Ethanol conversion into 1,3-butadiene over a mixed Hf-Zn catalyst:

a study of the reaction pathway and catalyst deactivation

G. M. Cabello González1, R. Murciano2, A. L. Villanueva Perales1*, A. Martínez2, F. Vidal-

Barrero1, M. Campoy1
1
Departamento de Ingeniería Química y Ambiental, Escuela Técnica Superior de Ingeniería,
Universidad de Sevilla, Camino de los Descubrimientos, s/n. 41092 Sevilla, Spain.
2
Instituto de Tecnología Química, Universitat Politècnica de València - Consejo Superior de
Investigaciones Científicas (UPV-CSIC), Avda. de los Naranjos s/n, 46022 Valencia, Spain.

* Corresponding author. Email: angelluisvp@us.es

Departamento de Ingeniería Química y Ambiental, Escuela Técnica Superior de Ingeniería,

Universidad de Sevilla, Camino de los Descubrimientos, s/n. 41092 Sevilla, Spain.

1
Abstract
Some fundamental and practical aspects of a Hf-Zn catalyst, with a nominal composition of 3.0
wt% Hf and 9.3 wt% Zn and prepared as a physical mixture of Hf/SiO2 (85 wt%) and the zinc-
silicate hemimorphite (15 wt%), have been studied for the one-step conversion of ethanol to 1,3-
butadiene. The elucidation of the main reactions leading to 1,3-butadiene and by-products was
made by means of kinetic curves and catalytic tests where intermediates were individually fed. In
addition, the convenience of by-product separation from unreacted ethanol in an industrial
process was studied by performing experiments where ethanol was co-fed with intermediates.
The causes of catalyst deactivation and the impact on catalyst structure and performance of
regeneration by calcination were also investigated. According to our results, the pathway to 1,3-
butadiene over the Hf-Zn catalyst includes ethanol dehydrogenation, acetaldehyde aldol
condensation, crotonaldehyde reduction with ethanol, and crotyl alcohol dehydration. The
recycling of the by-products butanal, acetone and 1-butanol into the reactor should be avoided, as
the first two are converted to heavy compounds by aldol condensation reactions, while 1-butanol
dehydration leads to butenes, which are difficult to separate from 1,3-butadiene. The suppression
of diethyl ether formation from ethanol by recycling to extinction is possible. It has been found
that catalyst deactivation is mainly caused by the retention of oxygenated aromatic-type coke
species, preferentially formed on the dehydrogenating Zn2+ sites associated with the
hemimorphite component of the catalyst, and probably also by a loss in Zn2+ sites due to the
reduction to Zn0 during catalysis. This reduction induces an imbalance between Hf4+ and Zn2+
sites, which changes catalyst selectivity. Regeneration by calcination with air removes coke and
re-oxidizes a fraction of Zn0 back to Zn2+, but it does not fully re-establish the original Zn2+/Hf4+
balance.

Keywords: Ethanol; 1,3-butadiene; hafnium-zinc catalyst; reaction pathway; deactivation-

regeneration.

2
1. Introduction

1,3-Butadiene is an important building block for the manufacture of synthetic rubber,

resins, and elastomers [1] and it is mostly produced as a by-product of the naphtha or gas oil

steam cracking process. In recent years, due to the recent shale gas revolution, the increasing use

of ethane as steam cracking feedstock instead of heavier fractions has caused a decrease in the

production of 1,3-butadiene, even though its demand continues to grow at a rate of 1-2% per

year. Therefore, the short-term demand for 1,3-butadiene is not expected to be satisfied by the

current capacity of oil refineries [2,3]. In that context, the production of 1,3-butadiene from

renewable sources, such as bioethanol, is receiving increasing attention as an environmentally

friendly alternative to fulfil the demand gap while contributing to mitigate CO2 emissions.

Moreover, bioethanol is currently the largest biofuel on the market and its production is expected

to continue to grow in the future [4-6].

The process to catalytically convert ethanol into 1,3-butadiene was industrially born in

the 20th century around the time of the Second World War. The so-called Lebedev’s process in

the former USSR was the first industrial process to produce 1,3-butadiene directly from ethanol

using a MgO-SiO2 catalyst. Later, in the USA, the Union Carbide and Carbon Chemicals

Corporation started the manufacture of 1,3-butadiene from ethanol based on the Ostromislensky

two-step process. In the first step of this process, acetaldehyde is generated by ethanol

dehydrogenation, while the second step involves the conversion of the acetaldehyde/ethanol

mixture into 1,3-butadiene over a Ta2O5/SiO2 catalyst [3]. However, these technologies were

abandoned around 1960 due to a lack of competitiveness against the petrochemical route [7-12].

Therefore, the development of novel catalytic materials with improved activity and selectivity

3
becomes essential to improve the techno-economic viability of the ethanol-to-1,3-butadiene

process.

The present study focuses on the direct or one-step process. Modifications to the

traditional MgO/SiO2 one-step catalyst aiming to improve its performance have been proposed in

the literature [3,7], such as modifying the properties of the silica support and/or the preparation

method [13-17]. Addition of a transition metal and/or replacement of MgO with one or more

transition metal oxides has also been thoroughly studied [1,4,8-10,18-24]. On this regard, this

work is based on the study by De Baerdemaeker et al. [25] on silica-supported catalysts of

bimetallic mixtures of Hf(IV)-Zn(II) and trimetallic mixtures of Cu(II)-Hf(IV)-Zn(II) and Cu(II)-

Zr(IV)-Zn(II). For those systems, they investigated the effect of the preparation method and

metal source and found that the system Hf(IV)-Zn(II) provided the highest butadiene yield (60-

70%) when hemimorphite, a crystalline zinc silicate (Zn4Si2O7(OH)2·H2O), is used as the source

of Zn and then physically mixed with HfO2/SiO2. Therefore, as a promising catalyst for the

conversion of ethanol into 1,3-butadiene, the present work aims to study aspects of the Hf(IV)-

Zn(II) catalyst which are relevant for its industrial application, such as the reaction pathway to

elucidate the effect of recycling by-products to the reactor and the catalyst

deactivation/regeneration behaviour.

Although significant research has been carried out in the past to elucidate the reaction

pathway of the one-step conversion of ethanol to 1,3-butadiene, this is still a subject of

controversy due, most probably, to differences in the type of catalyst and/or reaction conditions

employed in these studies. Nonetheless, the most widely accepted reaction pathway considers

crotonaldehyde as a key intermediate [26-28] and comprises, as main steps, the ethanol

dehydrogenation to acetaldehyde, the aldol condensation of acetaldehyde to 3-hydroxybutanal,

4
the dehydration of 3-hydroxybutanal to crotonaldehyde, the selective reduction of

crotonaldehyde by ethanol to crotyl alcohol through the Meerwein-Ponndorf-Verley-Oppenauer

(MPVO) reaction, and the dehydration of crotyl alcohol to 1,3-butadiene. ZnO-Al2O3 and MgO-

SiO2 have been the catalytic systems most commonly studied to unveil this pathway [26-35].

Recently, for the first time, a detailed reaction mechanism based on this pathway was proposed

for a one-step Ag-ZrO2-SiO2 catalyst by using information derived from kinetic measurements

and tracing techniques [29]. The main criticism against the aldol condensation reaction pathway

is that crotonaldehyde is found in trace amounts while 3-hydroxybutanal is not detected at all in

the reaction products. In this regard, some researchers argue that 3-hydroxybutanal is not

observed because it is very reactive and is rapidly dehydrated to crotonaldehyde once it is

formed [9,27,30,31,36,37]. Besides the aldol condensation pathway, others routes for the one-

step conversion of ethanol to 1-3 butadiene have been proposed in the literature [15,38,39].

Regardless of the catalytic system used, side products are always formed along with 1,3-

butadiene. The quantity and nature of these side products depend not only on the reaction

conditions, but also on the catalyst studied and are not usually mentioned in the literature. This

issue is of paramount importance from the viewpoint of process design for selecting the

appropriate technologies in the separation train. In addition, the reactions where side products

are involved should be clarified in order to anticipate the effect of recycling some of these

compounds back into the reactor. Makshina et al. [3] summarized all the studies by Quattlebaum

and Gorin et al. regarding the main side products formed in the conversion of ethanol to 1,3-

butadiene over Ta2O5/SiO2 and MgO-SiO2 catalysts. Setting aside the catalytic system, the main

side products reported in the literature are ethene and diethyl ether, directly produced by ethanol

dehydration, and butenes. In addition, ethyl acetate, butanal, n-butanol, acetone, propene, CO2
5
and CO are also typically formed, albeit in lesser amounts. At high ethanol conversions, heavy

polar and non-polar compounds (grouped as C6+) are also usually produced

[7,8,15,20,23,25,32,40-45].

The deactivation of the catalyst during the ethanol-to-1,3-butadiene reaction is an

important aspect to be considered as well. It is well known from the literature that the

commercial catalysts previously used in the one- and two-step processes suffered from a fast

deactivation that is believed to be due to coke deposition and, thus, regeneration was required to

recover their initial activity [27,28]. In spite of its relevance for the catalyst/process design, very

few studies have reported on the deactivation and regeneration of new catalytic systems

developed for this reaction. For instance, the deactivation of silica-supported bi- and tri-metallic

catalysts was related to pore blockage and, accordingly, its extent was lowered by increasing the

pore size of the silica support [9]. Similarly, Chae et al. [40] observed a lower deactivation for

Ta2O5/SiO2 catalysts when using an ordered mesoporous silica (OMS) as support instead of a

commercial silica. Catalyst regeneration by calcination resulting in full recovery of the catalyst

activity has been reported by Chae et al. for Ta2O5/OMS after one regeneration cycle and by

Cheong et al. [46] for zirconium oxide supported on mesocellular siliceous foam after two

regeneration cycles. In other cases, however, the regeneration failed for unknown reasons. For

instance, although regeneration studies with Na-doped Zn-Zr mixed-metal oxides [23] revealed

the recovery of their initial activity, the deactivation rate was seen to double after regeneration.

On the other hand, Simoni da Ros et al. [44] reported that regeneration by calcination of a

K2O:ZrO2:ZnO/MgO-SiO2 catalyst did not fully restore the catalyst activity. Unfortunately, even

if these observations might point towards a change in the catalyst structure during the

6
regeneration process, no studies addressing this issue have been reported so far for one-step

catalytic systems.

Based on the above premises, in this work we performed a kinetic study to derive, for the

first time, a detailed reaction pathway leading to 1,3-butadiene as well as to intermediate and

side products over the Hf-Zn catalyst. This catalytic system was selected for its high yield and

selectivity to 1,3-butadiene [25]. Next, the deactivation and the regeneration behaviour of the

catalyst were investigated. The likely causes of deactivation and of the changes in catalytic

performance observed upon regeneration by air calcination are discussed based on the

characterization of the Hf-Zn catalyst at different stages (fresh, used, and regenerated) employing

a set of techniques, including TGA-DTG, GC-MS, 1H-to-13C CP-MAS NMR, XRD, STEM-

EDX, and XPS.

2. Experimental

2.1. Catalyst preparation

The Hf-Zn catalyst employed in this study, with a nominal composition of 3.0 wt% Hf

and 9.3 wt% Zn, was prepared according to a methodology similar to that described for the

sample denoted as Hf3.0Zn9.3 (Table 3, entry 6) in the work by De Baerdemaeker et al. [25].

Briefly, the zinc silicate hemimorphite (Zn4Si2O7(OH)2·H2O, abbreviated as HM) was first

synthesized by precipitation at a pH of ca. 7 using a commercial aqueous Na2SiO3 solution (26.8

wt% SiO2, 10.6 wt% Na2O, 62.6 wt% H2O, Sigma-Aldrich) and Zn(NO3)2·6H2O (Sigma-

Aldrich, 98% purity) as silicon and zinc sources, respectively. Concurrently, a Hf/SiO2 solid was

prepared by wet impregnation (liquid/solid ratio of 20 cm3/g) of a commercial silica (Cab-O-Sil

M5, Cabot Corporation) with an aqueous solution containing the required amount of HfCl4
7
precursor (Acros Organics, 99% purity) to achieve a nominal Hf loading of 3.5 wt%. After

elimination of the water solvent in a rotary evaporator, the solid was dried at 100 ºC for 12 h and

subsequently calcined in flowing air at 500 ºC for 4 h (heating rate of 1 ºC/min). The final

catalyst, hereafter denoted within the text as Hf-Zn, was then produced by physically mixing the

HM and Hf/SiO2 components in the appropriate proportions (85 wt% Hf/SiO2, 15 wt% HM) to

achieve the desired nominal metal composition (3.0 wt% Hf, 9.3 wt% Zn).

2.2. Characterization methods

The content of Hf and Zn in the calcined catalyst was determined by Inductively Coupled

Plasma-Optical Emission Spectrometry (ICP-OES) in a Varian 715-ES spectrometer prior to

digestion of the solid in an acid mixture of HNO3:HF:HCl (1:1:3 volume ratio). Crystalline

phases in the materials were identified by powder X-ray diffraction (XRD) in a Panalytical

Cubix Pro diffractometer equipped with a graphite monochromator operating at 40 kV and 45

mA using nickel-filtered CuKα radiation (λ= 0.1542 nm). The textural properties of the

individual HM and Hf/SiO2 solids and the Hf-Zn catalyst were measured by N2 physisorption at

liquid N2 temperature (-196 °C) in a Micromeritics ASAP-2420 equipment. Prior the analyses,

the samples were degassed at 300 °C under vacuum overnight. The element distribution in fresh,

spent, and regenerated catalyst samples was obtained by energy-dispersive X-ray spectroscopy

(EDX) using an Oxford INCA Energy 2000 system coupled to a 200 kV JEOL JEM-2100F

microscope operating in scanning-transmission mode (STEM). The composition and nature of

the Hf, Zn, and C species present on the catalyst surface were studied by X-ray photoelectron

spectroscopy (XPS). XP spectra were recorded in a VG-Escalab-210 spectrometer using Al Kα

8
radiation (1486.6 eV) and a pass energy of the analyzer of 30 eV. Charge correction of the XP

spectra was made based on the Si2p peak located at 103.5 eV.

The acidity of the calcined Hf-Zn catalyst and its components (HM, Hf/SiO2) was studied

by FTIR spectroscopy of adsorbed pyridine in a Nicolet 710 FTIR spectrometer using self-

supported wafers (10 mg/cm2). Complementary characterization by low-temperature FTIR of CO

adsorption was performed to investigate in more detail the different types and strength of the

Brønsted and Lewis acid sites present on the surface of the Hf-Zn catalyst and its components

(HM and Hf/SiO2). Since the extinction coefficients of CO adsorbed on Brønsted and Lewis acid

sites in these materials are unknown, preventing quantification of the different types of acid sites,

the results of the FTIR-CO measurements will be discussed on a qualitative basis.

Solid-state 1H-to-13C cross-polarization (CP) MAS NMR spectroscopy was performed to

study the nature of the carbon compounds deposited on the catalyst during the reaction. The
1
H/13C CP-MAS NMR spectrum was recorded at room temperature on a Bruker AV III HD 400

WB spectrometer at a spinning ratio of 5 kHz with a 90º pulse for 1H of 5 μs, a contact time of

3.5 ms, and a recycle delay of 5 s. The 13C spectrum was referred to adamantane.

Thermogravimetric analysis (TGA-DTG) of the spent catalyst was performed in a Metler

Toledo TGA/SDTA 851E analyzer under flowing air by linearly increasing the temperature from

50 to 800 ºC at a rate of 10 ºC/min. The total amount of carbon in the spent and regenerated

catalyst was determined by elemental analysis in a Fisons Inst. EA1108 CHNS-O analyzer. The

nature of the carbon species present in the spent sample was assessed by mass spectrometry

(MS) after dissolving the solid in HF and subsequent extraction of the soluble coke species with

CH2Cl2.

9
 Further details of the experimental conditions employed in some of the above

characterization techniques are given in Supporting Information.

2.3. Experimental facility and catalytic tests

Catalytic tests were carried out in a continuous flow fixed-bed reactor at 360 ºC and

atmospheric pressure. Prior to reaction, the catalyst was pressed, crushed, and sieved to collect

the 0.3-0.5 mm fraction. The internal diameter of the stainless steel tube reactor was 0.83 cm and

its length 25.8 cm. The bed was divided into three sections. The top section was a SiC bed of 8

cm, the middle one comprised a bed of catalyst diluted in the necessary amount of SiC to obtain

a bed length of 13 cm, and the bottom section was also a SiC bed. All sections were separated by

a glass wool plug. The reactor output line was electrically traced at around 300 ºC and insulated

to avoid product condensations before analysis by on-line gas chromatography (GC). The online

analyses were performed in an Agilent 7890A gas chromatograph equipped with two flame

ionization detectors (FID), one for C1-C6 hydrocarbons and the other one for oxygenated

compounds (alcohols, aldehydes, ketones...), and a thermal conductivity detector (TCD) for the

separation and quantification of other compounds such as CO, CO2, H2, N2, H2O, etc. The start-

up procedure is detailed in Supporting Information.

Kinetics curves were obtained by carrying out several tests on the Hf-Zn catalyst at 360

ºC, varying the space velocity (WHSV) from 1.12 to 50 h-1 by changing either the amount of

catalyst or the ethanol flow rate so as to cover a wide range of ethanol conversions. Additional

experiments were carried out where relevant intermediate reaction products were individually

fed or co-fed in different ratios with ethanol to better understand their contribution in the overall

10
reaction scheme. The conditions employed in the catalytic tests using single and co-fed reactants

are summarized in Tables 1 and 2, respectively, and further explained in Supporting Information.

Table 1. Single reactant tests. All tests were carried out at 360 ºC and atmospheric pressure.

WHSV Reactant Partial Pressure

Feedstock (h-1) (bar)

Ethanol (Panreac, 99.8%) 1.12-50 0.21

Acetaldehyde (Sigma-Aldrich, 99.5%) 1.12-9.8 0.21

Crotonaldehyde (Sigma-Aldrich, 99.5%) 6.1 0.21

Crotyl alcohol (Sigma-Aldrich, 96%) 1.12 0.04

1,3-Butadiene (Air Liquide, 20% in N2) 1.3-0.65 0.04

1-Butanol (Indukern, 98%) 9-30 0.21

Acetone (Merck, 99.8%) 1.12 0.21

Butanal (Sigma-Aldrich, 99%) 1.12 0.21

Table 2. Co-feeding tests. All tests were carried out at 360 ºC and atmospheric pressure.

Ethanol Partial Concentration of

WHSV Pressure intermediate in
ethanol
Feedstock (h-1) (bar) (wt%)
Ethanol + Diethyl Ether 1.12 0.21 2.5, 5, 10
(Sigma-Aldrich, 99.5%)
Ethanol + Butanal 1.12 0.21 2.5, 5, 10
Ethanol + Acetone 1.12 0.21 2.5, 5, 10
Ethanol + Crotonaldehyde 9.8 0.21 5, 10
Ethanol + 1,3-Butadiene 3.2 0.21 3.2

11
 Finally, a test lasting 100 h was done to assess the deactivation behaviour of the catalyst

at WHSV of 1.12 gEtOH/(gcat·h), T= 360 ºC, PEtOH of 0.21 bar, and Ptotal of 1 bar. During the test,

the catalyst was submitted to two in-situ regeneration cycles by calcination at 500 ºC in flowing

air (15 cm3/min). The first regeneration lasted 6 h and was initiated once the catalyst lost ca. 20%

of its initial conversion (corresponding to a time on stream (TOS) of 35 h). The second

regeneration cycle, starting at a TOS of ca. 60 h, was also performed by air calcination at 500 ºC

and lasted 8 h. After each regeneration cycle, the evolution with TOS of the catalytic

performance was followed in order to study the consequences of deactivation and regeneration

on the activity and product yield/selectivity of the Hf-Zn catalyst.

The reactant conversion and the selectivity and yield (both expressed on a carbon basis)

to a given product were defined in Eq. 1.1-1.3.

𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟� 𝑚𝑚𝑚𝑚𝑚𝑚 �
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 (%) = 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 • 100 (1.1)
𝐹𝐹𝐹𝐹𝐹𝐹 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 � �
𝑚𝑚𝑚𝑚𝑚𝑚

𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚
𝐶𝐶 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑡𝑡𝑡𝑡 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝� 𝑚𝑚𝑚𝑚𝑚𝑚 �
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 (%) = 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 • 100 (1.2)
𝐶𝐶 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑖𝑖𝑖𝑖 𝑎𝑎𝑎𝑎𝑎𝑎 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 � �
𝑚𝑚𝑚𝑚𝑚𝑚

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 (%)•𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆(%)
𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌 (%) =
100
(1.3)

For single reactant tests the weight hourly space velocity (WHSV) was defined in Eq. 1.4

as the mass flow rate of the reactant divided by the mass of catalyst in the reactor. Only the mass

flow of ethanol was considered for calculating WHSV in co-feeding tests. Carbon mass balance

errors for the experiments reported in this study were below 10%.

12
 𝑔𝑔
𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 𝑜𝑜𝑜𝑜 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓� �
𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊(ℎ−1 ) =
𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜(𝑔𝑔)
ℎ
(1.4)

3. Results and discussion

3.1. Physicochemical properties of the materials

The commercial silica employed as support in the preparation of the Hf/SiO2 component

presented the following textural parameters: BET area of 212 m2/g, total pore volume of 1.30

cm3/g, and mean pore diameter of 26 nm. The results of the N2 physisorption measurements for

the individual HM and Hf/SiO2 solids and the Hf-Zn catalyst are gathered in Table 3. As seen

there, the textural properties of the Hf/SiO2 solid showed minor deviations with respect to those

of the bare SiO2 support, as expected from the low Hf loading (3.5 wt%). In comparison, the

hemimorphite (HM) exhibited much lower BET surface area and total pore volume albeit higher

average pore size. On the other hand, the Hf-Zn catalyst, prepared by physically mixing Hf/SiO2

and HM, displayed values of the textural properties which are intermediate to those of the

individual components, although much closer to the Hf/SiO2 solid, as anticipated by considering

the mass proportion of each component in the physical mixture (85 wt% Hf/SiO2 + 15 wt% HM)

Table 3. Textural properties of the individual HM (hemimorphite) and Hf/SiO2 solids and the
Hf-Zn catalyst as measured by N2 physisorption.
BET area Total pore volume Mean pore diameter
Sample (m2/g) (cm3/g) (nm)
Hf/SiO2 206 1.26 26

HM 51 0.46 42

Hf-Zn 179 1.10 28

13
 The XRD patterns of the hemimorphite (HM), the impregnated and calcined Hf/SiO2

solid, and the final Hf-Zn catalyst are presented in Fig. S1 (Supporting Information). The

synthesized HM sample exhibits only the characteristic diffractions for hemimorphite (JCPDS 5-

0555), and confirms its successful preparation and high purity. On the other hand, the absence of

reflections related to Hf compounds in the Hf/SiO2 component evidences its high dispersion on

the amorphous SiO2 carrier. The Hf-Zn catalyst shows only the X-ray diffractions associated to

the hemimorphite phase, albeit with a much attenuated intensity with respect to pure

hemimorphite, as expected by its high dilution with Hf/SiO2 in the physical mixture. According

to repeated ICP-OES analyses, the Hf and Zn contents in the catalyst are, respectively, 2.5 (±0.1)

wt% and 10.0 (±1.0) wt%, which are close to the nominal values. A representative STEM image

and the element EDX mapping for Si and Zn in the fresh Hf-Zn sample are shown in Fig. S2.

The brightest regions in the STEM image clearly correspond to Zn species in the hemimorphite

(corroborated by EDX) which appears as a separate phase in the physically mixed catalyst, as

expected.

The acid properties of hemimorphite (HM), Hf/SiO2, and Hf-Zn catalyst were first

assessed by FTIR spectroscopy of adsorbed pyridine. The respective FTIR-pyridine spectra at a

desorption temperature of 150 °C are compared in Fig. 1. According to this technique, the Hf-Zn

catalyst, as well as its components, exhibit essentially Lewis-type acidity, as characterized by the

IR band at ca. 1450 cm-1. The somewhat lower frequency of this band in Hf/SiO2 and Hf-Zn

catalyst (1449 cm-1) compared to HM (1452 cm-1) might suggest that the Lewis acid sites

associated with Zn2+ species in hemimorphite are of a slightly higher strength than those related

to Hf4+ sites. Moreover, the presence of a shoulder at 1452 cm-1 besides the main peak at 1449

cm-1 in the spectrum of the Hf-Zn catalyst (spectrum c) indicates that both Zn2+ and Hf4+ species
14
contribute to its total Lewis acidity. The concentration of Lewis acid sites amounts to 65, 39, and

41 µmol/g for HM, Hf/SiO2, and Hf-Zn catalyst, respectively. These results indicate that the

Lewis acidity of the catalyst is mostly determined by that of its main component Hf/SiO2.

Moreover, the absence of a band at ca. 1545 cm-1 in all three samples indicates that the amount

of Brønsted acid sites, if present, should be very low and/or that their strength is too low to be

detected by pyridine at 150 ºC. Overall, these results are in good agreement with those reported

for the equivalent Hf-Zn catalyst in the work by De Baerdemaeker et al. [25].

Figure 1. Normalized FTIR-pyridine spectra recorded at a desorption temperature of 150 ºC for

HM (a), Hf/SiO2 (b), and Hf-Zn catalyst (c).

In order to better discern the presence of both Lewis and Brønsted acid sites in the Hf-Zn

catalyst and its individual components, we also performed low temperature (-170 ºC) FTIR

spectroscopy of CO adsorption, where the C-O stretching frequency of the CO probe molecule is

15
highly sensitive to the nature and strength of the surface acid sites [47]. The FTIR-CO spectra at

increasing CO doses are presented in Fig. S3, where the presence of two main bands at 2190-

2188 cm-1 and 2159-2152 cm-1 assigned to CO adsorbed on surface Lewis (e.g. Hf4+ and Zn2+

species) and Brønsted (-OH groups) acid sites, respectively, is clearly noticed [47-49]. A more

detailed information on the nature and strength of the acid sites can be obtained by analysing the

FTIR-CO spectra at low CO doses (i.e. low surface coverages) where dipole-dipole interaction

effects are absent, enabling the resolution of several components. Thus, as shown in panel A of

Fig. 2, mainly the Lewis acid sites are titrated at 0.1 mbar of CO. The zinc silicate hemimorphite

(HM) shows a main band at 2195 cm-1 and a less intense one at 2209 cm-1 (spectrum a), while

only a band at 2195 cm-1 is seen for Hf/SiO2 (spectrum b). The presence of the IR band at a

higher frequency (2209 cm-1) in the hemimorphite indicates that this material contains some

Lewis acid sites of a higher strength than those present in Hf/SiO2, as also suggested from the

FTIR-pyridine results (Fig. 1). In turn, the Hf-Zn catalyst (spectrum c) exhibits a weak feature at

2195 cm-1 besides a more intense peak at 2190 cm-1. The lower frequency of the later band signs

for a lower overall strength of the Lewis acid sites in Hf-Zn catalyst relative to HM and Hf/SiO2.

On the other hand, the presence of OH groups (i.e. Brønsted acid sites) of different acid strength

becomes clearer at a slightly higher CO coverage of 0.3 mbar, as shown in panel B of Fig. 2.

Hence, the hemimorphite presents three bands at 2159, 2151, and 2141 cm-1 (spectrum a) while

Hf/SiO2 exhibits a main band at a higher frequency of 2167 cm-1 (spectrum b), evidencing that

the later sample contains stronger Brønsted acid sites. Although these four bands are also present

in Hf-Zn catalyst (spectrum c), the prevailing contribution of the peak at 2159 cm-1 suggests a

decrease in the average strength of the Brønsted acid sites in the catalyst compared to Hf/SiO2.

Therefore, even if both Lewis and Brønsted acid sites are unambiguously detected in the samples
16
by FTIR-CO, the results at low CO coverages clearly sign for a lower average strength of both

types of acid sites in the Hf-Zn catalyst compared to the individual components. This fact is

probably due to interactions arising during the physical mixing of the HM and Hf/SiO2

components at the catalyst preparation stage.

Figure 2. FTIR spectra of CO adsorption (-170 °C) recorded at a CO pressure of 0.1 mbar (panel
A) and 0.3 mbar (panel B) for HM (a), Hf/SiO2 (b), and Hf-Zn catalyst (c) The bands in the

17
Lewis acid sites region (2195-2188 cm-1) for Hf/SiO2 and HM in panel B are out of scale.

3.2. Study of the reaction pathway in the ethanol to 1,3-butadiene conversion

In order to study the reaction scheme of 1,3-butadiene formation from ethanol, kinetic

curves encompassing a wide range of ethanol conversions (∼ 20% - 90%) were obtained for the

main products, enabling them to be classified according to their primary/secondary and

stable/unstable nature [50]. The full data about ethanol conversions and product yields at 360 ºC,

0.21 bar of ethanol (1 bar total pressure), and variable WHSV are gathered in Table S1. We must

note that a direct comparison of the catalytic performance of our catalyst with that of the

equivalent catalyst reported in the original work by De Baerdemaeker et al. [25] is not

straightforward. This arises mainly from the much higher ethanol partial pressure employed in

our study (0.21 against ~0.025 bar) in order to reproduce more closely the operation of an

industrial reactor at atmospheric pressure. In general, lower butadiene selectivity is achieved in

the present work, as a consequence of a much larger formation of heavy products than reported

by De Baerdemaeker et al. [25] for the same reaction temperature and ethanol conversion. This

difference in performance is not critical as the purpose of the present work is to determine

qualitatively the effects of recycling by-products and catalyst regeneration by calcination.

3.2.1. Reaction path for 1,3-butadiene formation

The kinetic curves for 1,3-butadiene, acetaldehyde, and crotonaldehyde are given in Fig.

3. The shape of the curves is similar to those obtained by Sushkevich et al. [8] over

Ag/ZrO2/SiO2 and by Shylesh et al. [1] over Au/MgO-SiO2. As seen in Fig. 3, acetaldehyde is

the main side product of the ethanol-to-1,3-butadiene reaction. The acetaldehyde yield-

conversion curve resembles that of a primary unstable product, indicating that it is directly

18
generated by the dehydrogenation of ethanol on Lewis acid sites associated to Zn2+ species in the

hemimorphite [25]. In order to study in which reactions acetaldehyde is involved, a set of tests

was carried out feeding pure acetaldehyde (Table S2). As observed, acetaldehyde conversion

remained relatively low in all these tests, not exceeding 15% at the lowest WHSV of 1.12 h-1.

The main product obtained was crotonaldehyde, plausibly formed by the aldol condensation of

acetaldehyde on Lewis acid sites associated with Hf4+ species identified in this work by FTIR

spectroscopy of adsorbed pyridine (Fig. 1) and CO (Fig. 2). However, almost no traces of 3-

hydroxybutanal could be found among the products, probably as a consequence of its rapid

dehydration to crotonaldehyde. Carbon monoxide and methane were observed in a mole ratio

close to 1:1, except at low acetaldehyde conversion where the mole ratio was 1:2. This suggests

that they are mainly formed by the decomposition of acetaldehyde, although other minor sources

of methane (more relevant at low acetaldehyde conversions) cannot be discarded. Heavy

compounds (C6+) were produced in significant amounts, probably through the self- and cross-

condensation of acetaldehyde and crotonaldehyde.

40 1.6
35 1.4
Crotonaldehyde yield (%)

30 1.2
Yield (%)

25 1.0
20 0.8
15 0.6
10 0.4
5 0.2
0 0.0
0 20 40 60 80 100
Ethanol conversion (%)
Butadiene Acetaldehyde Crotonaldehyde

Figure 3. 1,3-butadiene, acetaldehyde, and crotonaldehyde yields as a function of ethanol

conversion.
19
 An experiment feeding pure crotonaldehyde was performed in order to study the

reactions in which this product is involved. Only qualitative results of this experiment regarding

the nature of the species formed will be commented as the large formation of unknown heavy

compounds impeded accurately determining the product yields. While the results of this

experiment revealed the predominant formation of heavy compounds, supporting their likely

formation from crotonaldehyde via self-condensation reactions, no traces of 1,3-butadiene were

found among the products. Conversely, an increase in 1,3-butadiene selectivity was observed in

experiments co-feeding crotonaldehyde with ethanol (Table S3). This observation supports the

pathway suggested in the previous literature where crotyl alcohol was formed through the

MPVO reduction of crotonaldehyde with ethanol and subsequently dehydrated to 1,3-butadiene

[29,32,36,37,41,51]. In fact, silica-anchored hafnium has proven to be an effective catalyst for

the MPV reduction of ketones [52] and alcohols [25] as it forms complexes that favor hydrogen

transfer reactions, so we expect MPVO reactions to be promoted by Hf species in the Hf-Zn

catalyst.

Almost no crotyl alcohol was detected during the test feeding only ethanol (Table S1),

probably due to its high reactivity. The formation of 1,3-butadiene by the dehydration of crotyl

alcohol and the rapid rate of this reaction were confirmed by individually feeding crotyl alcohol

(Table S4). The conversion of crotyl alcohol at 360 ºC and WHSV of 1.12 h-1was 99% with a

selectivity of 60% to 1,3-butadiene. Therefore, it can be concluded that, over the Hf-Zn catalyst,

both crotonaldehyde and crotyl alcohol are directly involved in the formation of 1,3-butadiene

from ethanol, which is in agreement with the most widely accepted reaction pathway (Fig. 1).

Indeed, the kinetic curve of 1,3-butadiene (Fig. 3) resembles that of a secondary stable product.
20
Its stable character was confirmed by the absence of reaction in an additional experiment

performed feeding only 1,3-butadiene (Table S5). Similarly, no signs for the occurrence of

reactions were evidenced when butadiene was co-fed with ethanol (Table S6).

3.2.2. Formation of by-products

Although ethene and diethyl ether are undesired primary products formed, in theory, by

the dehydration of ethanol catalyzed by Brønsted acid sites, there is some controversy in the

literature about whether ethene is directly generated by the dehydration of either ethanol or

diethyl ether or even from both [8,27,31,53,54]. According to the kinetic curves of diethyl ether

and ethene (Fig. 4), they both seem to be primary stable products, suggesting that ethene is not

produced from diethyl ether. As the kinetic curve of ethene rises exponentially, there must be

another reaction, besides ethanol dehydration, which leads to ethene that does not involve the

participation of diethyl ether. Furthermore, the Prins condensation route, whereby 1-3 butadiene

is formed by reaction of ethylene with acetaldehyde [38], is ruled out over this catalyst as ethene

and diethyl ether are not intermediate products of this reaction. This hypothesis was confirmed in

separate experiments co-feeding ethanol and diethyl ether (Table S7), where a nearly constant

selectivity (and yield) to 1,3-butadiene and ethene was observed when the concentration of

diethyl ether in the feed mixture was increased from 2.5 to 10 wt%. Furthermore, as diethyl ether

inlet concentration increases, there is a point where diethyl ether formation from ethanol is

suppressed, which could be expected due to the low chemical equilibrium constant of this

reaction at the reaction temperature [55].

21
 4.5
4.0
3.5
3.0

Yield (%)
2.5
2.0
1.5
1.0
0.5
0.0
0 10 20 30 40 50 60 70 80 90 100
Ethanol conversion (%)
Ethene Diethyl Ether

Figure 4. Ethene and diethyl ether yields vs ethanol conversion.

Among the main sub-products, C4 compounds (1-butanol, butanal, and butenes) represent

an important fraction. The sources of 1-butanol and butanal, however, remain to be clarified.

Proposed routes leading to 1-butanol include the hydrogenation of butanal, crotonaldehyde or

crotyl alcohol [8,29] and the direct condensation of ethanol [15]. Butanal, on the other hand, may

be produced through the isomerization of crotyl alcohol, reduction of crotonaldehyde, and the

dehydrogenation of 1-butanol [8,29,39]. The kinetic curve of 1-butanol (Fig. 5) clearly exhibits

the characteristic shape of a secondary unstable product. Although the curve of butanal also

resembles that of a secondary unstable product, this seems less clear than for 1-butanol, as a net

decrease in butanal yield above a certain conversion cannot be unambiguously concluded (Fig.

5). The fact that at low ethanol conversions (< 30%) 1-butanol is detected in the products, but not

butanal, may be an indication of the dehydrogenation of 1-butanol as a plausible route leading to

butanal. This assumption is compatible with the results obtained in an experiment feeding only

1-butanol (Table S8), where butanal was the main product formed, especially at low 1-butanol

conversions (butanal selectivity of 68% at 21% conversion), followed by butenes. Since butanal
22
is formed from 1-butanol, crotonaldehyde hydrogenation should play a minor role as a source of

butanal. Then, it seems that crotonaldehyde is mainly hydrogenated to crotyl alcohol by the

MPVO reaction, which then can be further hydrogenated to 1-butanol. In fact, when only crotyl

alcohol was fed (Table S4), the lumped selectivity to butanol and butenes, as a gross measure of

1-butanol formation, was significant. This indicates that over silica doped metal catalysts, and in

the absence of an external H2 source, 1-butanol likely comes from crotyl alcohol hydrogenation

by H fragments present on the catalyst surface originating from previous alcohol

dehydrogenations [8,29]. In this respect, Hayashi et al. [24] observed an increase in the

selectivity to acetaldehyde, 1-butanol, butanal, and butenes when the Zn content is increased in a

Zn/talc catalyst, suggesting that H species generated by ethanol dehydrogenation on Zn sites are

responsible for the hydrogenation of crotyl alcohol to 1-butanol, from which butanal and butenes

are subsequently formed. Finally, the total butene fraction comprising 1-butene, isobutene, and

cis- and trans-2-butene shows the characteristic kinetic curve of a secondary stable product (Fig.

5) which, as discussed before, is in line with a route involving the dehydration of 1-butanol to 1-

butene and its successive isomerization to cis- and trans-2-butene and isobutene [3].

23
 4.5
4.0
3.5
3.0

Yield (%)
2.5
2.0
1.5
1.0
0.5
0.0
0 10 20 30 40 50 60 70 80 90 100
Ethanol conversion (%)
Butanol Butenes Butanal

Figure 5. Yields of 1-butanol, butenes, and butanal vs ethanol conversion.

The kinetic curve of ethyl acetate (Fig. 6) indicates that it is a secondary unstable

product, probably formed from acetaldehyde by the Tischenko reaction. Indeed, ethyl acetate

was observed in the products of the experiment performed feeding only acetaldehyde, albeit in

relatively low selectivity (Table S2). Sushkevich et al. [8] have proposed that the further

transformation of ethyl acetate leads to acetic acid, which produces acetone through the release

of CO2 by decarboxylation. Finally, acetone can be reduced and dehydrated to propene. This

reaction sequence is consistent with the shape and position of the kinetic curves of ethyl acetate,

acetone, propene, and CO2, as depicted in Fig. 6. Acetic acid was observed in trace amounts as it

is rapidly transformed into acetone and CO2 is released. The formation of propene from acetone

was confirmed in an experiment feeding pure acetone (not shown), where isopropanol and heavy

compounds (C6+) were also found as main products, and in experiments co-feeding ethanol and

acetone (Table S9), where the yield of propene was seen to increase with the acetone content in

the feed. Besides acetone, the decarbonylation of crotonaldehyde is another plausible source of

propene, as inferred from the results of an experiment feeding only crotonaldehyde (not shown).
24
 2.0
1.8
1.6
1.4

Yield (%)
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0 10 20 30 40 50 60 70 80 90 100
Ethanol conversion (%)
Propene Ethyl acetate CO2 Acetone

Figure 6. Yields of ethyl acetate, acetone, propene, and CO2 vs ethanol conversion.

In addition, heavy compounds with more than six carbon atoms (C6+) were produced in

significant amounts when feeding pure ethanol (Table S1). The kinetic curve of C6+ compounds

unambiguously indicates their secondary and stable nature (Fig. 7). Some heavy products can be

formed by the self and cross condensation of aldehydes such as crotonaldehyde and butanal, as

can be observed in the experiments performed by feeding these compounds (Tables S3 and S10).

Indeed, feeding only butanal (not shown) lead to 2-ethyl-2-hexenal as the main product via self-

aldol condensation. Moreover, when butanal was co-fed with ethanol (Table S10), the selectivity

to heavy compounds increased significantly from ca. 13 to 20% when the concentration of

butanal in the feed mixture was raised from 2.5 to 10 wt%, as expected. Other minor by-products

like acetone have also proven to contribute to the formation of heavy compounds (Table S9). The

deposition of heavy compounds on the Hf-Zn catalyst during the course of the ethanol-to-1,3-

butadiene reaction may certainly contribute to its deactivation, as will be discussed in the next

section.

25
 25

20

15

Yield (%)
10

0
0 10 20 30 40 50 60 70 80 90 100
Ethanol conversion (%)
C6+

Figure 7. Yield of heavy compounds (C6+) vs ethanol conversion.

As a summary of the experimental observations discussed above, a detailed reaction

scheme accounting for the formation of 1,3-butadiene and the side products from ethanol over

the Hf-Zn catalyst is shown in Fig. 8.

26
Figure 8. Proposed reaction pathway for the formation of 1,3-butadiene and side products from
ethanol over the Hf-Zn catalyst.

3.3. Catalyst deactivation and regeneration

3.3.1. Deactivation-regeneration behaviour

In order to study the deactivation behaviour and regeneration of the Hf-Zn catalyst, a test

lasting in total 100 h with two intermediate in-situ regenerations was performed according to the

experimental procedure detailed in section 2.3. The evolution with the time-on-stream (TOS) of

the ethanol conversion and the yield and selectivity to 1,3-butadiene, acetaldehyde, and

ethene+diethyl ether is presented in Fig. 9. As observed, catalyst deactivation resulting in a

gradual loss of ethanol conversion with TOS occurred from the beginning of the reaction as well

as after regeneration. In turn, the decrease in conversion is accompanied by a progressive

27
increase in acetaldehyde selectivity and a slight decrease in 1,3-butadiene selectivity, which is

more evident in the period following the second regeneration cycle. Moreover, the combined

selectivity to dehydration products (ethene+diethyl ether) slightly increased after the

regenerations but remained nearly unaltered with TOS within each reaction period. Similar

trends in activity and selectivity during the ethanol-to-1,3-butadiene conversion were observed

by Baylon et al. [23] in a study reporting the deactivation and regeneration by calcination of a

Na-doped Zn-Zr-mixed oxide catalyst. According to these authors, the change in 1,3-butadiene

and acetaldehyde selectivity was related to the loss of active sites involved in the conversion of

acetaldehyde to 1,3-butadiene.

As seen in Fig. 9, the catalyst recovered most of its initial activity after the two

regenerations by calcination in flowing air at 500 ºC. It is worth noting that the activity recovery

was somewhat higher after the second regeneration lasting 8 h in contrast to the 6 h applied in

the first regeneration cycle. The almost complete recovery of catalyst activity upon calcination in

air points to the deposition of coke as a major source of deactivation of the Hf-Zn catalyst.

28
 1st Reg 2nd Reg

100

90

80

70

60

50
%

40

30

20

10

0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Time on stream (h)

Ethanol conversion Acetaldehyde selectivity

1,3-butadiene selectivity 1,3-butadiene yield
Diethyl ether + ethene selectivity Acetaldehyde yield

Figure 9. Effect of deactivation on ethanol conversion, main product selectivities and yields.
Operating conditions: WHSV= 1.12 gEtOH/(gcat·h), T= 360 ºC, PEtOH of 0.21 bar and P of 1bar.
Regenerations were carried out at 500 ºC with air flow of 15 cm3/min. First and second
regenerations lasted 6 and 8 hours, respectively.

3.3.2. Analysis of coke species

With the aim of determining the origin of the observed catalyst deactivation and changes

in selectivity, the Hf-Zn catalyst was characterized by different techniques in its fresh state

(HfZnfresh), after its use in the reaction during the first 35 h on stream (HfZnused), and after the

second regeneration step (HfZnregen). The total carbon contents, measured by the elemental

analysis, amounted to 2.56 and 0.17 wt% for the spent and regenerated samples, respectively.

This indicates that calcination in air at 500 ºC is effective in removing most of the carbon species

from the catalyst and provides support for coke deposition as a main source of deactivation. The

29
fact that still some carbonaceous material remained in the catalyst after regeneration could

explain the slightly lower initial conversion observed after the regeneration cycles with respect to

the fresh sample. In fact, the TGA-DTG analysis in air of the used sample (Fig. S4) indicated

that part of the carbon deposits could only be removed at temperatures well above 500 ºC.

The nature of the carbon species present in the used sample was investigated by 1H-to-13C

CP-MAS NMR spectroscopy and by the extraction of the coke with dichloromethane after the

dissolution of the solid in HF, and further analysis by GC-MS. The1H-to-13C CP-MAS NMR

spectrum (Fig. 10a) shows two intense, sharp bands at 16 and 59 ppm and three broad

components peaking at 76, 127, and 178 ppm. The sharp bands with chemical shifts of 16 and 59

ppm are attributed, respectively, to the methyl carbon (CH3-) and to the carbon linked to the -OH

group of ethanol [56], indicating that part of the carbon retained on the catalyst after the reaction

is associated with the ethanol that remained adsorbed on the catalyst. On the other hand, the

broad resonance with a maximum located at 127 ppm and the associated spinning side bands at

76 and 178 ppm are characteristic of aromatic carbons [57]. These results are compatible with

those reported by Cavani et al. [39], who detected signs of ethanol, aldehydes, and aromatic

compounds adsorbed on a MgO catalyst after the ethanol-to-1,3-butadiene reaction by DRIFTS

and mass spectrometry. On the other hand, the GC-MS analysis of the coke extracted with

dichloromethane confirmed the aromatic nature of the formed carbon species. The most probable

structures (> 80% probability) for four of the main compounds detected are shown in Fig. 10b.

Therefore, we postulate that the formation and adsorption of relatively bulky oxygenated

aromatic species are responsible, at least in part, for the deactivation observed for the Hf-Zn

catalyst in the one-step conversion of ethanol to 1,3-butadiene.

30
Figure 10. a) Solid-state 1H-to-13C CP-MAS NMR spectrum of spent Hf-Zn catalyst (HfZnused);
b) main carbonaceous species detected by GC-MS in the coke extracted from HfZnused: 1) 4-tert-
butylphenol, 2) diphenylketone, 3) 2-naphtaldehyde, and 4) dihexyl phthalate.

3.3.3. Structural and surface characterization of used catalyst

Structural changes in the crystalline hemimorphite (HM) phase in the Hf-Zn catalyst after

its use in the reaction and after regeneration were studied by X-ray diffraction (XRD). As shown

in Fig. S5a, the crystalline structure of HM was retained after the reaction and regeneration,

although a decrease in the intensity of the hemimorphite-related peaks was noticed for the

regenerated sample. This loss in crystallinity points to a partial amorphization of the HM phase

during the regenerations. In order to confirm this, the fresh Hf-Zn catalyst sample was submitted

to two consecutive calcinations in flowing air at 500 ºC for 6 and 8 h, respectively, mimicking

the conditions applied in the two in-situ regeneration treatments. As in the case of HfZnregen, a

decreased intensity of the HM peaks relative to those in the fresh sample was observed after the

two calcinations (Fig. S5b), supporting the partial damage to the crystalline structure of

hemimorphite during the in situ regeneration treatments. In the original work by De

Baerdemaeker et al. [25], it was demonstrated that using hemimorphite as source of Zn2+ species

31
is more effective in promoting the selective formation of 1,3-butadiene and in suppressing

unwanted side reactions on Hf-related acid sites compared to Zn2+ species generated from a Zn

nitrate precursor. Therefore, the loss of the HM phase during the regeneration may expectedly

induce an imbalance between Hf and Zn sites in the Hf-Zn catalyst that might account for, at

least in part, the changes in activity and selectivity observed during the post-regeneration

reaction periods (Fig. 9).

The nature of the surface Hf, Zn, and C species in the fresh, used, and regenerated

catalyst was studied by X-ray photoelectron spectroscopy (XPS). Fig. 11 shows the survey XP

spectra and the spectra in the Zn2p, Hf4f, and C1s regions. The survey spectra (Fig. S6a) exhibit

characteristic peaks at 103, 153, and 533 eV that are attributed to Si2p, Si2s and O1s,

respectively [58]. The signals corresponding to Zn2p (Zn2p1/2 and Zn2p3/2) are located in the

region of the spectrum comprised between 1023 and 1046 eV [25,58], while the small peak

around 18-19 eV is assigned to the Hf4f core level signal which, due to spin-orbit effects, splits

into two peaks at 17.6 and 19.3 eV corresponding, respectively, to the Hf4f7/2 (main peak) and

Hf4f5/2 components [59,60].

a) b)

32
Figure 11. XPS spectra for the Hf-Zn catalyst in its fresh state (HfZnfresh), after reaction
(HfZnused), and after regeneration (HfZnregen): a) Zn2p, and b) Hf4f regions.

The XPS spectrum in the Zn2p region (Fig. 11a) of the fresh catalyst exhibits a main

Zn2p3/2 band at a binding energy (BE) of ca. 1023 eV and a less intense Zn2p1/2 signal at 1046

eV which were attributed to the Zn2+ species in the crystalline hemimorphite structure.

Interestingly, the spectrum of the used sample reveals the contribution to the Zn2p3/2 peak of an

additional signal at lower BE (1021 eV) assigned to metallic Zn species (Zn0) [59]. The

contribution of the Zn0 species in HfZnused was estimated to be ca. 34% by deconvolution of the

Zn2p3/2 peak, indicating a reduction of a significant part of the Zn2+ species in the original

hemimorphite phase to Zn0 during the ethanol-to-1,3-butadiene reaction. The reduction of Zn2+ to

Zn0 would thus represent a loss in active sites and, thus, an additional source of deactivation, as

has also been reported for ZnO during its use as catalyst in dehydrogenation reactions [61,63].

On the other hand, the lower intensity of the peaks in the Zn2p spectrum of HfZnused in

comparison to the other samples (Fig. 11a) can be related to the deposition of carbon species on

the catalyst surface. It is also seen in Fig. 11a that the regeneration of the used catalyst by

calcination at 500 ºC caused the re-oxidation of some Zn0 species back to Zn2+, with a

concomitant decrease in the relative concentration of Zn0 from 34 to 11%. The reduction of Zn2+

is expected to result in the segregation of zinc from the HM structure as Zn0. Therefore, the re-

oxidation of Zn0 during regeneration is anticipated to result also in the formation of a segregated

ZnO phase, whose characteristic Zn2p3/2 peak would be expected at a BE of 1022 eV [59] and,

thus, hard to be differentiated from the peak of Zn2+ in HM (BE= 1023 eV). As discussed in the

work by De Baerdemaeker et al. [25] and supported by the acidity results concluded by our FTIR

33
study, Zn2+ cations in HM neutralized most of the Brønsted acid sites associated with Hf-OH

groups (generated by hydrolysis on Hf4+ species), reducing the formation of dehydration side

products (e.g. ethene) while promoting 1,3-butadiene. Considering this, it seems plausible that

the transformation of some Zn2+ species in the hemimorphite into Zn0/ZnO during catalysis and

regeneration would impair the balance between the different types of active sites (redox,

acid/base) in the Hf-Zn catalyst, thus worsening its catalytic performance. Therefore, the loss in

hemimorphite-related Zn2+ centres would make Hf4+ sites less effective for the aldol

condensation step while promoting ethanol dehydration, presumably on Hf-OH sites not

neutralized by Zn2+. This might account for the increase in selectivity to acetaldehyde and, to a

lesser extent, to ethene+diethyl ether in detriment of the 1,3-butadiene observed during the

deactivation test (Fig. 9) [61,62].

On the other hand, the XP spectrum of HfZnfresh in the Hf4f region shows a main peak of

Hf4f7/2 at ca. 17.6 eV (Fig. 11b) which is consistent with Hf4+species in HfO2 interacting with

silica [56-58]. Unlike Zn, the Hf4f XP spectra for HfZnused and HfZnregen were very similar to

that of HfZnfresh (Fig. 11b) and, therefore, it is concluded that the oxidation state and the nature

of the Hf species remain the same (i.e. HfO2) during the course of the catalytic reaction and after

regeneration.

The nature of the surface carbon species was studied by analyzing the XP spectra of the

samples in the C1s region (Fig. S6b). The fresh sample presents C1s bands related to

adventitious carbon, as commonly found in XPS analyses. The de-convoluted spectrum of the

spent catalyst (HfZnused) shows a main component at 284.5 eV that can be attributed to either

aliphatic or aromatic carbons [63-65]. An additional component at 285.9 eV, characteristic of

carbon linked to oxygen in C-O-H and C-O, is also evidenced, while the contribution at 287.7 eV
34
could be associated with C=O [67]. These features are essentially in agreement with the nature of

the CH2Cl2-soluble coke species determined by GC-MS (Fig. 10b).

Table 4 shows the surface composition of the catalyst in its different states as determined

by XPS. It can be seen that the carbon content on the catalyst surface almost doubles after

reaction, consistent with the deposition of carbonaceous compounds also inferred from the

previous characterizations. As a result, the surface concentration of Zn and Si in HfZnused

decreased by 50% and 40%, respectively, with respect to HfZnfresh, while the concentration of Hf

was much less affected (15% decrease). This suggests that the aromatic-type carbonaceous

species were preferentially formed and retained on the dehydrogenating Zn2+ species in the

hemimorphite component of the catalyst. After regeneration of the catalyst, the surface contents

of Zn, Si, and Hf increased, approaching the values of the fresh catalyst. Concomitantly, the

amount of carbon decreased to levels close to those in the fresh sample, supporting the removal

of most of the coke from the catalyst after regeneration by calcination in air at 500 ºC.

Table 4. Atomic surface composition and metal surface ratios determined by XPS.

Atomic surface concentration (%) Surface atomic ratios

Catalyst Zn Hf Si C Zn/Hf Zn/Si Hf/Si

HfZnfresh 0.50 0.20 32.0 24.3 2.6 0.015 0.006

HfZnused 0.25 0.17 19.3 40.0 1.4 0.013 0.009

HfZnregen 0.40 0.19 23.5 29.4 2.1 0.017 0.008

Finally, representative STEM images and element EDX maps (Si, Zn) for the used and

regenerated Hf-Zn samples are shown in Figs. S7a and S7b, respectively. No noticeable changes
35
in the dispersion of Zn species with respect to the fresh sample (HfZnfresh, Fig. S2) were inferred

from this technique after catalysis (HfZnused) and after regeneration (HfZnregen).

4. Conclusions

The one-step ethanol to 1,3-butadiene pathway was studied over a Hf-Zn catalyst

prepared as physical mixture of Hf/SiO2 and zinc silicate hemimorphite (HM). The mixed Hf-Zn

catalyst was seen to possess mainly Lewis-type acidity, as ascertained by FTIR spectroscopy of

adsorbed pyridine. Nonetheless, both Lewis and Brønsted acid sites of different strengths

associated with Zn and Hf species in the hemimorphite and Hf/SiO2 solids, respectively, were

detected by low-temperature FTIR of CO adsorption, a more surface-sensitive technique.

According to FTIR-CO, the hemimorphite phase contained stronger Lewis acid sites than

Hf/SiO2 while the later solid exhibited stronger Brønsted acid sites. Interestingly, the average

strength of both types of acid sites decreased upon physically mixing the two phases during the

preparation of the Hf-Zn catalyst. Kinetic curves were evaluated in a wide range of ethanol

conversions and allowed for elucidating the main reactions leading to 1,3-butadiene and by-

products. In addition, several intermediate compounds have been fed or co-fed with ethanol to

assess some of the intermediate steps. Our results support the generally accepted pathway to 1,3-

butadiene: ethanol dehydrogenation to acetaldehyde, acetaldehyde aldol-condensation to

crotonaldehyde, crotonaldehyde reduction with ethanol, and crotyl alcohol dehydration.

Regarding side products, ethene and diethyl ether seem to be primary stable products formed

from ethanol. The decarbonylation of aldehydes has been observed over the catalyst:

acetaldehyde to methane and crotonaldehyde to propene, with the concomitant release of carbon

monoxide. Ethyl acetate is likely produced via the Tishchenko reaction and then further
36
transformed into acetone and propene, which releases carbon dioxide. 1-Butanol seems to be

formed from crotyl alcohol hydrogenation with surface H fragments originating from previous

dehydrogenation reactions. 1-Butanol dehydrogenation and dehydration are the main sources of

butanal and 1-butene, respectively. 1-Butene can be isomerized into cis- and trans-2-butenes.

The source of heavy compounds appears to be the self- and cross-condensation of C4+ aldehydes,

like butanal and crotonaldehyde, and ketones, like acetone.

These results give some insight into which by-products should be removed from

unreacted ethanol before being recycled into the reactor. The recycling of butanal and acetone

should be avoided, as they are finally converted to heavy compounds by aldol condensation

reactions. Heavy product formation leads to the deactivation of the catalyst, the fouling of

downstream equipment and also makes it difficult to recover and recycle ethanol. Recycling 1-

butanol should also be avoided, as its dehydrogenation leads to butanal and then to heavy

products while 1-butanol dehydration leads to butenes, which are difficult to separate from 1-3

butadiene due to their similar boiling points, shape and polarity. Finally, the recycling of diethyl

ether can suppress unwanted diethyl ether formation from ethanol due to the low equilibrium

constant of that reaction at a high temperature. The economic viability of this measure will

depend on the cost of separating and recycling diethyl ether and the increase in 1,3-butadiene

selectivity due to the suppression of diethyl ether formation.

It has been found that catalyst deactivation in the course of the reaction is mostly caused

by the formation and retention of aromatic-type carbonaceous species preferentially on the

dehydrogenating Zn2+ sites of the hemimorphite phase of the catalyst and by the reduction of

some of these Zn2+ sites to Zn0. The loss in Zn2+ sites induces a misbalance between Hf4+ and

Zn2+ sites which also results in a change in catalyst selectivity. Regeneration by calcination with
37
air removes coke and re-oxidizes a fraction of Zn0 back to Zn2+, but it does not completely re-

establish the Zn2+/Hf4+ site balance as the re-oxidized Zn species are segregated from the original

hemimorphite phase and provide a less favorable interaction with Hf sites.

5. Acknowledgements

This work was carried out in the framework of the project “Biobutadiene production from

bioethanol” (CTQ2015-71427-R), which is funded by the Spanish Ministry of Economy,

Industry and Competitiveness (MINECO) through the European Regional Development Fund

(ERDF). The authors would like to thank Maria Jesús Plumed Freire (US) for her invaluable

collaboration, support and dedication in conducting the catalytic tests, and to Dr. Patricia

Concepción (ITQ, Valencia) for her assistance in the FTIR-CO study.

6. References

[1] S. Shylesh, A.A. Gokhale, C.D. Scown, D. Kim, C.R. Ho, A.T. Bell, ChemSusChem 9

(2016) 1462–1472.

[2] A.J. Nizamoff. On-purpose Butadiene. Report abstract. Chemsystem PERP Program.

Nexant. (2013).

[3] E. V. Makshina, M. Dusselier, W. Janssens, J. Degrève, P.A. Jacobs, B.F. Sels, Chem.

Soc. Rev. 43 (2014) 7917–7953.

[4] S. Da Ros, M.D. Jones, D. Mattia, J.C. Pinto, M. Schwaab, F.B. Noronha, S.A. Kondrat,

T.C. Clarke, S.H. Taylor, ChemCatChem 8 (2016) 2376–2386.

[5] R.A. Flach B., Lieberz A., USDA Foreign Agric. Serv. (2017) 1–44.
38
[6] REN21, Renewables 2017: Global Status Report, 2017.

[7] C. Angelici, B.M. Weckhuysen, P.C.A. Bruijnincx, ChemSusChem 6 (2013) 1595–1614.

[8] V.L. Sushkevich, I.I. Ivanova, V. V. Ordomsky, E. Taarning, ChemSusChem 7 (2014)

2527–2536.

[9] M.D. Jones, C.G. Keir, C. Di Iulio, R.A.M. Robertson, C. V. Williams, D.C. Apperley,

Catal. Sci. Technol. 1 (2011) 267.

[10] B.B. Corson, H.E. Jones, C.E. Welling, J.A. Hinckley, E.E. Stahly, Ind. Eng. Chem. 42

(1950) 359–373.

[11] I. Lebedev, Zh. Obs. Khim 3 (1933) 698–717.

[12] I. Ostromislensky, J. Russ. Phys. Chem. Soc. 47 (1915) 1494.

[13] C. Angelici, M.E.Z. Velthoen, B.M. Weckhuysen, P.C.A. Bruijnincx, ChemSusChem 7

(2014) 2505–2515.

[14] C. Angelici, M.E.Z. Velthoen, B.M. Weckhuysen, P.C.A. Bruijnincx, Catal. Sci. Technol.

5 (2015) 2869–2879.

[15] J.V. Ochoa, C. Bandinelli, O. Vozniuk, A. Chieregato, A. Malmusi, C. Recchi, F. Cavani,

Green Chem. 18 (2016) 1653–1663.

[16] S.H. Chung, C. Angelici, S.O.M. Hinterding, M. Weingarth, M. Baldus, K. Houben, B.M.

Weckhuysen, P.C.A. Bruijnincx, ACS Catal. 6 (2016) 4034–4045.

[17] X. Huang, Y. Men, J. Wang, W. An, Y. Wang, Catal. Sci. Technol. 7 (2017) 168–180.

[18] Y. Kitayama, M. Satoh, T. Kodama, Catal. Letters 36 (1996) 95–97.

[19] M. Lewandowski, G.S. Babu, M. Vezzoli, M.D. Jones, R.E. Owen, D. Mattia, P.

Plucinski, E. Mikolajska, A. Ochenduszko, D.C. Apperley, Catal. Commun. 49 (2014)

25–28.
39
[20] O. V. Larina, P.I. Kyriienko, S.O. Soloviev, Catal. Letters 145 (2015) 1162–1168.

[21] O. V. Larina, P.I. Kyriienko, V. V. Trachevskii, N. V. Vlasenko, S.O. Soloviev, Theor.

Exp. Chem. 51 (2016) 387–393.

[22] V.L. Dagle, M.D. Flake, T.L. Lemmon, J.S. Lopez, L. Kovarik, R.A. Dagle, Appl. Catal.

B Environ. 236 (2018) 576–587.

[23] R.A.L. Baylon, J. Sun, Y. Wang, Catal. Today 259 (2016) 446–452.

[24] Y. Hayashi, S. Akiyama, A. Miyaji, Y. Sekiguchi, Y. Sakamoto, A. Shiga, T. Koyama, K.

Motokura, T. Baba, Phys. Chem. Chem. Phys. 18 (2016) 25191–25209.

[25] T. De Baerdemaeker, M. Feyen, U. Müller, B. Yilmaz, F.S. Xiao, W. Zhang, T. Yokoi, X.

Bao, H. Gies, D.E. De Vos, ACS Catal. 5 (2015) 3393–3397.

[26] M.Y. Kagan, O.M. Lyubarskii, G. D. Podurovskaya, Izv. Akad. Nauk SSSR Ser Khim

(1947) 173–181.

[27] W.M. Quattlebaum, W.J. Toussaint, J.T. Dunn, J. Am. Chem. Soc. 69 (1947) 593–599.

[28] W.J. Toussaint, J.T. Dunn, D.R. Jachson, Ind. Eng. Chem. 39 (1947) 120–125.

[29] V.L. Sushkevich, I.I. Ivanova, S. Tolborg, E. Taarning, J. Catal. 316 (2014) 121–129.

[30] M. León, E. Díaz, S. Ordóñez, Catal. Today 164 (2011) 436–442.

[31] H. Niiyama, S. Morii, E. Echigoya, Bull. Chem. Soc. Jpn. 45 (1972) 655–659.

[32] M. Gao, Z. Liu, M. Zhang, L. Tong, Catal. Letters 144 (2014) 2071–2079.

[33] Z. Han, X. Li, M. Zhang, Z. Liu, M. Gao, RSC Adv. 5 (2015) 103982–103988.

[34] W.E. Taifan, T. Bučko, J. Baltrusaitis, J. Catal. 346 (2017) 78–91.

[35] W.E. Taifan, G.X. Yan, J. Baltrusaitis, Catal. Sci. Technol. 7 (2017) 4648–4668.

[36] S.K. Bhattacharyya, J. Catal. 7 (1967) 152–158.

[37] S.K. Bhattacharyya, B.N. Avasthi, Ind. Eng. Chem. Process Des. Dev. 2 (1963) 45–51.
40
[38] V. Gruver, A. Sun, J.J. Fripiat, Catal. Letters 34 (1995) 359–364.

[39] A. Chieregato, J.V. Ochoa, C. Bandinelli, G. Fornasari, F. Cavani, M. Mella,

ChemSusChem 8 (2015) 377–388.

[40] T.W. Kim, J.W. Kim, S.Y. Kim, H.J. Chae, J.R. Kim, S.Y. Jeong, C.U. Kim, Chem. Eng.

J. 278 (2015) 217–223.

[41] W. Janssens, E. V. Makshina, P. Vanelderen, F. De Clippel, K. Houthoofd, S. Kerkhofs,

J.A. Martens, P.A. Jacobs, B.F. Sels, ChemSusChem 8 (2015) 994–1008.

[42] M.D. Jones, Chem. Cent. J. 8 (2014) 1–5.

[43] P.I. Kyriienko, O. V. Larina, S.O. Soloviev, S.M. Orlyk, C. Calers, S. Dzwigaj, ACS

Sustain. Chem. Eng. 5 (2017) 2075–2083.

[44] S. Da Ros, M.D. Jones, D. Mattia, M. Schwaab, F.B. Noronha, J.C. Pinto, Appl. Catal. A

Gen. 530 (2017) 37–47.

[45] V.L. Sushkevich, I.I. Ivanova, E. Taarning, Green Chem. 17 (2015) 2552–2559.

[46] J.L. Cheong, Y. Shao, S.J.R. Tan, X. Li, Y. Zhang, S.S. Lee, ACS Sustain. Chem. Eng. 4

(2016) 4887–4894.

[47] K.I. Hadjiivanov, G.N. Vayssilov, Adv. Catal. 47 (2002) 307–511.

[48] C. Morterra, G. Cerrato, V. Bolis, B. Fubini, Spectrochim. Acta Part A Mol. Spectrosc. 49

(1993) 1269–1288.

[49] G. Ghiotti, F. Boccuzzi, R. Scala, J. Catal. 92 (1985) 79–97.

[50] B.W. Wojciechowski, A.-N. Ko, Int. J. Chem. Kinet. 15 (1983) 1249–1274.

[51] G. Natta, R. Rigamonti, Chim. Ind. 29 (1947) 195–200.

[52] F. Quignard, O. Graziani, A. Choplin, Appl. Catal. A Gen. 182 (1999) 29–40.

[53] M. Zhang, Y. Yu, Ind. Eng. Chem. Res. 52 (2013) 9505–9514.

41
[54] I. Takahara, M. Saito, M. Inaba, K. Murata, Catal. Letters 105 (2005) 249–252.

[55] J.E. Connett, J. Chem. Thermodyn. 4 (1972) 135–138.

[56] S. Radhakrishnan, P.J. Goossens, P.C.M.M. Magusin, S.P. Sree, C. Detavernier, E.

Breynaert, C. Martineau, F. Taulelle, J.A. Martens, J. Am. Chem. Soc. 138 (2016) 2802–

2808.

[57] M.A. Callejas, M.T. Martínez, T. Blasco, E. Sastre, Appl. Catal. A Gen. 218 (2001) 181–

188.

[58] M.P. Pachamuthu, K. Shanthi, R. Luque, A. Ramanathan, Green Chem. 15 (2013) 2158.

[59] S. Yokoyama, K. Tamura, T. Hatta, S. Nemoto, Y. Watanabe, H. Yamada, Clay Sci. 13

(2006) 75–80.

[60] V. Udayakumar, A. Pandurangan, J. Porous Mater. 24 (2017) 979–990.

[61] A. Ramadoss, K. Krishnamoorthy, S.J. Kim, Mater. Res. Bull. 47 (2012) 2680–2684.

[62] K.D. Jung, O.S. Joo, S.H. Han, S.J. Uhm, I.J. Chung, Catal. Letters 35 (1995) 303–311.

[63] G. Liu, L. Zeng, Z.J. Zhao, H. Tian, T. Wu, J. Gong, ACS Catal. 6 (2016) 2158–2162.

[64] B. Lin, J. Wang, Q. Huang, M. Ali, Y. Chi, J. Anal. Appl. Pyrolysis 128 (2017) 291–303.

[65] D. Zhang, R. Wang, L. Wang, X. Yang, J. Mol. Catal. A Chem. 366 (2013) 179–185.

[66] J.F. Moulder, W.F. Stickle, P.E. Sobol, K.D. Bomben, Surf. Interface Anal. (1992) 261.

42