed by Sullivan et al. [37] suggests
located inside and outside of the BEA pore. From all these results, a
alkyl side chain growth by ring contraction/expansion. Indeed,
proposal buildup of polyaromatic carbonaceous deposits on HBEA
propyl chain has been detected, among the coke molecules,
is shown within the zeolite pores in Fig. 9 and in the external sur-
evidenced that the ‘‘paring reaction’’ occurs inside the zeolite.
face in Fig. 10.
The paring mechanism leads predominantly to propene and
Mono-aromatic molecules can be formed successively through
isobutene by dealkylation reaction with concomitant formation
ethanol dehydration-ethylene oligomerization–cyclization, fol-
of less substituted polymethylbenzenes (xylenes, toluene, etc.).
lowed by dehydrogenation through hydrogen transfer (HT) to a
The alkyl side chain growth on benzene ring may lead also to butyl
hydrogen acceptor (olefin) [31]:
chain, which could cyclize by auto-alkylation reaction into indene
and then isomerize into dihydronaphthalene (2HN), family B
C2 H5 OH ! C¼2 þ H2 O (Scheme 1) [38–40]. Hydrogen transfer occurs from 2HN to naph-
thalene, family C. Moreover, all aromatic compounds may undergo
4C¼2 ! C¼8 a series of alkylation step. The cyclization reaction is favored on
296 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297

Auto-Alk. HT

F G H
Isom

Auto-Alk. Alk.

J
I I
Rearr. HT
Alk. Auto-Alk.

K K L

HT

Alk. Rearr. HT

N N M

Auto-Alk.

HT Alk.

Auto-Alk. P
O P

Fig. 10. Schematic illustrations buildup of polyaromatic deposits formed on HBEA zeolite external surface.

Paring Cyclisation
reaction reaction

Family A Family B Family B

Scheme 1. Mechanism of Family B formation.

large pore zeolite (i.e., HBEA), leading to a fast deactivation by cok- The growth sequence of coke is then as follows: (Fig. 9):
ing. On the other hand, the medium pore size zeolite, HZSM-5, pro-
motes dealkylation reactions and consequently increases the  alkyl side chain growth via Sullivan’s mechanism;
catalyst stability.  five-membered ring formation by auto-alkylation reactions;
 L. Pinard et al. / Journal of Catalysis 299 (2013) 284–297 297

 isomerization of five-membered ring into six-membered ring [2] F. Ferreira Madeira, N.S. Gnep, P. Magnoux, H. Vezin, S. Maury, N. Cadran,
Chem. Eng. J. 161 (2010) 403.
followed by hydrogen transfer.
[3] F. Ferreira Madeira, K. Ben Tayeb, L. Pinard, H. Vezin, S. Maury, N. Cadran, Appl.
Catal. A 443–444 (2012) 171.
In each step, the unsaturation number of molecules increases of [4] F. Ferreira Madeira, H. Vezin, N.S. Gnep, P. Magnoux, S. Maury, N. Cadran, ACS
one. The formation of families A, B, C, D, E, and F occurs inside the Catal. 1 (2011) 417.
[5] K. Ben Tayeb, L. Pinard, N. Touati, H. Vezin, S. Maury, O. Delpoux, Catal.
zeolite pore leading to the catalyst deactivation by poisoning acid Commun. 27 (2012) 119.
site and by blocking the pore access (Table 1). These molecules [6] M. Guisnet, F. Ramôa Ribeiro, in: Imperial College (Eds.), Deactivation and
could migrate to the external surface to continue to grow and Regeneration of Zeolite Catalysts, 2011.
[7] Y.T. Chua, P.C. Stair, J. Catal. 213 (2003) 39.
formed insoluble coke (families G–Q) (Fig. 10). [8] H.G. Karge, M. Laniecki, M. Ziolek, G. Onyestyak, A. Kiss, P. Kleinschmit, M.
Piray, Stud. Surf. Sci. Catal. 49 (1989) 1327.
[9] J.-P. Lange, A. Gutsze, J. Allgeier, H.G. Karge, Appl. Catal. 45 (1988) 345.
4. Conclusion [10] F. Bauer, H.G. Karge, in: H.G. Karge, J. Weitkamp (Eds.), Molecular Sieves, vol. 5,
Springer-Verlag, Berlin, Heidelberg, 2006, p. 249.
Coking occurs during ethanol transformation into hydrocarbons [11] P. Magnoux, P. Roger, C. Canaff, V. Fouché, N.S. Gnep, M. Guisnet, Stud. Surf. Sci.
Catal. 34 (1987) 317.
leading to fast catalyst deactivation. Coke poisons acid sites and [12] A. Feller, J.-O. Barth, A. Guzman, I. Zuazo, J.A. Lercher, J. Catal. 220 (2003)
blocks their access by filling pore system. Detailed coke features, 192.
in term of molecular structure and location, have been obtained [13] H.S. Cerqueira, C. Sievers, G. Joly, P. Magnoux, J.A. Lercher, Ind. Eng. Chem. Res.
44 (2005) 2069.
by combining a large range of techniques: matrix-assisted laser [14] F. Bauer, H.W. Chen, E. Bliz, A. Freyer, V. Sauerland, S.B. Liu, J. Catal. 251 (2007)
desorption/ionization time of flight mass spectrometry (MALDI- 258.
TOF MS) and gas chromatography mass spectrometry (GC–MS) [15] D. Meloni, S. Laforge, D. Martin, M. Guisnet, E. Rombi, V. Solinas, Appl. Catal. A
215 (2001) 55.
on extracted soluble coke, laser desorption/ionization time of flight [16] G. Jeschke, Chem. Phys. Chem. 3 (2002) 927.
mass spectrometry (LDI-TOF MS) and electron paramagnetic reso- [17] G. Jeschke, V. Chechik, A. Godt, H. Zimmermann, J. Banham, C.R. Timmel, D.
nance (EPR) spectroscopy directly on the spent catalyst. The coke Hilger, H. Jung, Appl. Magn. Reson. 30 (2006) 473.
[18] A.N. Tikhonov, V.Y. Arsenin, Solutions of Ill-posed Problems, John Wiley &
was composed of 17 families discernible by their unsaturation
Sons, New York, 1977. pp. 1–272.
number in range of 4–23. All these compounds were very alkylated [19] J.B. Higgins, R.B. Lapierre, J.L. Schlenker, A.C. Rohrman, J.D. Wood, G.T. Kerr,
with a number of alkyl fragments (CH2 group) up to 15. Inside of W.J. Rohrbaugh, Zeolites 8 (1988) 446.
[20] J.C. White, A.C. Hess, J. Phys. Chem. 97 (1993) 8703.
the zeolite pores, the main compounds were alkylbenzenes, more
[21] A.K. Rappé, W.A. Goddard III, J. Phys. Chem. 95 (1991) 3358.
particularly hexamethylbenzene and alkyl-pyrenes. Some of them [22] S.L. Mayo, B.D. Olafson, W.A. Goddard III, J. Phys. Chem. 94 (1990) 8897.
give radical cations separated of 6 Å, and others could leave the [23] H. Klein, H. Fuess, G. Schrimpf, J. Phys. Chem. 100 (1996) 1101.
zeolite pore and migrate on the external surface. These molecules [24] M. Trzpit, M. Soulard, J. Patarin, N. Desbien, F. Cailliez, A. Boutin, I. Demachy,
A.H. Fuchs, Langmuir 23 (2007) 10131.
could act as precursor for polyaromatic molecules possessing up to [25] A. Simon-Masseron, J.P. Marques, J.M. Lopes, F. Ramôa Ribeiro, M. Guisnet,
8 aromatic rings. A schematic buildup of carbonaceous residues Appl. Catal. A 316 (2007) 75.
has been proposed. The growth of the coke molecules on outer sur- [26] A. Janin, M. Maache, J.-C. Lavalley, J.-F. Joly, F. Raatz, N. Szydlowski, Zeolites 11
(1991) 391.
face was not limited by steric constraints leading to a filling almost [27] M. Trombetta, B. Busca, L. Storaro, M. Lenarda, M. Casagrande, A. Zambon,
complete of the pore system. Phys. Chem. Phys. 2 (2000) 3529.
[28] D. Meloni, S. Laforge, D. Martin, M. Guisnet, E. Rombi, V. Solinas, Appl. Catal.
A216 (2001) 55.
Acknowledgments [29] S.J. Kim, H.-G. Jang, J.K. Lee, H.-K. Min, S.B. Hong, G. Seo, Chem. Commun. 47
(2011) 9498.
[30] M. Guisnet, P. Magnoux, Appl. Catal. 54 (1989) 1.
The support of this work by IFPEN and Agence National de la
[31] P. Magnoux, A. Rabeharitsara, H.S. Cerqueria, Appl. Catal. A 304 (2006)
Recherche (Program Hizecoke) is gratefully acknowledged. 142.
[32] M. Bjørgen, S. Akyalcin, U. Olsbye, S. Benard, S. Kolboe, S. Svelle, J. Catal. 275
(2010) 170.
Appendix A. Supplementary material [33] M. Bjørgen, U. Olsbye, S. Svelle, S. Kolboe, Catal. Lett. 93 (2004) 37.
[34] W. von E. Doering, M. Saunders, H.G. Boyton, H.W. Earhart, E.F. Wadley, W.R.
Edwards, G. Laber, Tetrahedron 4 (1958) 178.
Supplementary data associated with this article can be found, in
[35] S. Svelle, M. Bjørgen, S. Kolboe, M. Letzel, U. Olsbye, O. Sekiguchi, E. Uggerud,
the online version, at http://dx.doi.org/10.1016/j.jcat.2012.12.018. Catal. Lett. 109 (2006) 25.
[36] S. Kolboe, S. Svelle, B. Arstad, J. Phys. Chem. A 113 (2009) 917.
[37] R.F. Sullivan, C.J. Egan, G.E. Langlois, R.P. Sieg, J. Am. Chem. Soc. 83 (1961) 1156.
References [38] A. Corma, P.J. Miguel, A.V. Ochilles, G.S. Koermer, J. Catal. 159 (1992) 45.
[39] A. Corma, P.J. Miguel, A.V. Ochilles, G.S. Koermer, J. Catal. 145 (1994) 181.
[1] F. Ferreira Madeira, N.S. Gnep, P. Magnoux, S. Maury, N. Cadran, Appl. Catal. A [40] R.F. Sullivan, C.J. Evan, G.E. Langlois, J. Catal. 3 (1964) 183.
367 (2009) 39.