Mechanism and Kinetics of Ethanol Coupling to

Butanol over Hydroxyapatite

Christopher R. Ho1,2, Sankaranarayanapillai Shylesh1, Alexis T. Bell1,2*

1Department of Chemical and Biomolecular Engineering
University of California
Berkeley, CA 94720-1462

and
2Chemical
Sciences Division
Lawrence Berkeley National Laboratory
Berekley, CA 94720

Submitted to

ACS Catalysis

October 17, 2017

*To whom correspondences should be addressed: [email protected]

1
ABSTRACT

The mechanism and kinetics for ethanol coupling to n-butanol over hydroxyapatite

(HAP) were investigated at 573 – 613 K. In-situ titration experiments show that the

active sites for acetaldehyde and butanol formation are different. In combination with

FTIR studies, it was found that ethanol dehydrogenation is catalyzed by Ca-O sites,

whereas condensation of acetaldehyde is catalyzed by CaO/PO43- pairs. Measurements

of the reaction kinetics at various ethanol (3.5 – 9.4 kPa) and acetaldehyde (0.055 –

0.12 kPa) partial pressures reveal that direct condensation involving two ethanol

molecules does not play a significant role in butanol formation; instead, n-butanol is

formed via a Guerbet pathway. At a constant acetaldehyde pressure, enolate formation

is rate-limiting and ethanol inhibits acetaldehyde condensation rates by competitive

adsorption. A model of the reaction kinetics consistent with all experimental

observations is developed.

Keywords: hydroxyapatite, ethanol coupling, acetaldehyde, butanol, Guerbet reaction

2
INTRODUCTION

Dramatic growth of U.S. ethanol production from biomass in recent years has motivated

the search for efficient methods to upgrade ethanol to more valuable chemicals such as

ethylene, 1,3-butadiene, and n-butanol.1,2,3 Butanol is of particular interest because of

its use for the production of products such as paints, solvents and polymers.4 Moreover,

because of its high energy density (29.2 MJ/L) and low miscibility with water, it is a

superior biofuel to ethanol.5 The traditional approach to producing butanol has been via

the hydroformylation of propene followed by hydrogenation of the resulting butanal.6,7

More recently several investigators have shown that butanol can be produced via the

coupling of ethanol over a variety of catalysts8,9, such as ruthenium complexes10,

MgO11,12,13, Mg-Al mixed oxides14,15,16,17, Cu/CeO218, basic zeolites19, and

hydroxyapatite20,21,22,23,24,25. Of these materials, hydroxyapatite is particularly noteworthy

because it exhibits an unusually high activity and selectivity (>70%) to butanol.21,23

Hydroxyapatite (HAP; Ca5(PO4)3OH) is a calcium phosphate that possesses both acid

and base functionalities. Tsuchida et al. have shown using CO2 TPD that increasing the

Ca/P ratio of HAP leads to an increase in the density and strength of basic sites on the

catalyst surface.21 This trend also correlates well with an enhancement in the catalytic

activity and selectivity towards butanol. Attempts to identify the catalytically active sites

on HAP have relied largely on spectroscopic studies.26,27,28,29 Diallo-Garcia et al. have

observed the existence of basic OH-, O2- and PO43- groups through FTIR of adsorbed

CO2.26,27 Further studies by these authors using acetylene as a probe have revealed an

additional acid-base (POH-OH) pair.27 Several authors have deduced from DRIFT

3
spectra of ethanol adsorption at various temperatures the formation of weakly-bound

ethoxide species on POH groups, but the reactivity of these ethoxide species is

currently unknown.24,29

Identification of acid sites has relied on FTIR spectroscopy of adsorbed pyridine. IR

bands have been observed at around 1620, 1600 and 1440 cm-1.28,30,31,32 These bands

are indicative of hydrogen- and coordinately-bonded pyridine on Lewis acid sites. The

absence of a band near 1540 cm-1 indicates the lack of Brønsted acid sites on HAP.

FTIR spectra recorded at different temperatures show that pyridine desorbs completely

by 473 K, implying that the Lewis acid sites to which pyridine is adsorbed are weak.28,32

Hill et al. have obtained DRIFTS spectra on HAP with a variety of acid, base, and

bifunctional probe molecules and found that both acid and base sites are weak

compared to traditional metal oxide catalysts.32 These authors have also shown, using

steady-state isotopic transient kinetic analysis (SSITKA), that the active sites

responsible for acetaldehyde formation are much fewer in number compared to the

active sites responsible for butanol formation.33

While much is now known about the properties of hydroxyapatite, the composition and

structure of the active sites responsible for ethanol coupling to butanol have not been

clearly defined and neither has the mechanism for the reaction. Several groups have

proposed that the reaction occurs via a Guerbet-type mechanism.21,23,25 Ethanol is

assumed to dehydrogenate to acetaldehyde, which then undergoes aldol condensation

with another molecule of acetaldehyde to form crotonaldehyde. Finally, the

crotonaldehyde is doubly hydrogenated to form butanol. It is known from literature that

4
hydroxyapatite is an active catalyst for alcohol dehydrogenation34,35, hydrogen

transfer35, and aldol condensation36, which supports the Guerbet mechanism. However,

other pathways have also been suggested, such as direct coupling of ethanol to either

acetaldehyde or another ethanol molecule.37 Gines et al. have shown that a Guerbet

mechanism involving gas-phase acetaldehyde is not the only coupling pathway over K–

CuyMg5CeOx by observing unlabeled butanal in the presence of gaseous 13C-labeled

acetaldehyde.38 More recently, Scalbert et al. have used thermodynamic arguments to

show that in the Guerbet mechanism, gas phase hydrogen cannot be responsible for

hydrogenation of the crotonaldehyde intermediate to butanol.39,40 Ogo et al. have

confirmed these results experimentally by demonstrating that crotonaldehyde does not

undergo hydrogenation on HAP when gaseous H2 is added to the feed.22 This has led

Scalbert et al. to support the direct condensation pathways, which are

thermodynamically allowed.39,40 However, the Guerbet mechanism is still a plausible

pathway if crotonaldehyde is hydrogenated through a hydrogen transfer process

involving ethanol.

The aim of this work is to elucidate the site requirements for ethanol coupling to butanol

over HAP and the mechanism by which this process occurs. Our work suggests that the

Guerbet pathway is the predominant route of ethanol coupling with the rate of reaction

being limited by enolate formation. In-situ titration and spectroscopic experiments point

to the existence of two types of sites: basic Ca-O species that catalyze ethanol

dehydrogenation and CaO/PO43- pairs that catalyze acetaldehyde condensation. The

reaction is shown to be autocatalytic with respect to acetaldehyde, but is inhibited by

5
ethanol adsorption over condensation sites. The data obtained in this study is used to

develop a model for ethanol coupling to butanol over HAP.

EXPERIMENTAL METHODS

Catalyst Synthesis

Hydroxyapatite (HAP) was synthesized using a modification of the procedure reported

by Tsuchida et al.21 and Hanspal et al.24 Aqueous solutions of 0.25 M Ca(NO3)2•4H2O

and 0.55 M (NH4)2HPO4 were prepared and brought to a pH of 11 by addition of

ammonium hydroxide. The calcium solution was added dropwise to the phosphorus

solution at room temperature and stirred for 0.5 h before heating to 353 K for an

additional 3 h. The resulting slurry was filtered, washed with DI water, and dried at 393

K overnight in air. The catalyst was calcined in 100 mL/min of air at 873 K for 2 h before

use.

Calcium oxide was synthesized by a precipitation method using aqueous solutions of

calcium chloride and sodium hydroxide to form the corresponding hydroxide before

calcining at 773 K for 1 h.

Characterization Techniques

Powder X-ray diffraction (PXRD) patterns were taken with a Bruker D8 GADDS

diffractometer equipped with a Cu-Kα source (40kV, 40mA). BET surface areas were

calculated from nitrogen adsorption isotherms obtained using a Micromeritics Gemini VII

2390 surface area analyzer. The Ca/P ratio of HAP was determined by ICP-OES using

an Optima 7000 DV instrument. Infrared spectra (FTIR) were acquired using a Thermo

6
Scientific Nicolet 6700 spectrometer equipped with a liquid-nitrogen-cooled MCT

detector. Samples were prepared by pressing 0.04 g of catalyst into a 20 mm-diameter

pellet and placing it into a custom-built transmission cell with CaF2 windows. All pellets

for FTIR were pretreated in helium at 723 K overnight.

Measurements of Catalytic Activity

Measurements of reaction rates were carried out in a quartz-tube, packed-bed reactor

(10 mm inner diameter). Quartz wool was placed below the catalyst bed to hold the

catalyst in place. The reactor temperature was maintained using a tube furnace

equipped with a Watlow temperature controller and a K-type thermocouple. Prior to

reaction, the catalyst was treated in 50 mL min-1 of air (Praxair, zero grade) at 773 K for

1 h before cooling to the reaction temperature. This method was also used for catalyst

regeneration.

In a typical experiment, ethanol (Sigma-Aldrich, 99.5%) and 10 vol% acetaldehyde

(Sigma-Aldrich, 99.5%) in ethanol were introduced into a He (Praxair, 99.999%) stream

flowing at 30 cm3 min-1 using two syringe pumps (Cole Parmer, 74900 series). Propanol

(Sigma-Aldrich, 99.8%), pyridine (Fisher Scientific, ACS grade), and propanoic acid

(Sigma-Aldrich, 99.5%) were also fed in a similar manner. All experiments were carried

out at a pressure of 1 atm. Product streams were analyzed by gas chromatography

using an Agilent 6890A GC fitted with a HP-5 capillary column (30 m x 0.32 mm x 0.25

μm) and a flame ionization detector.

7
For studies of the reaction kinetics, the system was allowed to reach steady-state under

a constant ethanol partial pressure of 5.7 kPa before introducing acetaldehyde into the

feed stream. Since the addition of acetaldehyde causes slow catalytic deactivation,

butanol formation rates were calculated by extrapolating rates to the time of

acetaldehyde introduction. The conversion was calculated by summing all product

yields. Selectivity was defined as the number of ethanol molecules consumed to form a

product divided by the total amount of ethanol consumed. Yield was also defined on the

basis of carbon number. Formation rates were calculated on the basis of moles of

product, such that a 1:1 formation rate of acetaldehyde and butanol would ideally give a

1:2 yield because of the difference in carbon number. The carbon balance at an ethanol

conversion of 17.1% was 95 ± 5%, but was not be precisely determined at very low

ethanol conversions due to small fluctuations (~5%) in the feed rate of the syringe

pumps. Acetaldehyde formation rates were determined by summing apparent

acetaldehyde and butadiene formation rates since one molecule of acetaldehyde is

consumed per molecule of butadiene produced. Reaction orders, model parameter

values and their associated errors were calculated by a least squares regression

analysis.

RESULTS AND DISCUSSION

The X-ray diffraction pattern of HAP shown in Figure 1a is characteristic of

nanocrystalline apatite.41,42 The IR spectrum of the catalyst presented in Figure 1b

shows characteristics bands at 3572 (columnar OH-) and 2200 – 1900 cm-1 (P-O

overtone/combination modes) along with small bands at 1444 and 1414 cm-1 related to

8
carbonate impurities.26,43 The Ca/P ratio as measured by ICP-OES is 1.67, indicating a

stoichiometric composition. The catalyst BET surface area is 81 m2 g-1.

Figure 1. (a) XRD pattern and (b) IR spectrum of HAP. The main diffraction peaks of

hydroxyapatite are labeled.

Spacetime studies of ethanol conversion over HAP are shown in Figure 2. The yield of

acetaldehyde reaches a plateau beyond 1/WHSV = 16 min, indicating that acetaldehyde

is an intermediate for other products. The selectivity towards butanol reaches a

maximum of 77% at an ethanol conversion of 8%. Above this conversion, the butanol

selectivity decreases due to further coupling to form C6 alcohols. As shown in Table 1,

several other products are observed in small quantities, most notably crotonaldehyde,

butanal and crotyl alcohol, which are key intermediates in the Guerbet mechanism.

9
Figure 2. Effect of space velocity on ethanol conversion and product yields. Inset shows

a vertically stretched region of the same plot. Solid lines are meant to guide the eye. T =

603 K; PEtOH = 5.7 kPa; WHSV = weight hourly space velocity.

Table 1. Initial ethanol conversion and product selectivity over HAP. T = 603 K; catalyst

mass = 0.3 g; PEtOH = 5.7 kPa; total gas flow rate at STP = 30 cm3 min-1.

Catalyst HAP
2
Surface Area (m /g) 81
Conversion (%) 17.1
Selectivity (%)

Ethylene 1.4

Acetaldehyde 2.9

Butadiene 3.5

Butanol 63.2

Crotonaldehyde 0.1

Butanal 0.2

10
 Crotyl Alcohol 0.7

C6+ Alcohols 24.5

To gain more insight into the reaction pathway, several Guerbet intermediates

(acetaldehyde, crotonaldehyde, butanal, crotyl alcohol) were fed into the reactor, and

the conversions and product yields are shown in Table 2. In all cases, the carbon

balance was poor, presumably due to the formation and deposition of heavy aldol

oligomers on the catalyst surface. Quantitative conversion of acetaldehyde is achieved,

with crotonaldehyde being the major observed product, demonstrating that aldol

condensation proceeds rapidly under reaction conditions. When butanal is fed into the

reactor, only trace amounts of crotyl alcohol and butanol are formed, indicating that

intermolecular hydrogen transfer between two butanal molecules is slow. Similarly, a

crotyl alcohol feed yields only trace amounts of crotonaldehyde and butanol. Instead,

butadiene is formed in significant amounts, suggesting that crotyl alcohol preferentially

undergoes a unimolecular dehydration reaction in the absence of a high concentration

of a hydrogen donor such as ethanol.

11
 Table 2. Conversion and yield of key Guerbet intermediates over HAP. T = 603 K;

catalyst mass = 0.05 g; PC2/C4 molecules = 0.1 kPa; P1-Propanol = 4.4 kPa; total gas flow rate

at STP = 30 cm3 min-1. tr = trace (<1% yield)

Yield (%)
Conversion
Reactant(s)
(%) Crotyl C5 C7
Ethanol Butadiene Crotonaldehyde Butanal Butanol Otherb
Alcohol Alcohols Alcohols

Acetaldehyde >99 0 0 17 0 0 0 - - 2

Crotonaldehyde 95 tr 0 - 0 0 0 - - 58

Butanal 90 tr 2 tr - tr tr - - 37

Crotyl alcohol 99 tr 32 tr tr - tr - - tr

Acetaldehyde/
>99a 0 0 0 0 0 83 7 0 -
1-Propanol
Crotonaldehyde/
>99a tr 4 - tr 2 70 - 12 -
1-Propanol
Butanal/
>99a tr 8 tr - tr 61 - 23 -
1-Propanol
Crotyl alcohol/
>99a tr 8 tr tr - 68 - 15 -
1-Propanol
aConversions and yields based on the C2 or C4 reactant.
bC6 aldehydes for acetaldehyde feed; C8 aldehydes for crotonaldehyde, butanal and
crotyl alcohol feeds.

Condensation and hydrogen transfer rates were compared by studying the reaction of

acetaldehyde with 1-propanol over HAP using a method similar to that reported by Ogo

et al.22 1-propanol was chosen as the hydrogen donor because it is chemically similar to

ethanol but does not convert to C2 products, which allows tracking of acetaldehyde

consumption. The yield of ethanol (hydrogen transfer product) was 83% while the yield

of C5 alcohols (aldol condensation products) was 7%, showing that hydrogen transfer
12
rates are much faster than aldol condensation rates, consistent with previous

observations.22 When 1-propanol is co-fed with crotonaldehyde, butanal or crotyl

alcohol, butanol is the major product observed, showing that all three Guerbet

intermediates can be readily reduced by a hydrogen donor. The fast hydrogen transfer

rates also reduce the selectivity towards unwanted side reactions such as dehydration

(in the case of crotyl alcohol) or further aldol condensation (in the case of

crotonaldehyde and butanal). In the context of ethanol coupling to butanol, the results of

these co-feed studies suggest that acetaldehyde formed by ethanol dehydrogenation

easily undergoes aldol condensation to form crotonaldehyde. Crotonaldehyde is then

rapidly hydrogenated to butanal, crotyl alcohol, and ultimately butanol, by hydrogen

transfer with ethanol.

The apparent rates of acetaldehyde and butanol formation were measured as a function

of ethanol partial pressure, and are presented in Figure 3. The conversion ranged from

1% at high ethanol partial pressures to 5% at low ethanol partial pressures. The

observed reaction order for acetaldehyde formation decreases from 1.0 to 0.3 ± 0.1 as

ethanol partial pressure increases from 0.1 to 9.4 kPa. The rate of butanol formation is

first order at low ethanol partial pressures (<1 kPa) and zero order at high ethanol

partial pressures (9.4 kPa).

13
Figure 3. Rates of acetaldehyde and butanol formation as a function of ethanol partial

pressure. The solid line for acetaldehyde represents the model fit from Eqn. 1. The solid

line for butanol represents the model fit from integrating Eqn. 2 over the length of the

reactor. T = 603 K; HAP mass = 0.05 g; total gas flow rate at STP = 30 cm3 min-1.

To assess the role of ethanol and acetaldehyde in the reaction more accurately, both

reactants were co-fed into the reactor to maintain constant ethanol and acetaldehyde

partial pressures throughout the catalyst bed. Ethanol and acetaldehyde conversions

were kept below 10% and 15% respectively to achieve differential conditions. Figure 4

shows butanol formation rates over HAP for a range of ethanol (3.5 – 9.4 kPa) and

acetaldehyde (0.055 – 0.12 kPa) partial pressures. For all reaction conditions tested,

ethanol exhibits a negative order dependence (-0.4 to -0.7 ± 0.1) while acetaldehyde

exhibits a positive order dependence (1.0 ± 0.2).

14
Figure 4. Butanol formation rates as a function of (a) ethanol partial pressure at

constant acetaldehyde pressures and (b) acetaldehyde partial pressure at constant

ethanol pressures. Solid lines represent the fitted kinetic model shown in Eqn. 2. T =

603 K; HAP mass = 0.05 g; total gas flow rate at STP = 30 cm3 min-1. Abbreviations:

EtOH = ethanol; AA = acetaldehyde.

A direct coupling pathway involving two ethanol molecules would be expected to exhibit

a zero to positive order in ethanol partial pressure based on a Langmuir–type

mechanism, and negative order dependence in acetaldehyde partial pressure due to

competitive adsorption on the active sites. Instead, the opposite behavior is observed,

indicating that the direct ethanol coupling pathway does not play a major role over HAP,

although a pathway involving condensation between ethanol and acetaldehyde cannot

be ruled out. In fact, linear extrapolation of butanol formation rates to PAA = 0 kPa

(Figure 4b) suggests that butanol does not form in the absence of gas-phase

acetaldehyde. The rate enhancement by the addition of acetaldehyde demonstrates that

acetaldehyde is an intermediate in the ethanol coupling reaction. When 87 μmol h-1 of

acetaldehyde (0.12 kPa) is introduced into the ethanol feed stream, an additional 170 to
15
300 μmol h-1 of butanol is produced, depending on the ethanol partial pressure. This

discrepancy in net acetaldehyde consumption and butanol formation rates strongly

suggests that ethanol coupling is autocatalytic. The acetaldehyde consumed during the

aldol condensation step must be regenerated during the reaction, implying that the

hydrogenation step in the Guerbet pathway occurs via a hydrogen transfer mechanism

between crotonaldehyde and ethanol. The first-order dependence on acetaldehyde

suggests that enolate formation (rather than C-C coupling) is rate-limiting in the aldol

condensation step.

In-situ titration experiments were performed over HAP to elucidate the site requirements

for various products and intermediates. Pyridine was used to probe acidic sites while

carbon dioxide and propanoic acid were used to probe basic sites.

The change in acetaldehyde and butanol formation rates for two pyridine partial

pressures is shown in Figure 5. Both rates decrease slightly upon pyridine introduction

and are rapidly restored by stopping the flow of titrant, suggesting that the acid sites on

HAP are weak in nature, in agreement with reported IR observations.28,30,31,32 Acetaldol

was not detected in the product stream during the pyridine titration experiment,

indicating that the dehydration step in acetaldehyde condensation to crotonaldehyde is

fast, even in the presence of a titrant. Instead, the decrease in the rates of acetaldehyde

and butanol formation upon pyridine addition suggest that weak acid sites such as Ca2+

or POH groups play a catalytic role by stabilizing reactive ethoxide and enolate

intermediates formed on adjacent basic oxygen atoms.

16
Figure 5. Transient formation rates of acetaldehyde and butanol over HAP during

introduction and removal of pyridine. Rates are normalized to their initial steady-state

values. T = 603 K; HAP mass = 0.05 g; PEtOH = 5.7 kPa; total gas flow rate at STP = 30

cm3 min-1.

The effects of in-situ titration using CO2 are shown in Figure 6. The acetaldehyde

formation rate is unaffected by CO2, whereas the butanol formation rate decreases by a

factor of five. Upon stopping the flow of CO2, the butanol rate recovers back to 40% of

its original value, indicating that a significant fraction of adsorbed CO2 remains on the

surface. When a stronger titrant such as propanoic acid is used, the rate of

acetaldehyde formation decreases six-fold and the rate of butanol formation is

completely suppressed (Figure S1). In addition, the butanol rate recovers much more

slowly than the acetaldehyde rate. If HAP possesses a single active site that performs

both unimolecular (dehydrogenation) and bimolecular (condensation) reaction, it would

be expected that the identity of the titrant would not affect the relative selectivity of

acetaldehyde and butanol as long as the same number of sites are titrated. However,
17
this is not observed. The formation rate of butanol is strongly affected by the choice of

titrant (Fig. 5 – pyridine titration vs. Fig.6 - CO2 titration) while the formation rate of

acetaldehyde does not change much. This suggests that the differences in transient

responses for each of these titrants are due to the different site requirements for the

formation of acetaldehyde and butanol.

Figure 6. Transient formation rates of acetaldehyde and butanol over HAP during

introduction and removal of CO2. Rates are normalized to their initial steady-state

values. T = 603 K; HAP mass = 0.1 g; PEtOH = 5.7 kPa; total gas flow rate at STP = 60

cm3 min-1.

FTIR spectra of HAP acquired under a continuous flow of CO2 (1 kPa) at 473 K and 603

K are shown in Figure 7. The bands observed at 1703, 1660, and 1626 cm-1 at 473 K

correspond to surface bicarbonate species formed by reaction of CO2 with Ca-OH

groups.27,32 The bands at 1482 and 1345 cm-1 can be assigned to surface carbonates

bound to phosphate groups, whereas the bands at 1420 and 1400 cm-1 are related to

18
carbonate species on Ca-O sites derived from condensation of two adjacent surface

Ca-OH groups.27,43 At 603 K, the bicarbonate species disappear, but the surface

carbonates remain, indicating that most Ca-OH surface groups have been converted to

Ca-O.43

Figure 7. FTIR difference spectra of HAP upon introduction of continuous flow of PCO2 =

1 kPa at (a) 473 K, (b) 603 K, and (c) 603 K under continuous flow of PEtOH = 5.7 kPa.

Figure 7 also shows difference spectra for HAP taken in the presence of a continuous

ethanol feed after introduction of CO2. The reaction conditions (T = 603 K, PEtOH = 5.7

kPa, PCO2 = 1.0 kPa) were identical to those used in the in-situ CO2 titration experiment.

The bands at 1420 and 1400 cm-1 disappear, demonstrating that ethanol effectively

displaces CO2 from Ca-O sites. This suggests that ethanol dehydrogenation occurs

over these sites because CO2 does not inhibit the formation of acetaldehyde. On the

19
other hand, the bands at 1482 and 1345 cm-1 remain, indicating that CO2 is still

adsorbed on phosphate groups. A new band at 1370 cm-1 is observed, which can also

be assigned to carbonates species associated with phosphate groups.27 This band is

likely also present in the two spectra of CO2-adsorbed HAP shown in Figure 7, but is

buried under the broad band at 1400 cm-1. Another new band at 1597 cm-1 is also seen,

which can be assigned to bidentate carbonate species that can be formed from acid-

base pairs.32 All of these bands point to the presence of a Ca-O-P site that can form

carbonates adsorbed on phosphate groups. The observed decrease in butanol

formation rate upon introduction of CO2 can thus be attributed to the blockage of these

Ca-O-P sites by formation of surface carbonate species.

To develop a mechanism for ethanol coupling to butanol over HAP, it is first necessary

to envision the structure of the catalyst surface. Prior studies have shown that the

morphology and surface functionality of HAP are highly dependent on synthesis

conditions.44 HAP synthesized by co-precipitation in basic media preferentially exposes

basic surface facets due to favorable solvation effects.45 Ospina et al. have synthesized

stoichiometric HAP at pH = 11 and observed the predominance of the basic hydroxyl-

terminated (010) plane using high-resolution transmission electron microscopy.41 Under

less basic synthesis conditions (pH < 10), HAP exposes fewer basic sites which leads to

a reduction in catalytic activity and selectivity towards butanol.21 This suggests that

basic crystal planes, such as the hydroxyl-terminated (010) plane, are responsible for

the chemistry. Figure 8 illustrates this surface possesses Ca-OH sites needed for

ethanol dehydrogenation as well as CaO/PO43- pairs needed for aldol condensation.

20
Figure 8. Top-down view of the hydroxyl-terminated (010) plane showing four unit cells.

The image was generated from diffraction data reported by Hughes et al.46 using the

molecular graphics program MECURY.47 Green is calcium, orange is phosphorus, red is

oxygen, and white is hydrogen. The blue-shaded region corresponds to

dehydrogenation/hydrogen transfer sites while the purple-shaded region corresponds to

aldol condensation sites. Black arrows indicate the surface atoms involved in aldol

condensation.

The first step of ethanol coupling to butanol involves the dehydrogenation of ethanol to

acetaldehyde via an E1cB mechanism, as proposed by Kibby et al. and Tsuchida et

al.21,35 As shown in Scheme 1, this mechanism starts with dissociative adsorption of

ethanol on Ca-O sites to form surface ethoxide and hydroxyl species (Step I, Scheme

1). This is followed by abstraction of an α-hydrogen atom by surface oxygen in what is

thought to be the rate-limiting step for dehydrogenation (Step II, Scheme 1) and electron

21
transfer to the Cα-O bond to form the adsorbed acetaldehyde which then desorbs (Steps

III-IV, Scheme 1). The remaining hydrogen adatoms combine to form H2 gas (Step V,

Scheme 1).

Scheme 1. Reaction mechanism for ethanol dehydrogenation to acetaldehyde over

HAP.

Once acetaldehyde is formed, it can migrate to the neighboring CaO/PO43- sites

responsible for aldol condensation. This mechanism is initiated by adsorption of

acetaldehyde onto the surface (Step I, Scheme 2). The next step is rate-limiting in the

aldol condensation process and involves the abstraction of an α-hydrogen by a basic

oxygen to form an enolate species (Step II, Scheme 2). The enolate is stabilized by the

nearby phosphate group which is weakly acidic. This acid-base cooperation accelerates

22
aldol condensation rates and has been observed over other catalytic systems such as

mixed Mg-Al oxides or grafted amines.16,48 The enolate adds to the carbonyl group of

another acetaldehyde molecule to form acetaldol (Step III, Scheme 2) which dehydrates

to form crotonaldehyde (Step IV, Scheme 2). Desorption of crotonaldehyde and water

(Steps V-VI, Scheme 2) complete the catalytic cycle. Ethanol does not participate

directly in the reaction, but can adsorb onto the active site and thereby inhibit the aldol

condensation process.

Scheme 2. Reaction mechanism for aldol condensation of acetaldehyde to

crotonaldehyde over HAP.

Crotonaldehyde is hydrogenated to butanol by hydrogen transfer from ethanol. The

possible mechanisms for hydrogen transfer can be divided into three categories based

23
on the form of the exchanged hydrogen (H2 gas, surface hydrogen adatoms, or direct

hydrogen transfer). Ogo et al. demonstrated that a crotonaldehyde/H2 feed does not

produce any hydrogen transfer products over HAP, ruling out H2 gas as the

hydrogenation source.22 Furthermore, the addition of acetaldehyde to an ethanol feed

increases crotonaldehyde hydrogen transfer rates over HAP (Figure 4), showing that

the reaction does not proceed by indirect transfer of surface hydrogen in which

deposition of surface hydrogen from ethanol is rate-limiting. Instead, hydrogenation of

crotonaldehyde is envisioned to occur by a direct hydrogen transfer mechanism. This

can proceed via a Meerwein-Ponndorf-Verley (MPV) reduction with ethanol to yield

crotyl alcohol, shown in Scheme 3. Crotyl alcohol subsequently isomerizes to 1-buten-1-

ol by abstraction and relocation of the α-hydrogen by basic oxygen to form an enol

which tautomerizes into butanal. A second MPV reduction converts butanal to butanol.

This hydrogen transfer mechanism is hypothesized to occur on Ca-O sites as calcium

oxide is known to catalyze C=C isomerization and MPV reduction.49,50 When 0.12 kPa

of crotonaldehyde is fed over calcium oxide in the presence of excess propanol at 330

°C, butanol is formed selectively (97% yield), providing further evidence that Ca-O is the

active site for hydrogen transfer over HAP (Table S2).

Scheme 3. Hydrogen transfer pathway between crotonaldehyde and ethanol.

24
The entire mechanism can be condensed into the reaction pathway shown in Scheme

4. This scheme depicts ethanol dehydrogenation to acetaldehyde on Ca-O sites,

acetaldehyde condensation to crotonaldehyde on CaO/PO43- pairs, and hydrogen

transfer between ethanol and crotonaldehyde to form butanol. The selectivity between

acetaldehyde and butanol is dictated by the relative rates of dehydrogenation and aldol

condensation. Acetaldehyde formation exhibits a positive but less than first order

dependence on ethanol partial pressure which is consistent with the E1cB mechanism.

Scheme 4. Proposed reaction pathway for the Guerbet coupling of ethanol to butanol

over HAP catalyst.

Using a Langmuir model, a rate expression for acetaldehyde formation can be derived

based on the proposed mechanism and is given in Eqn. 1.

(1)
25
The parameter 𝑘𝑘1 is an effective rate constant associated with the abstraction of the α-

hydrogen from ethanol, while 𝐾𝐾𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸,1 is the equilibrium constant for ethanol adsorption

onto Ca-O sites. The formation of butanol does not affect the rate expression for

acetaldehyde since the consumption of acetaldehyde in the aldol condensation step is

balanced by the generation of acetaldehyde in the subsequent hydrogen transfer

𝜇𝜇𝑚𝑚𝑚𝑚𝑚𝑚
reactions. Values of 𝑘𝑘1 = 18 ℎ𝑟𝑟∙𝑚𝑚2 and 𝐾𝐾𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸,1 = 0.48 𝑘𝑘𝑃𝑃𝑃𝑃−1 result in a very good

description of the rate of acetaldehyde formation, as shown in Figure 3.

Similarly, a rate expression for butanol formation can be derived based on the

mechanism presented in Scheme 2. The observed negative order dependence in

ethanol and first order dependence in acetaldehyde for butanol formation indicate that

empty sites and adsorbed ethanol molecules are the most predominant surface species

on the aldol condensation sites. Application of the pseudo-steady-state approximation to

all surface intermediates results in the simplified rate expression given by Eqn. 2.

(2)
PAA/EtOH is the partial pressure of acetaldehyde/ethanol, k2 is the rate constant for

enolate formation, and KAA,2/EtOH,2 are the equilibrium constants for acetaldehyde and

𝜇𝜇𝑚𝑚𝑚𝑚𝑚𝑚
ethanol adsorption presented in Scheme 4. Values of 𝑘𝑘2 𝐾𝐾𝐴𝐴𝐴𝐴,2 = 1500 ℎ𝑟𝑟∙𝑚𝑚2 ∙𝑘𝑘𝑘𝑘𝑘𝑘 and

𝐾𝐾𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸,2 = 0.30 𝑘𝑘𝑘𝑘𝑘𝑘−1 accurately capture the rate of butanol formation, as shown in

Figure 4.

The temperature dependence of all rate and adsorption equilibrium constants were

obtained by fitting Eqns. 1 and 2 to rate data obtained at various temperatures (573 –
26
613 K). These results are depicted in Figure 9. Apparent activation energies and heats

of adsorption calculated in this manner are given in Table 3. The enthalpies of

adsorption for ethanol over dehydrogenation �∆𝐻𝐻𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸,1 � and aldol condensation

�∆𝐻𝐻𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸,2 � sites are both approximately -25 kcal/mol, in good agreement with adsorption

microcalorimetric measurements reported by Hanspal et al. (-22 to -26 kcal/mol).24 The

lumped parameter 𝑘𝑘2 𝐾𝐾𝐴𝐴𝐴𝐴,2 has an apparent activation energy of 1.8 ± 2.5 kcal/mol,

indicating that the heat of adsorption of acetaldehyde is roughly equal to the activation

barrier for enolate formation on aldol condensation sites. The activation energy for

ethanol dehydrogenation (𝐸𝐸1 ) is 34 ± 2 kcal/mol, which is 12 kcal/mol higher than the

reported activation energy for 2-butanol dehydrogenation over HAP.35 This difference is

a direct consequence of the relative stability between acetaldehyde and 2-butanone.

Ketones formed from secondary alcohols have two methyl groups that stabilize the

carbonyl bond compared to the one methyl group for the more reactive aldehydes.

27
Figure 9. Arrhenius plots for ethanol dehydrogenation and aldol condensation reactions

over HAP.

Table 3. Apparent activation energies and pre-exponential factors for ethanol

dehydrogenation and aldol condensation over HAP based on a least squares regression

analysis with the data presented in Figure 9.

𝝁𝝁𝒎𝒎𝒎𝒎𝒎𝒎 𝑲𝑲𝑬𝑬𝑬𝑬𝑬𝑬𝑬𝑬,𝟏𝟏 𝝁𝝁𝒎𝒎𝒎𝒎𝒎𝒎 𝑲𝑲𝑬𝑬𝑬𝑬𝑬𝑬𝑬𝑬,𝟐𝟐

Parameter 𝒌𝒌𝟏𝟏 �𝒉𝒉𝒉𝒉∙𝒎𝒎𝟐𝟐 � 𝒌𝒌𝟐𝟐 𝑲𝑲𝑨𝑨𝑨𝑨,𝟐𝟐 �𝒉𝒉𝒉𝒉∙𝒎𝒎𝟐𝟐 ∙𝒌𝒌𝒌𝒌𝒌𝒌�
(𝒌𝒌𝒌𝒌𝒌𝒌−𝟏𝟏 ) (𝒌𝒌𝒌𝒌𝒌𝒌−𝟏𝟏 )

Apparent Ea or
34 ± 2 -23 ± 2 1.8 ± 2.5 -27 ± 3
ΔHads (kcal/mol)

ln(A) 31 ± 2 -19 ± 2 8.7 ± 2.1 -24 ± 3

Parity plots comparing measured formation rates and those obtained from Eqns. 1 and

2 are shown for acetaldehyde and butanol in Figure 10. The data points cluster along

the diagonal, which indicates a good description of the data by the model.

28
Figure 10. Parity plots for rates of (a) acetaldehyde and (b) butanol formation derived

from kinetic data and Eqns. 1 and 2. PEtOH = 2.4 – 9.4 kPa; PAA = 0.055 – 0.12 kPa; T =

573 – 613 K; HAP mass = 0.05 g; total gas flow rate = 30 cm3 min-1.

The model also explains the kinetics observed when only ethanol is fed into the reactor.

At low ethanol partial pressures (<4 kPa), acetaldehyde can freely adsorb on aldol

condensation sites. Thus, the positive order dependence for butanol formation is a

direct consequence of the positive order dependence for ethanol dehydrogenation. As

ethanol partial pressure increases, ethanol begins to compete with acetaldehyde for

adsorption on the condensation sites. As a result, aldol condensation rates decrease

and oppose the increasing dehydrogenation rates, leading to the observed zero order

dependence for butanol formation at high ethanol partial pressures.

The apparent rate of butanol formation from a pure ethanol feed can be predicted by

integrating the rate expression given in Eqn. 2 over the length of the catalyst bed, as

shown in Eqn. 3.

(3)

𝑟𝑟𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵,𝑎𝑎𝑎𝑎𝑎𝑎 is the apparent butanol formation rate, 𝑟𝑟𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 is the local rate of butanol

formation at a distance 𝐿𝐿 from the top of the catalyst bed, and 𝐿𝐿′ is the total length of

the bed. The partial pressure of ethanol was taken to be constant because the

experiments were performed under differential conditions (<5% conversion). The spatial

dependence of acetaldehyde partial pressure was determined experimentally through

spacetime studies (Figure 2) to account for minor acetaldehyde-consuming side

29
reactions such as the formation of butadiene. As shown in Figure 3, the model

developed from ethanol/acetaldehyde co-feed kinetics accurately predicts butanol

formation rates.

CONCLUSIONS

Butanol formation rates on HAP were determined over a range of ethanol (3.5 – 9.4

kPa) and acetaldehyde (0.055 – 0.12 kPa) partial pressures. A strong increase in the

rate butanol formation was observed upon addition of acetaldehyde to the feed,

demonstrating that the reaction is autocatalytic. For all reaction conditions tested,

ethanol exhibits a negative order dependence, while acetaldehyde exhibits a positive,

first-order dependence, implying that butanol is formed via a Guerbet-type mechanism

involving sequential dehydrogenation, aldol condensation, and hydrogen transfer

reactions.

Poisoning experiments with CO2 and propionic acid reveal that acetaldehyde and

butanol formation rates decrease to different extents, demonstrating that the sites for

the two reactions are not equivalent. Combined with in-situ FTIR results, it was found

that dehydrogenation is catalyzed by Ca-O groups while aldol condensation requires

calcium oxide/phosphate pairs. A detailed model of ethanol coupling reaction was

developed that accurately describes all experimental observations.

ACKNOWLEDGEMENTS

30
This work was funded by Director, Office of Science, Office of Basic Energy Sciences of

the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. The authors

would like to thank Joseph Gomes for helpful discussions.

REFERENCES

(1) Sun, J.; Wang, Y. ACS Catal. 2014, 4, 1078–1090.

(2) DeWilde, J. F.; Chiang, H.; Hickman, D. A.; Ho, C. R.; Bhan, A. ACS Catal. 2013,
3, 798–807.

(3) Jones, M. D. Chem. Cent. J. 2014, 8, 1–5.

(4) Ash, M.; Ash, I. Handbook of green chemicals; 2nd ed.; Synapse Information
Resources: Endicott, 2004; pp 645–646.

(5) Cheng, J. Biomass to Renewable Energy Processes; CRC Press: Boca Raton,
2010; pp 272–273.

(6) Van Leeuwen, P. W. N. M.; Claver, C. Rhodium Catalyzed Hydroformylation;

Kuwer Academic Publishers: Dordrecht, 2000; pp 6–8.

(7) Weissermel, K.; Arpe, H.-J. Industrial Organic Chemistry; 4th ed.; Wiley-VCH:
New York, 2003; pp 127–138.

(8) Kozlowski, J. T.; Davis, R. J. ACS Catal. 2013, 3, 1588–1600.

(9) Gabriëls, D.; Hernández, W. Y.; Sels, B.; Van Der Voort, P.; Verberckmoes, A.
Catal. Sci. Technol. 2015, 5, 3876–3902.

(10) Dowson, G. R. M.; Haddow, M. F.; Lee, J.; Wingad, R. L.; Wass, D. F. Angew.
Chemie - Int. Ed. 2013, 52, 9005–9008.

(11) Ueda, W.; Ohshida, T.; Kuwabara, T.; Morikawa, Y. Catal. Letters 1992, 12, 97–
104.

(12) Ndou, A. S.; Plint, N.; Coville, N. J. Appl. Catal. A Gen. 2003, 251, 337–345.

(13) Birky, T. W.; Kozlowski, J. T.; Davis, R. J. J. Catal. 2013, 298, 130–137.

31
(14) León, M.; Díaz, E.; Ordóñez, S. Catal. Today 2011, 164, 436–442.

(15) Carvalho, D. L.; de Avillez, R. R.; Rodrigues, M. T.; Borges, L. E. P.; Appel, L. G.
Appl. Catal. A Gen. 2012, 415-416, 96–100.

́ z, V. K.; Xu, M.; Iglesia, E.; Apesteguıa

(16) Di Cosimo, J. I.; Dıe ́ , C. R. J. Catal. 1998,
178, 499–510.

́ , C. R.; Ginés, M. J. L.; Iglesia, E. J. Catal. 2000, 190,

(17) Di Cosimo, J. I.; Apesteguıa
261–275.

(18) Earley, J. H.; Bourne, R.; Poliakoff, M.; Watson, M. Green Chem. 2015, 17, 3018–
3025.

(19) Yang, C.; Meng, Z. J. Catal. 1993, 142, 37–44.

(20) Tsuchida, T.; Sakuma, S.; Takeguchi, T.; Ueda, W. Ind. Eng. Chem. Res. 2006,
45, 8634–8642.

(21) Tsuchida, T.; Kubo, J.; Yoshioka, T.; Sakuma, S.; Takeguchi, T.; Ueda, W. J.
Catal. 2008, 259, 183–189.

(22) Ogo, S.; Onda, A.; Yanagisawa, K. Appl. Catal. A Gen. 2011, 402, 188–195.

(23) Ogo, S.; Onda, A.; Iwasa, Y.; Hara, K.; Fukuoka, A.; Yanagisawa, K. J. Catal.
2012, 296, 24–30.

(24) Hanspal, S.; Young, Z. D.; Shou, H.; Davis, R. J. ACS Catal. 2015, 5, 1737–1746.

(25) Silvester, L.; Lamonier, J.-F.; Faye, J.; Capron, M.; Vannier, R.-N.; Lamonier, C.;
Dubois, J.; Couturier, J.; Calais, C.; Dumeignil, F. Y. Catal. Sci. Technol. 2015, 5,
2994–3006.

(26) Diallo-Garcia, S.; Osman, M. B.; Krafft, J. M.; Boujday, S.; Guylène, C. Catal.
Today 2014, 226, 81–88.

(27) Diallo-Garcia, S.; Osman, M. B.; Krafft, J. M.; Casale, S.; Thomas, C.; Kubo, J.;
Costentin, G. J. Phys. Chem. C 2014, 118, 12744–12757.

(28) Stošić, D.; Bennici, S.; Sirotin, S.; Calais, C.; Couturier, J.-L.; Dubois, J.-L.;
Travert, A.; Auroux, A. Appl. Catal. A Gen. 2012, 447-448, 124–134.

(29) Ramesh, K.; Goh, E.; Ling, Y.; Gwie, C. G.; White, T. J.; Borgna, A. J. Phys.
Chem. C 2012, 116, 18736–18745.

32
(30) Parry, E. J. Catal. 1963, 2, 371–379.

(31) Tanaka, H.; Watanabe, T.; Chikazawa, M. J. Chem Soc., Faraday Trans. 1997,
93, 4377–4381.

(32) Hill, I. M.; Hanspal, S.; Young, Z. D.; Davis, R. J. J. Phys. Chem. C 2015, 119,
9186–9197.

(33) Hanspal, S.; Young, Z. D.; Shou, H.; Davis, R. J. ACS Catal. 2015, 1737–1746.

(34) Kibby, C. L.; Hall, W. J. Catal. 1973, 159, 144–159.

(35) Kibby, C. L.; Hall, W. J. Catal. 1973, 73, 65–73.

(36) Rodrigues, E. G.; Keller, T. C.; Mitchell, S.; Pérez-Ramírez, J. Green Chem.
2014, 16, 4870–4874.

(37) Chieregato, A.; Ochoa, J. V.; Bandinelli, C.; Fornasari, G.; Cavani, F.; Mella, M.
ChemSusChem 2015, 8, 377–388.

(38) Gines, M.; Iglesia, E. J. Catal. 1998, 176, 155–172.

(39) Scalbert, J.; Thibault-Starzyk, F.; Jacquot, R.; Morvan, D.; Meunier, F. J. Catal.
2014, 311, 28–32.

(40) C. Meunier, F.; Scalbert, J.; Thibault-Starzyk, F. Comptes Rendus Chim. 2015,
18, 345–350.

(41) Ospina, C. a; Terra, J.; Ramirez, A. J.; Farina, M.; Ellis, D. E.; Rossi, A. M.
Colloids Surf. B. Biointerfaces 2012, 89, 15–22.

(42) Mobasherpour, I.; Heshajin, M. S.; Kazemzadeh, a.; Zakeri, M. J. Alloys Compd.
2007, 430, 330–333.

(43) Cheng, Z.; Yasukawa, A.; Kandori, K.; Ishikawa, T. Langmuir 1998, 14, 6681–
6686.

(44) Liu, J.; Ye, X.; Wang, H.; Zhu, M.; Wang, B.; Yan, H. Ceram. Int. 2003, 29, 629–
633.

(45) Slepko, A.; Demkov, A. A. J. Chem. Phys. 2013, 139, 1–9.

(46) Hughes, J. M.; Cameron, M.; Crowley, K. D. Am. Mineral. 1989, 74, 870–876.

33
(47) Macrae, C. F.; Edgington, P. R.; McCabe, P.; Pidcock, E.; Shields, G. P.; Taylor,
R.; Towler, M.; Van De Streek, J. J. Appl. Crystallogr. 2006, 39, 453–457.

(48) Shylesh, S.; Hanna, D.; Gomes, J.; Krishna, S.; Canlas, C. G.; Head-Gordon, M.;
Bell, A. T. ChemCatChem 2014, 6, 1283–1290.

(49) Aramendía, M. A.; Borau, V.; Jiménez, C.; Marinas, J. M.; Ruiz, J. R.; Urbano, F.
J. J. Colloid Interface Sci. 2001, 238, 385–389.

(50) Schächter, Y.; Pines, H. J. Catal. 1968, 11, 147–158.

34
GRAPHICAL ABSTRACT

35